Pushpa Pandey,Hamed Haddad Khodaparast,Michael Ian Friswell,Tanmoy Chatterjee,Hadi Madinei,Tom Deighan
Abstract:The study presents a novel approach for stochastic nonlinear model updating in structural dynamics, employing a Bayesian framework integrated with Markov Chain Monte Carlo (MCMC) sampling for parameter estimation by using an approximated likelihood function. The proposed methodology is applied to both numerical and experimental cases. The paper commences by introducing Bayesian inference and its constituents: the likelihood function, prior distribution, and posterior distribution. The resonant decay method is employed to extract backbone curves, which capture the non-linear behaviour of the system. A mathematical model based on a single degree of freedom (SDOF) system is formulated, and backbone curves are obtained from time response data. Subsequently, MCMC sampling is employed to estimate the parameters using both numerical and experimental data. The obtained results demonstrate the convergence of the Markov chain, present parameter trace plots, and provide estimates of posterior distributions of updated parameters along with their uncertainties. Experimental validation is performed on a cantilever beam system equipped with permanent magnets and electromagnets. The proposed methodology demonstrates promising results in estimating parameters of stochastic non-linear dynamical systems. Through the use of the proposed likelihood functions using backbone curves, the probability distributions of both linear and non-linear parameters are simultaneously identified. Based on this view, the necessity to segregate stochastic linear and non-linear model updating is eliminated.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to solve the problem of stochastic nonlinear model updating in structural dynamics. Specifically, the paper proposes a novel method that uses the Bayesian framework combined with Markov Chain Monte Carlo (MCMC) sampling techniques to estimate parameters by approximating the likelihood function. This method is applicable to numerical and experimental cases and captures the nonlinear behavior of the system through backbone curves.
### Main contributions
1. **Proposed a new likelihood function**:
- Traditional Bayesian methods face problems such as complex likelihood functions, sensitivity to initial conditions, model misspecification, and data limitations when dealing with nonlinear systems. In this paper, by introducing a likelihood function based on backbone curves, the representation of nonlinear dynamic systems is simplified, thereby reducing the complexity of the likelihood function.
- This data - driven likelihood function approximation method improves the robustness to system behavior and reduces the dependence on theoretical assumptions.
2. **Combined Bayesian inference and MCMC techniques**:
- The Metropolis - Hastings algorithm is used to generate samples of the posterior distribution, thereby estimating parameters and their uncertainties.
- This method can handle high - dimensional complex distributions and is applicable to linear and nonlinear systems.
3. **Experimental verification**:
- The effectiveness of the proposed method is verified through experiments. The experimental system includes a cantilever beam system with permanent magnets and electromagnets.
- The experimental results show the convergence of the Markov chain, providing parameter trajectory diagrams and posterior distribution estimates of the updated parameters and their uncertainties.
### Mathematical models and methods
1. **Bayesian inference**:
- Bayesian inference uses Bayes' rule to update our beliefs about the distribution of system parameters. The main components include:
- **Likelihood function**: The probability of observing data given specific parameter values.
- **Prior distribution**: The initial belief about system parameters before observing any data.
- **Posterior distribution**: The updated distribution obtained by combining the prior distribution and the likelihood function.
- The mathematical expression for the posterior distribution is:
\[
P(\Theta | D, M)=\frac{P(D | \Theta, M)\cdot P(\Theta | M)}{P(D | M)}
\]
where \(\Theta\) is the parameter vector to be estimated, \(D\) is the observed data, and \(M\) is the system model.
2. **Establishment of the likelihood function**:
- Using backbone curves as measurement data, fit the probability density function (PDF) for each amplitude point, and then combine these PDFs into a joint PDF as the likelihood function.
- The likelihood function is defined as the joint distribution of all distributions at different amplitudes:
\[
L(D | \Theta, M)=\prod_{j = 1}^{N}P(d_j | \Theta, M)
\]
where \(P(d_j | \Theta, M)\) represents the probability density function of the measured data point \(d_j\) given the deterministic model and parameter values \(\Theta\).
3. **Markov Chain Monte Carlo (MCMC)**:
- MCMC is a framework for generating a sequence of samples from a complex probability distribution. Commonly used algorithms include the Metropolis - Hastings algorithm.
- The algorithm steps include initialization, generating candidate states, acceptance - rejection steps, updating the current state, iteration, and output.
- By repeatedly performing the above steps, samples of the target distribution are generated for parameter estimation.
### Conclusion
The method proposed in the paper performs well in handling stochastic model updating of nonlinear dynamic systems. By combining Bayesian inference and MCMC techniques and using backbone curves to simplify the likelihood function, it effectively solves the complexity and robustness problems in traditional methods. Experimental verification further proves the effectiveness and applicability of this method.