Finite-element-model Updating Using Computional Intelligence Techniques

Applications to Structural Dynamics

verfasst von: Tshilidzi Marwala

Verlag: Springer London

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

FEM updating allows FEMs to be tuned better to reflect measured data. It can be conducted using two different statistical frameworks: the maximum likelihood approach and Bayesian approaches. This book applies both strategies to the field of structural mechanics, using vibration data. Computational intelligence techniques including: multi-layer perceptron neural networks; particle swarm and GA-based optimization methods; simulated annealing; response surface methods; and expectation maximization algorithms, are proposed to facilitate the updating process. Based on these methods, the most appropriate updated FEM is selected, a problem that traditional FEM updating has not addressed. This is found to incorporate engineering judgment into finite elements through the formulations of prior distributions. Case studies, demonstrating the principles test the viability of the approaches, and. by critically analysing the state of the art in FEM updating, this book identifies new research directions.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Introduction to Finite-element-model Updating

Abstract

This chapter introduces finite-element-model updating. Direct and iterative updating procedures are explained. Some basic features on finite-element modeling are elucidated. Essential elements on vibration testing and analysis are explained and these include the domains in which data can be represented. These domains are in the modal, frequency and time–frequency spaces. Finite-element-model updating techniques are then reviewed and these can be broadly categorized into: matrix update methods, sensitivity-based techniques, iterative optimization procedures, Bayesian methods and computational intelligence techniques. Computational intelligence technques, which are the subject of this book, are then reviewed in detail.

Chapter 2. Finite-element-model Updating Using Nelder–Mead Simplex and Newton Broyden–Fletcher–Goldfarb–Shanno Methods

Abstract

This chapter presents the Nelder–Mead simplex method and the Broyden–Fletcher–Goldfarb–Shanno (BFGS) method for finite-element-model updating. The methods presented have been tested on a simple beam and an unsymmetrical H-shaped structure. It was noted that, on average, the Nelder–Mead simplex method gives more accurate results than did the BFGS method. This is mainly because the BFGS method requires the calculation of gradients, which is prone to numerical errors within the context of finite-element-model updating.

Chapter 3. Finite-element-model Updating Using Genetic Algorithm

Abstract

This chapter implements a genetic algorithm for finite-element-model updating. This method was tested on a simple beam and an unsymmetrical H-shaped structure and compared to a method based on the Nelder–Mead simplex method. It was observed on average that the genetic algorithm method gives more accurate results for modal properties than does the Nelder–Mead (NM) simplex method.

Chapter 4. Finite-element-model Updating Using Particle-swarm Optimization

Abstract

This chapter implements the particle-swarm-optimization method for finite-element-model updating. This method is tested on a simple beam and an unsymmetrical H-shaped structure and compared with a method that is based on a genetic-algorithm optimization technique. It is observed that, on average, the particle-swarm-optimization method gives a more accurately updated finite-element model than does the genetic-algorithm method.

Chapter 5. Finite-element-model Updating Using Simulated Annealing

Abstract

This chapter implements simulated annealing (SA) for updating of a finite-element-model using vibration data. This method was tested on a simple beam and an unsymmetrical H-shaped structure and was compared to a method that used the particle-swarm-optimization method (PSO). It was observed that, on average, the particle-swarm-optimization method gives more accurately updated finite elements than the simulated-annealing method. This is mainly due to the simplicity of its implementation.

Chapter 6. Finite-element-model Updating Using the Response-surface Method

Abstract

This chapter presents the response-surface method for finite-element-model updating. The response-surface method was implemented by approximating the finite-element surface-response equation by a multi-layer perceptron, which is a neural-network technique. The updated parameters of the finite-element model were calculated using a genetic algorithm to optimize the surface-response equation. The presented method is compared to existing methods that use simulated annealing and a genetic algorithm separately, with a full finite-element model for model updating. The presented method was tested on a simple and an unsymmetrical H-shaped structure. It was observed that the presented method gave the updated natural frequencies and mode shapes that were an improvement to the initial finite-element model. In general, the accuracy was of the same order of magnitude as those given by the simulated annealing and a genetic algorithm. Furthermore, it was observed that the response-surface method achieves these results with a computational speed that, on average, was at least twice as fast as a genetic algorithm and a full finite-element model and at least twenty times faster than simulated annealing.

Chapter 7. Finite-element-model Updating Using a Hybrid Optimization Method

Abstract

This chapter presents a hybrid of particle-swarm optimization and the Nelder–Mead simplex optimization method for finite-element-model updating. It was observed, on average, that the hybrid of particle-swarm optimization and the Nelder–Mead simplex optimization method gives more accurate results, followed by the particle-swarm optimization and then the Nelder–Mead simplex method.

Chapter 8. Finite-element-model Updating Using a Multi-criteria Method

Abstract

In this chapter a multiple criteria method (MCM) is presented and tested for finite-element-model updating using a simple beam with holes and an irregular H-shaped structure. The MCM minimizes the Euclidean norm of the error matrix that combines the modal property data with the frequency-response function data. The results given by the MCM are compared with the results from a frequency-response function method (FRFM), and those obtained from the modal property method (MPM). The three methods were compared based on their ability to reproduce the measured parameters. It was observed, on average, that the MCM gives the best results.

Chapter 9. Finite-element-model Updating Using Artificial Neural Networks

Abstract

This chapter implements Bayesian neural networks for finite-element-model updating. This method was tested on a simple beam and an unsymmetrical H-shaped structure and compared with an implementation that was based on the response-surface method. It was observed, on average, that the Bayesian neural-network approach gave more accurate results than the response-surface method did.

Chapter 10. Finite-element-model Updating Using a Bayesian Approach

Abstract

This chapter implements a Bayesian approach to finite-element-model updating. The Bayesian formulation was solved using the Markov chain Monte Carlo (MCMC) technique, as well as genetic programming based on the MCMC. These methods were tested on a simple beam and an unsymmetrical H-shaped structure. It was observed on average that the genetic programming MCMC performed better than the MCMC method.

Chapter 11. Finite-element-model Updating Applied in Damage Detection

Abstract

This chapter presents a multiple criterion method (MCM) that was tested in damage detection of a simple beam with holes and an irregular H-shaped structure. The MCM was compared with the frequency-response function method (FRFM) and the modal property method (MPM) in terms of their abilities to detect damage in structures. The MCM and FRFM methods were generally found to be able to identify damage better than the MPM.

Chapter 12. Conclusions and Emerging State-of-the-art

Abstract

This book dealt with the use of computational intelligence methods for finite-element-model updating. Finite-element-model updating is a process through which finite-element models are tuned to better reflect the measured data. This is based on the rational assumption that the measured data are more reliable than the finite-element-model’s predicted data.

Backmatter

Titel: Finite-element-model Updating Using Computional Intelligence Techniques
verfasst von: Tshilidzi Marwala
Verlag: Springer London
Electronic ISBN: 978-1-84996-323-7
Print ISBN: 978-1-84996-322-0
DOI: https://doi.org/10.1007/978-1-84996-323-7