Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Identification of Contact Failures in Composite Materials via Bayesian Inference and Reduced Models
Submission Author:
Eiji Watanabe , RJ , Brazil
Co-Authors:
Eiji Watanabe , Luiz A. S. Abreu, Diego Knupp
Presenter: Luiz A. S. Abreu
doi://10.26678/ABCM.COBEM2021.COB2021-1986
Abstract
In several engineering applications, heat conduction in a composite medium has been analyzed, such as thermal insulation, corrosion protection and layered compounds, which offers new opportunities to adapt structures to meet different requirements in modern building materials . In addition, the formulation and solution of problems that allow assessing the adherence between two or more materials is of great importance in several fields, such as electronics, telecommunications, aviation, defense and oil, among others. Given the importance of non-destructive detection of adhesion failures in laminated composites and the use of infrared thermographic images for this purpose, efforts have been made to ensure that the knowledge of heat transfer is applied so that quantitative analyzes can be made possible. It is not always possible to identify adhesion or adhesion failures using only qualitative tests, since temperature gradients in contact failure regions are generally very small. In some situations, the thickness of the material and its glass transition temperatures prevent the occurrence of large gradients and cause the flaws to be identified only by thermal image. In this work, a two-dimensional heat conduction problem is modeled in a single domain in a multilayered medium with isolated side edges, heat flux on the lower surface and heat exchange by natural convection on the upper surface. A reduced model, that represents an approximation of the complete model, is developed using the Improved Lumped Formulation and it involves only the adhesive layer that joins the two layers of the material analyzed in the original problem. This methodology aims to find a low computational cost reduced model and use it to estimate contact failures in multilayered materials. The forward problem solution gives us the temperature profile in one of its surfaces and it will be used in the inverse problem as simulated experimental measures, by adding a noise to the solution. The solution of the inverse problem is an estimated thermal conductivity function with spatial variation. The objective of this work is to detect contact failures in composite materials using a reduced model in the inverse problem to reduce the computational cost. Since the thermal conductivity of the adhesive is considerably grater than the air's, in the case of the presence of a contact failure, the estimated thermal conductivity function must present a significant variation around the failure position. The methodology adopted was able to estimate the contact failure.
Keywords
heat conduction, Laminated composites, contact failure, reduced models, markov chain monte carlo
