Artifical Neural Network Algorithms Improving And Application In The Hydraulic Measuring Technique

In the thesis, it is researched that artiflcal neural network algorithms is improved and applied to the hydraulic measuring technique. According to present artifical neural network mainly used to deal with the problem of complex plant modeling, prediction and control, it's algorithms has the larger time complexity, larger space complexity, slow learning speed, low accuracy capability of generalization and bad properties of convergence. The algorithms can't be directly applied to the hydraulic measuring technique, therefor, some improving new neural network algorithms is proposed, and these algorithms are used to the measuring technique for improving the precision and intelligences level of hydraulic measuring system. The main contents are as follows.(l)The particle swarm optimization(PSO) algorithm, is used to train neural network to solve the drawbacks of BP algorithms which is local minimum and slow convergence.(2)A Single Input-Single Output of Cerebllar Model Articulation Controller CMAC Algorithm (SISO-CMAC) is presented to solve the problem of low generalization precision and effected on convergence by hash coding. Furthermore, the two-dimensional and multi-dimensional nonlinear functions are approximated by using many SISO-CMAC.(3)Because the characteristic of sensors is non-ideal and has serious static nonlinear or slow dynamic response, there are some errors in measurement result. In order to solve these problem, it is researched that neural network is used to compensate the error in this thesis. A compensating block designed by neural network is cascaded to output of the sensor, the characteristic of sensors is improved, and the precision of measurement result is increased.(4)In the thesis, a new method that enhances the accuracy of grating transducer signals by using neural networks is studies. Furthermore, a method that combines FIR and median filter is introduced. The method can reduce random error of grating measuring system effectively.(5)The method which sensor error is compensated with the artifical neural network algorithms is applied to the hydraulic measurement technique.