基于阶次分析的发动机故障预测与识别

发布时间：2018-04-24 06:30

本文选题：发动机 + 阶次分析　；参考：《大连理工大学》2016年硕士论文

【摘要】：汽车在国民生活中扮演越来越重要的角色,发动机作为汽车的心脏,其故障在整车故障中占据较高比例。为了提高发动机安全性,提前预测故障,基于振动信号的发动机故障预测技术迅速发展起来。而发动机振动信号具有非平稳性、非线性等特点,传统的频域分析会产生频率模糊现象,因此催生了基于角域的阶次分析。耐久性试验是发动机研发中重要的一环,某企业致力于发动机耐久性试验,购进德尔塔分析仪用于检测发动机振动。企业希望能够进行早期故障预测,在故障初发时预警,避免发动机永久损坏造成重大损失。本文的研究即基于德尔塔分析仪检测发动机振动生成的阶次信号。本文首先分析了发动机、德尔塔分析仪的工作原理及阶次分析的原理等背景知识,根据当前对故障预测存在的弊端及企业需求进行了剖析,设计了一套完整的解决方案。(1)先分析故障原理及相应的故障频谱图,寻找故障可能对应的阶次,进行数据预处理。(2)将处理过的数据使用主成分分析法进行降维,进一步压缩数据,消除冗余,避免“维数灾难”。(3)采用模糊C-均值和支持向量机对降维后的数据进行分类,结果证明,支持向量机这套方案能够达到预期效果。(4)采用python按企业要求编写故障预测软件,综合以上方法。本文提出的解决方案能够及时对故障进行预测。最后韵预测软件已经应用于企业耐久性试验当中。
[Abstract]:Automobile plays a more and more important role in national life. Engine, as the heart of automobile, has a high proportion of faults in the whole vehicle. In order to improve engine safety and predict fault ahead of time, engine fault prediction technology based on vibration signal has developed rapidly. But the vibration signal of engine has the characteristics of non-stationary and nonlinear. The traditional frequency domain analysis will produce frequency ambiguity phenomenon, so it gives birth to order analysis based on angle domain. Durability test is an important part of engine research and development. A certain enterprise devotes to engine durability test and buys Delta Analyzer to detect engine vibration. Enterprises hope to be able to make early fault prediction and early warning to avoid serious damage caused by permanent damage to the engine. The research of this paper is based on Delta Analyzer to detect the order signal generated by engine vibration. In this paper, the working principle of engine, Delta analyzer and the principle of order analysis are analyzed, and the disadvantages of fault prediction and the demand of the enterprise are analyzed. A complete solution is designed. (1) the principle of the fault and the corresponding fault spectrum are analyzed first, and the order of the possible fault is found, and the data preprocessing is carried out. 2) the processed data are reduced by principal component analysis. Further compressing data, eliminating redundancy, avoiding "dimensionality disaster". 3) using fuzzy Cmean and support vector machine to classify the data after dimensionality reduction, the result shows that, Support Vector Machine (SVM) is able to achieve the desired results. (4) python is used to compile fault prediction software according to the requirements of enterprises, and the above methods are synthesized. The solution proposed in this paper can predict the fault in time. The final rhyme prediction software has been applied to enterprise durability test.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U472

【相似文献】