A rule-based heuristic finite capacity scheduling system for semiconductor backend assembly

Finite capacity scheduling (FCS) has been widely implemented in manufacturing companies for detailed shop floor operation management. Most of the traditional and artificial intelligence-based FCS systems lack the ability to provide a satisfying and practical solution for the large-scale scheduling problem in a complex environment. This paper describes a rule-based FCS system for the daily production scheduling (DPS) of a semiconductor backend assembly company. DPS is the FCS module of the genetic scheduling system (GSS), which has been enhanced and implemented successfully in a semiconductor backend assembly company, which is a world market leader in semiconductor packaging technology. Since the focus of this paper is on the DPS module, the overall framework and the other core modules of GSS are briefly covered. The detailed design of the six DPS modules, namely data input/integration, model configuration, data pre-processing, job prioritization (JP) executive, machine selection (MS) executive and result output, are presented in detail. The inference engines of JP and MS executive modules are discussed, which perform a unique rule-based job prioritization and machine selection process of the work order list and the assignments on the machines to achieve the desired scheduling objectives specified by the end user. Group technology is incorporated into the system to categorise the resources and parts for set-up/changeover reduction consideration. The successful deployment of the system in the industry endorsed the fact that the implemented DPS system has the capability to solve large-scale scheduling problems and performs well in the complex industrial environment. Favourable feedbacks received from the company include reduction in changeover and set-up, improved on-time delivery, increased machine utilization, decreased cumulative cycle time of work order, etc.