本文已被：浏览 806次 下载 353次
Received:June 09, 2020
Received:June 09, 2020
中文摘要: 涡轮叶片是燃气轮机中工作环境最恶劣的部件，叶片状态的实时监测与诊断至关重要。针对XGBoost（eXtreme gradient boosting）易受其超参数的影响，提出了一种改进萤火虫算法（IFA）优化XGBoost的故障诊断方法。将种群多样性的位置更新策略和动态步长更新措施引入萤火虫算法（FA）中，解决其收敛速度慢、易陷入局部最优等问题，使其能够更好地确定模型参数。实验结果表明，与未优化的XGBoost模型和FA-XGBoost模型相比，建立的IFA-XGBoost模型具有很好的识别效果，准确率达到96.87%，能更好地应用于叶片故障诊断。
Abstract:Turbine blade is the worst part of gas turbine working environment.The real-time detection and diagnosis of the blade state is crucial.In view of the susceptibility of eXtreme gradient boosting (XGBoost) to its hyper-parameters,a method based on improved firefly algorithm (IFA) and XGBoost is proposed.In order to solve the problems such as the slow convergence speed and easy falling into local optimum of the FA,the location update strategy and dynamic step update measures of population diversity are introduced,which can better determine the model parameters.The results show that compared with XGBoost and FA-XGBoost,IFA-XGBoost has characteristics of 96.87% classification accuracy,which can be better applied to gas turbine fault diagnosis.
keywords: gas turbine fault diagnosis firefly algorithm XGBoost
文章编号：20214010 中图分类号：TP181 文献标志码：
基金项目:上海市科技创新行动计划地方院校能力建设专项项目（19020500700）；上海市科委发电过程智能管控工程技术研究中心基金资助项目（14DZ2251100）。
Reference text: