Research and analysis of the prediction model of wiped film evaporation process based on PSO-SVR

Due to the relatively high cost of parameter measurement in the wiped film molecular distillation process, the relatively low accuracy obtained and the relatively long measurement time, the soft-sensing model established by traditional support vector machine regression has relatively poor predictive effect. So for the solution of these problems, this paper constructs a combined optimization model based on support vector machine regression and particle swarm algorithm, which uses Partical Swarm Optimization to optimize the penalty factor coefficient C and kernel function coefficient g of support vector machine, and then uses Support Vector Machine regression algorithm (SVR) to realize the prediction of process parameters in the distillation process. A the same time, this paper uses SVR and BP neural network to predict the process parameters. The prediction results are compared with PSO-SVR. The experimental results show that the effect of the soft sensor prediction model of PSO-SVR is much strongeh than that of the other two models. PSO-SVR has Higher accuracy, and it can be more economical and practical when faced with more complex measurement systems