An output-related Kernel Independent Component Model for Electro-fused Magnesia Process

In a complex industrial process, there are many production links with non-linear and non-Gaussian properties, making the sampled data highly correlated and difficult to separate. The traditional regression method based on the assumption of Gaussian distribution is not suitable for this method. In the production process of electro-fused magnesia, temperature is the index that can well reflect the production quality and operation safety. At the same time, it is difficult to measure the temperature because of the complex working conditions on site. To solve these problems, a kernel independent component regression (KICR) process monitoring method is proposed in this paper. The main contributions of the proposed method are as follows. A regression model is established between temperature and independent components. Different from traditional KICA, this paper selects the independent components that are most relevant to temperature in the regression. The operation status of production is monitored by the fluctuations of the output-related independent components. Finally, the smelting data of magnesia is used to verify the process monitoring effect of the proposed method.