基于遗传算法的电路故障诊断超参数优化算法框架

Hyper-parameters optimization framework of circuits fault diagnosis based on genetic algorithm

针对基于SVM的模拟电路故障诊断中诊断参数的调节是通过试凑法或按照全局最优的原则确定的,没有考虑实际诊断要求,无法进行各诊断环节参数同时调整优化的现状。提出一种适应度模型用于遗传算法参数寻优,把实际电路诊断要求量化成参数指标引入模拟电路故障诊断的优劣评估中;建立了基于遗传算法的电路诊断模型参数闭环寻优框架,对诊断系统的各部分参数优化进行整体度量,并分析了参数搜索算法的收敛性。通过实例诊断分析了闭环故障诊断参数寻优框架下各部分的参数制定对决策的影响,说明了建立的闭环故障诊断模型参数寻优框架和搜索算法的有效性和实用性。

Diagnostic parameters of analog circuits fault diagnosis based on SVM are adjusted in accordance with the principle to determine the global optimum or by trial.Parameter adjustment is not considered practical diagnostic system diagnostic requirements.It can not be part of various diagnostic parameters simultaneously adjust and optimize.The results are not satisfactory.The paper presents a model of fitness function for genetic algorithm parameter optimization,it will convert the actual circuit diagnosis requires fitness indicators in the evaluation results of analog circuit fault diagnosis.In this paper,a circuit diagnosis framework for closed-loop model parameters optimization based on genetic algorithm is presented.It＇s all part of the system parameters to optimize simultaneously,and analyzes the convergence of the algorithm parameter search.By example the closed-loop fault diagnosis parameter optimization framework developed under the parameters of the various parts of the impact on decision-making.This article describes the establishment of closed-loop fault diagnosis model parameter optimization framework and the search algorithm is effective and practical.