论文部分内容阅读
通过分析柴油机在磨合期、不同摩托小时和拉缸等典型状态下的振动信号样本,计算出各类样本在幅域、时域和频域的特征参量,按照类别可分离性判据进行特征选择,寻找出能够代表发动机不同状态的有效特征参数,同时降低特征向量空间的维数,最后利用自组织特征映射神经网络(Self-Organizing Feature Map)进行发动机不同状态的分类。分析结果表明,SOFM能够对各类模式进行有效的分类,准确率达到92%以上。
By analyzing the vibration signal samples of diesel engine under the typical conditions of running-in period, different motorized hours and pulling cylinders, the characteristic parameters of all kinds of samples in the amplitude, time and frequency domain are calculated, and the characteristics are selected according to the category separability criterion , Find the effective characteristic parameters that can represent different states of the engine, and reduce the dimension of the eigenvector space. Finally, the self-organizing Feature Map is used to classify the different states of the engine. The analysis results show that SOFM can effectively classify all kinds of patterns with an accuracy of over 92%.