摘要
考虑随机因素的影响,对叶片固有振动特性进行概率分析,并判断其对随机参数的敏感性是叶片动强度可靠性设计的基础。以某试验台用汽轮机等直叶片为研究对象,考虑几何参数(长度、宽度、厚度)、材料参数(弹性模量,密度)和转速的随机性,运用统计学习理论,将确定性有限元、径向基函数(RBF)神经网络和MonteCarlo模拟法相结合,得到了叶片静(动)频率的统计参数和累积分布函数。采用概率敏感性分析方法,定量地判断出叶片静、动频率对随机输入变量的敏感性,分析结果对工程实际具有一定的指导意义。此外,还将计算结果与响应面方法进行了比较,得出该方法较响应面方法更加快捷,可作为叶片动强度可靠性分析的可选方法。
With the chance factors being taken into account,a probabilistic analysis of the inherent vibration frequency of blades has been conducted and a proposition,made that the sensitivity of random parameters constitutes a basis for the dynamic strength reliability design of the blades.With the straight blades of a steam turbine in a test rig serving as an object of study,the randomness of geometrical parameters(including length,width,thickness),material parameters(elastic modulus,density) and rotating speed was taken into consideration.On this basis,a statistical learning theory was applied to obtain the statistical parameters and accumulative distribution function of static(dynamic) frequencies of blades by an integration of the deterministic finite element and radial basis function(RBF) with Monte Carlo simulation method.By adopting a probabilistic sensitivity analytic method,the authors have made a quantitative assessment of the sensitivity of blade static and dynamic frequencies to random input variables.The analytic results can provide positive guidelines for general engineering practice.Furthermore,the authors have compared the calculation results with those obtained by a response surface method,and concluded that the analytic method in question offers a quicker approach than the response surface method.It can serve as an alternative method for the dynamic strength reliability analysis of blades.
出处
《热能动力工程》
EI
CAS
CSCD
北大核心
2008年第5期453-458,共6页
Journal of Engineering for Thermal Energy and Power
基金
河北省教育厅科技基金资助项目(Z2001211)
关键词
统计学习理论
叶片
静(动)频
概率设计
概率敏感性分析
RBF神经网络
statistical learning theory,blade,static(dynamic) frequency,probabilistic design,probabilistic sensitivity analysis,radial basis function(RBF) neural network
