Research On GEP-based Process Monitoring And Intelligent Aided System For Chemical Process Safety Operation

The chemical processing industry has been one of the bridgestone industries in our country which has made a great contribution to the modernization and social prosperity, and supplied rich and colorful chemical products to the market. Meanwhile, the chemical industry is one of high-risk industries. (1) Most of chemical products are flammable, explosive, reaction active, toxic and caustic. (2) Chemical processes are always large scale and high integration, with high temperature, high pressure and high exothermicity. And accidents will happen even in case of slight variation of process parameters or improper operation as strongly nonlinearity of chemical process. (3) Chemical industry is a technology and capital intensive industry in which accidents will bring great damage to property and the environment. So, safety and stability of chemical process operation have been the critical issue enterprises and also one of the hot spots in the area of engineering research.Large amount of data containing process information have been collected and saved with the use of DCS (Distributed Control System) and computer technology. At same time, the data have not been used effectively. This phenomena is called"data overload, information poor". On the other hand, a great amount of experience and knowledge accumulated from production practice, especially the experience acquired by dealing with abnormal situation have not been well used. So how to convert the data to useful information and mine feature information which affect the safe operation of the process from the data, and then improve the long-term stable and safe operation of the process industry is the research focus in this paper.This paper is focused on the safety operation problem in the chemical process. By analyzing the process data with process modeling and artificial intelligence technology, studying and developing intelligent auxiliary system for safe operation of the chemical process, which can be used as assistant support tool for operator, especially assisting the operator to make decision when the occurrence of abnormal situation and reduce accidents and unqualified products. The main works in this paper are as follow:(1) Feature extraction algorithms of chemical processes. The chemical processes are always characteristic of high dimension, high noise and strongly nonlinearity. It is difficult to identify the process operation mode from these data. Using feature extraction, data can be decomposed, and removed redundant or correlative, and extracted from process feature information. Dimension of the input data will be reduced as well as, which original data set expressed with the minimum number of feature set. The identification accuracy of operation mode will be improved. The method based on kernel function can remove nonlinearity of data but only at the expense of data dimension. To this problem, a Principle Component Analysis (PCA) method based on Gene Expression Programming (GEP) has been introduced in this paper which extracts well the nonlinear feature of process and effectively reduce data dimension.(2) Classification methods of operation state for chemical processes. The essence of chemical process operation state monitoring is mode classification. Support vector machine (SVM) is one of the valid methods for classification. However, support vector machine is a method based on kernel function; its performance depends on the selection of kernel function and parameters. At present, there is no effective method for the selection of kernel function and its relative parameters but only from experience which is time consuming and not well performance. To this problem, an auto-search algorithm for kernel function by using of GEP has been introduced in this paper.(3) Knowledge acquisition for fault diagnosis. The key of expert system is expert knowledge which has always been the bottleneck problems. So knowledge refining is the key technique of developing fault diagnosis expert system for chemical processes. In this paper, an approach of GEP-based knowledge auto-refining technique for fault diagnosis of chemical process has been introduced, which can effectively extract the knowledge of fault diagnosis as the knowledge refining technique for fault diagnosis expert system.(4) An intelligent aided system for chemical process safety operation is developed. The system is integrated data acquisition, data pretreatment, state monitoring and fault diagnosis. The structure and the main function for the system are described in detail. The following problems are considered in the process of the design for the expert system. In the process monitoring, nonlinear PCA method is for feature extraction, and GEP-SVM method is for feature classification. In the process of knowledge acquisition, a method combined fault tree/experienced knowledge table, decision tree and GEP-based knowledge auto-acquired method is used to solve the bottleneck problems of knowledge acquisition. In the knowledge base organization, knowledge base is designed as global knowledge base or inner memory knowledge base, and introducing a knowledge selective strategy of inner memory knowledge base. The developed system has been applied to a lubricating oil process in a petroleum plant. The validity and practicability of the proposed system is proven in practice. Further, the frame of the proposed system as well as some key techniques has been confirmed.