Robot Joint Cycloidal Gear Tooth Fault Detection in Robot Working Conditions Based on Piecewise Mean Difference

In robot joints or rotary vector (RV) reducers, where cycloidal gears perform complex low-speed rotating motions involving reciprocating variable speed and incomplete rotation, conventional methods typically fail to detect their tooth fault. To address this issue, a piecewise mean difference (PMD) method is proposed in this article. First, rotary encoder outputs of the RV reducer are picked up and converted into the instantaneous angular speed (IAS) signals, which are separated according to the rotation direction justified by the lead or lag of the phases of the encoder outputs. Second, the PMD is used to adaptively truncate data blocks related to the approximately stable speed to suppress the influence of the trend term and amplitude modulation caused by the robot working conditions. Third, the cycloidal gear tooth (CGT) fault-related IAS components are extracted from the IAS signal by the rotation domain average (RDA), while the demean operation is implemented on the fault-related IAS components. Finally, the diversity entropy of the fault-related IAS components is calculated for the CGT fault detection. Experiments were carried out on an RV reducer test rig. Corresponding results show that the proposed method can effectively detect the tooth fault of the cycloidal gear in the robot joint under variable-speed rotating conditions.