Study On Applications Of Soft Sensor In Chloromethane Recovery

Chloromethane water content and recovery rate are two important indexes to estimate the performance of recovery device. However, the process of chloromethane recovery is very complex, that on-line measuring water content and recovery rate is very difficult. Soft-sensing Technique can effectively solve the difficult problem. Chloromethane recovery rate can be measured by mechanism modeling. According to the limitations of support vector machine incremental learning algorithm (ISVM), proposed improved ISVM, which can use to build soft sensor modeling of chloromethane water content.This article mainly includes the following:(1) Above all, introduced the background and cignificence of soft-sensing technology. Then, introduced the basic idea of Soft-sensing Technique and its implementation steps, and generally described the the industrial field.(2) Based on analyzing compress and condensation process, analyzed the change of mixed gas components in the process of recovery. Established chloromethane recovery prediction model by the mechanism modeling method, and simplified it. The forecasting results accorded with the actual conditions.(3) Based on introducing the support vector machine algorithm(SVM) in detail, according to the defects of ISVM, introduced the KKT conditions. Based on analysing and researching KKT conditions, found that the sample of violating the KKT in the incremental sample will change the initial support vector sample, and the initial not support vector sample maybe translate into support vector sample. So the improved ISVM was proposed. Firstly, the initial sample and the incremental sample are tested by KKT each other, then combine the sample of violating the KKT and support vector as new training sample. Last, Its feasibility was proved by a simulation example.(4) Because chloromethane water content can't be measured on-line, and the trying process mechanism isn't clear, build soft sensor modeling of chloromethane water by improved ISVM. First, analyzed the processes of chloromethane desiccation and determined the Secondary Variable. Then, pretreated these data, and build soft sensor modeling of chloromethane water content based on improved ISVM, which increased the prediction's accuracy compared with the support vector machine simple incremental learning algorithm.The soft measurement technique is applied to estimate indexes of the chloromethane recovery device performance. Successfully established chloromethane recovery soft sensor model based on the mechanism analysis and chloromethane water content soft sensor model based on support vector machine improvement increment learning. The research results is helpful for controlling chloromethane water content and recovery rate, and is helpful to improve the economic benefits of chloromethane recovery.