Research And Design Of Automatic Edge Seeking System Based On Neural Network

Along with the development of modern industrial manufacture, the integrated application of automatic control technology is taking more and more effects. Formed by strong mutual Permeating of disciplines such as pneumatic technology, hydraulic technology, transducer technology, PLC technology, the network and communication technology, etc, the electromechanical integrating technology has already nowadays become an important component of industrial science and technology.Edge seeker is a commonly used optical metrical instrument on measuring work piece in electromechanical field. This automatic edge seeking system receives signal from photoelectric reaction type edge seeker systematically originally, and outputs the catching signal to the digital display meter. This system contains the host computer and the subordinate machine. On the subordinate machine, it is set up by using 16 ultra low consumption processor MSP430F147 made by TI CO., and it is configured by using 4 keyboard input,4 LED display, serial port communication, numerical control potentiometer, Simulation voltage comparing modules. On the host computer, the software of special PC man-machine conversation and the communication software with the subordinate machine have been designed for training, writing, read parameter of neural network.This article describes the chip characteristics, On-Chip module, and every peripheral circuit module of this system in detailed introduction. BP neural network employs a kind of more neural network structure at present. It can approach arbitrary nonlinear function with the arbitrary precision, and it has good approaching performance, and by its simple structure, it is a kind of good performance neural network. The mathematics theory of BP neural network has been studied, and the basic principle, learning method, development trend of the neural network has been also introduced in detail, and the neural network algorithm source code of the system has been provided in this article. In order to adapt to different environments, the system obtains the reference voltage of the intersection between the bright district and the dark district by training neural network, and it is compared with the voltage signal from transducer to judge whether it has already reached the intersection.This article describes the experimental data of the original no neural network system under different working environment, and the experimental data of this system under the same working environment as the original system. It is proved by the experiment that there are better adaptability and accuracy on this system comparing the original system, and the adaptive ability to the working environment changing has been improved greatly.