基于瞬时转速和遗传算法的内燃机故障诊断方法

Fault Diagnosis Method for ICE Based on Instant Rotational Speed and Genetic Algorithm

内燃机瞬时转速信号中蕴含着大量有关机器运行的信息 ,非常适合用于内燃机的故障诊断。但瞬时转速中蕴藏的故障信息比较微弱 ,加上测量的随机性影响 ,必须提取多个故障特征进行综合诊断 ,而如何从中构造理想的特征参数 ,是一项非常困难的工作。对此 ,本文使用遗传算法对故障特征进行组合优化 ,构造最佳特征参数。以 6-1 3 5型柴油机单缸少量减油故障为例 ,从多个瞬时转速数据样本的时域、频域和波形参数中提取了 6个原始特征 ,利用遗传算法使这些参数在各种运算组合的过程中优胜劣汰 ,最后进化成为一个综合了几个原始特征的综合诊断参数 ,使用该参数进行故障状态的判别 ,提高了诊断的正确率 ,并简化了诊断过程。

The instant rotational speed signals of Internal Combustion Engine (ICE) con tain lot of information about machine states, and fit for fault diagnosis of IC E. The availability of modern Data Acquisition System (DAS) makes it very easy t o measure the instantaneous rotational speed of crankshaft of ICE, and it is our goal to detect some early faults of ICE using the instantaneous rotational spee d. But due to the faintness fault information in the speed data, the diagnosis p recision is very low. Therefore, it becomes necessary to extract multi-features for synthesis diagnosis, but at the same time the amount of information needed processing increases sharply, and some conflicts might exist in different featur es, so in a certain extent the diagnosis becomes more difficult. How to construc t the best diagnosis characteristic parameter from so many original features is a very difficult job. Genetic algorithm is a randomization searching method with fast development. The algorithm uses the ideas from natural selection and genetic mechanism in biol ogical universe for reference, and has higher performance than traditional searc h methods in intelligent search. It may efficiently find a solution to a problem in a large space of candidate solutions, and very fits for fault feature optimi zation. In this paper, genetic algorithm is used in features' combinatorial opti mization and construction of the best characteristic parameter. The simulated fa ult experiments are made at 6-135 type diesel, in which the fuel distributive va lue of one cylinder is reduced 15%. Through pre-processing to the instant rota tional speed data at the fault state, six original characteristic parameters are selected from time domain, frequency domain and waveform features. Using geneti c algorithm, a reconstructed characteristic parameter is obtained from almost in finite operation (add, subtract, multiply, divide and power) sets of original fe atures, which could effectively distinguish machine's status, enhance diagnosis' s accuracy, and simplify diagnosis procedure.