西安建筑科技大学学报(自然科学版)

为了准确识别预应力混凝土结构的损伤程度,制作预应力钢筋混凝土实验梁,进行三点弯曲加载实验,收集损伤全过程声发射(AE)信号.绘制声发射振铃计数与持续时间的特征参数关联分布图,以揭示梁的损伤演化过程.借鉴加卸载响应比理论进一步将梁的损伤破坏过程划分为4个典型阶段.构建Elman神经网络,基于Elman神经网络采用局部搜索算法,难以达到全局最优的缺点,提出用蝙蝠算法(BA)对其进行优化.设计BA-Elman神经网络模型训练识别试验梁各损伤阶段AE信号特征参数数据,准确率达到93%,相较于基础Elman神经网络准确率提高了6%左右.定型BA-Elman网络结构并识别同种工况下的其他梁AE信号,识别准确率达到92%左右.

In order to accurately identify the damage degree of prestressed concrete structure, the prestressed reinforced concrete experimental beam was made, and the three-point bending loading experiment was carried out to collect the acoustic emission( AE )signal of the whole damage process. The correlation distribution diagram of characteristic parameters of acoustic emission ringing count and duration was drawn to reveal the damage evolution process of the beam. According to the load-unload response ratio theory, the damage process of the beam was further divided into four typical stages. The Elman neural network was constructed, but based on the local search algorithm adopted by the Elman neural network, it was difficult to achieve the global optimum, so the bat algorithm( BA )was proposed to optimize it. The BA-Elman neural network model was designed to train and identify the AE signal characteristic parameter data of experimental beam at various damage stages, and the accuracy rate reached 93%, which was about 6% higher than that of the basic Elman neural network. The BA-Elman network structure was established and AE signals of other beams under the same working condition were identified with an accuracy of about 92%.

引言
1 实验方案
2 预应力钢筋混凝土梁损伤演化分析
3 预应力钢筋混凝土梁典型损伤阶段划分
4 基于BA-Elman神经网络识别AE信号
5 结论

图1 试验梁几何尺寸及配筋<br/>Fig.1 Geometric dimensions and reinforcement of test beams

图1 试验梁几何尺寸及配筋
Fig.1 Geometric dimensions and reinforcement of test beams

图2 浇筑好的预应力钢筋混凝土梁<br/>Fig.2 Cast prestressed reinforced concrete beam

图2 浇筑好的预应力钢筋混凝土梁
Fig.2 Cast prestressed reinforced concrete beam

图3 多通道声发射仪<br/>Fig.3 Multichannel acoustic emission instrument

图3 多通道声发射仪
Fig.3 Multichannel acoustic emission instrument

图4 三点弯曲加载<br/>Fig.4 Three-point bending loading

图4 三点弯曲加载
Fig.4 Three-point bending loading

图5 循环加载示意图<br/>Fig.5 Diagram of cyclic loading

图5 循环加载示意图
Fig.5 Diagram of cyclic loading

表1 AE特征参数的含义及用途[13]<br/>Tab.1 Meaning and purpose of AE characteristic parameters

表1 AE特征参数的含义及用途[13]
Tab.1 Meaning and purpose of AE characteristic parameters

图6 梁C50-1振铃计数和持续时间关联分布图<br/>Fig.6 Correlation distribution of ring count and duration of beam C50-1

图6 梁C50-1振铃计数和持续时间关联分布图
Fig.6 Correlation distribution of ring count and duration of beam C50-1

图7 梁C50-1不同加载阶段状态<br/>Fig.7 State of beam C50-1 in different loading stages

图7 梁C50-1不同加载阶段状态
Fig.7 State of beam C50-1 in different loading stages

图8 加卸载响应比变化图<br/>Fig.8 Change of loading and unloading response ratio

图8 加卸载响应比变化图
Fig.8 Change of loading and unloading response ratio

图9 Elman神经网络结构<br/>Fig.9 Elman neural network structure

图9 Elman神经网络结构
Fig.9 Elman neural network structure

图10 BA-Elman神经网络识别流程图<br/>Fig.10 BA Elman neural network recognition flow chart

图10 BA-Elman神经网络识别流程图
Fig.10 BA Elman neural network recognition flow chart

表2 训练样本阶段分布情况<br/>Tab.2 Stage distribution of training samples

表2 训练样本阶段分布情况
Tab.2 Stage distribution of training samples

表3 BA-Elman神经网络训练参数<br/>Tab.3 Training parameters of BA Elman neural network

表3 BA-Elman神经网络训练参数
Tab.3 Training parameters of BA Elman neural network

图11 均方误差迭代<br/>Fig.11 Mean square error iteration of Elman neural network

图11 均方误差迭代
Fig.11 Mean square error iteration of Elman neural network

图12 梯度和验证失败次数迭代<br/>Fig.12 Gradient and verification failure iterations

图12 梯度和验证失败次数迭代
Fig.12 Gradient and verification failure iterations

图13 BA-Elman神经网络识别混淆矩阵<br/>Fig.13 Identification confusion matrix of BA-Elman neural network

图13 BA-Elman神经网络识别混淆矩阵
Fig.13 Identification confusion matrix of BA-Elman neural network

图14 BA-Elman神经网络识别ROC曲线<br/>Fig.14 ROC curve recognized by BA-Elman neural network

图14 BA-Elman神经网络识别ROC曲线
Fig.14 ROC curve recognized by BA-Elman neural network

表4 二分类问题识别情况[24]<br/>Tab.4 Identification of dichotomous problems

表4 二分类问题识别情况[24]
Tab.4 Identification of dichotomous problems

表5 梁C50-2损伤程度识别准确率<br/>Tab.5 Accuracy of damage degree identification of beam C50-2

表5 梁C50-2损伤程度识别准确率
Tab.5 Accuracy of damage degree identification of beam C50-2

[1]刘岩. 预应力混凝土结构发展综述[J]. 混凝土与水泥制品,2008(3): 52-55.
[2]ROBERTS T M,TALEBZADEH M. Acoustic emission monitoring of fatigue crack propagation[J]. Journal of Constructional Steel Research, 2003, 59(6): 695-712.
[3]MACHORRO-LOPEZ J M, HERNANDEZ-FIGUEROA J A, CARRION-VIRAMONTES F J, et al. Analysis of acoustic emission signals processed with wavelet transform for structural damage detection in concrete beams[J]. Mathematics, 2023, 11(3): 719.
[4]ZHANG F, YANG Y, NAAKTGEBOREN M, et al. Probability density field of acoustic emission events: Damage identification in concrete structures[J]. Construction and Building Materials, 2022, 327: 126984.
[5]陈忠购. 基于声发射技术的钢筋混凝土损伤识别与劣化评价[D]. 杭州:浙江大学, 2018.
[6]赵云鹏. 基于动力和声学特性的混凝土简支梁桥损伤识别方法研究[D]. 哈尔滨: 东北林业大学,2019.
[7]DAS K, BEHERA R N. A survey on machine learning: concept, algorithms and applications[J]. International Journal of Innovative Research in Computer and Communication Engineering, 2017, 5(2): 1301-1309.
[8]THIRUMALAISELVI A, SASMAL S. Pattern recognition enabled acoustic emission signatures for crack characterization during damage progression in large concrete structures[J]. Applied Acoustics, 2021, 175: 107797.
[9]MORFIDIS K, KOSTINAKIS K. Approaches to the rapid seismic damage prediction of R/C buildings using artificial neural networks[J]. Engineering Structures, 2018, 165: 120-141.
[10]苏三庆,韦璐茜,王威,等.基于支持向量机的钢结构隐性损伤磁记忆识别研究[J]. 西安建筑科技大学学报(自然科学版), 2019,51(1): 1-6.
[11]崔凤坤,杜岳涛,徐岳,等. 基于混合算法的大跨度钢管混凝土拱桥正常使用可靠度评估[J]. 西安建筑科技大学学报(自然科学版), 2016,48(6): 874-880.
[12]李舵, 董超群, 司品超, 等. 神经网络验证和测试技术研究综述[J]. 计算机工程与应用, 2021, 57(22): 53-67.
[13]沈功田,耿荣生,刘时风. 声发射信号的参数分析方法[J]. 无损检测,2002(2): 72-77.
[14]周俊临. 自适应自组织映射网络在模式识别中的应用研究[D]. 成都:电子科技大学, 2005.
[15]于江,皮滟杰,秦拥军. 循环载荷下再生混凝土损伤声发射特性[J]. 材料导报, 2021,35(13): 13011-13017.
[16]DING S, ZHANG Y, CHEN J, et al. Research on using genetic algorithms to optimize Elman neural networks[J]. Neural Computing and Applications, 2013, 23(2): 293-297.
[17]JIA W, ZHAO D, ZHENG Y, et al. A novel optimized GA-Elman neural network algorithm[J]. Neural Computing and Applications, 2019, 31(2): 449-459.
[18]YANG Xinshe. A new metaheuristic bat-inspired algorithm[C]//Nature Inspired Cooperative Strategies for Optimization(NICSO 2010). Berlin, Heidelberg: Springer Berlin Heidelberg, 2010: 65-7.
[19]玄登影,王福林,高敏慧,等. 一种改进适应度函数的遗传算法[J]. 数学的实践与认识, 2015,45(16): 232-238.
[20]李晓峰,徐玖平,王荫清,等. BP人工神经网络自适应学习算法的建立及其应用[J]. 系统工程理论与实践, 2004(5): 1-8.
[21]余华鸿,周凤艳,陈毛毛. 基于机器学习的KDD-CUP99网络入侵检测数据集的分析[J]. 计算机工程与科学, 2019,41(S1): 91-97.
[22]HAY A M. The derivation of global estimates from a confusion matrix[J]. International Journal of Remote Sensing, 1988, 9(8): 1395-1398.
[23]TAFIADIS D,CHRONOPOULOS S K, KOSMA E I, et al. Using receiver operating characteristic curve to define the cutoff points of voice handicap index applied to young adult male smokers[J]. Journal of Voice, 2018, 32(4): 443-448.
[24]乔宁. 多元逻辑回归在实时竞价中的应用研究[D].天津:河北工业大学, 2015.

备注

引言

1 实验方案

2 预应力钢筋混凝土梁损伤演化分析

3 预应力钢筋混凝土梁典型损伤阶段划分

4 基于BA-Elman神经网络识别AE信号

5 结论

期刊信息