Study Of Distillation Column Fault Dignosis Based On Support Vector Machine

Distillation column occupies very important position in petroleum & chemical industry. As a hot field of distillation research, fault diagnosis is of significance both in theory and in practice.Support Vector Machines (SVM) is based on structural risk minimization principle of statistical learning theory, and it can solve the problem of small sample learning. It can establish classification machine with good generalization ability under the condition of small sample. Therefore, it is suitable for distillation column fault diagnosis, which is a engineering problem of small sample learning. This paper gives a deep research on the fault diagnosis in distillation process. Completed work here is summarized as follows:Firstly, based on large related paper, the comprehensive contrast on the different methods of fault diagnosis is made, in which the advantage and disadvangage are pointed out. Then the process simulation software Aspen Plus is simply introduced.Then, the core concept of statistical learning theory is introduced. The statistical learning theory, which is a machine learning theory of finite sample, has a quite solid theoretical foundation. Then the introduction of Support Vector Machines and multi-valued classification SVM is made.Next, because the nature of most chemical Process is nonlinear, according to the nonlinear properties of chemical process data, the kernel principal component analysis (KPCA) method has been researched. In this paper, the KPCA methods was used to monitor the acetic acid dehydration azeotropic distillation process, the simulation shows that the method of KPCA is effective in the fault detection of distillation process.Lastly, in order to solve the difficulty in selecting free parameters of SVM, this paper proposed a method of using improved genetic algorithm (IGA) to optimize the parameters of SVM. The IGA algorithm uses generation gap selection and alterable crossover rate, which ensure the most adaptive individuals pass from the current generation to the next and enable the optimized object to be stable easily in an advanced state of evolution, can raise calculation efficiency. Finally, in this paper, based on IGA, an integrated algorithm of KPCA-IGA-SVM is proposed. A case study on the the acetic acid dehydration azeotropic distillation process illustrates the effectiveness of the proposed method.