Research On Tool Health Monitoring Technology Based On Multi-source Heterogeneous Industrial Big Data

Numerical control system is an important equipment in intelligent manufacturing system,towards to the direction of unmanned,integrated and automatic.Tool health monitoring technology plays an important role in improving the utilization rate of NC machine tools,ensuring machining quality and reducing cutting cost.With the rapid development of automation,information technology,data acquisition technology and sensing technology,tool health monitoring system can easily obtain massive multisource heterogeneous data comtaining much wear imformation.Traditional methods mainly depends on the signal feature extraction and feature selection to achieve the ecognition of tool state,which not only relies on a large amount of expert knowledge and strong signal analysis theory,but also takes a lot of time and energy,and is easily affected by subjective factors.In this paper,deep learning algorithm is proposed to process tool data to make full use of multi-sensor information of tool health monitoring system and implement tool multi-source heterogeneous big data fusion.First,this paper designed cutting tool signal acquisition and wear quantity measurement experiment.Based on analyzing the characteristics of different monitoring signals of the tool,this paper selects the multi-source heterogeneous signals,which are combined with sound signals,vibration signals and infrared thermal images.The massive monitoring data of the whole cutting process is collected.Then,multi-source heterogeneous big data was fused at desional level in this paper.Deep Belief Network(DBN)is used to directly process structured sound and vibration signals to achieve extract wear features adaptively while the Convolutional Neural Network(CNN),which is good at processing image information,completes the training process of large amounts of infrared thermal images and predicting new images.Finally,the decision level fusion of DBN and CNN neural network,combining the obfuscation matrix of tool wear state,is realized based on D-S evidence theory to fully integrate multi-sensor information and achieve the expected goal of tool health monitoring.The deep learning fusion method in this paper provides a new idea for tool health monitoring.Finally,a tool health monitoring system was established based on JavaWeb,which can realize real-time call of SQLServer database data and analyse data to implement tool health monitoring.