Stability Analysis Of End Milling Process And Research On On-line Detection Method Of Cutting Chatter

With the continuous development of manufacturing technology, the problem of machining stability, especially cutting chatter has become one of the main factors which restrict the improvement of machining efficiency. Currently, many researchers have studied the generation mechanism and on-line detection technique of cutting chatter, as well as calculation method of stability Lobes. However, because of the complexity of machine tool and machining processes, the guiding role of existing research results on actual machining processes is limited. Therefore, based on the summary of research status of related fields, with end-milling process as the research objective, stability analysis of end milling process and research on on-line detection technique of cutting chatter is fouced in this thesis. The prediction method of tool point frequency response function (FRF) is focused, which is necessary for stability lobe diagram. Moreover, the identification method of parameters at the spindle-holder and holder-tool interfaces is studied. And the effects of the connection parameters at the two interfaces on stability lobe diagram are invesgated. The deterioration trend of relative dynamic compliance of machine tool is studied, and the effects of the deterioration trend on the stability lobe diagram are also analyzed. And an on-line detection method of cutting chatter is proposed.With spindle-holder-tool system as the research objective, the transfer matrix of Timoshenko beam and elastic support are given. And the prediction method of the end FRPs of spindle, holder and tool is provided. The receptane method of spindle, holder and tool is given, and the prediction model of tool point FRF is proposed. Compared with modal superposition method, the maximum error of the first seven natural frequencies derived from this prediction model proposed in the dissertation is decreased by35%, and the calculation efficiency improves by50%.Research shows that the connection parameters at the spindle-holder and holder-tool interfaces are one of the main factors which affect the prediction precision of machine tool tool point FRF. In order to improve the prediction precion of tool point FRF, the end point FRFs of holder and tool are obtained based on Timoshenko beam theories and transfer matrix method. Based on receptance coupling substructure analysis (RCSA) and improved particle swarm optimization (PSO), the identification method of connection parameters at the spindle-holder and holder-tool interfaces is focused on. Experimental results show that the identification model of interface parameters proposed in the dissertation has high precision and universality, which means that the model can be employed in the parameter identification of fixed interfaces extensively.Based on the four-freedom-degree dynamic model of end milling process, the calculation formulas of machine tool relative dynamic compliance are given. The relative excitation experimental scheme is put forward, and the evaluation criterions of relative dynamic compliances are proposed. Based on Generalized Hidden Markov Model (GHMM) and gravity method, the deterioration trend of relative dynamic compliances of machine tool is predicted. Moreover, the effects of the deterioration trend of relative dynamic compliances on the stability lobe are also analyzed. Research results show that this prediction model can solve the problem of small sample effectively. This prediction model has high prediction precision, which can provide dynamic stability lobe diagrams for the end milling processes.Based on wavelet packet decomposition and singular spectrum analysis, the feature extraction method of cutting chatter is put forward. Combining PSO and local neighbourhood search algorithms, the improved BP Artificial Neural Network (ANN) is proposed, which is used to recognize cutting states, and the on-line detection of cutting chatter is realized. Research results show that the method proposed in this dissertation can identify the stable, transition and chatter cutting states effectively, and the recognition ratio reaches95%.