In order to improve the function of soft sensor to conduct variable selection, fault detection and model structure identification in the case of faulty state, a design method of new soft sensor is studied though the v...
详细信息
In order to improve the function of soft sensor to conduct variable selection, fault detection and model structure identification in the case of faulty state, a design method of new soft sensor is studied though the variable selection algorithm. A non-stationary time serial is introduced to describe the process output not being reflected by sensor variables and to detect whether the process enters the faulty state. A non-negative garrote method is adopted to identify the model structure and a modeling method for new soft sensors is presented. The obtained model can be used for both prediction, and detection of structural model change and the emergence of disturbance. Compared with the ordinary soft sensor based on partial least square algorithm, the advantages of the proposed method are demonstrated by a simulation example and an industrial application to temperature prediction of a blast furnace hearth.
暂无评论