Real time optimization of an extrusion cooking process using a first principles model

Food extruders present a challenging multivariable control problem mainly because the product quality attributes cannot be measured on-line. The proposed strategy is based on a first principles model and is implemented in a cascade scheme with two control levels. The inner level consists of two decentralized PID controllers. The outer level is a controller-optimizer that generates the set points of the inner PID loop but also directly manipulates one process input variable. The controller-optimizer objectives are to reach the desired product quality and to maximize the throughput rate. To ensure that the first principles model used in the controller-optimizer always adequately represents the plant in its current state, a model-update algorithm is implemented. The cost function minima, in the controller-optimizer and in the model-update procedures, are found with the IMC-optimization algorithm. Simulations illustrate the good performance of the proposed strategy.