CAJ | 학술논문

提出基于优胜劣汰、步步选择的粒子群优化算法（SSPSO），弥补了一般粒子群优化算法容易陷入局部极值、早熟收敛或停滞的缺陷。并运用 SSPSO 对广义回归神经网络（GRNN）平滑参数 P 进行优化，充分利用 SSPSO 寻优能力强及径向基函数调整参数少的优点，建立厂房结构的振动响应预测模型，对某厂顶溢流式水电站的厂坝结构振动响应问题展开预测研究。通过分析预测效果得出：与一般的粒子群算法相比，所提出的 SSPSO 算法的寻优能力得到了很大的提高。与此同时，基于 SSPSO 优化的广义回归神经网络（SSPSO －GRNN）与其他网络相比，在预测精度、收敛性能、泛化能力等各个方面得到了很大提升。为水电站厂房振动响应预测提供了新的方法和思路，为增强厂房结构的智能化监测提供了保障。
제출기우우성렬태、보보선택적입자군우화산법（SSPSO），미보료일반입자군우화산법용역함입국부겁치、조숙수렴혹정체적결함。병운용 SSPSO 대엄의회귀신경망락（GRNN）평활삼수 P 진행우화，충분이용 SSPSO 심우능력강급경향기함수조정삼수소적우점，건립엄방결구적진동향응예측모형，대모엄정일류식수전참적엄패결구진동향응문제전개예측연구。통과분석예측효과득출：여일반적입자군산법상비，소제출적 SSPSO 산법적심우능력득도료흔대적제고。여차동시，기우 SSPSO 우화적엄의회귀신경망락（SSPSO －GRNN）여기타망락상비，재예측정도、수렴성능、범화능력등각개방면득도료흔대제승。위수전참엄방진동향응예측제공료신적방법화사로，위증강엄방결구적지능화감측제공료보장。
Particle swarm optimization (PSO ) algorithm is easy to fall into local extremum and premature convergence.To overcome defects of PSO,a new kind of PSO based on the survival of the fittest and step by step selection (SSPSO)was proposed here.Then,SSPSO was used to optimize smoothness parameter P of generalized regression neural network(GRNN).The advantages of the strong optimization ability of SSPSO and fewer parameters of GRNN were fully used.Then,the vibration response prediction model based on SSPSO-GRNN for a power-house structure was constructed based on the study data of a certain crest overflow hydropower station.The predicted results showed that the optimization capability of SSPSO is greatly improved compared with PSO;at the same time,the prediction accuracy,convergence performance and generalization ability of SSPSO-GRNN are better than those of other networks.The study results provided a new method for vibration response prediction of hydropower station houses to enhance their intelligent monitoring.