Abstract:The analytical solution of the one-dimensional N-layer thermal model (NLM) was successfully employed to generate training data for a machine learning (ML) based procedure for the nondestructive inspection of carbon fiber reinforced composite parts with infrared thermography (IRT). The main objective was to identify a reliable correlation between the experimental data of a pulsed IRT experiment and the NLM prediction, thereby enabling the use of simulated data for ML training. This paper focuses on the initial stages of this process, in more detail on the analytical modelling and experimental data preprocessing, such as normalization, correction of experimental shortcomings and feature selection for machine learning. For pulse phase thermography (PPT) simulated and experimentally derived phase data was compared directly in the frequency domain. Therefore, the features for training and validation of ML were defined from those phase spectra in frequency domain. The suitability of these features for automated and reliable defect depth and/or defect material detection was investigated in both simulated and measured IRT test data. As a basis for feature selection, we used referenced and normalized phase-frequency curves as a function of defect depth. A correlation was identified between the results of the experimental and the simulated feature sets, both qualitatively and quantitatively. To demonstrate the practical applicability of this method, two different, generic ML techniques, multilayer perceptron and random forest regression, were tested as examples. The investigation was performed on plates made of multidirectional carbon fiber reinforced polymer (CFRP) with artificial defects made from three different materials.

Generative Manifold Learning Thermography for Non-Destructive Evaluation of Defects in Composite Materials

Data-Augmented Manifold Learning Thermography for Defect Detection and Evaluation of Polymer Composites

Thermal Deformation Defect Prediction for Layered Printing Using Convolutional Generative Adversarial Network

Enhanced Defect Detection in Carbon Fiber Reinforced Polymer Composites via Generative Kernel Principal Component Thermography

Generative Independent Component Thermography for Improved Defect Detection of Carbon Fiber Composites

Spatial-Neighborhood Manifold Learning for Nondestructive Testing of Defects in Polymer Composites

Deep Autoencoder Thermography for Defect Detection of Carbon Fiber Composites

Convolutional Graph Thermography for Subsurface Defect Detection in Polymer Composites

Generative Deep Learning-Based Thermographic Inspection of Artwork

Multiview Learning for Subsurface Defect Detection in Composite Products: A Challenge on Thermographic Data Analysis

Nondestructive detection and analysis based on data enhanced thermography

Slow Feature Thermography for Non-Destructive Evaluation of Defects in Composite Materials

One-dimensional N-layer thermal modelling as a basis for effective machine learning training data generation for nondestructive testing of composite parts.

Deep Autoencoder for Non-destructive Testing of Defects in Polymer Composites

Independent component thermography for non-destructive testing of defects in polymer composites

Manifold learning and segmentation for ultrasonic inspection of defects in polymer composites

Use of line laser scanning thermography for the defect detection and evaluation of composite material

Spatio-Temporal 3-D Residual Networks for Simultaneous Detection and Depth Estimation of CFRP Subsurface Defects in Lock-In Thermography

Automatic Detection of CFRP Subsurface Defects via Thermal Signals in Long Pulse and Lock-In Thermography

One-dimensional deep convolutional autoencoder active infrared thermography: Enhanced visualization of internal defects in FRP composites

Predictive Models for the Characterization of Internal Defects in Additive Materials from Active Thermography Sequences Supported by Machine Learning Methods