Bayesian Inference for 3D Volumetric Heat Sources Reconstruction from Surfacic IR Imaging

Abstract

The domain of non-destructive testing (NDT) or thermal characterization is currently often done by using contactless methods based on the use of an IR camera to monitor the transient temperature response of a system or sample warmed by using any heat source. Though many techniques use optical excitation (flash lamps, lasers, etc.), some techniques use volumetric sources such as acoustic or induction waves. In this paper, we propose a new inverse processing method, which allows for the estimation of 3D fields of heat sources from surface temperature measurements. This method should be associated with volumetric heat source generation. To validate the method, a volumetric source was generated by the Joule effect in a homogeneous PVC sample using an electrical thin cylindrical wire molded in the material. The inverse processing allows us to retrieve the depth of the wire and its geometrical shape and size. This tool could be a new procedure for retrieving 3D defects on NDT.

Keywords

• bayesian inference
• inverse problem
• tomography
• infrared thermography
• non-destructive testing
• 3d reconstruction