Research On Online Evolutionary Fault Repair Of Circuits Based On Improved Simulated Annealing

Evolutionary hardware,as a new research field,integrates the thoughts of biological evolution in nature into the design of circuit evolution,enabling it to adjust its structure in time when encountering sudden failures in complex and harsh environments such as space and deep sea,thus realizing self-adaptation,selforganization and self-repair of circuits.Circuit evolutionary design is an important research direction of evolutionary hardware.The target circuit is constructed by combining evolutionary algorithms with programmable architecture,and programmable logic devices are used as the hardware carrier of circuit evolution to realize autonomous circuit evolution without human intervention.However,this technology is faced with stall effect and other problems,so it is very important to improve the evolutionary efficiency of programmable architecture and the computational efficiency of evolutionary algorithm in circuit evolutionary design.In addition,once the electronic equipment in the extremely harsh environment breaks down,it is difficult to achieve manual repair,which requires the electronic equipment to have a strong anti-interference ability in normal operation.The main research contents are as follows:(1)The programmable logic array based on feedforward neural network is applied to the architecture of virtual reconfigurable VRC,which further optimizes the programmable architecture,not only saves the internal resources of hardware,but also improves the scale and efficiency of circuit evolution.Based on the programmable architecture,a modularized matrix coding scheme is proposed,which can effectively map the circuit and improve the rate and convergence of circuit evolution.(2)An improved simulated annealing circuit evolutionary algorithm is proposed in this paper.In evolutionary algorithm,a neighborhood perturbation method based on modular matrix coding is proposed to increase the diversity of progeny through perturbation,so that the correct chromosome individuals can be found quickly and effectively,and the efficiency of circuit evolution can be improved.(3)Aiming at FPGA single point failure caused by single particle effect,a fault location and diagnosis technology based on feedforward neural network was proposed.This technology can locate the fault node accurately,and combined with the redundancy repair technology,can realize the location redundancy repair of the circuit fault node.The circuit online evolution design,fault location diagnosis and redundancy repair experiment are carried out on the FPGA platform.