Analytical And Gm-Ar Models Based Fault Pre-Diagnostic Method For Chemical Equippments

The petrochemical industry is more dangerous than other industries, and its accident impact is also more severe due to its attribute in nature. The fault diagnosis and prediction technology are the main methods to improve the safety essence of petrochemical industry. Because the two methods deal with the problems from different views, they have their own advantages and disadvantages. In actual production, if operators can know the trend of equipment condition in advance, and detect and separate the faults which may cause the equipment deviate from its normal state, they can well ensure the safety of the equipment and avoid the occurrence of major accidents.The analytical model can efficiently detect faults and isolate the fault source. However, it can’t detect and exclude the system potential fault. The prediction model summarizes and analyzes the data to get the trend. Most of devices belong to the gray system, so this paper chooses the gray prediction model to study. The gray model requires a small amount of data and predicts the macro trends very well, but lacks adaptability to the data fluctuation. The gray model cannot obtain the causes and consequences merely from the data changing.In this paper, the gray model is improved by combining it with time-series model in series. The residuals of the gray model are corrected to enhance the adaptability to the data oscillations. One fusion method about diagnosis and prediction methods is studied and the hybrid pre-diagnosis model is built in the paper. The gray prediction model can amplify the early data and the analytical model can perform fault detection and isolation using the predicted values. The pre-diagnosis model can not only qualitatively detect fault, but also quantitatively analyze the developing level and speed of the fault. There are two ways to combine the gray model with analytical model:1) Predictive and analytical model. Firstly, the prediction model predicts the variables related with the fault parameter. Then the analytical model deals with the prediction values and obtains the fault parameter.2) Analytical and predictive model. Firstly, the analytical model solves the historical values of the fault parameter. Then, prediction model predicts the trend and developing speed of the fault parameter.Finally, pre-diagnostic method is used in acetic acid distillation column of the TA process. The prediction method can obtain the data trend of distillation column, while the analytical model calculate load performance curve of the plate to determine whether the plate is in the normal operating range. Operators can set the parameters of the model through the man-machine interface. The system automatically pre-diagnoses the state of plate and provides reliable guidance for the actual production.