Differential Evolution for Finite Element Model Updating: Algorithm and Application in Structural Analysis

• Published in: E-journal of Nondestructive Testing Vol. 29; no. 7
• Main Author: Abu Abed, Wassim
• Format: Journal Article
• Language: English
• Published: 01.07.2024
• ISSN: 1435-4934; 1435-4934
• DOI: 10.58286/29656

The availability of a numerical model that accurately reproduces the behaviour of an existing structure is paramount for diagnostic and prognostic purposes in Structural Health Monitoring. An important procedure to this end is updating the parameters of the numerical model on the basis of measured data. This process ideally involves the use of an optimisation method that adapts the parameters of the model so that the discrepancy between the numerically calculated structural response and the measured one is minimised. In this paper a procedure for updating finite element models for the application in structural analysis is presented. The updating algorithm uses a simple and efficient adaptive population based method for global optimization over continuous spaces known as Differential Evolution. Continuous interpolation functions are utilised in order to reduce the dimension of the search space resulting in speeding up the convergence without jeopardising the reliability of finding the solution or its accuracy. Three different examples with artificial datasets are used to verify the updating algorithm. First, a static finite element model of a 10 m single span beam consisting of beam elements with linear elastic material behaviour is considered. The second example consists of a two-dimensional single-span reinforced concrete road bridge consisting of slab elements. In both examples, the distribution of the modulus of elasticity of the elements are updated out of displacement data. The third example involves the dynamic analysis of a train crossing. A load case of an ICE1 crossing over a 32 m single-span bridge consisting of beam elements is assumed. The objective of this example is updating the distribution of the modulus of elasticity of the elements out of acceleration data. The paper concludes with a discussion of the advantages and disadvantages of the developed updating algorithm. An overview of outstanding issues and improvements is also given.