UKF-based soft sensor design for joint estimation of chemical processes with multi-sensor information fusion and infrequent measurements

摘要

Having a proper and real-time process monitoring and control requires accurate and frequent measurement of process variables. However, using a particular sensor for an important variable has high cost and low precision limitations. In such cases, laboratory analyses are an excellent choice for their high precision, but as their measurements are obtained manually and infrequently, they are not practical for every application in industries. This study presents an advanced soft sensor for a highly non-linear continuous stirred-tank reactor (CSTR) system which is observed with multiple sensors with different sampling rates under the assumption that there are different kinds of non-idealities in data acquisition for sensors. Some rules are assigned in order to vanquish these non-idealities in joint estimation algorithm. Therefore, the problem of simultaneous state and parameter estimation based on data fusion technique and unscented Kalman filter (UKF) is presented, and the effectiveness of the proposed method is investigated. Moreover, the proposed method is such that the effects of the inaccurate sensor on the parameter estimation are reduced. Simulation results on the estimation of four states and two parameters in a typical CSTR process show the proficiency of the proposed approach.