Optimal multiaxial sensor placement for modal identification of large structures

Research on optimal sensor placement has become a very important topic because of the need to obtain effective testing results with limited testing resources in modal identification and structural health monitoring. An integer-encoding multi-swarm particle swarm optimisation (IMPSO) algorithm is proposed to place multiaxial sensors optimally on large structures for modal identification. The concepts of grade evaluation and migration strategy and the mutation operators of genetic algorithm are introduced into the integer-encoding particle swarm optimisation algorithm. Three different fitness functions for optimal multiaxial sensor placement (OMSP) are investigated. The second fitness function considers spatial correlation based on Moran's I to solve the information redundancy of multiaxial sensor placement, whereas the other two functions evolve from the existing methods for comparison with the second fitness function. The novel algorithm and three fitness functions are further applied to the Laxiwa arch dam for verifications. The results show that the proposed IMPSO outperforms two existing algorithms in its global optimisation capability. The results also prove that the second fitness function has advantages in sensor distribution and ensuring the well-conditioned information matrix and observability of multidimensional modal shapes. The multiaxial sensor placement scheme determined by the proposed method is applied to the modal test of the Laxiwa arch dam under simulative ambient excitation. The results show that the scheme determined by the second fitness function can identify the frequencies and multidimensional mode shapes accurately, indicating that this method may be used to provide guidance for OMSP in various types of large structures. Copyright (c) 2013 John Wiley & Sons, Ltd.