Adaptive noise cancellation based on beehive pattern evolutionary digital filter

• Beehive pattern evolutionary rules are developed to overcome low convergence speed during global optimum estimation with cloning and mating methods.
• The adaptive noise cancellation algorithm based on the Beehive Pattern Evolutionary Digital Filter (BP-EDF) is proposed.
• The proposed algorithm enables the global optimum estimation when detecting feature impulse fault signal from the heavy noise contaminated condition.
• Convergence speed and the noise cancellation performance of the proposed algorithm BP-EDF are validated by simulation and experiment.

Evolutionary digital filtering (EDF) exhibits the advantage of avoiding the local optimum problem by using cloning and mating searching rules in an adaptive noise cancellation system. However, convergence performance is restricted by the large population of individuals and the low level of information communication among them. The special beehive structure enables the individuals on neighbour beehive nodes to communicate with each other and thus enhance the information spread and random search ability of the algorithm. By introducing the beehive pattern evolutionary rules into the original EDF, this paper proposes an improved beehive pattern evolutionary digital filter (BP-EDF) to overcome the defects of the original EDF. In the proposed algorithm, a new evolutionary rule which combines competing cloning, complete cloning and assistance mating methods is constructed to enable the individuals distributed on the beehive to communicate with their neighbours. Simulation results are used to demonstrate the improved performance of the proposed algorithm in terms of convergence speed to the global optimum compared with the original methods. Experimental results also verify the effectiveness of the proposed algorithm in extracting feature signals that are contaminated by significant amounts of noise during the fault diagnosis task.