Artificial neural network application to a process time planning problem for palm oil production

Abstract

The demand for palm oil is rapidly growing, but its production is faced with unreliable manufacturing indices, including processing time standards. In this study, an artificial neural network (ANN) model was developed as a solution for the prediction of standard time. The primary data were obtained through an industry survey and a direct time study measured at an oil mill using a stopwatch for each process and recorded on a standard time observation sheet. These direct time data collected over a 12-month period were standardized into numeric input data for ANN. A standard multilayered, feed-forward back-propagation type of neural network architecture was proposed. Direct time study data involving eleven different operations for a 22.5 tonne capacity Roche palm oil mill in Ohaji-Egbema were used in training, testing and validating the network. Time Processor software was developed in FORTRAN for investigating the quality of the trained network's output and standard time. Also, the labour and cost requirements of the mill were effectively optimized using linear programming (LP). Results from LP showed that optimal cost requirement of the mill was 6,330.16USD per month. This amounts to a savings of 86.92%, compared with current requirement of 48,395.14USD per month. The ANN model output was 423.666mins compared with the current time of 540 mins for processing the same palm fruit. This shows that time standardization through ANN provides a savings of 21.54%. Thus, the developed ANN model has a reliable and good prediction capacity. It can be applied in a timely manner to medium and large scale oil mills or similar processes.

Keywords

• artificial neural network
• capacity planning
• labour requirementscosting
• process control
• process standard time
• palm oil