Kölzow, Felix (2021)
Application of Probabilistic Methods for Lifetime Prediction of High Temperature Components under Creep-Fatigue Loading.
Technische Universität Darmstadt
doi: 10.26083/tuprints-00018616
Dissertation, Erstveröffentlichung, Verlagsversion

Kurzbeschreibung (Abstract)

Due to intermittent renewable energy share, precise lifetime assessment procedures are necessary to increase the power plant reliability and flexibility, while steam turbine components are exposed to creep-fatigue loading. Simultaneously, approaches are necessary for an objective risk assessment while quantifying the failure probability. Different probabilistic methods exist which could allow to estimate material uncertainties, compute the probability of crack initiation and generate optimised experimental designs. Another component is the falsification for a stepwise improvement of the lifetime approaches and probabilistic methods provide tools for this kind of procedure. This is the reason why probabilistic methods are tried to combine with accumulative lifetime methods for creep-fatigue loading. The generalised damage accumulation rule is combined with probabilistic methods to quantify the probability of crack initiation for high temperature components under creep-fatigue loading, while considering three types of materials (X12CrMoWVNbN10-1-1, GX12CrMoVNbN9-1, 30CrMoNiV5-11). A temperature modified Manson-Coffin-Basquin relationship is proposed to take a broader temperature range during the low-cycle fatigue assessment into account and to gain statistical information. The Wilshire-Scharning creep rupture model is examined and, in combination with the maximum likelihood method, allows the creep assessment procedure to take into account ongoing experiments. The reliability method FORM can be applied in conjunction with the generalised damage accumulation rule to compute the failure probability. The used HRLF algorithm does not indicate convergence problems. A Taylor series perspective and a sensitivity analysis indicates that the critical damage threshold is one of the most crucial parameters. Additionally, different workloads are compared to each other and a comparison between a deterministic and probabilistic lifetime assessment approach is shown. Conceptional difficulties of accumulative lifetime assessment methods and the missing possibility to obtain statistical information leads to the investigation of the continuum damage mechanics approach to combine it with sophisticated probabilistic methods. Different end-of-life criteria (load drop, change of elastic behaviour) are compared to each other. A concise definition of damage is necessary and the temperature dependency of critical damage threshold is investigated. To obtain probabilistic results while using damage mechanics lifetime predictions, the polynomial chaos expansion is employed. The Sermage damage model in combination with a non-intrusive regression technique is used to show exemplary probabilistic results and potential problems, while a probabilistic service-type cycle is considered. Long-term creep rupture experiments show partially low influence measures in the linear regression analysis. Thus, the last part of this work considers the sequential and exchange algorithm proposed by Fedorov to obtain an experimental design that leads to a lower number of experiments while reducing the uncertainty in the parameter estimation. The construction of nested experimental design spaces is introduced to achieve this goal.

Frage zum Eintrag

Redaktionelle Details anzeigen