Optimizing the product-based availability of a buffered industrial process

Many industrial processes for discrete consumable products consist of a series (or set) of sequential process operations (or subsystems) which are de-coupled by means of in-process storage buffers. Each subsystem of such a process contains one or more parallel coupled or uncoupled operating lanes. We describe the use of a discrete-event simulation model for determining the availability of such a process. We likewise define and use a genetic algorithm to determine process designs and operating rules that have high availability. A 65-variable example, consisting of four operating subsystems with at most four lanes per subsystem, is used to illustrate the method. The results for this and similar real-world applications indicate that, by applying this methodology, it is possible to design buffered industrial processes having high availability.