A Soft Sensor Modeling Method Based on Double-Layer Support Vector Machine

In order to effectively monitor the production status and achieve reliable control in industrial production, the timely detection of key quality parameters is very important. However, due to the complex conditions of the industrial site and limited detection technology, real-time on-line detection of some key quality parameters cannot be achieved in the actual production process. With the extensive application of various information systems represented by distributed control systems in the production process, the process data information has been greatly enriched, making data-driven soft measurement technology an important means for online detection of key variables of industrial processes. A two-layer soft measurement modeling method based on support vector machine (SVM) is proposed. One layer is used to analyze the relationship of industrial data in time series and solve the correlation problem of time series. The other layer is used to model and analyze soft measurement to solve the robustness of nonlinear regression model. The simulation results show that the soft sensor model has good performance.