< Back to previous page

Project

Model-based active thermography for contactless quality assessment in the agro-food industry

In the agro-food industry, there is an increasing demand for non-destructive quality inspection tools that allow to inspect every individual product. This is the result of an increased focus on food safety and quality over the past decades. Several fast, non-destructive and contactless technologies have been studied for food quality inspection, many of them based on the interaction of electromagnetic radiation with the food product. Nevertheless, in some situations the search for a suitable technique remains. This is for example the case when there is an interest in detecting subsurface cavities or inhomogeneities which are not visible with the naked eye or by using classical machine vision.

It was hypothesized in this PhD research that active thermography has exactly the right properties to fill in this gap in the field of non-destructive food quality inspection. The technique is based on recording and analyzing the surface temperature of an object as a function of time in response to an external thermal excitation. Defect detection is based on the principle that (sub)surface regions with different thermo-physical properties result in differences in this temperature-time output. Active thermography is a versatile technology with respect to the experimental configuration and the data processing techniques. Although active thermography is being studied and applied extensively in several domains (e.g. automotive and aerospace engineering and building applications), only a limited number of studies on the active thermography inspection of agro-food products have been performed. The overall objective of this PhD research was to explore the potential of active thermography for rapid, non-destructive and contactless quality inspection in the agro-food sector. In this study, the focus was on pulsed thermography, because this method allows for high inspection speeds and poses a low risk on thermal damage of the inspected products. Both aspects are crucial for quality inspection in the agro-food chain.

An active thermography experimental set-up was elaborated to be able to perform high quality experiments. It was shown that it is important to minimize reflections and to optimize the spatial uniformity of the applied thermal excitation. A long-pass infrared filter was installed in front of the thermal camera to avoid direct reflections of a part of the excitation energy via the sample surface into the camera’s detector. Moreover, the excitation sources were equipped with frosted glass diffusers to increase the spatial uniformity of the excitation. Next, an industrial application of active thermography for the detection of seal contamination in heat-sealed food packages was studied. The heat of the seal bars was used as the thermal excitation source. The detection performances of six thermal image processing methods were compared. High resolution digital images served as a reference to quantify seal contamination. The lowest detection limit was obtained for the method based on a fit of the cooling profiles, although this limit was only slightly lower than that of the method based on a single frame. In this application, the excitation was predetermined by the industrial context and was found not to be ideal with respect to defect detection. This case highlights the importance of optimizing the excitation and recording parameters to improve defect detection. Although model-based active thermography was studied in other domains, it was not yet explored for the agro-food domain. Therefore, a methodology for the model-based optimization of active thermography experiments was developed next.

Five finite element models of the heat transfer occurring during an active thermography inspection were proposed and experimentally validated. The five models differed with respect to their spatial dimension (2D or 3D), the way in which the excitation was handled (as a boundary heat flux or as a prescribed radiosity from an external heat source) and the way in which the convection of air around the sample was included (through correlation formulas based on the Rayleigh number or through explicitly modeling the non-isothermal flow physics). Although all proposed models were able to capture the general aspects of the temperature-time profiles obtained at sound and defective regions, the model including the non-isothermal flow physics showed the best correspondence to the experimental output. All proposed models were found to be computationally intensive. Therefore, a framework to rapidly and efficiently predict the temperature-time output of a heat transfer model for a wide range of input parameters was developed. The framework combines a Principal Component Analysis (PCA) with the concept of Design and Analysis of Computer Experiments (DACE). It involves setting up a meta-model that interpolates between the output of a limited number of simulations that were selected based on an efficient, space-filling experimental design. This PCA+DACE framework is a generic tool that can be modified with respect to the heat transfer model and the input parameters considered. The PCA+DACE framework was illustrated for a case in which the thermo-physical properties were the inputs and the temperature-time output of sound and defective regions in a PVC sample with subsurface defects was predicted.

Next, the PCA+DACE framework was applied to determine the optimal pulse duration for the pulsed thermography detection of subsurface defects located at different depths in a PVC sample. In this application, the pulse duration and the defect depth were considered as the input parameters of the PCA+DACE approach. The principle of pulsed phase thermography was applied to calculate the phase contrast between defective and sound areas in the PVC sample. The simulation results provide us with insight into the more narrow region of the design-space within which the experimental optimum is situated. The largest phase contrast was obtained for the most shallow (1 mm) subsurface defect excited for the shortest time (0.5 s). This indicates that applications in which surface or shallow subsurface defects are considered are the most interesting for active thermography inspections since only a short thermal excitation is required.

Finally, the PCA+DACE framework was applied to optimize the pulsed thermography settings for a confidential, practical case study in the agro-food sector. In this study, it was shown that the in silico approach combining PCA and DACE is a powerful tool to pinpoint the most promising region for the experimental settings. It thus proves that computer-based simulations can help the practitioner to limit the time-consuming and often expensive physical experiments.

Based on the obtained results, it can be concluded that the developed finite element model combined with the proposed PCA+DACE framework has the potential to optimize inspection applications in which there is an interest in measuring (defects related to) differences in (sub)surface thermo-physical properties of agro-food products. Interesting paths for future research are to further improve the accuracy of the finite element heat transfer model, to derive information on the defects directly from the principal components of the PCA+DACE framework, to tailor the experimental set-up to the needs of specific agro-food inspection applications, to increase the inspection and data processing speed, to develop image processing algorithms for the active thermography inspection of objects in motion and to extend the application scope of the proposed methodological framework within the agro-food domain.

Date:1 Oct 2012 →  7 Jul 2017
Keywords:food quality monitoring, active thermography, pulsed thermography, heat transfer simulations
Disciplines:Other chemical sciences, Nutrition and dietetics, Agricultural animal production, Food sciences and (bio)technology, Analytical chemistry, Macromolecular and materials chemistry, Agriculture, land and farm management, Biotechnology for agriculture, forestry, fisheries and allied sciences, Fisheries sciences
Project type:PhD project