Stress Intensity Factor Range Research Articles

Fatigue prediction through quantification of critical defects and crack growth behaviour in additively manufactured Ti-6Al-4V alloy

The study presents a methodology for predicting the fatigue life and identifying critical defects in additively manufactured (AM) Ti-6Al-4V alloy processed via laser powder bed fusion (L-PBF). Predictions were made on L-PBF Ti-6Al-4V alloy in two conditions: as-built and heat-treated by using X-ray μ-Computed Tomography (μ-CT) for the quantification of the defects and fatigue crack growth (FCG) data. For validation, fatigue life predictions were made on the same specimens on which μ-CT was conducted prior to fatigue testing. FCG and fatigue tests (S-N) highlighted differences in the performance between the as-built and heat-treated conditions: the as-built condition had a lower threshold stress intensity factor range (ΔKth) of 2.16 MPa m1/2 than that of the heat-treated condition, 4.96 MPa m1/2. Differences in fatigue limit were attributed to differences in ΔKth in as-built and heat-treated specimens. To gain mechanistic understanding of these differences near ΔKth cracks were examined using Electron Backscatter Diffraction - Kernel Average Misorientation (EBSD-KAM). It was found that upon heat treatment, the dislocation density increased around near threshold fatigue cracks. Increases in ΔKth are attributed to the increases in β-phase content and α-lath thickness caused by heat treatment.

Fatigue behaviour and life prediction of self-piercing riveted joints in DP590/AA5754 dissimilar sheets

This paper investigates the fatigue properties of self-piercing riveted joints made by DP590 steel and AA5754 aluminium alloy. The fatigue failure mechanism and fatigue fretting mechanism of the joint were well revealed by analysing the microstructure features of fracture surfaces of sheets and rivets as well as viewing and measuring the distribution and chemical composition of fretting debris. The fatigue life of the joint was also predicted based on the stress intensity factor of the kinked crack of the lower sheet. The results confirmed that two failure modes occurred during fatigue: lower sheet fracture and rivet fracture. During the fatigue failure, cracks appeared in both the lower sheet and the rivets, indicating a competitive relationship between the propagation of cracks in the lower sheet and the rivets. In the range of 40 % and 50 % load level, fatigue crack propagation of the lower sheet is dominant and fracture failure of the lower sheet occurs. Rivet fatigue crack propagation has a significant advantage in the 60 % load range and rivet fracture failure often occurs. It was found that the fatigue failure of the lower sheet is mainly influenced by the main crack, which produces two different morphological sections. Fretting wear of the contact surface was the source of crack initiation, and the fretting debris promoted crack initiation and propagation. Cracks propagate along the width and thickness of the plate, as well as the thickness of the rivet neck. The stress intensity factor range can properly predict the fatigue life of joints.

Modeling of crack tip fields and fatigue crack growth in fcc crystals

Predicting the mechanical behavior of polycrystalline materials containing a crack under both monotonic and cyclic loading conditions is crucial for accurately assessing the integrity of engineering materials. This study focuses on the deformation characteristics of face-centered cubic (fcc) grains within the crack tip field and their significant role in governing the driving force for fatigue crack growth during cyclic loading. We employ a cyclic crystal plasticity finite element (CPFE) model to analyze the mechanical response of austenitic 316L stainless steel polycrystals by accounting for nonlinear kinematic hardening effects. Through CPFE simulations, we investigate the deformation fields in 316L grains at the crack tip, considering two different grain orientations under plane strain conditions. Our CPFE results under monotonic loading align consistently with previous theoretical and experimental findings, particularly in comparing CPFE-simulated and experimentally observed plastic sectors consisting of different slip traces on the specimen surface near the crack tip. Based on a critical plastic work criterion for crack advancement, cyclic CPFE simulations are used to determine the fatigue crack growth rate as a function of stress intensity factor range for the two crack tip grain orientations in stainless steel. The simulated Paris law exponent matches experimental values. Furthermore, we compare cyclic CPFE results with those from cyclic J2 plasticity finite element simulations. This study demonstrates a cyclic CPFE approach for determining crack tip fields, accounting for crystallographic effects on plastic deformation of crack tip grains. Our approach can be applied to effectively evaluate fatigue crack growth rates in fcc polycrystalline metals.

Fatigue Crack Growth Rates and Crack Tip Opening Loads in CT Specimens Made of SDSS and Manufactured Using WAAM.

Additive manufacturing offers greater flexibility in the design and fabrication of structural components with complex shapes. However, the use of additively manufactured parts for load-bearing structural applications, specifically involving cyclic loading, requires a thorough investigation of material fatigue properties. These properties can be affected by many factors, including residual stresses and crack tip shielding mechanisms, which can be very different from those of conventionally manufactured materials. This research focuses on super duplex stainless steels (SDSSs) fabricated with wire arc additive manufacturing (WAAM) and investigates their fatigue crack growth rates and the net effect of crack tip shielding mechanisms. Using the compliance-based method, we measured crack tip opening loads in compact tension (CT) specimens with cracks propagating longitudinally and transversely to the WAAM deposition direction. It was found that fatigue crack growth rates were very similar in both directions when correlated by the effective stress intensity factor range. However, the differences in crack tip opening loads explain a quite significant influence of the deposition direction on the fatigue life.

Friction stir welded joints in aluminum highway bridge decks: a quality control framework

Friction stir welding (FSW) has shown considerable promise for highway aluminum bridge deck fabrication but lacks specific quality control guidelines for fit-up defects. This study conducts a performance-based (PB) quality control assessment of butt-lap FSW joints for highway bridge decks. Five FSW conditions were simulated, including a standard control welding condition, fit-up defects (gaps and tool offset), and a welding tool’s rotational direction inversion from the standard condition. A rigorous prequalification process established acceptable tolerance levels for fit-up defects: a 3 mm positive offset, a 1.5 mm negative offset, and a 1.5 mm gap. Subsequently, specimens from real aluminium deck extrusions, incorporating the various welding conditions were subjected to fatigue testing. Results showed that the fatigue strength and failure mode were primarily influenced by the weld root microstructure. The FSW tool’s rotational direction significantly influenced fatigue strength due to its impact on the nucleation of the hooking defect in the weld root area, while the impact of fit-up defects on fatigue strength was comparatively lesser. Furthermore, finite element analysis examined the impact of the geometrical features of the root microstructure and the direction of the initial crack propagation on the stress intensity factor range. These findings have practical implications for setting tolerance levels for fit-up defects in aluminum butt-lap FSW joints and optimizing of the quality and fatigue strength of butt-lap FSW joints.

Unified SIF at different stress ratios and the physical insight on the exponent of the Paris law

The Paris law has been broadly employed to describe fatigue crack growth (FCG) at steady state. It establishes the relationship between FCG rate and stress intensity factor (SIF) range. However, its physical insight remains an issue of controversy. The study aims to employ dislocation dynamics to explore two aspects associated with the Paris law: how does the stress-ratio affect the application of the Paris law? What is the microcosmic mechanism behind the Paris exponent in Paris law? We limit our exploration on the scenarios that dislocation nucleation and dislocation glide dominate FCG in an ideally isotropic single crystallite. Our exploration revealed that (1) a unified SIF range without empirical parameters could be adopted to account for the stress ratio effect, and this observation is in line with existing experiments, and (2) SIF affects FCG by influencing the nucleation of dislocation sources and activity of dislocations before crack tips. The stress-dependent dislocation nucleation time is correlated with the exponent of the Paris law.

Cutting-edge insights: Crack growth rates in 9% Ni steel from room to cryogenic conditions influenced by crack closure and effective stress intensity factors

This study investigates the fatigue crack growth rate of 9% Ni steel at varying stress ratios under both room and cryogenic temperatures, attributing different growth rates to the phenomenon of crack closure. Traditional methods of studying crack closure, including the opening force, compliance ratio, and adjusted compliance ratio techniques, are assessed. Our results are juxtaposed with previous formulae for the opening force technique, leading to the introduction of a new formula that incorporates temperature effects by normalising the stress intensity factor with the yield strength at the respective test temperature. We highlight the adjusted compliance ratio (ACR) technique's limitations at cryogenic temperatures due to challenges in measuring notch open-crack compliance. As a solution, we present an innovative method that determines compliance using the effective elastic modulus. The mechanics behind the ACR technique are elucidated through the lens of the wedging mechanism. Our insights reveal that the ACR technique precisely calculates the effective stress intensity factor range (ΔKeff) under specific conditions. We address existing limitations by introducing a novel method, the triple compliance ratio (TCR) technique. A comparative analysis of effective stress intensity factors from all four techniques reveals the TCR technique's superior ability to correlate crack closure with crack growth. By eliminating the crack closure effect, the TCR technique offers a more accurate prediction of fatigue crack growth rate across diverse load ratios.

Microstructural influences on simultaneous strength and fatigue crack resistance in advanced high-strength steels

In this study, we investigate the relation between microstructure and fatigue crack propagation mechanisms in three commercial hot-rolled thick-plate advanced high-strength steels (AHSSs), namely 800CP, 700MC, and 700MCPlus, and compare them with a conventional high-strength low-alloy (HSLA) steel, 500MC, commonly used in heavy-duty vehicle chassis production. Tensile testing, and fatigue crack growth rate (FCGR) assessments have been conducted, and mechanisms controlling the performance of these steels are comprehensively examined through microstructure characterization before and after fatigue testing. Notably, FCGR results reveal that, despite having the highest yield strength among the investigated steels, 700MCPlus exhibits the slowest FCGR in both near-threshold and stable crack growth regimes. This improved fatigue crack propagation resistance of 700MCPlus in terms of its threshold stress intensity factor range (ΔKth) is attributed to its unique texture, which restricts slip activity, and the presence of martensite at grain boundaries, contributing to fatigue crack deflection. This martensite-induced crack deflection becomes more significant in the stable crack growth regime, where fatigue crack primarily propagates intergranularly as the stress intensity factor range (ΔK) increases. These mechanistic findings offer new design possibilities for AHSSs, combining excellent strength and fatigue performance.

Assessing fatigue crack growth thresholds for a Ti-6Al-4V (STOA) alloy using two experimental methods

Fatigue crack growth rate behavior in the threshold and near-threshold regimes for a Ti-6Al-4V solution-treated and over-aged (STOA) alloy was investigated using two test procedures: i) accelerated load reduction (ALR) utilizing faster load-shed rates than recommended in the ASTM standard E647 load-reduction (LR) test procedure, and ii) compression pre-cracking constant-amplitude (CPCA) or load-increasing (CPLI) and load-reduction (CPLR). Fatigue crack growth tests were conducted on compact, C(T), specimens (B = 6.35 mm) with two different widths (W = 51 and 76 mm) at two stress ratios (R = 0.1 and 0.7). The crack-growth-rate test data were compared to previous test data produced from the same batch of material using the ASTM LR and CPCA test procedures that were tested on C(T) specimens (W = 25, 51 and 76 mm) over a wide range in stress ratios (R = 0.1, 0.4, and 0.7).Although none of the load-reduction test procedures provided a good representation of the threshold value (ΔKth) at low R, the CPCA test procedure was the most promising. The LR, ALR, and CPLR test procedures were influenced by prior loading history, leading to higher ΔKth values and slower crack growth rates in the threshold regime due to a rise in the crack-opening loads. The LR and ALR tests at R = 0.1 and shed rates of −0.08, −0.32, and −0.95 mm−1 yielded ΔKth values that were 15 to 38 % higher than the estimated threshold stress-intensity factor range (ΔK*th)R=0.1 based on the crack-closure model (FASTRAN). The compression pre–cracking test procedures are more accurate alternatives for developing threshold and near-threshold fatigue crack growth rates. Specifically, the CPLR test procedure produced a ΔKth value that was 8–18 % higher than (ΔK*th)R=0.1. Additionally, LR and ALR tests produced different ΔKth values as a function of the specimen width for a given load ratio, which violates classical fracture mechanics principles. In addition, the spread in the ΔK-rate data in the low-rate regime was larger than the spread in the mid-rate regime as a function of R, a term referred to as “fanning”. In contrast, the CPCA test procedure produced consistent crack growth rates over the same range of ΔK values examined, independent of the specimen width and these data did not exhibit “fanning” with R. Further research is necessary to develop standard test procedures capable of providing a more definitive representation of the ΔKth value in the threshold regime. However, FASTRAN can serve as a crucial tool in aligning the actual ΔKth values with those obtained through CP testing, thereby bridging the gap and enhancing the reliability of the results.

A new method for prediction of fatigue crack propagation life under variable amplitude spectrum loading

This paper proposes a new method for predicting propagation lives for through-thickness cracks under variable amplitude mode-I spectrum fatigue loads. Only the crack opening, i.e., the increasing parts of the load cycles within spectrum profiles are considered for cycle counting and computing their stress ratios. Moreover, a cycle-by-cycle analysis is also performed throughout the spectrum to compute the so-called Spectrum Overload Index (SOL-Index), which is a multiplier applied on stress intensity factor range used within different crack growth retardation models, in which effects of stress ratio are also included. The SOL-Index is calculated using the average of peak loads for all cycles within the spectrum and their percent deviations from the peaks’ average value. Simulation results using the proposed cycle counting method, the integrated crack growth model and the SOL-Index are compared with those from multiple experiments under two different loading spectrums and another study in the literature. The results show that usage of Wheeler's model embodying the Forman or Walker equation and the SOL-Index together yields the closest results to the experiments.

Fatigue and fracture of small cracks in superelastic Nitinol

The present study develops a hybrid total life and damage-tolerant fatigue methodology for the determination of a small crack growth threshold in Nitinol. Axial tension–tension/Mode I fatigue testing was performed on superelastic Nitinol wire specimens with starter cracks milled into the specimens using focused ion beam (FIB). Fatigue testing was performed in displacement control at various strain amplitudes, typical for Nitinol fatigue characterization given its superelastic behavior. The simplistic geometry of the test specimen and axial loading conditions also allowed for the calculation of stress amplitude and initial stress intensity factor range. A logistic regression analysis was used to determine strain/stress-based fatigue limits. The small crack growth threshold for superelastic Nitinol was obtained for the first time from physically small cracks with sizes comparable to native inclusions. Lastly, a fractographic analysis was performed to examine crack initiation and propagation from the FIB-milled starter cracks and to identify relevant fractographic features. While the proposed methodology is centered around the use case of Nitinol medical devices, it could be adapted for other materials and applications where small crack initiation and propagation play an important role in fatigue lifetime prediction.

Investigation of Austempering Effect on Fatigue Crack Growth of AISI 4140 Steel

The compact tension (CT) and tensile specimens of the AISI 4140 steel in cold rolling condition (untreated steel) were austempered by immersing it into the salt bath at 362°C for 60 minutes. The tensile strength properties and the fatigue crack growth (FCG) resistance were performed to investigate the effect of the austempering process in AISI 4140 steel. A significant increase in the yield strength for austempered steel is about 8.7 % and the elastic strain energy increases by 55.7 %. Austempered steel's fatigue crack cycle is longer than that of untreated steel. Data of stress intensity factor range (ΔK, MPa.m1/2) and FCG rate (da/dN, m/cycle) was constructed in double log plot x-y axes for determining the materials constants m and C according to Paris’s law equation using a linear regression method. From the curve of ΔK versus da/dN, the constant m value for austempered steel (m = 3.45) shows better resistance than untreated steel (m = 3.77). On the other hand, the constant C value of 1.409×10-12 for austempered steel is one order magnitude higher than that of untreated steel (C = 4.151×10-13). The resistance of austempered steel against fatigue crack growth can be attributed to the formation of a bainite structure.

Fatigue Crack Propagation in 5754 Aluminum Alloy under Four-Point Bending

The aluminum 5754 alloy is one of the widely used engineering materials in shipping, rivet making, tread plates and automotive industries. These engineering structures envisage variable loading conditions during their service. In addition to it, it is also experiencing seismic vibrations. Hence, the engineering components made from such aluminum alloy are susceptible to fatigue fracture. In the current study, the prediction of fatigue crack growth (FCG) in 5754 aluminum alloy was made using the exponential function. The beam specimen comes up with a cross-section of 25X25 mm2, a span length of 300 mm with a mechanical notch length of 2.70 mm at the centre was subjected to four-point bending (FPB) employing hydraulic INSTRON 8800 tensile testing apparatus. The periodic loading condition deformed the material up to large plastic deformation. The applied load was further down the elasticity of the material. The experimental data provided the relation between crack length (a) to the number of cycles (N) to failure. The response surface methodology (RSM) and modified exponential equation were used to predict the FCG. In RSM, when “stress intensity factor (K)” and “number of the cycle (N)" were considered independent variables, the response (a) was optimum (maximum) as compared to when “stress intensity factor range (del K)” and “fatigue crack growth rate (da/dN)” were considered independent variables. Hence, for designing the aluminum 5754 alloys as engineering structures, it was the number of cycles which provides a safe design as compared to da/dN. The modified exponential equation using an exponential function predicted the FCG for aluminum 5754 alloy in the form of a beam specimen. The anticipated results agreed with experimental data as the prediction ratio was 1.20 and the % deviation was 3.7.

Numerical prediction of fretting fatigue crack growth in scaled railway axles considering fretting wear evolution

To predict the fretting fatigue crack growth (FFCG) life of railway axles, a series of interrupted fatigue experiments were conducted on scaled railway axles. Subsequently, the evolutions of fretting wear and fatigue cracks in the press-fitted region were analyzed. Based on the test results, finite element models incorporating fretting wear evolution were established, and the FFCG was investigated using the maximum tangential stress criterion, cyclic resistance curve, and the modified NASGRO equation. The analysis revealed that the fretting wear evolution leads to stress redistribution at the press-fitted region, thereby promoting FFCG. When considering fretting wear evolution, the equivalent stress intensity factor (SIF) range of the crack remains above the threshold value throughout the short crack stage. However, neglecting fretting wear evolution results in the SIF range being below the threshold value for cracks shallower than 0.30 mm. This implies that considering fretting wear evolution enables life prediction throughout the short crack stage. As the crack length increases, the influence of fretting wear evolution on crack growth gradually diminishes. By accounting for fretting wear, a more accurate stress distribution in the press-fitted region can be obtained, leading to a more precise and conservative prediction of crack growth life.

Advances in Machine Learning Techniques Used in Fatigue Life Prediction of Welded Structures

In the shipbuilding, construction, automotive, and aerospace industries, welding is still a crucial manufacturing process because it can be utilized to create massive, intricate structures with exact dimensional specifications. These kinds of structures are essential for urbanization considering they are used in applications such as tanks, ships, and bridges. However, one of the most important types of structural damage in welding continues to be fatigue. Therefore, it is necessary to take this phenomenon into account when designing and to assess it while a structure is in use. Although traditional methodologies including strain life, linear elastic fracture mechanics, and stress-based procedures are useful for diagnosing fatigue failures, these techniques are typically geometry restricted, require a lot of computing time, are not self-improving, and have limited automation capabilities. Meanwhile, following the conception of machine learning, which can swiftly discover failure trends, cut costs, and time while also paving the way for automation, many damage problems have shown promise in receiving exceptional solutions. This study seeks to provide a thorough overview of how algorithms of machine learning are utilized to forecast the life span of structures joined with welding. It will also go through their drawbacks and advantages. Specifically, the perspectives examined are from the views of the material type, application, welding method, input parameters, and output parameters. It is seen that input parameters such as arc voltage, welding speed, stress intensity factor range, crack growth parameters, stress histories, thickness, and nugget size influence output parameters in the manner of residual stress, number of cycles to failure, impact strength, and stress concentration factors, amongst others. Steel (including high strength steel and stainless steel) accounted for the highest frequency of material usage, while bridges were the most desired area of application. Meanwhile, the predominant taxonomy of machine learning was the random/hybrid-based type. Thus, the selection of the most appropriate and reliable algorithm for any requisite matter in this area could ultimately be determined, opening new research and development opportunities for automation, testing, structural integrity, structural health monitoring, and damage-tolerant design of welded structures.

A hybrid modeling method of fatigue crack growth for gas turbine blades under combined high and low cycle fatigue loadings

Fatigue crack is one of the major faults of the gas turbine blades. Previous modeling studies on fatigue crack growth of blades mainly focus on a single numerical or analytical approach considering different sizes of cracks under various fatigue loadings. However, the actual crack propagation process is time-varying. The iteration and update of the model are necessary to evaluate the fatigue life of blades precisely. This paper proposes a hybrid modeling method to study blade crack growth under combined high and low cycle fatigue (CCF) loadings. The method emphasizes the combination and interaction between the finite element (FE) numerical simulation and the analytical calculation based on the fracture mechanics model. The current crack propagation length of the blade FE model is calculated from the crack growth rate obtained by the stress intensity factor (SIF) range according to Paris law. Then, a new SIF range is resolved from the FE model with the updated crack length. The proposed method is verified by the first-stage compressor blade from an in-service gas turbine. Results show that the crack growth is faster using the proposed hybrid modeling method than the traditional method. The update of the SIF range under CCF loadings cannot be ignored when predicting the fatigue life of the blade. Also, a sensitivity analysis is carried out. Suggestions are given on how to set the crack extension length at each stage during modeling, especially when the blade approaches failure.

Experimental Data-Driven approach for the evaluation of crack tip opening loads under variable amplitude loading

In this paper, we present a new data-driven approach for the evaluation of crack tip opening load (Pop) values under variable amplitude loading. These values are analysed for CT specimens manufactured from common aircraft grade aluminium alloys and subjected to several load sequences, all of which have a relatively high R-ratio content without significant overloads and underloads. Direct experimental measurements and outcomes of numerical simulations indicate that the Pop change abruptly from cycle to cycle. For this type of fatigue loading (including a military transport aircraft load spectrum with more than 400,000 turning points) and materials, it was found that Pop is well described by a linear function of the preceding minimum and maximum values of the applied loading. This finding provides a new simple way to evaluate the effective stress intensity factor range, which is often considered as a fatigue crack driving force. It is also verified via the use of machine learning that the proposed approach is capable of predicting the mean stress (R-ratio) effect on Pop for block loading. Therefore, the developed data-driven approach is a promising alternative to the computational methods, which currently dominate advanced fatigue life assessment procedures.

Influence of solution heating rate on grain structures and fatigue crack propagation behaviors of an aviation Al-Li alloy sheet

In this study, the effects of grain structure on the fatigue crack propagation (FCP) behaviors in the aviation Al-Li alloy sheets prepared by varying solution heating rates are discussed systematically. The results reveal that both the intensity of Goss texture and grain dimensions are improved with the decreased solution heating rates. Upon an identical stress intensity factor range (ΔK), the increased Goss texture of the sample with a solution heating rate of 2 ℃/min (2HR) promotes crack deflection, causing a decrease in FCP rates compared to that of the sample with a solution heating rate of about 100 ℃/min (FHR). The dislocations clustered near grain boundaries are more significant with increasing grain size during cyclic loading, which intensifies the strain localization. The accumulation of dislocations increases the possibility of micro-voids formation and creates the conditions for crack deflection, which results in a decrease in FCP rate. Besides, based on the calculation of the reversible plastic zone (RPZ) size, the initial ΔK entering the rapid propagation region increases with decreasing solution heating rates and exhibits a strong linear relationship with grain size.

Fatigue Crack Growth Analysis in Modified Compact Tension Specimen with Varying Stress Ratios: A Finite Element Study

In this study, the primary objective is to analyze fatigue crack propagation in linear elastic fracture mechanics using the SMART crack growth module in the ANSYS Workbench, employing the finite element method. The investigation encompasses several crucial steps, including the computation of stress intensity factors (SIFs), determination of crack paths, and estimation of remaining fatigue life. To thoroughly understand crack behavior under various loading conditions, a wide range of stress ratios, ranging from R = 0.1 to R = 0.9, is considered. The research findings highlight the significant impact of the stress ratio on the equivalent range of SIFs, fatigue life cycles, and distribution of deformation. As the stress ratio increases, there is a consistent reduction in the magnitude of the equivalent range of stress intensity factor. Additionally, a reciprocal relationship is observed between the level of X-directional deformation and the number of cycles to failure. This indicates that components experiencing lower levels of deformation tend to exhibit longer fatigue life cycles, as evidenced by the specimens studied. To verify the findings, the computational results are matched with the crack paths and fatigue life data obtained from both experimental and numerical sources available in the open literature. The extensive comparison carried out reveals a remarkable level of agreement between the computed outcomes and both the experimental and numerical results.

Atomistic fracture in bcc iron revealed by active learning of Gaussian approximation potential

The prediction of atomistic fracture mechanisms in body-centred cubic (bcc) iron is essential for understanding its semi-brittle nature. Existing atomistic simulations of the crack-tip under mode-I loading based on empirical interatomic potentials yield contradicting predictions and artificial mechanisms. To enable fracture prediction with quantum accuracy, we develop a Gaussian approximation potential (GAP) using an active learning strategy by extending a density functional theory (DFT) database of ferromagnetic bcc iron. We apply the active learning algorithm and obtain a Fe GAP model with a converged model uncertainty over a broad range of stress intensity factors (SIFs) and for four crack systems. The learning efficiency of the approach is analysed, and the predicted critical SIFs are compared with Griffith and Rice theories. The simulations reveal that cleavage along the original crack plane is the atomistic fracture mechanism for {100} and {110} crack planes at T = 0 K, thus settling a long-standing issue. Our work also highlights the need for a multiscale approach to predicting fracture and intrinsic ductility, whereby finite temperature, finite loading rate effects and pre-existing defects (e.g., nanovoids, dislocations) should be taken explicitly into account.