Crack-like Features Research Articles

Loss function inversion for improved crack segmentation in steel bridges using a CNN framework

Automating bridge visual inspection using deep learning algorithms for crack detection in images is a prominent way to make these inspections more effective. This paper addresses several challenges associated with crack detection: (1) data imbalance, caused by a small crack area as compared to the background, and (2) a high false positive rate, due to a large amount of crack-like features in the background. First, a new benchmark dataset is presented, containing images of cracks in steel bridges along with pixel-wise annotations. Secondly, the importance of incorporating background patches is examined to assess their impact on network performance when applied to high resolution images of cracks in steel bridges. Finally, a loss function is introduced that enables the use of a relatively large number of background patches in neural network training. The proposed approaches yield a significant reduction in false positive rates, thereby improving the overall performance of crack segmentation.

CNN-based automated approach to crack-feature detection in steam cycle components

Periodic manual inspection by trained specialists is an important element of asset management in the nuclear industry. Detection of cracks caused by stress corrosion is an important element of remote visual inspection (RVI) in power plant steam generator components such as boilers, superheaters and reheaters. Challenges exist in the interpretation of RVI footage, such as high degree of concentration for reviewing lengthy and disorienting footage due to narrow field of view offered by endoscope. Deep learning is considered useful to automate crack detection process for improved efficiency and accuracy, and has enjoyed success in related applications. This article utilises a new application of automated crack feature detection in steam cycle components to demonstrate a transferrable data-driven framework for a variety of anomaly inspections in such structures. Specifically, a case study of superheater (a type of reactor pressure vessel head) anomaly inspection is presented to automatically detect regions of crack-like features in inspection footage with a good accuracy of 92.97 % using convolutional neural network (CNN), even in challenging cases. Due to the black-box nature of the CNN classification, the explicability of the classification results is discussed to enhance the trustworthiness of the detection system.

Comparing the oxidation of FeCrAl alloys in isothermal and cyclic high-temperature steam environments

This study investigates the oxidation behavior of FeCrAl (APMT and C26M) alloys in high-temperature steam at 1000 °C using isothermal and cyclic conditions. The FeCrAl alloys exhibited a fully ferritic body-centered cubic structure with no significant phase transformation, and the manufacturing procedures of the alloys resulted in varying grain sizes and porosity. In the 5 h isothermal test, the protective alumina film provided corrosion resistance without cracking in the high-temperature steam environment. However, thermal cycling seemed to compromise passivity; expansion and contraction during heating and cooling allowed oxygen penetration into the underlying metal in local regions, leading to transgranular crack-like features. Spallation of oxides was also observed. Post-exposure analysis of the oxide film formed on the FeCrAl alloys revealed two distinct oxide layers on the alloy: an outer Al-rich oxide film and an inner mixed oxide enriched in Cr and Fe.

Identification and development of a new local corrosion mechanism in a Laser Engineered Net Shaped (LENS) biomedical Co-Cr-Mo alloy in Hank’s solution

In this work, the microstructure and the corrosion behavior of a biomedical grade Co-Cr-Mo (CCM) alloy manufactured by laser engineered net shaping (LENS) was characterized. Electron microprobe maps identified the micro-segregation of Cr and Mo to the cell boundaries in the microstructure of the LENS alloys. Cyclic polarization in Hank’s solution and study of the corroded surface under scanning electron microscopy (SEM) revealed the development of micro-galvanic cells and new crack-like features which served as active sites of localized dissolution of the alloy. Based on this evidence, a new corrosion mechanism that constituted of selective dissolution of Cr/Mo-depleted regions and development of crack-like features at junctions between the various cell types was proposed for LENS CCM alloys.

Shear banding instability in multicomponent metallic glasses: Interplay of composition and short-range order

The shear-banding instability in quasistatically driven bulk metallic glasses emerges from collective dynamics, mediated by shear transformation zones and associated nonlocal elastic interactions. It is also phenomenologically known that sharp structural features of shear bands are typically correlated to the sharpness of the plastic yielding transition, being predominant in commonly studied alloys composed of multiple different elements, that have very different atomic radii. However, in the opposite limit where elements' radii are relatively similar, plastic yielding of bulk metallic glasses is highly dependent on compositional and ordering features. In particular, a known mechanism at play involves the formation of short-range order dominated by icosahedra-based clusters. Here, we report on atomistic simulations of multicomponent metallic glasses with different chemical compositions showing that the degree of strain localization is largely controlled by the interplay between composition-driven icosahedra-ordering and collectively-driven shear transformation zones. By altering compositions, strain localization ranges from diffuse homogenized patterns to singular crack-like features. We quantify the dynamical yielding transition by measuring the atoms' susceptibility to plastic rearrangements, strongly correlated to the local atomic structure. We find that the abundance of short-range ordering of icosahedra within rearranging zones increases glassy materials' capacity to delocalize strain. This could be understood on the basis of structural heterogeneities that are enhanced by the presence of local order. The kind of plastic yielding can be often qualitatively inferred by the commonly used compositional descriptor that characterizes element associations, the misfit parameter ${\ensuremath{\delta}}_{a}$, and also by uncommon ones, such as shear-band width and shear-band dynamics' correlation parameters.

Experimental and numerical analysis of fatigue cracks emanating from internal defects in Ti6Al4V SLM

In the present work a series of fatigue tests on Ti6Al4V SLM parts are analyzed via both SEM and confocal microscopy. On the one hand, fracture surfaces are studied, and a common pattern is found, formed by a series of different textures which show the complex crack front evolution from crack initiation in a particular internal defect to complete failure. On the other hand, fatigue strength is observed to highly depend on the defect where initiation takes place, so experimental observation of that critical entity is carried out. Both defect location within the specimen and shape are studied, considering the crack-like or blunt feature of every defect. Once experimental analysis is complete, numerical simulation is attempted. By making use of critical defect and residual stress measurements obtained experimentally, both fatigue strength and crack front evolution are estimated.

Sensitivity of material failure to surface roughness: A study on titanium alloys Ti64 and Ti407

The relationship between material failure and surface roughness has been investigated using two titanium alloys: Ti64 and the more ductile Ti407. Three surface types were created (polished, sandblasted and scratched) with instances spanning a wide range of average roughness. The surfaces were tested in three-point bending with the imparted roughness on the tensile under-surface of a rectangular beam specimen. Results showed failure of Ti64 to be highly sensitive to both magnitude and orientation of roughness. High roughness in the maximum tensile stress direction (and scratch like features perpendicular to this direction) were most detrimental. Thus, strain-to-failure (and work-to-failure) in Ti64 dropped off significantly with increasing surface roughness in the tensile direction. Finite element modelling of the test indicated that cracks initiate at zones of high plastic strain at the tips of roughness valleys due to high local surface curvature. Thus, roughness can be considered as a series of blunt crack-like features where larger crack tip curvature induces greater likelihood of crack propagation. Contrastingly, the mechanical response of Ti407 was insensitive to surface roughness owing to its significantly greater ductility. Thus, designers need to be aware of the sensitivity of failure of particular materials to surface roughness. The insensitivity of Ti407 is advantageous, but the sensitivity of failure to surface roughness in a material like Ti64 is potentially serious if not properly accounted for.

SDDNet: Real-Time Crack Segmentation

This article reports the development of a pure deep learning method for segmenting concrete cracks in images. The objectives are to achieve the real-time performance while effectively negating a wide range of various complex backgrounds and crack-like features. To achieve the goals, an original convolutional neural network is proposed. The model consists of standard convolutions, densely connected separable convolution modules, a modified atrous spatial pyramid pooling module, and a decoder module. The semantic damage detection network (SDDNet) is trained on a manually created crack dataset, and the trained network records the mean intersection-over-union of 0.846 on the test set. Each test image is analyzed, and the representative segmentation results are presented. The results show that the SDDNet segments cracks effectively unless the features are too faint. The proposed model is also compared with the most recent models, which show that it returns better evaluation metrics even though its number of parameters is 88 times less than in the compared models. In addition, the model processes in real-time (36 FPS) images at 1025 × 512 pixels, which is 46 times faster than in a recent work.

The effect of fluid and solid properties on the auxetic behavior of porous materials having rock-like microstructures

Materials with a negative Poisson's Ratio (PR), known as auxetics, exhibit the counterintuitive behavior of becoming wider when uniaxially stretched and thinner when compressed. Though negative PR is characteristic of polymer foams or cellular solids, tight as well as highly porous rocks have also been reported to exhibit negative PR. The paper proposes a novel auxetic structure based on pore-space configuration observed in rocks. We developed a theoretical auxetic 3D model consisting of rotating rigid bodies. To alleviate the mechanical assumption of rotating bodies, the theoretical model was modified to include crack-like features being represented by intersecting, elliptic cylinders. We then used a 3D printer to create a physical version of the modified model, whose PR was tested. We also numerically explored how the compressibility of fluids located in the pore-space of the modified model as well as how the elastic properties of the material from which the model is made of affect its auxetic behavior. We conclude that for a porous medium composed of a single material saturated with a single fluid (a) the more compliant the fluid is and (b) the lower the PR of the solid material, the lower the PR value of the composite material.

Effect of random defects on dynamic fracture in quasi-brittle materials

We propose an asynchronous spacetime discontinuous Galerkin (aSDG) method combined with a novel rate-dependent interfacial damage model as a means to simulate crack nucleation and propagation in quasi-brittle materials. Damage acts in the new model to smoothly transition the aSDG jump conditions on fracture surfaces between Riemann solutions for bonded and debonded conditions. We use the aSDG method’s powerful adaptive meshing capabilities to ensure solution accuracy without resorting to crack-tip enrichment functions and extend those capabilities to support fracture nucleation, extension and intersection. Precise alignment of inter-element boundaries with flaw orientations and crack-propagation directions ensures mesh-independent crack-path predictions. We demonstrate these capabilities in a study of crack-path convergence as adaptive error tolerances tend to zero. The fracture response of quasi-brittle materials is highly sensitive to the presence and properties of microstructural defects. We propose two approaches to model these inhomogeneities. In the first, we represent defects explicitly as crack-like features in the analysis domain’s geometry with random distributions of size, location, and orientation. In the second, we model microscopic flaws implicitly, with probabilistic distributions of strength and orientation, to drive nucleation of macroscopic fractures. Crack-path oscillation, microcracking, and crack branching make numerical simulation of dynamic fracture particularly challenging. We present numerical examples that explore the influence of model parameters and inhomogeneities on fracture patterns and the aSDG model’s ability to capture complex fracture patterns and interactions.

Interaction of a Sliding Wedge with a Metallic Substrate Containing a Single Inhomogeneity

Surface folding is a recently discovered unsteady plastic flow mode in metal sliding, caused by microstructure-related spatial heterogeneity. In this work, we simulate a wedge sliding against a metallic specimen with a single, near-surface inhomogeneity or pseudograin, representative of unit asperity–grain interactions. The inhomogeneity is either plastically softer or harder than the surrounding substrate and is perfectly bonded to it. Remarkably, this simple model is able to reproduce numerous experimentally observed aspects of unsteady sliding, including the development of bumps and depressions on the prow, surface self-contact (fold) formation and the development of crack-like damage features on the residual surface. It is found that a hard inhomogeneity causes surface depressions, while a soft inhomogeneity causes surface bumps. Both features develop into folds, and subsequently, crack-like damage features. The model also reproduces the effects of sliding incidence angle, friction and size of inhomogeneity on the sliding response. The propensity to bump and fold formation decreases on increasing the depth at which the inhomogeneity is embedded. Analysis reveals that plastic buckling is a plausible mechanism for the development of perturbations on the surface of the prow. The present model differs from the classical triboplastic models of Oxley and Torrance mainly by way of the added inhomogeneity and is a minimal model to produce folding and associated damage in a single sliding pass. Implications for wear and deformation processing are discussed.

Stress assisted corrosion in a crude oil production tube made from L80 steel

Abstract Failure investigation was carried out on a crude oil production tube made of L80 steel. The failure appeared like a long crack, but on closer examination, it was found to have formed by localized thinning along the longitudinal axis of the tube. Remaining areas of the tube was unaffected and without any appreciable corrosion. Deep depressions and crack-like features were observed in the cross section of the sample near the location of thinning. Analysis of the corrosion products taken from an affected location revealed that the corrosion products were comprised of sulphides of iron. Microhardness measurements made on the tube indicated relatively higher hardness at the affected locations. The results of electrochemical corrosion tests conducted in 0.2 M sodium chloride and 0.2 M sodium chloride + 0.01 M sodium thiosulphate solutions showed a higher corrosion rate for the sample at the affected location. The investigation concluded that the failure of the tubing was due to stress assisted corrosion (SAC). The tube at the failed location was at an elevated stress due to leaning or resting of the pump rod, which resulted in the accelerated corrosion.

Surface folding in metals: a mechanism for delamination wear in sliding.

Using high-resolution, in situ imaging of a hard, wedge-shaped model asperity sliding against a metal surface, we demonstrate a new mechanism for particle formation and delamination wear. Damage to the residual surface is caused by the occurrence of folds on the free surface of the prow-shaped region ahead of the wedge. This damage manifests itself as shallow crack-like features and surface tears, which are inclined at very acute angles to the surface. The transformation of folds into cracks, tears and particles is directly captured. Notably, a single sliding pass is sufficient to damage the surface, and subsequent passes result in the generation of platelet-like wear particles. Tracking the folding process at every stage from surface bumps to folds to cracks/tears/particles ensures that there is no ambiguity in capturing the mechanism of wear. Because fold formation and consequent delamination are quite general, our findings have broad applicability beyond wear itself, including implications for design of surface generation and conditioning processes.

Correlations between formation properties and induced seismicity during high pressure injection into granitic rock

We reviewed published results from six projects where hydraulic stimulation was performed in granitic rock. At each project, fractures in the formation were well-oriented to slip at the injection pressures used during stimulation. In all but one case, thousands of cubic meters of water were injected, and in every case, flow rates on the order of tens of liters per second were used. Despite these similarities, there was a large variation in the severity of induced seismicity that occurred in response to injection. At the three projects where induced seismicity was significant, observations at the wellbore showed evidence of well-developed brittle fault zones. At the three projects where induced seismicity was less significant, observations at the wellbore indicated only crack-like features and did not suggest significant fault development. These results suggest that assessments of the degree of fault development at the wellbore may be useful for predicting induced seismicity hazard. We cannot rule out that the differences were caused by variations in frictional properties that were unrelated to the degree of fault development (and it is possible that there is a relationship between these two parameters). The projects with more significant seismicity tended to be deeper, and if this is a meaningful correlation, it is unclear whether depth influenced seismic hazard through the degree of fault development, frictional properties, or some other variable. The results of this paper are not conclusive, but they suggest that there may be significant opportunity for future research on identifying geological conditions that increase induced seismicity hazard.

The Micromechanics of Biological and Biomimetic Staggered Composites

Natural materials such as bone, tooth and nacre achieve attractive properties through the “staggered structure”, which consists of stiff, parallel inclusions of large aspect ratio bonded together by a more ductile and tougher matrix. This seemingly simple structure displays sophisticated micromechanics which lead to unique combinations of stiffness, strength and toughness. In this article we modeled the staggered structure using finite elements and small Representative Volume Elements (RVEs) in order to explore microstructure-property relationships. Larger aspect ratio of inclusions results in greater stiffness and strength, and also significant amounts of energy dissipation provided the inclusions do not fracture in a brittle fashion. Interestingly the ends of the inclusions (the junctions) behave as crack-like features, generating theoretically infinite stresses in the adjacent inclusions. A fracture mechanics criterion was therefore used to predict the failure of the inclusions, which led to new insights into how the interfaces act as a “soft wrap” for the inclusions, completely shielding them from excessive stresses. The effect of statistics on the mechanics of the staggered structure was also assessed using larger scale RVEs. Variations in the microstructure did not change the modulus of the material, but slightly decreased the strength and significantly decreased the failure strain. This is explained by strain localization, which can in turn be delayed by incorporating waviness to the inclusions. In addition, we show that the columnar and random arrangements, displaying different deformation mechanisms, lead to similar overall properties. The guidelines presented in this study can be used to optimize the design of staggered synthetic composites to achieve mechanical performances comparable to natural materials.

From viscous fingering to elastic instabilities

An analytical and numerical study of the linear Saffman–Taylor instability for a Maxwell viscoelastic fluid is presented. Results obtained in a rectangular Hele–Shaw cell are complemented by experiments in a circular cell corroborating the universality of our main result: The base flow becomes unstable and the propagating disturbances develop into crack-like features. The full hydrodynamics equations in a regime where viscoelasticity dominates show that perturbations to the pressure remain Laplacian. Darcy’s law is expressed as an infinite series in the cell thickness. An unique dimensionless parameter λ¯, equivalent to a relaxation time, controls the growth rate of the perturbation. λ¯ depends on the applied gradient of pressure, the surface tension, the cell thickness, and the elastic modulus of the fluid. For small values of λ¯, Newtonian behavior dominates whereas for higher values of λ¯ viscoelastic effects appear. For the critical value λ¯=λc¯≃10 a blowup is predicted and fracture-like patterns are observed.

Laser Generated Crack-Like Features Developed for Assessment of Fatigue Threshold Behavior

Abstract Developing methods to evaluate damage tolerance in high cycle fatigue is an important issue for the propeller industry. Traditional methods to establish threshold stress intensity factors for aluminum and martensitic steel have relied mainly on load shedding through-crack test specimens resulting in a pre-cracking history and the associated closure effects. For threshold determination of real life defects, a surface-crack test method, devoid of pre-cracking and closure issues, is needed. To this end, laser generated crack-like features (LGCLFs) were created and have been used successfully to evaluate the threshold crack growth properties of cracks in 7075-T651 with and without shot peening in axial R = 0.1 loading. For this work, a laser part marking system was used to generate a LGCLF semi-circular in shape, and measuring ∼0.015 in. (0.381 mm) deep by ∼0.030 in. (0.762 mm) long, with a notch height of 0.0007–0.001 in. (17.8–25.4 μm) with a tip radius of 0.000 14 in. (3.5 μm). When evaluated metallographically, these laser notches exhibited a very thin recast layer and a negligible heat affected zone. Specimens produced with these flaws were tested in R=0.1 tension-tension fatigue to produce a stress versus life plot of the data. A NASGRO crack growth model was used to generate stress versus cycle data points for the same geometry over the same stress range with a 7075-T651 material model and a crack growth threshold of 1.35 ksi ⋅(in.)0.5 resulting in very good agreement with the test data. This threshold compares favorably with the small crack threshold of 1.27–1.33 ksi⋅(in.)0.5 produced by da/dN testing performed by NASA at R=0.1 using a compressive pre-cracking method. Testing with samples shot peened prior to generating a LGCLF exhibited significantly higher endurance stresses and failure initiations were not from the LGCLF.

Continuum and atomistic modeling of electromechanically-induced failure of ductile metallic thin films

A multiscale modeling approach is presented for the analysis of electromechanically-induced void morphological evolution and failure in ductile metallic thin films, which are used for device interconnections in integrated circuits. Self-consistent mesoscopic simulations of surface morphological evolution are combined with atomistic calculations of surface properties and molecular-dynamics (MD) simulations of plastic deformation mechanisms in the vicinity of void surfaces. Results are presented that demonstrate a coupled mode of surface instability that leads to formation of electromigration-induced slit-like features and stress-induced crack-like features on void surfaces. In addition, MD results are presented for the dislocation-mediated mechanism and the kinetics of void growth in ductile metallic systems subject to hydrostatic and biaxial tensile strains. The incorporation of MD-derived constitutive information for plastic deformation into the mesoscopic analysis and simulation framework also is discussed.

Damage propagation in multidirectional composites subjected to compressive loading

The mode of compressive damage growth in multidirectional composites is investigated. A detailed systematic study of the nature of damage evolution is presented and the microstructural basis for treating compressive damage as a cracklike feature is explained. Experimental and theoretical findings conclude that there is a strong scientific basis for treating compressive damage growth in fiber-reinforced laminated polymer matrix composites as a bridged mode-I crack in compression. A large-scale crack-bridging model has been successfully used to predict the steady-state damage propagation stress as a function of damage growth in notched composites. These predictions are in good agreement with experimental observations.

Modeling of electromechanically-induced failure of passivated metallic thin films used in device interconnections

A theoretical analysis is presented of the failure of metallic thin films used for device interconnections in integrated circuits. Failure is mediated by void dynamics, which is driven by surface electromigration and processing related residual thermal stresses in the films. The analysis is based on surface mass transport modeling coupled strongly with the electrostatic and elastic deformation problems in the metallic films. Special emphasis is placed on the combined effects on void dynamics of anisotropy both in surface diffusivity along a void surface and in the applied stress tensor. A systematic parametric study is carried out based on self-consistent numerical simulations of surface morphological evolution. Void dynamics is analyzed and results are presented for void morphological stability in terms of critical stress levels as a function of stress state and surface mobility anisotropy. Finally, the role of plastic deformation is discussed around crack-like features emanating from void surfaces in ductile metallic films based on results of molecular-dynamics simulations in Cu.