Linking ICME to Component Life Management During Design

Transcription

1 Linking ICME to Component Life Management During Design TMS 2014 Annual Meeting San Diego, California February 17-20, 2014 Craig McClung, Michael Enright, John McFarland, Kwai Chan Southwest Research Institute Wei-Tsu Wu, Ravi Shankar Scientific Forming Technologies Corporation

2 Acknowledgments Funding for this effort was provided by US Air Force Research Laboratory Small Business Innovative Research (SBIR) Projects Topic No. AF Phase I Contract FA M-5110 Phase II and IIE Contract FA C-5105 Rollie Dutton and Patrick Golden, AFRL Program Monitors Federal Aviation Administration Grant 11-G-009 Joseph Wilson and David Galella, FAA Program Monitors Other colleagues made invaluable contributions Jonathan Moody (SwRI) Simeon Fitch (Elder Research) Weiqi Luo, Jinyong Oh (SFTC) 2

3 Goals The goal of ICME is to optimize materials, manufacturing processes, and component design through integration of computational processes into a holistic system The specific goal of this effort is to link manufacturing process simulation directly to a critical measure of component reliability using production software Manufacturing Process Simulation Residual Stresses Microstructure Material Anomalies Probabilistic Damage Tolerance Analysis Risk of Component Fracture 3

4 DARWIN Overview Design Assessment of Reliability With INspection 4

5 DEFORM Integrated Process and Material Modeling System Cogging Forging Machining Ring Rolling Heat treatment Inertia Welding Milling Casting Rolling Sheet Forming Spin Pit Testing Life Furnace Heating Spot Welding Machining Distortion Extrusion Hot Press Forming Stir Welding SPF Induction Heating 5

6 Numerical Simulation of Material Processing Residual Stresses Microstructure Anomaly Tracking and Deformation 6

7 DARWIN Overview Design Assessment of Reliability With INspection Anomaly location and orientation Residual stresses Microstructure 7

8 DARWIN-DEFORM Links Residual Stresses Microstructure Anomaly Tracking and Deformation 8

9 DARWIN Stress Superposition Approach for Residual Stresses Service Stress Neutral file Residual Stress Neutral file stress gradient Normalized Stress Service Stress Combined stress Residual Stress DARWIN Stress Extraction Normalized Distance Residual stress analysis Arbitrary stress gradients are used to calculate crack driving force with weight function stress intensity factors 9

10 Automated Calculation of Crack Growth Life and Risk DARWIN can perform full-field automated calculation of location-specific fatigue crack growth life and fracture risk Automatically generate idealized fracture geometry model for any crack location in an arbitrary geometry Automatically extract stresses from FE models and calculate stress intensity factors Automatically calculate FCG lifetime from a common initial crack size at every location Automatically calculate the risk of component fracture with a probabilistic FCG analysis Considering uncertainties in anomaly size & frequency, stress scatter, life scatter, NDE inspection time, and NDE POD 10

11 Demonstration Example: Effect of Material Processing Residual Stress on FCG Life Without Residual Stress With Residual Stress Stress Life 11

12 Effect of Material Processing Residual Stress on Fracture Risk Without Residual Stress With Residual Stress Life Risk 12

13 Modeling Random Residual Stresses in DARWIN DEFORM DOE Stress Results Files DARWIN residual stress DOE 1 contour NESSUS Gaussian Process Response Surface Model residual stress DOE n contour 13

14 Demonstration Example: Random Residual Stresses crack path crack path 14

15 DEFORM Random Variables Table 1. Application Example Manufacturing Process Parameters Variable Description Mean Standard Deviation 1 Conv. coeff. factor Flow stress factor Heat cap. factor Object temp Pass 1 offset factor Poisson ratio factor Therm. con. factor Transfer time Young mod. factor DOE with 100 residual stress training points using LHS 15

16 Principal Components Analysis (PCA) for Residual Stresses Along Crack Path Training data Mode shapes 16

17 Effect of Random Residual Stress on Risk Without Residual Stress With Random Residual Stress 17

18 DARWIN-DEFORM Links Residual Stresses Microstructure Anomaly Tracking and Deformation 18

19 Influence of Forging Strain on Orientation of 3D Anomalies Circles represent relative seed area Lines represent relative major and minor axis lengths Angle of major axis is seed orientation relative to forging Kantzos et al. 2003, Effects of Forging Strain on Ceramic Inclusions in a Disk Superalloy, Adv. Matls and Proc. for Gas Turbines, TMS 19

20 Importing Residual Strain data from DEFORM 20

21 Viewing Principal Strain Orientations in DARWIN 21

22 Visualizing the Influence of Forging Strains on Anomaly Orientation First Principal Forging Strain Anomaly Orientation Computed in DARWIN Note alignment with principal strains 22

23 Planned DARWIN Enhancements for Random Anomalies Import results from multiple (DOE) DEFORM runs containing residual strain and anomaly occurrence rate information at FE nodes Define input random variables associated with DEFORM computations Create GP response surface models of residual strain and anomaly occurrence rate Response surfaces at initial crack locations only Compute anomaly occurrence rate scaling factors and apply to occurrence rates associated with zone anomaly distributions Compute random residual strain and anomaly occurrence rate via Monte Carlo simulation Design of Experiments Response Surface 23

24 DARWIN-DEFORM Links Residual Stresses Microstructure Anomaly Tracking and Deformation 24

25 Grain Size Modeling in DEFORM Empirical JMAK Method Input: Initial average grain size distribution Strain, temperature, strain rate history Grain growth equations Recrystallization kinetics Output: Dynamic Metadynamic Static Location-specific grain size contours Percentage recrystallization d. m10 h n10 drx = a10d 0 ε ε exp / + ( Q10 RT ) c 25

26 Microstructure-Based Fatigue Crack Growth Model da dn ξ = = ξ Es 4σ εd ' y 1/b ' f ( 2s) 1 1/ b d K E d 2 / b D = 0 D0 1/3 K: Stress Intensity Range E: Young s Modulus s: Dislocation Cell Size d: Dislocation Barrier Spacing σ y : Cyclic Yield Stress ε f : Fatigue Ductility b: Fatigue Exponent D: Grain Size 26

27 Practical Implementation of Micromechanical Models in DARWIN User provides standard fatigue crack growth properties and a single average grain size associated with these properties DEFORM calculates average grain sizes at each FE node DARWIN computes crack growth rate at selected locations by scaling micromechanical models based on grain size da D da = f dn D* dn A similar paradigm can be used to calculate fatigue crack initiation lifetimes * 27

28 Demonstration Example: Influence of Grain Size Scaling on Life & Risk ANSYS ABAQUS DEFORM Stress Results Files DARWIN service stress contours DEFORM Grain Size Results File grain size contours 28

29 Influence of Grain Size Scaling on Crack Growth Rate da D da = f dn D* dn * grain size contours Nominal values: C=1.56 x n 2 =3.66 crack growth rate multiplier 29

30 Effect of Location-Specific Grain Size Scaling on FCG Life Without Grain Size Scaling With Grain Size Scaling a=0.01 a=

31 Effect of Location-Specific Grain Size Scaling on Fracture Risk Without Grain Size Scaling With Grain Size Scaling Life a=0.01 Risk 31

32 Planned DARWIN Enhancements for Random Microstructure Import results from multiple (DOE) DEFORM runs containing: Average grain sizes at all finite element nodes ALA grain sizes at selected finite element nodes Define input random variables associated with DEFORM computations Create GP response surface models of average and ALA grain size Response surfaces for average grain size at all initial crack locations (all FE nodes) Response surfaces for ALA grain size at initial crack locations identified by DEFORM (selected FE nodes) Compute random average and ALA grain size via Monte Carlo simulation Design of Experiments Response Surface 32

33 Planned DARWIN Enhancements for Crack Initiation Simulate 3D grains based on 6DOF grain information from DEFORM Build GP response surfaces from 6DOF grain results at selected locations from multiple DEFORM runs containing: Average & ALA grain sizes & aspect ratios Grain orientation (Euler angles) Represent local microstructure as a 3D volume element Ellipsoid containing grain is simulated using the ALA 3D grain model Number of facets is sampled from a facet distribution Characteristics of individual grain boundary facets are sampled from a misorientation angle distribution 33

34 Planned DARWIN Enhancements for Crack Initiation Implement enhanced micromechanics-based crack initiation model in DARWIN Formation module Treatment of pile-up length: The pile-up length is computed based on the ALA grain size and the misorientation angle of the neighboring grains If the misorientation angle is less than a critical value (e.g., 15 ), slip across the grain boundary is assumed to occur and the length of the neighboring grain is added to the pile-up length This process is repeated until the slipband is blocked by a neighboring grain with a misorientation angle greaten the critical value Once the pile-up length is determined, the number of fatigue cycles for crack initiation at a slipband and at an inclusion can be computed ( ) 1/2 2 1/2 α 8Mµ h c i σ 2Mk N = λπ ( 1 ν ) D D 34

35 Planned DARWIN Enhancements for Time-Dependent Crack Growth Address concurrent damage mechanisms involving cycle-dependent crack growth due to fatigue and time-dependent crack growth due to corrosion, oxidation, and creep in Ni-based alloys Couple microstructure-based time-dependent crack growth models with corresponding cycle-based crack growth model to address effects of long dwell times on component life 10 1 ME da/dt, mm/sec R = o C 760 o C 704 o C 649 o C 538 o C da/dn, mm/cycle o C 649 o C 538 o C ME3, R =0.5 K = 16.5 MPa(m) 1/2 204 o C 649 o C 538 o C 704 o C K, MPa(m) 1/ o C Frequency, Hz 35

36 Planned DARWIN Enhancements for Time-Dependent Crack Growth Location-specific lifing for a generic ME3 disk: Gayda et al, Superalloys 2004 coarse grain size at rim fine grain size at bore mixed grain size in transition zone. Specify tertiary gamma prime size distribution at various disk locations Assess the roles of grain size and tertiary gamma size on disk life. Gabb et al., Int. J. Fatigue

37 Linking Materials and Lifing: Some Specific Needs Link microstructure and lifing properties Link processing analysis with life analysis Probabilistic models linking material/microstructural variability at relevant length scales to variability in fatigue/fracture/life properties and risk Microstructure-property models that are computationally efficient and robust Suitable for integration into the overall optimization process, including linkages to probabilistic lifing codes Processing Microstructure Lifing Properties Life Prediction Reliability 37

38 Summary Interfaces between DEFORM and DARWIN have been developed for full-field, location-specific bulk residual stresses, forging residual strains, and average grain size. These interfaces permit full-field results from manufacturing process simulations to be incorporated in predictions of fracture life and reliability. Deterministic and probabilistic approaches were presented and demonstrated for modeling the effects of these parameters on crack growth behavior. Further work is underway to develop improved deterministic and probabilistic approaches to address microstructural effects on crack initiation and growth. The program demonstrates the practical potential for ICME that directly addresses component integrity. 38