基于复小波包分形理论的爬壁机器人故障检测

通过研究爬壁式机器人的控制和运动特征，提出一种基于复小波包分形理论的故障检测方法．利用复小波包的平移不变性，将爬壁式机器人传感器输出信号分解成独立的频谱，并进行阀值处理和重构，从而有效去除高频噪音并提取故障的特征频率；依据信号分形维数的多尺度不变性，在嵌入维数空间，采用维数最大距离法，确定复小波包域故障信号的关联雏数．仿真实验表明，爬壁式机器人在各种异常工作模式下的故障信号关联维数能比较真实地反映其故障状态空间，同时也验证了故障信号的关联维数低于正常信号的关联维数作为故障发生与否的定量判据的正确性．

Fault Detection for Wall-Climbing Robot Using Complex Wavelet Packets Transform and Fractal Theory

A novel fault detection method for Walbclimbing robot is presented based on complex wavelet packets analysis and fraetal theory. It employs complex wavelet packets transform to obtain the real and imaginary parts complex wavelet coefficients of the Wall-climbing robot sensor output signals. The high frequency noise in the output signals is excluded and the characteristic frequency of fault signal is abstracted via shrinking and reconstructing the complex wavelet coefficients by using hard-thresholding method. Furthermore,The multi-scale spectrum correlation dimensions of the fault signals are computed out by using fraetal theory. It employs the dimension furthest distance method to define the fault sensitive dimension of system state at a series of fixed dimension, so the nonstationary characteristic of the noise fault signals is picked up when the some faults happen. The simulation experiment shows that the characteristic space of the noise fault signals is accord with the fault state space of Wall-Climbing Robot well. The results also show that the correlation dimension of fault signal is bigger than that of normal signal. This conclusion is a good quantitative evidence to judge whether the fault occurs or not.