粒子群優化演算法應用於感測器最佳化配置問題

Abstract

摘要 結構物因材料老化或受外力作用而造成損壞。所以持續監測結構物的健康狀況是有其必要性，也是近來土木結構工程領域重要的研究議題之一。而感測器系統為結構健康監測系統之關鍵設備。故如何有效的規劃與佈設感測器亦成為結構健康監測系統關鍵探討議題。如何使用有限的感測器便能得知所欲量測的結構行為，此類問題延伸出感測器最佳化配置問題。在眾多仿生演算法中，粒子群優化演算法(Particle Swarm Optimization, PSO)是一個收斂快速的最佳化演算法。本研究之目的即以PSO演算法探討加速度型感測器於結構安全監測系統之最佳化配置問題。研究將探討在固定數量下，如何選定最佳的感測器布設位置。研究將利用FIM (Fisher information matrix)當作粒子群優化演算法的感測器位置適合度指標。為增進PSO演算法之搜尋效能，本研究發展出三種改良式搜尋策略來增加粒子群優化演算法的搜尋能力。研究中藉由四個案例來驗證改良式PSO之效能。研究結果顯示，PSO演算法可求得感測器之最佳布設位置，且適合度指標均分別達18.494、2.1587 x 106、6.3635 x 105與2.1587 x 106。

關鍵字: 結構健康監測系統、感測器最佳化配置問題、粒子群優化演 算法、感測器位置適合度指標FIM (Fisher information matrix)

Abstract Structures may be damaged due to ageing of material or by external force. Continued monitoring of structures of health is a necessity and is recently one of the important research topics in the field of structural engineering. Sensor system for structural health monitoring (SHM) system is significant equipment. Hence, how to effectively plan and arrangement of sensors has become the key research topic in SHM. Meanwhile, how to use limited and want sensors to measure structural behavior is an optimal sensor configuration issues. In numerous bionic algorithms, particle swarm optimization (PSO) algorithm is a fast convergence algorithm for solving optimization problems. The objective of the present study is based on PSO algorithm to optimal placement of acceleration sensors for structural health monitoring system. Under a fixed number of sensors, the study will explore how to select the best sensor location emplacement. The FIM (Fisher information matrix) is employed as a fitness index for PSO algorithm for optimal locating sensors. For improving PSO algorithm, three strategies for improving search were developed. Four case studies were utilized to verify the effectiveness of modified PSO. Simulation results reveal that PSO algorithm can effectively seek out the optimal location emplacement of sensors. Keywords: structural health monitoring (SHM), optimal sensor configuration, particle swarm optimization (PSO) algorithm, FIM (Fisher information matrix)