Research On Hybrid Probabilistic Physics Of Failure Modeling And Fatigue Life Estimation Of High-Temperature Structures

With the rapid development of larger, more complex and precise modern mechanical equipments, some special requirements for designing these complex structures, i.e. mechanisms and reliability, have been brought forward correspondingly. Examples of such structures are the turbine disks used in gas turbine aero-engines and/or heavy duty gas turbines. During operation these disks are subjected to high temperature, pressure and rotating speed conditions, for which a failure could lead to catastrophic results. The life and reliability of these complex structures have been one of the primary factors that restricted the development of major mechanical equipments.For turbine disks during service, the demands of the start-up, normal operation and shut-down phases result in different failure modes, like fatigue and creep which are affected by many uncertainties typically exhibit random behavior. Low cycle fatigue (LCF) at high temperature is a key failure mode of these structures. In order to reduce weight and improve the working life while keeping or increasing reliability, an accurate algorithm for probabilistic LCF life prediction of complex structures is essential, which is the main purpose of this contribution.LCF at high temperature is an interactive mechanism of different processes such as time-independent plastic strain, time-dependent creep and environmental corrosion, and the complex interaction between them. These damage mechanisms under multi-environmental factors make it difficult to predict LCF life using a unified model that can make accurate life prediction for fatigue-creep interaction. According to the complexity, uncertainty and scatter in the fatigue failure mechanism, the overall objective of this dissertation is to address key challenges and critical issues in life prediction and reliability analysis on major mechanical equipments. In this dissertation, probabilistic LCF life prediction models and uncertainty assessment methodologies were studied systematically in both theoretical and engineering applications. The studied alloys included Ni-base Superalloy GH4133, GH4698 (typically used as the turbine disk material in gas turbine engines) and pearlitic heat resisting steel (which is used in steam turbine items). The main work and innovative contributions of this dissertation are as follows:(1) Development of a Fuzzy Miner’s rule considering damaging and strengthening of low-amplitude loads under different load sequencesDue to the shortcomings of the traditional Miner’s rule, a Fuzzy Miner’s rule is developed to consider the strengthening and damaging of low-amplitude loads with different load sequences. This model improves the application of the traditional Miner’s rule, by considering not only the damaging and strengthening of low-amplitude loads, but also the load sequence effects. To apply the Fuzzy Miner’s rule, the law of selecting membership functions for different load spectrum is found and different membership functions are investigated to show the important influence on estimating fatigue life. Applicability of the method was validated using experimental and real-time data gained from aircrafts. It was also found that the predicted fatigue life by the proposed rule is more accurate and reliable than that by the traditional methods. In addition, the Fuzzy Miner’s rule provides a criterion for judging if“the loads cause damage or not”and a theoretical basis for explaining the“coaxing effect”phenomenon in engineering.(2) Development of a generalized strain energy damage function model for fatigue-creep life predictionThe fatigue-creep interaction is a key factor for the failures of many complex structures under high temperature and cyclic loading. These fatigue-creep life prediction issues are significant in selection, design and safety assessments of those components. In order to describe the accumulation and development of damage uniformly and accurately, a generalized strain energy damage function model was developed for fatigue-creep life prediction under different loading waveforms. The approach used in this model to reflect the effects of time-dependent damaging mechanisms on fatigue-creep life is different from those used in all earlier models. In addition, an improved generalized strain energy damage function was used to reduce the difference between the approximate strain energy and real strain energy absorbed during the damage process. This proposed model can describe the effects of different time-dependent damaging mechanisms on fatigue-creep life more accurately than others, which makes it widely applicable and a precise method to predict the life of fatigue-creep interaction.(3) Development of a generalized energy-based damage parameter for fatigue-creep life predictionBased on the plastic strain energy density (PSED) and Physics of Failure (PoF) analysis, a generalized energy-based fatigue-creep damage parameter was developed to account for the creep and mean strain/stress effects in the LCF regime. Moreover, the mechanism of cyclic hardening is taken into account within this model. On this basis, it is assumed that damage accrues by means of viscous flow and ductility consumption is only related to plastic strain and creep strain under high temperature LCF conditions. Based on the ductility exhaustion (DE) theory, a new viscosity-based life prediction model was introduced by using dynamic viscosity to describe the flow behavior. Both the proposed damage parameter and viscosity-based model provided a better prediction of GH4133’s fatigue behavior when compared with the SWT and PSED methods. Under mean strain conditions, these two models provide a more accurate life prediction of GH4133 than that under zero-mean strain conditions, which provides an effective, reliable and new way for accessing the remaining life of complex structures.(4) Construction of a hybrid probabilistic PoF-based LCF life prediction frameworkProbabilistic life prediction of complex structures, such as aircraft turbine disks, requires the modeling of multiple complex random phenomenas. The Bayesian approach can potentially give more complete estimates by combining test data with technical knowledge available from theoretical analysis and/or previous experimental results, and provides many practical features such as a fair coverage of uncertainty and the updating concept that reduces costs and saves time. A hybrid probabilistic PoF-based framework for life prediction using Bayes’theorem was developed to quantify the uncertainty of material properties, total inputs and model uncertainty resulting from creation of different deterministic models in a LCF regime. In practice, the statistical inferences in this framework involve high-dimensional integrations that usually are very computationally intensive. The Bayesian inference was solved using the MATLAB platform to run the necessary MCMC simulation, which greatly improves the utility of this framework. Through this framework, a basis for safety assessment and life cycle management of major mechanical equipments was offered.(5) Development of a hybrid probabilistic PoF-based fatigue life prediction methodology under Bayesian information updating and uncertainty Under the hybrid probabilistic PoF-based fatigue life prediction framework, a white-box approach for modeling total uncertainty, including physical uncertainty, statistical uncertainty and model uncertainty, was developed for LCF life prediction based on Bayesian information updating. In addition, the black-box approach was expanded to quantify the model uncertainty in LCF life prediction. Compared with the white-box approach, there is no consideration of the uncertainties associated with the inner workings of the model in the black-box approach. The development of white-box approach, the improvement of methodology and expanding of research on black-box approach provided reliable theoretical basis and scientific method for model selection and comparison. The proposed probabilistic PoF-based fatigue life prediction methodology intrinsically explains the scatter of service life of complex structures subjected to the same conditions and has certain conductive significance for structural health monitoring, design, safety evaluation and remaining life assessment of major mechanical equipments.