CAJ | 학술논문

针对混合电路故障检测率低、诊断速度慢等问题,采用一种全新的诊断方法来应对这些故障诊断技术发展中的瓶颈问题.通过提取有效的故障特征和选取不同的核函数,支持向量机(SVM)能够很好的改善 BP 神经网络的收敛速度慢和容易陷入局部极小值等不足,具有更好的模式识别能力.以动态电流IDDT为参数的数/模混合正弦波信号发生器为例,建立了以支持向量机为基础的混合电路故障诊断模型,说明了支持向量机在小样本、非线性等情况下具有更强的泛化能力,得出了支持向量机方法的故障诊断率明显高于 BP 神经网络,为混合电路故障诊断提供一个新的方向.
침대혼합전로고장검측솔저、진단속도만등문제,채용일충전신적진단방법래응대저사고장진단기술발전중적병경문제.통과제취유효적고장특정화선취불동적핵함수,지지향량궤(SVM)능구흔호적개선 BP 신경망락적수렴속도만화용역함입국부겁소치등불족,구유경호적모식식별능력.이동태전류IDDT위삼수적수/모혼합정현파신호발생기위례,건립료이지지향량궤위기출적혼합전로고장진단모형,설명료지지향량궤재소양본、비선성등정황하구유경강적범화능력,득출료지지향량궤방법적고장진단솔명현고우 BP 신경망락,위혼합전로고장진단제공일개신적방향.
Mixed circuit has low fault detection rate and slow fault diagnosis speed. Here,the paper presents a novel fault diagnosis technology to settle the bottleneck in the development of fault diagnosis technologies. By introducing the extraction of fault feature and optional kernel functions,support vector machine(SVM)can rationally overcome the problems of slow conver?gence speed and local minimum,etc,and has better performance of pattern recognition. An example of sine wave signal genera?tor with dynamic current IDDT as the parameter is utilized to establish the mixed circuit fault diagnosis model based on SVM, which explains that SVM has strong generalization ability of learning method in small samples and nonlinear cases. Results show that the proposed method based on SVM has higher fault diagnosis rate than BP neural network method,which is a new orienta?tion in the fault diagnosis of mixed circuit.