Fault Diagnosis And State Evaluation Of CTA Hydrogenation Process

Pure Terephthalic Acid(PTA) is an important chemical raw material. PTA production process is mainly divided into two parts. First Crude Terephthalic acid(CTA) is produced by liquid phase catalytic oxidation reaction using P-xylene(PX). Then PTA is obtained by CTA hydrogenation process. In CTA hydrogenation process CTA is dissolved in the deionized water by increasing the heat and the pressure. Then catalytic reaction with hydrogen in the fixed bed reactor which is full of Pd/C catalyst is occurred. The main impurity4-carboxybenzaldehyde (4-CBA) will be removed after the above reaction. At the same time, non-ferrous material will be deoxidized. At last, fiber grade PTA through crystallization, separation and dry units is got.CTA hydrogenation process is a complex industrial system which has strong nonlinearity, uncertainty, large pure delay and strong coupling characteristics. The safe and stable operation not only influences the product quality, but also affects the consumption of raw materials, energy and other technical and economic indexes. So CTA hydrogenation process fault diagnosis and operating state evaluation research has very important significance. In this paper, the main work is as follows:(1) This paper establishes the normal statistical monitoring model by using the Kernel Principal Component Analysis (KPCA) with the historical data which represent the normal operates condition first. But the outliers in data sets will destroy the covariance structure of KPCA. It will make the model cannot truly reflect the real normal condition. So in this paper a method which combining outlier detection algorithm eliminates and KPCA algorithm is put forward. It eliminates the effects of outliers on the statistical monitoring model. At last CTA hydrogenation process fault is diagnosed using the squared prediction error(SPE) and Hotelling T2.(2) Least Squares Support Vector Machine (LSSVM) algorithm is applied to fault identification of CTA hydrogenation process. Because the parameters have large influence on LSSVM’s performance. In this paper, combined with the advantages of the Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), a hybrid Genetic Partial Swarm(GAPSO) is put forward to optimize the parameters of LSSVM. Since it is often difficult to collect the abundant fault data in the factory, ASPEN PLUS based CTA hydrofining process model is used to generate the needed data. The simulation results show that the method of fault identification precision has better performance when it is applied to the CTA hydrogenation process.(3) In order to guarantee the CTA hydrogenation process to stay pin the good operating state.This paper uses the charts quality control (SPC) and process capability index to analysis the important production index of CTA hydrogenation process. Finally the operating state of the whole production process is evaluated and marked by the excellent, the normal and the bad conditions. Thus it can improve the production quality and reduce the consumption of material and energy though maintaining excellent operating condition, improving the normal operating state and adjusting the bad operating condition.