基于粗集约简预处理的模拟电路故障诊断

基于人工神经网络的智能故障诊断系统作为人工智能技术在模拟电路故障诊断领域的应用,在实践中取得了一定的成效。但将神经网络用于电路故障诊断时,知识具有隐含性,可解释性差,且对输入数据的冗余难以约简,获得每一个训练样本都要进行一次测试或模拟计算,样本花费代价很大,而粗糙集理论作为处理不确定、不完整、不精确知识的有力工具,具有强大的知识约简和定性分析能力。因此,该文提出了对模拟电路的故障特征进行粗集约简预处理研究的智能诊断方法。并举例说明诊断系统的具体实现方法,仿真结果表明:在相同的精度要求下,该算法的训练时间远小于普通的进化神经网络,对模拟电路的故障诊断有一定的实际意义。

Rough set reduction pretreatment analogous circuit fault diagnosis

Based on the artificial neural network intelligent fault diagnosis system took the artificial intelligence technology in the analog circuit fault diagnosis field,has obtained better result in the practice.However,the neural network is used for circuit fault diagnosis,implicit knowledge may be poor explanation,and the redundancy of input data reduction difficult.And rough as the strong tool dealing with uncertain,incomplete,inaccuracy knowledge,has strong knowledge reduction and qualitative analysis.Therefore,this paper has proposed rough set reduction pretreatment to the fault characteristic of analog circuit intelligent diagnosis method.Example circuits are provided to illustrate the proposed method.The simulation result indicated that,under the same precision request,this algorithm training time far is smaller than the ordinary evolution neural network.