Datei im Netzwerk der BAM verfügbar ("Closed Access")
Filtern
Dokumenttyp
Sprache
- Englisch (3)
Schlagworte
- Aktive Thermografie (2)
- Active thermography (1)
- Bayesian Uncertainty Quantification (1)
- CFK (1)
- CFRP (1)
- Carbon fiber reinforced polymer (1)
- Defect shape reconstruction (1)
- Digital Twins (1)
- Finite element simulation (1)
- Gaussian Processes (1)
Organisationseinheit der BAM
Flash thermography is a fast and reliable non-destructive testing method for the investigation of defects in carbon fiber reinforced polymer (CFRP) materials. In this paper numerical simulations of transient thermography data are presented, calculated for a quasi-isotropic flat bottom hole sample. They are compared to experimental data. These simulations are one important step towards the quantitative reconstruction of a flaw by assessing thermographic data. The applied numerical model is based on the finite-element method, extended by a semi-analytical treatment of the boundary of the sample, which is heated by the flash light. A crucial part for a reliable numerical model is the prior determination of the material parameters of the specimen as well as of the experimental parameters of the set-up. The material parameters in plane and in depth diffusivity are measured using laser line excitation. In addition, the absorption and heat transfer process of the first layers is investigated using an IR microscopic lens. The performance of the two distinct components of CFRP during heating – epoxy resin and carbon fibers – is examined. Finally, the material parameters are optimized by variation and comparison of the simulation results to the experimental data. The optimized parameters are compared to the measured ones and further methods to ensure precise material parameter measurements are discussed.
Pulse thermography is a non-destructive testing method based on infrared imaging of transient thermal patterns. Heating the surface of the structure under test for a short period of time generates a non-stationary temperature distribution and thus a thermal contrast between the defect and the sound material. In modern NDT, a quantitative characterization of hidden imperfections in materials is desired. In particular, defect depth and shape are of interest. The reconstruction of the defect from thermography data is a nonlinear inverse problem, and ill-posed. We propose an algorithm for the identification of subsurface defects based on the travel time of the reflected thermal pulse. Our work extends results by Lugin and Netzelmann, taking lateral thermal flows directly into account while retrieving the defect depth. This requires significantly less computational work. Quantitative information about the defect shape and depth is obtained. Application of our method to both thermography data generated by a finite element simulation and experimental heating of PVC test specimens with different defects yields good reconstruction of the actual defects.
In recent years, the use of simulation-based digital twins for monitoring and assessment of complex mechanical systems has greatly expanded. Their potential to increase the information obtained from limited data makes them an invaluable tool for a broad range of real-world applications. Nonetheless, there usually exists a discrepancy between the predicted response and the measurements of the system once built. One of the main contributors to this difference in addition to miscalibrated model parameters is the model error. Quantifying this socalled model bias (as well as proper values for the model parameters) is critical for the reliable performance of digital twins. Model bias identification is ultimately an inverse problem where information from measurements is used to update the original model. Bayesian formulations can tackle this task. Including the model bias as a parameter to be inferred enables the use of a Bayesian framework to obtain a probability distribution that represents the uncertainty between the measurements and the model. Simultaneously, this procedure can be combined with a classic parameter updating scheme to account for the trainable parameters in the original model.
This study evaluates the effectiveness of different model bias identification approaches based on Bayesian inference methods. This includes more classical approaches such as direct parameter estimation using MCMC in a Bayesian setup, as well as more recent proposals such as stat-FEM or orthogonal Gaussian Processes. Their potential use in digital twins, generalization capabilities, and computational cost is extensively analyzed.