A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering of the Asian Institute of Technology, Bangkok, Thailand

Thesis (M.Eng.) - Asian Institute of Technology, 1978

A sugar milling district consisting of a milling facility and an area producing cane in central Punjab, Pakistan was selected for this study. The objectives of the study were to develop meteorological and production models for future management studies and to produce descriptive statistics of the system. The purpose of these model s was to provide inputs for the simulation of Sugar-cane systems to synchronize the operational policies of the mill and harvesting policies of the crop. For rainfall amounts weekly lag one serial correlation coefficient was 0.52, significant at 99% level of probability, therefore lag one Markov Chain with the principle of Monte Carlo technique was used for data generation. The data genera ted with this technique when compared with historical data indicates close co- operation for mean, standard deviation, skewness coefficient and lag one serial correlation coefficient. The frost data was also generated with the same technique and when compared with historical; indicated a good agreement for mean and standard deviation, however the skewness coefficient was not preserved. This study indicates that Markov Chain technique can be used effectively for rainfall and frost data generation. Linear and logarithmic regression were used for production modeling using number of plowings, number of irrigations, amount of fertilizer, number of hoeing, ground water table depth, age of the ·crop a t harvest, and month of harvest as inputs . The linear and logarithmic models used explains 33%-68% variability in the yield and the results include only those inputs which were significant at 99% level of probability. With those factors of productions the yield of Sugar-cane crop gives comparatively better fit with linear regression as compared with logarithmic regression. The mill performance models were developed by linear and poly the mill performance models are exceptionally good. nominal regression. The results indicate 98%- 99% variability, indicating that the mill performance models are exceptionally good.