On Model Updating Using Neural Networks

Abstract

Key parameters in dynamic systems often change during their life cycle due to repair and replacement of parts or environmental changes. This paper presents a new approach to account for these changes by updating the system models. Current iterative methods developed to solve the model updating problem rely on minimisation techniques to find the set of model parameters that yield the best match between experimental and analytical responses. These minimisation procedures require considerable computation time, making the existing techniques infeasible for some applications, such as in an adaptive control scheme, correcting the model parameters as the system changes. The proposed approach uses frequency domain data and a neural network to estimate the updated parameters quickly, yielding a model representative of the measured data. Besides control-related applications, this may also be of use for manufacturing systems, where parameters change during operation requiring repeated updates of the nominal model. Numerical simulations and experimental results show that the neural network updating method (NNUM) has good accuracy and generalisation properties, and it is therefore a suitable alternative for the solution of the model updating problem of this class of systems.

Selection of Training Samples for Model Updating Using Neural Networks

“…The generated samples should re#ect the probability distribution of the parameters to be considered. Atalla and Inman [4] reported that a random generation of samples yields the best result for the NN training.…”

“…They are the full factorial (FF), the hypercube (HC) [22], the linear (LI) [16], and the random (RA) method [4,16].…”

“…Using NN methods for model updating and damage detection, Atalla and Inman [4] suggested that a random generation (within the assumed boundaries of variation) of the model parameters yields the best results. The training samples should re#ect the probability distribution of the model parameters being updated if this information is known.…”

“…The behaviors of concrete in the state of plane stress under monotonic biaxial loading and compressive uniaxial cycle loading are modelled by Ghaboussi et al [3] using a BP NN model. The use of frequency domain data and NN's to update structural parameters has been studied by, among others, Atalla and Inman [4] and Levin and Lieven [5]. Stephens and VanLuchene [6] used NN to assess the safety condition of civil engineering structures following strong-motion earthquakes.…”

Structural Damage Detection Using Artificial Neural Networks and Measured FRF Data Reduced via Principal Component Projection

“…To avoid large neural networks, one can use a subset of the available FRF data by considering fewer measurement points and by de"ning frequency windows. Atalla and Inman [6] updated a 15-DOF lumped-parameter system and a #exible frame structure using an RBF neural network, the input to which consisted of pre-selected frequency points from a number of frequency windows. An alternative approach, based on modal parameter input, is described by Levin and Lieven [7].…”