基于改进复杂追踪算法的结构模态参数识别
|
摘要
以固定梯度的复杂追踪算法进行目标函数寻优时,收敛速度慢,易陷入局部极值,且存在针对不同的模型需人为选择合适的学习步长等不足,限制了该算法的实用性。为此,将最优步长思想引入复杂追踪算法,根据实时分离度自动调整步长,并根据分离信号的峭度值自适应地选择不同的非线性函数,以提高算法的计算精度,进而增强其实用性。为验证该改进复杂追踪算法识别结构模态参数识别的可行性与优越性,采用该方法分别识别了简支梁的数值模型和三层框架试验模型的模态参数,并与原方法进行识别结果对比,结果表明该改进算法可以较准确地识别结构模态参数。
Constant gradient based complex pursuit(CP) is prone to fall into local extremum, has slow convergence speed, and needs to artificially choose the appropriate learning step for different models. These shortcomings greatly limit the performance in practical application of this method. To solve the defects above, the idea of optimal step is introduced into the CP. In this idea, the step is automatically adjusted by the real-time degree of separation. At the same time, different nonlinear function can be chosen according to the kurtosis of the separated signal. Then, the computation accuracy of the CP algorithm is raised and its practical applicability is strengthened. To verify its practicability and reliability, the improved CP is applied to identify the modal parameters of a concrete simply-supported beam and an experimental three-story frame. The results are compared with those using the original CP algorithm. The accuracy of the improved CP in structural modal identification is demonstrated.
Constant gradient based complex pursuit(CP) is prone to fall into local extremum, has slow convergence speed, and needs to artificially choose the appropriate learning step for different models. These shortcomings greatly limit the performance in practical application of this method. To solve the defects above, the idea of optimal step is introduced into the CP. In this idea, the step is automatically adjusted by the real-time degree of separation. At the same time, different nonlinear function can be chosen according to the kurtosis of the separated signal. Then, the computation accuracy of the CP algorithm is raised and its practical applicability is strengthened. To verify its practicability and reliability, the improved CP is applied to identify the modal parameters of a concrete simply-supported beam and an experimental three-story frame. The results are compared with those using the original CP algorithm. The accuracy of the improved CP in structural modal identification is demonstrated.
引文
[1]张晓丹.基于盲源分离技术的工程结构模态参数识别方法研究[D].北京:北京交通大学,2010.
[2]静行,袁海庆,赵毅.基于独立分量分析的结构模态参数识别[J].振动与冲击,2010,29(3):137-141.
[3] SHI Z, TANG H, TANG Y. A fast fixed-point algorithm for complexity pursuit[J]. Neurocomputing, 2005, 64(1):529-536.
[4] KERSCHEN G, PONCELET F, GOLINVAL J C. Physical interpretation of independent component analysis in structural dynamics[J]. Mechanical Systems and Signal Processing, 2007, 21(4):1561-1575.
[5] MCNEIL S. Modal identification using blind source separation techniques[D]. Houston:University of Houston, 2007.
[6]张晓丹,姚谦峰,刘佩.基于快速独立分量分析的模态振型识别方法研究[J].振动与冲击,2009,28(7):158-161.
[7] ANTONI J, CASTIGLIONE R, GARIBALDI L.Interpretation and generalization of complexity pursuit for the blind separation of modal contributions[J].Mechanical Systems&Signal Processing, 2017, 85:773-788.
[8] HYV?RINEN A, OJA E. A Fast Fixed-Point Algorithm Independent Component Analysis[J]. Neural Computation, 1997, 9(7):1483-1492.
[9]吕淑平,方兴杰,杨丽微.独立分量分析的算法分析与改进[J].噪声与振动控制,2013,33(6):153-157.
[10] PONCELET F, KERSCHENA G, GOLINVALA J C, et al.Output-only modal analysis using blind source separation techniques[J]. Mechanical Systems and Signal Processing, 2007, 21(6):2335-2358.
[11]胡皞,常军,刘文波,等.基于独立成分分析的结构模态参数识别[J].苏州科技学院学报,2014,27(3):40-45.
[12]曹树谦,张文德,等.振动结构模态分析-理论、实验与应用[M].天津:天津大学出版社,2001.
[2]静行,袁海庆,赵毅.基于独立分量分析的结构模态参数识别[J].振动与冲击,2010,29(3):137-141.
[3] SHI Z, TANG H, TANG Y. A fast fixed-point algorithm for complexity pursuit[J]. Neurocomputing, 2005, 64(1):529-536.
[4] KERSCHEN G, PONCELET F, GOLINVAL J C. Physical interpretation of independent component analysis in structural dynamics[J]. Mechanical Systems and Signal Processing, 2007, 21(4):1561-1575.
[5] MCNEIL S. Modal identification using blind source separation techniques[D]. Houston:University of Houston, 2007.
[6]张晓丹,姚谦峰,刘佩.基于快速独立分量分析的模态振型识别方法研究[J].振动与冲击,2009,28(7):158-161.
[7] ANTONI J, CASTIGLIONE R, GARIBALDI L.Interpretation and generalization of complexity pursuit for the blind separation of modal contributions[J].Mechanical Systems&Signal Processing, 2017, 85:773-788.
[8] HYV?RINEN A, OJA E. A Fast Fixed-Point Algorithm Independent Component Analysis[J]. Neural Computation, 1997, 9(7):1483-1492.
[9]吕淑平,方兴杰,杨丽微.独立分量分析的算法分析与改进[J].噪声与振动控制,2013,33(6):153-157.
[10] PONCELET F, KERSCHENA G, GOLINVALA J C, et al.Output-only modal analysis using blind source separation techniques[J]. Mechanical Systems and Signal Processing, 2007, 21(6):2335-2358.
[11]胡皞,常军,刘文波,等.基于独立成分分析的结构模态参数识别[J].苏州科技学院学报,2014,27(3):40-45.
[12]曹树谦,张文德,等.振动结构模态分析-理论、实验与应用[M].天津:天津大学出版社,2001.