作者：林旭梅, 刘帅, 石智梁

作者单位：青岛理工大学信息与控制工程学院，山东 青岛 266000

关键词：钢筋混凝土腐蚀程度;改进PSO;模糊神经网络;惯性因子;检测精度;pH 值传感器

摘要：

针对传统粒子群算法（PSO）在处理复杂搜索问题中容易产生早熟收敛，局部寻优能力较差等问题，提出PSO算法中惯性因子的自适应调整方法，将改进的PSO算法优化模糊神经网络(FNN)，并将改进的PSO-FNN算法应用于多传感器信息融合的钢筋混凝土腐蚀检测中，检测系统包括pH值传感器、氯离子传感器和湿度传感器。通过改进的PSO算法得到优化的神经网络连接权值，提高算法的搜索速度和训练效率，避免模糊神经网络易陷入局部极小值的问题。利用改进PSO-FNN算法对钢筋腐蚀的样本数据进行训练及测试，结果表明，改进的PSO-FNN腐蚀检测模型算法性能优于PSO-FNN算法，收敛速度快，可有效提高钢筋混凝土腐蚀检测的精度。

Research on reinforced concrete corrosion detection based on improved PSO-FNN algorithm
LIN Xumei, LIU Shuai, SHI Zhiliang
School of Information and Control Engineering, Qingdao University of Technology, Qingdao 266000, China
Abstract: Aiming at the problems that traditional particle swarm optimization (PSO) is prone to premature convergence and poor local optimization ability when dealing with complex search problems, an adaptive adjustment method of inertia factor in PSO algorithm is proposed. The improved PSO algorithm is used to optimize the fuzzy neural network (FNN), and the improved PSO-FNN algorithm is applied to the corrosion detection of reinforced concrete based on multi-sensor information fusion. The detection system includes pH sensor, chloride ion sensor and humidity sensor. The optimized neural network connection weights are obtained through the improved PSO algorithm, which improves the search speed and training efficiency of the algorithm, and avoids the problem of fuzzy neural networks easily falling into local minimums. The improved PSO-FNN algorithm is used to train and test the sample data of steel corrosion. The results show that the performance of the improved PSO-FNN corrosion detection model algorithm is better than the PSO-FNN algorithm, the convergence speed is faster, which improves the accuracy of reinforced concrete corrosion detection effectively.
Keywords: corrosion degree of reinforced concrete;improved PSO;fuzzy neural network;inertial factor;detection accuracy;pH sensor
2020, 46(12):149-155  收稿日期: 2020-07-20;收到修改稿日期: 2020-08-25
基金项目: 国家重点基础研究发展计划“973”项目（2015CB655100）
作者简介: 林旭梅(1971-),女,安徽桐城市人,教授,博士,研究方向为控制理论及应用、自动检测技术
参考文献
[1] MUSTAPHA S, LU Y, LI J, et al. Damage detection in rebar-reinforced concrete beams based on time reversal of guided waves[J]. Structural Health Monitoring, 2014, 13(4): 347-358
[2] 何世钦, 曹泽阳, 刘伟杰, 等. 长期荷载和氯盐环境耦合作用对钢筋混凝土梁挠度的影响[J]. 清华大学学报(自然科学版), 2019, 59(11): 902-909
[3] 余亮, 何旸, 吕健, 等. 腐蚀环境下带裂钢筋混凝土的侵蚀作用研究[J]. 混凝土与水泥制品, 2018(1): 27-31
[4] ZENG D, HAO B H, ZENG Q H. The study for non-destructive quantification method of reinforcement corrosion degree based on electrochemical detection and finite analysis technology[J]. Applied Mechanics and Materials, 2014, 527: 31-36
[5] 蒋萌, 蔡宁生, 寇新建, 等. 氯盐腐蚀环境下混凝土结构耐久性检测技术及研究[J]. 混凝土, 2013(6): 22-24+28
[6] HODHOD O, AHMED H. Modeling the corrosion initiation time of slag concrete using the artificial neural network[J]. HBRC Journal, 2014, 10(3): 231-234
[7] WANG C, SONG W. A modified particle swarm optimization algorithm based on velocity updating mechanism[J]. Ain Shams Engineering Journal, 2019, 10(4): 847-866
[8] 崔萌洁, 卢文科, 左锋, 等. 基于PSO-BP模型的扩散硅压力传感器温度补偿[J]. 中国测试, 2019, 45(11): 95-100+125
[9] 司景萍, 马继昌, 牛家骅, 等. 基于模糊神经网络的智能故障诊断专家系统[J]. 振动与冲击, 2017, 36(4): 164-171
[10] DOOSTDAR P, KEIGHOBADI J, HAMED M A. INS/GNSS integration using recurrent fuzzy wavelet neural networks[J]. GPS Solutions, 2020, 24(1): 1-15
[11] LIN H Y, LIN C J. Using a hybrid of fuzzy theory and neural network filter for single image dehazing[J]. Applied Intelligence, 2017, 47(4): 1099-1114
[12] 韩红桂, 林征来, 乔俊飞. 一种基于混合梯度下降算法的模糊神经网络设计及应用[J]. 控制与决策, 2017, 32(9): 1635-1641
[13] 沈贵阳, 刘远祥, 樊俊江. 既有污水混凝土构筑物表层及内部混凝土腐蚀情况研究[J]. 混凝土与水泥制品, 2019(9): 6-9
[14] 张立业, 孙利民. 混凝土梁桥动力特性对耐久性指标的灵敏度[J]. 同济大学学报(自然科学版), 2015, 43(11): 1613-1618
[15] 王娜, 胡超芳. 基于最近邻模糊聚类的T-S模糊辨识方法[J]. 控制工程, 2019, 26(6): 1068-1073