Structural health monitoring for damage detection and fatigue life estimation: a frequency-domain approach (#87)
D. E. Teixeira Marques1, D. Vandepitte2, V. Tita1
1 University of Sao Paulo, Department of Aeronautical Engineering, Sao Carlos, Brazil
Regarding structural design, the aeronautical industry is constantly looking for a compromise between weight reduction and robustness, the latter also being a consequence of the strong safety requirements imposed by regulatory agencies.
Structural Health Monitoring (SHM) systems have emerged as an alternative to this scenario. The main idea behind an SHM system is to monitor in real time the structural conditions of a component, which creates the obvious advantage of immediate detection of possible damages, as well as reducing or even eliminating stops for inspection, saving time and money for airline operators. Despite being the object of studies for several decades, SHM systems have a good prospect for an increase in its use and application in the next few years. Recent advances in sensors, signal processing, machine learning and artificial intelligence point to a future of widespread use of sensing and monitoring technologies for all kinds of purpose, with applications varying from big structures, such as airplanes and helicopters, to cars and even home appliances. In this context, the development of reliable and cost-effective monitoring systems is undoubtedly an important step towards that future.
The problem of estimating the service life of structures is also an old interest of researchers and industry, and it is particularly related to fatigue damage. Fatigue is a failure mode associated with cyclic loading of a component, on which the cumulation of damage leads to an ever-increasing degradation of its material properties, causing premature failure. As a consequence of the accumulation of fatigue damage, the material properties are reduced, and the structure may fail under load conditions that are well below its yielding or ultimate strength limit. By its turn, changes in the material properties will necessarily affect the structure’s vibrational response, as the latter is a function of structural properties such as its mass, rigidity, damping, etc. It is to be expected thus that Vibration Based Methods (VBM) may be able to detect and monitor fatigue damage.
It is worth noticing that VBM are usually performed in the frequency domain, as it provides more useful insight into the structure’s vibrational response. Alternatively, the procedure for fatigue analysis of structures can also be performed both in time or frequency domain. However, it is recognized that fatigue analysis in the frequency domain are especially suitable for structures that are subjected to random loads and for cases where the stress response occurs mainly as a consequence of rich structural dynamics (MRSNIK; SLAVIC; BOLTEZAR, 2018), which is the case for several aeronautical components.
Based on the exposed above, this work proposes to further investigate how VBM (and damage metrics) may correlate with fatigue damage and fatigue life, which may prove useful for an SHM system capable of using the structure’s vibrational response for both damage detection and estimation of its remaining service life. The research starts with the selection of the structure under analysis and a load case scenario. This allows for the calculation of the fatigue damage that is imposed on the structure and also to obtain the stress amplitude distribution from the load spectrum using the concepts of vibration-fatigue. Next, the stress distribution that is acting upon the structure is combined with concepts from Linear Elastic Fracture Mechanics to estimate the damage growth. Finally, experimental modal analyses are carried out in the structure in two conditions - prior and posterior of the load application - and a damage metric (DI – Damage Index) is used to obtain a quantitative estimation of the damage, which is then correlated to the structure’s fatigue life.
MRSNIK, M.; SLAVIC, J.; BOLTEZAR, M. Vibration fatigue using modal decomposition. Mechanical Systems and Signal Processing, v.98, p.548-556, 2018.
Keywords: Structural Health Monitoring, Vibration Based Methods, Fatigue
Estimation of Road Roughness Based on Vehicle Measurements (#90)
M. Burger1, M. Speckert1, K. Dreßler1
1 Fraunhofer ITWM, Kaiserslautern, Germany
In the vehicle development process, it is highly important to know as much as possible about the customer-specific market situation. The vehicle is designed to meet region- and customer-specific durability requirements. Thus, it is a primary goal to collect data and information about the environment, the vehicle will be used in. In order to assess vertical loads, the vertical road input is the most decisive impact factor; it is, additionally, roughly independent from the vehicles that travels on it. It is common practice to describe road roughness by scalar indices, which can be computed from the corresponding road profiles. A widely used index is the International Roughness Index (IRI). Here, we focus on this index concept as well as on the ISO roughness coefficient, denoted by C.
A straightforward way to obtain road profiles and, whence, roughness information is to directly measure the profile. This is possible, e.g., using expensive and complex laser scanners and the resulting resolution decreases with increasing speed of the measurement vehicle. Consequently, it is desirable to develop alternatives. In contrast to direct profile measurement, we propose and discuss an algorithmic approach to derive road roughness indices on the basis of comparably simple vehicle models (e.g. a quarter-car model) and easy-to-measure vehicle quantities, e.g., axle and frame accelerations.
The algorithms use a stochastic modelling approach that gives a road profile on a certain segment as realization of a stochastic process. The stochastic process, typically a Gaussian or Laplacian processes, is, in turn, characterized by the roughness indices. Thus, the latter are computed by solving a suitable inverse problem directly for the roughness coefficients. We present several variants of suited inverse problem formulation and discuss benefits and drawbacks. We also propose an approach that relies on the definition of a new index concept: the vehicle-speed-dependent roughness index (VSRI), which can be seen as a straightforward generalization of IRI. Additionally, we present a treatment of the inverse problems in the Bayesian way, this allows the quantification of uncertainty of the derived roughness indices in terms of assumed model and measurement errors.
The methods are highly efficient from a computational viewpoint and, thus, they are applicable for large databases (e.g., existing campaigns) or even in on-board and realtime scenarios during customer use. Following the presented approach, it is possible to obtain statistical and distributional information of road roughness in certain regions or markets that can be used in the vehicle development process. Moreover, used in an on-board application, the obtained information may be used as well for predictive maintenance purposes.
Keywords: Road Roughness, Inverse Problems, Bayesian Methods, Georeferenced Data
A metro rail axle multi-tiered integrity concept (#113)
R. Heim1, K. Liedgens2
1 Fraunhofer LBF, R&D Division Structural Durability, Darmstadt, Germany
For both, passenger and freight, rail transport is significantly safer than road transport. In Germany the average dangerous goods accident rate in the years 2004 to 2013 was on-road 42 times higher than that on the rail,and the average fatality risk in the years 2008 to 2017 for people in cars was 56 times higher than in railwayvehicles . Statistically a person can travel more than 33 billion kilometer with trains in Germany before afatal injury may occur - in EU28 average that number is still almost 7.7 billion. This extraordinary safety level is given by a self-contained transportation system having tracks, signals and vehicles which all operatesafely for a huge number of years. The major structures of a rail vehicle - such as wheel set, bogie and carbody - may have a useful life of up to 50 years which is then equivalent to very high numbers of load cycles.Though carefully designed and manufactured, material fatigue because of stress concentrations at cross-section transitions or non-metallic inclusions, corrosion pits or flying ballast impacts may arise on wheelsetshafts.
On July 2008 a hollow shaft of a German high speed train cracked at a mileage of 3.09 million kilometers  which refers to more than 10e9 load cycles adressing the VHCF (very high cycle fatigue) regime. On Christmas 2016 a total number of 7 freight wagons derailed in Austria due to a double broken shaft in leading position at the second wagon. That shaft was manufactured in 1984 and in February 2015 an ultrasonic NDT (non-destructive testing) for crack checking was performed without any issues. The axle cracked at both tran-sition radii to the middle part because of corrosion pits which were underneath the protective coating. According to the federal safety investigating body the amount of damage was about 2.1 mio. EUR . Hence maintaining a heavily loaded structure over a long period makesan integrity concept necessary, because even a well-designed component cannot exclude effects which maylead to crack initiation and propagation.
In this paper a multi-tiered integrity concept for metro rail axles is described - including a fracture mechanicsbased inspection interval, which was developed by using spectrum loading for full scale crack growth experiments and computational engineering. This integrity concept was introduced for a specific metro vehicle typein 2008 and had its focus on the definition of an NDT inspection interval for aged axles at a mileage of >2.5mio. kilometers to extend their ‘safe-life period’ without compromising vehicle safety. Therefore, the maximum operational loading for the full laden metro train was analyzed and transferred into an efficient full scale test lab environment to evaluate the most critical crack growth scenario. By using computational stress intensity factors together with a generalized form of the Paris equation to account for the effect of the stress ratio R on the crack growth rate, the Walker equation, the full laden crack propagation was conveyed to the true operational conditions which are characterized by an utilization rate significantly below 30 percent.Then the computed operational crack growth rate was aligned to the PoD (probability of detection) of manual ultrasonic NDT using far end scans to avoid any risk for unstable crack growth even in cases of possible NDT errors.
The paper finally reviews the value of the integrity concept by looking at fatigue incidents which were reported by the transport organization in 2010, 2015 and 2017.
 Allianz Pro Schiene | Statistisches Bundesamt - 2018
 Klinger, C.; Bettge, D.: Axle fracture of an ICE3 high speed train; Engineering Failure Analysis, Vol. 35,Pages 66-81, December 2013
 Bundesministerium für Verkehr, Innovation und Technologie: Untersuchungsbericht - Entgleisung Z48600; Bericht 795.376_EUB_1.0_final, 2018
Keywords: rail vehicle, crack growth, inspection interval, NDT, full scale testing
Fatigue Life Estimation under Variable Amplitude Loading including a Consideration of Size Effects (#104)
M. Hell1, R. Wagener1, T. Melz1
1 Fraunhofer-Institute for Structural Durability and System Reliability LBF, Component Related Materialbehaviour, Darmstadt, Germany
The majority of safety relevant components is subjected to variable amplitude loading, containing loads in the High and Very High Cycle (HCF and VHCF) Fatigue Regime due to regular use or admissible overloading, as well as loads in the Low Cycle Fatigue (LCF) Regime originating from misuse. In presence of stress gradients, for example at notches, bearing shoulders etc. also macroscopically elastic loading may cause cyclic plasticity whithin a highly stressed volume in the vicinity of the maximum stress, thus inducing residual mean stresses. In order to be able to treat purely elastic as well as elasto-plastic loading at the same time, local strain-based approaches with elasto-plastic material behaviour are the method of choice. The cyclic material behaviour is characterized using strain-controlled fatigue tests on un-notched specimens, deriving a stress-strain relation for the computation of the stress or strain response to an external load and a strain-life relation, which is used to determine the fatigue life corresponding to the local stress-strain state. The cyclic stress-strain behaviour is linked to the microstructure of the material and the evolution of dislocation structures under cyclic loading influences the cyclically transient behaviour in terms of cyclic hardening or softening as well as the cyclically stabilized stress-strain relation. Depending on the slip character of the material, the stress-strain relation may depend significantly on the applied load-time function. With respect to the fatigue life estimation under variable amplitude loading, it is therefore necessary, to assess the stress-strain behaviour under constant amplitude loading as well as under variable amplitude loading, using, for example, incremental step tests.
The basic assumption regarding the experimental material characterization for the application of strain-based fatigue design approaches with elasto-plastic material behaviour is the equivalency of the material behaviour, which may be measured within a homogeneously loaded specimen, to the material behaviour of an infinitesimal small material volume at the stress-concentration. Depending on the stress gradient, gradients in material properties, for example in welded or additively manufactured components, and the microstructure of the material, the stress at the notch root may be averaged over a certain microstructural length and may also change in size by plastic deformation, especially during variable amplitude loading. As the extension of the highly stressed material volume, which influences the fatigue life, depends on the component geometry and the load magnitude, it seems appropriate, by analogy with the nominal stress approach, to treat those geometry related effects as size effects. In order to be able to understand possible size effects in local strain-based fatigue design approaches, it is necessary to redefine the interfaces for the consideration of the influence of the component geometry and material properties on fatigue life on the basis of elasto-plastic material behaviour. The presented work will show, how the mechanical, the statistical, the technological and the surface related size effects, being defined phenomenologically by Kloos, may be transferred to the fatigue life estimation of components under variable amplitude loading with strain-based fatigue design approaches. A comparison between numerical and experimental data will evaluate, which of the size effects play a dominant role and quantify the potential for improvement of the accuracy of the fatigue life estimation by a consideration of size effects.
Keywords: Fatigue, Life Estimation, Variable Amplitude Loading