电路板卡故障红外成像智能诊断的分析与应用

摘要

针对电路板故障诊断精度低、误差大的问题，提出了一种红外智能诊断方法.首先利用红外成像技术计算出板卡固有温度成像图和故障温度成像图，根据温差成像得到异常区域初步定位；然后结合BP 神经网络中梯度下降算法局部搜索能力强、遗传算法全局搜索能力强的特点，形成2种算法于一体的ND GA-BP神经网络算法.引进“移民算子”增加种群多样性，构造新生代种群，将遗传操作和生物学“优胜劣汰”思想结合，发挥生物学的“顶端优势”，实现种群中最优个体的筛选.将 ND GA-BP 神经网络算法应用到电路板故障红外成像智能诊断，结果表明，其诊断精度提高了31．8％，误差降低了50．01％.%In order to solve the problems of low accuracy and large error in circuit board fault diagnosis,a method of infrared intelligent diagnosis was proposed.Firstly,infrared imaging technique was used to cal-culate the intrinsic temperature imaging and failure infrared imaging,and the initial identification of the abnormal areas was obtained according to the temperature differences between them.Then combining the strong local search ability of gradient descent in BP neural network algorithm with the strong global search ability of genetic algrithm(GA),a new double GA-BP(ND GA-BP)neural network algorithm was pro-posed.The algorithm introduced"immigration operator"to increase the diversity of population to con-struct new generation populations.Genetic manipulation was combined with biological"survival of the fit-test"to achieve the best individuals population screening.Both theory and practice prove that when the ND GA-BP neural network algorithm is applied to the infrared imaging intelligent diagnosis of circuit board fault,the diagnostic accuracy is increased by 31.8% and error is reduced by 50.01%.