DEVELOPMENT OH AN ACCELERATED WEAR TEST HOA TILLAGE TOOL MATERIALS AND ITS APPLICATION TO 0ADI WARY CAST IRON AND WO DULAR IRON By Nuredin Nuri Mohsenin A THESIS Submitted to the School of Advanced Graduate Studies of Michigan State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of DOCTOR OR PHILOSOPHY Department of Agricultural Engineering 1956 ProQuest Number: 10008658 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest. ProQuest 10008658 Published by ProQuest LLC (2016). Copyright of the Dissertation is held by the Author. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code Microform Edition © ProQuest LLC. ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 4 8 1 0 6 - 1346 ii AGKN 0 wLEDG-idEA 'I.1S The author wishes to express his appreciation for the helpful suggestions and timely guidance of Dr. W.M. Garleton of the Department of Agricultural Engineering of Michigan State University. He also wishes to thank the other members of the guid­ ance committe, Doctors H.L. Womochel, D.J. Montgomery, and W.D. Baten for their suggestions and guidance during the inve stigation. The author is particularly thankful to Doctors H.L. Womochel and D. J. Harvey of the Metallurgical Engineering Department for their cooperation and assistance in melting and pouring the cast iron specimens, the chemical analysis of the irons, and making possible the use of foundry and \ metallurgical equipment. ' The writer also appreciates the efforts of Dr. A.W. Parrail, Head of Agricultural Engineering Department, for obtaining funds from the Agricultural Experiment Station . The help received from members and persons in charge of the research laboratory is also much appreciated. The writer is especially grateful to his wife, Iran, who typed the manuscript and gave much encouragement during the investigation. iii Nuredin Nuri Mohsenin candidate for the degree of Doctor of Philosophy Final examination, May 8, 1958, 3:00 p.m., Room 218, Agricultural Engineering Department Dissertation: Development of an Accelerated Wear Test for Tillage Tool Materials and its Application to Ordinary Cast Iron and Nodular Iron Outline of Studies Major Subjects: Minor Subjects: Agricultural Engineering Mechanical Engineering and Physics Biographical Items Born, September 15, 1923, Teheran, Iran Undergraduate Studies, Karaj Agricultural College, Univsity of Teheran, Iran, 1944“47, Utah State Agricultural College, 1947-4§, California State Polytechnic, 1948-49, Oklahoma Agricultural and Mechanical College, 1949-51* Graduate Studies, University of California, 1951-52, Michigan State University, 1952-56. Experience: Summer experience on California farms, 1948, 1949, Machinist, Chicago Auto Parts, summer 1951, Order Clerk, Chicago Tractor Supply Company, summer 1952, G-raduate Teaching As­ sistance, Michigan State University, 1953-55, Graduate Research Assistant, 1955-56. Member of American Society of Agricultural engineers; Sigma Tau, Engineering Honor Fraternity; Sigma Pi Sigma, Physics Honor Society; Society of Sigma Xi. vlil TABLE OF CONTENTS I. INTRODUCTION.................................... Page 1 IJI. REVIEW OF LITERATURE............................ 6 III. MECHANISM OF WEAR PHENOMENA AND FACTORS INFLUENCING WEAR OF METALS IN SOIL............ 15 The Mechanism of Wear........................ 15 Wear of Metal Against.. Metal............. 15 Wear of Metal Against...... Soil.......... 17 Factors Influencing Wear of Metals in Soil.. 18 Soil Factors Affecting Wear...... 19 Metal Factors Affecting Wear.......... . 25 Effect of Operating Speed and CuttingAngle33 IV. DEVELOPMENT OF A METHOD FOR LABORATORY AND FIELD WEAR TESTS........ . Development of a Laboratory Wear Test Machi ne....... 35 35 Requirements for a Laboratory Wear Test Machine. ............................ 35 Devices for Laboratory Abrasion Tests.... 37 Design and Construction of the Wear Test Machine............... •••••........ 40 Development and Casting of the Experimental Plow Points. ••...••••• ........ 50 Selection of the Irons..................... 50 Investigations for a Test Specimen of Suitable Shape and Size.................. 55 DEVELOPMENT OP AN ACCELERATED WEAR TEST FOR 0 TILLAGE TOOL MATERIALS AND ITS APPLICATION TO ORDINARY CAST IRON AND NODULAR IRON By Nuredin Nuri Mohsenin AN ABSTRACT Submitted to the School of Advanced Graduate Studies of Michigan State University of Agriculture and Applied Science In partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Engineering 1956 Approved_ Uj oM l hi._ V NUREDIN MOHSENIN ABSTRACT Experiences of the past had indicated the need for a wear test .r© thod whereby a satisfactory correlation could be established between the results of the field and the laboratory tests. The object of such a method was to r e ­ lieve the experimenter from a large part of the variables and dif i'icult ies encountered in elaborate and costly field tests. Determination of the role of carbon and other al­ loying elements as a factor in wear resistance, and the possibilities of using nodular iron as a tillage tool m aterialwere additional reasons for undertaking this re­ search. Studies were made of the mechanism of wear of metals in soil. Soil friction, soil moisture, and soil texture; composition, heat treatment, hardness, and microstructure of metal; and operating speed and cutting angle were found to be among the most important factors ai fee ting wear of tillage tools. Wear experiments were conducted both in the laboratory and in the field. A wear test machine was developed for determining the wear resistance of materaisl u m Ler speed and load conditions which approximated very nearly the field conditions. Wear tests can be made on small samples in less than ten minutes. Those materials that show superior wear NUREDIN rluHSENIN ABSTRACT resistance can be tested for imoact resistance, another im­ portant factor in the wear of tillage tools* The field tests were conducted in four series to in­ clude 22 individual tests. There were 122 miniature -plow j point specimens made from 11 types of irons to include or­ dinary cast iron and nodular iron of high carbon or low carbon and alloyed or unalloyed compos, tions. The specimens were made both in the chilled and unchilled conditions. Correlation between the laboratory and field wear tests was highly significant. This high degree of correlation plus the simplicity and the rapidity with which tests could be made indicated the suitability of this method of wear test for the selection of tillage tool materials. Comparison of chilled and unchilled irons of the same composition and treatment under the same field test conditions Indicated that an unchilled iron will wear four to five times as much as the corresponding iron in the chilled condition. As for the suitability of nodular iron as a material for cast iron plowshares, It was concluded that if a plowshare Is to consist of both chilled and unchilled Irons, a high carbon nodular iron promises to be as good a material for this purpose as can be made available at the present time. In the unchilled condition, nodular Iron was superior to any of the ordinary cast iron samples. In the chilled condition, nodular iron NUREDIN MOHSENIN ABSTRACT was intermediate in wear resistance between low and high carbon ordinary cast irons. In addition, it has greater ductility and impact resistance, tendency, a controllable chilling and greater hardness in the unchilled condition which aids scouring. Correlation between Brinell hardness of the miniature plow points and their weight losses in the field proved that the present methods of hardness measurements which do not differemtiate between the microstruetures of irons with the same hardness, cannot be taken as an index of wear resistance It was proposed that perhaps a scratch test might give result of more practical value than the present methods of hardness me asurement s . ix Page V. Casting of the Miniature Plow Points..... £3 Examination of the Castings and Preparation of the Test Pieces for Laboratory and Field Tests....... ... 60 Adaptation of a Cultivator for the Accelerated Field Tests..,................ 66 Preliminary Field Tests to Determine Reproducibility of Results................ 68 Experimental Designs for the Accelerated Field Tests....... 70 Controlled Experiments.................... 70 Latin Square Experiments................. 71 PiUiiSENTATION AND ANALYSIS OF FIELD AND LABORATORY WEAR TEST DATA..................... 75 Small Scale Accelerated Field Tests........ 75 First Series of Field Tests............. 76 Second Series of Field Tests............. 73 Third Series of Field Tests....... 89 Fourth Series of Field Tests............. 914- Laboratory Wear Tests of the Field-Tested Specimens and Correlation of the Field and Laboratory Data...... 100 Large Scale Field Tests Using Actual Plow share s............... 113 Correlation Between Hardness and Wear Resistance of the Cast Miniature Plow Polnt3. .................................... 12.2, z Page VI. SUMMARY AND CONCLUSIONS........................ 129 Summary.......................... 129 Conclusions. 133 .............. LITERATURE CITED.................................... 138 xi LIST OP TABLES Table Page I* Analysis of regular commercial cast iron plow­ shares during the early stages of the project..., 26 XI, . Effect of load on weight losses of steel specimens tested on the wear test machine........ I4.5 III. Typical weight losses in milligrams obtained in four replicated tests on each cast iron specimen. ........................ .. 14-8 IV. Chemical analysis of irons used in the casting of the miniature plow points....................... 61 V. Field test of chilled ordinary and nodular iron plow points of equal composition and treatment... 69 VI* First series of field tests. Weight losses of chilled iron specimen in milligrams (controlled, design)............................................. 77 VII. Rating of all the iron specimens in order of decreasing resistance to wear..................... 99 VIII. First series of tests. Correlation data for the field and laboratory wear tests... ,. ,J.01 IX. Second series of tests. Correlation data of the laboratory and field wear tests..............X02 X. Third series of tests. Correlation data of the laboratory and field wear tests............. .103 XI. Fourth series of tests (right side). Correlation data of the laboratory and field wear tests J.OI4. XII. Fourth series of tests (left side). Correlation data of the laboratory and field wear tests.. ... .105 XIII. Weight losses of high carbon (HC) and low carbon (LC) cast iron ploivrshares In actual field tests. J.16 XIV. Results of measurements of the outline tracings of the cast iron plowshares before and after field tests.................................... .119 xii Table XV. XVI. Page Correlation data Loss in field of for Brinell hardness and weight the chilled specimens.............#123 Correlation data for Brinell hardness and weight loss in field for theunchilled specimens........... ^2.5 xiii LIST OF Flaunts .FIGURE 1. 2* 3# . PAGE General view of the wear test machine showing the sand, hopper and the vacuum cleaner...... . . Close-up of the wear test machine showing the carriage, the rotating disk with a rubber band, the sand tube, the specimen held in place, and the plastic window in inset........ Jj.2 Ordinary gray iron with graphite in the form of flakes............. ......................... 53 Nodular iron with graphite in the form of nodules or spheroidal carbon 5* .. Groups of unchilled (top) and chilled (bottom) cast iron plow points before cutting and finishing................................ 6# Original and final designs of the patterns and castings for the chilled iron specimens...... 7. 8. 22. 23* 53 57 57 Wood pattern and one of the machined castings used as a chill block........................ 59 Match plate used originally for casting of the unchilled iron specimens.............. 59 9-20.Photomicrograph of ordinary and nodular cast irons................... . . . ........ 21. 2 83-65 Side view of the cultivator adapted for field tests................. 67 Right side of the cultivator showing the five shanks used for the 5*5 Latin square design.,. 67 Test series No. 1. Correlation of "controlled" field and laboratory wear tests of chilled iron specimens ...... 2li, Test series No. 2. Correlation of "randomized" field and laboratory wear tests of chilled iron specimens.................................. 108 109 ixv BIGURE 25. 26. 27. 28. PAGE Test series No. 3. Correlation between labor­ atory and field wear of the unchilled Iron specimens........................................ HO Test series No. I}, (right side). Correlation of laboratory and field wear tests of the chilled and unchilled specixnens........ HI Test series No. lj_ (left side). Correlation of laboratory and field wear tests of the chilled and unchilled specimens......................... 112 Method of sectioning the point of the cast iron shares for comparison of chill patterns.... ___ 111*. 29. Cross-section of the point of the cast iron plow­ 111}. shares showing the chill pattern............... 30. Correlation between Brinell hardness and weight loss in field for the chilled specimens........ 12lp Correlation between Brinell hardness and weight loss in field for the unchilled specimens...... 126 31. I. INTRODUCTION Agricultural tools provide one of the sustaining links between the good earth and a hungry world. Not the least important of these tools are the tillage tools. Wear in soil is the destroyer of a tillage tool which may happen to be a moldboard plow, a disk plow, a disk harrow, a cultiva­ tor, or some other soil-working face of an agricultural im­ plement. This oroblem of wear of tillage tools, especially the plowshare, which cuts under the surface of the soil and acts as a wedge to open the soil, has been one of the agri­ c u l t u r i s t s foremost problems from the time when he used a crooked stick plated with iron to this age of modern plows equipped with razor-blade plowshares. Gallwit z (13 r calculated that every year 5*272 tons of steel and iron are lost in German agricultural soils owing to wear of plowshares. Beresford and Humphrey (2) estimated that in Idaho the cost of servicing plowshares was as much as twenty percent of the total plowing cost. Carleton and Martin (9) in Kansas found that an ordinary solid-steel plow­ share can be sharoened from four to seven times before being discarded, and each sharpening lasting only eight to ten acres of plowing. They concluded that the cost of servicing plow­ shares amounted to ten to fifteen percent of the total plow“'Numbers in parentheses refer to the "Literature Cited" on pages 138 to lljXU 2 ing cost. When the cost of maintaining a suitable cutting edge on other tillage tools was added to the cost of plow­ share service, they estimated a total cost of two to three million dollars which the Kansas farmers had to bear annually because of the wear of tillage tools* In Michigan, rough estimates by farm implement dealers indicate that more than one-half million shares may be used annually (18). The above estimates do not include lost time or travel expense in changing or renewing the plowshares. Other items of importance also are the factors of waste of power through the use of dull tools, faulty machine operation, and ineffi­ cient results from the implement. When the annual cash out­ lay charged to the wear of tillage tools is added to these other factors, it apoears likely that a small reduction in wear would bring about a considerable saving in time and ex­ penses per year to the farmer. Many improvements have been made in metals used in var­ ious industries and the tillage implement materials have, undoubtedly, kept pace with other developments in metallurgy. Even so, at the present time, there Is no agreement as to the best material for tillage tools. Plowshares, for example, are produced from a variety of materials such as soft-center steel, low carbon steel, high carbon steel, carburized steel, cast iron with the point and cubbing edge chilled, and a new m a ­ terial called nodular or ductile Iron. There are few speci- 3 fications for the composition, heat treatment, and hardness of tillage tool materials which meet with universal aporoval by manufacturers* specifications. These manufacturers differ widely in their There are no basic or scientific data sup­ porting present practice on the selection of tillage implement materials. Present oractice is either traditional or based on the experience of the particular manufacturer. To make a fundamental study of the requirements of tillage tool cutting edges, some workers in this country and abroad have tried to evaluate wear of tillage tools in various manners. Although their vo rk has not yet resulted in any solid conclu­ sions, these investigations have been valuable in ejp loring the method of approach, in establishing a working hypothesis, and indicating certain relationships. The work at Michigan State University started in 1950 when a project was set up in cooperation with the Departments of Mechanical Engineering and Metallurgical Engineering to investigate the possibilities for wear reduction of cast iron plowshares through controlled addition of alloying elements. Cast iron was used as a basic material because it was estima­ ted that in Michigan, owing to the abrasive nature of the soil, there were three cast iron shares in m e for each steel share used. Approximately 200 shares were cast in the Mechanical Engineering foundry. These were of various carbon contents 14- and with various inoculants for controlling chill. Many troubles were experienced, the principal one being T,hot tears'* along the chilled cutting edge of the low carbon compostion share 3 . This particular difficulty was a very serious one because the primary objective was to determine the role of carbon content as a factor in wear resistance. A second serious trouble had to do with maintaining the same chill pattern on comparative shares. Shares could not be expected to show the same resistance to wear unless the depth and quantity of chill on each one was approximately the same* Some actual field tests were made with the experimental shares. Some trouble was encountered with breakage due to rocky conditions. A summary of tests for the high and low carbon shares is shown in Table XIII. Discouragements in foundry work and field tests plus the lack of sufficient personnel caused the project to become in­ active and to remain so until the beginning of winter term, 1954* when the project was reactivated and continued as a thesis problem. From the experience of the past, it was realized at the beginning of this work that evaluating wear of tillage tools in the field is a difficult task. A limited number of plows can be tested at one time, and variation in soil and moisture conditions present constantly changing variables. In Michigan, 5 weather conditions present quite a problem also in conducting out-door field tests. Realizing these limitations, it was de­ cided to develop a laboratory wear test method whereby it was hoped, a satisfactory correlation could be obtained from the results of the field tests and the laboratory tests. It was felt that such a method of wear test would be most valuable for the development of better materials for tillage tools. It would relieve the experimenter from all the variables, dif­ ficulties, and expenses encountered in field tests. To attempt to obtain such a correlation, it was decided to conduct the experiment in three stages: (1) laboratory tests, where a suitable wear test machine could be used; (2) small scale field tests, where a large number of test specimens could be subjected to the same tilling conditions in a comparatively s hort time; and (3) a large scale field test, where actual plowshares or other tillage tools would be subjected to actual tillage practices. In addition to the development of an accelerated wear test method, additional reasons for undertaking this research were to determine the role of carbon as a factor in wear re­ sistance, and to test the new nodular iron, which promised to be an Ideal material for plowshares because it combines chil­ ling tendency, ductility, and impact resistance. 6 II. REVIEW OR LITERATURE The problem of wear of metal against metal has been studied quite extensively. Although the literature is a- bundant with various wear test methods and discussions on the mechanism of wear phenomena, the problem appears to b© still a subject of considerable controversy. The available information on the wear of metals in soil, however, is meager. The first published information on the wear of tillage tools was reported as early as 1922* Excel­ lent wearing of manganese steel jaws in rock crushers, tractor gears and track links, and dredger machinery led Hoffman (19) to the hope that this material might provide a better plow­ share than those ordinarily used. He tried manganese steel, soft center steel, and chilled cast iron plowshares on a three bottom tractor-plow, rotating the shares after approximately one-third of each test had been run. Data taken included weights, dimensions, and outline tracings of each share. Least wear was shown by the chilled cast iron shares, slightly more by the soft center steel, and considerably more by manganese steel. During the years of 1927 and 1928, a study was conducted by the Institut fur Maschinenkunde der Landwlrtschaftllchen Hochschule Berlin under the direction of Karl Gallwitz (13) to determine some of the factors which affect the wear resist­ ance of materials in soil. Some 37 steel plowshares of r various compositions and heat treatments were subjected to both laboratory and field tests. The laboratory machine consisted of a large concrete pan with an outer diameter of four meters and an inner diameter of 2.34 meters. The oan was filled with s&nd of known composition and moisture content to a depth of forty centimeters, A drive consisting of a beveled gear and pinion, powered by a thirty kilowatt electric motor, rotated a framework to which plowshares were attached at any desired angle or depth. Moisture of the sand was con­ trolled very closely by adding water at regular intervals. Under these laboratory conditions, the wear of each plowshare was determined as to the effect of soeed, type of abrasive, moisture content, operating angle, operating depth, composi­ tion and hardness of the material. The same plowshares were then used in the field for actual plowing. The results of two years of work indicated certain relationships, in some cases a satisfactory correlation between the field and labor­ atory wear tests. The effect of alloying elements in steel remained a question to be studied later. In 1936, Zink, Sellers, and Roberts (38) reported the results of their studies for the cutting edges of tillage Implements. They had devised a laboratory wear test machine similar to that of Gallwitz except on a smaller scale and with some Improvements. Here the soil in a trough-shaped pan was rotated against the specimens. The specimens were given an 8 oscillating motion by a cam and push rods* By this means, the specimens not only moved back and forth in a simple har­ monic motion across and at right angles to the rotating pan, but also they remained stationary a part or the time. This oscillating motion was to prevent the specimens from travel­ ling in the same path. The wear specimens, instead of being actual plowshares, as in Gallwitz*s tests, consisted of twoinch by four-inch steel plates cut from one-fourth or fivesixteenth inch stock. The specimens were carefully sharpened to a cutting edge and heat-treated. conditions, Under these laboratory studies were conducted over a 100-hour test per­ iod to determine the effects of various heat treatments, difdifferent soils at different moisture contents, direction of flow lines in rolled steel, and effect of hardness upon the wear resistance of materials. Along with laboratory work, some field work with actual plowshares was carried on. No attempt was made, however, to study the correlation between the results of the laboratory and field work. In this work again, no definite concluaions were drawn with reference to the desirable material for tillage tools or a suitable laboratory method of wear test. The methods used both by Gallwitz, Zink ejt al, although they may be ef­ fective for tests of materials for tillage tools, have the following disadvantages: 1. Preparation of the wear test specimens requires 9 excessive time* In the first method, actual plowshares were to be made 1'rom the material to be tested. In the second method, cutting edges had to be shaped and sharpened very carefully. The preparation of a number of these cutting edges for large replications and higher significance of the results of the laboratory tests, would require a considerable amount of time* 2* To obtain a relatively large amount of weight loss or change of dimension of the specimen, a long test period is required (100 hours in the second method). 3* Variation in the moisture content of the abrading medium during the long period of the test affects the rate of wear* A precise control of this moisture content during the test period was indicated to be a task of great difficulty, ij.* The apparatus is bulky, and the running of a few spec­ imens for a long period in a large mass of sand or soil re ­ quires considerable power. In view of the above shortcomings, it appears that this form of laboratory test is not a completely desirable method* Another interesting work in connection with wear of til­ lage tools is reported by Carleton and Martin (9)* They con­ ducted field tests to determine the comparative wearing prop­ erties of various types of plowshares, share treatments, and hard-surfacing materials. They found that soft-center shares, despite their higher costs, were very little better than 10 ordinary solid-steel shares unless properly heat treated. They also concluded that cast iron shares gave cheap service if scouring was not a problem. As to the wear resistance of hard-surfaced plowshares, they concluded that the method did not appear to be practical for use by the farmer himself or even the ordinary blacksmith shop. The matter of hard-surfacing the cutting edges of tillage tools by applying a layer or coating of special alloy, resist­ ant to abrasion, has attracted the attention of the farmer eyer since he became concerned about the wear of his tillage tools. A hard-surfacing material consists generally of a chromiumcobalt-tungsten or mollybdenum alloy and is applied to the cutting edge of the tillage tool by electric arc or acetylene torch. Tests have proved its importance in increasing the life of cutting edges of tillage tools but owing to its rather high cost and the skill which is required in applyint it, it has not proved to be the solution to the wear problem. The skill with which the welder applies the hard-suri'acing material is the secret of success or failure of the treatment. For this reason the farmers* opinions on the value of hard-surfacing vary widely. In the years after World War II, some Japanese scientists engaged in the study of wear of steel against soil. work, Tetsutaro Mitsuhasi soil moisture, velocity, In this e_t_ al (33) studied the effect of cutting angle, and load In the normal 11 direction upon a piece of steel dragged in soil. Some field work was also carried on to see the effect of these various factors, particularly moisture, on wear resistance of plow­ shares used in Japan. No attempts were made to correlate the results of the laboratory work and the field work. The most recent literature on wear of tillage tools appeared in 1951* Reed and Gordon (28) reported the results of their studies at the U. S.D.A. Tillage Laboratory. These workers devised a laboratory wear test machine which incor­ porated the principles used originally by Brineil (6) for the study of abrasive wear of metals. The Brineil method of wear test utilized a specimen clamped in a holder against which a disk of open-hearth iron 100 mm in diamter and four mm thick was rotated* Sand from a hopper above the device was fed con­ tinuously between the specimen ard the rotating disk. Wear resistance of a material was deteririined by loading the spec­ imen with a certain weight and rotating the disk against it at a predetermined speed and time. Using the above principles, the work at U.S.D.A. Tillage Laboratory was carried on to determine the relative wear resistance of 16-inch disks made from five types of s teel and 18-inch disks made from four types of steel, Field tests were made by weighing and measuring four or five disks of each type of steel and size, then placing them on farmer-owned tan­ dem disk harrows to be used during the regular seedbed prep- 12 aration season. The locations of the disks on each harrow were changed periodically so that no disk remained in a f a ­ vorable position throughout the test. The weight loss of each disk was determined at the end of each test. Specimens for wear test, by use of the wear test machine, were cut from two or three disks of each type of steel. loss The average weight for three runs on each sample was taken as the weight loss of that sample under the laboratory conditions. The results of laboratory and field tests for the dif­ ferent types of steels were represented in the form of bar graphs and the authors concluded that the data for the two types of tests correlated closely. To examine this correla­ tion more carefully, the writer paired off the data for both 16-inch and l8-inch disks and calculated the correlation co­ efficients nr ” as shown on the following page* a table of distribution Reference to of correlation co­ efficient (3 1 ) showed that at five percent level of signifi­ cance, the results of the two tests for the 16-inch disks cor­ relate closely whereas those of the l8-inch disk fall short of the required ”r ” by five and one-half points. The fact that a satisfactory correlation was obtained at least for one set of the data, however, was an indication that the Brineil method of wear test might have some possibilities for laboratory test of materials for tillage tools. The writer concluded that those possibilities deserved further investigations. 13 CORRELATION ANALYSIS AND 18-INCH DISKS Let X^ z. weight Xg 2. weight OF LABORATORY WEAR TESTED BY U.S.D.A. TILLAOE LABORATORY loss in laboratory X]_ A 93.78 96.22 98.89 100.67 106.33 a c F D 18-inch Disks Type of Steel (s x ) 2 (SXX )2 - - 39,139.6593 39,11^.9506 N 2 - 1.05310 (SX0 ) . SX„ - - 0.765625 N N SX1X 2 0 .7 6 7 6 5 0 (SX2 ) 1 01+882 2 2 7 .7 0 0 0 - 1 7 3 . 25315 SX1 X 2 SX1SX2 SX1SX2 N 1 0 0 .0 0 1 0 2 .2 2 - 1+9,181.378 H SX2 97.78 0 .5 0 0 SX,1 _ *1 95.55 E B H J -O.i4.lO 0 . 14.50 o .w o 0 . 14.60 SX1 - 1+9272.727 1 in milligrams. loss in field in pound. 16-inch Disks Type of Steel AND FIELD WEAR OF 16-INCH - 2 2 7 .1 1 7 6 N 1 7 3 .0 5 3 1 3 *2 o.lp.5 0 .1+30 o . i +30 0.1+75 ih b 21 — Regression of - sxi ►^2 " on X^ SXi SXq N (3Xi )2 N — Regression of Xi on X 2 S X 1X 2 " SX1 S X 2 N 2 sx2 - (sx2 ) 2 N r Correlation coefficient Geometric mean of the two regression coefficients - V /b2ibl2 16-inch Disks b 2l ■=• - 18-inch. Disks ^ 0.006376 b ^ 91 • 3*1-9 bl2 “ -“ * 0 .0 0 i|2 8 r - 0.930“ At %% level and 3 degrees of freedoms r ^ 0.878 0,20002 ^ 0.0081 2^.7087 bl2 - 0.20002 0.002025 - 98.7753 r ^ 0.895 At %% level and 2 degrees of freedom: r - 0.9?0 III. MECHAklbM OF WEAR PiiEROMEi\iA Ai\i> FACTORS Iio?LUk.wdIiw wE a H OF METALS liM SOIL The Mechanism of Wear Metallic wear can be divided Into two distinct categories 1 ) wear of metal against metal, 2 ) wear of a metal against a non-metal or abrasive wear. Wear of Metal Against Metal Although the techniques of grinding and polishing have Improved in the past few years, it is still a difficult task to prepare surfaces of appreciable size which are flat to within one hundred to one thousand Angstroms. Most of the surfaces used in engineering practice have surface irregular­ ities which are much greater than this. Since the range of molecular attraction is only a few Angstroms, we may expect that the area of Intimate contact will, even for the most carefully prepared surfaces, be quite small. lic surfaces are in apparent contact, When two metal­ the solids are supported only on the summits of the highest of these irregularities. The real area of contact has been found to be almost indepen­ dent of the size of the surfaces and is determined by the load W and the flow pressure Pm as indie 8 ted by Burwell (8 ) and is shown in the following relationship: A - w/Pm (1) Under the intense pressure at tue localized points of contacts 16 plastic defuriaauion and flow occur until the area is sufficient­ ly great to support the load* Therefor© in the above expression Pm is actually the mean oressure at a load where metal is plastic and yields. There are two principal ways in which wear occurs when one metallic surface is sliding over another metallic surface. the first method, In the asperities of one of the surfaces either catch on the asperities of the other, or these asperities may penetrate into the hollows of the mating surface, and the par­ ticles of either or both surfaces are torn or sheared off when the surfaces slide away from one another. In the second method of wear, there Is a real adhesion and welding together at the points of molecular contacts and thus formation of the so-called metallic junction. With stationary surfaces or at low speeds of sliding, this cold welding is produced by the intense pres­ sure in the region of contact and at higher speeds it is assisted by a high temperature softening or melting of the metal. It is the shearing, deformation, and plucking away of these metallic junctions which constitutes the physical wear of metala The frictional force, as discussed by Burwell (3) is in a large measure, the force required to shear these junctions as shown in the following expression: F - (5)(A) where 3 is the average (2) shear strength of metallic junctions and A is the true area of contact as shown in the previous 17 relationship, When equations (1) and (2) are combined, the familiar expression for the friction coefficient is obtained as follows: f - F/W = S/Pm (3) where S and Pm are plastic properties of the material con­ cerned. Wear of Metal Against Soil (Abrasive Wear) In the case of metal against metal, wear, as indicated above, is primarily a matter of formation and shearing of metallic junctions. In the abrasive type of wear, however, the existence of such phenomena has not been proved. Wear by abrasive is referred to the removal of metal by cutting or scrubbing action of the abrasive material. Wear of tillage tools in soil is the abrasive type of wear and particularly a sliding abrasion, In the presence of minor or infrequent heavy impacts which are not sufficient to work-harden the sur­ face of the metal but are extremely important in crushing, tear­ ing and chipping off minute metallic particles of the tillage tool. Because of the differences in physical properties and chemical compositions of the sliding surfaces, there probably exists here nothing akin to s eizure and welding together at the points of moleuclar contacts. The abrasives which are present in soil usually have rather very sharp corners so that very high local pressure occur when such corners move against the metal. 18 Factors Influencing Wear of Metals in Soil Apparently there are two distinct variable relationships in studying the wear of* metals in soil. complex and relates to soil, The first one is very dome of the factors met in this phase are soil physical and chemical composition, soil moisture, soil pressure, compactness, presence of roots, trash, rocks, and possibly chemical action of soil constituents. The second group of factors relates to the metals used primarily at the cutting edges of the tillage tool. hardness, strength dome of these factors are and resistance to impact, microstructure, chemical composition, heat treatment, grain size, surface fin­ ish, and direction of rolling. In addition to these two dis­ tinct groups of variables, operating conditions such as speed and cutting angle are also factors of considerable importance. Not all the factors mentioned above have been studied as to their effects upon the wear of metals in soil. these factors are Interrelated. Most of The effects of most soil var­ iables on wear can be represented, for example, by friction between the soil and the metal surface. By the same token, hardness and toughness of medals, which apparently influence wear to a great extent, are the results of certain chemical composition, microstructure, and heat treatment. Hardness seems to exert its beneficial effect by preventing the pene­ tration cf the abrasive particles of soil. cannot take a toehold, they just slide off. If such particles If they do get 19 a toehold, then toughness is the property that prevents a chunk being torn out* Soil friction, the amount and type of abrasive particles in soil, and moisture content of soil, certainly play a great part in the matter of getting a toehold. In the light of the preceding discussion and the available experimental evidence, the factors influencing weo.r of metals in soil can be divided into the three following categories: soil factors, 2) metal factors, 1) and 3) effect of operating conditions. Soil Factors Affecting Wear Soil friction* As mentioned above, soil friction plays an important part in the wear of tillage tools* The general laws of friction between soils and metals have been described by Nichols (22), (23) * (21^.) , (25) • He determined th© coefficient of kinetic friction by pulling a flat piece of metal being studied across smooth surfaces of soils uniformly moistened to various moisture contents. The amount of pull divided by the weight of the slid.er gave the coefficient of kinetic fric­ tion. This friction coefficient varied from about O.ip to 0.6 as the percent of water in soil was increased from oven-dry conditions to about twenty percent moisture. On the basis of this variation of sliding friction with soil moisture con­ tent, Nichol divided the frictional resistance Into the fol­ lowing four phases: 20 A. Compression phase. This is the condition when water does not adhere to the metal and when the bearing power or re­ sistance to compression of a soil is less than the pressure (the metal sinks in). In this phase, the friction coefficient’ varied with speed, pressure per unit area, and with t he smooth­ ness and material of the surface. This reaction is one com­ monly occuring in the tillage of sandy dry soils not containing any appreciable amount of colloids. B. Friction phase. When the bearing power of a soil is greater than the pressure per unit area but the soil water is still not sufficient to adhere to the metal, true friction is obtained. In this phase, friction was found to be proportional to the total pressure between the two surfaces, the colloidal content of the soil and roughness and material of the surfaces, but independent of the soil moisture content. C. Adhesion phase. This phase occurs where there is enough moisture oresent to cause the soil to adhere to the sliding surface. As the moisture films around soil particles become thicker, the moisture is held less firmly. Finally * there comes a point when the attraction of the metal is suf­ ficient to attach the moisture to it. The graphical represen­ tation of the moisture-friction curve showed clearly that be­ ginning with the C-phase, the curve begins to rise and reaches a maximum v alue at the boundary between C phase and D phase. This sudden rise was explained to be due to the attraction 21 force between metal and moisture film which must be overcome in sliding soil across metal. Since this phase involves the highest friction obtained in a tillage operation and high friction constitutes a waste inpower and causes excessive wear, the principles underlying the phenomenon of adhesion with emphasis on soil and metal surface were investigated by Kummer (21). The friction coefficient varies in the G phase with speed, area of contact, pressure per unit area, surface tension of the film moisture, surface and kind of metal, and percent of soil moisture contact. D. Lubrication phase. This phase occurs when there is enough moisture oresent to give a lubricating effect. The friction coefficient varies with speed, pressure per unit area, and surface and kind of metal. Beginning with this phase, the coefficient of friction decreases as soil moisture increases. Soils containing enough moisture to give this effect puddle badly and therefore no tillage operation is performed under this condition. It was this reduction of friction due to lubricating effect which led Crother and Haines (10) to the use of electric current on plow mold-boards to aid in scouring and reducing the draft in plowing. These workers thought that if a current was passed through the soil having the raoldboard of a plow as anelectrode, then the film of water formed at the soil metal surface should act as a lubricant and reduce the plow­ ing draft. Using this principle, large reduction in friction were obtained in laboratory tests, but tests in 22 the field showed that the magnitude of reduction obtained with this arrangement was too small to have immediate prac­ tical value. This work was carried out in 1921]. when the use of electricity on the farm was quite limited. With the pre­ sent developments in the field of electricity and the advent of such devices as the ,fElectralr* on the Farmall-l|_00 tractors, which places a powerful source of electricity at the disposal of the farmer anywhere on the farm, perhaps the idea of elec­ tric current on the plow moldboard is not too impractical and deserves further investigation. Soil moisture. The effect of soil moisture on the wear of tillage tools is to a certain extent, manifested through the soil-metal friction which was covered briefly in the pre­ ceding paragraphs. The ordinary range of soil moisture con­ tent for tillage operations is represented by "A”, "B11, and "C11 phases depending upon the texture, organic matter, struc­ ture, and other factors affecting the water-holding capacity of the soil. Zink jet al (38) it und that in their laboratory wear tests, the weight loss of the steel specimens in sand containing 1.85 percent moisture amounted to about twenty times that of dry sand, and the width loss about three times that of dry sand. This experience was confirmed by Tetsutaro et al (33) who found both under laboratory and field conditions that increasing moisture of sand resulted in increase of wear of their specimens. This wear reached a maximum at a certain 23 moisture content and then began to fall. This critical point was found to be about twenty percent for a loam type of soil and about II4. percent for a sandy soil (33)* point of maximum, the lubricating phase Apparently at this began to show up and the coefficient of soil and metal friction started to decrease upon further increase of moisture. Soil texture. The effect of soil texture on the wear of tillage tool was also studied by fetsutaro, e_t al. These Jap­ anese workers found that from the standpoint of plowshare wear the soils can be classified into three types: and clay. sandy, loam, Field tests of the ordinary Japanese plowshare, made from either white cast iron or gray cast iron, showed that in both cases, sandy soil proved to be the most abrasive type of soil. In the case of the g r a y cast iron, the clay-type of soil resulted in greater wear than the loam-type of soil, but vice-versa for the white cast iron. The following table shows the relative wear of white and gray cast iron in the three types of soils. The numerical values were roughly estimated from the graph in the original paper. Relative Wear Type of Soil White Cast Iron Gray Cast Iron 11 2$ Loam 7 12 Clay 5 15 Sandy 2 k The reason for clay being more abrasive than loam in the case of soft gray iron was not explained. It is known, however, that the hardest constituent of clay Is the aluminum oxide (Al203 ) which has an average Knoop micro hardness value (100gram load) of about 2,000 compared to a value of 820 for quartz as measured by Thibault and Nyquist (3^-). It is also known that metals consisting of micro-constituents softer than the abrasive will not resist abrasion as effectively as metals with micro constituent s harder than the abrasive. One of the micro­ constituents of the white iron is iron carbide with a Knoop hardness of about 800 to 1000. These hard microconstituents of the white iron are responsible for the better wear resistance of this material compared with the gray cast iron which contains soft graphite flakes in place of iron carbides. Examination of the above table shows that the gray cast iron was worn three times as much as the white iron in the case of the clay soil compared with a little more than twice as much in the case of the sandy soil. This greater relative wear of the two types of iron in the two types of soils perhaps can be traced back to the aluminum oxide constituent of the clay soil which is more abrasive than quartz, the predominating constituent of sandy soils. If the percentage of aluminum oxide in the clay soil were as much as the percentage of quartz in the sandy soil, wear. the clay soil would have probably resulted in the highest The reason that the clay soil was more abrasive than 25 the loam soil in the case of the gray cast iron can he now explained by considering the greater percentage of aluminumoxide in the clay and the low wear resistance of the gray cast iron* The latter factor was perhaps mainly responsible for the effect of aluminum oxide being magnified only in the case of gray cast iron* Metal Factors Affecting Wear Composition and heat treatment* Many of the physical properties of metals which have direct or indirect influence upon wear resistance of metals against soil are the results of composition and heat treatment* ture, hardness, Such properties as microstruc strength, and impact resistance, grain size, and depth of chill (in the case of white cast iron) are cer­ tainly dependent upon the original compositon and the subse­ quent heat treatments* Carbon, nickel, and chromium are the most important elements considered in the composition of metals intended to combat abrasive wear. Carbon is the cheapest and the most prevalent hardening element used in cast irons and steels for tillage tools. pite its importance, Des­ no scientific investigation has been con­ ducted to evaluate the role of carbon in the wear of tillage tools. Owing to this lack of information, there are few specifications for carbon composition of materials which meet with universal approval by the manufacturers of tillage 26 tools. During the first year of the plowshare experiments at Michigan State University, several cast iron plowshares were obtained for chemical analysis. The following table shows the percentage of carbon, silicon, and carbon equivalent (carbon plus one-third of silicon) of these shares. TABLE I ANALYSIS OF REGULAR COMMERCIAL CAST IRON PLOW shares Share Number PO-l PO-2 p o -3 PO-4 p o -5 during the Carbon 3.70 3*68 2.95 3.hk2.85 early stages Silicon 1.56 1.45 1.80 1.35 1.75 of the project Carbon Equivalent 4. 22 4.16 3.55 3.89 3.43 The results of this analysis showed that the cast iron shares available on the market are apparently either highcarbon shares (PO-1, PO-2, PO-tp or low-carbon shares (PO-3, PO-5). The choice of a low-carbon or a high-carbon iron in plowshares Is probably not based on any scientific data, but is either traditional or based on the experience of the par­ ticular manufacturer. For example, the casting of low-carbon shares, as indicated in the introduction of this work, in­ volved various difficulties, the principal one being ,!hot tears11 along the chilled cutting edge. One major company, due to tradition or personal experience, has preferred low- 27 carbon plowshares to such an extent that apparently it avoided the problem of "hot tears” by chilling only the point. This practice, naturally, has resulted in rapid wear along the cut­ ting edge of the shares made by this manufacturer. Another illustration on this matter of high-carbon or low-carbon shares was found in the following quotation from Bornstein (If.)* "There are some peculiar problems that arise from time to time in connection with wear of plowshares. About a dozen years ago, we were supplying a plowshare made of rolled high carbon steel for certain territories in South America. We were advised that these shares did not wear as well as some other shares sold in that locality. Examination of these shares showed them to be low carbon steel casti ng. We then supplied them a low carbon steel casting for plowshares and we were told that the results were satisfactory. However, we cannot understand, even to this day, why the higher carbon steel share did not give satisfactory wear." The importance of carbon as an alloying element and the lack of scientific data as to its effect in the wear of til­ lage tools were reasons for selecting carbon as one of the major variables in casting of the test pieces for this project. The combined effect of chromium and nickel upon the abra­ sive wear resistance of cast iron has been studied by several workers. The nearest example to the wear of tillage tools was found in the studies of wear-resistant materials for the cera­ mic Industry. tires, Vanick (35) reports the use of Ni-H&rd on muller scrapers, knives, and augers in tne ceramic industry. Presence of nickel hardens and strengthens the matrix in 28 cast iron by depressing the formation of soft oearlit©. But since nickel is a graphitizer, it will act as a softening agent if added to excess. This tendency must, in order to in­ sure hardness, be balanced by the simultaneous addition of chromium or some other carbide-forming elements. The applica­ tion of this principle has been used in the development of a patented material known as Ni-Hard. The Ni-Hard which was tested for apolication in the cera­ mic industry was a hard, that was toughened, abrasion-resisting white cast iron strengthened, and hardened by alloying with approximately Jp-J- percent of nickel and lp percent chro­ mium (35). The pearlitic matrix was changed to martensite which substantially increased the hardness and resistance to abrasion. By an appropriate low-temperatur© (i|.00 to I4.50 d e ­ grees Fahrenheit) heat treatment, the metal was tnen strengthened and to ughe ne d . As to the direct application of Ni-Hard in the tillage tool industry, Sanders ejt al (30) report their experiences as follows: "During the autumn of 19i+-7* plowshares made of chilled cast iron became very scarce because of an excessive num­ ber of breakages due to the hard and dry condition of the soil. The microstructure of these plowdiares consisted of cementite and pearlite which break off or wear away leaving a soft background which rapidly disappeared in service. As a solution to this problem, Ni-Hard type of plowshares were used. The structure of these shares consisted of a carbide— martensite formation throughout and gave approximately twice the life with considerably fewer failures due to breakage. ff 29 The superior wear resistance of manganese steel in bull­ dozer blades and grader blades has drawn the attention of some workers to the possibilities of manganese as an alloying ele­ ment in materials of tillage tools, especially the plowshare. Ransome (27), in a renort to the Institution of British Ag­ ricultural Engineers, has indicated that his experience with manganese steel plow^iares has shown that manganese steel will wear just as much as a plain carbon steel share. This recent experience confirmed the results obtained earlier by Hoffman (see cage 6 )» The manganese steel used in rock crushers, bulldozer and grader blades, cower shovel buckets, and similar equipment, which are e;xposed to heavy impacts in service, are usually austenitic steel of 12 to Ilf percent manganese. Under he avy impacts or pressures, austenitic manganese steel work-hardens on the surface and resists wear excellently. The fact that the wear resistance of this material when used in a plow share has proved to be inferior or at the best only equal to that of plain carbon steel, indicates that perhaps the minor or moderate shocks which exist in ordinary plowing operations are not sufficient enough to work-harden the soft austenitic steel* Hardness* The combined effect of such factors as chemical composition, heat treatment and cooling rat© of metal is intro — 30 duced in a physical prcmerty known as hardness* Attempts have been mad® from time to time to correlate hardness and wear resistance. In general, some correlation can be obtained between hardness and the loss of weight in service. certain conditions, however, breaks down. Under this relationship definitely Study of literature on this subject of corre­ lation between hardness and wear in service indicated that for similar materials and treatments, it Is possible to estab­ lish a correlation in specific instances. However, it is not possible to choose two irons of different composition and treatments and attempt to correlate their wear resisting properties on the basis of hardness tests. Bornstein (ij.) states his experiences in this roatter as follows: "In the case of steels, we have frequently found a lower Brineil hardness material to give us better wear resistance than a higher Brineil hardness material. How­ ever, in this case, there was s considerable difference in the composition of the materials. For the same mater­ ial, we have usually found the higher Brineil hardness to give the better wear resistance. In the case of gray cast iron, w© have found the structure to be Important and hardness only one factor pointing towards the wear resistance. In the case of white cast iron, we have found the higher hardness to be disirable in resisting we ar . 14 In the wear tests of the nine types of steel disks at the U.S.D.A. Tillage Laboratory, Reed and Gordon (28) found that their data for hardness of the disk blades failed to correlate with th© results of the field and laboratory tests. The results of th© laboratory wear tests by Zink e_t &1_ (38), 31 however, correlated rather closely with the hardness tests. In this case, th© steel specimens were cut from th© same bars but were given different heat treatments. The high degree of hardness seems to exert Its influence, as indicated before, by preventing an abrasive from getting a "bite”. If such particles, by reason of inherent hardness cannot get a grip, they simply slide off; hence the value of hardness. If however, they penetrate the metal surface, the property of ductility becomes effective because the surface will flow under Impact and will not crack and chip off as with hard material. (7) states, On this property of plastic flow Brown "Conditions exist that even a mild steel will give a surprising degree of wear resistance due to its prop­ erty of plastic flow. n To summarize th© effect of hardness on wear of metals, it appears that although hardness is of major importance in relation to wear resisting properties, it is not an index of the wear resistance of a material when used In service. If hardness tests were an indication of relative wear of a material in service, there would have been no need for all the complicated wear devices which have been developed for each specific wear problem. Microstructure♦ As mentioned before, microstructure is the result of composition and heat treatment. As far as th® 32 wear of metals in soil is concerned, it apoears that the prooerties associated with microstructure may be divided into three parts: (a) the properties contributed by the harder ohases such as martensite and cementite; (b) th® properties contributed by the softer phases such as ferrite and flake graphite;'and (c) properties dependent upon the the physical constitution of the aggregate; that is, the form, size, and distribution of the harder regions in the matrix and their relative volumes* In general, it seems that an important requirement of the harder constituent and the matrix, if possible, is a hardiness level greater or at least equal to that of the most dominating constituents of the soil. No general rule can be given for a microstructure to be satisfactory for all conditions of soils. In a sandy soil, In the presence of rocks and stones, a microstructure con­ sisting of hard cementite and a tough oearlitic matrix will probably be the most desirable structure. On the other hand, for an abrasive soil where shocks are minor and infrequent, a microstructure consisting of carbide-martensite throughout will perhaps be preferred to any other structures. perience of Sanders The ex­ et al (30) with Ni-Hard plowshares, as quoted on page 28 is an example of a desirable microstructure for certain soil conditions. 33 Effect of Operating Speed and Cutting Angle The effects of operating speed and cutting angle were studied by Tetsutaro et al (33) i*i their laboratory studies of the wear of steel against soil* These investigators found that as velocity was increased from J^3 centimeters per second to 127 centimeters per second, the wear of steel specimens increased from twenty milligrams to sixty milligrams* The increase of speed also resulted in increase of cutting re­ sistance fromone-half kilogram to about two and one-half kilograms* They also found that as the cutting angle was increased fronfpO degrees to 80 degrees, the wear was decreased by three times while the cutting resistance was increased by four times. The decrease of wear with increasing in cutting angle was attributed to the fact that the sand velocity on the wearing surface decreased as the cutting angle increased. They concluded that the element which gives the wear phenomenon is not the cutting resistance but the load in the normal dir­ ection to the wearing surface of the metal. As velocity ard cutting angle increase this normal load and thus wear in­ creases linearly* The work of these Japanese scientists seems to be in agreement with the finding of Nichols (25) on the friction between soil and sliding raetals. Nichols found, as discussed before, that the coefficient of friction between metal and % soil increases as speed increases. This increase of co­ efficient of friction and thus force of friction seems to he the reason for the increase of wear due to increase of speed. 35 IV. DEVELOPMENT OF A METHOD FOR LABORATORY AND FIELD wEAH TESTS Owing to the limitations and difficulties involved in evaluating the wear of tillage tools in the field, as men­ tioned in the introduction of this paper, it was decided to conduct the experiment in three stages; where a wear test machine could be used; (1 ) laboratory tests, (2 ) small scale field tests, where a large number of test specimens could be subjected to the same tilling conditions in a comparatively short time; and (3 ) a large scale field test, where actual plowshares would be subjected to actual tillage practice. Most of the writer*s efforts were focused on the first two stages of the experiment. The field test data which were obtained previously from the wear of the plowshares cast at the Michigan State foundry were analyzed and used for the third stage of the experiment. Development of a Laboratory Wear Test Machine Requirements for a Laboratory Wear Test Machine As discussed in the preceding chapter, wear of tillage tools in soil is the resultant of the material itself, soil as the mating material, the and the operating conditions. Ommission of any of these three important elements in the force polygon of wear gives a resultant of wrong length and 36 pointing in the wrong direction. We find the literature full of wear test methods apparently motivated by the idea that wear resistance is an inherent property of metals, measurable by almost any way of tearing off particles. G-illet (15) states, As "Wear resistance is not an inherent property to be measured in a bsolute units. exist apart from the conditions of service." It does not The purpose of a wear test is to evaluate wear resistance in use. Since uses vary widely we can have no universal wear test but in­ stead must have many types of wear tests each corresponding truly with the specific conditions of use. Again as Gillet puts it, "a universal wear test is in the same category as a perpetual motion machine" and further "to obtain wear test results that mean anything about the performance of a metal in service, that test must be conducted under conditions which duplicate, as nearly as possible, the conditions in the field." A useful wear test method must produce wear by the same type of attack as occurs in service. The authorities on wear agree (15) that the appearance of the specimen at suc­ cessive stages of wear in the test ought to be indistinguish­ able from that of the same material at the same stage in ser­ vice. The test must give some response, within a reasonable range, on duplicate tests of oruly duplicate samples. And, the rate of wear should be measurable by accurate weighing 37 or some indirect measurement. With, the above factors in mind, a comprehensive study was made of the various laboratory devices used for deter­ mining abrasion wear resistance of metals. Devices for Laboratory Abrasion Tests Haworth (17) has classified abrasion-testing devices into the following groups: Fixed abrasive. Typical of this class are endless belt grinders and grinding wheels. The principal disadvan­ tages are that (a) grinding takes place at a uniform rate with no recognition of heterogeneity of the structure which, in loose abrasive applications, may result in undercutting of hard constituents in relatively soft matrices, and (b) the methods lack versatility since abrasive types and grits are limited to those available in bonded form. Sand blast. This method of test, according to Rosenberg (29) does not differentiate between many materials which are known to have widely different wear-resisting characteristics. The air pressure in this type of test was found to be very critical. The diameter of the sand-blast nozzle was also difficult to control. Therefore, variations in line pressure and wear of the blast-nozzle exercise significant effects on the results obtained. When pieces of chilled iron and gray iron were subjected to a light flow of sand under no pressure 38 other than the force of gravity, it was found that one metal was about as resistant to wear as the other. With a 30 psi air blast pressure no difference was noted between a chroniiummolybdenum steel, a 12 percent manganese steel, and a white iron* When air blast pressure was increased to 60 psi, no difference was again noted between the steels, but the white iron showed a decidedly higher loss of weight. Ball mill* In this method of wear test, specimens are ground and lapped to form one-inch spheres. The specimens are placed in a cylinder containing 3and or other abrasive which revolve about its longitudinal axis at a predetermined speed. Rosenberg (29) found that different results were u- sually obtained on the same material during repeated tests. In order to obtain some idea of the trends shown by the tests, he averaged the last ten 2 l4.-h.our runs with each abrasive. He found that the differences between various steels were not very marked. Also, contrary to what might have been expected, in this type of test, the sharper sand caused less wear than the rounded sand. Beside the lack of sensitivity and the slowness of the tests, ball mill tests require specimens of definite shapes that cannot be produced readily from some of the better ma ­ terials used for tillage tools. Abrasive lap* Blake (3) and Weiss (3 6 ) used the abrasive lap type of tester to determine relative wear of metals due 39 to abrasion. Their devices were similar and consisted es­ sentially of specimens loaded against a circular track ro­ tating in a mixture of water and abrasive. The principal comment on the data is the relatively small difference among some of the materials tested. For example, Avery (1) who tested a number of materials with the device described by Blake, found that white iron and 3AE 1020 steel had approxi­ mately the same degree of abrasion resistance. In the device used by Weiss, silicon carbide produced a vastly accelerated test but it was felt that its results cannot always be used with complete assurance for applications involving less ab­ rasive material. The wear test method developed by Zink ej; al (3^), may also be classified as an abrasive-lap method. As discussed in detail on Page 9> this method has several disadvantages and is probably not a desirable wear test method for materials of tillage tools. Brine11 rotating wheel. The fifth classification of abrasion testing device is represented by the Brinell abra­ sion testing machine. The principles of the Brinell wear test machine were discussed briefly on Page 11. Although none of the devices described previously were felt to be satisfactory for determining wear resistance of materials for tillage tools* the Brine3.1 abrasion machine appeared to approach the required type of abrading action and to have 2*0 merit both as to simplicity of design and operation. In view of the simplicity of the machine, its effectiveness, and the rapidity with which wear tests could be made, the Brinell principles, with some modifications, were used in developing the laboratory wear test machine used in this study* It should be kept In mind that the abrasive laboratory t© 3 t does not provide for the effect of occasional impacts experienced by the tillage tools in the field. To select a material for tillage tools, the results of abrasive laboratory tests must be coupled with information gained from Independent impact tests. Design aftd Construction of the Wear Test Machine Make-up and essential parts of the machine* In devel­ oping the wear test machine, the spindle of a milling machine was used to drive a steel disk one-half inch thick and four inches in diameter (Figures 1 and 2). The disk was covered with a soft rubber belt to form the abrasive wheel. belt was changed after each twenty tests. The Dry sand flowing down the hopper was carried between the rubber face of the rotating wheel and the test sample. A load, hanging down the milling machine table, held the specimen in contact with the flowing sand and the rotating wheel. A vacuum cleaner blower was used to carry the sand and the dust away from the rotating wheel by means of a carriage mounted on four ball 41 bearings and sliding in and out in the slots or the milling machine table. A plastic window provided a means for watching the flow of sand. In this way, a uniform flow of the abrasive particles was assured by manipulating a control gate under the hopper. The loss of weight was determined by weighing the specimens before and after the test on an analytical balance reading to one-tenth of one milligram. The use of a rubber belt around the abrading disk was found to have a number of advantages over the steel wheel used by Brinell. 0 One advantage was the lack of high pressure on the in­ dividual abrasive particles, since the resiliency of the rubber permitted the particles to imbed in the face of the rubber belt. When a steel disk was used with sand, the lat­ ter was pulverized as it was carried between the specimen and the steel disk* In tests with the rubber belt no ap­ preciable crushing of the sand particle was observed. Another advantage of the rubber belt was that, owing to the resiliency of the rubber, the Initial contact area between the specimen and the rubber-faced disk increased only slightly during the test. The unit pressure on the specimen, therefore, was relatively constant throughout the test. With a steel disk, on the other hand, the area of contact between disk and specimen gradually increased and the unit pressure correspondingly decreased. Pig. 1. General view of the wear test machine. Pig. 2. Close-up of the wear test machine showing the carriage, the rotating disk with the rubber band, the specimen held in place, and the plastic window in inset. 1+3 In addition to reducing the crushing of sand and pro­ viding a uniform loading, the rubber belt was found to give a cushion action to the sand particles striking the surface of the test specimen. This cushion action, which was similar to the action of soil on the surface of tillage tools, produced a wear appearance of considerable Improvement over that of the steel disk used without the rubber belt. In order to set up standard conditions for testing of all the field-tested specimens, the effects of several var­ iable factors were determined as discussed in the following sections. Effect of speed and heat developed by friction. When sand slides over the surface of the iron specimen most of the work done against the frictional force opposing the m o ­ tion will be liberated as heat between the surfaces. If all the frictional work Is liberated as heat, then the rate of heat developed is given by: Q, - fWv/j where f is the coefficient of kinetic friction, W is load on the sand particle, J is the mechanical equivalent of heat, and v is the velocity of sand sliding over the surface of the metal. At high velocities and heavy loads, this frictional heating causes surface melting and surface flow and conse­ quently rapid Increase in wear. Haworth (1?) measured the kb temperatures developed by means of a thermocouple spotwelded to the bottom of a groove machined on the back of the specimen. Tests made at 250 revolutions per minute and 20 pounds of load resulted in a temperature of about 2 l|0 degrees Fahrenheit for a linear distance of 500 feet. Since these operating conditions were very close to the standard conditions used in the present work, for all the laboratory wear tests, the writer assumed that development of such small heating was prob­ ably not sufficient to cause appreciable amount of surface melting and surface flow of the metal being tested. More­ over, since all the specimens were subjected to the same condi­ tions of testing, they were equally affected by the same amount of heating. Effect of length of run» It was found that the loss of weight was directly prooortlonal to the length of the test. Although a longer period of each test resulted in greater loss of weight and therefore more reliable test results, it required a great amount of sand. The work and the time in­ volved in drying and sieving the sand plus the time required to run individual tests on the 172 specimens would have slowed down the experiment considerably* After several trials a minimum time of 1 -g- minutes was chosen as the standard time for all comparative tests. In order to simulate the field conditions as nearly as possible, the speed of the rotating disk was chosen at 46 harder materials, such as white cast iron, heavier loads were required in order to obtain a measurable weight loss during the short period of testing. If the test period had been increased, as mentioned before, a considerable amount of sand would have been required and the tests would have been slowed down. For this same reason, Brinell (6 ) applied different loads and test periods with different materials # under investigation. The obvious disadvantage of his method was that the hard and the soft materials could not be clas­ sified according to a definite wear scale. For the study of the wear resistance of materials of tillage tools, the es­ tablishment of such a comparative wear scale was essential* In the light of these factors, after a considerable number of trials, a twenty pound load was chosen as the standard for all the comparative tests. Effect of size of the abrasive particles. Experiences by other workers have proved that the size of the abrasive particle is a factor of major Importance in obtaining re­ producibility of results in duplicate tests. Finer particle si2es present a greater multiplicity of cutting edges for an equal volume of abrasive and therefore cause more rapid wear than larger particles. In all the comparative wear tests in the present work, dry mortar sand was the abrasive material. After drying, the sand was screened and only the fraction that passed through a 2 8 -mesh screen and over a 1 0 0 -mesh 45 275 revolutions per minute, corresponding to a plowing speed of about 3i miles per hour. When the four-inch wheel with its one-eighth of an inch thick rubber belt was operated at this speed, a point on its circumference travelled a lineal distance of I4.6 O feet in l| minutes. Effect of load. The following table shows the effect of increasing load upon the weight losses of strips of struc­ tural steel one-eighth inch in thickness in ten minutes at 275 revolutions per minute. TABLE II EFFECT OF LOAD ON WEIdHT LOSS OF STEEL 3PECIflEN3 TESTED ON THE WEAK TEST MACHINE Specimen 1 0 3 k 5 6 7 Lo ad (pounds) 10 15 21 25 30.75 35 ko Bel*ore 3 8 .0 1 1 0 39.9920 3 9 .8614.0 3 8 .5 6 2 0 37.854° 39.7822 3 9 .6 2 0 8 Nfsight (grams) After 3 7 .3 5 4 0 3 9 .7 8 2 2 3 9 .6 2 0 8 3 8 .2 6 0 6 3 7 .4 7 3 5 3 9 .2 8 5 7 — Loss 0 .1 5 7 0 0 .2 0 9 8 0 .2 4 3 2 0 .3 0 1 4 0 .3 8 0 5 0.4965 — As seen from Table II the loss of weight of a test sample Increased as the weight holding the sample against the abrading wheel increased. At forty pounds of load, the rubber belt could not stand further pressure and either came off the wheel or was torn by the high pressure. lAihen the load exceeded even twenty pounds, difficulties were exper­ ienced in running the comoarative wear tests smoothly. For k7 screen was used. Four tests were made on Tour different positions on each specimen, and the mean of the four tests was reported as weight loss for that specimen* Use of sand in the range of 2 8 -mesh to 1 0 0 -mesh particle size resulted in weight loss values which deviated not more than ten per­ cent from the mean of the four tests. Table III shows typical values obtained for several of the specimens in each of the four replicated tests. As seen from Table III, the maximum percent deviations from the mean of four tests for white cast iron Specimens were higher than those for the gray cast iron. This higher percent of deviation could be expected because of the small weight losses obtained when the hard white iron specimens were subjected to wear test. The percent deviation could be reduced by choosing &. narrower range of particle size for the abrasive material. A series of tests made on white iron using dry sand with particle size in the range of 2 8 -mesh to JpO-mesh resulted in maximum percent deviations not exceeding five percent of the mean of the four replicated tests. Pre­ paration of sand with this narrow range of particle distri­ bution, however, required a considerable amount of time. In­ asmuch as the weight loss values used in correlating the r e ­ sults of the laboratory tests with those of the field tests were the mean of I4.8 tests In the case of white iron and Ij.0 tests in the case of gray iron specimens, the degree of TABLE III TYPICAL WEIGHT LOSSES IN MILLIGRAMS OBTAINED IN EODR REPLICATED TESTS ON EACH CAST IRON SPECIMEN, Spec­ imen Type of Iron le s t IviO *► II Ill IV Maximum % Deviation From Mean 9.0 7. S 6.9 7.5 8.30 9.65 White C12 6.7 6.9 6.5 6.0 6 .5 3 8.12. Cast Elj. 9.6 9.6 8.6 8.2 9 .0 0 8.90 Iron J6 5.5 5.5 6.0 6 .14. 5.85 9.1+0 PI4. 3.2 8.1+ 9.14. 8.65 8.66 T j ----- .... B7 i+5 .o 1+5.0 ; I4 4 .6 14-6.3 00 00 Cast i I . £ Gray . I 1 Ai2 CO • I Mean of Four Tests 2 .3 8 B5 59.9 63.3 6 3 .8 5 9 .1+ 61.60 3.57 FI 35.6 37.0 3k-k 3 7 .7 36.17 1+.90 1+9.9 1+8.5 ! lj.8.8 1+8 . 2 1+3.85 2.15 37.5 36.0 : 37.8 3 6 .5 36.95 2.30 Iron RIO K9 reproducibility shown in Table III was thought to be satis­ factory for practical purposes in correlating laboratory and field data* Standard conditions for laboratory tests* The standard test conditions for comparative wear tests by means of the wear test machine consisted of a wheel speed of 275 revolu­ tions per minute (peripheral speed about 3 J- miles per hour), twenty pounds of weight, and a time period of 1-J minutes. During this period the surface of the specimen was subjected to a normal load of twenty pounds and a path of abrasive 1|_60 feet in length. As the rubber belt wore out, obviously the distance traveled in 1^ minutes decreased* At the end of twenty tests, the rubber belt was worn by one-sixteenth of ail- inch. Simple calculations showed that in order to have all the specimens subjected to approximately the sarae number of revolutions of the abrasive wheel, the test period was to be increased by one second at the end of each ten wear tests. At the end of each twenty tests, the rubber belt was discarded and a new belt was installed. This procedure was thought to give results sufficiently accurate. however, For more exact work, a counter could have been Installed on the machine, and the number of revolutions increased by one revolution for each two tests. The sand must be dry for use in the wear test machine. If sand is not completely dry, it may cake together in the 5o supply tub© and may cut off, partly or entirely, the re­ quired continuous stream of sand supply* At the beginning or each test the sand hopper was filled with dry mortar sand (100-28 mesh Tyler Standard). Although the crushing of sand was not as serious as in the case of the steel wheel without a rubber belt, nevertheless the sand was used only once in order to have all the specimens subjected to the same abrasive of the same particle size distribution. Development and Casting of the Experimental Plow Points Selection of the Irons Ordinary cast iron. Cast iron was chosen as one of the test materials for the following reasons: 1. In Michigan, because of abrasive nature of the soil, about three cast iron plowshares were used for every steel plowshare used. Any finding in connection with re­ sistance of cast iron against wear in soil would bring about a considerable saving in plowshare investment by the farmer. 2« The specimens could be cast at the University foundry where the direct supervision of persons directly interested in the project could assure care In obtaining uniform composition and treatment of all the similar spec­ imens* 51 3. If specimens were made from castings, the size and shape oT the specimens could be chosen such that the same specimens could be used both in the field and in the labora­ tory. It was believed that this arrangement would result in a better correlation between field and laboratory tests. The ordinary cast iron specimens included (a) low car­ bon unalloyed, (b) low carbon alloyed, alloyed, and (d) high carbon alloyed. cast in both chilled and unchilled. (c) high carbon un­ These specimens were The principal alloying elements were nickel, chromium, and molybdenum. The effect of nickel and chromium upon wear resistance of cast iron was discussed on Page 27* Molybdenum is a beneficial element in cast Iron in that it refines the grain structure of the iron and improves tensile strength and impact resistance. Nodular or ductile cast iron. The cast iron industry has continually attempted to improve the properties of cast iron through advances in melting and heat treatment practices and through the use of alloys and inoeulants. Investigators have devoted their efforts towards finding a simple ladle ad­ dition to gray iron vihich would change the flake form of graph­ ite to some other pattern, preferably to round shape such as that of temper carbon in malleable Irons. It was thought that this process would result in overcoming some of the brittleness inherent in flake graphite structure of gray cast iron, would 52 strengthen the iron, and possibly render it ductile. The International Nickel Company in 19lq9 (II4.) announced the discovery of such ladle addition. They found that mag­ nesium, although insoluble in iron, when properly added to cast iron would convert the graphite flake to a spheroidal form on solidification. strength, This process Imparts ductility and and increases the rigidity of the part in the as- cast condition. Figures 3 &nd 4 show the transformation of gray iron (flake graphite) to ductile cast iron (spheroidal carbon) when gray iron Is treated with magnesium. These pictures reveal the reason for the superior impact strength of the nodular iron. In this type of iron, the graphite in­ clusions are in the form of spheres which do not interrupt the continuity of the matrix as do the flakes in ordinary gray iron. dray cast iron in the untreated condition will develop a tensile strength of not over 3 5 *0 0 0 pounds per square inch. After proper conversion of flake graphite to spheroidal car­ bon by the addition of magnesium, the resulted nodular Iron will develop tensile strengths in the neighborhood of 8 0 ,0 0 0 to 90,000 pounds per square Inch, a yield point of 55*000 to 70,000 pounds per square Inch, and elongations which fall somewhere between three and eight percent as cast. The duc­ tile iron also can develop a chill surface of controlled depth by casting against a metal chill. By suitable composition 53 F ig * 3- Ordinary gray iron with graphite in the f orm of flakes. Fig. l+l Nodular iron with form of nodules or spheroidal carbon. the Sk control, a carbide abrasion-resistance surface with a mod­ erately ductile backing metal can be obtained. Proper heat treatments toughen the chilled surface and improve the duc­ tility of the unchilled portion. It is claimed that surface hardnesses of 50 to 60 Rockwell G in the chilled area can be readily obtained with two to ten percent elongation and a tensile value of about 8 5 ,0 0 0 pounds per square inch in the backing metal (Uf). One manufacturer of plowshares has taken advantage of these characteristics of ductile cast iron by applying it to his chilled plowshare production. Since its introduction, the Company claims many thousands of plowshares were sold with an unconditional guarantee on breakage (26). Stott (32) reports on the application of nodular iron In sprockets that are used on a trenching machine. Up to the time that ductile iron was adopted, the manufacturers had tried many materials, but were always faced with breakage if they heat treated for maximum hardness, or wear, if sufficient toughness were maintained. Since these sprockets have been changed to ductile Iron, this condition has been remedied. Wearing quality on the sprockets has increased to the point that the chains are wearing out faster than the sprockets. The manufacturers have worked out a heat treatment of differential hardening so that the teeth are about 60 Rockwell C while the hub is still machineable. 55 The advent of this new ductile cast iron, which seemed to combine the properties desired in a cast iron plowshare, such as chilling tendency, ductility, and impact resistance, was the reason for choosing the nodular iron as one of the test materials. Investigations for a Test Specimen of Suitable Shape and Size Figure 6 (first row) shows the original aluminum pattern on the final casting which was designed to be used for wear experiments both in the field and in the laboratory. The dis­ advantages of this original design were as follows: 1. The specimens were too heavy and the weight losses could not be measured accurately by means of an analytical balance having a maximum capacity of 200 grams. 2. The size and thickness of the specimen were such that complete chilling through the casting could not be assured. The thickness of the specimen had to be designed such that the microstructure of the casting revealed no soft ferrite in the chilled specimens and no massive cementite in the unchilled specimens. 3. The extension in the steel bolt from the face of the casting proved to be a problem in evaluating the weight loss due to the cast iron specimen alone. If a loose bolt were used in connecting the specimen to the shank of the tillage tool, the head of the bolt would have introduced 56 problems in scouring of the plow point* When the steel bolt was cast right into the specimen and the end was ground and smoothed off with the surface of the specimen, the ground end of the bolt wore out more or less than the body of the speci­ men, depending whether the specimen was made of chilled or unchilled iron* After a few other preliminary designs, the final design shown in Figure 6 (second row) was chosen for the miniature plow points to be used in all the experiments. Eight aluminum patterns were made from aluminum bars which were cast at the University foundry. cut and machined to the final shape. The bars were The finished pattern and the cast plow points were 2 7 /8 inch long, one inch wide, and 3/8 inch thick. The weight of the specimens including the bolt was 190 grams. Thickness of the specimens was a d ­ justed to stand the impact encountered in the field, and to permit complete and uniform chill throughout the sample. The 5/1 6 inch stove bolts were cast deep into the casting to about one-eighth of an inch from the face of the specimen. This technique eliminated the problems of scouring and uni­ formity of wear, because of the exposed head of the steel bolt. The flat head of the stove bolt served as a support for keeping the bolt in the sand mold. finishing, the heads were cut off. After shake-out and 57 Fig* 5* Groups or unchilled (top) and chilled (bottom) cast iron plow points before cutting and finishing. Fig. 6 . Original and final design of the patterns and castings for the chilled iron specimens. Second row from left to riglt : Pattern, as-cast, ready for field test, after field and laboratory tests. 58 Casting of the Miniature Plow Points After the aluminum patterns for the miniature plow points were designed and constructed, a few other equip­ ment were necessary before the casting of the specimens could be attempted. Two chill blocks had to be made for the casting of chilled specimens. Figure 7 shows the wood chill pattern and the machined castings used as one of the/blocks. These chill blocks formed the drags of the molds and oroduced eight specimens per mold at a time with one heat. A wooden match plate was constructed to make molds for pouring eight unchilled specimens at a time with one heat. Figure 8 shows the match plate with patterns out of position. Preliminary plate experiments with this match/however, proved that, owing to the sharp points of the miniature plow points, gray iron casting could not be made without having the points chilled. To avoid this trouble, unchilled castings were made in the form of 1 2 -inch bars as shown in Figure 5* These bars were then cut and machined into four miniature plowpoints. Because of the lack of special facilities at the Uni­ versity foundry, the nodular iron specimens were provided through the cooperation of the Foundry Section of the Manu­ facturing Research Division of the International Harvester Company in Chicago. The Section furnished lOij. nodular iron miniature plow points complete with data covering the chem­ ical composition, physical properties, and photomicrographs. Fie. T • Wood oattern and one of the machined castings used as a chill block. Fig.8-‘* Match plate used originally for cast­ ing of the unchilled iron specimens. 60 They also furnished the unchilled nodular iron bars for making the unchilled specimens. The remainder of the specimens were made at the Michigan State University foundry. An induction furnace, holding 30 pounds of metal, was used in melting and pouring the various kinds of irons. A total of ten heats were made to pour all the required specimens. Table IV shows the chemical analyses of all the irons used in the ordinary and nodular cast iron miniature plow points. The identifications for the various types of irons used In this Investigation are shown in the following: Type of Iron Chilled irons Low carbon unalloyed High carbon unalloyed Low carbon alloyed High carbon alloyed High carbon nodular alloyed High carbon nodular unalloyed Low carbon unalloyed High carbon unalloyed TJnchilled Low carbon alloyed irons High carbon alloyed High carbon unalloyed nodular High carbon alloyed nodular Identification A C E J -P Q, B U F H H S Examination of the Castings and Preparation of the Test Pieces for Laboratory and Field Tests Visual examination. Figure £ shows the castings after shake-out and sand blasting. The unchilled bars were frac­ tured at a point which was farthest from the gate and which POINT 61 U\ ON rH C FLOW UN MINIATURE o o o O on O rH o o vO ON O CO ON O o H o o I —I o I —I O USED o CM OF IKONS ANALYSIS CO o on •H co rH CHEMICAL vO O IN THE CASTING OF THE VO UN CVJ O • • o o O o CO ro o O rH O CO Ch O O U t-H <—1 CM CN ON vO O rH UN o vo c\l rH rH CO UN vO ro o o UN I —I o o UN CM H r—i rH o O O •H O C\J CO vO o o o CM C-~ co t"— ON o o O CO o CO O vO H o o o UN ON o co C"- CM V0 GO C— o ON CO o on O ( —I CM CM O vo ON ON CM CO vo co ON CM CO O ON vO rH CO -Tf CM rH ON o CO o 62 had the smallest section size* II any white section was present in the casting, the whole set was discarded and a new heat was made with a charge balanced to give an iron with a lower chilling tendency. The chilled castings were also fractured and examined for the presence of gray iron in the heavy sections* Microstructure studies* In addition to visual examina­ tions of the fractures, representative samples of each heat were sectioned and polished for the s tudy of the microstruc­ ture. If any abnormal type of graphite flake or traces of massive cementite was present in the sample of the unchilled casting, the casting was discarded. The typical microstruc- tures of the irons used in this study are shown in the pho­ tomicrographs on the following pages. The unchilled alloyed nodular iron specimens furnished by the International Harvester Company contained traces of massive cementite and thus could not be used for comparative wear studies. A ohotomicrograph of this iron, which was identified as "S” In Table IV, is shown in Figure 20. The study of the microstructure of the cross sections of a number of specimens showed that the structure was uniform from the outside surface miniature plow noints. quite to the inside surfaceof the This study assured a uniform structure of the experimental plow points from one side to t he other throughout the thickness of the casting. 63 • '-Vvi'-.* Pig. Iron A (low carbon unalloyed), chilled sample at 150X . Pig. 10 Iron E (low carbon alloyed), chilled sample at 150X, Pig. 11. Iron C (high carbon unalloyed), chilled sample at 250X. ’ Pig. 12. Iron J (high carbon alloyed), chilled sample at 2£0X. 6l± Pig. 13. Iron P (high carbon unalloyed nodular), chilled sample at $00X. Pig. II4.. Iron Q, (high carbon alloyed nodular), chilled sample at 500X. Pig. 15. Iron B (low carbon unalloyed), unchilled sample at 3 $ 0 X, Pig. 16. Iron P (low carbon alloyed), unchilled sample at 350X. 65 Pig. 17* Iron D (high, carbon ■unalloyed), unchilled sample at 150X. Fig. 19. unalloyed sample at Iron R (high carbon nodular), unchilled 150X. Pig. 18. Iron H (high carbon alloyed), unchilled sample at 150X. Pig. 20. Iron 3 (high carbon alloyed nodular), unchilled sample at 150X. 66 Grinding and finishing of the specimens. One or the major problems encountered in finishing the test pieces was the presence of pittings on the surface of castings, especial­ ly the chilled samples. These pittings not only reduced the effective surface area for wear, but also provided cavities for collection of particles of soil and thereby affected the net weight loss. Consequently, all the test pieces had to be surface-ground to a uniform depth sufficient to eliminate the pittings. Study of the microstructure of the ground specimens showed no sign of martensite formation due to over­ heating as long as the samples were kept cool during the grinding operation. Another problem was the rusting of the specimen due to the great tendency of csst iron to corrosion. Because of this problem, a rain of weight was detected when the speci­ mens were left out in the air. Therefore, it was necessary to keen the samples in an oven at 212 degrees Fahrenheit af­ ter they had been surf ace-ground, marked, and weighed. This technique proved to be effective. Adaptation of a Cultivator for the Accelerated Field Tests A Ferguson N-KO-21 cultivator was adapted for field testing of the miniature plow points (Figures 21 and 22). One-f-inch steel bars were used in place of the shank of the Fig. 21. Side view of the cultivator adapted for field tests. Fig. 22. Right side of the cultivator show­ ing the five shanks used for the 5 x5 Latin square design. 68 cultivator sweep. The steel bars were machined and shaped at the point to provide a backing for the cast specimens. The arrangement on the cultivator was such that any number of specimens up to 12 plow pointy could be placed on the shanks and dragged in soil for wear tests. The spacing of i the holes on the frame of the cultivator permitted practical­ ly any desired arrangements of the experimental points. A Ferguson or Ford tractor was used for dragging the cultivator in the soil. The depth and operation of the til­ lage tool was regulated by means of the hydraulic control on the tractor. Preliminary Field Tests to Determine Reproducibility of Results To check the reproducibility of wear tests in the field, a series of field tests were conducted with some of the or­ dinary and nodular iron miniature plow points. Table V shows the results of two of these tests each with 12 specimens of chilled ordinary cast iron and chilled nodular iron. The length of the tests was one hour for each test and the plow points were rotated every twenty minutes. Although the 12 specimens in each case were made from one type of iron and one heat, the reproducibility of the weight losses under the same testing conditions was quite poor. 'The data for the nodular iron show variations in weight losses by more than a factor of two and the data for 69 TABLE V FIELD TEST OP CHILLED ORDINARY AND NODULAR IRON PLOW POINTS OP EQUAL COMPOSITION AND TitEATMENT Sample No. We ight (grams) Nodular Iron Ordinary Iron Before After Loss Before After Loss 1 206.9080 205.871+6 0.3314-0 192.580 192.230 0.350 2 190.0060 189.1+660 0 .51+00 192.572 192.167 0.1+05 3 197.9650 197.1(413 0.5237 197.071 196.678 0.393 k 192.2590 191.5330 0.7260 190.082 189.718 0.361+ 5 193.1500 192.5890 0.5610 185.971+ 185.712 0 .2 6 2 6 1 9 6 .5010 196.2623 0 .2 3 8 7 186.619 186.055 0.561+ 7 1 9 3 .9 5 3 0 193.7213 0 .2 3 1 7 189.1J-77 189.183 0.291+ 8 1 9 0 .7 7 ^ 0 190.0920 0 .6 8 2 0 188.032 187.596 0.1+36 9 2 1 2 .5 0 9 0 2 1 2 .2 2 8 0 0 .2 8 1 0 189.1+714- 189.21+8 0 .2 2 6 10 1 9 3 .i1.070 193.0810 0 .3 2 6 0 193.715 193.271 0 .1+1+1+ 11 2 0 0 .14.870 199.8810 0.6060 198.659 198.251 0.1+08 12 2 0 0 .101+0 199.72l|1+ 0.3796 191+.000 .. 193.l4.63 0.537 71 of the six pairs of the cultivator shanks. These samples were dragged in undisturbed soil for a period of one hour. At the end of each twenty minutes, the points v/ere rotated so that every specimen had a chance to stay in one of the three rows once during the test, A more reliable but less practical method for rotating the plow points was the changing of positions of the specimen after a certain length of time, such that every plow point was in one of the 12 positions on the cultivator once during the test. This method would have taken much time since 12 changes wore necessary for each of the six tests. Assuming 15 minutes for each changing, 12 changed would have taken three hours. Adding the one hour period for actual working in the soil, each test would have taken four hours. This meant that no more than two tests could be made in one day. Assuming that weather conditions remained favorable for the three days per­ iod necessary to run all the six tests, the changing of soil moisture content and its effect on wear resistance of speci­ mens could not be neglected. Besides, even with this labor­ ious arrangement, the presence of accidental rocks and stones which knocked small chips off the miniature plow points could not be predicted. Latin Square Experiments In this type of experiment, the total number of specimens 70 ordinary cast iron test pieces contain variations by more than a factor of three* Apparently no comparative wear study could be made on the basis of such data with such variation in wear for spe­ cimens of similar materials and similar treatments. The variables which caused such variations in weight losses of the specimens, such as soil variations, lay of the land, presence of accidental and unpredictable impacts against rocks and stones buried in the ground, are characteristic of agricultural soils beyond the control of the operator* The only solution to this problem appeared to be the ran­ domized placing of the plow points on the cultivator according to a statistical design and analysis of the data by the method of the analysis of variance. Experimental Designs for the Accelerated Field Tests To show the value of a statistical design in such an agricultural engineering experiment where control of the field variables was difficult if not impossible, a series of so-called controlled experiments were conducted along with the randomized experiments. Controlled Experiments Two similar miniature plow points from each of the six types of chilled irons were places next to each other on each 72 used in each series of tests were arranged in two 6 x 6 Latin squares for the two sides of the cultivator* In the third and fourth series of the field tests, where only five types of irons were selected for comparison, the specimens were arranged In two 5 x 5 Latin squares* The rows of the squares represented positions of the plow points on the cul­ tivator. The columns of the squares represented individual one-hour tests. After randomizing the rows and the columns, the specimens were placed on the cultivator according to the scheme indicated by the two Latin squares* The final arrange­ ment of the chilled specimens in the 6 x 6 squares and their relative positions on the cultivator for the first test In this series are shown on the following page* hour tests, After six one- the data were assembled in the Latin squares and analysed by the method of analysis of variance. The advantages of the randomized design were as follows: 1. The specimens were placed in squares, according to a table of random numbers, in such a manner that any treat­ ment could occur once and only once in a row (position) or column (individual tests)* This condition, which is the basic requirement for a Latin square design, mud.© it possible, as stated by Fisher (35)» for each specimen to have an equal probability of receiving any of the possible treatments, and each pair of specimens not In the same row or column to have the same probability of being treated alike. 73 FIELD TE3T SERIES NO. 2 FINAL ARRANGEMENT OF THE CHILLED IRON PLOW POINTS IN THE 6 x 6 LATIN SQUARES AND THEIR RELATIVE POSITIONS ON THE CULTIVATOR ■ Test No. 1 2 3 Test No. 6 5 1 2 Sample No. sion 1 3 5 7 9 1L c E Q A 2L Q A J 3L E J LjX J 5L 6L 3 k 6 5 Sample No. 11 Posi­ tion 2 k 6 8 10 1; p J 1R J P |A Q c E P E C 2R E J j.Q P A C Q P 3R P E J A Q I|R C J Q A P Q E J C A 5R A Q i jc E !P ! C jJ A P A — r0 C E E p Q A c P Q J E 6R Q A |E C J P f KEY TO THE POSITIONS OF THE PLOW POINTS ON THE CULTIVATOR C ▲ 1L Q E 2L 2R I A i+L 3L A 5L J ▲ 1R A A6L A 3R A i+R Jt 6 r A 5R 7k 2# The conditions outlined on the previous pages elim­ inated the need for changing the position of the specimens during the individual tests. In addition, as will be shown in the analysis of the data, the effects of position, soil variations in individual tests, and other sources of errors could be eliminated before the irons were compared for rela­ tive wear resistance. The presence of all these variables obviously reduced the number of degrees of freedom, but this was overcome by repeating the squares in the two sides of the cultivator. 3. The fact that no changing of positions was necessary during the one-hour tests permitted all the six tests of one series In one day. Thus the experiment was less subject to weather conditions and variations in soil moisture from day to day. 75 V. PRESENTATION AND ANALYSIS OF FIELD AND LABORATORY WEAR TEST DATA Small Scale Accelerated Field Tests After some preliminary field tests to determine the optimum operating conditions in the field, four series of field tests were conducted on the experimental miniature plow points* The first and the second series, each consis­ ting of six individual one-hour tests, and the third and fourth series each of five one-hour tests* In all the 22 field tests the following procedure was followed: The selected types of plow points were mounted at the specified positions on the cultivator shanks. The cultiva­ tor was dragged in undisturbed soil for one hour. of the tractor was kept at about 3i miles per hour. erating period wasmeasured with The speed The op­ a stop watch from the time the cultivator was lowered into the soil up to the time when it was lifted from the soil at the end of each run. The num­ ber of runs and the time for each run were recorded until the plow points had been operating for exactly sixty minutes. At the end of the test, a new set of samples was placed on the cultivator and the next test of the series was started. ter washing and drying, an analytical balance* Af­ the loss of weight was determined by 76 First Series of Field Tests (Controlled Design) Table VI shows weight losses for the 72 samples of chilled iron plow points during the one-hour tests in the field. The samples were not randomized, but were rotated every twenty minutes. As seen from this table, three samples broke during the test and two others were chipped so badly that their weight losses could not be averaged out with the rest of the samples in that group* In the case of randomized experiments, such missing values in the data, as will be shown later, could be estimated. But in this case there was no statistical method available either for analysing the data or for estimating the missing values. The means of the 12 replications in each group of irons are shown at the bottom of the table* A comparison of wear resistance on the basis of the mean of weight loss per hour showed the following rating in order of most resistance to wear: J - High carbon alloyed C - High carbon unalloyed Q - High carbon unalloyed nodular P - High carbon alloyed nodular A - Low carbon unalloyed E - Low carbon alloyed Now the question may arise, since the differences in the 77 TABLE VI FIELD TEST SERIES NO.l - WEIGHT LOSSES OF THE CHILLED IRON SAMPLES IN MILLIGRAMS ________ ( CONTROLLED DESIGN )_________ Test • Sample No. No. Type of Iron A c E J P Q 1 1 2 1+16.33 1+3 8 .0 0 31+9.33 317.33 458.00 469.00 379.35 331+. 33 1+91.66 I+OO.3 3 3 8 3 .6 6 5 0 2 .0 0 2 3 5- 530.66 1+8 2 .6 6 1+2 1 .6 6 1+11.00 1+92.33 1+56.33 317.00 31+1 .0 0 1+29.33 1+31.33 355.00 1+29.61+ 3 5 6 507.66 1+76.00 1+22.00 1+28.33 1+8 1 .0 0 509.66 370.33 31+0.33 637.66* 1+55.66 631+. 33w Broke 4 7 8 679.33 1+71.33 531+.00 ¥+7.31+ 787.33 639.33 391.33 51+1.33 513.32 Broke 460.66 5 6 3 .3 3 5 9 10 556.00 5 3 8 .0 0 1+06.66 398.00 510.00 1+72.00 51+5.33 1+18.00 6 8 2 .0 0 1+78.00 551+.00 1+3 8 .0 0 6 11 12 711+.00 6 8 0 .6 6 Broke 554*65 675.33 626.00 1+16.66 582.00 1+70.66 ;1+99.33 1+1 2 .6 6 1+71+.00 — '"'Chipped badly at the point- . . . 78 mea n of weight losses for the various groups of iron is rather small, what should, bo the difference in the means of any two types of irons in order to be significantly dif­ ferent from each other? In other words, is the difference in the means of Irons C and A due to better wear resistance of iron C or perhaps due to some variables in the field? In­ asmuch as the samples were not put on the cultivator at ran­ dom, there was no way to determine the level of significance for these differences. As will be shown In the analysis of the randomized experiments, there was enough evidence to show that the variation In soils from one test to another and posi­ tions of plow points on the cultivator had greater effect on the final results than the differences In irons. Until the effects of these variables are compensated for, we cannot say that Iron C is always better than iron A. Since the control of all the field variables was almost an impossible task, the only alternative seemed to be the randomization of the sam­ ples and analysis of results by statistical methods. Second Series of Field Tests (Randomized Design) In the second series of field tests the wear tests of the chilled iron specimens were repeated with the exception of having them put on the cultivator snanks according to tne Latin square designs described on Page 73# The data for these tests were assembled in the two 79 Latin squares (for the two sides of the cultivator) and an­ alysed as shown on the following oages. The analysis of var­ iance for each side of the cultivator is shown below the cor­ responding square* The underlined figures in the squares were missing val­ ues which could be estimated either by the method of minimi­ zing the sum square error, or by a trial and error method* The method of minimizing the sum square error consists of setting the first derivative of the sum square error E with respect to each of the missing values equal to zero and then solving the resulting equations simultaneously. This method, as shown on Page 8Ip , is lengthy and time consuming. The method chosen is instead the trial and error method based on the formula shown on Page 83 . The derivation of this for­ mula, which is based on a method of covariance analysis, is described by G-oulden (38). When more than one value is raiss- ing, repeated application is made of the above mentioned for­ mula until the difference between the computed and the assumed values become negligible. The computations shown on Page 83 were presented as an illustrative example for finding the best estimate for the missing values in the square for the left side of the cultivator, which involved three missing values. The inis sing values, as mentioned before, were the weight 60 FIELD TEST SERIES NO. 2 WEIGHT LOSSES OF THE CHILLED IRON PLOW POINTS IN MILLIGRAMS ASSEMBLED IN THE RIGHT SIDE LATIN SQUARE FOR ANALYSIS Test No, 1 2 . 3 5 6 10 12 533 E 1+77 k Sample No. >031;ion 2 k 6 8 1R J Uo6 P 670 A 866 Q. 578 c E J 61+8 Q 765 P 703 A 873 2R 603 c 1+83 . 3R UR 5R 6R P 211 Q 308 C A 307 322 C E 525 P 1+09 J 193 A C 251+ 267 J 257 Q 162 A E 21+8 ■ 200 E 3k7 J 227 Q 317 A 322 E P 281 Q 251+ C J 193 11+1 P 118 P Q 307 21+9 TREATMENTS A Mean ,1+68.7 C E J 329.3 1+09.5 333*5 398.7 397.3 ANALYSIS OF VARIANCE D.F* Total 33* Position (Rows) 5 Tests (columns) 5 Tre atment 5 18 Erroro P Sum Square Mean Square 1,1+98,599 1+7.0 1,177,586 2 3 5 ,5 1 7 5.96 29,867 11+9,331+ 3.25 16,299 81,1+93 5,010 90,186 P @ 2.77 2.77 2.77 *Two degrees of Freedom taken out ior the two estimates which were made In this analysis* 81 FIELD TEST SERIES NO. 2 WEIGHT LOSSES OP THE CHI LIED IRON PLOW POINTS IN MILLIGRAMS ' ASSEMBLED IN THE LEFT SIDE LATIN SQUARE FOR ANALYSIS Test No. 1 2 1*. 5 6 Sample No. Posi;ion 1 3 5 7 9 11 1L c 588 E 860 Q 1039 A U16 P 782 J 5W> 2L Q 52k A 775 J 911+ P 1168 E 8^7 c 627 3E E 386 J a>6 A 671+ C 306 Q LpLQ p 24-07 i|i J 162 ip 1+07 0 367 E A 371 Q 239 5L p 359 0, 357 E k96 J 188 C 302 A 308 6L A 190 C 200 P 261 Q 1*£7 J 162 E 335 P Q TREATMENTS c A Mean E J 14-55.7 398.3 6it5.3 399.6 561*..0 509.3 ANALYSIS OF VARIANCE Total Position (Rows Tests 9columns Tre atrnent Error^ D.P. Sum Square Mean Square .32* 2,1+61+, 193 1,366,706 287,061 285,271+ 525,152 273,31+1 57,1+12 57,055 30,891 )5 )5 5 17 P F© 5% 8.85 2.81 1.86 2.81 1.85 2.81 “Three degrees of freedom taken out for the three estimates which were made in this analysis. 82 FIELD TEST SERIES NO. 2 TWO-WAY TABLE AND THE COMBINED ANALYSIS OF THE TWO LATIN SQUARES Side of Treatment tivator A Riglit 2812 C E 1976 6 Left 2001 2459 6 P j 6 2392 6 Sum Q 14024 238 4 6 6 Mean 390 36 2390 3872 2398 338k 3056 1783i| 6 6 6 6 6 36 u .. . 5546 14.366 6331 i+399 5776 5kko 31858 12 12 12 12 12 12 72 2734 6 Sum 495 , Mean 462 36k D*F. 11 Total Side 1 Treatment 5 Side x Treat, 5 * 528 i 367 Sum Square £ 6 8 ,3 8 6 1 2 0 1 ,6 1 3 254,438 112,329 481 it-53 Mean Square 2 0 1 , 613 5 0 ,8 8 8 22,466 F F @ 3% 8.97 2.27 6 .6 1 5.06 COMBINED ANALYSIS OF THE TWO LATIN SQUARES D.F. 66 Total 10 Tests (Columns) Position (R o w s ) 10 1 Side Treatment 5 Side x Treat. 5 Error** * Errorr 35 Error L 5.0 Sum Square 4, 16k, 4 0 4 436,395 2,5144,292 2 0 1 ,6 1 3 254,438 1 1 2 ,3 2 9 615,338 727,667 Mean Square 43, 6 I4.O 254A29 201,613 5 0 ,8 8 8 F F © $% 2 .4 14 11 2.8 2 .0 7 2 .0 7 k .0 8 2 .45 1 8 ,1 9 2 - --- .-— . 83 FIELD TEST SERIES NO. 2 ESTIMATING MISSING VALUES FOR THE LEFT SIDE SQUARE The formula for the estimation of tbe missing value, as described by Goulden (36), is r(R C + T) - 2G (r -1)(r - 2) where r is the number of rows or columns, and R, G, T, and G, are the row total, column total, treatment total, and grand total containing the missing value. First Estimates Estimate E: Assume A - G - 15572 ¥ 1+33 “ “ 35 R - 3371 ; C - 1370 + 1+33 ^ 1 803 T - 3012 ; G - 15572 ♦ 2 X 1+33 - I 6 R 38 E - - 6(3371 ♦ 1803 + 3012) - 2 X 161+38 (6 - 1 ) (6 - 2 ) 812 Estimate A: R - 3J+53+ k33 z 3866 T ^ 1959 C ^ 1370 + 8 1 2 z 2 1 8 2 G - 15572 * 8 1 2 * 14-33 z 16817 A z 727 Estimate C: R z 3453 + 727 - 41®° c - l83^ T _ 1763 G - 15572 + 812 + 727 Z 17111 G - 622 8[j_ The values obtained in the first estimate were used in a second estimate and those found in the second estimate were used in the third estimate. Since the values found in the third estimate were about the same as those found in the second estimate, they were used as 'the final estimates and are shown as underlined figures in the data for the left side square. The best estimates for the missing values could be found by the method of minimizing sum square error as indicated in the following: Let: R ^ row total, Or grand total, SS S G ^ column total, x sum square, N T treatment total, weight loss of a single plow point total number of plow points, sum, E ^ sura square error, P and Q, missing values. Total SS ^ Row SS + Column SS + Treatment SS + Error SS Or : 2 Sx2+ P2+ Q2- (G + P + a)2 - SR^ + LB 2 _ i l +_£_) 5“ 5 35-------- - (G + P fft) 35 + SC 2 + {G 3 + Q ) 2 + (Oil. + p ) 2 ~T~ ^ -------6- (G + P 4 Q) 36 + ST 2 ♦ (TQ ■t_d)2 E B + (TP + P )2 - (G + P +Q)2 + S 5----55 For minimizing the error sum square, set the first derivative of E with respect to P and Q, equal to zero. dE/dP - 0 ^ Solve the two equations for P and Q. dE/dQ ^ 0 J 85 losses for the specimens which, during the field operation, either had broken off at their points or had chipped off so badly that their weight losses could not be integrated with the rest of the data. After analysing the squares for the right side and the left side of the cultivator, the two squares were combined by setting up a two-way table and finding the error due to the two sides of the cultivator (see Page 82). The combined analysis of variance shown on Page 82, indicated the following results. 1. Differences in individual tests were significant. 2. Differences in positions of the specimens on the cultivator shanks were very highly significant. 3. Differences between the two sides of the cultivator were highly significant. if. After the effects of the individual tests, the dif­ ferences in position, and the sides of the cultivator were taken out, the differences in treatments were significant. To evaluate the wear resistance of the various types of irons, a comparison of the means of the weight losses of the six types of irons was made by using the standard t-test, as shown on Page 86. For comparing the means containing misss' sing values, the method described by Yates (37) was used as indicated in these computations. The theory behind the t-test is involved and is certainly 86 FIELD TEST SERIES NO. 2 COMPARISON OF THE MEANS USING THE STANDARD " t " TEST Sample Mean A C E j p q lj.62 364 528 367 481 453 Cfdiff. » Standard error of difference between two means? S2 ( l/r1 + 1/r ) where S is the estimate of variance, shown as the mean square of error in the table of analysis ofjvarijnce. 'r*is the number of replicates to be assigned to any treatment mean for compar­ ison with another treatment mean. For testing the significance of mean differences between treatments containing missing values, the rule given by fates (37) was applied. In comparing the means of A and C, for ex­ effective ample, the/number of replications r were determined by apply­ ing the rule as follows: For each value of A determine C present in row and column, replications _- 1 C present in a row or column, replications 2/3 G missing, replications ^ 1/3 A missing, replications ^ 0 Similarly calculate the effective replication for C based on the presence or absence of A in corresponding rows and columns, Using the above rule• rA - 6 + 1 + 0 + 1 -f 1 + 2 / 3 ■+ 1 2 32/3 87 ji C 6 f l + 0 + l + l + l + 2/3 ^ 32/3 18192 ( 3/32 + 3/32 ) ^ At A - C 58.k and ij.0 degrees of freedom t - diff> X t - $8.1|_ X 2.02 i 2.02 118 In order that samples A and C to be significantly different at least 95 percent of the time, the difference between the means of the two samples should be 116 or more. Since the difference is 562 - 365 - 98 we can say that iron A and G are not different , as far as the resistance to wear in soil is concerned. At 106 and 56 degrees of freedom t A - 0 1.68 - 98 Ai^ this level of significance, the difference is just sufficient to make the two irons different. Following the above procedure, the test proved that the only irons which were significantly dilierent were irons C and E. The difference between these Irons was sig­ nificant even at the I/o level. 88 beyond the scope of this work* A great deal of theoretical work has been done in advance and tables and formulas have been prepared for the actual performance of the test (16)* In testing the difference between any two treatments, the basic principles involved in a t-test require that the dif­ ference between the means of the two treatments to be equal or greater than the product of the standard deviation for the mean difference and a r,t M value obtained from the t-table for the number of degrees of freedom and at the desired probabil­ ity. The desired probability in most engineering and agricul­ tural experiments Is usually taken at not less than 95 percent. At this level of significance, it was found that the only irons which were different from one another were Irons C and J from iron E. In other words, the wear resistance of all the irons were about the same except for the Irons C and J (high carbon chilled unalloyed and high carbon chilled alloyed, re­ spectively) which had greater iron E resistance to wear than the (low carbon chilled alloyed). According to this test, the difference in weight losses for irons A and G (low carbon unalloyed and high carbon unalloyed, respectively) should have been 118 or greater in order to say that at leastfor 95 per­ cent of the time iron G is superior to iron A in wear resist­ ance. Since the actual difference in weight losses came out to be 98 milligrams, we c a n say that for this level of prob­ ability the two irons were not significantly different, 89 despite the fact that iron A wore out 98 milligrams more than iron C* If we are willing to take a chance of being wrong ten percent of the time, however, the t-test shows that the difference of 98 was just sufficient to make the two irons different* When the results of these randomized tests are coiapared with those of the controlled field tests, the values of sta­ tistical design become evident. It shows that in this type of experiment where so many variables affect the wear of the iron pieces, a mere comparison of the means of several repli­ cated tests does not prove the superiority of one type of iron over the other. The correlation analysis of the field and the laboratory data, as will be shown later, also proved that the randomized experiments resulted in better correla­ tion than the controlled experiments. findings, it tests by the In the light of these was decided to run all the rest of the method of the Latin square field design* Third Series of Field Tests In this series of field tests, the miniature plow points were made from the five types of irons which were discussed on Page 60 . All these irons were made in the unchilled condition in order to obtain some information on relative wear resistance of gray cast iron with respect to that of the chilled cast iron. This information was 90 needed in analysing the results of wear tests of the full size cast iron plow shares chilled at the cutting edges. A total number of fifty specimens were tested in this series in five one-hour field tests. The five tyoes of irons were placed on the cultivator shanks according to two 5 ^ 5 Latin squares with the same randomization mentioned in the case of the unchilled samples. The weight losses were assembled in the two Latin squares for analysis. The data for the right and the left side of the cultivator plus the analysis of variance for each square are 4 shown on the following pages. Specimen in the right side of the cultivator broke off during the test and therefore an estimate had to be made for the missing value, following the method discussed previously. The final analysis of the combined data of the two sides of the cultivator is shown in the following: * THIRD SERIES Oh E1ELD TESTS ANALYSIS Oh VARIANCE EOR COMBINATION OF RIGHT SIDE A IW LEET SIDE SQUARES D.P. s .S. 37.50!| 1+8* Total 8 Tests (Columns) l.3¥> 8 Positions (Rows) 15.373 5.798 1 Squares 9.850 Treatments k 1.281 3q x treat k U.056 Error^ +- Error ^ 23 5.337 27 Error *'0ne estimated value involved M.S. __ 0.163 1.922 5.798 2. /+12 0.198 P 0.85 9.70 29.29 12.18 . . F @ -5J 2.30 2.30 U. 21 2.73 91 FIELD TEST SERIES N O . 3 WEIGHT LOSSES OP THE UNCHILLED IRON PLOW POINTS IN MILLIGRAMS PER SQUARE CENTIMETER ASSEMBLED IN THE' RIGHT SIDE LATIN SQUARE FOR ANALYSIS Test No. 1 2 3 k 5 Sample No. Posi­ tion 1R 2R 3R kR 5R 2 1* 6 8 10 D 154.44 R H 97.62 1 0 8 .kk F 1 0 6 .3 2 B 125.07 B 94.44 H 103.94 D 179.82 R F 72.29 R 83.26 F 94.73 B 1 0 2 .ko H 1 1 0 .kk D H B 103.32 F 90.05 D 135.26 R D R 53. ki-- B 75.67 H 7k. 08 108.06 ... ... P 71.94 79.76 9k. 00 119.35 95.03 TREATMENTS Mean B D F H R 101.18 135.01 87.07 100.99 8k. 66 ANALYSIS OF VARIANCE D.F. Total Position (Rows) Tests (Columns) Treatment Error^ 23* 4 4 4 11 Sum Square 20.545 7.512 0.482 8.850 3.701 Me an Square 1.878 0.120 2.212 0.336 P 5.59 0.357 6.58 p ® 5?£ 3.36 3.36 3.36 *One degree of freedom taken out for the one estimate which was made in this analysis. 92 FIELD TEST SERIES NO. 3 WEIGHT LOSSES OF THE UNCHILLED IRON PLOW POINTS IN MILLIGRAMS PER SQUARE CENTIMETER ASSEMBLED IN THE LEFT SIDE LATIN SQUARE FOR ANALYSIS Test No. 1 2 3 • Posi­ tion k 5 Sample No. 1 3 5 7 9 1L H 97.02 D 98.09 F 90.31+ R 81*..88 B 71+.58 2L D 108.52 R 8 3 .1 2 H 98.12 B 98.38 F 77-02 3L R 81+.03 P 81.62 B 90.55 D 116.91+ H 81.79 1+L F 76.76 B 81.38 R 70.26 H 100.65 D 96.20 SL B 1+8.65 H If4.79 D 61.61 F 1+9.61 R 39.1+1 H R 81+. 1+8 72.31+ TREATMENTS Mean I J B D F 78.70 96.27 75-07 1 j ANALYSIS OF VARIANCE D.P. Sian Square Total Position (Rows) Test (Columns) Treatment Error L 2 lf. If. Ifk 12 1 1 .1 6 1 7 .8 6 1 0,861f. 2 .0 8 1 0 .355- Mean Square F 1.97 0 .2 2 0 .3 2 0.0295 67 7.3 1 7 .6 F ® 5% 3 .2 6 3 .2 6 3 .2 6 93 The final analysis indicated the following results: 1* The difference between the individual tests was not significant. Apparently, the test plots were more uniform in soil structure and composition than the plots used in the previous tests* 2. The effect of positions on the cultivator was still highly significant. 3* The difference in the two sides of the cultivator was very highly significant. Such difference could be the result of improper adjustment of the cultivator. Perhaps the right side of the cultivator was operating deeper in the soil than the left side, with consequent higher weight losses for this side of the tillage tool. ip. After the effects of position on the cultivator shank and the side of the cultivator were taken out, the effect of treatments was highly significant. The means of weight losses for the ten specimens in each type of the gray irons are shown in the following: Type of Iron R Mean Weight Loss (grams) 2.669 P 2.757 B H 3 •0lp8 3.153 D 3.910 To find the relative wear resistance of the irons, a t-test was made as in the case of the chilled irons. In order that any two of the irons be significantly different from each 9k other (for 95 percent of the time), the required difference was lound to be 0.290 gram for any two irons which did not involve iron D. If iron D were one of the two irons to be compared, the required difference for significance was found to be O.ij.30 gram. This variation arose from the fact that one of the specimens of iron D broke off during the test and thus its missing value had to be estimated. When these fig­ ures were added to the weight losses of the irons to be com­ pared, the comparison revealed the following results. 1. Irons R and P (high carbon gray unalloyed rodular and low carbon gray alloyed ordinary iron, respectively) were not different from each other, but were significantly ^superior to all the rest of the gray irons. 2. found to Iron D, the high carbon unalloyed gray iron, was be inferior to the rest of the gray irons as far as its wear resistance in soil was concerned. Fourth Series of Field Tests The most significant results obtained from the three series of field tests conducted so far were as follows: 1. iron was In chilled Irons, wear resistance of a high carbon better than that of a low carbon iron. This higher resistance to wear could be attributed to the higher percentage of the hard massive cementite or iron carbide constituent of the high carbon white iron. 9$ 2. Alloying elements, nickel and chromium, did not seem to improve wear resistance of the chilled irons. In gray irons, however, the alloying elements contributed con­ siderably to the wear resistance of the Irons. 3* Gray nodular iron proved to be superior to all the other gray irons. White nodular iron, however, was inferior to white high carbon ordinary iron In alloyed or unalloyed conditions. A fourth series of H e l d tests was conducted for the two following reasons: 1. To see if the results obtained in the previous tests were reproducible. 2. To determine the relative wear resistance of chilled and unchilled irons when they were working under the same field conditions. After all the chilled and unchilled miniature plow points were subjected to the wear test machine in the lab­ oratory, fifty plow points, made up from ten types of irons, were selected for the fourth series of tests in the field. The right side of the cultivator was used for testing five types of the irons in a 5 x 5 Latin square design. The left side of the cultivator was used for testing the other five types of irons in another 5 ^ 5 Latin square. Each side of the cultivator contained the samples for a series of tests by itself. The results of the two squares were to be analyzed 96 independently. Five one-half hour field tests were conducted and the results were assembled in squares as shown on the following pages. The analysis of variance for each square showed that the effect of positions on the cultivator was significant, but apparently the soil plots were not significantly different. After the effect of nosition was taken out, the difference in treatments was very highly significant. To see if the results obtained in the previous tests were reproduced in this last series of field tests, the irons were grouped in order of increasing weight loss in each series of tests. These groupings are shown in tabulated form on Page 99. As seen from Table VII, the relative positions of the iron specimens in the wear scale were reproduced during the fourth series of field tests. For example, irons G and J which proved to be superior to the .rest of the chilled irons, repeated their positions at the top of the scale during the repeated tests. Similarly, iron E, which was at the bottom of the list, repeated its position at the bottom of the scale during the repeated tests. Other results of the previous tests were also reproduced during the fourth series of field tests. When the t-test was applied for comparison of the means of weight losses of the irons, the results checked very closely with those obtained in the two previous series of tests. 97 infirm ™ „ EIELD TEST SERIE^ NO. 1*. ( RIGHT SIDE ) WEIGHT LOSSES OP THE CHILLED AND UNCHILLED IRON PLOW POINTS IN MILLIGRAMS PER SQUARE CENTIMETER ASSEMBLED IN LATIN SQUARE FOR ANALYSIS Test No 1 2 3 k 5 ?5 3 0 .2 2 B3 144.09 A5 12.73 012 1 3 .8 0 R6 14.8.76 2R A8 13.11)- R8 35.65 B6 39.71 F2 1)4 . 2 2 09 111-.51+ 3R B7 38.51 A10 8 .8 3 0147.55 R2 26.5 F7 35.62 i|H Qll 1 3 .2 0 F6 55.80 R5 14-3.27 A7 1 8 .7 6 B5 8 2 .6 0 5R R1 l]-2 .7 0 Q 10 10.95 FI 33.59 B2 1)4.144 A12 14-.1+1 Posi­ tion 1R TREATMENTS Mean A B F Q R 13.57 14-9.87 i+3.89 1200 39.38 ANALYSIS OP VARIANCE i D*F. | Sum Square Mean Square Total Position (Rows) Test (Columns) Treatment Error 2 lj. 1+ It. 5 12 ! ! j _ _ 8.305 1.025 0.396 6 .2 2 1 0 .6 6 3 ..... 0 .2 5 6 0.099 1.555 0.055 F F ffi 5% 4.65 1 .8 0 2 8 .2 7 3.26 3.26 3 .2 6 98 FIELD TEST SERIES NO. ij. ( LEFT SIDE ) WEIGHT LOSSES OF THE CHILLED AND UNCHILLED IRON PLOW POINTS IN MILLIGRAMS PER SQUARE CENTIMETER ASSEMBLED IN LATIN SQUARE FOR ANALYSIS, Test No. 1 2 3 4 5 H6 8 I4..09 E6 34.12 j6 20.50 P3 31.47 C3 2 7 .8 0 2L Cl 2 1 3 .1 7 PI 18.15 E9 17.40 H7 75.87 J12 21.02 3L J2 18.86 Cl 25.^9 P9 2298 E2 34-06 HI 133.61 UL El 16.57 H10 64.70 CIO 1 2 .3 8 Jll 15.87 Pll 27.33 5L P4 12.31 J4 1 2 .5 7 H3 57.36 04 14.53 E7 23.07 Posi­ tion 1L TREATMENTS Mean C E H j P 18.67 25.04 83.13 17.76 22.45 ANALYSIS OF VARIANCE Total Position (Rows) Test (Columns) Treatment Error D.F* Sum Square 2k 19.862 It 4 12 1 .8 2 1 1 .2 5 2 15.498 1 .2 9 1 Mean Square 0.455 0 .3 1 3 3.874 0 .1 0 8 P 4.21 2.90 35.87 F 3 .2 6 3 .2 6 3 .2 6 99 TABLE VII RATING- OF ALL THE IR0.i\ SPECIMENS IN ORDER OF DECREASING RESISTANCE TO WEAR Identifi­ cation Description of Iron Mean Weight Loss (mg/cm^) Field Test Series No. 2 C J Q, * A P E Chilled Chilled Chilled Chilled Chilled Chilled high carbon unalloyed high carbon alloyed high carbon nodular unalloyed low carbon unalloyed high carbon nodular alloyed low carbon alloyed 1 0 .6 2 10,70 13.23 13-50 lil-05 15.1|.0 Field Test Series No. 3 R F B H D Unchilled Unchllled Unchilled Unchilled Unchilled high carbon nodular unalloyed 78.50 low carbon alloyed 81,07 89.64 low carbon unalloyed high carbon alloyed 92,73 1 1 5 .0 0 high carbon unalloyed Field Test Series No. 1+ (Right Side) Q A R F B Chilled high carbon nodular unalloyed Chilled low carbon unalloyed Unchllled high carbon nodular unalloyed Unchilled low carbon alloyed Unchllled low carbon unalloyed 12.00 13.57 39.38 14-3.89 1i9.87 Field Test Series No. ip (Left Side) J c P E H Chilled high carbon alloyed Chilled high carbon unalloyed Chilled high carbon nodular alloyed Chilled low carbon alloyed Unchilled high carbon alloyed 17.76 1 8 .6 7 22*1+5 25.01+ 83.13 100 As for the relative wear resistance of chilled and un­ chilled irons, reference to Table VII shows that an unchilled Iron wore about four times as much as the corresponding iron in the chilled condition. Comparison of irons A and B and also J and H shown in Table Vll are examples for relative wear resistance of chilled and unchilled irons with similar composition and tested under similar conditions. Laboratory Wear Tests of the Field-Tested Specimens and Correlation of the Field and Laboratory Data All the iron specimens tested in the field were tested by the wear test machine in the laboratory, following the procedure described on Page 75. side of the plow point specimens. Two tests were made on each The value reported for the wear of each sample in the laboratory was the average of four tests. Tables VIII, IX, X, XI, and XII show the results of the laboratory wear tests assembled with the field data for cor­ relation analysis. The mean of all the laboratory wear tests on all the iron samples for each group of Iron was indicated by X1# The mean of all the field wear tests on all the iron samples for the corresponding iron was indicated by X 2» The valutes of X± and X 2 for each series of tests were analysed to determine the correlation between the laboratory and field wear tests. The details of the correlation analysis for each 101 TABLE VIII FIRST SERIES OF TESTS - CORRELATION DATA i^GR FIELD AND LABORATORY WEAR TESTS Type of Iron -xi 1 X1 x2 x2 7.73 12.16 5 .7 8 1 0 .2 0 j, — .... J ..... Q x 2 X 1 X1 X1 x2 x2 X1 x2 9.05 1337 6.50 1 1 .0 7 6.65 1435 8.30 11.20 2 6.60 12.79 4 . 3 3 9,27 8.73 13.69 6.63 9.76 6.35 11.68 6.88 1 4 .6 6 7.15 15.49 6 .0 5 1 2 .3 1 8 ,14.0 1437 5.30 9.25i 8.35 12.53 5.85 10.36 k 7.18 lit.09 5.53 1 2 .0 0 9.95{8.65 12.59 7.38 12.54 5 8.33 1 4 .8 2 4.45 1 2 .3 2 7.90 14.04 5.73 10.81 6 7.30 13.90 5.10 1 2 .5 1 9.18 1488 5.85 9.94 7 6.90 19.83 5.55 15.59 7.4-8 2^98 6 .0 0 11.42 13.95 14.99 4 .8 3 13.44 8 7.25 13.76 4-45 1 3 .0 6 8.93 18.67 5.88 9.97 9 7.23 1 6 .2 3 5.30 1 1 .8 7 8.70 1489 6.60 15.92 7.18 19.91 5.83 1 6 .1 8 10 8.70 15.71 5.18 1 6 .2 0 7.50 13.78 6,60 12.20 8.43 13.96 6.60 12.79 11 6 .2 8 12 7.00 19.87 6.53 1 6 .1 9 8.55 1827 5.98 13.74 4.40 14.58 5.45 13.83 pie No. c A 20.85 — _ E _ _ 9.00 13.32 5.70 ------------ ' -------- 5.40 1 3 .3 0 --- --- — 6.55 jl6.44 9.05 19,71 7.15 12.17 1310 16.99 6 .8 0 12.05 Sum 87.65 189.50 5 8 2 5 |lija 5 2 10441 '191.977392 13420 7707 13458 69.87 146.79 Mean 7.30 15.79 5.30 1 2 .8 7 8,53 16£)0 6.16 11.35 8.56 1 4 .6 2 6.35 13.34 . r Weight loss in laboratory (milligrams) 2 x2 - Weight loss in field (mg/cm ) 102 r"A P±OoO CMCM CMC O A A A CMA A A A CO_rfCM. A CM A CM-d*D- o lA rA A OT-=T rH CMCOo r'H.J’ArH rHHi rH rH 1A A CO o A g o O O A O Yf\A ni WH ''A O A A -d-co.-d;AcO, CO CO GO, A A C^-sO A A A g o Av A A A - 173 GO 05 o i o co i— i A 5CO rH A ps ■B o EH l O h - ^ O A rH -Ar^r q -chA Q. A CO CO IIs— A A AcerA A c— C" A 3 CM O rH HtCMA P±0O A A o A A A. 1 ^ 4 rH HHHHCOACMCVlH cm rH a rH A A O O AoO O O A G O O A * O CM rH rH rHA CM_cjA rH rH C—A A A § 05 o o O A A CMCMcr^r^c—'C^-rH CMH H oom rH O rH O O A O O CO O A A ACO A A r^r CMO A rH Ol rH A CMCMC'T 12 X (Ago C° h-vO 1“A IS-OO fi'0' P~* • O CM (—I O d A rHCOrH CM O-A A CT* rH rH Pm Pm Pm PSPm Pm d d h-1 0 J-3 O rH CMA 4 - A A rH CMA -=frA A SH bO •H rH < —i CM so & O aJ O A cS rH T* rt £ -P rH CD •H ■H CO O AACO A CM C\J_rtO C" CM_d-rHA A A. A A A A-d-CM _d-CM CM03 O* O OlA X Posi­ tion PQ EH 1070 Eh CM < a cm. A A A _-h AOoOOOcQ CM 'la'A A OiOlAcO rHcO CO _d*CM A W A O rH rH CM^ CM rH rH C?A, %' i—1CMMl (—1CO corCMA rH i—1rH C^* t—1 PH ’ 1AO i—1 A A A O AOCO A rHA A O !>A CMA C M A rH A A A A A K A CO a3 vOCO0O*jfb= H H COM) M) CO* O CM rH J-CO CMMl rH rH CM> rH A H A A CM _d"0 A CM CO.01 A O O CMAMD CO rH A A H c O v a 4 A H A O fH A_^M>W rH H A AC^-pj-rH CMrH LAA A EH X i—1 -=t O CM rH COA -d*rH rHCMArHArHAC— Eh ■ cd •o O pp 125 <2 h-Q no M 1 925 H PS <12 f/31Q _ c eh r*3 A CM. A J CO CO o 1 —f H o A A Hr CM A c a n a c d Q O H H O ’H )K O O H O A f M oO H 0 4 rH rH rH rH H—0 s A CO rH _TT X| X Eh *=aj id A O CQ CM O '-O ilA lA O l A H CM A O Q _z±CO.HO,CNJ A A C M .O CM r-fcO A r-f A A I>~ CM_zf A A _ d A A A -d -_ d -A A A Oi si i -d c O C\| H H \ A C M H rO • <£j o A £ o M TABLE X ^ Q- * c0 -CM CA O CM riC Q . O S O 'v O A _d A -d A A A O J lA O r O H H Q O h* O 1A H H oOOcO O H v D H l 1—I 1—I 1—I t—1 r rH 1—I *rl e _H* CM "A A CM _ d CM _d:cO A CM A A A A O H ! d A ' cd A id o ,c o ' A " CMA j X O &, <*J A- » O. | (Xs rH 1—I CH A O O i 1—i A co -=k | o CM CM CM IA CM H - M A O H CO ( ^ 1 0 ,0 0 . O . CM (A CO. !>=■A_ X I t H H i r f r l C ^ O ' f ^ O C\JrT A A A A A A A A A A CM S o 1—1 A r>s JH O -P Crf JH O A erf 1—I 'd 1—1 0 *H u Hi EH O 12! <1 J d o o CM A rH o O -d 'C A rH C —A A | C— O CM i 0 44cP H aD coA coA i -d CM A 3 - CM A A A A A A A I O J d O O A -X C O O H C O ) X (H O H H H H X O C O _ r ± i A , A Pi •H *H CO CO O 1— X o |2i [— CM O CM O A CM A CM O -! O CA O < A H CM CM CM, r-ioO^i 03 A CO’ A O ^ O ’ TA -d^M G O ..... rH CMA “ ■ 1 T-d--d-i _dcO (Xs O - A A rH + P« co o o »h d -P ddddddi^Pi rH CM A J - 1 A H >-3 CM A - d A W CO o ! CO -P PQ •H Pi A £• -d PJ Crf •a -p A! bD •H ■H O 0 ts >1 >1 CM X X 101^ OO rA • O rA A. X PC! K* o Ss? Eh X OS «3j '— ■fccj iS & ~ Q Q l-H i-5 co pq M t-f &4 ft er Q m Js« Q5 rA A • >• • ■* • 1 X vQ ^O A- A-co ft f ( i | ! j tsO CM M M EH < CA CO » rA -=* dMA • CO CA a- ; MA • OA rH MA CM A- •H £ o JH O -P aj o Pd ft 05 ft ft rH CM cA-d-lA * I3 rcf rH ,a © ctf »H rH in ft £ •H CQ W © CQ o O i —i H -p -P •H fci! 0 fty> © 1A°0 O A- CM rH rH cm li il i—I c\j < « 105 T. rH K • o a CM CM O O U\GQ o H A A A CM A A 00 CO A A CM O rH (—1 • rH O j—i _M" CO H O CMv£) A A 'iA O c O C O 'lA • • • ■ CM O H CO IA CM X CM CM H H H • 1 7 .7 6 <\j W A A c O A rH _£ rH A A A • • t • • rH co C\J r— CM A H CM CM rH J rH • o XII a Zh rH X 0 ft « o i23 EH O • I A CM H sO H CM rH H * o a a o o O CM CM O O • • - • ■* • O JC M O H dA A A A A A i—1 • A CO O A • A A £ hO t—I i —1 O A H H cO A •H Cvj C\J O vO C— A H 4 o ia o CM ■• * ■* • * _IT A - M f A cO K A H A H CM O • A CM O "LA O CM O O ir\v£) H cvj • t • • • co a a a g o A M3 » A r—i m »H pq A £ O -P CCS U *0 O H A oj 0 i —1 O IA • r~| * ITwO A CO M3 CO CM O A H iH H - d - H PI H i-H£1 CM A _ z t A § :8 j 1 £ ♦H *H GQ OQ CQ ca o O r—1 i —1 -P -p ,£ bb bD •H •rH 0 o U .H X II C\i W 106 series of tests are shown* The graphical representations of the correlation be* tween laboratory and field data are shown on Pages 108,109,110, 111* and 112* On each graph, weight losses in the laboratory were plotted against weight losses in the field for each type of iron included in that series of tests. To make the com­ parison of the degree of correlation easier, it is customary to show two regression lire s. To determine these lines, in one case X-^ was taken as the independent variable and X^ as the*dependent variable. In the other case, X£ was taken as the independent variable and X-j_ as the dependent variable. The degree of divergence of the two lines is an indication of the degree of correlation* 100 percent correlation. The line will coincide for a For a zero percent of correlation, the two lines will intercept each other at 90 degrees. To assure linearity of the regression, eash set of data was tested for linearity before the equations of the regression lines were determined* The details of the linare earity test for the data of page 103 / shown in the follow­ ing* The procedure followed is that described by Dixon and Massey (11), on the basis of an analysis of variance. In this method, a test is made to show whether or not the mean for each group of iron is located on a straight line. A comparison is made of the variance within groups with the variance of the deviations of the group means from the es- 107 timated regression line* Total Between Within Regression About regression D*R. Sum Square Mean Square 49 45 k 32,1440 8,349 24,091 1 3 7,701 648 535.36 216 About regression = SS between - SS regression F r 216/535.36 . O .4 0 4 and is to be compared with P = 2.815 at the 95 percent level. Since P is not significant, we accept the hypothesis that the best estimate of the regression curve is a straight line. The study of the regression lines and the correlation coefficient for each series of tests indicated the high degree of correlation which existed between the results of the laboratory tests and the field tests. The data for the fourth series of field tests resulted in a correlation co* efficient which was significant even at the one oercent le­ vel. It was also Interesting to note the difference in the correlation coefficient for the first series of field tests, where the plow’ ooints were nor pur v>n the cultivator at ran­ dom, and the second series of field tests, where the Latin square design was.used. Xn t he former case, the correlation factor was just short of the five percent level, while In the latter, the correlation was significant at nearly ohe one percent level. This is another example showing the 108 16 Figure Test Series No* 1 Correlation of " Controlled " Field and Laboratory Wear Tests of Chilled Iron Specimens 1$ r -. 0 .8 1 CV2 0 0 CiD e 1 23 1U- rH © 13 to 03 O * •P /I CtO •H A. Low Carbon Unalloyed 12 C. High Carbon una3.1oyed •H m co o 13 -p W) •H .tt> A. Low Carbon Unalloyed 12 C. High Carbon unalloyed - E* Low Carbon Alloyed 02 J. High Carbon Alloyed 11 P. High Carbon Nodular Alloyed . High Carbon Nodular Unalloyed 10 X, 7 5 X 8 - Weight Loss in Laboratory - mg 110 115 Do Figure 25 110 Test series No, 3 Correlation of Laboratory and Field Wear Tests "of Unchilled Iron Specimens - 0,93 105 w a 0 (30 100 a 1 'd [ ^ 0 M3 M3 L A M3 L A L A M3 L A M3 L A L A M3 CA CM CA' O O MA CO M3 ~ ztO C\J O J I A H O H O N i—Ii—I 3 P ; ! i ^ CAM3 O O I A P O O C O C O C O C\j L A LQ\ r—I G> CM _ rtC O P „C O C Q -z tv Q O ^ r-J CM CM ( O H H H LA CO 'LA C— CM CO CO CM C— CM P P l A A l O P C— O CM O CM L A 1 A M 3 M3 L A L A M3 M3 M3 M3 O o M3 CA CO r0 0 p i —I EH C9fOOl-^Pr?a'P _cfCO CM M3 O ! c\] C M N H ( A H I A i —ii —I W i —i i —I -~ 4 P CO L A CM oA O O C M 4 C M A CM ! CO. CAvO - c j - l A P CAvO.OQ H O f A P co*m ? O ’ P _ ^ f L A M 3 P C A -C ^ IP . P P P CM p 'L A 'L A _ r j- pq 3 j X Q PI ' 3 o P AJ CME— CM P P CM P C— I P CM CM O O CM L A CM CM O CM CM CM O _j_M3 'O M3 M3 M3 M3 M3 M3 M3 M3 M3 M3 <3 r=q 0-1 cO CO cO AJ _ ± C \)C O M3 Q P L A ^ A P P P P 03 & ^ o a CQ C A P ca O .M 3 P c A A— O p P i—I P 03 03 P <3 W P O A P i 03 PQ 03 o Oh <3 EH <3 P p CM M3 CA P C O CO CO P » CM M3 _ ± P -Z T P P - z r O CM i—I CM i—I O LA LA CM i-AP I P p P O , IA LA LA LA LA I \ A LA LA LA LA I \ A LA LA LA LA ' CO - r M A X A C O CO C— cA 'L A 'L A E— P 1LA LA LA LA LA LA x P P t Lxj M j 1■> I t j I CAM3 _ r ± P ° A O CO, r-L IA C O L A L A i—) AS o o O -p -C O M3 < A O p CM i—1 p L A p *“H j M3 O M3 A O ^ jl A P O A C M A I-^ J * (M 1 L A r i O M 3 CO M 3 . P CA A jC O kCO M a a a d L A P l a P c o ao o o o l a ’ Si Eh i i —ti—i o i —ii —i it AJ L A CO CO CO A -C O O L A P P P CM P vO L A L A L A L A M3 L A o S 3 O H Eh <3 P : O 1 P AJM 3 CM_d-CO p P CO CM CM CM P O O O I L A M3 M3 M3 p P lA O C A 1 I ! CO O O t P C M -c fL A CACO r A O L A CM' CM _ c rp -C M -c f-C M P M 3 .M 3 C O . O L A : x i ; lt \L A | CM CM o fp p a - cm cm ca o co l a i—1 CM i—I i—I 1 CA O CA p c A O CA LA-P*LA LA LA-rrLA LALALALA P CALA o p p M3 O C O CM cm p 03 O o i S3 CO o P *d P -P © & O P S3 *H © Ph * P *H IPn M n a H I h—i ; O , A- i °.| -d M O CM X A CA XJ w 95 Figure 3 X 90 85 Correlation between Brinell Hardness and Weight Loss in Field for Unchilled Iron Specimens r „ 0.95 G 80 - o U 190 210 230 250 Brinell Hardness 270 128 me asure me nt. The problem would be the design of a scratching device which reads directly the load units necessary to pro­ duce a groove of predetermined width across the metal speci­ men. In such a method, the stylus of the device would come in contact with a large secti on of the specimen and thus the effect of variations in micros true ture would be minimized. 127 harder than one with a normal H a k e graphite* The wear re­ sistance or an iron with normal Tlake graphite, however, has been proved to be about four times as much as the iron with abnormal flake graphite. In the case of the unchilled irons, it is possible that the composition and treatment of the iron specimens were such that a micro structure consisting of pearlite and normal graph­ ite had existed uniformly in all the specimens. The hardness measurements in this case were correlated closely with the wear resistance of the irons because of this uniform struc­ ture. In the case of the chilled irons, however, variations In microstructure, which may have been caused by such factors as differences In distribution of iron carbide in the matrix of pearlite, were probably responsible for the lack of cor­ relation between hardness and wear resistance. When the worn surfaces of the miniature plow points as well as several cast iron plowshares were examined with a magnifying glass, it was noticed that the worn surfaces con­ sisted of a number of scratches which appeared as deep and shallow channels on the surface of the metal* It seemed that the weight loss of the plow points was due mainly to the re­ moval of metal from these narrow and wide channels by the abrasive soil. It was though that perhaps a scratch test, carried out with certain precautions, might give results of more practical value than the present methods of nardness VI - SUMMARY AND CONCLUSIONS Summary The wear of tillage tools, especially the plowshare, has long been recognized as one of the agriculturist's fore most problems. It has been estimated that the cost of plow share maintenance constitutes ten to twenty percent of the plowing cost, depending upon the abrasiveness of the soil. A small reduction in wear would bring about a considerable saving In plowshare investment per year to the farmer. There are no basic or scientific data supporting pre­ sent metallurgical practice as to the composition, heat treatment, and hardness of materials used in tillage tools. To make a fundamental and basic study of the require­ ments of tillage tool cutting edges, some workers in this country and abroad have tried to evaluate wear of tillage tools under laboratory or field conditions. The wear test methods proposed by these workers apparently have not been the most desirable method for the selection of material for tillage tools. These methods require either a test sample of special share and size, which must be subjected to wear for periods exceeding 100 hour* s, or they have failed to re­ sult in satisfactory correlation between field and labora­ tory data. The vrork at Michigan State University was started in 19^0 to investigate the possibilities for wear reduction 130 or cast iron plowshares through controlled addition of al­ loying elements. Approximately 200 plowshares were cast at the University foundry. Some actual field tests were made utilizing the ex­ perimental shares. Discouragements in foundry work and field tests plus the lack of sufficient personnel caused the project to be­ come inactive until the beginning of winter term 1954 when the project was reactivated and continued as a thesis problem. Realizing the limitations involved in evaluating wear of tillage tools in the field, it was decided to develop a laboratory wear test method whereby a satisfactory correla­ tion could be established between t h e results of the field tests and the laboratory tests. The object of such a method was to relieve the experimenter from a large part of the var­ iables and difficulties encountered in elaborate and costly field tests. In addition to the development of an accelerated wear test method, determination of the role of carbon as a factor in wear resistance, and the possibilities of using nodular iron as a tillage tool material were additional reasons for underaking the research. In attempting to explore the method of approach to the solution of the problem, studies were made of the mechanism of wear and factors affecting the wear of metals in soil. 131 It was indicated that the wear of metal against metal Is primarily a matter of formation and shearing of metallic junctions. Wear of metal in soil, however, was indicated to be the removal of metal by cutting or scrubbing action of the abrasive particles in soil in the presence of minor or int requent heavy impacts, which are important in crushing, tearing, and chipping off minute metallic particles of the tillage tool. Factors influencing wear of metal in soil were divided into three groups of soil, metal, and operating factors, Among the soil factors, the effects of soil friction, soil moisture, and soil texture were disclassed. Some of the more Important metal factors were indicated to be composition, heat treatment, hardness, and microstructure. Among the operating factors, operating speed and cutting angle were considered to be the most important ones. Other conditions being equal, it was indicated that hardness and impact resistance of the metal are the deter­ mining factors in the wear of metal in soil. Hardness exerts its influence by preventing an abrasive from getting a grip on the surface of the metal. If particles fail to get a grip, they simply slide off. If they penetrate the metal surface, however, the impact resistance of metal becomes effective because the surface will flow under impact and will not crack and chip off as with hard material. 132 The wear experiments were conducted both in the labora­ tory and in small scale field tests. A wear test machine was developed for determining the wear resistance of materials under speed and load conditions which approximated very nearly the field conditions* The addition of a rubber belt around the original type of steel disk as used by Brinell was b e ­ lieved to be responsible for the high degree of correlation obtained between the field and laboratory wear test data. The rubber belt not only reduced the crushing of sand and pro­ vided a uniform loading; it also gave a cushion action to the sand particles similar to the action of soil on surface of tillage tools* This condition produced a wear appearance of considerable improvement over that of the steel disk used without a rubber belt* After the laboratory tests, the same spec huen was placed on a cultivator and tested under actual field conditions* Up to 12 specimens could be tested in the field at one time* Preliminary field tests of duplicate samples indicated variations in weight losses exceeding 200 percent. Since the control of field variables was extremely difficult, if not impossible, the experiments were conducted according to 6 x 6 or 5 x 5 Latin square designs, and the results were analysed by the method of analysis of variance* Experimental samples were made to include ordinary cast iron and nodular iron of high carbon or low carbon and 133 alloyed or unalloyed compositions in the chilled or unchilled conditions. There were 122 miniature plow point specimens 0 made from 11 types of irons* The field tests were conducted in four series which included 22 individual field tests. 'The first and the second series of field tests were conducted to determine the relative wear resistance of the chilled iron specimens and also to compare the apparently controlled and the randomized experiments* The third series of field tests was set up to determine the relative wear resistance of the unchilled irons* The fourth seriep of field tests contained both chilled and unchilled iron specimens to determine the relative wear resistance of the gray and white irons and the reproducibility of results obtained in the previous field tests. Correlation studies were made to find the degree of correlation between the results of laboratory tests, small scale field tests, and the large scale field tests of the cast iron plowshares made during the early stages of the project. Also, correlation between hardness of the plowpoints and the weight losses in the field viere determined. Conclusions The results and conclusions of the laboratory and field tests as to the evaluation of the wear test method, the de­ sign of the experiments, the comparative wear resistance of 13k the irons, and the hardness studies can be summarized as f ollows: 1* Data showing the correlation between the laboratory wear tests and the Tie Id wear tests resulted in correlation coefficients which were significant up to the one percent level. This high degree of correlation plus the simplicity and the rapidity with which tests could be made indicated the suitability of this method of wear test for the selection of tillage tool materials. Laboratory wear tests can be made on small samples of materials in less than ten minutes, where as field tests require hours of field operation, subjected to weather conditions, and a much higher cost of labor and machinery. Those materials that show superior wear resist­ ance can be tested for impact resistance, scouring, and other characteristics necessary for superior tillage tools. 2. The most critical factor influencing the reproduci­ bility of weight losses by means of the wear test machine was found to be the range of the sand particle size* For a high degree of correlation and maximum reproducibility of results, the us© of dry mortar sand wi th particle size in the range of 2 8 -mesh to [|.0-mesh is recommended in addition to the use of a rubber belt, and the specified load and the speed of the rotating wheel. 3* In such an agricultural engineering problem where 135 control of all the field variables is almost an impossible task, the design of the experiment according to some sta­ tistical method appears to be the most logical aporoach. The randomized experiment not only offers more reliable data and considerable savings in time and expenses, it also provides a method for finding the best estimates for the missing data of the specimens which may have been damaged or lost during the field operations. !(-. Comparison of the wear resistance of the irons in soil under both field and laboratory conditions resulted in the following rating in the order of decreasing resistance to wear. Chilled high carbon unalloyed Chilled high carbon alloyed Chilled high carbon nodular unalloyed Chilled low carbon unalloyed Chilled high carbon nodular alloyed Chilled low carbon alloyed Unchilled high carbon nodular unalloyed Unchilled low carbon alloyed Unchilled low carbon unalloyed Unchilled high carbon alloyed Unchilled high carbon unalloyed When another series of field and laboratory tests were con­ ducted to check the reproducibility of the most significant results obtained in the previous tests, the irons were placed again in the same order shown above. Comparison of chilled and unchilled irons of the same composition and treatment under the same field test conditions indicated that an unchilled iron will wear out 136 four to Tive times as much as the corresponding iron in the chilled condition. 6. As far as the correlation between the results of the laboratory and the large scale field tests is concerned, when the loss of material at the chilled sections of the plowshares, that is, the point and the cutting edge, was considered, the high carbon iron was superior or at least equal to low” carbon iron in wear resistance. This result was in perfect agree­ ment with what was found in the case of laboratory and field tests of the miniature plow points. 7. Correlation between Brinell hardness of the minia­ ture plow points and their weight losses in the field resulted in a rather high degree of significance in the case of un­ chilled irons but the correlation was far from significant in the case of chilled iron specimens. This inconsistency proved again that the present methods of hardness measure­ ments which do not differentiate between the microstructures of irons with the same hardness, cannot be taken as an index of wear resistance of cast iron in soil. It was proposed that perhaps a scratch test might give results of more prac­ tical value than the present methods of hardness measurements. 8. As for the suitability of nodular iron as a material for cast iron plowshares, it was concluded that if a plow­ share is to consist of both chilled and unchilled irons, a high carbon unalloyed nodular iron promises to be as good 137 a material for this purpose as car: be made availabe at the present time. This conclusion was made on the basis of the experimental data showing such properties of nodular iron as: its moderate resistance to wear in the chilled condi­ tion, its highest resistance to wear in the unchilled condi­ tion, its ductility and impact resistance, its property of developing a chilled surface of controlled depth, and its greater hardness in the unchilled condition which aids scouring of the plowshare. 138 Literature Cited 1 . Avery, H.3. Hard Surfacing by Fusion Welding. Brake Shoe Company, 19l|_7. American 2. Beresford, H. and E.N. Humphrey. Prolonging Plowshare Service. Idaho Agricultural Experiment"station, Moscow. Bull. 202, 1934. 3. Blake, J.M. Wear Testing of Various Types of Steels. Proceedings, American Society for Testing Materials. Vol. 28, 1 9 2 8 , pp. 3 41-355. Ip# Bornstein, H. Ferrous Metals: Their Treatments and Properties for Agricultural Machinery. Agricultural Engineering. l5(August 1934), PP. 21&-2TT. 5. , The Wear of Farm Tools. 9 (November 1938), pp. 241-245. Metals and Alloys. 6. Brinell, J.A. Researches on the Resistance of Iron, Steel, and Some Other Materials for Wear. Jernkontotests Annaler. 78(1921), pp. 347-398. 7. Brown, E.J. Castings to Resist Abrasive Wear. Trade Journal. 92(1952)> PP. 587-595. 8. 9# 10. 11. Foundry Burwell, J.T. and C.D. Strang. On the Emperical Law of Adhesive Wear. Journal of Applied Physics. 23(January 1952), pp. 18-287 Carle ton, W.M. and J.W. Martin. Sharpening and Hardsurfacing Plow and Lister Shares, Kansas Engineering Experiment Station, Manhattan, Bull, 44, 1945* Crother, E.M. and W.B. Haines. An Electrical Method for Reduction of Draught in Plowing. Journal of Agricultural Science. 14(1J24), pp. 221-231. Dixon, J.W. and F.J. Massey. Analysis. Introduction to Statistical McGraw-Hill Book Company, Inc., New York, 1951, p. 1 6 0 . 12. Fisher, R.A. The Design of Experiments. and Boyd, Edinburg and London, 194-7• 13. Gallwitz, K. Werkstoffe und Abnutzung von Pflugscharen. Landwirtschftllche Yahrbucher. 72(1930), pp. 1-50. ifth ed. Oliver 139 l^J-* Geiger, H.L. and h.w. Northrup, a New Metal for Farm Tool Components. Agricultural Engineerinc. 32fMarch lQ^i ). pp. 114-3-11+7. ** ' 15. Gillet, H.W. Considerations Involved in the Wear Testing of Metals. Symposium on Wear of Metals. American Society for Testing Materials, (1937), pp. 3-23. 16. Goulden, O.K. Methods of Statistical Analysis. 2nd. ed. John Wiley and Sons, Inc., New York, 1952, p. 32l+. 17. Haworth, R.D.The Abrasion Resistance of Metals. Trans­ actions of American Society for Metols. Vol. lj_l, 191+9, pp. 8 1 9 -8 6 9 . 18. Hill, E.B. and R.G. Tlawby. Census Maps of Michigan 1950. Agricultural Economics Department, Michigan State College, East Lansing, July 1952. 19. Hoffman, A.H. Engineering. 20. Holz, Herman. Brinell Researches on the Resistance of Iron, Steel, and Some Other Materials to Wear. Testing. 1(February 1921+), pp. 101+-11+6. 21. Rummer, F.A. The Dynamic Properties of Soil, VII. A Study of the Nature of Physical Forces Governing the Adhesion Between Soil and Metal Surfaces. Agricultural Engineering. Vol. 19, pp. 73-78 (1938). 22. Nichols, M.L. Dynamic Properties of Soil Affecting Im­ plement Design. Agricultural Engineering. 1 1 (June 1930), pp. 201-201).. Wearing Tests of Plowshares. 3(May 1922), p. 8 3 . Agricultural 23 • , The Dynamic Properties of Soil, I. An Exolanation "of the Dynamic Properties of Soils by Means of Colloidal Films. Agricultural Engineering. 1 2 (April 1931), PP. 259-261).. 2l+. , The Dynamic Properties ofSoils, II. Soil and Metal Friction. Agricultural Engineering. 12(August 1931), PP. 321-322+. 25. , The Sliding of Metal Over Soil. Engineering, 6 (April 1925), PP. 80-81+. 26. Otaco Limited. May 3, 1951. Orillia, Ontario. Agricultural Personal correspondance, li+o 27. Ransom©, David. Report to the Institution or British Agricultural Engineers. The Engineer. 193(Februarv 1992). P. 191. 28. Reed, I.P. and E.D. Gordon. Determining the Relative Wear Resistance of Metals. Agricultural Engineering. 3 2 (February, 1951), PP. 98-100. 29. Rosenberg, S.J. The Resistance of Steels to Abrasion by Sand. U.S. Bureau of Standards Journal ol Research. Vol. 5, 1930. pp. 553-574* 30. Sanders, Kain, Hardy, and Gordon. Institute of British Foundrymen Proceedings. 1949, p. 1&9. 31. Snedecor, G.W. Statistical Methods. 4th ed. State College Press, Ames, 1950, p. 149. 32. Stott, B.L. Ductile Iron as an Engineering Material. Society of Automotive Engineers. Preprint Wo. 16, New York: 1953, p. 2. 33. Tetsutaro Mitsuhasi, Yutaki Imai, and Shin Yokoi. Study of Wear of Steel Against Soils. Journal of Mechanical Laboratory. 4(^aY an8 December 19507, pp. 91-96 and! 291-298, 5 (November 1951), pp. 294"298. 34* Thibault, N.W. and H.L. Nyquist. Knoop Hardness Testing. Transactions, American Society for Metals. 38(1938), pp. 271-330. 35. Vanick, J.S. Reducing Metal Wear with Abrasion Resisting Castings. The American Ceramic Society. 26(April 1947), pp. 109-116. 36. Weiss, C.R. Iron Age. The Xowa Relative Wear of Metals Due to Abrasion. Vol. 129, 1932, p. 1166. 37. Yates, F. Empire Journal of Experimental Agriculture. Vol. L, 1933, pp". 129-1427 38. Zink, F.J. , G.A. Sellers, and J. Roberts. Results of Studies of the Cutting Edges of Tillage Implements. Agricultural Engineering. l?(March 1936), pp. 93-97.