基于BP神经网络的钛合金加工切削力预测模型研究

发布时间：2018-11-27 19:05

【摘要】：钛合金以其良好的耐热性、耐腐蚀性,以及优异的力学性能等,广泛应用于航空航天、医疗、化工等领域,具有广阔的科研空间和应用前景。但是由于钛合金的加工性较差,在加工过程中存在切削温度高、刀具磨损严重,加工效率低下等问题。在钛合金切削过程中,切削力是一个非常重要的物理量,它对刀具寿命、加工质量等都有显著影响,因此切削力的研究对提高刀具寿命,改善加工质量,提高加工效率具有非常重要的意义。本文选取钛合金Ti6Al4V作为研究对象,研究分析该型号钛合金切削力的预测。本文主要内容包括:首先,根据金属切削原理,采用有限元分析方法结合ABAQUS建立了钛合金切削仿真模型。通过将Kistler9257B三向车削测力仪测得的主切削力与通过仿真模型获得的主切削力进行对比,验证了仿真模型的正确性。其次,基于python语言通过ABAQUS二次开发建立了钛合金切削仿真参数化模型。该参数化模型不仅可以提高建模的准确性,而且能够省去大量的重复建模时间。然后设计了关于进给量、切削速度、刀具前角的单因素仿真试验,探究三者对主切削力的影响规律。再次,建立了主切削力的BP神经网络预测模型。首先设计三因素三水平的全因素试验方案,利用切削仿真参数化模型获取样本数据。对神经网络内部各参数进行设计,进行神经网络预测模型的训练与验证。最后,基于响应曲面法建立了主切削力的公式预测模型,并通过显著性检验验证了公式模型对样本数据的拟合效果,数据表明本文所建公式模型具有良好的拟合效果,可用于主切削力的预测。输入相同的切削参数,利用公式预测模型可以获得主切削力的预测结果,通过BP神经网络模型也可以获得相应的主切削力预测结果。将两组预测结果分别与通过切削仿真参数化模型获得的主切削力进行对比,可以得到两种预测模型的预测误差。结果表明:BP神经网络预测模型的预测精度高于公式模型的预测精度,从而说明本文所建的BP神经网络切削力预测模型具有一定的优越性,可以对生产实际提供指导。
[Abstract]:Titanium alloys are widely used in aerospace, medical, chemical and other fields for their good heat resistance, corrosion resistance and excellent mechanical properties. However, due to the poor machinability of titanium alloy, the cutting temperature is high, the tool wear is serious, and the machining efficiency is low. In the process of titanium alloy cutting, cutting force is a very important physical quantity, which has a remarkable influence on tool life and machining quality, so the study of cutting force can improve tool life and machining quality. It is very important to improve the processing efficiency. In this paper, titanium alloy Ti6Al4V is selected as the research object, and the prediction of cutting force of the titanium alloy is studied and analyzed. The main contents of this paper are as follows: firstly, according to the principle of metal cutting, the simulation model of titanium alloy cutting is established by using finite element analysis method and ABAQUS. The correctness of the simulation model is verified by comparing the main cutting force measured by the Kistler9257B three-way turning dynamometer with the main cutting force obtained by the simulation model. Secondly, the parameterized model of titanium alloy cutting simulation is established by ABAQUS secondary development based on python language. The parameterized model can not only improve the accuracy of modeling, but also save a lot of repeated modeling time. Then, a single factor simulation experiment on feed rate, cutting speed and cutting tool front angle is designed to explore the influence of the three factors on the main cutting force. Thirdly, the main cutting force prediction model based on BP neural network is established. First, a three-factor and three-level full-factor test scheme is designed, and the sample data are obtained by using the cutting simulation parameterized model. The parameters of neural network are designed, and the prediction model of neural network is trained and verified. Finally, based on the response surface method, the formula prediction model of the main cutting force is established, and the fitting effect of the formula model to the sample data is verified by the significance test. The data show that the formula model established in this paper has a good fitting effect. It can be used to predict the main cutting force. With the input of the same cutting parameters, the prediction results of the main cutting force can be obtained by using the formula prediction model, and the corresponding main cutting force prediction results can be obtained by the BP neural network model. By comparing the two sets of prediction results with the main cutting forces obtained by the parameterized cutting simulation model, the prediction errors of the two kinds of prediction models can be obtained. The results show that the prediction accuracy of the BP neural network model is higher than that of the formula model, which shows that the BP neural network cutting force prediction model built in this paper has some advantages and can provide guidance for the production practice.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG501;TP183

【相似文献】