摘要
针对转子裂纹参数定量识别精度不足的问题,提出一种基于Kriging组合模型和非支配解排序遗传算法Ⅲ型(NSGA-Ⅲ)的转子裂纹参数识别方法。首先,基于动力学模型响应建立不同相关函数的Kriging代理模型作为候选模型库;其次,通过逐步回归模型筛选策略剔除性能不佳的模型并依据启发式权重加权获得最佳Kriging组合模型;再次,应用所提Kriging组合模型建立裂纹参数与裂纹转子系统响应的关系;最后,通过Kriging组合模型预测的响应幅值与实际响应幅值的差异构成多目标函数,利用NSGA-Ⅲ得到识别结果。结果表明:采用该方法识别裂纹参数的最大相对误差为2.5%;相比于基于单个Kriging模型和指数函数-高斯函数组合模型的识别方法,所提方法具有较高的适用性和精确性。
In order to solve the problem of insufficient quantitative identification accuracy of rotor crack parameters,a rotor crack parameter identification method based on Kriging combined model and non-dominated solution sequencing genetic algorithm type III(NSGA-III)was proposed.Firstly,Kriging conbined models with different correlation functions were established as the candidate model library based on the dynamic model response.Secondly,the model with poor performance was eliminated by the stepwise regression model screening strategy,and the optimal Kriging combined model was obtained according to the heuristic weight weighting.Thirdly,the relationship between the crack parameters and the response of the cracked rotor system was established by using the proposed Kriging combined model.Finally,a multi-objective function was constituted according to the difference between the predicted response amplitude of Kriging combined model and the actual response,and the recognition results were obtained by NSGA-III.Results show that the maximum error of crack parameter identification by this method is 2.5%.Compared with the recognition method based on single Kriging model and exponential function-Gaussian function combined model,the proposed method has higher applicability and accuracy.
作者
胡楷
马军
王晓东
陈虹潮
HU Kai;MA Jun;WANG Xiaodong;CHEN Hongchao(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650500,China)
出处
《动力工程学报》
CAS
CSCD
北大核心
2023年第11期1506-1514,共9页
Journal of Chinese Society of Power Engineering
基金
国家自然科学基金资助项目(62163020、62173168)
云南省基础研究计划资助项目(202101BE070001-055)。
关键词
转子系统
Kriging代理模型
逐步回归
裂纹
参数识别
三代遗传算法
rotor system
Kriging surrogate model
stepwise regression
crack
parameter identification
the third generation genetic algorithm
