Tool remaining life prediction based on edge computing and AT-LSTM recurrent neural network

Intelligent operation and maintenance of key components in industrial manufacturing through real-time condition monitoring and predictive maintenance technology can improve equipment operational efficiency. Addressing the tool life prediction issue in CNC machine tools, this paper proposes an edge computing attention mechanism long short-term memory (AT-LSTM) recurrent neural network model, utilizing spindle load data features during CNC machining processes to predict tool life. Edge controller embedded computing devices are developed to encapsulate the AT-LSTM model, achieving data acquisition and tool life prediction. Real-time data transmission to the cloud enables cloud-based training to update model parameters and firmware, which are then remotely downloaded to the edge controller. Experimental results demonstrate the reliability of the AT-LSTM model in predicting tool remaining useful life. The cloud-edge collaborative architecture enhances the flexibility and real-time capability of life prediction.