Probabilistic quality control in non-gaussian process control applications

In this paper, a multi-objective control algorithm is proposed for non-Gaussian stochastic batch process control and quality monitoring. As for batch process control, the controllable system parameters shall meet the required levels in pre-specified times so that an intermediate product satisfies the required product quality attributes. As such, a joint problem of output and so called Dwelling time control is introduced. However, as non-Gaussian input/output disturbances exist, the system outputs and subsequently the dwelling time are non-Gaussian random processes. Therefore, the idea of minimum entropy control idea is applied to develop a new solution for the above multi-objective control problem. The process runs in time domain intervals called Batches. Within each batch the process is controlled by a set of fixed control parameters. Between any two adjacent batches, the controller parameters are tuned by an evolutionary algorithm (EA) so that the integral sum of tracking error (ISTE), entropy of tracking error (ETE), integral of absolute dwelling time error (IADTE), and entropy of dwelling time error (EDTE) are minimised. In addition, the physical and operational constrains of the process plant such as control signal level, etc. shall be satisfied. The proposed algorithm is applied to a batch Continuously Stirred Tank Reactor (CSTR) model where promising results have been obtained. The method can be used as a basis for joint probabilistic process/quality control where more complicated quality measures are applied.