Coded Thermal Wave Imaging based Defect Detection in Composites using Neural Networks

Abstract

Industry 4.0 focuses on the deployment of artificial intelligence in various fields for automation of variety of industrial applications like aerospace, defence, material manufacturing, etc. Application of these principles to active thermography, facilitates automatic defect detection without human intervention and helps in automation in assessing the integrity and product quality. This paper employs artificial neural network (ANN) based classification post-processing modality for exploring subsurface anomalies with improved resolution and enhanced detectability. A modified bi-phase seven-bit barker coded thermal wave imaging is used to simulate the specimens. Experimentation has been carried over carbon fiber reinforced plastic (CFRP) and glass fiber reinforced plastic (GFRP) specimens using artificially made flat bottom holes of various sizes and depths. A phase based theoretical model also developed for quantitative assessment of depth of the anomaly and experimentally cross verified with a maximum depth error of 3%. Additionally, subsurface anomalies are compared based on probability of detection (POD) and signal to noise ratio (SNR). ANN provides better visualization of defects with 96% probability of detection even for small aspect ratio in contrast to conventional post processing modalities.

Multimodal Spatiotemporal Feature Map for Dynamic Gesture Recognition from Real Time Video Sequences

“…TL has been the mostpopular approach in CNN models in recent years [19]. The 1D convolutional layers have a larger kernel size than the typical 3x3 kernel used in spatio-temporal convolutions, which allows them to capture longer temporal information [20]. Long-Term Temporal Convolutions have shown to improve the performance of action recognition on datasets that involve long-term action [21].…”

The Behaviour of Wave Propagation for Structural System Identification: A Comparative Study

“…Thermal waves have also been used by many researchers for damage assessment. Parvez et al [7] have shown that, as this wave is generated in the system, the temperature rises and this response is further quantified for building relation for damage assessment. The medium must have a minimal amount of friction between its particles as well as elasticity and inertia to produce mechanical waves.…”