CAJ | 학술논문

为了准确诊断直升机旋翼不平衡故障，提出了一种基于粒子群算法和广义回归神经网络模型(PSO-GRNN)的故障诊断方法。将交叉验证得到的平均均方误差作为粒子群的适应度函数，运用粒子群算法搜寻最优的GRNN光滑因子，建立最优的故障诊断模型。结果表明：采用PSO-GRNN模型可实现直升机旋翼不平衡的类型和程度的有效诊断，故障类型准确率高达94.29%，故障程度的诊断最大误差仅6.54%，满足工程需求。
위료준학진단직승궤선익불평형고장，제출료일충기우입자군산법화엄의회귀신경망락모형(PSO-GRNN)적고장진단방법。장교차험증득도적평균균방오차작위입자군적괄응도함수，운용입자군산법수심최우적GRNN광활인자，건립최우적고장진단모형。결과표명：채용PSO-GRNN모형가실현직승궤선익불평형적류형화정도적유효진단，고장류형준학솔고체94.29%，고장정도적진단최대오차부6.54%，만족공정수구。
In order to diagnose the helicopter rotor’s unbalance fault accurately, a method based on particle swarm opti-mization (PSO) algorithm and generalized regression neural network (PSO-GRNN) was proposed. The average mean square error obtained from cross validation was used as the fitness function of the particle swarm. Then, the optimal GRNN smooth factor was attained by using the PSO algorithm, and an optimal model for fault diagnosis was achieved. It can be concluded that based on the PSO-GRNN model, the type and the extent of the helicopter rotor’s unbalance can be diagnosed effective-ly, the accuracy rate of fault type is up to 94.29%and the maximum error of fault degree is only 6.54%, which satisfies the requirement of engineering projects perfectly.