Diagnosis Of Disturbance In Distillation Process Based On Inversion Method

Distillation is one of the most widely used operations in the process of petrochemical production.Minor disturbance to the distillation equipment and operation will lead to huge economic loss.So it is necessary to identify the cause of abnormal disturbance and find out the degrading tendency of parameters in abnormal state in time,to prevent the occurrence of distillation accident and guarantee the safe and stable operation of the distillation unit.In this paper,with disturbances of single and double variables as examples,the inversion problem of disturbance cause in distillation process is investigated.The dynamic mathematical model of distillation column is established by means of mechanism modeling to simulate normal and abnormal samples,which is based on the nonlinear and non-steady nature of dynamic distillation process.Correspondingly,the methods of artificial neural network(ANN),negative selection method(NSA)and support vector machine(SVM)are used to determine the type of disturbance.Then the inversion model of disturbance and feature expression quantity is established combining genetic algorithm(GA).So the disturbance could be analyzed by mathematic calculation.In addition,considering the complexity and uncertainty of the extracted features in the above diagnostic methods,the deep learning(DL)is used to identify the type of disturbance,which enhances the intelligence of the recognition process.Finally,the depropanization process is selected as the research object based on a dynamic simulator in this thesis.Then the inversion model of disturbances is established,and the common disturbances in the process are diagnosed.The proposed methods are verified with the common disturbances in this thesis.Case studies show that the proposed methods can quickly locate the disturbance type,obtain the disturbance quantity accurately with the perturbation data,and realize the further identification of the disturbance causes of dynamic distillation system.