Natural algorithms for choosing source locations in active control systems

Abstract

Natural algorithms have been considered as a method of finding optimal secondary loudspeaker positions in an active noise control system. Several genetic algorithms and simulated annealing algorithms were developed and tested for different size searching problems. The performance of each algorithm was investigated and compared with a simple random search. The performance was also tested by using the results obtained from exhaustive searching. It was found that the speed of convergence of the best genetic algorithm and that of the best simulated annealing algorithm were similar, with the latter generally being slightly faster on average. In the case of choosing eight loudspeaker positions from a possible 32 locations, for which there are more than[formula]possible combinations, the simulated annealing program was reliably able to find a set of positions which gave an overall reduction at the microphones which was within 0.5 dB of the best achievable, by searching only about 2000 loudspeaker combinations. The sets of secondary source positions which gave the greatest reduction, without excessive control effort, were then used in experiments in an enclosure, and the results showed good agreement with the theoretical predictions. These secondary source positions were also used in experiments in which the frequency of the primary source was changed slightly and good performance was again observed, indicating that the search procedure is reasonably robust.