点火线圈绝缘故障脉冲分类算法  

Ignition coil insulation fault pulse classification algorithm

作　　者：李木天[1] 王启松[1] 赵永平[1] 

机构地区：[1]哈尔滨工业大学自动化测试与控制系,哈尔滨150001

出　　处：《仪器仪表学报》2013年第5期1121-1130,共10页Chinese Journal of Scientific Instrument

基　　金：黑龙江省自然科学基金(F200813);中央高校基本科研业务费专项资金(HIT.NSRIF.2013015)资助项目

摘　　要：点火线圈的绝缘故障使发动机扭矩下降、动力不足,影响汽车的驾驶感受与安全性。绝缘故障发生时,点火线圈初级电流和次级电压中0.5～2.5μs的绝缘故障脉冲中包含故障源的信息。本文提出一种点火线圈绝缘故障脉冲分类算法,根据绝缘故障脉冲特征确定故障类型。首先提取初级电流和次级电压波形中反映各类故障特点的11个原始特征量,用主成分分析法去除特征量间的重叠信息,简化特征量。对前5个主分量进行统一尺度的缩放,减少极端数据对分类结果的影响,使用支持向量机对点火线圈的实际故障进行分类。实验结果表明:相对于BP神经网络算法和标准支持向量机算法,本文的分类算法迭代次数少、时间短。在相同的信噪比环境下,本文算法的分类准确度更高,均方根误差更小。Ignition coil insulation fault causes torque descent and power insufficiency, which affects vehicle driving experience and safety. When the insulation fault occurs, the pulses of 0. 5 - 2.5 μs in the primary current and second- ary voltage contain the information of the fault source. The paper proposes an ignition coil insulation fault pulse classi- fication algorithm, which determines the fault type based on the pulse characteristics. Firstly, 11 original features of primary current and secondary voltage pulse waveforms that reflect the characteristics of various fault types are extrac- ted. Then the principal component analysis is used to remove the overlap information among the original features and simplify the features. The first 5 principal components are scaled using unified scale to reduce the influence of ex- treme data on classification results. Finally support vector machine classifier is used to classify the actual faults of the ignition coil. The experiment results show that compared with BP neural network algorithm and standard support vec- tor machine algorithm, the proposed classification algorithm requires less number of iterations and calculation time; and has higher accuracy and smaller root mean square error under the same signal to noise ratio.

关 键 词：点火线圈 绝缘故障 分类 主成分分析 支持向量机 

分 类 号：TN911.7[电子电信—通信与信息系统]