摘要
Binary particle swarm optimization algorithm(BPSOA) has the excellent characters such as easy to implement and few set parameters.But it is tendentious to stick in the local optimal solutions and has slow convergence rate when the problem is complex.Cultural algorithm(CA) can exploit knowledge extracted during the search to improve the performance of an evolutionary algorithm and show higher intelligence in treating complicated problems.So it is proposed that integrating binary particle swarm algorithm into cultural algorithm frame to develop a more efficient cultural binary particle swarm algorithm (CBPSOA) for fault feature selection.In CBPSOA,BPSOA is used as the population space of CA;the evolution of belief space adopts crossover,mutation and selection operations;the designs of acceptance function and influence function are improved according to the evolution character of BPSOA.The tests of optimizing functions show the proposed algorithm is valid and effective.Finally,CBPSOA is applied for fault feature selection.The simulations on Tennessee Eastman process (TEP) show the CBPSOA can perform better and more quickly converge than initial BPSOA.And with fault feature selection,more satisfied performance of fault diagnosis is obtained.
Binary particle swarm optimization algorithm （BPSOA） has the excellent characters such as easy to implement and few set parameters. But it is tendentious to stick in the local optimal solutions and has slow convergence rate when the problem is complex. Cultural algorithm（CA） can exploit knowledge extracted during the search to improve the performance of an evolutionary algorithm and show higher intelligence in treating complicated problems. So it is proposed that integrating binary particle swarm algorithm into cultural algorithm frame to develop a more efficient cultural binary particle swarm algorithm （CBPSOA） for fault feature selection. In CBPSOA, BPSOA is used as the population space of CA; the evolution of belief space adopts crossover, mutation and selection operations; the designs of acceptance function and influence function are improved according to the evolution character of BPSOA. The tests of optimizing functions show the proposed algorithm is valid and effective. Finally, CBPSOA is applied for fault feature selection. The simulations on Tennessee Eastman process （TEP） show the CBPSOA can perform better and more quickly converge than initial BPSOA. And with fault feature selection, more satisfied performance of fault diagnosis is obtained.
基金
National High Technology Research and Development Program of China(No.2007AA04Z171)
