Enhancement of Quality of Modal Test Results of an Unmanned Aerial Vehicle Wing by Implementing a Multi-Objective Genetic Algorithm Optimization

Abstract

Due the fact that aircraft structures work in an environment with lots of dynamic forces, it is of vital importance to perform a dynamic analysis to understand dynamic characteristics of aircraft in that specific environment. These characteristics are usually obtained using numerical methods (finite element analysis) or experimental methods (classical modal analysis). In classical modal analysis, quality of test equipment plays a critical role in final results' accuracy and completeness. There is another important factor which is expertise of a test engineer. Test engineer uses his/her experience to find sufficient/optimum numbers, types and locations of transducers. This process sometimes would be time consuming and exhausting which results in degradation of test results quality. In this paper an algorithm is developed and implemented to find numbers, types and locations of transducers in a modal test which will make results of test more reliable. In this study, an unmanned aerial vehicle used as dummy structure to test functionality of developed algorithm. This algorithm utilized two toolboxes from MATLAB (multi-objective genetic algorithm toolbox and parallel computing toolbox) and MSC (c) NASTRAN finite element solver. A genetic algorithm based optimization is performed in which MSC (c) NASTRAN was used to calculate dynamic characteristics of UAV wing. Since this was a time and resource consuming process a parallel computing cluster is also utilized which decreased run times at least fourfold. In algorithm it was tried to find optimum numbers, types and locations of transducers which will result in minimum cost and error in test results. Error was defined as a summation of mode shape observability error, mass loading error and optimum driving point error. At the end of study optimization results are presented and validated by classical modal analysis. (C) 2017 Elsevier Masson SAS. All rights reserved.