An iterative equivalent linearization approach for stochastic sensitivity analysis of hysteretic systems under seismic excitations based on explicit time-domain method
Introduction
Sensitivity analysis of responses with respect to structural parameters is of great importance to structural optimization, structural health monitoring and model updating, etc. The early works [1], [2], [3] on the sensitivity analysis of structures mainly concentrated on deterministic dynamic loads. However, the deterministic sensitivity analysis cannot capture reasonable sensitivity results from a probabilistic point of view when structures are subjected to random excitations such as earthquake, wind and wave. Therefore, in this sense, sensitivity analysis methods should take into account the uncertainties of dynamic excitations.
The concept of stochastic sensitivity analysis, which is aimed at finding the sensitivities of statistical moments of structural responses with respect to structural parameters, was proposed by Szopa [4] and Socha [5]. Afterwards, Benfratello et al. [6] developed a time-domain approach for the sensitivity evaluation of linear structural systems subjected to stationary random excitations. Cacciola et al. [7] proposed a semi-analytical approach to investigate the sensitivities of random responses of both classically and non-classically damped linear systems under nonstationary random excitations. Chaudhuri and Chakraborty [8] dealt with the response sensitivity evaluation of linear structures in the frequency domain under nonstationary random excitations. Based on the pseudo-excitation method [9], Xu et al. [10] presented a new approach for the first and second order sensitivity analysis of the nonstationary random responses of coupled vehicle-bridge linear systems, and Liu [11] developed a numerical method for calculation of the sensitivity of the evolutionary response power spectral density functions of linear structures subjected to nonstationary random excitations. Motivated by the explicit time-domain method (ETDM) proposed by Su and Xu [12] for nonstationary random response analysis of linear structures, Hu et al. [13], [14] further extended this method to the sensitivity analysis of variance responses and dynamic reliability of linear structures subjected to nonstationary random excitations. More recently, Chun et al. [15], Zhu et al. [16] and Gomez and Spencer [17] derived the sensitivities of stationary random responses of linear systems in different ways and further employed the sensitivity results for topology optimization. More research on the sensitivity analysis of linear systems under random excitations can be found in [18], [19].
As compared with the research on the sensitivity analysis of linear systems under random excitations, only a few works were presented in the literatures on the sensitivity analysis of stochastically excited nonlinear systems. As a pioneer in the field of stochastic sensitivity analysis of nonlinear systems, Socha [5] studied the response sensitivity of a single-degree-of-freedom (SDOF) Duffing oscillator and a SDOF Van der Pol oscillator under stationary white noise and colored noise excitations based on the moment equation method. Later, Socha and Zasucha [20] derived the approximate analytical sensitivities with respect to model parameters for the stationary responses of a SDOF nonlinear hysteretic system subjected to Gaussian white noise excitation via the statistical linearization approach. Socha and Soong [21] discussed the relationship between sensitivity analysis and equivalent linearization as applied to an oscillator with an odd order polynomial stiffness under stationary Gaussian white noise excitation. As can be seen from the above literatures, these works are limited to SDOF nonlinear systems with stationary random excitations.
The main purpose of this paper is to present an explicit time-domain approach for sensitivity analysis of multi-degree-of-freedom (MDOF) nonlinear systems under nonstationary seismic excitations. The recently proposed ETDM [12], [13], [14] has proven to be an efficient method for nonstationary stochastic response and sensitivity analysis of linear systems owing to the use of the explicit time-domain expressions of dynamic responses and their sensitivities. And for nonlinear systems, the ETDM has been used in conjunction with the equivalent linearization method (ELM) for the nonstationary response analysis of Duffing systems and hysteretic systems with high efficiency [22]. On this basis, in the present study, the ETDM-based ELM is further extended to the sensitivity analysis of nonstationary stochastic responses of hysteretic systems under seismic excitations. For each time instant, an overall equivalent linear equation consisting of the equivalent linear equation of motion and the corresponding sensitivity equation as well for the original hysteretic system is first derived based on the statistical linearization technique. Then, the above overall equivalent linear equation is solved on an iterative basis, in which the ETDM is used to conduct the series of nonstationary stochastic sensitivity analyses of the iterative linearized systems. A 100-degree-of-freedom hysteretic system is investigated to illustrate the accuracy and efficiency of the proposed method, and a 10-degree-of-freedom base-isolated system is further analyzed to show the feasibility of the proposed method for stochastic structural optimization.
Section snippets
Equivalent linear system of hysteretic system under nonstationary seismic excitation
Various methods have been developed for random vibration analysis of nonlinear systems, such as the Fokker-Planck-Kolmogorov equation method [23], the stochastic averaging method [24], the moment equation method [25], the ELM [26], [27], the probability density evolution method [28] and the Wiener path integral method [29]. Among them, the ELM has gained wide popularity for its versatility in engineering application, while most of the other methods are limited to the size of the problem and
ETDM-based ELM for stochastic sensitivity analysis of hysteretic systems
Given and with previous iterative values of statistical moments and sensitivities, Eq. (15) is in the same form as the equation of motion for a linear system. Therefore, Eq. (15) can be solved using linear random vibration methods. The ETDM has been found to be an efficient approach for nonstationary random vibration problems of linear systems owing to the use of explicit time-domain expressions of dynamic responses [12], [22]. In this section, the ETDM is incorporated into the
Numerical examples
Two numerical examples are investigated in this section. In the first example, different methods, including the ETDM-based ELM, the PSM-based ELM and the Monte-Carlo simulation (MCS), are used to obtain the statistical moments and sensitivities of responses, through which the accuracy and efficiency of the proposed method are validated. In the second example, the proposed ETDM-based ELM is further used to conduct the stochastic optimal design of base isolators so as to show the feasibility of
Conclusions
An efficient time-domain approach has been proposed for the stochastic sensitivity analysis of hysteretic systems subjected to nonstationary seismic excitations in the frame of ELM with fusion of ETDM. For each time instant, the equivalent linear equation of motion is first established by the statistical linearization technique, and the corresponding sensitivity equation is then determined analytically through direct differentiation. For convenience of formulation and direct use of the existing
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The research is funded by the National Natural Science Foundation of China (51678252) and the Science and Technology Program of Guangzhou, China (201804020069).
References (41)
- et al.
Design sensitivity analysis and optimization of dynamic response
Comput Methods Appl Mech Eng
(1984) - et al.
Efficient design sensitivities of structures subjected to dynamic loading
Int J Solids Struct
(2006) Sensitivity of stochastic systems to initial conditions
J Sound Vib
(1984)The sensitivity analysis of stochastic non-linear dynamical systems
J Sound Vib
(1986)- et al.
Gaussian and non-Gaussian stochastic sensitivity analysis of discrete structural system
Comput Struct
(2000) - et al.
A modal approach for the evaluation of the response sensitivity of structural systems subjected to non-stationary random processes
Comput Methods Appl Mech Eng
(2005) - et al.
Sensitivity evaluation in seismic reliability analysis of structures
Comput Methods Appl Mech Eng
(2004) - et al.
Accurate and highly efficient algorithms for structural stationary/non-stationary random responses
Comput Methods Appl Mech Eng
(2001) - et al.
Sensitivity analysis and optimization of vehicle–bridge systems based on combined PEM–PIM strategy
Comput Struct
(2011) Sensitivity and Hessian matrix analysis of power spectral density functions for uniformly modulated evolutionary random seismic responses
Finite Elem Anal Des
(2012)
An explicit time-domain approach for sensitivity analysis of non-stationary random vibration problems
J Sound Vib
Topology optimization for linear stationary stochastic dynamics: applications to frame structures
Struct Saf
Sensitivity analysis of optimum stochastic nonstationary response spectra under uncertain soil parameters
Soil Dyn Earthquake Eng
Explicit sensitivities of the response of discretized structures under stationary random processes
Probab Eng Mech
Sensitivity and linearization techniques in analysis of non-linear stochastic systems
J Sound Vib
Stochastic averaging of n-dimensional non-linear dynamical systems subject to non-Gaussian wide-band random excitations
Int J Non Linear Mech
Gaussian closure technique applied to the hysteretic Bouc model with non-zero mean white noise excitation
J Sound Vib
Equivalent linearization method using Gaussian mixture (GM-ELM) for nonlinear random vibration analysis
Struct Saf
Probability density evolution method: background, significance and recent developments
Probab Eng Mech
Equivalent linearization and Monte Carlo simulation in stochastic dynamics
Probab Eng Mech
Cited by (20)
A generalized method for the stationary probabilistic response of nonlinear dynamical system
2023, Communications in Nonlinear Science and Numerical SimulationRandom response analysis of nonlinear structures with inerter system
2023, Soil Dynamics and Earthquake EngineeringCitation Excerpt :Such methods basically obtain the response PDF by solving the Fokker–Planck–Kolmogorov (FPK) equation derived from the motion equation [24]. Recently, Su combined the equivalent linearization method with the explicit time-domain method to develop a method for stochastic sensitivity analysis [25]. Jiang proposed a method to renormalize the FPK equation [26].
Stochastic heat transfer analysis and reliability assessment under non-stationary random thermal load using the explicit time-domain method
2022, International Journal of Heat and Mass TransferCitation Excerpt :It would be exciting for future studies to harness the power of ETDM to solve other types of stochastic problems, including electrodynamics [52], fluid dynamics [53] and, in particular, multiphysics problems [54,55]. Moreover, as many practical heat transfer problems are nonlinear, another interesting extension of the present study would be to adopt the methods developed for nonlinear random vibrations [37,56] to treat nonlinear stochastic heat transfer problems. Yingxiong Li: Methodology, Investigation, Data curation, Writing – original draft.
Spectral decomposition-based explicit integration method for fully non-stationary seismic responses of large-scale structures
2022, Mechanical Systems and Signal ProcessingCitation Excerpt :Su et al. [40,41] derived explicit expressions for the stochastic state responses of structures using several direct integration schemes, which allows the structural response statistics to be directly evaluated in the time domain. Such a method is usually termed in the literature as the explicit time-domain method (ETDM) [42,43]. In practical application, the ETDM requires the calculation of excitation correlation matrix at a series of time instants, which is usually achieved via numerical integration of the excitation PSD matrix that can be very time-consuming, especially for addressing the fully non-stationary excitations.