涡轮轴低周疲劳寿命可靠性分析及优化设计方法研究

Probability Analysis and Reliability Based Design Optimization Methods for Low Cycle Fatigue Life of Turbine Shaft

关键词:  涡轮轴  低周疲劳  结构分析  可靠性分析  优化设计  参数化平台

Key words:  Turbine shaft  Low cycle fatigue  Structure analysis  Reliability analysis  Optimal design  Parameterized platform

摘要:

为保障涡轮轴在多种不确定性因素影响下的可靠性和寿命性能，建立了涡轮轴低周疲劳寿命可靠性分析与优化设计方法。搭建了涡轮轴结构分析、可靠性分析和可靠性优化设计的参数化平台，实现了设计参数的不同访问值处结构分析和可靠性分析的自主调用。提出了改进Monte Carlo结合自适应Kriging的算法（A-MCS-AK），通过多点加点和样本池缩减大幅提高了可靠性分析效率。针对极大化涡轮轴低周疲劳寿命均值和极小化涡轮轴低周疲劳寿命失效概率的可靠性优化模型，建立了基于自适应协作代理策略的类序列解耦算法。对选定涡轮轴的分析结果表明，搭建的参数化平台实现了分析数据的自主流动，提出的A-MCS-AK算法较传统的自适应Kriging结合MCS方法计算效率更高，建立的类序列解耦法实现了两种优化模型的高效求解，得到的优化方案显著提高了某型涡轮轴的寿命均值并降低了失效概率。

Abstract:

In order to ensure reliability and increase low cycle fatigue life （LCFL） of turbine shaft under various random uncertainties， the reliability analysis and reliability based design optimization （RBDO） were studied. A parameter platform is established for structure analysis， reliability analysis and RBDO of the LCFL of turbine shaft. Based on this platform， calling structure finite element software and reliability analysis can be automatically realized at different design parameters visited by RBDO iteration. For improving efficiency of reliability analysis greatly， an advanced Monte Carlo simulation combined with adaptive Kriging model （A-MCS-AK） is proposed by strategy of multi-training-point at one updating and candidate sample pool reduction. For two RBDO models of respectively maximizing average expectation of LCFL and minimizing failure probability of LCFL of the turbine shaft， a quasi-sequential decoupling method is presented by combining cooperatively adaptive surrogate. The analysis results of the turbine shaft show that the parameterized platform completes the automatic and orderly transmission of data， and the proposed A-MCS-AK is more efficient than traditional adaptive Kriging combined with MCS （AK-MCS） for reliability analysis. The solutions of two RBDO models of the turbine shaft LCFL show that cooperatively adaptive surrogate strategy can improve the efficiency of solving RBDO under acceptable precision， and the average LCFL and the reliability can be improved simultaneously.