摘要
采用钢纤维、玻璃纤维、铜纤维、矿物纤维等增强纤维,石油焦碳、人造石墨、天然石墨等摩擦调节组元,以及树脂、丁腈橡胶、丁苯橡胶等粘接剂制备汽车摩擦材料。选用BP神经网络建模,以原材料配方为输入变量、摩擦磨损试验数据为输出变量,采用L-M算法对网络进行训练,并进行摩擦磨损性能预测和配方优化。结果表明,隐层神经元为4的单隐层神经网络结构模拟效果较好,性能曲面预测图能表现出原材料的组合性能,采用该网络优化试样的性能测试结果与预测值的相对误差小于20%。
Reinforcing fibers（steel fiber, glass fiber, copper fiber, mineral fiber）, friction modifiers（petroleum coke, synthetic graphite, flake graphite）, binders（resin, NBR rubber, SBR rubber）, and fillers are used to design and fabricate the automotive brake linings. BP neural networks are tried to simulate the relationship between the raw material formulations and the performance of automotive friction materials. The research results show that the optimized architecture of BP neural network is 3 layers with 4 neural in hidden layer. After being trained, BP neural network can forecast the tribological and wear performance, and optimize the formulation of friction materials. The resuits show that the forecasting performance error of the automotive friction material under the optimized condition is less than 20%.
出处
《材料导报》
EI
CAS
CSCD
北大核心
2010年第10期74-78,共5页
Materials Reports
基金
中央高校基本业务费专项资金资助项目(CUGL090223)
教育部留学回国人员科研启动基金
湖北省教育厅重点研究项目(2009114)
地质过程与矿产资源国家重点实验室开放基金项目(GPMR200918)
关键词
增强纤维
摩擦材料
摩擦磨损性能
BP神经网络
配方优化
reinforcing fiber, friction material, friction and wear performance, BP neural network, formulation optimization
