变分模态分解与极限梯度提升树融合的高速轴向柱塞泵空化等级识别

针对高速轴向柱塞泵在不同空化程度下故障特征不明显导致识别准确率低的问题,提出了一种变分模态分解和极限梯度提升树融合的识别方法。在不同空化等级下进行高速轴向柱塞泵空化试验,采集壳体的振动加速度信号,对信号采用变分模态分解方法并从中提取故障特征以构造特征数据集,最后利用极限梯度提升树进行空化等级的识别。为证明所提方法的抗噪性能,在测试集中加入了随机高斯白噪声。结果表明,加入不同信噪比的噪声后,该识别模型仍能准确地识别出高速轴向柱塞泵的空化等级。

Identification of Cavitation Intensity of High-speed Axial Piston Pumps Based on Variational Mode Decomposition and XGBoost

Aiming at the problems of implicit fault features and low identification accuracy of high-speed axial piston pumps, a recognition method based on Variational Mode Decomposition (VMD) and XGBoost is proposed in this paper. Cavitation experiments are carried out to collect the vibration signals on the pump housing under different levels of cavitation intensity. The vibration signals are decomposed by variational mode to obtain fault features and XGBoost model is used to identify cavitation intensity levels for high-speed axial piston pumps. In addition, random Gaussian white noise is added to the test dataset to prove the anti-noise performance of the proposed method. The results show that the proposed model can still accurately identify cavitation conditions on the noisy signals with different signal-to-noise ratios.