Soft sensor modeling based on PCA and LS-SVM for strip thickness in cold steel rolling mills

Accurate and timely measurement of key performance variables is crucial for safe and economic operation of industrial processes. In the cold steel rolling process, strip thickness is the key performance indicator. However, due to time delay, it can be difficult to measure accurate and timely values. Therefore, a soft sensor can be used to provide timely and accurate values. This paper proposes a new approach to soft sensor design that combines principal component analysis (PCA) and least-squares, support vector machines (LS-SVM) that allows for a parsimonious model to be built. Using this approach, a soft sensor is developed using industrial data for the steel thickness in the cold steel rolling process. The soft sensor provides accurate estimation of the steel thickness.