Improved Gravitational Search Algorithm And Its Application In Structural Damage Identification

With the rapid development of artificial intelligence technology in past decades, a great number of new intelligent optimization algorithms have been proposed and widely used in various fields. As a newest heuristic intelligent optimization algorithm, Gravitational Search Algorithm(GSA) is characterized as a simple concept that is easy to implement with higher computational efficiency and small numbers of parameters to adjust. Since its introduction GSA has been analyzed over time by researchers and modified for solving complex problems in both scientific research world and industrial application fields. On the other hand, the finite element model updating technique has been widely studied and used in the field of structural damage detection in the past 20 years, while the conventional optimization algorithm it used till suffer from premature, low efficiency, and so on. Based on this background, this paper first proposed an improved GSA which is achieved by considering memory strategy and better neighborhood topology. Then for the first time GSA is applied to detection of structural damage, and a finite element model updating method based on the improved GSA is proposed.The research work mentioned in the paper mainly include the following aspects:(1)First,this paper briefly reviewed the research of model-based method of damage detection and the technology of finite element model updating based on intelligent algorithm. Then, from the aspects of theory analysis, algorithm improvement and fusion algorit hm, this paper reviewed the domestic and international research of the gravitational search algorithm.(2)This paper introduces the basic theory, framework and process of GSA. The algorithm is analyzed, and some key factors that affect the performance of the algorithm are analyzed.(3)Memory strategy is introduced into GSA after the analysis of equilibrium point in standard GSA. The results of experiments on six benchmark functions show that GSA with personal memory(GSAP) yields better performance than sta ndard GSA.(4)Based on the analysis of information sharing mechanism of swarm intelligence, the star neighborhood topology of standard GSAP is replaced with the non-clique neighborhood topology(NCNT). First, the performance of NCNT in GSAP is compared with three other typical neighborhood, including the star topology in standard GSAP in the experiments on six benchmark functions. Then, the influence of the number of average neighbors on the performance of NCNT-GSAP is analyzed by the function optimization test. Finally, according to the conclusion proposed of previous test an improved GSAP based on dynamic non-clique neighborhood topology(DNCNT-GSAP) is proposed, related test show that DNCNT-GSAP can significantly improve the ability to escape from the local optimal and enhance the reliability of the algorithm, which improves the potential of engineering application of GSA.(5)In view of the importance of the objective function in the finite element model updating, this paper compares the performance of an objective function based on modal flexibility assurance criterion and the traditional objective function by simulation experiment, in which the damage identification test of a simple supported beam with 10 elements and a Cantilever truss with 30 elements are carried out respectively. The results show that the former objective function can identify the damage extent of damaged elements precisely but fail to locate the damaged elements correctly when measurement data is exposed to noise, while the traditional objective function is more balanced in identifying the position and the damage extent of the damaged elements with better noise immunity. Besides, the two simulation experiment s both suggest that the DNCNT-GSAP yield good performance in damage detection and has a certain engineering application potential.