微型电机声学质量检测小样本机器学习方法研究

Research on Small Sample Machine Learning Method for Acoustic Quality Detection of Micro Motors

为了解决微型电机声学质量检测人工手摸及听诊方法存在的主观误判率高、效率低下等问题,同时兼顾检测结果准确率和检测模型构建的快速性,提出了一种小样本机器学习检测方法,其根据微型电机传动链物理模型进行多维声学故障特征提取,在此基础上,采用粒子群优化算法对支持向量机这种小样本学习方法的核心参数进行优化,从而提高模型判别的准确率。试验结果表明,该方法能够有效判别微型电机异常振动和声音,准确率达到95%以上。

In order to solve the problems of high subjective misjudgment rate and low efficiency in manual hand touch and auscultation methods for acoustic quality detection of micro motors,while taking into account the accuracy of detection results and the fast construction of detection models,a small sample machine learning detection method was proposed.Based on the physical model of micro motor transmission chain,multi-dimensional acoustic fault features were extracted,particle swarm optimization was used to optimize the core parameters of support vector machine,a small sample learning method,so as to improve the accuracy of model discrimination.The experimental results show that this method can effectively distinguish abnormal vibration and sound of micro motors,with an accuracy rate of over 95%.