Application of data reconciliation and gross error detection in mixing tank for battery production process

Abstract

Process measurements are taken in chemical plant for the purpose of evaluating process control or process performance. In general, measured data inherently contain inaccurate information because measurements are obtained with imperfect instruments. The error in measured data can lead to significant deterioration in process performance. Data reconciliation and gross errors detection techniques are the techniques widely applied in various production processes for reducing the effect of measurement error. Therefore, the application of these techniques in mixing tank for battery production process is very attractive. This work is divided into two case studies. The first case study is to determine the reconciled data from the data reconciliation problem formulated by using three different methods: Contaminated normal distribution, Lorentzian distribution, and Hampel's redescending M-estimator. The other case study is to test the performance of these three methods under the condition of measurement with different degree of error. The results of these methods for nominal case which contain only random errors the three different methods used provide good reconciled value and in the test case which contain both gross errors and random errors, the [rho] function of Hampel’s redescending M-estimator is the most appropriated function for data reconciliation in mixing tank of battery production process.