Bioprocess soft sensor modeling using SVR based on improved PSO

In order to solve the problems of slow convergence and high complexity in dealing with large number of samples when modeling a bioprocess soft sensor using SVR, an improved PSO is used to solve the quadratic optimization problem to determine support vectors and the weights of them in this paper. By introducing constraint relationship in parameters, PSO could converge to the optimal point faster. Simulation results demonstrate that the proposed method could be simple to implement as a bioprocess soft sensor and has better convergence, which is presents as an alternative to current SVR training methods.