论文部分内容阅读
对机匣振动加速度信号进行采集与分析是航空发动机振动故障诊断的重要方法。对振动信号进行处理与特征提取后,可以利用神经网络非线性映射的能力,对振动故障实现分类。利用转子-轴承-机匣耦合振动实验台模拟了5种风扇机匣的振动故障,从频谱的初步分析中并未能够实现对故障的准确判断。对振动数据进行了处理,提取了振动波形的频域与幅域参数,采用概率神经网络的方法实现单一故障的分类,并对不同参数所训练的网络进行了比较,检验了该诊断方法对于机匣振动故障的可行性。
It is an important method to diagnose the vibration of aeroengine vibration to collect and analyze the vibration acceleration signal of the casing. After the vibration signal is processed and the feature is extracted, the vibration fault can be classified according to the ability of nonlinear mapping of neural network. The rotor-bearing-casing coupling vibration experiment platform was used to simulate the vibration of five kinds of fan casing. The preliminary analysis of the spectrum did not make it possible to judge the fault accurately. The vibration data are processed, the frequency and amplitude parameters of the vibration waveform are extracted, the single fault classification is realized by the method of probabilistic neural network, and the networks trained by different parameters are compared, and the diagnostic method for the machine Feasibility of box vibration failure.