A dynamic meshing transmission dataset for manufacturing quality inspection of electric vehicle reducer gears

The NVH (Noise, Vibration, Harshness) performance of electric vehicle reducer gears directly affects the NVH level of the whole vehicle. However, the existing single gear quality detection methods based on tooth surface waviness are faced with two major challenges, which are unable to detect quickly and cannot fully characterize the real performance. This work introduces the first real industrial dataset for manufacturing quality inspection of electric vehicle reducer gears to solve these two challenges and provide data benchmarks. This dataset covers the dynamic meshing transmission data of five types of manufacturing offline gears (healthy gears, slight bump gears, leaky grinding gears, and two kinds of ghost order whine gears). And a variety of uncertain factors in the manufacturing process (different machine tools, different batches, different device control parameters, etc.) are introduced to avoid the influence of laboratory ideal conditions on signal characteristics. The data quality of this dataset was evaluated using a 1D-CNN classifier and benchmarked against WTGCM, MCC5-THU, and AGFD, where it exhibited superior performance.