Study On Technique Of Soft Sensor Based On Rbf Network Optimized By Medo Pso Algorithm In Denimic Flow System

It focused on the soft-measure theory, method and practical technology. It is a complex problem to meet the need of real-time and accuracy in the measurement of dynamic flow in hydraulic pressure servo systems.The proposed RBF network was used to solve this problem. The work has very important roles in theory and practice for soft measurement method of dynamic flow.Firstly, introduced the theory of neural network, analyzed the RBF algorithm and its various improved algorithms. In order to improve the training speed and the forecast precision synchronously of the RBF network, the method applying the optimized sensitivity analysis with the subtractive clustering algorithm was used to train the network. The training speed and precision can be improved synchronously based on the optimization algorithm.Secondly, introduced the theory of particle swarm optimization, analyzed particle swarm optimization and its various improved algorithms. In order to solve the problem of the PSO getting trapped easily in a local optimal solution caused by the initial distribution of particle swarm and the search method, the improved particle swarm optimization(IPSO) was proposed. The proposed IPSO method consists of the modified evolutionary direction operator(MEDO) and the traditional PSO. The MEDO can make particle swarm escape from local optimum, and improves the global search solution capability of the traditional PSO.Thirdly, the disadvantage of RBF is that its ability of searching global solution is poor. The problem was caused by parameters space of local information setting. In order to improve the ablility of searching global solution of RBF the IPSO that has well real-time and accuracy was used to optimize RBF network. We could obtain the better network structure and parameters through applicating IPSO in RBF neural network, it improve the ability of searching global solution of RBF.Finally, the whole system for soft sensor of dynamic fluid was designed and realized based on the optimized RBFNN with IPSO used Matlab and VC.