Research on a rapid pre-delivery inspection strategy for harmonic reducers

Harmonic reducers, which serve as integral components in the kinetic transmission of industrial robots, significantly influence the comprehensive performance of the equipment. Before integration into the distribution network, meticulous quality assurance is imperative. A novel strategy for rapid pre-delivery inspection (PDI) of harmonic reducers is proposed to address the inefficiency in detection caused by the complexity of the assembly process. The core concept centers around integrating variational mode decomposition (VMD) and multi-scale one-dimensional convolutional neural network (MS-1DCNN) based on acoustic pressure signals. Initially, a test rig is devised to capture one-dimensional raw acoustic pressure signals from harmonic reducers. Furthermore, sparrow search algorithm is employed to determine the optimal parameter combination [ K, α] for VMD. The acoustic pressure signals can be decomposed into intrinsic mode functions (IMFs) of diverse characteristic scales. Subsequently, effective IMFs are selected by Pearson correlation coefficient, and all such components are superimposed to form the reconstructed signals. Finally, the MS-1DCNN discriminant model is established for rapid PDI of harmonic reducers. The experimental results show that VMD-MS-1DCNN achieves an outstanding recognition accuracy of 99.45% in identifying harmonic reducers, offering a unique perspective and promising industrial applications.