Research On Compensation Method For Dynamic Balance Prediction Of Mechanical Electric Spindle

With the introduction of Industry 4.0,industrial technology is developing in an efficient,sophisticated,and cutting-edge direction,and the precision requirements for machining in various industrial fields are also very strict.The vibration generated by the mechanical spindle during the production process is the main cause of the lathe machining error,which seriously affects the quality accuracy of the production parts.Due to the uneven distribution of the rotor mass,the spindle vibration will be caused,which will affect the machining accuracy.Therefore,the dynamic balance method is often used to reduce the vibration caused by such causes.The device eliminates the unbalance of the rotor in real time and achieves the purpose of reducing the vibration of the spindle,thereby improving The processing precision in the industrial production process increases the profit for the enterprise.This paper relies on the Liaoning Provincial Natural Science Foundation(20170540764)project to model the dynamic balance process of the mechanical spindle rotor for the mechanical spindle built-in balance block device.The specific research work is as follows:1.The mechanical spindle vibration amplitude prediction model based on RNN(Recurrent Neural Network)recurrent neural network is established,which provides an effective decision basis for the effective control of the mechanical spindle amplitude.Due to the complex mechanism of the mechanical spindle vibration prediction model,the vibration amplitude has nonlinear characteristics with the change of the rotational speed,and it is difficult to establish an accurate mechanical spindle vibration prediction model.Moreover,the built-in balance block position parameters in the existing model ignore the influence of the changing working conditions,resulting in low precision of the mechanical spindle vibration prediction model.In this paper,RNN(Recurrent Neural Network)recurrent neural network is used to establish the mechanical spindle vibration prediction model,and the vibration amplitudes of the built-in balance block at different positions and spindle speeds are predicted,and HS(Harmony Search)and acoustic search algorithm are introduced to balance block position.The parameters are updated by self-learning to improve the accuracy of the mechanical spindle amplitude prediction model.2.The method of dynamic balance process control of mechanical spindle combined with the method of mechanical spindle balance block position presetting by case-based reasoning technology and PID prediction compensation method based on particle swarm optimization algorithm is proposed.According to the amplitude feature value extraction in the case library,the position of the mechanical spindle balance block is pre-set to realize the function of the initial balance block position of the given mechanical spindle;the PID parameters are optimized according to the position parameters introduced by the particle swarm algorithm,and the mechanical spindle built-in balance is performed.Block pre-set position compensation.A method of amplitude compensation for mechanical spindle based on recurrent neural network(RNN)is proposed to compensate and control the position of the balance block of the mechanical spindle to improve the accuracy of compensation control.3.An experimental study of the method proposed in this paper was carried out.With the mechanical spindle vibration experimental platform in the national and local joint engineering laboratory of high-grade stone CNC machining equipment and technology of Shenyang Jianzhu University,the amplitude of the spindle vibration of the mechanical spindle at different speeds and different balance block positions is measured,which is the dynamic balance of the mechanical spindle.The study of forecast compensation methods provides data samples.The experimental research on the mechanical spindle amplitude prediction model and the mechanical spindle balance block position compensation method is carried out.The experimental results show that the proposed HS-RNN based mechanical spindle vibration prediction method can automatically determine the network structure,and accurately predict the amplitude of the mechanical spindle.Then through the mechanical spindle balance block position error prediction compensation model,the spindle vibration amplitude can be effectively reduced to within the error range,achieving a satisfactory balance effect.