基于V-detector算法的滚动轴承故障诊断方法

针对实值阴性选择(RNS)算法的检测器尺寸不能自适应变化的问题,提出基于变检测半径的RNS算法(V detector)的轴承故障诊断方法.将计算轴承振动信号局域波分解后各基本模式分量的关联维数作为特征向量,并根据故障模式将其划分为多个自体样本集,采用V detector算法训练多个检测器集,用其对轴承故障进行诊断.结果表明：自体半径过小则误诊率高,自体半径过大则检测器灵敏度低,这都将导致准确率减小；覆盖率越高,则准确率越高、计算花费越大,当覆盖率≥95%时,覆盖率对准确率的影响远小于其对计算花费的影响；相对于基于RNS的诊断方法,V detector算法具有同样高的准确率,且计算花费显著减小、稳定性更高,可有效地识别轴承故障.

Fault diagnosis approach for rolling bearing based on V-detector algorithm

Aimed at the detectors size of real valued negative selection  (RNS)  algorithm is in capable  of adaptive adjustment, a fault diagnosis method for rolling bearing was proposed based on RNS with variable sized detectors (V detector). Firstly, correlation dimensions of intrinsic mode functions decomposed from bearing vibration signal were calculated as feature vectors. Secondly, the feature vectors were divided into different self samples sets according to fault modes, and then V detector algorithm was used to train and generate corresponding detector sets, which used for fault diagnosis. The experimental results show that the misdiagnosis rate will be high if self radius is too small, while the detectors sensitivity will be low conversely, either of which results in low diagnostic accuracy; higher estimated coverage corresponds to higher accuracy and time cost, and the estimated coverage has far less effects on the accuracy than on the time cost when it is higher than 95%. Compared with the diagnosis method based on RNS, the proposed method has far lower time cost and higher stability with the same high accuracy, and can identify the rolling bearing fault patterns effectively.