Research On Indirect Measurement Method Of Milling Force Based On Multi-signal Fusion Features

CNC machines tools are important processing equipment in the modern manufacturing industry,cutting force signal can effectively reflect the real-time cutting state of the machine tool,has a great significance in improving the intelligent monitoring level of machine tools and the optimization of processing parameters.The method of direct measurement of cutting force has many limitations in application,is difficult to promote the application in the actual production processing.Thus,this paper studies the indirect measurement method of milling force during machining.The specific work is as follows:Aiming at the problems that the existing cutting force prediction model has poor anti-interference ability and low prediction accuracy,start with model input features and milling force modeling methods,and proposed a milling force modeling method based on multi-signal fusion features.Aiming at the non-linear problem of signal response in the cutting process,based on the response mechanism of the signal,made the studied of the transmission process of the signal in the mechanical system,obtained the theoretical composition of each machining signal and the correlation characteristics with the cutting load.Built a milling machining experiment platform for related research.The noise interference in the signal will seriously affect the prediction accuracy of the milling force model.A study on the signal preprocessing method has been carried out in response to this problem.By analyzing the time-frequency domain characteristics of the original signal,calibrated the noise disturbance components in the signal,and through the wavelet packet filtering noise reduction,zero mean,correlation analysis and other processing methods to remove the interference components in the signal.Improve the real-time correspondence between signalsFinally,proposed a milling force modeling method based on multi-signal fusion features of stacked sparse self-encoding deep network(SSAE).Extracted signal fusion features though sparse self-encoder and constructed a deep network architecture for model training.In order to improve the prediction effect of the model,studied the influence of network structure and internal parameters on the training effect of the model,obtained an indirect measurement model of milling force with good comprehensive effect.By compared with the prediction effect of the model based on single-signal features,it is proved that the multi-input fusion feature can express the milling force more effectively.