Productivity analysis of a large multiproduct batch processing facility

Abstract

Effective and efficient operation of multiproduct batch processing plants is often difficult. Such facilities are subject to uncertainties, delays, interferences and bottlenecks that shift with the product mix. Predicting the performance of such facilities is increasingly important as competitive forces both increase the number of products and shorten the lead times. This paper deals with the analysis, modeling and productivity investigation of a five-stage batch processing facility making over 100 different product types on more than 80 pieces of equipment. The dynamic simulation model explicitly deals with batch time variations, equipment cleanouts, delays due to operators, interferences due to batches competing for equipment, and periodic planned maintenance shutdowns. The model has been successfully used to help evaluate a number of changes, and it has now been turned over to plant personnel to help in facility planning, production planning and production scheduling. I do not intend to get into further details; this is not required to tell our story of how the model was developed or what it accomplished. I will note that the model was done in SIMSCRIPT II.5 ( Russell, 1983) and is now quite large. The effort to get a reasonable first stage prototype model developed and ready for review was roughly 2 man-months spread over a 4-month period. Roughly half was for data collection and operations description development and half was for model development (designing—writing—debugging the model). The final model with all its refinements and modifications has taken almost 1 man-yr of effort and the model is over 9000 lines of SIMSCRIPT code; almost half of this is for reading standard data files and for developing customized reports. It has been used to evaluate many cases. Documentation both on the modeling assumptions and logic and on how to use the model has been prepared. The model was developed on VAX computers, and the executable has been moved to the plant VAX 8300; this transfer was relatively easy. Plant engineers have indicated a keen interest in using the model. Since they were part of the development team, they feel it is valid, useful, and not someone's else's invention. As noted earlier, we were asked to develop a model that could help answer a number of detailed questions about the capacity and scheduling of a highly dynamic, large-scale batch production system. Since we needed to be able to determine the impact of rather small-scale changes in this system we developed a highly-detailed discrete simulation model of the total plant. We found out that if it had just focused on scheduling the filling operations (which some people advocated early on) we would have not captured the full magnitude of the shifting bottleneck characteristic nor would we have really addressed the full problem. This model is viewed as a good representation of the plant and is capable of addressing the questions originally asked. Results are having an impact on plant operations. Recently we have been working with the plant scheduler to: (1) have the model read several data files generated from off-plant sources; and (2) make it easier to routinely run the model to investigate a variety of questions which deal with the capability of the plant to meet specific tasks and how to schedule around difficulties. We feel this model clearly shows that simulation models can realistically represent large complex multiproduct batch plants and further shows that simulation based scheduling tools can provide a productivity gain and a resulting competitive edge.