A Bee Swarm Algorithm for Optimising Sensor Distributions for Impact Detection on a Composite Panel

This paper describes the development of a bee swarm algorithm for optimising the distribution of impact detection sensors on a composite plate. The algorithm was initially developed and tested on a travelling salesman problem, it was then adapted to solve the sensor placement problem using an artificial neural network to assess the fitness of sensor distributions. It performed well, managing to quickly find optimum distributions within a constrained set of neural network parameters. The algorithm was modified further, to optimise the neural network parameters alongside the sensor distributions and the range that the network parameters could take was substantially increased. This led to significant increases in accuracy. The algorithm showed itself to be a very effective tool in the sensor optimisation problem as well as demonstrating the benefits of thoroughly optimising the neural network parameters.