Genetic Algorithm–Based Acoustic-Source Inversion Approach to Detect Multiple Moving Wave Sources of an Arbitrary Number

Acoustic-source inversion in the time domain can be used for a broad range of engineering applications [for example, detecting speeds and weights of overweight trucks on urban roads and locating underwater moving acoustic sources (i.e., naval surveillance)]. However, past studies have not shown whether it is feasible to determine the profiles of moving wave sources without knowing the number of target sources. To fill this knowledge gap, this work presents a genetic algorithm (GA)-based acoustic-source inversion algorithm to identify the amplitudes, positions, and speeds of multiple moving sources by using wave motions recorded at sensors within a heterogeneous one-dimensional solid bar system without knowing the number of targets. To this end, the authors cast this problem into an inverse-source problem, in which solutions can be obtained via minimization of a misfit between measured wave responses and numerically computed responses found using estimated parameters. The wave responses due to moving sources are computed by using the finite element method (FEM) in the time domain. The presented FEM modeling successfully captures the shifts of frequency contents (i.e., the Doppler effect) of wave responses that are generated by moving wave sources. The numerical results, for which the GA is used for the minimization process, show that it is possible to detect the parameters of moving sources in a heterogeneous host without knowing how many targeted sources exist in advance. It is also observed that greater computational cost (i.e., iteration number of the GA) and denser distribution of sensors are required to detect a larger number of moving wave sources. Similarly, larger computational costs are required for a heterogeneous host than a homogeneous one.