多路阀阀芯节流槽拓扑结构组合的神经网络模型

发布时间：2018-02-15 10:46

本文关键词： 多路阀节流槽拓扑结构神经网络数字化设计　出处：《西安交通大学学报》2016年06期 　论文类型：期刊论文

【摘要】：针对多路阀设计中阀芯节流槽结构的拓扑形态表达和其流固耦合响应的大组合、大自由度解析问题,提出多路阀阀芯节流槽拓扑结构组合的神经网络模型。在利用多路阀动态特性台架实验验证其三维流体解析结果的基础上,将节流槽拓扑结构分类为由半圆槽、U型槽、圆孔等结构的参数化组合构成,通过正交实验法得到各参数组合条件下的多路阀三维流体解析响应,作为反向传播神经网络的训练样本,实现其节流槽拓扑结构组合的神经网络表达;采用进化神经网络优化训练过程的初始权重和阈值,优化后的神经网络能够对非训练样本集合的多路阀三维流体解析响应实现准确预测。研究结果表明,此模型为阀芯节流槽结构设计中拓扑形态表达提供了一种新的思路,对多路阀数字化设计具有实际意义。
[Abstract]:Aiming at the large combination of the topology form expression of the valve core throttle structure and the fluid-structure coupling response in the design of the multi-way valve, a large degree of freedom analytical problem is proposed. A neural network model for the topology combination of valve core throttle is presented. Based on the experimental results of three-dimensional fluid analysis of the valve, the topology of the throttle is classified as a semicircular groove with U-shaped groove. The parameterized composition of circular orifice and other structures is obtained by orthogonal experiment. The three-dimensional fluid analytical response of multi-way valve under the condition of each parameter combination is obtained, which is used as the training sample of backpropagation neural network. The neural network representation of the topological structure combination of the throttling slot is realized, and the initial weight and threshold of the training process are optimized by the evolutionary neural network. The optimized neural network can accurately predict the three-dimensional fluid analytical response of multi-way valve with non-training sample set. The research results show that the model provides a new way for the expression of topological shape in the design of valve core throttle structure. It has practical significance for digital design of multi-way valve.
【作者单位】：同济大学机械与能源工程学院;
【基金】：工业和信息化部2011年科技成果转化资助项目(财建〔2011〕30号)
【分类号】：TH137.52

【相似文献】