Parametric identification of flexible beam system using evolutionary algorithm

Abstract

An application of flexible structures in engineering is spread extensively due to lightweight property and technical importance. Before implementing the system, the dynamic behavior of the system needs to be studied by developing a mathematical model. The model-based approach like finite element method which commonly used in modeling usually required a wide knowledge on the system to be studied and involves complex equations. In system identification technique, the conventional parameter estimation is commonly applied and the limitation is, it may cause the solution trapped in local optima and reduce the efficiency of the model. Therefore, in this study, the model of flexible beam is developed by using system identification method which is based on experimental data collected from the experimental rig and using evolutionary algorithm (EAs) as estimation technique. This research provides a new platform for other researcher to develop a model based on system identification technique using EA’s for other system or application. Other than that, it also delivers a basis for future study on analysis of the performance EAs in terms of different parameter settings in comparison with other algorithms. An attempt of obtaining the linear model is accomplished by developing an experimental rig of flexible beam using square wave signal with mixing resonance frequency to collect input-output data. Auto-regressive with exogenous inputs (ARX) is chosen as a model structure of the system. The coefficient parameters of model structure are estimated via EAs such as firefly algorithm and bat algorithm. A few sets of parameter settings for FA and BA are tested to examine the effect of settings to the performance of model. The best models obtained from each estimation method are compared with least squares algorithm and validated using mean square error (MSE) and one step-ahead prediction (OSA). The main result shows that FA-estimation has MSE of 9.46E-5 which is the lowest among all the estimation method and BA-estimation also outperformed LS-estimation (1.16E-2) by getting lower MSE which is 2.70E-4. Overall of this study proved that evolutionary algorithm able to produce better performance than conventional algorithm. Firefly algorithm and bat algorithm is effective and capable to be used in this area of study like other engineering application