Position-dependent Dynamics Prediction Based On Deep Learning For Five-axis Machine Tools

Five-axis machine tool is an advanced manufacturing equipment that can be applied for high-precision manufacturing,and is widely used in nautical,aerospace,advanced medical equipment and other fields.Compared to traditional three-axis machine tool,five-axis machine tool introduces two rotation axes,which expands the machining workspace,making it possible to machining complex curved surfaces and high-precision components.The problem of weakening dynamic behavior is carried out during five-axis machining due to the constant changes of the rotation axes.Machining stability,producing accuracy and surface roughness are great influenced by the swing axis and rotation axis which are the core components of the five-axis machine tool.The dynamics of five-axis machine tool usually changes significantly with the continuously varying position of center mass of rotation axis during the whole machining process.Therefore,researching the dynamic behavior of five-axis machine tool and constructing the distribution of dynamics in the workspace are of great significance for improving machining stability and accuracy.The main research contents of this paper are as follows:(1)Dynamic behavior in the workspace of a five-axis machine tool is analyzed.The tilting table of a type five-axis machine tool is taken as the research object,and the dynamic model is built according to its topology structure,and the dynamic equations are derived to obtain the distribution of dynamics in the machining workspace of the fiveaxis machine tool.The finite element modeling and analysis method is applied to ideally approximate the connection points of the machine tool.Through the finite element simulation experiment,the distribution of dynamics in the whole workspace of the fiveaxis machine tool is qualitatively analyzed.(2)Experimental excited method for evaluating dynamic behavior in the workspace of five-axis machine tool is proposed.According to the structural features of the five-axis machine tool,the target parameters of the machine tool dynamics are clarified,the design of the excitation signal in the workspace of the machine tool is studied,and the excited instrument which is used to the five-axis machine tool is manufactured.The influence caused by excited instrument on the measurement results is analyzed in detail,and the design and construction of experimental excited system for dynamics of a five-axis machine tool is completed.(3)Dynamics modeling algorithm based on deep learning is researched carefully.According to the deep learning theory,the dynamics prediction model is constructed.The neural network architecture,the type of neural network and the neural network algorithms are clarified specifically.The deep learning model of dynamic behavior prediction algorithm is constructed to obtain the distribution of dynamics of the five-axis machine tool in the whole workspace.The transfer learning model is used to obtain the distribution of dynamics of another type of machine tool.(4)A detailed excitation experiment validation of a five-axis machine tool is conducted.The experimental excited system is constructed according to the workspace of the five-axis machine tool,and the data processing process and system parameter identification method are improved to acquire the more accuracy results.The excitation experiment of the five-axis machine tool is completed by using the excitation instrument,and the distribution of dynamics of the experimental machine tool is obtained.