1汇报安排

题 目：参加 IMECE 2015国际会议总结报告会

时 间：2015年11月26日（周四）下午19: 00–21: 00

地 点：西五楼 A328

报 告 人：博1321班 马驰 学号：4113001094 指导教师：梅雪松 教授

2参会论文信息

会议名称：IMECE 2015

会议日期：13-19 November, 2015

会议地点：Houston, Texas, USA

会议简介：ASME’s International Mechanical Engineering Congress and Exposition (IMECE) is the largest interdisciplinary mechanical engineering meeting in the world. IMECE plays a significant role in stimulating innovation from basic discovery to translational application. It fosters new collaborations that engage stakeholders and partners not only from academia, but also from national laboratories, industry, research settings, and funding bodies. Among the 4,000 attendees from 75+ countries are mechanical engineers in advanced manufacturing, aerospace, advanced energy, fluids engineering, heat transfer, design engineering, materials and energy recovery, applied mechanics, power, rail transportation, nanotechnology, bioengineering, internal combustion engines, environmental engineering, and more.

3 参会论文信息

Title: Thermal Error modeling Based on Genetic Algorithm and BP Neural Network of High-speed Spindle System

Author: Chi Ma, Liang Zhao, Hu Shi, Xuesong Mei, Jun Yang

Abstract- In order to improve the prediction accuracy of the thermal error models, the five-point method was used to measure the thermal deformations of the high-speed spindle system at different rotational speeds to prepare the data for the thermal error modeling. Then, grey cluster grouping and correlation analysis were used to optimize and select the heat-sensitive points to improve the efficiency, stability and robustness of the model and minimize the independent variables to reduce modeling cost. Subsequently, the neural network with back propagation (BP) algorithm was used to construct the strongly nonlinear relationship between spindle thermal errors and typical temperature variables. Then, considering the BP network, which converged slowly and was difficult to obtain the global optimal solution, a genetic algorithm (GA) was applied to determine the structure and initial values of the BP neural network, namely, the number of the nodes in the hidden layer, the corresponding weights and the thresholds of the traditional BP network were optimized by the GA to obtain the optimal values of above parameters. The number of the nodes in the hidden layer can be determined by performing such operations of GAs as the selection, crossover and mutation. The reciprocal of the sum square of the difference between the predicted and expected outputs of individuals was regarded as the fitness function and the weights and the thresholds of the BP network can be optimized by setting the control parameters of the GA. Then, the high-speed spindle thermal error models based on BP and GA-BP networks were proposed and the fitting and prediction abilities were compared. The results showed that the five-point method can effectively analyze the variation of spindle’s position pose, and the grey cluster grouping and correlation analysis could depress the multicollinearity among temperature variables and improve the stability and accuracy of the thermal error models. Moreover, although the traditional BP network had better fitting ability, its convergence and generality were far worse than the GA-BP model. While the GA-BP model showed its effectiveness in quickly solving the global minimum searching problem with excellent robustness and convergence under different cutting conditions and it is more suitable to use the GA-BP neural network as the thermal error modeling method in the compensation system. And it is reasonable to develop the mapping relationship between thermal error and temperature variables because the nonlinear processing capacity of BP and GA-BP neural networks. Besides, the nonlinear relationship between the thermal error and temperature variable can be constructed by the BP and GA-BP neural networks and it is essential to optimize the structure and the weights and thresholds of traditional BP neural network to improve the accuracy, convergence and generality of thermal error models. And the prediction accuracy of GA-BP network is better than of traditional BP network.

欢迎有兴趣的同学届时参加。