Datei im Netzwerk der BAM verfügbar ("Closed Access")
Filtern
Dokumenttyp
- Vortrag (20)
Referierte Publikation
- nein (20)
Schlagworte
- Concrete (4)
- Stress wave propagation (3)
- Damage (2)
- Impact (2)
- Modellkalibrierung (2)
- Multiscale (2)
- Variational Bayesian (2)
- Bayesian Inferenz (1)
- Bayesian inference (1)
- Bridge Monitoring (1)
Organisationseinheit der BAM
Eingeladener Vortrag
- nein (20) (entfernen)
FAIR (findable, accessible, interoperable and reusable) data usage is one of the main principals that many of the research and funding organizations include in their strategic plans, which means that following the main principals of FAIR data is required in many research projects. The definition of data being FAIR is very general, and when implementing that for a specific application or project or even setting a standardized procedure within a working group, a company or a research community, many challenges arise. In this contribution, an overview about our experience with different methods, tools and procedures is outlined.
We begin with a motivation on potential use cases for the applications of FAIR data with increasing complexity starting from a reproducible research paper over collaborative projects with multiple participants such as Round-Robin tests up to data-based models within standardization codes, applications in machine learning or parameter estimation of physics-based simulation models.
In a second part, different options for structuring the data are discussed. On the one hand, this includes a discussion on how to define actual data structures and in particular metadata schema, and on the other hand, two different systems for storing the data are discussed. The first one is the openBIS system, which is an open-source Lab notebook and PostgreSQL based data management system. A second option are a semantic representations using RDF based ontologies for the domain of interest.
In a third section, requirements for workflow tools to automate data processing are discussed and their integration into reproducible data analysis is presented with an outlook on required information to be stored as metadata in the database.
Finally, the presented procedures are exemplarily demonstrated for the calibration of a temperature dependent constitutive model for additively manufactured mortar. Metadata schemata for a rheological measurement setup are derived and implemented in an openBIS database. After a short review of a potential numerical model predicting the structural build-up behaviour, the automatic workflow to use the stored data for model parameter estimation is demonstrated.
Data provenance - from experimental data to trust worthy simulation models and standards Jörg F. Unger, Annika Robens-Radermacher, Erik Tamsen Bundesanstalt für Materialforschung und -prüfung (BAM). Unter den Eichen 87, 12205 Berlin, Germany FAIR (findable, accessible, interoperable and reusable) data usage is one of the main principals that many of the research and funding organizations include in their strategic plans, which means that following the main principals of FAIR data is required in many research projects. The definition of data being FAIR is very general, and when implementing that for a specific application or project or even setting a standardized procedure within a working group, a company or a research community, many challenges arise. In this contribution, an overview about our experience with different methods, tools and procedures is outlined. We begin with a motivation on potential use cases for the applications of FAIR data with increasing complexity starting from a reproducible research paper over collaborative projects with multiple participants such as Round-Robin tests up to data-based models within standardization codes, applications in machine learning or parameter estimation of physics-based simulation models. In a second part, different options for structuring the data are discussed. On the one hand, this includes a discussion on how to define actual data structures and in particular metadata schema, and on the other hand, two different systems for storing the data are discussed. The first one is the open BIS system, which is an opensource Lab notebook and Postgre SQL based data management system. A second option are a semantic representations using RDF based ontologies for the domain of interest. In a third section, requirements for workflow tools to automate data processing are discussed and their integration into reproducible data analysis is presented with an outlook on required information to be stored as metadata in the database.
Numerical simulators, such as finite element models, have become increasingly capable of predicting the behaviour of structures and components owing to more sophisticated underlying mathematical models and advanced computing power. A common challenge lies, however, in calibrating these models in terms of their unknown/uncertain parameters. When measurements exist, this can be achieved by comparing the model response against measured data. Besides uncertain model parameters, phenomena like damage can give rise to further uncertainties; in particular, quasi-brittle materials, like concrete, experience damage in a heterogeneous manner due to various imperfections, e.g. in geometry and boundary conditions. This hardens an accurate prediction of the damaged behaviour of real structures that comprise such materials.
In this study, which draws from a data-driven approach, we use the force-version of the finite element model updating method (FEMU-F) to incorporate measured displacements into the identification of the damage parameters, in order to cope with heterogeneity. In this method, instead of conducting a forward evaluation of the model and comparing the model response (displacements) against the data, we impose displacements to the model and compare the resulting force residuals with measured reaction forces. To account for uncertainties in the measurement of displacements, we endow this approach with a penalty term, which reflects the discrepancy between measured and imposed displacements, where the latter is assumed as unknown random variables to be identified as well. A Variational Bayesian approach is used as an approximating tool for computing posterior parameters. The underlying damage model considered in this work is a gradient-enhanced damage model.
We first establish the identification procedure through two virtual examples, where synthetic data (displacements) are generated over a certain spatially-dense set of points over the domain. The procedure is then validated on an experimental case-study; namely a 3-point bending experiment with displacement measurements resulting from a digital image correlation (DIC) analysis.
A safe and robust performance is a key criterion when building and maintaining structures and component. Ensuring this criterion at different stages of the lifetime can be supported by applying continuous monitoring concepts. The latter usually can serve multiple purposes, including the determination of material parameters for the design phase, the evaluation of the actual loading/environmental conditions (instead of using conservative estimates that are usually larger) and evaluating or predicting the true performance of the structure (thus decreasing the model bias). In this context, a digital twin of the structure has many benefits. It allows to introduce virtual sensors to “measure” sensor information that is e.g. inaccessible or unmeasureable. In order to efficiently use monitoring techniques in the context of a digital twin, it is important to consider the complete chain of information including the choice of sensors, the data processing and structuring, the modelling assumptions, the numerical simulation and finally the stochastic nature of the model prediction. In this presentation, challenges in this context are discussed with a specific focus on Bayesian model updating of the digital twin, accounting for both parameter updates as well as model bias that results from the limitations of modelling assumption. A bottleneck in this approach is the computational effort related to sampling methods such as Markov chain Monte Carlo methods that require many evaluations of the forward model. An alternative to the expensive computation of the forward model for updating the digital twin is the combination with model reduction techniques such as the Proper General Decomposition [1, 2]. The results are illustrated for several examples and scale, ranging from digitals twin for material tests in the lab over lab scale structural digital twins up to damage identification in field experiments.
A safe and robust performance is a key criterion when building and maintaining structures and components. Ensuring this criterion at different stages of the lifetime can be supported by applying continuous monitoring concepts. The latter usually can serve multiple purposes, including the determination of material parameters for the design phase, the evaluation of the actual loading/environmental conditions (instead of using conservative estimates that are usually larger) and evaluating or predicting the true performance of the structure (thus decreasing the model bias). In this context, a digital twin of the structure has many benefits. In addition, it allows to introduce virtual sensors to “measure” sensor information that is e.g. inaccessible or unmeasureable. In the limit, the remaining useful life of a structure can be interpreted as a property that can be “measured” indirectly via the numerical model in combination with real sensor data. In order to efficiently use monitoring techniques in the context of a digital twin, it is important to consider the complete chain of information including the choice of sensors, the data processing and structuring, the modelling assumptions, the numerical simulation and finally the stochastic nature of the model prediction. In this presentation, challenges in this context are discussed with a specific focus on Bayesian model updating of the digital twin, accounting for both parameter updates as well as model bias that results from the limitations of modelling assumption. A bottleneck in this approach is the computational effort related to sampling methods such as Markov chain Monte Carlo methods that require many evaluations of the forward model. An alternative to the expensive computation of the forward model for updating the digital twin is the combination with model reduction techniques such as the Proper General Decomposition. The results are illustrated for several examples and scales, ranging from digitals twin for material tests in the lab over lab scale structural digital twins up to damage identification in field experiments.
In materials and component research, artificial intelligence methodologies will lead to massive upheavals in the coming years. The processes of material development, material processing, lifetime prediction and material characterization will change significantly. By combining AI methods and new forms of knowledge representation, the data-based management of product life cycles will take on new qualities. To address this emerging field of research Fraunhofer IWM set up the online workshop »AI Methods for Fatigue Behavior Assessment and Component Lifetime Prediction«
Combination of model reduction and adaptive subset simulation for structural reliability problems
(2019)
A safe and robust design is a key criterion when building a structure or a component. Ensuring this criterion can either be performed by fullfilling prescribed safety margins, or by using a full probabilistic approach with a computation of the failure probability. The latter approach is particularly well suited for complex Problems with an interaction of different physical penomena that can be described in a numerical model. The bottleneck in this approach is the computational effort. Sampling methods such as Markov chain Monte Carlo methods are often used to evaluate the system reliability. Due to small failure probabilities (e.g. 10^6) and complex physical models with already and extensive computational effort for a single set of parameters, these methods a prohibitively expensive. The focus of this contribution is to demonstrate the advantages of combining model reduction techniques within the concept a variance reducing adaptive sampling procedures. In the developed method, a modification of the adaptive subset simulation based on Papaioannou et al. 2015 is used and coupled with a limit state function based on Proper Generalized Decomposition (PGD) (Chinesta et al. 2011). In the subset simulation the failure probability is expressed as a product of larger conditional failure probabilities. The intermediate failure events are chosen as a decreasing sequence. Instead of solving each conditional probability with a Markov chain approach, an importance sampling approach is used. It is be shown that the accuracy of the estimation depends mainly on the number of samples in the last sub-problem. For model reduction, the PGD approach is used to solve the structural problem a priori for a given Parameter space (physical space plus all random parameters). The PGD approach results in an approximation of the problem output within a prescribed range of all input Parameters (load factor, material properties, ..). The approximation of the solution by a separated form allows an evaluation of the limit state function within the sampling algorithm with almost no cost. This coupled PGD – adaptive subset Simulation approach is used to estimate the failure probability of examples with different complexity. The convergence, the error propagation as well as the reduction in computational time is discussed.