The intelligent control terminal with SCM based on constructive neural networks

Abstract

The constructive neural networks(CNN) based on the FP(Forward Propagation) algorithms can be used to realize an intelligent machine with the SCM(single-chip microcomputer). High performance of error-toleration and error-correction can be found when it communicated with the others .The system has the ability of self-learning and expansion of knowledge on line, it can also run safely by setting the rejection schemes. Index Terms—constructive neural networks(CNN), SCM( single-chip microcomputer), intelligent terminal. I. INTRODUCTION he control objects in modern industry are mostly complex, nonlinear, gray and long-time-lag system, it is difficult to control and establish exact model. In order to get satisfactory control results, people try to construct real-time control system with neural networks, and a series of production (4,5,6) have been made. But, with the development of modern industry, people hope that the intelligent control terminal can recognize not only the existing control model, but also the newly process models. The control terminal can evolve through the circle of application, learning, and more advanced application. But the existing neural networks are difficult to learning and evolution on line. In this paper, the constructive neural networks to be used in the intelligent control terminal with the SCM and CNN can not only learn on line but also perform the error-toleration and error-correction. The system can be safely and effectively used in the filed of industrial control under severe condition. You can also see the following from this paper: (1) The CNN enable the control terminal with well self-learning function. It's learning complexity is only O(n), more less than other neural networks, it can be realized by SCM. (2) When the control terminal is used in pattern recognizing or communicating with the others, the CNN can automatically perform the error-toleration and the error-correction. CNN can perform top intelligent control.