a?“ k I llumyuzuyuaumumw"will: ' A? Y ‘ Midi;mn :6. 3 4% Umver 51 t y This is to certify that the thesis entitled System Modeling on Rice Milling Technology in Indonesia presented by Eriyatno has been accepted towards fulfillment of the requirements for Ph.D. degree in Agricultural Engineering 777%,59“? DateF/EOS/97? 0-7639 OVERDUE FINES ARE 25¢ PER DAY PER ITEM Return to book drop to remove this checkout from your record. :1 l . I r 3 13“" W 5: 100 “£218 '7'?“ A 5‘ My ifiz-z'ss’g; . O “A“ a,“ . ‘ 20 A 326 1fi§£”' Y. UQM' 16, (2‘54 SYSTEM MODELING ON RICE MILLING TECHNOLOGY IN INDONESIA BY Eriyatno A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Engineering 1979 ABSTRACT SYSTEM MODELING ON RICE MILLING TECHNOLOGY IN INDONESIA BY Eriyatno Three computer models were developed to simulate the regional rice mill operations. The first model was designed to estimate rice mill losses. Field and labor- atory measurements of rice milling performances were an- alyzed and equations were developed for constructing the systems model. The rice mill model estimated the average mill losses in Indonesia to be 4.8 percent of rough rice production or about 3.1 percent of milled rice. The model showed that rice milling mechanization could appreciably reduce mill losses while also increasing mill yields. Im- provement of pre-mill drying and grain cleaning facilities could improve the milling yeild by 2 percent and reduce the mill loss by 2.2 percent. The second systems model was designed to forecast the regional rice mill production. Secondary data on har- vested areas and yield per hectare were used in generating the dynamic model. The average annual increase of milled rice in Indonesia was projected to be 2.6 percent with the assumption that the post harvest technology was fixed. The total post harvest losses were estimated to be from Eriyatno 15 to 20 percent. Half of these losses were due to stor- age practices. The third systems model was an optimization compu- ter program designed to evaluate the regional mill ca- pacity and distribution within a region, based on a fin- ancial analysis. The model was applied to the subdis- tricts of Ciomas and Ciawi in West Java province. Field study was conducted to compile data for model verifica- tion. The model indicated an over capacity of mechanized mills in Ciomas and an under capacity in Ciawi. The sim- ulation results showed that the simplest technology, the Engelberg Steel Huller and Rubber Roll Huller were gener- ally more favorable for a mill mechanization plan. How- ever, the milling quality consideration could shift the technology. Once the optimal condition was achieved the model forecast that the regional mill capacity in Ciomas should be increased by 3 percent and in Ciawi by 8 per- cent per year in consequence of the annual increase of regional rice production. Approved by Major Professor Department Chairman Dedicated to my late mother ii ACKNOWLEDGMENTS The author wishes to express his sincere gratitude to his major professor, Dr. Merle L. Esmay for his guid- ance and help throughout the course of this study. His capable counsel, unfailing interest, and continuing en- couragement made the work particularly enjoyable. Appreciation is also extended to other members of his guidance committee: Dr. Herman E. Koenig, Dr. Law- rence O. Copeland and Dr. Robert H. Wilkinson. The author feels deeply indebted to Mr. Abdulmad- jid, Mr. Soegito and Mr. Kartono of the Central Bureau of Statistics in Indonesia, for the research assistant- ship that enabled him to undertake the investigation. Special appreciation is also extended to U.S. Agency for International Development for their finan- cial support. Thankful acknowledgment is also due to a wide var- iety of people too numerous too mention from the Insti- tut Pertanian Bogor, the Biro Pusat Statistik and the Badan Urusan Logistik. The understanding and patience of the author's wife, Ayi Sukarti, is sincerely appreciated. Her assis- tance during the years have been invaluable. iii TABLE OF CONTENTS Page LIST or TABLES . . . . . . . . . . .l. . . . . . . . vii LIST OF FIGURES . . . . . . . . . . . . . . . . . . ix CHAPTER I INTRODUCTION . . . . . . . . . . . . . . . . 1.1 Objective . . . . . . . . . . . . . . . 1.2 Methodology . . I I “VIEW OF L ITERATURE O O O O O C O 0 C O O O o o o o o o o o o o o 0 ON \1 U143 H 2.1 Post Harvest Loss . . . . . . . . . . . 2.2 Rice Production Estimation . . . . . . . 2.3 Rice Milling . . . . . . . . . . . . . . 12 2.3.1 Equipment and Practices . . . . . 13 2.3.2 Selective Mechanization . . . . . 14 2.3.3 Selection Criterion . . . . . . . 19 III MILLING LOSS MODEL . . . . . . . . . . . . . 24 3.1 System Constraints and Desirable Model Characteristics . . . . . . . . . . . . 25 3.2 Input Data . . . . . . . . . . . . . . . 26 3.2.1 Weather Factors . . . . . . . . . 26 3.2.2 Moisture Content of Rough Rice . 27 3 . 2. 3 Mill Conversion Curve . . . . . . 27 3.2.4 Milling Yield . . . . . . . . . . 29 3.3 System Design . . . . . . . . . . . . . 33 3.4 Model Implementation . . . . . . . . . . 41 3.4.1 Rough Rice Moisture Content . . . 41 3.4.2 Milling Loss Level . . . . . . . 42 3. 4.3 Milling Technology . . . . . . . 44 3.4.4 Pre-Mill Operation . . . . . . . 45 iv IV REGIONAL RICE PRODUCTION MODEL . . . . 4.1 4.2 4.4 System Constraints and Desireable Model Characteristics . . . . . . . . . Input Data 0 O O O O O O O O O 0 4.2.1 Harvested Area . . . . . . 4.2.2 Yield Per Hectare . . . . 4.2.3 Post Harvest Technology . System Design . . . . . . . . . . 4.3.1 Growth Model Approach . . 4.3.2 Curve Fitting Procedure . 4.3.3 Model Construction . . . . Model Implementation . . . . . . 4.4.1 Growth Model . . . Harvested Area . . Yield Per Hectare 4.4.2 Forecasting Model Harvested Area . . Yield Per Hectare . . Milled Rice Production REGIONAL RICE MILL MECHANIZATION MODEL 5.1 System Constraints and Desirable Model 5.2 5.4 Characteristics . . . . . . . . . Input Data . . . . . . . . . . . 5.2.1 Harvesting Frequency . . . 2 Rice Mill Classification - 3 Regional Prices . . . . . System Design . . . . . . . . . . 5.3.1 Model Structure . . . . . 5.3.2 Variable Search Method . . 5.3.3 Model Construction . . . . Model Implementation . . . . . . 5.4.1 Mechanized Mill Regional Capa- city . . . . . . . . . . . 5.4.2 Milling Size Selection . . Regional Mill Distribution Milling Class Comparison . Plan Projection . . . . . Page 47 48 49 49 49 50 51 51 52 55 55 55 55 60 62 62 62 69 71 74 74 74 76 77 78 78 88 88 88 91 91 93 94 Page VI CONCLUSIONS . . . . . . . . . . . . . . . . . 96 VII SUGGESTIONS FOR FURTHER STUDY . . . . . . . . 100 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . 102 APPENDIX A: POST HARVEST SURVEY IN INDONESIA, 1977/1978 C C I O C O O C O O C O O O C 107 Vi Table 2.1 2.2 2.3 LIST OF TABLES Estimated Post Harvest Losses in Indonesia in Percent of Milled Rice . . . . . . . . . Rice Production in Indonesia, Estimated in 1000 Tons of Milled Rice . . . . . . . . . Milling Process Output in Percent of Rough Rice 0 O O O O O O I O I O O O O I O O O 0 Classes of Milling Facilities . . . . . . . Regional Climatic Factors, Yearly Average . Milling Yield Performance in Percent of R0 ugh Ri c e O O O O O O O O O O O O O O O 0 Distribution of Mill Technology . . . . . . Rough Rice Moisture Content, Yearly Aver- age in Percent wet Basis . . . . . . . . . Estimated Mill Loss . . . . . . . . . . . . Milling Practices Performances in Percent Of Rough Rice 0 O O O O O O O O O O O O O 0 Simulated Pre-Mill Operation . . . . . . . Parameters for Harvested Area Growth Func tions 0 O O O O O O O O O O O O O O O 0 Simulation of Rice Production in Indonesia, 1978-1988 in Thousand Tons of Milled Rice . Mechanized Rice Mill Criteria . . . . . . . Mill Technology Cost Input . . . . . . . . Results of Regional Mill Capacity Simula- tion 0 O O I O I O O O O I O O O O O O O 0 vii Page 11 15 '16 26 37 38 41 42 44 46 57 70 76 78 89 Table 5.4 Regional Mill Distribution . 5.5 Mill Class Comparison Score viii LIST OF FIGURES Figure 3.1 Equilibrium Moisture Content Curve Ambient Temperature = 25°C . . . . . . . . . . . . . 3.2 Milling Yield Curve . . . . . . . . . . . . 3.3 Schematic Illustration of an Engelberg Type Huller with a Polisher . . . . . . . . . . . 3.4 Schematic Illustration of Rubber Roller Huller O O O O O O I O O O O O O O I O O O O 3.5 Schematic Layout of Rubber Roll-Steel Huller Mill Rice Mill Unit . . . . . . . . . . . 3.6 Schematic Lay-Out of Stone Disc-Steel Huller Mill, Rice Mill Plant . . . . . . . . 3.7 Mill Loss Model . . . . . . .l. . . . . . . 3.8 Simplified General Flow Diagram, Mill Loss MOdel O O O O O O O O O O O O O O O O O O O 4.1 Simplified Flow Diagram of Forecasting Model for Regional Rice Production . . . . . 4.2 Area Change Trends for Harvested Rice in Java 0 O I O O O O O O O O O O O O I O O O O 4.3 Area Change Trends for Harvested Rice in Off-Java o o o o o o o o o o o o o o o o o o 4.4 Rice Productivity Trends Yield per Hectare . 4.5 Forecasting Model Verification for Harvested Area of Wet Land Rice, in Java . . . . . . . 4.6 Forecasting Model Verification for Harvested Area of Dry Land Rice in Java . . . . . . . 4.7 Forecasting Model Verification for Harvested Area of Wet Land Rice, in Ofwaava . . . . . ix Page 28 30 32 33 34 35 39 40 56 58 59 61 63 64 65 Figure 4.8 4.10 5.1 Forecasting Model Verification for Area of Dry Land Rice, in Off-Java Forecasting Model Verification for Rice Productivity . . . . . . . . Forecasting Model Verification for Rice Productivity . . . . . . . . Frequency Distribution of Farmers' ing Tim 0 O O I O I O O I O O O O Fibonacci Logic Diagram . . . . . Page Harvested O O I O O 66 Wet Land O O O I O 67 Dry Land . . . . . 68 Harvest- . O O O O 75 . . . . . 86 Simplified Flow Diagram for Regional Rice Mill Mechanization Model . . . . . O O O O O 87 Forecasting Simulation of Rough Rice Pro- duction in Ciomas and CiaWi . . . . . . . . 95 CHAPTER I INTRODUCTION Since 1950 Indonesia has been importing from 500:000 to 2,000,000 tons of rice per year. With the population growth rate at an average of 2 percent per year and more people preferring rice as their staple food, the rice con- sumption consequently increased from between 3.6 to 4.6 percent per year. The application of high yielding variety, fertilizer, better irrigation and expansion of new planting areas, was estimated by the Central Bureau of Statistics (BPS) of Indonesia to contribute only about 1 to 2 percent in increasing harvested paddy fields and 2 to 3 percent in increasing the yield per hectare. Considering this trend, it seems that the future situation will never be enough to reach a self supporting level for this 130 million pe0ple nation. Hence, alternative efforts must be introduced, and one of them is the improvement of a post harvest technology. The main objectives of the post harvest technology im- provements are the reduction of grain loss, better food quality and by-product utilization. The grain loss is a major issue due to the fact that substantial amounts of food may be saved. The losses occur in field harvesting, threshing, drying, milling, storage and transportation. Various studies have been done to assess post harvest losses at different stages of operation (IRRI, 1977; Ilangantileke, 1978). In Asia, the losses have been estimated from 8 to 30 percent of the rice production and potential savings through improvement of post harvest practices may be up to 10 percent (Esmay gt 31., 1977). The magnitude of losses depends upon plant variety, specific regional condition, technoloqy practices and equipment. A reliable estimate of grain losses is required so that new technology improvement may be applied effectively. Grain milling operations have been treated as a major part of the overall post harvest development program. Rice milling includes the process of removing the hulls and bran from the rough rice to produce edible rice. The milling operation is closely related to rough rice drying and stor- age facilities. The milling output determines the food quality; therefore it highly affects the sales price. The business of rice milling has an important role within the regional food marketing and distribution systems. Depend- able planning and control of the milling development program is critical in order to provide better food quality at a price within the peOples' purchasing power. The milling process equipment ranges from the simplest pestle and mortar to complex automated systems. In Indon- esia more than half of the rice milling was done by hand pounding in 1947. The introduction of small mechanical milling units in 1960 reduced hand pounding to about 18 percent in 1974 (Institut Pertanian Bogor, 1975) and to about 6 percent in 1977 (Post Harvest Survey - BPS, 1978). These changes may induce an unfavorable impact on labor dis- placement or fuel energy utilization. The output of large mechanized rice mills with cap- acities of more than 1.0 ton of rough rice per hour has been decreased by nearly 50 percent in 1947 to 12 percent of the national rice production (IPB, 1975). The Post Har- vest Survey done by BPS (1978) indicated that there has been a strong tendency to increase the small milling facil- ities in the rural areas such as: the Engelberg Steel Huller and the Rubber Roll Huller with capacities of 0.2 to 0.6 tons of rough rice per hour. The tradeoff between operational cost and milled rice quality might be the factor in evaluat- ing this kind of development. The survey by BPS (1978) indicated that almost 70 per- cent of the mechanical mill facilities in Indonesia were operated less than half of their yearly potential capacity. This low productivity indicates improper planning distribu- tion of technology facilities. The size and site of most milling facilities was not appropriately proportioned to the given regional rice production. This situation has produced an extremely high competitive atmoSphere among the millers with the result that many are going out of business. Among the survivors, more than half operate at a very low profit margin. Selective mechanization must be introduced to improve technology for the needs of farmers, millers and consumers. The goal is to achieve such an appropriate technOIOgy within particular regional areas. Adequate regional factors and parameters must be identified for the construction and oper- ation of goal-seeking systems. The systems approach can help provide useful results for the planning agencies. The millers also would be assured of a viable business. The complexity of rice milling problems made it diffi- cult to compare experimental treatments. To compare differ- ent technology systems under field conditions all variables should be held constant except the one being observed. As this is impossible to do in the real world, a computer model- ing approach was used. 1.1 Objectives 1) Evaluate existing post harvest practices and prob- lems, and identify the regional factors that affect the development of milling technology. 2) Develop a computer model for assessment of regional rice mill losses for various post harvest tech- nology and under different weather conditions. 3) Construct a regional rice production model based upon post harvest losses, harvesting area, yield per hectare and various technology practices. 4) Design a regional mechanization model for rice milling with respect to regional rice production, seasonal harvesting time, machinery performance and associated costs. 1.2 Methodology One year of in-country research was carried out in Indonesia from July 1977 to July 1978. Two national field surveys were conducted during the dry and wet harvesting seasons to compile first hand information on recent post harvest practices and problems. Each survey consisted of data collection and on-the-spot observations. The survey- ors interviewed farmers as well as mill operators and man- agers (see Appendix A). Respondents were selected by random sampling. This survey was sponsored by the Central Bureau of Statistics (BPS) of Indonesia and was executed by a multi- disciplinary team consisting of 5 agricultural engineers, 6 food technologists, 5 staticians and an agricultural econ- omist. The survey included 11 rice production provinces in Java, Sumatra, Kalimantan, Sulawesi and Bali. A total of 1433 farmers were interviewed and 323 milling facilities were visited. Questionnaires were completed and samples of rough rice and milled rice were collected from farm storages, millers and markets. These samples were analyzed as to physical and nutritional quality at the Food Labora- tory of Bogor Agricultural University (IPB). Several experiments were done at the Rice Processing Center at Tambun and various processing facilities in West and East Java to measure machinery performances and identify milling losses. A regional mechanization program focused on the sub-districts of Ciawi and Ciomas near Bogor. Secondary data were gathered from research and governmental institu- tions. Technical discussion, consultation and seminars were also held. The data were processed partly in Indonesia. The sim- ulation modeling and computer application was done at Mich- igan State University. Numerical methods for polynomial regression was extensively used in formulating inter-rela- tionships among variables. Optimization techniques were used in the selective mechanization model. The Fibonacci search method for minimizing functions with constraints was applied. CHAPTER II REVIEW OF LITERATURE The design of rice milling technology systems involves the determination of post harvest losses at mill facilities, estimation of regional rice production and selection of appr0priate mill machinery. 2.1 Post Harvest Loss_ An FAO survey in 1977 indicated a lack of adequate quantitative and qualitative post harvest loss data. Re- ports from sixteen countries showed that no information was available. Reports from eighteen other countries pro- vided only gross estimates, frequently with considerable disparity between minimum and maximum loss estimates; thus, reflecting conflicts of opinions and unrealibility of loss levels. Post harvest losses of rice in Indonesia have been estimated from limited case studies by several institutions. The results have been quite unreliable. Losses have been found to range from 10 to 30 percent of total production. Table 2.1 shows some of the sketchy results. No current es- timation of threshing losses was available. Table 2.1: Estimated Post Harvest Losses in Indonesia in Percent of Milled Rice Operation A* B** C *** l. harvesting 8.0 8.0 - 2. threshing - - - 3. drying 2.0 2.0 - 4. milling 4.5 4.5 - 5. farm storage 5.0 4.0 - 6. non-farm storage - 1.0 - 7. in-field transportation — 2.0 2.72 8. off-field transportation 5.5 1.48-3.12 2.0 9. others - 0.26-0.62 2.0 *USAID, 1971 **National Bureau of Logistics, 1974 ***Central Bureau of Statistics, 1977 Grain losses may be classified into four catagories: First is mechanical loss from the mechanical operations of threshing and milling. The losses consist of broken kernels, unhusked rice, scattered and unthreshed grain. Second is loss due to rodent, birds, insects, chickens and other an- imals that intervene during the post harvest period. Third is loss caused by microorganisms such as fungi. Fourth is loss due to the environmental facturs such as temperature and humidity. Fermentation and other chem- ical deteriorations are included here. The wide variation within each catagory makes it necessary to have many measure- ments to be representative of actual losses (Brooker 3E 31., 1975; Tuquero 33 31., 1977). The available data on levels of loss usually are in- sufficiently detailed and inconsistently representative to reliably indicate all losses at all post harvest stages. It has been established that losses in farm storage are rela- . tively high for corn and rice and that physical and proces- sing losses are considerable for rice (FAO, 1977). Esmay (1978) stated that attempts to reduce post pro- duction losses have been somewhat piecemeal through the in- troduction of various technological changes in different areas under various conditions. Seldom have these trial introductions considered the effect of all post harvest operations, the possible social cost of labor displacement, equitable income distribution and capital requirement. 2.2 Rice Production Estimation Generally a rice production estimation is used to support the food policy and regional development program. Several methods have been introduced to calculate present and projectory production of rice within a regional boundary. The Central Bureau of Statistics of Indonesia (BPS) used the statistical approach forecasting method through various regression analysis of past data (BPS, 1978). Basically, the BPS formula for predicting rice production at a given time is: 10 RPROD (t) = AVHST (t) x YPHA (t) x (CONV - LOSS) where, RPROD = estimation of rice production, in milled rice AVHST = harvested area, in hectares per unit time YPHA = yield per hectare, in kilograms of dry stalk paddy CONV = conversion from dry stalk paddy to milled rice (0.68) LOSS = combined losses constant, transportation (0.0472) and feed (0.0136) The milling conversion parameter which covered from dry rough rice to milled rice varied from 0.60 to 0.74 depending upon the moisture content of rough rice, milling techniques and quality of milled rice being pro? duced (IPB, 1975; IRRI, 1977). Research conducted by Bogor Agricultural University (IPB) together with the BPS-proposed average mill conversion was 0.6455 with re- spect to the technological level in 1964. Table 2.2 shows some results of rice production estimation. Techniques of forecasting usually include data on the smoothing process and on the stochastic model. Most of the agricultural production has seasonal fluctuations which should be considered in evaluating its historical data (Hughes, 1974; Tulu 3E 31., 1974). Several growth patterns used in forecasting analysis have been implemented specifically (Forrester, 1964; Pimentel 33 31., 1976; 11 Table 2.2: Rice Production in Indonesia, Estimated in 1000 Tons of Milled Rice 1* Year 2** Low High 1978 14,925 15,840 -- 1979 15,344 16,382 -- 1980 15,729 17,058 15,290 1981 16,123 17,728 -- 1982 16,522 18,508 -- 1983 16,925 19,195 -- 1984 17,335 19,963 15,880 *BPS, 1978 **Schmidt, 1976 Goodman, Kormondy (1969) explained that although growth trends are difficult to come by, there are enough studies on a spectrum of different kinds of plants and animals to permit the statement that most species show a sigmoidal pattern during the initial stages of their population growth. of growth, There is, in such cases, an initial slow rate in absolute numbers, followed by an increase rate to a maximum, at which point the curve begins to be deflected downdard. It terminates in a rate that grad- ually lessens to zero, as the population more or less 12 stabilizes itself with respect to its environment (For- rester, 1968; Goodman, 1974). 2.3 Rice Milling In order to visualize the effect of the milling process on rice, a brief understanding of the physical structure of the rice grain is very essential. In spite of varietal differences, such as short, medium and long grain varieties, the structure of the grain is virtually the same. The hull (lemma and palea) which is loosely attached to the edible rice grain within, is siliceous, hard and hairy. Directly beneath, but separated from it, and firmly attached to the starch body (endosperm) of the grain, is the bran coat or layer which is arranged in enveloping seven types of layers around the endosperm in the following order from outside in: (a) epicarp, (b) mesocarp, (0) cross cells, (d) tube cells, (e) sperm- odern, (f) perisperm and (g) alleurone layer. The actual weight portion of rough rice is 0.65 - 0.70 kernel (endOSperm), 0.17 - 0.22 for hull and 0.08 - 0.09 for bran (Wratten 33 31., 1964; Myasnikova, 1969; Pandya, 1969). During the process of milling, most of the hull and the outer six layers of bran i4; removed while some por- tion of the aleurone layer stayed with the edible grain. Grain with hull removed is known as brown rice. It con- tains more protein and vitamins than milled rice but it 13 was reported to cause digestive disturbances. Milled rice is more attractive in appearance, requires less time to cook, and maintains quality longer in storage than brown or undermilled rice, which becomes rancid easily. Nowadays, the milling business is an essential part of food production, either as a rural industry or on a greater commercial scale (Camacho 33 31., 1978; Yasuma, 1972). 2.3.1 Equipment and Practices A good rice milling operation should 1) produce the maximum yield of edible rice, 2) obtain the best possible quality, 3) minimize losses and 4) minimize the Processing cost. Modern rice mills will produce some 5 to 10 per- cent higher output of milled rice than traditional mills. Furthermore, the head yield of milled rice increases and the losses are minimized (Esmay 33 31., 1977). However, there are other factors involved. These are: moisture content of the rough rice, the condition of the grain as to checks and cracks previously incurred by improper harvesting, threshing, handling and drying methods. The most neglected basic requirement for improv- ing the output of milled rice in the tropics is to have a good condition of the input rough rice (Van Ruiten, 1975; Timmer, 1972). Percentage of seedless grain, which is approximately 2.0 to 5.5 percent, and foreign material content which is approximately 0.5 to 2 percent, also 14 affect the mill output. Proper grain drying and cleaning may improve this condition (IPB, 1975). The heterogenity of mill types used in the tropics causes considerable variability in the ratios of milled rice recovery and head rice yield. Table 2.3 presents some data on the efficiency of five milling operation systems. U Thet Zin 33 31. (1974) suggested that the milling efficiency should more apprOpriately be made on the milling process rather than on huller types such as the steel huller, disc huller or rubber roll huller. Many of the rice mill systems in the tropics do not have a complete line of standard components such as rough rice cleaner and separators. Table 2.4 shows the separate milling processes on various combinations that may be found in the system. The recovery of milled rice generally in- creases with the addition of component parts. Column J represents a complete and complicated milling found in the modern rice milling industry, which generally has a capacity of more than 3 tons of rough rice per hour. 2.3.3 Selective Mechanization Giles (1973) stated that only some machines when used in some farming situations and under some management expertises contribute significantly to increase yields. Giles introduced selective mechanization term as that is selective in machines, farming situations and managers. 15 eeaa .mdm one mmHA my . a mnma mm QANV .HTEEH mnma .BAHV Incas . H.He I: lmem.mm n- u: I: panda Hag: moam lmea.eo-~.mm.1~.os Awem lmewe lmeom Amos lmemm umaasm Haom umnnsm Amem.ae ANCG.HH lmem.mm Amem.~m .. .. Hag: meoum omen lmee.me.lave.mm AHCQ.GH Idem.ee lace.em .. u- “made: Hmmum mumnammcm lmee.eenm.oe.laeoe lavew laces laces .. It meanesom came Houoe :oxoum pom: coum+xmsm comm xmom :oflumoflwwommm deuce COMM OOHHHZ poem nmoom mo unmoumm ”Hm usauso mmmoonm mcflaaflz "m.~ magma 16 Table 2.4: Classes of Milling Facilities Class Standard Process A B C D E F G H I J a. Grain Cleaner x x x x x x b. Stalk paddy thresher x x c. Rough Rice grader x x d. Huller x x x x x x x x x x e. Husk separator x x x x x x x x x f. Rough rice separator x x x x x x x x g. Polisher, first stage x x x x x x x h. Polisher, second stage x x x x i. Milled rice grading x x x x x j. Grain conveyor x Source: Esmay 33 31., 1977. Pinches (1956) stressed that an agricultural engineer concerned with the integration and balancing of a system will be involved in economic relationships as well as physical and biological judgment. There is evidence that many rice post production systems in the tropics have been improperly planned (Rawnsley, 1972; Weitz-Hettelsater, 1972). There are a total of 31,698 rice milling facilities in Indonesia in 1978, with 940 having about 12.8 milled rice tons per year capacity and 30,758 small mills having 1.2 milled 17 rice tons per year. Recently, about 40 percent of the large mill facilities were out of business and most of the remainder were not operating in full capacity (Post Harvest Survey, BPS, 1978). The mill location, type and size of equipment are critical for effective planning. Weitz-Hettelsater con- sulting engineers recommended only large capacity, cap- ital rich, labor efficient mills for Indonesia, which has limited capital and a surplus of labor. Also, most developing countries have a very limited supply of manag- erial and skilled labor necessary to operate complex plants. Planning for post-production operations must be done in a systematic way to realistically consider all available human, physical and economic resources. The total trans- fer of complex rice mill systems from developed countries has seldom proved satisfactory (Esmay, 1977). An effective method for evaluating development pro- grams is required which incorporate a comprehensive anal- ysis of all relevant factors and parameters, and which provide for iterative redesign of the development program as changes occur in its operating environment. System analysis including abstract modeling and computer simula- tion have a potential for meeting the requirements (Faid- ley and Esmay, 1974). Through such a system, bottlenecks can be more readily identified, knowledge and action gaps disclosed, interactions estimated, institutional and organ- izational defficiencies pinpointed and alternate pathways 18 considered (Cox, 1975). Koenig (1976) mentioned that the theoretical con- cepts in simulation, stability, control, optimization and other basic concepts of system science depends only upon the mathematical forms of the model and not the particular identification on meaning associated with the variables in the model. Manetsch (1976) stated that in developing simulation models there are three critical factors to consider: 1) the appropriateness of the problem for modeling; 2) the relevancy to the real world and whether the problem is worthy of investigation; and 3) sufficient familiarity with the subject matter of the problem to make the study feasible. Validation is required once the modeling is com- pleted to determine whether the model represents the real world satisfactorily. Optimization results in the speci- fication of the best combinations of system parameters and controllable inputs which satisfy the needs given the con- straints placed upon the system (Beveridge, 1970). The primary objectives of modeling an existing system is to provide a method of evaluating alternative management de- cisions and thus develOp an optimum strategy foroperat- ing or altering the system in a rapidly changing economic and physical environment (Bowers, 1970; Hare, 1967). 19 2.3.3 Selection Criteria The selection criteria used in machinery systems design is most often an economic one, i.e. least cost or maximum profit (Singh, 1978; Burrow and Siemens, 1974). To be considered are machinery, labor and timeliness. Estimating methods will vary somewhat. Different assump- tions will be made regarding equipment life, repair cost and trade-in values. Cost estimates will not all agree (Hinz, 1972). The choice of milling technique criteria employed by Timmer in Indonesia (1972) was: e Xk + 2° kakt MIN k t=1 (1 + 1) k kak - ngg kt kt x: - capital cost of the total investment in the k-th technique, assumed to be fully incurred in year zero W - wage paid in technique k total number of workers employed by the k-th tech- 5?" nique in year t - quantity of rough rice used by k-th technique in year t, assumed constant over time Xk - quantity of milled rice of type k produced by tech- nique k in year t, assumed constant over time - market price of type k rice - market price of rough rice 20 In order to more explicitly explore the implication of social discount rate on the choice of the technique in rice processing in Indonesia, Warr (1975) used the follow- ing decision criterion: MAX -E% k Xk N - net present value of the stream of agregate consump- tion generated by production using technique k X - quantity of rough rice milled by technique k Spencer 33 31. (1976) used a continuous linear pro- gramming model to determine the optimum technique, size and location of rice processing facilities in Sierra Leone, Africa, under various conditions. Processing, as well as assembly and distribution or transportation costs, are explicitly taken into consideration in examining the effects of alternative policy choices on employment, mill- ing output and income. Timeliness is a measure of ability to perform a job at a time that gives optimum quality and quantity of product. If the machine system does not have enough cap- acity to perform the job with desired results, the value of production loss is considered an economic penalty for poor timeliness (Singh, 1978; Tulu 33 31., 1974). Esmay (1977) proposed four steps of planning rice processing systems which are: mill site selection, equipment selec- tion, storage facilities, and marketing and distribution. 21 1. Mechanical Mill Production Capacity. The paddy harvesting pattern is of great importance in planning a procurement program. Milling, handling, storage and trans- portation must all be balanced properly to provide a con- stant milling and market supply. Storage requirements must be planned based upon the seasonal availability of paddy. The more distributed the harvest is throughout the year, the lower are the storage requirements and vice versa (Spencer 33 31., 1976; King 33 31., 1964). Excess capacity in the rice milling industry is the most marked feature. Because rice production is seasonal it is very difficult to determine the capacity of rice millers. In Indonesia, capacity utilization for medium and large mills was estimated on a basis of 200, 12 hour working days a year. In the mid 1950's, at this level, operation was about 50 percent of potential capacity (IPB, 1975). Excess capacity affects profitability and hence new investment. The medium and large mills sometimes compete for the farmers trade by favorable credit treatment and higher prices for paddy to keep the mills in operation and retain skilled workers; but where the government ad- ministers food prices, the opportunities for such compe- tition are limited. From time to time available paddy in each region is shared out among mills (IBRD, 1969; Colli- eer 33 31., 1974). 2. Distribution of Mill Technology. If a specific 22 set of machinery is used to perform a set number of opera- tions on a farm level of a given size, the annual use of each machine, as well as its fixed and variable costs, can be estimated. Several sizes of farm machinery sets could be simultaneously evaluated. As the size of the machinery set increases both productivity and initial price increase together with the annual machinery costs (Hughes, 1974). As a means of prediction, budgeting and control of costs, the concepts of fixed and variable costs can be helpful (Smith, 1973). Variable costs increase proportionately with the use of machines, while fixed costs are independent of use. The costs of interest on machinery investment, taxes, housing and insurance are dependent on calendar year time and is independent of use. The cost of fuel, lubrication, daily service and maintenance are associated with use. Depreciation and cost of repairing seem to be a function of both use and time. But, most often depreciation is included in the fixed cost catagory, and repair cost in the variable cost catagory (Hinz, 1972; Burrows, 1974) There are many levels of sophistication in calculat- ing approximate costs. For predicting future costs of either new and used farm machinery which are intended to be used normally, the use of annual fixed cost percentage method is adequate. Total cost per year of farm machinery could be computed in several interrelated equations (Bowers, 1970): where, ACOST FCP DEP XINT ACOST FCP BUY WORK REPAIR- LABOR OIL FUEL DEP XINT SALV XLIFE IRATE TIS 23 FCP x BUY + (REPAIR + LABOR + FUEL + OIL) x WORK DEP + XINT + TIS (BUY - SALV) / XLIFE ((BUY + SALV) / 2.) X IRATE annual cost for operating the machines amount of annual fixed cost percentage, in decimal initial purchase of the machine annual use of machine, in unit time repair and maintenance costs, in decimal of purchasing price per unit time labor wage lubrication cost fuel and electricity cost depreciation cost, straight line method interest of investment salvage value, usually 0.1 x BUY lifetime of the machine interest rate combined cost of taxes, insurance and shelter, in’decimal CHAPTER III MILLING LOSS MODEL A grain loss for a specific stage of the post har- vest process is defined as a percent by weight of the in- put and is expressed as follows: E(OUT) - AOUT (LOSS)n = ( INPUT )n where, E(OUT) = expected weight of output or maximum con- version AOUT = actual weight of output n = stage of post harvest process The milling of rice operations has an input of rough rice and an output of edible polished rice. The mill loss simplified relationship can be represented as: E(MILOSS) = E(CONVER) - YREND where, E(CONVER) = expected value of mill conversion, which relates to the condition of input mater- ial YREND = actual mill yield, with regard to mill technology and milled rice quality Direct measurement of mill loss is impossible as one 24 25 sample of rice cannot be milled two different ways. The system modeling then is used as the estimation method. Based on the mechanics of mill and better grain handling and separator devices, the more modern the rice milling tech- nology the less mill loss expected. Pre-milling opera- tions such as grain drying and grain cleaning practices affect the condition of rough rice entering the mill facility and thus the mill loss. 3.1 System Constraints and Desirable Model Characteristics l. The model should be able to represent a wide range of regional conditions with respect to geographic boundaries and the climatic factors of relative humidity and dry-bulb temperature. 2. The model should be able to take into consider- ation the probabilitistic nature of the weather factors that effect the mill conversion, and the distribution of input materials to the mill facilities. 3. The model should be able to handle any practical mix of mill practices and equipment. 4. The model should include the effects of pre-mill operation, which are the drying and cleaning practices. 5. The regional mill loss should consider the cli- matic condition differences between lowland (up to 10 meters from sea level) and upland regions. 26 3.2 Input Data 3.2.1 Weather Factors The monthly averages for relative humidity and temp- erature were computed from several weather stations in each province from time series data between 1967 and 1977, and were compiled by the National Meterological Institute of Indonesia. Monthly variation in each province is ex- pressed in polynomial regression equations. Table 3.1 shows that the climatic factors differ from region to region, therefore the climatic data will be based on regional input. The system modeling assumed that the probabilistic behavior in weather simulation will be on a normal distribution pattern. Table 3.1: Regional Climatic Factors, Yearly Average Low Region High Region Province Temp. RH Temp RH (9C) (9.) (°C) (%> North Sumatra 26.85 84.15 24.79 84.35 Mid-Sumatra 26.03 85.19 25.21 83.21 South Sumatra 26.73 84.34 25.86 82.92 West Java 26.40 78.74 23.88 81.19 Middle Java 26.28 79.78 24.63 84.42 East Java/Bali 27.41 76.92 24.61 79.81 South Sulawesi 26.04 80.09 23.38 80.89 North Sulawesi 26.16 83.76 25.93 83.95 South Kalimantan 27.13 76.00 27.08 84.57 27 3.2.2 Moisture Content of Rough Rice The moisture content of rough rice, when exposed to a given environment, will reach an equilibrium value with respect to the ambient temperature and humidity. The Hend- erson equation (1952) fixes the equilibrium moisture con- tent relationship with environment as: EQDB = EXP(ALOG(ALOG(1. - RH)/(-c x TEMP))/n) where, RH = relative humidity, in decimal TEMP = absolute temperature, in oK EQDB = grain moisture content, in percent dry basis c and n = parameters, grain characteristics, dimen- sionless Rough rice has c and n values of 8.82 x 10"6 and 2.22 respectively (IPB, 1976). Figure 3.1 illustrates the equilibrium moisture content curve for rough rice at 25°C ambient temperature. The model was verified with labor- atory experimental data and shows that the curve has a good fit between 45 and 95 percent relative humidity. 3.2.3 Mill Conversion Curve Mill conversion was evaluated as percentage of head rice or whole kernel and broken rice to total input. Broken rice was defined as less than three quarters whole kernels. Fine broken rice which is less than a quarter whole kernels and known as brewer rice or "menir," was EQUILIBRIUM MOISTURE CONTENT, WET BASIS (PERCENTS) 28 2110 15.0 . o 1&0 ‘4/" I // / / / ’7 / / / / 5.0 ,‘ / / ll zoos - EXP (ALOG (1.— emu-c: TEMP))/n) I EMCH =- 5008/ (1oo+eooe) c =88 . 2x 10-5 n- 2.22 TEMP =25° c M 1 l 100 0 20 40 60 80 RELATIVE HUMIDITY I PERCENTS) Figure 3.1: Equilibrium Moisture Content Curve Ambi- ent Temperature = 25°C. 29 considered as a by-product of the mill operation. The mathematical relationship between rough rice moisture con- tent and mill conversion was developed from laboratory ex- periment data. These equations were generated by non- linear numerical methods. CRHEAD = -0.288790 + 18.715072 x RRMC - 123.970912 x RRMC2 + 257.996278 x RRMC3 CRBROK = 1.297751 - 26.421624 x RRMC + 181.129922 x RRMC2 - 392.976923 x RRMC3 where, CRHEAD = percentage of head rice, in percent of rough rice CRBROK = percentage of broken rice, in percent of rough rice RRMC = rough rice moisture content, in percent wet basis Figure 3.2 shows that these equations were valid be- tween the moisture content range of 9 to 19 percent. The rough rice moisture content normally falls within this range for milling. The simulation stated CRHEAD has a standard deviation of 0.5 percent and CRBROK has 0.6 per— cent . 3.2.4 Milling Yield Five levels of milling technology are considered for the evaluation of milling yield performance. They are: 1) The traditional method of Hand Pounding (HP). 2) The Engelberg Steel Huller (ETH). The hulls are HEAD RICE, CRHEAD I IN PERCENTS OF ROUGH RICE) 30 35 \ / ‘ \ CRHEAD- -0.288790+ 18.715072 'RRMC \ -123.970912 ’RRMC"2 \ / + 257.996278'RRMC"3 CRBROK- 1.297751-2&421624'RRMC +181.129922’RRMC°‘2 / \ -32.976423‘ RRMC"3 d CI A A 10 p/ 10 15 MOISTURE CONTENT, WET BASIS (IN PERCENT) Figure 3.2: Milling Yield Curve :1: I33“! H9008 $0 SINSOHBJ NI I )IOUCHO 'BOIU NBNOUC 31 removed and some polishing takes place by the shearing action between the kernels and the movement of the steel roller. Axial flow move- ment of rough rice is accomplished with a trun- cated screw conveyor. A mix of milled rice, bran and hulls is delivered from the huller output. I 3) The Rubber Roll Huller (RRH). The hulling ele- ment consists of two closely spaced rubber rollers which rotate in the same direction and at different speeds. The machine usually is combined with a steel polisher. 4) The Rice Mill Unit (RMU). The unit consists gen- erally of a rubber roll husker, a steel polisher, grain separators, a husk separator and a brewer rice grader. The rice mill unit usually has grain drying and storage facilities. 5) The Rice Mill Plant (RMP). The rice mill plants generally include mechanical drying, mechanical handling equipment, and storage and modern rice milling machinery. Modern mill machinery con- sists of efficient grain cleaners, rubber roll hullers, various separators, cone type abrasive polishers, bran separators and grinding equip- ment. Schematic illustrations of various types of mechan- ical mill machinery are shown in Figure 3.3 to Figure 3.6. 32 Amhma ..Hm no moEmmv .A30Hmmv umnmfiaom o one: Am>ondv umaaom mohatmumoawmcm no mo coHuoHumoHHH oeumEooom "m.m ouomfim e. :ozusm . .ullli @/ 82:0 .552 Ba 6 :5 seem Seem 85:8 m 2.5 ses..\ .///,.\ hi, Seem :55 85:8 max/w %% 82m 123: , :mzm 8:: m>_m> 923381 83 \\. / ago: Ben. 33 Amped ..Hm um >m5mme .Hoaao: uoaaom Honoom mo cofluouumoHHH oeumfiwnom “¢.m muomwm L250 :3: w SE Evan: fl 520m 3m”— mct m 5:5: :8. .853..ch Elissa—E 8:285 :8. 34 .Amema .HmmHe Deco Haw: mowm Haw: Hmaao: accumuaaom Honoom mo unchoq owuoEmnom "m.m muomwm 10.53453 NOE Kuhn—E NOE 931:2 4...: mun—15:43.5 §84... «EL-m mus. 8.3.: «9.5154 x9! 5 — _ 3 8t 22; o W e2 89.65 :95 ace 8: with“ @ SB 388 o out {.395 o 53 8: o 838 . out 223 0 ® 2.3: o 53 8.80 o 29.2.. :25 o .68.. 3.25 o 8.. .520 o 36 The effect of mill technology level on the milling yields for both wet and dry seasOns is given in Table 3.2. These data were generated from several experimental studies by Bogor Agricultural University and the Central Bureau of Statistics from 1974 to 1977 (IPB, 1975; BPS, 1977). The average value was computed as (Hundsberger, 1973): k .2 ni x Xi SE: 1:]. where, n = numbers of sample on experiment i >4 II average value on experiment i index of experiment number P. II The standard deviation joint estimate is: k 2 Z (n. - l.) x Si S2 _ 1 1 1 N-k where, k N - 1:1 ni S = standard deviation on experiment i 7“ II numbers of experiment being considered Table 3.2 shows that the milling yield is less in the wet season. This may be caused by a higher moisture content of rough rice due to humid weather. Rice losses during processing would have to be related to the method of drying and moisture levels and the precise type of 37 Table 3.2: Milling Yield Performance 1J1 Percent of Rough Rice Wet Season Dry Season Average Technology 2' so i so if so HP 60.51 7.86 64.37 8.72 61.86 8.16 ETH 65.67 5.89 65.33 5.30 65.02 3.52 RRH 65.23 7.20 66.09 4.87 65.45 6.69 RMU 62.13 6.22 63.80 6.91 62.61 6.59 RMP 62.10 5.21 61.39 3.06 61.90 4.66 equipment used (FAO, 1977). The distribution of rice mill technology in each province is presented in Table 3.3 based on the data col- lected fromthe Post Harvest Survey done by BPS (1977-1978). The table values represent the percentages of rough rice milled by each technology within the region. These values were used as distribution factors in the model to compute the regional milling yield. Seedless grains (YNOSED) and foreign materials con- tent (YDIRT) were also accounted for in the input mater- ials. The survey data (see Appendix A) determined that the average value of seedless grains content for Sumatra was 5.1 percent, for Java 5.5 percent, and for Kalimantan and Sulawesi 6.4 percent. The average value of foreign material content for Sumatra was 0.5 percent, for Java 0.6 percent and for Kalimantan and Sulawesi 0.7 percent. Table 3.3: (in percent) Distribution of Mill Technology Province HP ETH RRH RMU RMP North Sumatra 1.7 65.0 3.3 0.5 29.5 Middle Sumatra 9.8 63.0 12.5 1.6 13.1 South Sumatra 13.8 70.1 7.7 2.1 v6.3 West Java 1.5 23.8 51.0 10.0 13.7 Middle Java 2.9 12.0 63.1 13.3 8.7 East Java/Bali 3.1 27.3 41.6 12.5 15.5 South Sulawesi 0.7 43.8 50.0 0.6 4.9 North Sulawesi 0.0 82.8 17.9 0.0 0.0 South Kalimantan 0.0 79.9 10.6 1.3 8.2 3.3 System Design The stochastic of weather inputs and milling per- formance were captured in the system model from the ran- dom variable generator with Normal and Gamma distribution. The averaging process of simulation results was done with the Monte Carlo technique (Manetsch, 1977). A black box concept of the mill loss model as designed is illustrated in Figure 3.7. Computer programming consists of the main program (MILOSS) and five subroutines (CONVERT, RENDEMN, MOISTC, MILCON and GAMMA). The general flow diagram is shown in The system was operated under the assumption Figure 3.8. that normal mill operation existed, hence normal 39 Eco: $3 :2 £6. 83E ZOHBZOU AAHE AAHZ A¢ZOHUmm .BDmBDO WOHm moz¢2mommmm QAHE BDQZH 9mm>0 seHono: mwmwwmmmummmr, oneHozoo meczHoo. zHamo qaonomm mBDmZH maozmwoxm 40 STAR MILOSS Read simulation status Number of regions (NREG) Number of Mill Technology (NTECH) Number of simulation run (NSIMUL) ‘MILOSS Read exogenous inputs Monthly temp., region Monthly humidity, region % seedless grain, % dirt MILOSS Determine policy inputs MOISTC compute T_____§ Distribution of mill tech. moisture Drying before mill _3_ content Cleaning before_mill ID = ; 16 = 1 2L CONVERT compute mill conversion compute :‘ mill yield for each GAMMA Egggégé mill technology 1 variable MILLOSS compute mill loss etermined % head rice % broken rice ~yes r—__———_-_I no .Lz”7“~ Change on policy 6—7 atle ' ‘ inputs T @9 Figure 3.8: Simplified General Flow Diagram, Mill Loss Model 41 distribution patterns could be introduced for the mill performance. 3.4 Model Implementation 3.4.1 Rough Rice Moisture Content The model simulated the climatic factors in each region and then computed the monthly average for rough rice moisture content. Model verification was done with actual laboratory moisture content measurements of samples taken from each province. Table 3.4 shows that the sim- ulation data closely represents the actual condition for modeling purposes. As expected, humid regions do produce higher moisture content rough rice. Table 3.4: Rough Rice Moisture Content, Yearly Average in Percent Wet Basis Simulation Samples Province Ave. SD Ave. SD n* North Sumatra 15.36 0.12 15.56 1.24 122 Middle Sumatra 15.12 0.01 14.74 1.64 125 South Sumatra 15.17 0.07 13.41 0.88 143 West Java 14.48 0.04 14.47 0.49 137 Middle Java 14.35 0.04 14.40 1.33 118 East Java/Bali 14.26 0.04 14.25 0.89 281 South Sulawesi 14.40 0.04 14.54 0.74 105 North Sulawesi 15.24 0.03 15.99 0.97 80 South Kalimantan 15.51 0.17 14.87 1.19 136 *Number of samples collected. 42 3.4.2 Mill Loss Levels The mill loss was estimated from 50 simulation runs. Table 3.5 shows the results by region. With the present distribution of mill technology, the rice production prov- inces in Indonesia have a yearly average mill loss esti- mated at about 4.8 percent of the rough rice with a stand- ard deviation of 0.23 percent; or about 3.1 percent of milled rice with a standard deviation 0.16 percent. Table 3.5: Estimated Mill Loss % of Rough Rice %of Milled Rice Province Low Land High Land Low Land High Land North Sumatra 4.47 4.56 2.82 2.87 Middle Sumatra 4.47 4.49 2.81 2.82 South Sumatra 4.94 4.99 3.07 3.10 West Java 4.59 4.72 2.95 3.06 Middle Java 4.52 4.72 2.98 3.08 East Java/Bali 4.77 4.81 3.04 3.09 South Sulawesi 4.88 4.96 3.12 3.22 North Sulawesi 5.07 5.13 3.26 3.30 South Kalimantan 5.02 5.07 3.20 3.23 The arid regions, such as Middle Java, tend to have less mill loss. mill loss than low lands, however this difference is not High land regions had an 8 percent higher 43 significant. The model simulated the regional mill conversion .(CONVER). For all the regions the average head rice per- centage was found to be 60.85 percent of rough rice with a standard deviation of 0.12 percent. The average broken rice percentage was 7.25 percent for rough rice with a standard deviation of 0.18 percent. The highest mill con- version value was found in Middle Java and the lowest in North Sumatra and North Sulawesi with both having more humid weather than Middle Java. The annual average of actual milling yield (YREND) per region was also generated from the simulation model. For all the regions, the average mill yield was found to be 64.69 percent for rough rice with a standard deviation of 1.07 percent. The highest mill yield was found in Middle Java (65.26 percent), while the lowest one was in South Sumatra (62.13 percent). These differences were mainly due to the distribution of mill technology. The average mill yield in Indonesia was found to be lower than other South East Asian Countries which have re- ported ranges from 64 to 68 percent (IRRI, 1977; IPB, 1975). Hand pounding is practiced relatively little even though most mill facilities are in poor condition and have aging machines that lack spare parts (Post Harvest Survey, BPS, 1977). The Rice Mill Plant had low mill loss but also pro- duced the lowest mill yield. Mbst of the Rice Mill Plants 44 in Indonesia are either old or operated inefficiently, they therefore have reduced mill performance (IPB, 1975). The milling output of the Rice Mill Plant is usually more polished than others which means more bran layers are removed (Post Harvest Survey, BPS, 1977). These conditions could produce less weight of the milled rice output per kilogram of rough rice inputs. 3.4.3 Milling Technology The interrelationships of mill technology were de- veloped from the model with runs based on several input policies concerning mill technology distribution. Sys- tem performance indicators were limited to mill losses and mill yields. The results are shown in Table 3.6. Table 3.6: Milling Practices Performances, in Percent of Rough Rice Specification Mill Loss Mill Yield Existing Condition 4.79 64.95 All HP 6.93 62.66 Mechanized, 4.74 64.15 Equal Distribution All ETH 4.87 64.03 All RRH 2.87 68.53 All RMU 4.25 66.50 All RMP 3.49 61.17 45 The traditional hand pounding generated the high- est mill loss of about 7 percent of rough rice with a relatively low mill yield of about 62.7 percent of rough rice. Improvement milling equipment with additional pre- mill drying and grain winnower facilities could reduce the mill loss by 2.2 percent and improve the mill yield by 2 percent. The Rubber Roll Huller had the best performance of the mechanized mill systems while the Engelberg Steel Huller had the highest mill loss. The Post Harvest Sur- vey (BPS, 1977) indicated that most of the Engelberg Hullers have a lack of pre-mill drying and grain cleaning facility. 3.4.4 Pre-Mill Operation. Pre-Mill practices include all of the pre-milling grain treatments carried out at the milling facility. The major pre-milling operations are grain drying and cleaning. The pre-milling effects on mill losses are presented in Table 3.7. The simulation results showed that the improvement of pre-mill facilities, specifically grain drying and cleaning practices could reduce mill loss by providing a better grain condition for the milling process. The model predicted that an increase of about 15 percent in pre-mill facilities reduced mill loss by ap- proximately one percent. The Post Harvest Survey (BPS, 46 Table 3.7: Simulated Pre-Mill Operation Mill Loss, in Per- Drying Practices Cleaning Practices cent Of Rough Rice in Percent in Percent Ave. SD 50 50 7.62 1.30 50 100 3.65 1.35 100 50 4.79 0.23 100 100 2.82 0.21 1977) indicated that the pre-mill operation is mostly inadequate in rural areas. The survey estimated that about 70 percent of rough rice was dried within the mill site, and less than 50 percent was cleaned before mill- ing. Most of the rough rice drying is done by sun- drying and grain cleaning is still done manually. CHAPTER IV REGIONAL RICE PRODUCTION MODEL The development of rice post harvest technology is dependent upon the regional rice production system. Rice mill technology improvement plans should be based on the amount and time of rough rice available in the surround- ing area of mill sites. Good planning involves a future trend study of the dynamic behavior of the existing systems. A regional rice production dynamic model has been de- veloped to provide a quantitative analysis of the post har- vest Operation alternatives. Various policy alternatives may be generated for the purpose of future production plan- ning. The model may also be used for analyzing various regional food policies and for food price forecasting. The simplified dynamic model for regional rice pro- duction is: YPROD(t) = (((AVHST(t) x UBIN(t)) x (1. - FLOSS) X (CTHRES - TLOSS)) x (MILCON - MILOSS)) - FSTORL - TRLOSS where, YPROD = regional rice production, in milled rice, tons/unit time 47 48 AVHST = harvested area of paddy rice, hectares/unit time UBIN = yield per hectare, in tons of dry stalk paddy/hectare FLOSS = in-field losses, in percent of rough rice CTHRES = threshing conversion from dry stalk paddy to rough rice, in percent TLOSS = threshing loss, in percent of rough rice MILCON = milling conversion from rough rice to pol- ished rice, in percent MILOSS = milling loss, in percent rough rice FSTORL = storage loss, in percent of milled rice TRLOSS = transportation loss, in percent of milled rice 4.1 System Constraints and Desirable Model Characteristics 1) 2) 3) 4) The model should capture the behavior mode of time dependent variables involved in the system based upon the derivation of available time series data. The model should be able to represent such pro- duction constraints as land, especially that under cultivation and various post harvest practices. Five year forecasts should be possible as re- quired for regional development planning. The model concentrates mainly on post production parameters up to the time the milled rice reaches the market. 49 4.2 Input Data 4.2.1 Harvested Area The harvested area data should include the rice cul- tivation practices of irrigation. The harvested area sea- sonal variation should simulate not only the area planted but also the effects of plant diseases (KASS, 1972). Java and off-Java region boundaries were considered as they have specific rice production trends. In Java, land constraint is a critical factor; whereas, off-Java there is still ample unused land. Cultural practices were divided into wet and dry rice production. The time series data for each district in both re- gions were compiled monthly. The data were obtained from the Central Bureau of Statistics (CBS) in Indonesia. The model requires at least ten years of input data for reliable sim- ulation modeling. 4.2.2 Yield per Hectare The yield per hectare varies with the rice production cultural practices, rice variety and soil fertility. The increase of rice productivity in Java has been different from the off-Java region because of the various cultiva- tion practices by farmers (IPB, 1974; Glassburner, 1978). Besides the regiOnal classification, the input data had also been classified as dry and wet land rice production and seasonal variation. Yields have been expressed in tons of stalk paddy per hectare. These data were obtained from the Central Bureau of Statistics in Indonesia. 50 4.2.3 Post Harvest Technology Physical conversion data for rice during the various post production operations have been taken from tropical study sources (IPB, 1974; Badan Urusan Logistik, 1976; Tuquero EE.E£-r 1977; Ilangantileke, 1978; Wanders, 1978), except for the milling operation data obtained as a part of this research. Most of the parameters are assumed to have a normal frequency distribution which is significant at 0.05 probability level. Threshing conversion from stalk paddy to rough rice was found to be 88.7 percent with standard deviation of 0.3 percent (IPB and CBS, 1975). The manual threshing method had been estimated to have a loss of 2.3 percent of rough rice with a 0.9 standard deviation. The pedal thresher has 1.2 percent with a 0.5 standard deviation and the mechanical thresher has 0.5 percent with 0.4 standard de- viation. Grain damage from the threshing operation was estimated at 0.4 percent with a 0.2 standard deviation. The damaged kernels were not recovered in the milled rice output, therefore, they were considered as by-products or waste (IPB, 1975; Wanders, 1978; Ilangantileke, 1978). In-field losses for harvest loss and in-field trans- portation loss were estimated to be 2.5 percent of the rough rice with 0.9 standard deviation (Ilangantileke, 1978). It was found that the present mill conversion rate for Java was 64.8 percent with a 0.5 standard deviation, 51 and for off-Java 63.5 percent with a 1.0 standard devia- tion. Average mill loss was estimated for Java at 4.6 percent of rough rice with a 0.5 standard deviation and for off-Java at 4.8 percent of rough rice with 0.3 stand- ard deviation (see data presented in Chapter III). Rural transportation for paddy by bullock carriage and small boat had an average loss of 0.7 percent of milled rice; whereas the mechanical transportation had an 0.3 percent of milled rice (IPB, 1974; Badan Urusan Logistik, 1976). There are few studies on grain storage losses in Indonesia. The Post Harvest Survey by CBS (1978) indica- ted that farmers used to store the rough rice grain as long as 4 months (see Appendix A). Jindal (1978) pre- dicted 0.5 percent dry matter decomposition of paddy on storage occurred within 20 days with rough rice moisture content at 17 percent wet basis. 4.3 System Design 4.3.1 Growth Model Approach The available time series data for the harvested area and the yield per hectare had a relatively low auto- correlation coefficient of about 0.6 to 0.7. Consequently the dynamic linear regression approach with stochastic parameters (Hillier, 1967; Hundsberger, 1973) does not provide for a reliable mathematical model construction. The simulation model used the growth trend approach 52 (Goodman, 1974) to capture the dynamic behavior of the rice production system. Both harvested area and yield per hectare information were used. Basically, the growth trend as a time dependent variable is described as: GROWTH = VARt + 1 - VARt t + l VARt where, GROWTH = growth variables VAR, in percent VAR = variables under study, in unit VAR t = index time Various mathematical expressions were tested during construction of the model to obtain the best fit for the available growth data. The growth might be negative and thus would equal zero. Given these inputs, the simulation model was run through time from 1978 to 1988. 4.3.2 Curve Fitting Procedure The growth model consisted of non—linear equations as time dependent functions. Model verification was based on a curve fitting procedure using the Gauss Newton method (Marquardt, 1963). The general model is: Y = F(Xl, x2, . . . xk; Bl, B2, . . . , Bm) The computer programs were develoPed to solve for the coefficients (B) utilizing N data points for Y and X, i = 1, 2, . . . N. The procedure was generated from lin- earization of the proposed model with the utilization of a least squares objective function. 53 Starting estimates of the unknown coefficients were required in the model. The model was linearized by ex- panding Y in a Taylor series about current trial values for the coefficients and retaining the linear term only, Ai = A * -3Yo-* A 3Y. *A BY. * A Y Yi + E l] ABi + [ lJABZ + o o o + [ l] ABm BBi 332 33M where, AB [§ §*] ' 1 2 M a" j j ' 3 ‘ ' ' ° ° ' The asterisk designates quantities evaluated at the initial trial values. A least square objective function is formulated as, (Y. - I.) minimize LS = l 1 "NZ i The linearized model was substituted into the objec- tive function and the "normal equation" formed by setting the partial derrivatives of the objective function with respect to each coefficient equal to zero; Egg = 0 ; j = l, 2, . . . . M BB. 3 The resulting normal equation was in the form of (§?§) AB = BT(Y - §*) where, 54 r P :31 BY1 . . . BY17 3B1 afiz 8133M aY2 BY2 . . . . . . BY2 T —A_- A 3131 3132 313M .3. = ° SYN 3YN......3YN -w— -7— -7— _331 3132 313M; (Bl ' 31*) (Y1 ‘ Y1*) (BZ " 32*) A (Y2 - Y2*) A}; = (Y ' Y” = . A _ _ * L (BM BM") (YN YN) B? is the transpose of the B matrix. The derriva- tives in the B matrix may be evaluated analytically or numerically. The normal equations are a system of linear algebraic equations and may be solved by an appropriate technique for B. The B vector and LS approach zero as convergence is achieved, and the final coefficients were calculated from, B. = B.* + B. ; ' = 1, 2 . . . M J J J 3 ' If convergence is not achieved, B may be updated by SS replacing the old values by the new values and the process repeated. 4.3.3 Model Construction The simplified flow diagram of the computer program- ming is illustrated in Figure 4.1. This model has two main programs. Program GROWTH was designed to construct the mathematical model for simulation of harvested areas and yield per hectare growth trends. This program used non-linear numerical analysis of the Marquardt technique (Kuester, 1973; Pennington, 1965), as an extension of the Gauss Newton method. Program FRECAST was developed to compute the re- gional rice production and estimate the post harvest loss that occurred in each post production stage. There were two additional sub-programs used. Program MILOSS provided the mill conversion and mill loss (refer to Chapter III). NORM provided the stochastic variables with normal fre- quency distribution. A ten year forecast was the output of the FRECAST program. 4.4 Model Implementation 4.4.1 Growth Model 1. Harvested Area. Several trials of the mathe- matical models were carried out. The most favorable moded for the yearly growth trend of the harvested area in Indo- nesia was found to be the Log Log Inverse function: GROWTH = EXP(B(1) - (B(2)/TIME) - B(3) * ALOG(TIME)) 56 ® READ = Number of Regions Length of Simulation Regional Time Mathematical GROWTH: Generates Dynamic Series Data Models Models,. _ e_ - Harvested Curve Fitting Tes Area - Land Productivity V a emazical Expression of Growth Model T -Mn Satisfy? MILOS: Provides FRECAST= igiigzstmg C°mP“‘ - Mill Conver- -Harvested Area é__ _ 31?? Loss -Land Productivity é_______ No V Post Harvest FRECAST: -Compute Regional Inputs iii: Rice Produc—p__ _ Conversion l - Loss T NORM: Provides Random Variables Figure 4.1. Simplified Flow Diagram of Forecasting Model for Regional Rice Production. 57 where TIME = 1, 2, 3, . . . k. Parameters §_for both Java and off-Java regions and for the dry and wet fields paddy are given in Table 4.1. Figure 4.2 and Figure 4.2 illustrate the yearly growth trends of harvested areas in the two regions for the period from 1964 to 1985. Table 4.1: Parameters for Harvested Area Growth Functions Specification B(1) 8(2) 3(3) Java - wet land -0.014704 -0.095774 -0.005226 — dry land 0.143548 0.091592 0.080025 Off-Java - wet land 0.018865 -0.049476 0.000091 dry land -0.035668 -0.105399 -0.002887 Figure 4.2 shows that after 1976, the wet land har- vested rice area in Java increased linearly but at a rel- atively low rate of 0.8 percent per year. Land constraints in the densely populated land of Java evidently affects this growth. The dry land harvested area in Java had a negative growth since 1968. Some of the dry land paddy field decrease may be accounted for by a shift through the improvement of irrigation systems. Also some rice land has been utilized for industrial sites and possibly for commercial crops such as sugar cane and tobacco. Ur- banization may also contribute to the decrease of rice land as more and more rural people move to the cities 58 10 .— uza \ c: 5 m . nu m E \ u; \ IAN g o I <9 2" 5 D ._ 2 E RY LAND -10 l l l 64 70 75 80 85 YEAR Figure 4.2: Area Change Trends for Harvested Rice in Java 59 10 5 \ h. \. :2 \~ nu ‘~“ g ‘~—.~___ WET LAND m h———-———— -- a. E u. 0 <9 <2: DRY LAND I c: .< 36 < -5 .4 .< D Z Z .< -12 l l J 1 64 70 75 80 85 YEAR Figure 4.3: Area Change Trends for Harvested Rice in Off-Java. 60 (Glasburner, 1978). Figure 4.3 shows that the change in off-Java region wet land paddy field has been almost constantly increasing at 2 percent per year. Good rural irrigation and trans- migration projects will be needed to maintain even the 2 percent annual growth. The decrease of dry land rice is greater than in Java. This decrease may be caused by shifts of crops and rice cultivation methods. Since un- cultivated land is quite available in off-Java regions there should be some positive growth trends if the exten- sification programs are improved. Presently, any increase trends tend to be off-set by declining numbers of farmers in off-Java due to migration to Java (IPB, 1976). 2. Yield Per Hectare. Several exponential functions represent the changes in yield per hectare: GUBJS = EXP (0.115724 - (0.13885/(TIME)3) ' (0.034221) x ALOG(TIME) GUBJG = EXP (0.024405 - ((0.02496/(TIME)2) GUBOJS = EXP (0.021602 - ((-0.046051)/(TIME)5) GUBOJG = EXP (0.008036 - ((-0.048967)/(TIME)2) where, GUBJS = growth of yield per hectare in Java, wet land GUBJG = growth of yield per hectare in Java, dry land GUBOJG = growth of yield per hectare in off-Java, wet land 61 1010 WET LAND, JAVA \ , \ Emma, OFF JAVA \\\ ~ \-_D_R_Y_LA_ND ANNUAL CHANGE IN PERCENT 00 l L J l 64 70 75 80 » 85 YEAR Figure 4.4: Rice Productivity Trends, Yield Per Hectare. 62 GUBOJG = growth of yield per hectare in off-Java, dry land Figure 4.4 illustrates rice production trends per hectare. A11 classifications show a positive increase in yield. The introduction of the High Yielding Varieties of rice and better cultivation practices have no doubt con- tributed to these favorable trends. Wet land rice in Java shows a decreasing rate of yield growth while other class- ifications show a constant rate of annual increase. The dry land rice yield growth rate was about 1 percent per year while wet land in off-Java regions was about 2 per- cent per year. 4.2.2 Forecasting Model 1. Harvested Area. Simulation results are presented in Figure 4.5 to Figure 4.8. The simulation data curves were compared with yearly average data from the CBS (BPS, 1978). The simulated forecasting graphs show a relatively good fit with the past data and are significant at 0.05 probability level. Java wet land paddy cultivation is increasing more slowly while off-Java regions show a higher growth rate. Dry land rice production areas reached maximum points in the period from 1965 to 1970, and has declined ever since. 2. Yield Per Hectare. Rice yield simulation curves are presented in Figure 4.9 and 4.10, together with yearly average data compiled by the CBS from 1963 to 1977 (BPS, 63 .m>mn cw ¢ on .moflm coma uoz mo mmufi ooumm>umd How coaumoHMHum> Hmooz msflummomuom um.v Gunman 7mm . 20:23:25 u II. :5: ago 5.3 u d 88 come (SiHVlOSH/SGNVSflOl-Ll.) NI 'VBHV OBLSSAHVH 64 oOAm ocmq who no mmufi Umumm>umm new GOAHMUAMAHm> Hobo: mcflummomuom ma ca ms m .m>mn ca ch "o.v musmflm 8p 2036.43.25 H II ska: mmo mnummo a“ .mon tang no: mo mmum cmumm>umz How :oflumo«MHHm> Hone: mcflummomuom “h.v musoflm :5; 20.5.5325 I I ‘ gnu: mac (5‘0 I ‘ 65 SBUVLOSH/SONVSOOHJ. NI 'VSHV OBBBAUVH 8D.” 66 .m>me:mmo :A .mon Gama who mo mmud nmumm>umz MOM COHDMOHMHum> Hobo: mcflummomuom "m.v musmflm m om mm on mm :8 H /ll 14, 8a W / a / N“ / 3 4. a / W I coop 3 4 v D H O \f m 8: w O 5 295.525 "I 4 \\ .H: 3 ES: moo <55 u 4 w. , 82 w“ m 4. .4 camp 67 .>....>....UDDOCA. moi 024.. km; :0... ZO.... Juno—z 02_Pm<.. Z. (#40 ‘ 32$ .93. <><.. “7.0 <...<0 4 m ms Oh mm M... b ON mN om mm cc 04 cm. 'CI'IBIA BOIH VH/ACIOVd )IOVlS EDI 00L .>....>....QDOO....A. mu... DZ<.. >¢Q con. ZO.... Jun—0.2 02.hm 8 an 2 8 8 _ p . A . on. 4 4 zofifizsal . 85. .2! <22. 2. <55 4 4 4 4 4 a: 8:: .93. 43:. to <53 4 4 4 4 4 4 4 4 4 to..." 4 4 4 4 -3: 4 4 4 4 \\\ re: \ WWO \\\ 474.. 5.2 4 \ \ a», 5.2 \ 5» b7 \ \ (3.4. \\ 'O'IBIA 30m VH/AOCIVd )IOVLS EDI OOI 69 1978). The simulation results have a relatively good fit with the past data and are significant at 0.05 probability level. The simulated yields per hectare all show an increas~ ing trend. However, on Java wet land rice a decreas- ing rate. In the year of 1985 the productivity of wet land rice in Java was predicted to reach 5 tons of dry stalk paddy per hectare while in off-Java region it may reach 4.5 tons. By that time, the dry field rice will reach just 1.8 tons per hectare. 3. Milled Rice Production. The model disaggregates the production of milled rice according to the regions and also according to the post harvest operations. Forecasted production is presented in Table 4.2 along with the esti- mated post harvest losses. The estimated rice consumption predicted by the CBS was also in Table 4.2 (BPS, 1978). A ten year period has been covered with the assumption that the post harvest technology was fixed. Improvement of post production practices should either increase the post harvest conversion or decrease the losses. Simulation predicted that the annual growth rate of rice production during the period of 1978 - 1988 would be averaging 2.6 percent. Since the rice consumption rate is about 3 percent, imported rice must then also grow pro- gresSively. The simulation model also predicted the post harvest 70 Table 4.2: Simulation of Rice Production in Indonesia, 1978 - 1988 in Thousand Tons of Milled Rice Post Harvest Year Rice Production Losses Rice Consumption* 1978 16,580 2,997 17,014 1979 17,210 3,468 17,731 1980 17,613 3,437 18,934 1981 17,696 3,373 19,268 1982 18,561 3,681 20,095 1983 18,774 3,199 20,959 1984 19,562 3,698 21,862 1985 19,848 3,759 -- 1986 20,684 4,104 -- 1987 20,564 3,704 -- 1988 21,490 4,028 -- *Source: BPS (1978) losses with technology conditions in 1978. It was esti- mated that the post harvest losses were about 15 to 20 per- cent or around 3,000,000 tons of milled rice per year. Almost half occurred in grain storage. The Post Harvest Survey (1978) indicated poor storage facilitiesin rural areas (see Appendix A). The simulation model predicted that a 40 percent reduction of current post harvest losses would reduce the present defficiency of rice. CHAPTER V REGIONAL RICE MILL MECHANIZATION MODEL Literature contains much data on costs of operating rice milling equipment, but few publications have con- sidered the problem of capacity and size selection (Spen- cer 33 31., 1976). This system model simulates a regional mechanization system for the rice milling operations based on the regional rice production constraints. The mechan- ization systems design procedure basically consists of a series of subsystem optimizations by an iterative search process. Each search is subject to the constraints of a specific time period upon the completion of each opera- tion (King and Logan, 1964; Box, 1965; Singh, 1978). The system model separates the selective mechaniza- tion into two programs: The first sub-model deals with the capacity of the mills based upon optimal solutions of the annual mill capacity. The regional mill capacity is expressed in tons of rough rice per year and related to the local rough rice production within that period. The second sub-model deals with mill size selection. Several alternative mill sizes are considered for the region. Size selection is based upon minimization of regional milling costs. 71 72 It was pointed out in Chapter I that many regions in Indonesia have an over-capacity of rice milling facilities. This over-capacity produces undesirable idle time for millers. The Post Harvest Survey (1978) carried out as a part of this study indicated that mills averaged 60 percent idle production time throughout the year. The millers' profits were cut drastically due to the unavailability of rough rice. The total time of operation is a major factor for the survival of millers. A reducing of idle time can therefore be realized through a reduction of excess capac- ity within the region. On the other hand, some parts of Indonesia did not have sufficient milling facilities and labor for hand pounding was also short. Portions of rough rice were stored before it could be processed. The Post Harvest Survey found that farmers in South Sumatra and South Kal- imantan stored their rough rice as long as four months. Grain storage losses are high in farm areas, and have been estimated at 2 - 5 percent of total production (FAO, 1977). The reduction of persistent under-capacity of mills should be related to the storage losses. The first part of the model was designed to opti- mize the regional mill capacity through the minimization of idle time and farm storage losses. The timeliness factor (Tulu 33 31., 1974) was used as an indicator of both idle time cost and storage loss cost. The monthly volume fluctuation of the rice harvest was considered as 73 the main factor affecting the timeliness factor. Produc- tion directly effects the distribution of input materials to each mill facility. The second part of the model determines the selection of suitable mill sizes within the region. The selection criterion for optimization was based on a financial anal- ysis of fixed and variable costs (Hinz, 1972; Hughes, 1974). The field study was carried out to implement the regional study. Data were collected on machine performance and farmer interviews were made in two subdistricts or Kecamatan in the West Java province. Twenty mill facili- ties were observed and 50 farmers were interviewed in each subdistrict. The field study was conducted during two harvesting peaks and supplemented with secondary data from various government and research institutions. The subdistrict of Ciawi had about 7,500 hectares of land and 110,000 people of which 40 percent were farmers. The subdistrict of Ciomas had about 8,500 hectares of land and a population of 172,000, of which 25 percent were farmers. Both districts appeared to be representative of rice production areas of the province. Ciomas had 8,296 hectares of wet land paddy and 596 hectares of dry land paddy, while Ciawi had 3,907 hectares of wet land only. High yielding varieties of rice and improved cul- tivation methods had been introduced in these areas. 74 5.1 System Constraints and Desirable Model Characteristics 1) The model was designed to render a quantitative analysis for a development plan of regional mill mechanization. It included the dynamics of rice production. 2) The range of sizes considered for each milling class was related to the range found in the re- gions under study and to the range which was sufficient to evaluate relevant financial factors. 3) The model was designed to capture an apprOpriate optimization technique for selecting mill tech- nology based on the assumption that no competi- tion for labor and machinery from other farm enterprises occurred. 5.2 Input Data 5.2.1 Harvesting Frequency The frequency distribution of harvesting time (PROHVT) consisted of the percentage of farmers harvesting their paddy within a specific month. Data were generated from two farmer interviews in one year in each subdistrict. Samples were taken at random. Figure 5.1 presents the monthly fluctuation of harvesting time for each district. There were two harvesting peaks; one in March and April and another in August. The August peak spread some into July and September and a secondary peak was evident in 75 CI CIOMAS DISTRICT CIAWI DISTRICT 50') FREQUENCY IN PERCENT 10" Aug Sep Oct ov Dec on I... not 0.0-4'4 4 Apr May 7 Jun Jul Jan Feb Mar Figure 5.1: Frequency Distribution of Farmers' Harvest- ing Time. 76 November. 5.2.2 Rice Mill Classification Mill technology was classified into four catagor- ies. The criteria of each mechanical mill class is pre- sented in Table 5.1. The allocation of rice production to each mill class was developed from field observation data. ‘The rice production allocation percentages might be viewed as input targets for further regional development alterna- tives. The variation in the allocation percentages (XPOL) provides a basis for reevaluating mill technology distri- bution. The criteria for mechanized mill classifications are: 1) power required per unit rice mill machinery (HSP) in horsepower, 2) potential mill capacity (CAP) in tons of rough rice per hour, 3) skilled operator required (OPERT) in man per unit of rice mill machinery, 4) time period re- quired for repair and maintenance (TIRP) in hours, and 5) expected machine life (XLIFE) in years. Table 5.1: Mechanized Rice Mill Criteria Classification HSP CAP OPERT TIRP XLIFE l. Engelberg Steel Huller 5 0.3 2 600 5 (EGH) 2. Rubber Roll Huller 5 0.5 2 50 6 (RRH) 3. Rice Mill Unit 15 1.0 4 100 8 (RMU) 4. Rice Mill Plant 25 2.5 6 400 10 (RMP) 77 The portion of the rough rice production milled by hand pounding (HP) was also determined from the field study. Both subdistricts averaged two percent. The rest of the rice production was milled mechanically. Since transportation cost was also considered, the portion of rough rice transported (PORT) either from the paddy field to the mill or from the mill facility to the market was specified in each mill class. Fuel consumption for stationary machines (CONSE) was estimated at 0.12 liters per HP per hour. Lubrication oil (CONSG) was estimated at 0.035 liters per HP per hour. Re- pair and maintenance cost (YREP) was estimated at 1.2 percent of machine price (BUY) over the time period re- quired (TIRP) times machine life (XLIFE). 5.2.3 Regional Prices The prices used in this study were fixed over the evaluation period. Price distortions will undoubtedly affect the model output and thus the selective mechaniza- tion (Warr, 1975; Collier 33 21., 1974), but these changes were not included in this model. Table 5.3 shows the average cost and item price for each mill class operation. The currency rate in July 1978 was 450 Rupiah per US Dollar. Diesel fuel price (SOLAR) was 25 Rupiah per liter. Lubrication oil (GREAP) averaged 350 Rupia per liter. The regional rough rice price (RRPRI) at the mill site was 78 77.50 Rupiah per kilogram in Ciomas and 76.30 Rupiah at Ciawi. The milling fee (XMILFE) was 4.50 Rupiah per kilo- gram of rough rice in Ciomas and 5.00 Rupiah in Ciawi. Table 5.2: Mill Technology Cost Input Mill Labor Cost (Wage) Machine Price Transport Cost . . _ (BUY) in Rupiah/kg Class in Rupiah/man hour in Rupiah Rough Rice EGH 400. 300,000 5. RRH 400. 825,000 5. RMU 500. 1,625,000 10. RNP 500. 3,000,000 10. 5.3 System Design 5.3.1 Model Structure 1. Regional Mechanized Mill Capacity. The deter- ministic model was designed to compute the regional mech- anized mill capacity (XP) through minimization of the ob- jective function which consisted of the timeliness cost of storage losses and idle time. A simplified mathematical expression of the model objective function was minimize Y = f (DISCOl, DISCOZ) subject to XP 3 0 DISCOl is timeliness cost due under capacity, DISCOl = DAMAGE x RRPRI where DAMAGE was storage loss in tons of rough rice at 2 79 percent per month. In the model, DAMAGE was accumulated yearly using the DELAY subroutine with the EULER integra- tion technique (Manetsch, 1976). DAMAGE was a function of the regional rice production per unit time (XRRY) and the mill capacity (XP). The average delay was one week. DISC02 was the penalty cost caused by over capacity which lead to a loss of benefit as DISC02 = (XP - XRRY) x CIDLE CIDLE = XMILFE x (EEC/(1. + EBC)) x CF where, XP designated level of regional mechanized mill capacity, in tons of rough rice per year XRRY = regional rough rice production, in tons per year CIDLE = penalty cost due to idle time, in Rupiah XMILFE = expected milling fee, in Rupiah per ton rough rice EBC = intended benefit cost ratio of mill business, dimensionless CF is correction factor and defined as XP - XRRY CF = 1 + XRRY If there was no mechanization (XP = 0), the model showed that CIDLE would equal zero. If the regional mill capacity equaled rice production the CIDLE would also equal zero which means there was no idle Operation. If the XP 80 was less than the XRRY, the DISC02 becomes negative but on the other hand increased the value of DISCOl. This tradeoff situation should be optimized. Weekly simulation was done to compute yearly regional mill capacity. Comparison with existing conditions was carried out, where the present mill capacity in the region (ACCAP) was computed as 4 ACCAP = ( Z A. x C.) x WORK . . 1 1 l=l where, C = average capacity of the mechanical mill, in tons of rough rice/hour A = amount of specific mill types in the region WORK = machine operating time per year, in hours 2. Mill Technology Selection. The deterministic model was designed to provide information on the feasibil- ity of each mill class for the region. Two factors were included in the system modeling. First was the output from the regional mill capacity program (XP) which acted as a physical constraint. Second was minimization of the regional mill costs based upon the operational cost of each mill class as compared to the distribution of rough rice regional production for each mill class. Basically the model consisted of a series of equaL tions for each mill technology fixed and variable costs. The objective function was: 81 minimize Y = ACCOSTi/XPMi subject to XPM i 0 and ACCOST Z 0 where, i = mill class index ACCOST = regional mill cost per year, in Rupiah XPM = portion of regional production distributed into mill class i, in tons of rough rice per year Several equations supported the regional mill cost function. Some notations have been described earlier in this chapter. The Xi is the designated amount of each mill class within the region. WORKHi = XPMi/(CAPi x Xi x EFF) where, WORKH = operation time per year, in hours EFF production efficiency, dimensionless ENERGYi = HSPi x WORKHi x CONSE x SOLAR where, ENERGY = cost of fuel per year, in Rupiah GREASEi = HSPi x WORKHi x CONSG x GREAP where, cost of lubrication per year, in Rupiah GREASE LABOR. OPERT. x WAGE. x WORKH. 1 1 1 1 where, LABOR = cost of labor per year, in Rupiah REPAIRi = YREP x BUYi x (WORKHi/TIRPi) where, 82 REPAIR = cost of repair and maintenance per year, in Rupiah RRANSi = TRCi x XPMi X PORTi/Xi where, TRANS = cost of rough rice transportation per year, in Rupiah VARCOi = ENERGYi + GREASEi + REPAIRi + LABORi + TRANSi where, VARCO = regional variable cost per year for mill class i, in Rupiah DISCi = 0.9/XLIFEi + 0.076 where, DISC = annual discount percentage, with Salvage value = 0.10 0.076 = coefficient which represented a straight line depreciation cost, taxes, insurance and shelter FCOSTi = DISCi x BUYi X Xi where, FCOST = regional fixed cost for specific mill tech- nology, in Rupiah per year ACCOSTi = VARCOi + FCOSTi where, ACCOST = regional mill cost for specific mill class, in Rupiah per year 83 The comparison method used in this analysis was sim- ilar to the Bayer decision procedure (Hillier, 1967), which selects an alternative such that minimize E(1(a,0)) = g l(a,k)We(k) k=l where, 1(a,0) = comparison level, in Rupiah a = mill class alternatives 0 = indicators W = weighing policy for indicator, in percent The indicator (0)in this evaluation was the milling cost, operation time, fuel consumption and labor require- ment. The comparison level was converted to Rupiah. 5.3.2 Variable Search Method This system modeling determined the minimum of a single variable non-linear function subject to constraints. The search technique used was the Fibonacci method which was an interval elimination procedure. The region in which the optimum lies was sequentially reduced by the search procedure (Rosenbock, 1960; Beveridge, 1970; Kuester, 1973). Starting with the original boundaries of the inde- pendent variable, the interval in which the Optimum value of the function occurred was reduced to some final value, the magnitude of which depended on the desired accuracy. In this model, the accuracy was set at 10 percent of the value of the independent variable. The location of points for function evaluations was based on the use of positive 84 integers known as the Fibonacci numbers (Fn). NO deriva- tives were required. -A specification of the desired accuracy determined the number of function evaluations. The variable search procedure started with a deter- mination of the original search interval as Ll with bound- aries al and b1' The required Fibonacci number was com- puted as: "2| II '11 "*1 Z ll H F = + F n 1 2 n-2 7 where d was the desired accuracy. The algorithm proceeds as follows: 1) Place the first two points, X1 and X2 (X1 < X2) within L1 at a distance dl from each boundary, d1 = g-Z x Ll N x1 = a1 + dl x2=bl-d1 2) Evaluate the objective function at X1 and X2. Designate the functions as F(Xl) and F(X2). Narrow the search interval as follows: a1 3 x* 1 X2 X1 _<_ x* 1 b1 1) 2) where x* was the location of the optimum. The new search for F(X < F(X2) l) for F(X > F(X interval was given by 85 with boundaries a2 and b2. 3) Place the third point in the new L subinterval, symmetric about the remaining point, F d2 = FN-3 x L2 N-l X3 = a2 + d2 or X3 = b2 - d2 4) Evaluate the Objective function f(X3), compared with the function for the point remaining in the interval and reduce the interval to 5) The process was continued with the preceeding rules for N evaluations. The general equations were: = FN-(k+1) d FN-(k-l) x L k k Xk-I-l = ak + dk' or xk+l = bk - dk (symmetric about mid-point). _ FN-(k-l) k FN L x L 1 = k-l k-l After N-l evaluations and discarding the appropriate interval at each step, the remaining point was precisely in the center fo the remaining interval. The Objective function was then evaluated and the final interval where the Optimum was located was determined. A simplified dia- gram illustrates the procedure in Figure 5.2. 86 Define Original Search Constraints and Required Accuracy + J. Place First Two Points 1 Evaluate Objective Fugction(s) _--§> and Narrow Search - Interval Place New Point in Search Interval Symmetric About Midpoint with Respect to Point Remaining in Interval Both ‘ Points +‘\‘at Midpoin NO Side of Midpoint and Move t distance to One I Evaluate_Function Calculate Final Interval Figure 5.2: Fibonacci Logic Diagram 87 TAR / READ: Regional Input Rice Production , —‘ Compute XRRY _Harvest Monthly Production Pattern 5P Set = ALPHA = DEGREE of Accuracy _g , QL [Initial Gue§§:_zgf | XP "'>' XRR'YJ l XP < XR' 'R—flfi ______1L__7fr__. \L - Compute I e l Compute Storage Time Indicator (Loss Indicator *Minimize: ‘Timeliness Cost Yl <‘—_—" < ALP Feasible Regional , Mill Capacity: XP READ = 1 Facility l Charactistics Initiai Guess: 2| I M111 Machinery A Performance _ («6 e Minimize: ’ 7 Regional Fixed and Variable Mill Cost (Y2) Regional Rough Rice Distribu- tion Policy Figure 5.3: Simplified Flow Diagram for Regional Rice Mill Mechanization Model. 88 5.3.3 Model Construction The model consisted of the two main parts as de- scribed previously; the MECMIL for regional mill capacity and the SEEFAC for size Selection. Figure 5.4 gives the simplified flow diagram for the computer programming. 5.4 Model Implementation 5.4.1 Mechanized Mill Regional Capacity The yearly production in subdistrict Ciawi was 22,304 tons of rough rice and in Ciomas 27,843 tons of rough rice. Ciomas had 16 Engelberg Steel Huller (EGH); 45 Rubber Roll Huller (RRH) and 2 Rice Mill Unit. Ciawi had 13 Engelberg Steel Huller and 5 Rubber Roll Huller, and part of its rough rice production was transported outside of the district for milling. Ciomas however had to close some mill facilities due to lack of input materials. Small mill facilities tended to go out of business first, but some medium Rice Milling Units were also found closed. The system model simulated the Optimal mechanized mill capacity in each subdistrict based on input data and simultaneously compared that with conditions at the year 1977 - 1978. The results are given in Table 5.5. The field study carried out in 1978 indicated that Ciomas had an over-capacity of rice milling mechanized operations. Most of the mills Operated less than 500 hours per year. The simulation projected that for optimum conditions the regional mill capacity should be decreased 89 Table 5.3: Results of Regional Mill Capacity Simulation Subdistrict Specification Ciomas Ciawi 1. Regional mill capacity, in tons of rough rice/year a. 1977-1978 41,580 8,960 b. Simulated, optimal 12,594 10,133 2. Idle time indicator, in percent a. 1977-1978 (Standard) 100 100 b. simulated 23 116 3. Storage loss indicator, in tons of rough rice/month a. 1977-1978 3.4 7.1 b. simulated 6.9 6.8 by 60 percent. This would reduce the idle operation by 77 percent. It would however produce more storage losses due to under-capacity during the harvesting peak month. The storage losses would be increased by 3.5 tons of rough rice per month during the harvesting peak. Storage losses might be offset some by increasing hand pounding practices and by increasing the man-hours per day at the mill facil- ity during the harvest month. Regular man-hour per day was 5 to 6 hours in Ciomas. The model simulation showed the subdistrict Ciawi had an under-capacity in mechanized rice mills. The sim- ulation results suggested at least a 15 percent increase in mill capacity in order to improve the millers profit 90 while also reducing the farm storage loss chance. The 15 percent increase indicated a storage loss reduction of about 4 percent. Idle operation time would increase 17 percent due to over capacity during non harvesting. Never- theless, the idle time in Ciawi would still be relatively low compared to Ciomas. The Optimum condition in Ciawi was for a 20 percent less idle time than in Ciomas. The simulation model indicated a trade-off between the idle operation time and storage losses. Since that was not a simple linear interrelationship between these indi- cators, complex results occurred. In general, the idle time indicator dominated the system. Storage loss was less effective in the optimization process. Further analysis was done in parameter sensitivity. The percentage of hand pounding practices (HP) was found to have a linear relation with the model output. The optimum regional mill capacity did not change with alterna- tive price inputs since the model did not include the ef— fect of price distortion on mill technology performance. The variation Of several parameters do however affect the final value of the objective function (Y). Ten percent step increase in each parameter was generated to evaluate the effects on objective function. Most parameters did not have a linear relationship with the timeliness cost. Ten percent increases in rough rice price (RRPRI) reduced Y value by 0.4 percent, while ten percent reduction of the milling fee (XMILFE) reduced 91 the Y value by 9.5 percent. The milling fee showed a greater influence particularly in the penalty cost structure due to idle time. There was reduction of the Y value of about 4 percent for every ten percent increased of the Benefit-Cost ratio (BBC). The model therefore indicated a benefit to the rice mill business as affected by the sys- tem and vica versa. The BBC was viewed as other inputs to the system modeling. 5.4.2 Mill Size Selection 1. Regional Mill Distribution Distribution of mill facilities in each subdistrict was simulated through optimization Of the objective func- tion, the regional mill cost. The computation was based on the output of the regional mill capacity program as discussed previously in this chapter. The optimum mill capacity was computed in Ciomas as 13,643 tons of rough rice per year and in Ciawi as 10,133 tons of rough rice per year. Two system alternatives of rough rice distribution of each mill class were considered in the model. First Alternative (ALT 1) was to follow the 1977-1978 situation in both subdistricts which was EGH at 60 percent, RRH at 30 percent, RMU at 10 percent and none for RMP facil- ity. The second alternative (ALT 2) considered was for equal distribution among mill facilities, which was 25 percent for each mill class. The average operation time 92 for milling activities per year is 1000 hours. The simu- lation results along with 1977-1978 mill facilities in the subdistrict is given in Table 5.4. Table 5.4: Regional Mill Distribution Mill Ciomas C1aw1 Class 1977-1978 Alt 1 Alt 2 1977-1978 Alt 1 Alt 2 EGH 16 27 11 13 20 9 RRH 45 8 6 5 6 5 RMU 2 1 3 0 1 2 RMP 0 0 1 0 0 1 Alternative I indicated that 80 percent of RRH should be reduced in Ciomas subdistrict. This reduction was mainly due to the lower level of regional mill capacity in 1978. The shift to more EGH might be favorable since it is the simplest technology and required less cost and spare parts than others. Alternative 1 in Ciawi indicated an increase of all mill facilities, as higher levels of regional mill capacity was predicted. The output in Alternative 2 showed the difference between mill class due to the mill cost variation and mill performance of each technology. The result showed more of the EGH facility. However, the simulation model did not include the quality of milling output as a system 93 variable. Generally, the EGH milling output has more broken and less polishing of rice (IPB, 1975; IRRI, 1977). As discussed in Chapter III, the EGH has a higher mill loss than others. Milling quality consideration might shift the technology. 2. Mill Class Comparison. To compare the mill tech- nology the model was run for subdistrict Ciomas with dif- ferent inputs on rough rice distribution. Each trial placed the whole rice production into each mill class. A relative comparison scoring method was then formulated. The lowest comparison indicator was set as a 100 score. Table 5.5: Mill Class Comparison Score Mill Class (a) Indicator (0) EGH RRH RMU RMP Milling cost 100 124 196 222 Operation time 286 334 176 100 Fuel 100 117 185 175 Skilled Operator 195 100 189 200 The mill cost indicator favored the EGH and RRH. The operating time was most favorable for the RMP and RMU. An energy conservation program would favor the simplest technology. The smallest labor cost was for the RRH facility. 94 3. Plan Projection. A future plan for milling technology development must consider the predicted input of regional rough rice production. Simulation forecast- ing based on data from two subdistricts was carried out using the Regional Rice Production Model. It was assumed that the trend of harvest area change and yield per hectare, for both wet and dry land rice followed the gen- eral pattern in Java (see Chapter IV). The model was run with the assumed 1977-1978 post harvest technology condition. The simulation results as illustrated in Figure 5.4 combine past data (1972-1976) collected by the district Office of Ciawi and Ciomas. The simulation predicted an increase of rough rice production in Ciomas of 400 tons per year in Ciawi, 800 tons per year for the period of 1975 to 1985. Consequently the regional mill capacity should be adjusted to changes. Computations showed that after the Optimal condition was achieved the regional mill capacity in Ciomas should be increased 3 percent and in Ciawi 8 percent per year after the optimal condition was achieved. These figures may be altered as different post harvest technology are introduced to the simulation model. 95 .wzmwo pew mmEoHO cfl :Ofluospoum moflm nmsom mo :oHumHDEHm mcflummomuom "v.m musmflm mm 3 8 Na .b cm as uh hm 05 ms 45 Mb Nu. Nb 202443.25 II 3th wot—".0 ...0.m....m.n. 5.0.1....— <... mm Hm mm OOH flmmzmasm susom HH> ma 44 O4 OOH wamm Ono m>mh ummm H> ON ON O4 OOH m>mO Dance: > Hm mm OOH Add m>mn ummz >H OODQEOA can Ha he O4 vow muumesm cusom HHH ON ON O4 OHM muumfism ummz HH OH OH mm Hm muumesm nuuoz H .HHm. .Hmv .HHmO .Hmv HH >O>Hsm H >0>wsm HH >m>usm H >m>usm mocfi>onm .Oz suflaaomm Haas moam quHmm Damagedmmm H.4 manna 109 To gather information on post harvest operation out- puts, samples Of rough rice and milled rice were collected from the survey area. Each sample weighed about 25 grams, was examined for its physical and chemical characteris- tics in the Food Laboratory of Bogor Agricultural University. Rough rice samples were collected from farmers and mill sites where milled rice samples were collected from mill sites and several markets. Amounts of samples collected is given in Table A.2. In order to gain a general view on post harvest Of rice in Indonesia several results from the survey are pre- sented in Table A.3 to Table A.8. Some of the results have been related to the System Modeling under study. The Survey Report will be published by the Central Bureau of Statistics of Indonesia. 110 OO O4 4m O4 mm mm O4 Hm xH O4 I O4 O4 am I O4 I HHH> OO 4O O4 O4 O4 m O4 OH HH> OO mm mm mm mm mm O4 OO H> OO 54 O4 4N O4 OH O4 ON > ONH mm m HH O m OOH 4N >H O4 Ob NH O4 NH mm O4 Om HHH O4 O4 O4 OH ON ON O4 mm HH Om O4 Om OH Om OH O4 mm H HHO Hm HHm Hm HHO Hm HHm Hm mumxumz HHHS OOHm HHHS OOHm mHmEumm mocH>oum . .Oz OOHm OOHHHZ OOHm smsom mOHmEmm :Hmuu .N.4 OHQOB .mocmHumdxw mumEHmw co pmmmm« 111 NO OO O 4 OO NN OO HO O4 mszOO O .4 4 OO H NH 4H ON O O OH 0;\OOO OI4 .0 N OH O N4 OO ON N OO ON OO\OOO OIN .n N O 4O NO O OO O 4N NH O;\OOO H .0 :OHumuso umm>um= .m HH O 4O ON NO NO O O NN OO\O02 ION .4 OH O O OH O 4 O O NH O:\242 ONIHH .0 44 OH N OH O N NO N4 NH 0:\:02 OHIO .3 ON NO H O O N HO N4 4 On\cme OIH .O Henna .4 O OOH OO NO O O4 O OOH OO waoHO .n OOH O O O4 OOH OO NO O N Oc< I Hc< .O ucwEmHsvm umm>um: .m OH NO HO OO OO HO OO ON OO OwOmHmO .0 OO O O O NO O N ON O .OEHO EOEHOOo OH .2 O OO O NH ON O O H O O0HHO0m .m OEHB mcHumm>umz .N O O O O H O OO NH O 000 I >02 .O ON OH OH O 4O N OO OH OH H0O I amm .0 NO O O4 O4 4H 44 ON HN OO O=< I HOO .O O HH N N HH 4 N HN H can I O0: .0 OH N HO ON OO O4 H 4H H Hac I HO: .2 O NO O OH N N O 4H N4 amp I :44 .m cucoz OCHumw>umz .H xH HHH> HH> H> > >H HHH HH H . OOHDOOHOHOOOO mocH>Opm .mucwoumd :H :oHuanHuumHQ >Ocmsvmum. mmOHuomHm OCHumm>umx m.< OHOOE 112 O 4 O ON O OO O O4 O OoO0H 000HO .n OOH OO OO NO NO NH OOH OO OOH 0:0: .0 HOOOH mcH>Ho O OO OH OO OH OO NH OH O . 08: OOHOOO 0O000000 .n OOH ON OO N4 OO OO OO 4O NO O0O0H 4003 00 oon50n 00>0 .0 mcH>Ho OOHm cocoa O O O ON O N O O O 002000;» H0001 .0 OOH O O NN ON OO OO 4O O OOO00000 H0sc0a .0 O OOH OOH O4 N O O 4O OO O0HO0020 .n O O O O ON NH NN N ON HH0HO On .0 acmEmHsvm mcHzmmuze O4 O O O O O O O H O00 IO .4 HH O O O O O O H N O00 OI4 .0 OO O O OH NH O ON OH NH O00 OIN .O 4H OOH OOH OO OO OO 4O OO OO O00 HI .0 mchmwuze OHOOOO OOH m>mo OO O O ON O OH N H H HooHOIOOHO00 OH .0 ON O O O O O O O H 0H0HOIOH .n NN OOH OOH OO OO OO OO OO OO 0:0: .0 OOOOOO OOO0O HH0OO xH HHH> HH> H> > >H HHH HH H coHumOHwHomdm OOCH>OHO .mucmoumm CH mcoHuanHumHo >ocmskumv mmoHuomum mcHnmmune UGO mcH>HQ r}. I II 4.4 OHQMB 113 ON ON 4 OH N HN OO NO OO HHOO .O O O O OH OO NH O O H xonIoonE00 .0 4O OH NO O O4 ON O N N4 xonI0003 .0 O4 OH OO OO NN 4O OO O4 HO O0O .0 OOOEOHOOM OOOOODO .m H.N 4.H O.N O.H O.N O.O O.O O.O O.O OOOOoe OH .OHOO0H 0O0O0>< OOOHOHO EHOO .4 OH OH OH 4 OH NH NN OH OH 000 I >02 .O OH NH OH NN NH OH OH OH OH 00o I 00m .0 OH OH OH OH OH OH OH OH OH OOO I HOO .0 NH OH OH NN OH OH OH NH OH OOO I >0: .0 NH OH OH OH NH OH OH OH OH OOO I O0: .n 4H OH 4H O OH OH NH OH NH O0O I O0O .0 OOOoz OOHHHHz .O O N O NH O O O O O OOOoe\00OHH IOH .0 O ON H OH OH O O O O O0O02\00OHH OHIO .0 NO O4 H4 HN NO O4 O ON O OHO05\00EHH OIO .0 OO NO OO O4 OO OO NO NN NO O0O02\00EHH NIH .0 Oocmsvmum HHHZ .m OOH OOH OOH OO OO NO OO OO OO HHHz H00HO0O00: .0 O O O 4 N O N NH N OOH0O00O 0O0= .0 mmstccowB mcHHHHS .H xH HHH> HH> H> > >H HHH HH H OOHOOOHOHOOOO OOOH>OHm .mucmouwd :H OOHHOQHHDOHQ >ocmswmumv mOoHuomHm OOOHODO new mzHHHHz Eumm m.< OHOOB 114 mums OHOOUOUOCHO O.NNH . O.OOH 0.0HH O.OO H.OOH N.NOH O.OHH O.OOH OO\O0HOOO OH OOHHO OOHO OOHHHS uoxumz Hausa .O . H.ON H.O4 O.OO O.NO O.OO 0.00 O OO.HO OONO0HOOO OH OOHHO mon cmsom Eumm .O O.N 0.0 0.0 O.H O.H O.N N.H 0.0 4 OOHH mx\zmHO:m OH umou uuoamcmub .O 0.0 « O.4 O.4 4 0.0 0.0 N.O 0.0 OOHH OOHHHE OO mucmoumm OH .O 4 4.0 N.O « 0.0 4.0 0.0 0.0H O OOHH £OOOH Ox\:0HO:m OH .0 00m OOHHHHz .O O H.N O.N O O O.O O.O . O.H . N000O OO\O0HOOO OH umoo OCHOHQ .4 O.O O.N H.N N.N 0.0 O.N O.N 4.0 4.4 OOO0O OO\O0HOOO OH umou mcHnmmuce .m O O O O O H O OH O 0O\O0HOOO IOOO.OO .O OH O O O O N O OO O 0;\:0HOOO OOO.OmIOOO.ON .O OO O O ON m O4 OH OH 4N 0c\:0HO:m OOO.ONIOOO.OH .O OH OOH NO OO NO O4 NO m4 OO 0:\:OHOOO OOO.OHI .0 mucmoumm OH COHODDHHHOHQ >ocmsvmum umOO OcHumO>H02 .N H.O4 4.N4 O.NO H.NO 0.04 0.04 0.0N 4.04 4.Hm mucmoumm OH msHQHOO OHOODOHHOZ OOHO cmsom .H xH HHH> HH> H> > >H HHH HH H OOHDOOHOHOOQO OOOH>OHO umoo umO>H02 umom new OOHHOHHOZ O.4 OHOOO 115 mumo mumswmnmcH4 OO.H 4 OO.H HO.~ 40.H ~O.~ OO.N mO.~ 40.H mmmconzu .o OO.~ 4 NO.~ OO.~ 40.~ mO.m OO.m OO.~ OO.N nu©H3 .n O0.0 4 N0.0 O4.0 «0.0 Hm.O O0.0 O0.0 O0.0 numcmH .m as CH mNHm :Hmuo .4 O0.0 4 O0.0 O0.0 m4.0 ON.O 40.0 O~.O 4~.O muHm HHHE :H .n N0.0 4 H0.0 O0.0 H0.0 Om.O m0.0 ~m.o mm.O Eumm :H .m mucmoumm :H ucmucoo mHmHumumz :OHmuom .m 4N.O 4 Om.m NN.~H N0.0 OO.m O~.4 Hm.m OO.4 muHm HHHE OH .2 H0.0 4 OH.m O~.OH 4m.O Om.m Om.m 4m.m mO.m Eumw OH .O mucmuumm :H :Hmuu mmchme .m Hm.mH 4m.4H OO.mH 4m.4H OO.mH O4.0H H4.mH Om.mH Om.OH HH >m>usm a mm.4H 4 OO.mH HO.4H O4.4H O4.4H m4.~H 45.4H mn.4H H >m>usm I "muHm HHHE :H .n OO.mH OO.mH OO.4H OO.4H O4.0H ~4.0H OO.mH Om.4H OH.OH HH >m>uam I NO.4H 4 4N.mH mw.mH OO.4H H4.4H mm.mH mO.4H NO.4H H >m>usm l "Eumm :H .m mHmmm uwz mnemoumm :H ucmucou musumHoz .H xH HHH> HH> H> > >H HHH HH H coHumoHMHoomm mocH>oum :oHqucou mon cmsom O.O mHnme 2116 name mumsvmnmcH4 OH.O 4 «O.N mm.4 mn.H m4.0 OO.H mO.N NO.H muHm HHHz :H .n m4.H 4 O4.H Om.H m4.~ HO.~ NO.H HH.m OO.H umxumz uoHuumHo CH .0 Om.m 4 4 O0.0 OO.N O4.0 40.m 40.m Om.O umxumz uoHuuchnsm :H .2 4 4 O0.0 OO.H mm.H mO.m mm.4 4 mH.H umxumz Hausm CH .m onH @mHHHE mo mucmoumm CH mon Hmsmum .O OO.>O H4.mm OH.OO H~.mO «0.00 mm.~h OO.HO NO.mO OO.HO muHm HHHE :H .c ~0.00 40.NO m4.00 H0.00 mm.4> Hm.mO H4.mO 4m.4O N0.00 umxumz uoHuumHo :H .o 4m.HO OO.~O O0.00 Hm.OO 4m.OO 4H.HO Om.~O O4.00 Om.OO umxumz uoHHumHnnsm OH .3 4 Om.OO Hm.4h O0.00 O4.00 O0.00 40.NO 4 O0.00 umxumz Hausa :H .m QOHH UmHHHE no mucwouma cH mon 660: .m OO.~N 4 m~.m~ Om.m~ 40.- OO.m~ OH.mm OO.mm m~.mm mHmmn Own mucwoumm :H ucmucoo mmoH>C4 .4 OO.HO 40.NO 4 Hm.mn 4m.~> HO.~O OO.mO OO.HO 4 mHmmn xun mucmouwm :H unaucoo mumu©>nonumu .m «O.O ON.O OO.O HO.O OH.OH NO.O OH.O HO.O Om.m mHmmn >OO muOmoumm OH ucmucoo chuoum .N mm.mH OO.HH OH.~H «O.NH O~.~H OH.mH O0.0H ~O.HH 4m.mH muHm HHHz CH .6 OO.MH OO.mH OH.~H ON.NH NO.~H «0.0H OO.HH Hm.~H OO.NH umxumz uoHuumHo :H .0 OO.mH NN.~H «0.0H HH.NH OO.~H OH.MH O4.HH m4.mH 4m.NH umxumz uoHuumansm :H .n 4 mm.~H NO.NH Om.~H OO.mH m~.mH OO.HH 4 O4.NH umxumz Hmusm :H .m mHmmn umz mucmuumm :H ucwucoo musumHoz .H xH HHH> HH> H> > >H HHH HH H :oHumoHHHoQO mocH>oum OoHOHOOou mon OmHHHz O.O mHOOe ICHIGRN STATE UNIV. LIBRARIES "WWI”HmNIHWIWIWIW"WWNNIIINWI 31293100633860