基于遗传算法的神经网络对大体积混凝土监测优化研究

打开文本图片集

中图分类号：TU755.4 文献标志码：A 文章编号：1005-8249（2025）03-0134-07

DOI：10.19860/j.cnki.issn1005-8249.2025.03.024

Abstract：Inorder tostudytheoptimallocationof temperatureandstressmonitoringofmassconcreteunderthecircumstanceof limited numberof sensors，thetemperatureand stressvaluesof measuring pointswereobtainedbyfiniteelement simulationto buildadatabase，andthehighfitofneuralnetworkandtheoptimizationcharacteristicsof geneticalgoritmwereusedtofitthe temperature fieldand stressfieldofmassconcrete andoptimizethelocationof sensors.Theresultsshowthat theRMSEbetween thepredictedandmeasuredvaluesofeachnodedecreasesfrom5.21to3.56，andthecoeficientofdeterminationincreasesfrom 0.71to0.91，andtheoptimizationefectisremarkable.Throughoptimization，thecondition monitoringrequirementsof mass concrete can be realized under the limited number of sensors，and the specific applicability is better.

Key words：neural network；genetic algorithm；mass concrete；temperature field；stressfield；monitoring position

0 引言

随着我国经济高速发展，大型基础设施数量日益增长，大体积混凝土成为不可或缺的组成部分[1]大体积混凝土存在水化热高、散热困难、裂缝控制难度高等问题，且浇筑仓块不同位置的温升规律存在一定差异[2]，通过合理有效的状态监测，可及时发现并解决混凝土温升异常及裂缝问题。（剩余8244字）