Study On Online Detection Of Concentration Of Acid Pickling Solution Based On Soft Sensing Method

In strip hydrochloric acid pickling process, hydrochloric acid was added into tanks at regular intervals to ensure the pickling effect based on the remaining acid concentration of hydrochloric acid. Since the pickling solution contains a variety of mixed components, manual measurement by chemical titration method can't achieve hydrochloric acid supplementation process automatically. Based on the actual situation of a steel mill in Nanjing, this paper presents a soft-sensing method for the on-line estimation of the concentration of the components in the pickling solution. Based on the study of the pickling mechanism, the acid solution is mainly composed by FeCl2 and HC1. Thus the solution temperature, conductivity and density were chosen as auxiliary variables, and an on-line detection device was designed. Based on the characteristics of component concentration in acid solution and the variation range of actual component concentration, a soft-sensing modeling method combining mechanism analysis and neural network was proposed to estimate the components concentration, And the neural network was used to compensate the mechanism model to deal with the dynamic change of the production process and reduce the prediction on component concentration. In order to guarantee the reliability and stability of the model, the test data were preprocessed by Fast-MCD to remove the outlier data before model training. At last, the software system of on-line measurement device was designed, and the on-line industrial experiment is carried out by using the online measuring device. The results showed that the root mean square error for the prediction of hydrochloric acid and ferrous ion concentration are 5.53 and 5.83, respectively. The absolute value of the deviation is 5.4g/L and 5.7g/L respectively, which can meet the needs of practical industrial production.