摘要
遗传算法全局搜索能力强,而BP神经网络擅长局部精确搜索,采用遗传算法优化神经网络初始权值的方法,实现两种算法的结合,达到优势互补,并首次将内浇口类型及其厚度作为设计输出参数引入设计系统,构建更实用的压铸工艺参数优化设计多输入/多输出双隐层GA-BP神经网络。通过实例,验证了本系统所给出的压铸工艺设计结果的合理性。
Genetic algorithms （GA） are powerful approaches to solve global optimization problems; while BP neural networks are effective methods for searching local optimum. In this paper, to overcome the local convergence problem of BP network a genetic algorithm is proposed to optimize the initial weights of BP network and for the first time the style and thickness of ingate are included in output parameters of the optimizing system. The validation and effectiveness of this multi-input/multi-output double hidden layers GA-BP neural network for die-casting parameters design system are tested and approved by example.
出处
《铸造》
CAS
CSCD
北大核心
2009年第1期49-52,共4页
Foundry
关键词
GA-BP网络
压铸工艺参数
遗传算法
优化
GA-BP neural network
die-casting parameters
genetic algorithms
optimization
