摘要
以热塑性聚酰亚胺多孔材料制备工艺为研究对象,考察主要工艺参数(冷压压力、烧结温度及保温时间)对多孔材料关键指标(含油率和油保持率)的影响;采用反向传播神经网络(BPNN)和径向基神经网络(RBFNN),建立其油保持率预测模型,分别考察了Levenberg-Marquardt算法和拟牛顿算法优化网络模型的运算量和精度。结果表明:随着压力、温度和时间的提高,材料孔隙率降低,从而含油率呈现下降趋势;而油保持率由材料孔径和孔隙率共同决定,正交实验表明其与工艺参数关系较复杂;同时RBFNN模型因采用径向基函数,在小输入量范围内可产生高响应,为此,更适合热塑性聚酰亚胺多孔材料冷压烧结工艺特点。
The preparation technology of thermoplastic polyimide porous material was researched. The effects of cold press pressure, sintering temperature and sintering time on oil reservation and oil retain were investigated ; the model of pre- dictive oil retain was designed by Back-Propagation Neural Network （BPNNN） and Radial Basis Function Neural Network （RBFNNN） ;training epochs and training goal of neural network were optimized with quasi-Newton method and Levenberg- Marquardt method. It is showed that porosity and oil reservation are decreased with the increase of pressure, temperature and time. Oil retain is affected on both pore diameter and porosity, orthogonal designing indicates that the relation between oil retain and technology parameter is complex. Meanwhile, the model of RBFNNN has better general ability in narrow independent variable range by radial basis function, therefore, it is appropriate to cold press and sintering technology of thermoplastic polyimide porous material.
出处
《润滑与密封》
CAS
CSCD
北大核心
2007年第1期79-82,104,共5页
Lubrication Engineering
基金
江苏省材料摩擦学重点实验室开放基金项目资助(kjsmcx04001)
