随机干扰下AUV推进器故障特征提取与融合

Feature extraction and fusion for thruster faults of AUV with random disturbance

针对水下传感器自身噪声等随机干扰影响水下机器人推进器故障诊断结果的准确性问题,为降低随机干扰影响,提出了基于小波近似分量提取故障特征、基于控制信号变化率提取故障特征以及带有归一化处理的特征融合方法.将速度信号进行小波分解,对分解后的尺度系数进行小波重构得到小波近似分量;对控制信号进行求导,得到控制信号变化率.基于修正贝叶斯算法,分别从小波近似分量和控制信号变化率中提取故障特征.基于证据理论对提取到的两个单一特征进行融合,并将融合结果进行归一化处理.水下机器人实验样机的水池实验结果验证了所提方法的有效性.

The correctness of fault diagnosis results for thrusters of AUV（autonomous underwater vehicle）was frequently influenced by random disturbance,which was caused by the internal noise of underwater sensors.To decrease the influence,two feature extraction methods that extracting fault feature from the wavelet approximate component of longitudinal velocity and from the changing rate of control voltage,and a feature fusion method with normalization were proposed.After the wavelet reconstruction of scale coefficients for wavelet decomposition of longitudinal velocity,the wavelet approximate component was obtained.After the derivation of control voltage,the changing rate was acquired.Two kinds of fault feature were extracted from the wavelet approximate component and the changing rate based on modified Bayes′classification algorithm separately.Following the feature fusion of the two kinds of fault feature based on evidence theory,the fusion result were normalized.The effectiveness of the proposed methods was verified by the experiments of AUV,which were carried out in the pool.