模糊神经网络在SU-8胶内应力研究中的应用  

Application of Fuzzy Neural Network in the Internal Stress of SU-8 Photoresist

作　　者：朱神渺[1] 杜立群[1] 刘冲[1] 喻立川[2] 王治宏[1] 

机构地区：[1]大连理工大学精密与特种加工教育部重点实验室,大连116023 [2]大连理工大学辽宁省微纳米技术及系统重点实验室,大连116023

出　　处：《纳米技术与精密工程》2007年第4期323-326,共4页Nanotechnology and Precision Engineering

基　　金：国家科技支撑计划项目(2006BAF04B13);国家自然科学基金资助项目(50675025)

摘　　要：以薄膜内应力为指标,对SU-8负性光刻胶的工艺参数进行了优化.根据基片曲率法的原理,通过三因素三水平的正交实验,测量了9组不同工艺条件下SU-8胶层内应力的大小.引入模糊神经网络,对影响SU-8胶内应力的工艺参数进行了优化仿真研究.以正交实验数据为样本,对模糊神经网络进行训练,建立了SU-8胶内应力的大小与前烘温度、曝光剂量和后烘温度3个主要参数之间的预测模型.同时对网络预测结果进行了实验验证,两者结果基本吻合.利用网络优化结果,有效地降低了胶层内应力.With the application of neural network, process parameters of negative SU-8 photoresist were optimized in terms of internal stress. Based on the principle of curvature method, a 33 factorial orthogonal array technique was designed to measure the internal stress of SU-8 photoresist with nine groups of different process parameters. Employing fuzzy neural network, a simulative model was established to investigate the influence of process parameters on the internal stress. Results of orthogonal experiment were trained by fuzzy neural network, and the prediction model was built between the internal stress and three main process parameters： soft baking （SB） temperature, exposure energy and post exposure baking （PEB） temperature. The prediction resuits were in good agreement with the experimental results. The internal stress was greatly reduced by using the optimized results of neural network.

关 键 词：SU-8胶 内应力 正交实验 BP神经网络 

分 类 号：TG492[金属学及工艺—焊接]