A new metaheuristic optimization based on K-means clustering algorithm and its application to structural damage identification

Abstract

This paper develops a new metaheuristic optimization algorithm named K-means Optimizer (KO) to solve a wide range of optimization problems from numerical functions to real-design challenges. First, the centroid vectors of clustering regions are established at each iteration using K-means algorithm, then KO proposes two movement strategies to create a balance between the ability of exploitation and exploration. The decision on the movement strategy for exploration or exploitation at each iteration depends on a parameter that will be designed to recognize if each search agent is too long in the region visited with no self-improvement. To demonstrate the effectiveness and reliability of KO, twenty-three classical benchmark functions, CEC2005 and CEC2014 benchmark functions, are employed as a first example and compared with other algorithms. Then, three well-known engineering problems are also considered and their results are compared to the results obtained by the other algorithms. Finally, KO is applied to structural damage identification (SDI) problem of a complex 3D concrete structure including seven stories building having a 25.2 m total height. For this purpose, SAP2000 is used to solve the finite element (FE) model of this structure. Then, for the first time, we successfully developed a sub-program that allows two-way data exchange between SAP2000 and MATLAB through the Open Application Programming Interface (OAPI) library to update the FE model. From the results, we found that KO has the best performance for the considered benchmark functions based on the Wilcoxon rank-sum test and Friedman ranking test. The results obtained in this work have proved the effectiveness and reliability of KO in solving optimization problems, especially for SDI. Source codes of KO is publicly available at http://goldensolutionrs.com/codes.html.