断丝状态下的钢丝绳故障诊断

Fault Diagnosis of Wire Rope in Broken Wire State

针对传统钢丝绳断丝损伤定量检测模型泛化性能不足的问题,提出一种基于极限学习机的钢丝绳断丝损伤故障诊断方法。采用距离可分离性判据对常用的特征参数进行可分性比较,并选取区分效果较佳的特征参数进行特征融合。实验结果表明,峰值、功率谱熵、小波奇异值熵的区分效果明显优于波宽、波形下面积;将多特征融合与极限学习机相结合的方法可以有效地对钢丝绳断丝损伤进行分类识别,比传统的BP神经网络具有更高的准确率和更短的训练时间。

Aiming at the problem of insufficient generalization performance of traditional quantitative identification model of wire rope breakage, a fault diagnosis method of wire rope breakage based on extreme learning machine was proposed. The distance separability criterion was used to compare the common feature parameters, and the characteristic parameters with better distinguishing effect were selected for feature fusion. The experimental results show that the peak, power spectrum entropy and wavelet singular value entropy are better than wave width and waveform area, and the method of combining multi-feature fusion with extreme learning machine can effectively classify and identify the broken wire damage of wire rope, which has higher accuracy and shorter training time than the traditional BP neural network.