Evaluation of service conditions of high pressure turbine blades made of DS Ni-base superalloy by artificial neural networks

https://doi.org/10.1016/j.mtcomm.2019.100838Get rights and content

Abstract

Evaluating the service conditions of high pressure turbine blades in areoengine, which is the key point for component design and remaining life prediction, has always been a great challenge. In this study, we report a method, which is developed by back propagation artificial neural networks (BPANN), for service condition evaluation of turbine blades by establishing a quantitative correlation between the microstructural evolution and the temperature, stress and time. This method was successfully applied to a directionally solidified superalloy DZ125 based on microstructural datasets obtained from temperature-stress-time (T-σ-t) simulations. As an example, the service condition of a turbine blade made of DZ125 superalloy were evaluated by the BPANN model using the microstructural descriptors, including γ′ volume fraction (Vf), γ′ rafting degree (Ω) and thickness of the rafted γ′ precipitates (D), and service time. This study shows great potential in accurately assessing the degradation and predicting the remaining life for hot section components made from directionally-solidified and single crystal nickel-based superalloys.

Introduction

Quantitative relationships between composition, processing, microstructure and performance are very important for materials development. First principles calculations, molecular dynamics, phase field modelling and other methods have employed for accurate modeling of the relationships between composition, processing and microstructure [[1], [2], [3], [4]]. However, establishing a quantitative relationship that links the microstructure to the mechanical properties still remains a challenge, especially for components with complex compositions and/or subjected operating conditions, such as turbine blades.

For turbine blades in-use in gas turbine engines, typically fabricated from superalloys, creep deformation is usually considered the primary degradation and failure mode [5]. During creep deformation, not only plastic strain accumulate, but also microstructural evolution and degradation takes place. These two process are often interlinked, making behavior predictions challenging. Directly measuring the local creep deformation of turbine blades during overhaul is not feasible, due to the complex service loads and intricate component designs. As such, a few blades might be removed from the jet engine during an overhaul, in order to directly check the microstructural evolution by-using metallographic analysis. Typically, different microstructural evolutions lead to different deformation behaviors, therefore a complete microstructural characterization is the most efficient way to explore the deformation mechanisms of superalloys at different conditions. In previous studies, special attention has been paid to the morphology of different phases, such as γ′ precipitates, grains, carbides, TCP phase and dislocations [6]. For instance, Miura et al. [7,8] qualitatively analyzed the service conditions of turbine blades made of polycrystalline and single crystal nickel-base superalloys based on the morphology of the γ′ precipitates. However, researchers have realized that the qualitative characterization of microstructural evolution is not sufficient to completely explain the deformation mechanism and degradation degree of the creep properties.

Developing quantitative relationships between creep conditions (temperature, stress and time) and the microstructural evolution is advantageous in evaluating the local loading conditions of turbine blades [9], and such relationships can be used to evaluate the damage degree of turbine blades, and guide engine design and material selection. Nevertheless, quantitative relationship are very complex and non-linear due to synergistic effects among temperature, stress and time on the microstructural evolution during creep. In our previous studies, the evaluation of the service temperature and stress of a directional solidified superalloy, DZ125, was investigated [10,11]. Chen et al. [10,11] used T-σ-t coupling simulation experiments to establish a qualitative relationship between the operating temperature, stress, time and quantitative microstructural descriptors - the γ′ volume fraction (Vf) and the rafting degree (Ω). In their study, the evaluation method of the service conditions was developed by 3D maps and hardness to provide an accurate assessment result, that is descriptor value matching between the observed microstructure against a database of values for known conditions. Although useful, this method is qualities in nature, hence it needs a large dataset of known values because it does not permit extrapolation, like a quantitative relationship . Meanwhile, Huang et al. [12] also proposed a method to evaluate the service conditions of DZ125 turbine blades by thermal dynamical equations combining parameters such as the γ′ width and γ channel width. They proposed a unified analytical model for the γ′ precipitate evolution of DZ125 based on the Lifshitz-Slyozov-Wagner (LSW) theory, and used it to predict the experimentally observed increase in the extent of coarsening associated with the increase in temperature, time, or applied stress. However, the equations are only suitable for simulating a single variable, and have difficulty to capture all the relevant relationships between microstructures and macro-conditions as a whole. Hence, a new method is critically needed.

In summary, considering that the existing research mostly focuses on qualitative analyses and constitutive equations, it is difficult to model the complex relationships between the microstructure and macroscopic properties. This study attempts to establish such a relationship between the microstructure and creep conditions for DZ125, using an artificial neural network (ANN) model, capable of mining complex nonlinear relationships. Artificial neural networks, as a simple and popular machine learning model, have been employed to predict different properties in various materials by establishing numerical relationships [[13], [14], [15]]. For example, Conduit et al. [16] developed a back propagation artificial neural network (BPANN) tool to design new nickel base superalloys by optimizing 11 properties at the same time, including the cost, density, γ′ phase content, etc. Meanwhile, Palavar et al. [13] successfully established an ANN model to predict the weight loss of IN706 superalloy due to wear.

In this study, different degrees of microstructural evolution were obtained through a large set of temperature-stress-time (T-σ-t) coupling experiments of DZ125, and the resulting microstructures were quantified through selected microstructural descriptors. Based on these descriptors, two BPANN models, one for the temperature and one for the stress assessment, were established, that can quantitatively link service conditions to microstructure. Finally, the application of this method was demonstrated on a real turbine blade that has been in service for 900 h.

Section snippets

Materials and experiments

The directionally solidified nickel-based superalloy DZ125 was developed based on René DS 125 by a reduction of the Ti and C contents and an increase of the Hf content [17]. Directionally solidified crystal bars (Φ15 mm × 170 mm) were manufactured by the Beijing Institute of Aeronautical Materials (BIAM), and were subjected to the standard heat treatment (SHT): 1180 °C/2 h + 1230 °C/3 h, AC + 1100 °C/4 h, AC + 870 °C /20 h, AC [18]. The chemical composition of the bars was measured by

Results and discussion

The microstructural evolution of superalloys during actual service or simulation experiments is strongly influenced by the coupling of temperature, stress and time [11,23,24]. In the past decades, the creep deformation behavior of superalloys has been systematically studied mostly in the temperature range of 700–1200 °C, and linked to the microstructural evolution [25]. Many of these studies have mainly focused on qualitative analysis at high temperature-low stress or low temperature-high

Summary

The microstructural evolution of a directionally solidified superalloy, DZ125, was investigated during temperature-stress-time coupling service simulation experiments in the range of 900–1100 °C, 0–400 MPa, and 25–1200 h. Two BPANN models for evaluating the service condition of a given microstructure were established through the use of quantitatively characterized descriptors of the microstructure. The results are summarized as follows:

  • 1

    To establish a quantitative relationship between given

Declaration of Competing Interest

We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled.

Acknowledgements

The supports provided by the National Key Research and Development Program of China (Grant No. 2016YFB0701403), the Key-Area Research and Development Program of GuangDong Province (Grant No. 2019B 010943001) and the National High Technology Research and Development Program of China (Grant No. 2012AA030513) are gratefully acknowledged.

References (39)

  • Y. Yu et al.

    A high strength and elastic carbon containing near-α Ti alloy prepared by hot isostatic pressing process

    Mater. Sci. Eng. A

    (2016)
  • P. Kontis et al.

    The effect of chromium and cobalt segregation at dislocations on nickel-based superalloys

    Scr. Mater.

    (2018)
  • H. Mughrabi

    The importance of sign and magnitude of γ/γ′ lattice misfit in superalloys—with special reference to the new γ′-hardened cobalt-base superalloys

    Acta Mater.

    (2014)
  • W.J. Boettinger et al.

    Phase-field simulation of solidification

    Annu. Rev. Mater. Res.

    (2002)
  • L.Q. Chen

    Phase-field models for microstructure evolution

    Annu. Rev. Mater. Res.

    (2002)
  • R.G. Parr

    Density functional theory of atoms and molecules

    Horizons of Quantum Chemistry

    (1980)
  • K.C. Antony et al.

    Aircraft gas turbine blade and vane repair

    Superalloys

    (1988)
  • J. Liburdi et al.

    Practical experience with the development of superalloy rejuvenation

    ASME Turbo Expo 2009: Power for Land, Sea, and Air

    (2009)
  • N. Miura et al.

    Morphology of γ’ precipitates in second stage high pressure turbine blade of single crystal nickel-based superalloy after serviced

    Materials Science Forum

    (2010)
  • Cited by (24)

    • Duration-affected creep behaviors of Ni-based single crystal superalloys with/without rhenium addition designed for IGT application

      2023, Materials Science and Engineering: A
      Citation Excerpt :

      Recently, the Re enrichment on dislocations was also confirmed to impose a drag effect on dislocation movement, which also contribute to improve the creep property [23]. Commonly, the rafting of γ′ phase during high-temperature creep of Ni-based superalloys exhibits the following characteristics: (1) the early formation of the rafting structure at the initial creep stage [43]; (2) the onset of the rafting regime once the macroscopic strain exceeds the threshold strain [6]; (3) the accelerated rafting process promoted by increasing the applied stress [32]. However, in this study, sluggish formation of the rafting structure can be observed in both alloys regardless of the strain.

    • Structural degradation and elemental variations in an ex-service first-stage gas turbine blade

      2023, Materials Characterization
      Citation Excerpt :

      Typical structural degradation is γ' phase coarsening and rafting, and it usually exhibits disc-shaped γ' precipitates in the stress perpendicular direction and band-shaped precipitates in the stress parallel direction [26,27]. High temperatures usually prompt the coarsening of the structure, while high-speed rotation induces the stress prompt rafting of the γ' phase [28,29]. The hypergravity force has an appropriate proportion with the distance from the shank (fixed part) to the tip of the blade; thus, stresses in a turbine blade are dependent on its locations [14].

    View all citing articles on Scopus
    View full text