基于相关向量机的小样本故障诊断预测研究

发布时间：2018-06-09 16:58

本文选题：相关向量机 + 小样本　；参考：《西安工程大学》2016年硕士论文

【摘要】：目前常用的一些基本的故障诊断、故障预测方法都将大样本数据作为基础,但在实际问题中常常能得到的故障数据都属于小样本类型.传统的故障诊断、故障预测方法已不适于用来解决小样本类型的故障问题.相关向量机(Relevance Vector Machine,简称RVM)是新提出的以支持向量机(Support Vector Machine,简称SVM)为基础的模型,该模型更适应于解决小样本问题.已经被应用于语音及图像处理、医学诊断、模式分类等很多领域.本文针对RVM在小样本故障诊断、故障预测中的应用问题展开研究,主要包含以下方面:(1)分析故障诊断、故障预测的研究现状;(2)RVM理论基础及其在故障诊断、故障预测中的可行性;(3)针对RVM中核函数选择的盲目性及核函数中相关参数对RVM性能的影响,通过组合高斯核函数和柯西核函数构造混合核函数,利用布谷鸟搜索算法(Cuckoo Search Algorithm,简称CS)对混合核函数的参数进行优化选择,建立基于布谷鸟算法的混合核函数相关向量机模型(CS-RVM),并通过仿真测试验证模型的有效性;(4)将CS-RVM应用到航空发动机的气路诊断和柴油机的油路诊断中,通过与RVM和基于差分进化(Differential Evolution,简称DE)的混合核函数相关向量机模型(DE-RVM)比较,充分说明该方法在小样本故障诊断、故障预测中的可行性.
[Abstract]:At present, some basic fault diagnosis methods are based on large sample data, but the fault data often obtained in practical problems belong to small sample types. Traditional fault diagnosis and fault prediction methods are not suitable to solve the problem of small sample types of faults. Correlation vector machine (RVM) is a new model based on support vector machine support Vector Machine (SVM), which is more suitable for solving small sample problem. It has been used in many fields, such as speech and image processing, medical diagnosis, pattern classification and so on. In this paper, the application of RVM in small sample fault diagnosis and fault prediction is studied, which mainly includes the following aspects: 1) Analysis of fault diagnosis, present situation of research on fault prediction and its theoretical basis and its application in fault diagnosis. According to the blindness of kernel function selection in RVM and the influence of relevant parameters in kernel function on the performance of RVM, the hybrid kernel function is constructed by combining Gao Si kernel function and Cauchy kernel function. Cuckoo search algorithm (CSA) is used to optimize the selection of parameters of hybrid kernel function. A hybrid kernel function correlation vector machine model based on cuckoo algorithm is established, and the validity of the model is verified by simulation tests. The CS-RVM is applied to the gas path diagnosis of aero-engine and the oil path diagnosis of diesel engine. Compared with RVM and DE-RVM-based hybrid kernel function correlation vector machine model based on differential evolution (DEV), the feasibility of this method in small sample fault diagnosis and fault prediction is fully demonstrated.
【学位授予单位】：西安工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：V263.6;TP18

【参考文献】