Soft sensing of alumina concentration in aluminum electrolysis industry based on deep belief network

In the increasingly large-scale, complex and networked industrial process environment, some key process variables which are difficult to realize on-line measurement can be reconstructed in real-time and accurately by soft sensing technology. Alumina concentration is an important parameter in the production process of aluminum electrolysis industry. However, due to the high temperature, high magnetic and strong coupling field environment of aluminum electrolysis industry, it is impossible to carry out on-line measurement and real-time monitoring. In order to solve this problem, we propose a soft sensing model of alumina concentration based on deep belief network (DBN), and introduce time series to optimize the input parameters of the model to obtain more accurate results. The accuracy and effectiveness of the soft sensor model are verified by simulation.