基于BP神经网络的曲轴润滑特性全局优化  

Global Optimization of Crankshaft Lubrication Characteristics Based on BP Neural Network

作　　者：王振[1] 白杨 郝长利[1] 杜建秋[1] 刘为 刘丽娜 李腾 Wang Zhen;Bai Yang;Hao Changli;Du JianQiu;Liu Wei;Liu Lina;Li Teng(Tianjin Internal Combustion Engine Research Institute, Tianjin University Tianjin, 300072, China;Tianjin HiRain Technologies Co., Ltd.)

机构地区：[1]天津大学内燃机研究所,天津300072 [2]天津经纬恒润科技有限公司

出　　处：《小型内燃机与车辆技术》2018年第4期42-48,共7页Small Internal Combustion Engine and Vehicle Technique

基　　金：国家重点研发计划(2016YFD0700400)

摘　　要：针对某增压柴油机缸内最大爆发压力提升后出现的曲轴润滑恶化现象,提出了一种基于最优拉丁超立方实验设计与BP神经网络相结合的优化方法,通过建立弹性流体动力润滑近似模型,对曲轴各主轴颈的峰值总压、最小油膜厚度、峰值粗糙接触压力、摩擦损失、剪切率、轴心运动等参数进行全局寻优,以获取最优的轴承间隙、轴颈和轴瓦表面粗糙度、供油孔角度和大小以及供油压力的设计参数。对比最优方案与原机弹性流体动力润滑计算结果发现:优化方案18 MPa在总摩擦损失方面只比原机16 MPa高约1.5%,粗糙接触比略优,剩余评价指标结果接近原机,所有指标全面优于原机18 MPa,总体达到预期优化目标。In order to extenuate the deterioration of crankshaft lubrication after increasing the maximum explosion pressure in a supercharged diesel engine, an optimization method based on the optimal Latin hy- percube experimental design combined with BP neural network was proposed. The best design parameters about the bearing clearance, journal and bearing surface roughness, oil supply hole angle and size, as well as the oil supply pressure were obtained by searching the global optimum of the peak total pressure, mini- mum oil film thickness, peak asperity contact pressure, friction loss, maximum shear rate, axle centre movement and other parameters of the crankshaft through the approximate model of the elastic hydrody- namic lubrication. Compared with the results of the original design which is under a load of maximum ex- plosion pressure of 160 bar, the total friction loss of the optimized project under the load of maximum ex- plosion pressure of 180 bar is just 1.5% higher, the asperity contact ratio is slightly lower and the remain- ing results are in good agreement with the original＇s. On the whole, the optimized project is superior to o- riginal design under the load of maximum explosion pressure of 180 bar and it achieves the desired opti- mization objective.

关 键 词：最优拉丁超立方 BP神经网络 弹性流体动力润滑 全局优化 

分 类 号：TK421.9[动力工程及工程热物理—动力机械及工程]