The Prediction Of Tool Wear Condition And RUL Based On Multi-sensor Information Fusion

The cutting tool is known as an executor of the cutting machining,whose real-time condition is able to directly affect the machining quality,precision and efficiency of the components,and even seriously lead to the fault of machining system.Moreover,the realtime determination of tool wear condition and the prediction of remaining useful life can ensure the quality of the parts produced by machine tools,save machining cost and labor cost,as well as promote the rapid development of intelligent machining technology.Based on the multi-sensor information fusion technology,a method for real-time prediction of tool wear and remaining useful life is proposed in this paper.Its main contents are as follows:Firstly,the experimental platform is set up using an acceleration sensor and a dynamometer in order to collect the vibration signals and cutting force signals generated during machine tool processing.And then these signals are preprocessed to support the subsequent tests,including signal analysis and state recognition.Secondly,a step-by-step method of signal de-nosing and feature extraction is proposed through time domain analysis,frequency domain analysis,and time frequency domain analysis of the monitoring signals.With contrast experiments,the fast independent component analysis which ensures the characteristic information of the monitoring signal clear,is used to achieve signal noise reduction,and the wavelet packet frequency band energy analysis is used to achieve signal feature extraction in full states.After that,an improved Elman neural network model was established to estimate tool wear and remaining useful life according to the multi-sensor information fusion technology.To be exact,by optimizing the learning algorithm and network structure,and using genetic algorithm to initialize the weights and thresholds of the Elman neural network,the learning ability and convergence speed of the network are improved.Finally,the tool wear condition and remaining useful life are able to be predicted by using this improved Elman neural network model.Compared with the experimental results,that of neural network prediction model has fast convergence speed and high estimation accuracy,which effectively guarantees the prediction effect.Meanwhile,the all-time tool wear and remaining useful life can be continuously estimated.Overall,a method for simultaneous on-line prediction of tool wear condition and remaining useful life is proposed by using modern sensor technology,computer technology and signal processing technology.The research results in this thesis have very important academic significance and practical value for promoting the rapid development of adaptive intelligent machining technology.