Maintenance-free soft sensor models with time difference of process variables

In industrial plants, soft sensors are widely used to estimate process variables that are difficult to measure online. However, their predictive accuracy gradually decreases with changes in the state of the plants. Although regression models are reconstructed with database which includes newest data to solve this problem, some problems remain in practice. Therefore, we have attempted to reduce the effects of deterioration with age on soft sensor models without maintenance of the models. By constructing models based upon the time difference of an objective variable and that of explanatory variables, the effects of drift and gradual changes can be handled. We verified the superiority of the proposed method over traditional ones with simulation data and applied this method to actual industrial data. It was confirmed that the proposed method could achieve almost the same predictive accuracy as the updating model for 3 years without reconstruction of the model.