The Implementation Of Artificial Neural Networks Control System Based On Fpga

In the recent years, Artificial Neural Networks is more widely used in the fields of intelligent control, pattern identification and in other areas. The traditional implementation of Artificial Neural Networks method—basing the general processor's software neural network control has two problems:first, the parallel speed and comuting speed are slow; the second,it is difficult to meet the cost of many embedded systems,power and size aspects of the stringent requirements. To this end, nuural network-specific hardware implementation methods are prososed. Neural netwok hardware implementation is devided into the following categories:Based on field programmable gate array (FPGA) implementation; based on digital signal processor (DSP) implementation; based on specific integrated circuit (ASIC) implementation. FPGA-based hardware in which neural network due to its unique high-speed parallel and can be repeated as the most suitable characteristics of flash hardware realization of neural network controller chip.This paper studies on the hardware implementation of neural network problems, comparing the feasibility of several neural network based on BP neural network as the selected neural network controller hardware implementation of neural network model. Limited resources on the FPGA as the hardware, FPGA realization of neural network model depends on the accuracy of the data in the FPGA side display and activation functions that the approximation accuracy. This topic is in the FPGA, data representation and activation functions SIMOID typical function approximation and the multiplier on the design of optimized FPGA design methods to achieve a rational allocation of limited hardware resources and utilization. Finally, the topics to Altera's FPGA chip as the core of BP neural network controller in the control of three-tank applications, and realize the accurate control of flow of three tanks.Paper shows the on deep study of the above work and we produce the corresponding test module, the simulation verifies the feasibility of the subject.