Modeling Research Of Thermal Error Prediction For High-speed Motorized Spindle

The high-speed motorized spindle is the core component of CNC machine tools,and the motorized spindle is widely used in various machining situations due to its compact structure.However,due to the compact structure of the motorized spindle,it is difficult to dissipate the heat inside the motorized spindle,and the heat accumulates,which leads to thermal deformation of motorized spindle parts and thus generates machining errors.It is difficult to set up cooling water channels inside the motorized spindle,so building a motorized spindle thermal deformation prediction model supplemented by a compensation strategy becomes the primary means to improve the machining accuracy of the motorized spindle.In this thesis,we focus on improving the accuracy of the prediction model of motorized spindle thermal error,analyze the impact of inputting different temperature data on the prediction accuracy of the motorized spindle thermal deformation prediction model,and optimize the prediction model for its shortcomings.The main research elements of this thesis include the following:First,the finite element steady-state temperature field simulation analysis of the motorized spindle is carried out.Calculation of motorized spindle motor loss heat generation and bearing friction heat generation.Based on the cooling strategy and heat transfer mechanism of the motorized spindle,the heat flow inside the motorized spindle is analyzed.Solid Works software is used to model the motorized spindle in combination with the structural parameters of the model A0304 motorized spindle.Workbench software is used to simulate the temperature field distribution of the model A0304 motorized spindle at rated speed,which provided the basis for the modeling of the thermal resistance network and the division of thermal nodes.Secondly,the motorized spindle speed-up experiment platform is built,and the thermal displacement of the nose and flange end surfaces of the motorized spindle is collected using eddy current displacement sensors,and the temperature data of the outer ring of the motorized spindle bearing and the motor stator are collected using platinum RTD temperature sensors,and the experimental results are analyzed to provide data support for thermal resistance network calculation and thermal error prediction modeling.After that,for the problem that it is difficult to collect the temperature data of the bearing inner ring and roller experimentally,the thermal resistance network method is used to model the motorized spindle,and the temperature of the bearing inner circle and roller is calculated by combining the experimentally collected temperature data.Based on the experimentally collected data and the data calculated by the thermal resistance network method,a BP neural network motorized spindle thermal error prediction model is established,and the temperatures of the outer bearing ring,bearing roller and bearing inner circle are used as the training set of the prediction model,and the effect of using the temperature data of different bearing positions as the training set on the accuracy of the prediction model is compared and analyzed.Finally,the BP neural network is optimized using the Whale Optimization Algorithm(WOA)and Bird Swarm Optimization Algorithm(BSA)to optimize the weights and thresholds of the BP neural network.The thermal error prediction models of WOA-BP and BSA-BP motorized spindle are established,and the prediction accuracy of the two prediction models is compared and analyzed by using the fitting index.