Structural Optimization Of Cylinder Based On Machine Learning

Carding machine is the key equipment in textile production,mainly used for processing cotton,wool,hemp,silk and other textile raw materials.It can comb raw materials into flat fibers and remove impurities and short fibers,thereby improving the quality and performance of textiles.In order to increase the output of the carding machine in the design,the cylinder,the key component,usually has the structural characteristics of large width and large diameter,which will cause concave deformation when it rotates at high speed,resulting in dynamic unbalance.When the dynamic unbalance is too large,the vibration generated by the rotation of the cylinder will have a great impact on the fiber carding of the carding machine.In this paper,the machine learning algorithm is used to optimize the geometric structure of the cylinder,aiming to reduce the mass of the cylinder while reducing the deformation of the cylinder to reduce the dynamic imbalance of the cylinder and improve the effect of the carding machine.The main content of this paper is as follows:Firstly,the structure and working procedure of the carding machine are introduced.The load distribution of the cylinder during work is analyzed and the reasons for the vibration of the cylinder and the relationship between the dynamic unbalance and the vibration of the cylinder is given,which determines the key points of the geometric optimization of the cylinder.The statics of the cylinder is analyzed and the approximate equation of the deflection curve of the cylinder roller is established.Secondly,the static simulation of the cylinder is carried out through the finite element method,and the relationship between the geometry parameters of the cylinder and its maximum deformation and the maximum local stress is analyzed.The modal of the cylinder,with and without the prestress,is analyzed also,which determines the critical speed when it is working,and an improvement plan is proposed.Then,the cylinder is parametrically modeled based on its key geometric parameters,and test points are generated based on Latin hypercube sampling.The data sets of the mechanical performance parameters of the cylinder are generated through the calculation of DOE.Finally,GA-BP neural network and RBF neural network is used to build the prediction model of the mechanical performance parameters of the cylinder respectively.The results show that the RBF neural network has a better prediction effect on the mechanical performance parameters of the cylinder.The NSGA-Ⅱ algorithm is used to carry out multi-objective optimization of the geometric parameters of the cylinder.With the Pareto front being iterated,the ideal parameters are found.The results show that,under the premise of not exceeding the allowable stress,the mass of the cylinder is reduced by 11.09%,and the maximum deformation of the cylinder is reduced by 15.17%,which meets the requirement of working.