2018年第7期文章——基于BP神经网络算法的聚乙烯管材寿命预测

本期栏目

综 述

材料与性能

加工与应用

助 剂

塑料与环境

机械与模具

文章推荐

基于BP神经网络算法的聚乙烯管材寿命预测

谷亚新，赵梓怡

（沈阳建筑大学材料科学与工程学院，沈阳 110168）

摘要: 通过分析聚乙烯管材在实际应用中寿命的影响因素，选取其中可控因素加以分析及相关性预测。以MATLAB计算软件为平台，根据具体试验建立寿命预测模型，参考试验变量设计模型可变因素。实验设计部分，以静液压实验为基础，选取实验温度、实验用水有效氯含量以及实验压力作为3因素变量进行静液压实验，共进行实验192组次，试验样品384个，模拟聚乙烯管材在输配水系统中的实际使用。在建模过程中，进行神经网络学习后，对所建模网络进行训练，然后对其中所选择的3因素变量BP神经网络疲劳方程模型进行寿命预测。结果表明，BP神经网络计算出聚乙烯管材的使用寿命具有科学性，运行后的网络模型概率学上拟合优度R2为0.87；预测结果相对于聚乙烯管材的通常算法具有计算简洁、样本可扩充、模型阀值及权重可自行更改等一系列优势。

关键词:BP神经网络；静液压实验；失效分析；寿命预测

Lifespan Prediction of Polyethylene Pipe Based on BP Neural Network Algorithm

GU Yaxin, ZHAO Ziyi

（School of Materials Science and Engineering, Shenyang Jianzhu University, Shenyang 110168, China）

Abstract:The influence factors for the lifespan of polyethylene (PE) pipe in practical application were analyzed, and the controllable factors were selected to conduct a lifespan prediction. A lifespan prediction model was established according to the experimental results, and the changeable variables of model were designed by reference to the experimental variables using MATLAB software. For the experimental design, a certain numerical gradient was selected according to the hydrostatic tests. The experimental design was according to the hydrostatic tests. Select the experimental temperatures, the effective chlorine contents in the experimental water, and the experimental pressures for totally 192 groups of experiments and 384 test samples. The PE pipe was simulated in the practical use of water distribution systems. In the computational modeling process, the BP neural network fatigue equation model with three-factor variables was selected for life prediction. The obtained results indicated that the BP neural network has a scientific significance to calculate the service lifespan of PE pipe, and the goodness of fit, R2, was up to 0.87 in probability according to this network model. Compared to the conventional algorithms of PE pipe, the prediction results have superiority in calculation simplification, specimen expansion, model threshold and importance variability.

Key words:BP neural network; hydrostatic test; failure analysis; lifespan prediction

【文章信息】

DOI:10.19491/j.issn.1001-9278.2018.07.016

收稿日期：2018-02-08

联系人，1594750270@qq.com

本期文章导读

行业会议推荐