工程设计学报2025，Vol.32Issue(2)：141-150,10.
基于聚类椭球模型的机械系统不确定性分析方法
Uncertainty analysis method for mechanical system based on multi-cluster ellipsoidal model
摘要
Abstract
Multi-source uncertainties,such as random loads,manufacturing errors,and installation inaccuracies,influence mechanical system during the service processes,resulting in the deviation of system response,which seriously affects the operation stability and system reliability.In practical engineering,it is very difficult to collect high-quality data of complex mechanical systems on a large scale.For this reason,the sample size of the obtained data is usually small,and its uncertainty is difficult to quantify using probability models.Furthermore,when mechanical systems operate under certain specific conditions,the data may cluster in certain areas,exhibiting distinct clustering characteristics,which leads to challenges for the uncertainty analysis and measurement of mechanical system under these conditions.To address the above issues,an uncertainty analysis method for mechanical systems based on a multi-cluster ellipsoidal model was proposed to accurately quantify the uncertainties of system parameters with clustering characteristics,so as to achieve the rapid evaluation of system response uncertainty.In order to accurately quantify the small sample data with clustering characteristics,the multi-cluster ellipsoidal model was employed for uncertainty modeling.Based on the parameter interval information,the sensitivity analysis of the mechanical system was conducted to determine the main parameters that affected the performance of the mechanical system.The upper and lower boundaries of the mechanical system response were obtained by using multi-cluster ellipsoidal model and the sequential quadratic programming algorithm to achieve the uncertainty propagation of system parameters.The accuracy and effectiveness of the proposed method were verified through three numerical examples and one radar system engineering example.The research results indicate that the hyper-ellipsoidal clustering method has high computational efficiency and accuracy when addressing the uncertainty of mechanical system performance under limited samples.关键词
不确定性传播/聚类椭球模型/功能度量法/敏感性分析Key words
uncertainty propagation/multi-cluster ellipsoidal model/performance measure approach/sensitivity analysis引用本文复制引用
马正琰,欧阳衡,郝志杰,高硕,刘宝会..基于聚类椭球模型的机械系统不确定性分析方法[J].工程设计学报,2025,32(2):141-150,10.基金项目
国家自然科学基金资助项目(12102022,52475252) （12102022,52475252）
河北省高等学校科学研究项目产学研合作专项(CXY2024005) （CXY2024005）
河北省自然科学基金创新群体项目(E2024202298) （E2024202298）
