17 results
Now showing1 - 10 of 17
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Prognose von Anrissorten und Lebensdauer von Stumpfstößen mittels explainable artificial intelligence(Deutscher Verband für Schweißen und verwandte Verfahren e.V. (DVS), 2022-09-19); Stumpfstöße sind in vielen Branchen üblich. Das Ermüdungsverhalten solcher Verbindungen hängt von zahlreichen Faktoren ab wie z. B. von der Höhe der Belastung, der lokalen Schweißnahtgeometrie oder der Festigkeit des Grundmaterials. Erschwerend kommt hinzu, dass diese Faktoren häufig in Wechselwirkung miteinander stehen, ihr gegenseitiger Einfluss jedoch aufgrund der großen Anzahl der erforderlichen Prüfungen und der statistischen Natur der Variablen mit üblichen statistischen Verfahren kaum quantifiziert werden kann. Darüber hinaus ist die Bewertung der Betriebsfestigkeit mittels dreidimensionalen Finite Elemente Berechnungen zu zeitintensiv, um Bewertungen direkt während der Fertigung durchzuführen. Folglich weicht das Ermüdungsverhalten solcher Verbindungen oft erheblich von Vorhersagen und experimentellen Ergebnissen ab. Daher sind alternative Ansätze erstrebenswert, welche es ermöglichen verschiedene Einflussfaktoren und ihre Wechselwirkungen zu berücksichtigen. Zu diesem Zweck werden in dieser Studie Algorithmen des maschinellen Lernens eingesetzt, um Versagensorte und die Anzahl der Lastzyklen zum Bruch mehrerer hundert Wöhlerversuche von Stumpfstößen vorherzusagen. Zusätzlich zu genauen Vorhersagen ist ein Verständnis der Relevanz und des gegenseitigen Einflusses der Faktoren erwünscht, z. B. eine Rangfolge der wichtigsten Faktoren. Die Erfassung des Einflusses mehrerer möglicherweise interagierender Faktoren erfordert jedoch in der Regel komplexe nichtlineare maschinelle Lernmodelle. Wir haben Gradient Boosted Trees verwendet. Da es sich hierbei um Black-Box-Modelle handelt, wurde das SHAPley Additive eXplanations (SHAP)-framework verwendet, um die Vorhersagen zu erklären, d. h. um einflussreiche Merkmale und ihre Wechselwirkungen zu bestimmen. Schließlich werden die Modellerklärungen mit dem Domänenwissen verknüpft.Publicationtype: Conference Paper28 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Comparison of machine learning and stress concentration factors-based fatigue failure prediction in small-scale butt-welded jointsFatigue behavior of welded joints is significantly influenced by numerous factors, for example, local weld geometry. A representative quantity for the influence of the notch effect created by the local weld geometry is the stress concentration factor (SCF). Thus, SCFs are often used to estimate fatigue failure locations and fatigue strength; however, this simplifies the mutual effect of other influencing factors. Consequently, fatigue strength estimates for welded joints may deviate from experimental results. Machine learning techniques offer an alternative to traditional fatigue assessment approaches based on SCFs. This study presents a comparison of failure location predictions and number of cycles to failure for 621 fatigue tests of small-scale butt-welded joints. In addition, an understanding of importance and mutual influence of the factors is desired. We used gradient boosted trees in combination with the SHapley Additive exPlanation framework to identify influential features and their interactions.Publicationtype: Journal ArticleScopus© Citations 11 85 336 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Predicting compressive strength and behavior of ice and analyzing feature importance with explainable machine learning modelsBuilding and using ice-related models is challenging due to the complexity of the material. A common issue, shared by both material models and (semi-)empirical approaches, is estimating unknown input parameters such as compressive strength. This is often done with additional empirical formulas which have drawbacks, e.g., they are based on a limited amount of data. Regarding material modeling, a strongly related problem is the prioritization of effects to include in the model. This is mostly done based on a subjective mix of knowledge, model purpose, and experimental studies limited to that purpose, which risks overlooking effects or interaction of effects, and limits transferability of material models to other scenarios. To tackle these issues, a hybrid approach of domain knowledge and explainable machine learning was used. A large ice test database was compiled to train machine learning models to predict compressive strength and behavior type. The machine learning models’ predictions were more accurate than existing empirical or analytical approaches and can thus be used as an alternative, though less straightforward, tool for such predictions. Further, the SHAP explainable AI method was applied to the predictions. Impact rankings of experimental parameters and interaction effects between parameters were analyzed and discussed in terms of ice mechanics. Top features were strain rate, triaxial stress state, temperature, and loading direction, but impact rankings were highly dependent on prediction target and type of ice. Few interaction effects were found. The approach adds objectivity to the prioritization of effects for material modeling and generated further insights into ice mechanics. It is also considered useful for other natural materials or generally when there is more data than knowledge.Publicationtype: Journal Article53 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Statistical characterization of stress concentrations along butt joint weld seams using deep neural networks(Multidisciplinary Digital Publishing Institute (MDPI), 2022-06-15); ; ; ; In order to ensure high weld qualities and structural integrity of engineering structures, it is crucial to detect areas of high stress concentrations along weld seams. Traditional inspection methods rely on visual inspection and manual weld geometry measurements. Recent advances in the field of automated measurement techniques allow virtually unrestricted numbers of inspections by laser measurements of weld profiles; however, in order to compare weld qualities of different welding processes and manufacturers, a deeper understanding of statistical distributions of stress concentrations along weld seams is required. Hence, this study presents an approach to statistically characterize different types of butt joint weld seams. For this purpose, an artificial neural network is created from 945 finite element simulations to determine stress concentration factors at butt joints. Besides higher quality of predictions compared to empirical estimation functions, the new approach can directly be applied to all types welded structures, including arc- and laser-welded butt joints, and coupled with all types of 3D-measurement devices. Furthermore, sheet thickness ranging from 1 mm to 100 mm can be assessed.Publicationtype: Journal ArticleScopus© Citations 10 86 147 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Ice pressure measurements with tekscan sensors-behaviour, calibration and application limits(American Society of Mechanical Engineers (ASME), 2022-06); ; ; ; Ships and offshore structures operating in ice-covered sea areas have to withstand ice loads. In particular, the distribution of ice pressure in the contact area is critical. The pressure dis-tribution changes temporally and spatially and depends on the ice failure behaviour. Experimental measurements of ice pres-sures are possible with pressure mapping systems. A commonly used system in academia uses sensors manufactured by TekScan. The resulting measurements can be used to gain insight into ice failure behaviour, to establish design pressure patches, and to develop numerical models. However, various issues emerge when using these pressure measurement systems, e.g., equilibra-tion and calibration, which result in measurement uncertainties. Hence, this paper summarize and discuss possible issues and gives recommendations to address them.Publicationtype: Conference Paper39 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Analyzing flexural strength data of ice: how useful is explainable machine learning?(American Society of Mechanical Engineers (ASME), 2022-06); ; ; The climate crisis results in a rapid sea ice decline, making shipping routes accessible for longer durations throughout the year and therefore increasing maritime traffic. At the same time, ice-structure interaction is known to cause damage to ships and structures. In this context, the flexural strength is a key property of the ice. It is also an important factor in the process of the formation of ice ridges which then act as obstacles to marine transit.Publicationtype: Conference Paper50 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Vorhersage des Ermüdungsfestigkeitsverhaltens von Stumpfstößen mittels Explainable Artificial IntelligenceButt joints are one of the most common welded joints in welded structures. It is well known that the fatigue behavior of such joints depends on many factors such as loading conditions, local weld geometry, etc., which often interact with each other; however these are difficult to quantify due to the statistical nature of the variables. Machine leaming algorithms can capture and interpret influential factors and their interactions For this purpose, the SHAPley Additive e Xplanations (SHAP) framework was used to explain the predictions. Additionally. anomaly detection methods were used to improve the robustness of the predictions.Publicationtype: Conference Paper57 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Determination of weld geometry parameter and stress concentration distributions for butt joints using deep neural networksPublicationtype: Conference Paper42 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Analyzing the complexity of ice with explainable machine learning for the development of an ice material model(2022-01); Seasonally or partially ice-covered sea areas are attractive to many stakeholders. Ships and structures operating in these areas must be designed to withstand ice-induced loads. This is done with empirical methods and model-scale ice tests, but these have various drawbacks. Numerical simulations are a desirable remedy, but their accuracy is limited because current ice material models only partially capture ice behavior. Improving ice material models is difficult due to gaps in ice mechanics theory and full-scale measurements. Small-scale experimental data is analyzed with explainable machine learning for an improved understanding of ice. An elastic-brittle material model is developed and applied in several studies.Publicationtype: Doctoral Thesis514 2711 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Prediction of fatigue failure in small-scale butt-welded joints with explainable machine learningButt-welded joints are common in many industries. The fatigue behavior of such joints depends on numerous factors, e.g. load level, local weld geometry, or parent material strength. To make things worse, these factors often interact, yet mutual influence can hardly be quantified by multivariate studies, i.e. varying one factor at a time out of many factors, due to the large number of required tests and the statistical nature of weld geometry. Consequently, fatigue assessment of such joints often deviates significantly between prediction and experimental result. Thus, alternative methods are desirable in order to take various influencing factors into account. To this end, machine learning techniques were used to predict failure locations and number of cycles to failure of fatigue tests performed on small-scale butt-welded joint specimens. In addition to accurate predictions, an understanding of importance and mutual influence of the factors is desired, e.g. a ranking of the most important factors; however, capturing the influence of several possibly interacting factors usually requires complex nonlinear machine learning models. We used gradient boosted trees. Since these are black box models, the SHapley Additive exPlanations (SHAP) framework was used to explain the predictions, i.e. identify influential features and their interactions. Lastly, the model explanations are linked back to domain knowledge.Publicationtype: Journal ArticleScopus© Citations 13 67 148