Research On Thermal Elasticity Effect And Thermal Dynamics Characteristics Of NC Machine Tool Feed System

With the increase of the speed and acceleration, the machining precision is more and more obviously influenced by the thermal characteristic of NC machine tool. To improve the thermal characteristic of NC machine tool has become an urgent problem in the development of manufacturing technology. In this dissertation, according to heat transfer, thermoelasticity dynamics, control theory and system identification, the characterictics and the identifycating modl of thermoelastic effect, the thermal dynamic characteristics and the influencing rules of feed system, the modeling and forecasting methods of the thermal dynamic characteristics of feed system are systematically investigated by combining theoretical analysis with engineering practice.Thermoelastic phenomenon of object is brought by interchanging heat source. According to theory of heat transfer and Group Explicit numeric method, the thermoelastic effect of unidimensional heat transferring is studied. The being and changing rule of Thermoelastic phenomenon of rod are analysed. Three characteristics of the thermoelastic effect including frequency holding, real-time and integrative thermal deformation and phase shifting, are brought forward.Because it is difficult to get the theoretic model of thermoelastic effect, two kinds of nonlinear time series neural network models (NARX_NN and NARMAX_NN) are put forward, which are simple in structure, rapid in converging, dynamic fit for modeling of thermoelastic effect. Compared with multi-layer feedforward neural network (MFN), recurrent neural network(RNN) and Generalized least squares-output error (GLS-OE), these two models are proved valid in identifying thermoelastic effect and precise in calculateing nonlinear, dynamic and time varying thermoelastic process.The thermoelastic effect is one of the expressions of thermal dynamic characteristics. On the basis of theoretic analyse, GE calculating and finite element simulating of unidimensional heat transferring of ballscrew impacted by single heat source, the traits of lead and lag in temperature response and thermal deformation response are studied, and three characteristic parameters including frequency holding, amplitude attenuating and phase shifting are put forward, which can be used for describing thermal dynamic characteristics. By analysis of thermal dynamic characteristics of ballscrew impacted by multiple heat sources, Laplacian arithmetic is brought out to get the relation between thermal deformation on arbitrary location and the temperature of heat sources, and the steady-state thermal deformation response of ballscrew is researched.Using HUST-FS-001 sub high-speed feed system as an object,changing rules of thermal errors including axial, radial and return thermal errors in certain feed velocity are investigated. Rules of temperature rise and thermal error in changing machining conditions (feed velocity, cutting load and preload) are also investigaged. It is the foundation of identifying and modeling the thermal dynamic characteristics. The results of test show: temperature rises on bearing boxes are positive correlate with feed velocities, cutting loads and preloads; temperature rises on guides housings and screw nut boxes are not correlate with preloads, but they can reflect the changing rules of cutting loads and feed velocities; temperature rises on motor housing are positive correlate with cutting loads, but not correlate with preloads and feed velocity; Under three changing machining conditions ,the changing of radial thermal errors is little, stochastic and ruleless;The axial and return thermal errors are positive correlate with the machining conditions in certain rules.The thermal dynamic characteristics of feed system are complicated and nonlinear because of the complexity in structure, the multiformity in heat sources and the time varying in machining conditions. On the basis of analyzing the modeling ability of the two nonlinear time series neural network (NARX_NN and NARMAX_NN) on multiple input-multiple output (MIMO) system, the thermal dynamic models of HUST-FS-001 feed system are achieved by these two neural network.Compared with multi-variable regressive analysis (MRA), recurrent neural network (RNN) and Generalized least squares-output error (GLS-OE), these two models are proved more precise, robust and valid in identification of thermoelastic effect.The achievements of this dissertation provide significant theoretic foundation of real-time compensating for the thermal error of feed system. It is practical and valuable to apply these achievements to production practise. Also, it is the foundation of further research in thermal dynamic characteristics and evolving rules of integrated machine tools.