A model predictive control approach for fab-wide scheduling in semiconductor manufacturing facilities

In a semiconductor fabrication facility, complex product flows involving hundreds of machines, many reentrant loops and various uncertainties (e.g., diverse equipment characteristics) pose a challenge in generating a production schedule that will ensure meeting of the product targets without excessive cycle time. In order to overcome these difficulties, a model predictive control (MPC) approach based on an aggregated flow model is developed. The MPC method is used to determine aggregated machine utilization schedules for each machine group at every shift. This optimization-based MPC allows the scheduler to simultaneously solve the constraint-aware production optimization and in-process inventory control problems at each scheduling instance. The performance of the proposed method is evaluated on a modified Intel mini fab case considering multiple products and changes in demands for each product.