Coke Deposition Measurement Based On The Analysis Of Catalyst Images

Automatic control of multiphase flow reactors is a key issue to be addressed in the catalyst based continuous production process. To ensure the selectivity and efficiency of the reaction, it is of significance to detect and analyze the active state of catalyst. Coke deposition behavior is a main cause of the catalyst deactivation, e.g. in the processes of MTG (methanol to gasoline), MTO (methanol to olefins), FCC (fluid catalytic cracking) and GTO (natural gas to olefins). Coke deposition on catalyst will not only reduce catalytic activity and selectivity, but also affect the product yield, the reaction residence time, the regenerator temperature and so on. Thus, the amount of coke deposition on catalyst is one of the most important indexes of catalytic property and service life. As a result, it is essential to measure the amount of coke deposition and analyze the active state of catalyst during the continuous production process.On account of the limitations on all existing methods, the aim of this paper is to propose a new method to measure the amount of coke deposition on catalyst through image analysis. An image acquisition system consisting of a flatbed scanner and an opaque cover is used to obtain catalyst images. Catalyst pellets regions are segmented successfully using a series of image processing algorithms, such as denoising, binarization, dilation and so on. After imaging processing, eleven colour features and six GLCM (Gray Level Co-occurrence Matrix) texture features are extracted from each catalyst pellet region and the Gray layer is selected as the most effective colour layer. Finally, twelve features are selected as effective features by prediction tests and two best feature sets F1, F2 can be obtained from the testing results.Back Propagation Neural Network with Particle Swarm Optimization algorithm (PSO-BP) and Support Vector Machine algorithm (SVM) are used to establish the prediction model of coke amount when F1, F2 are taken as the feature sets respectively. Comparing the prediction effects of these two algorithms, the PSO-BP algorithm performs better. The root mean square errors (RMSE) of the prediction values are all below 0.021 and the coefficients of determination, R2 for the model are all above 78.71%. The prediction results show that the PSO-BP prediction model is precise for measuring the coke deposition amount at various datasets. Therefore, a feasible, effective and precise method is demonstrated, which may be applied to realize the real-time measurement of coke deposition based on on-line sampling and fast image analysis.