Mist-edge-fog-cloud computing system for geometric and thermal error prediction and compensation of worm gear machine tools based on ONT-GCN spatial–temporal model

The geometric precision of worm gears (WGs) determines the service performance and life of precision machine tools, indexing turntables and other equipment. The machining accuracy of worm gear machine tools (WGMTs) is the core to guarantee the geometric precision of WGs, and is greatly affected by the thermal and geometric errors. To improve the machining accuracy of WGMTs, the thermal and geometric errors should be controlled and compensated. But the control system has a poor real-time performance, and the synchronous control of the geometric and thermal errors cannot be currently achieved, and the thermal error model has a low prediction accuracy and low robustness. To make up for the above gap, a mist-edge-fog-cloud computing system is designed for the error prediction and compensation to relieve the bandwidth pressure of the industrial Internet. Moreover, a sensor network composed of multiple sensors is constructed to obtain the thermal information, and then the ordered neuron temporal-graph convolutional network (ONT-GCN) is proposed based on the ordered neuron-long short term memory network (ON-LSTMN) and graph convolutional network (GCN) for the first time to conduct the spatial and temporal modeling of the thermal error data. The interaction among multiple sensors is explicitly considered, and the dependence of the temporal information of the thermal error data on and spatial information of sensors is taken into account. Besides, to realize the error control, the mapping relationship between the tooth surface error and geometric-thermal errors is established. The error mapping model converts 51 geometric errors and 4 thermal errors into the spatial errors of the hob. Moreover, the sensitivity of errors is analyzed, and then the key error items that affect the geometric precision of the tooth surface are identified and compensated. The results show that the ONT-GCN is superior to traditional time-series modeling methods and that the mist-edge-fog-cloud computing system can effectively shorten the executing time compared with other system frameworks, and can improve the machining accuracy of WGMTs.