Experimental Analysis of a Stochastic Backward Simulation Approach Under the Specifics of Semiconductor Manufacturing

Abstract

Manufacturing companies are experiencing many challenges with regard to customer-oriented and on-time production, especially in the context of an intensified global business and while advancing the digital transformation. Accordingly, the ongoing development and deployment of Industry 4.0 solutions for customisable products in small batch sizes, not only pose new problems for work preparation, but also for operative production planning in connection with high cost, time and quality pressure. Modern, complex and highly automated production systems (in this case from the semiconductor domain) must be operated close to an optimal operating state in order to be economically reasonable. Promised delivery dates and throughput times specified in contracted service agreements must be ensured and require a permanent and effective adjustment of production planning and control in daily execution. All other general conditions of an economic production remain unchanged. Today, the question of whether the production target is realistic and whether all promised delivery dates are met are today still answered with rather simple backward-oriented approaches, mostly without taking into account uncertainties, stochastic behaviour of the manufacturing system or alternatives that arise during operation. As shown in previous publications, these questions can be answered in more detail and more resilient using a backward-oriented discrete event-based simulation approach (SimBack). This article presents additional findings. The results show and deepen the impression, that the SimBack-approach can be successfully solve scheduling questions for customer-specific orders in a real-world environment.