A Discriminant Method For Activator Invalidation Based On The Bayes Estimation

Petrochemical industry is one of the mainstay industries in our national economy. However, it is also the most contaminative. In the process of production, there is much contamination which is easy to burn, explode and be poisonous.It is necessary for a petrochemical enterprise to be highly safe and environmentally protective. Aimming at this goal, the activity status of activator is the most important in the inspection and control, especially for the chemical equipment with mass oxidizing reaction, because it is very significant to ensure the safty production, reduce air pollution, improve the effect of activator, and decrease the cost.In this paper, the sulphur recovery device is investigated. Based upon the Bayes estimation, a discriminant method is put forward for the activity status of activator. First,the chemical process is briefly analyzed, and then the influence parameters are concluded and the key inspect points are selected due to process analysis and expert experience. In accordance with the characteristics of device, the parameter influences on the process are analyzed. Second, the estimates for the influence parameters are formed and the models to predict the tendencies for the key inspect points are built up by means of multiple regression. Finally, the predictive values for the process parameters and related probabilities to occur are obtained on the key points by the regression model, and the activity status of activator is then discriminated by the layered Bayes estimation.Furthermore, a self-learning mechanism is elementary built up based upon the characteristics of the chemical device. The mechanism includes two aspects. One is for the predictions of process parameters, they are selected using the stepwise regression analysis to improve the estimateprecision. The other is for the selection of the key points, the common fluctuation characteristics of the data, which will happen once the activator is invalid, are concluded by the expert experience.At last, the research work is summarized, and the further research issues are also pointed out.