摘要
传统的实验设计与分析方法为首先进行正交实验设计,然后对实验结果进行回归分析和方差分析以确定最佳工艺条件。文章提出的基于BP网络和遗传算法的正交实验分析方法,利用BP网络的高度非线性拟合特性对复杂的多输入多输出问题进行较高精度的回归,运用遗传算法优越的全局并行随机搜索及对适应度函数广泛的适应性等特性进行最优工艺条件的搜索,克服了传统分析方法系统模型辨识困难、后续实验工作量大以及最佳工艺条件搜索困难等缺点,大大提高了实验工作的效率和质量。
Traditional methods for experiment design and analysis is to design orthogonal experiment first,and then make regression analysis and variance analysis to search the optimal technological condition.This paper presents a new orthogonal experiment analysis method based on BP net and Genetic Algorithm,which makes use of perfect performance of function approximation of BP net to construct high precision regression function of complex multiple input and output problem,and effective concurrent stochastic global search and broad adaptability to fitness function of Genetic Algorithm to find the optimal technological condition.This method overcomes the limitations of traditional analytical methods such as hardness to identify system models and large quantity of workload of latter experiment as well as great difficulty to find the optimal technological condition and so on,and greatly improves the quality and effectiveness of experiment work.
出处
《计算机工程与应用》
CSCD
北大核心
2001年第20期16-18,共3页
Computer Engineering and Applications
基金
河北省教委博士科研启动基金资助
