摘要
为了快速、准确地诊断井架钢结构的损伤位置和程度,提出仅基于测试精度高的频率数据和BP神经网络的识别方法。首先,选择频率变化比和频率平方变化比组合参数作为损伤位置识别因子,频率变化率作为损伤程度识别因子;然后,分步构建损伤位置和损伤程度识别的BP神经网络;最后,利用前10阶频率数据和BP神经网络对现场某井架钢结构的损伤位置和程度进行识别。分析结果表明,在测试噪声为10%时,采用前6阶损伤位置识别因子,能够清楚识别损伤位置,识别结果分别是1,5,9,15,19号单元损伤;采用前10阶损伤程度识别因子,1号单元的损伤程度识别结果分别为0.106 9,0.318 2,0.505 4,0.710 2,0.915 9,识别误差均不超过10%。
For the sake of diagnosing damage location and extent for derrick steel structures quickly and accurately, a novel identification approach was proposed only based on frequency dada with high testing accuracy and BP neural network. Firstly, the ratio of frequency changes and the square of frequency chan- ges were selected as identification indexes of damage location, and the rate of frequency changes was selected as indentification index of damage extent. Secondly, BP neural networks were etablished step by step for identificating damage location and extent. Finally, damage location and extent of a certain in-serv- ice derrick steel structure were identified only using first ten frequencies and BP neural network. Results show that with 10% measurement noise, damage location is clearly identified, respectively locating in No. 1, No. 5, No. 9, No. 15 and No. 19 element by taking the first six-order identification indexes of damage location, and damage extent of No. 1 element is respectively identified as 0. 106 9,0.318 2,0.505 4, O. 710 2 and 0. 915 9 by taking first ten-order identification indexes of damage extent, the recognition error is no more than 10%.
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2012年第8期118-123,共6页
China Safety Science Journal
基金
国家自然科学基金资助(51104129)
河北省自然科学基金青年基金资助(E2011203153)
关键词
井架钢结构
频率
BP神经网络
损伤识别
损伤系数
derrick steel structure
frequency
BP neural network
damage identification
damage coefficient
