Optimal sensor placement in large‐scale dome trusses via Q‐learning‐based water strider algorithm.

Summary: In this study, the Q‐learning algorithm is integrated into the binary water strider algorithm to adaptively control the search operators and repair strategies. The proposed algorithm (QWSA) is applied to optimal sensor placement in structures with numerous candidate locations. To handle the constraint related to the number of sensors, an efficient strategy based on the triaxial driving‐point residue (DPR3) coefficient is developed, in addition to the random solution repair strategy. The reinforcement learning algorithm with softmax policy determines which strategy should be adopted for the generated solutions. Since the utilized sensors are triaxial, the tridimensional modal assurance criterion (TMAC) is applied as the objective function. To verify the effectiveness of the proposed method, two large‐scale dome‐shaped trusses are investigated, and the results are compared with the basic water strider algorithm and several other well‐known and modern metaheuristics in the literature. The results demonstrate that placements obtained by QWSA have in average lower than half of the cost value obtained by compared methods. Additionally, average DPR3 (DPR3¯) of the QWSA is higher than the rest of the algorithms. The Wilcoxon test indicates that, at a 95% confidence level, the results of QWSA are superior to all algorithms. Additionally, the proposed method outperforms them in terms of convergence speed. [ABSTRACT FROM AUTHOR]