During a previous internship, a tool for simulating batch execution in a clean room was developed. This tool takes into account processing times on equipment, equipment failures, and certain holds related to integration. The batches injected into this simulator come from the actual history of the clean room.
The goal of the PhD is to develop a simulator that can prospectively simulate batch execution based on the POR routes of the main themes present or upcoming in the clean room. Based on the POR routes, the tool should be able to generate development batches for technology bricks (short loops), as well as functional batches including test plates and pilot plates. A nomenclature and enrichment of the routes through metadata will need to be carried out to enable the tool to generate batches realistically, both in terms of process and project scheduling.
Different simulation engines will be compared in terms of performance and accuracy. Classical resolution engines (discrete simulation, event-driven, conjunctive graph-based) as well as innovative approaches (primarily reinforcement learning, but also supervised learning) will be studied.
The development and publication of a methodology for creating simulation instances (testbed) will also be carried out during this PhD work.