Search

Deep learning-based sustainable subsurface for anomaly detection in different materials is an objective to improve the reliability of thermographic inspection. This article aims to describe a method that uses Barker-coded thermal wave imaging to identify subsurface anomalies in materials. The novelty of the proposed methodology is to detect smaller defects at a higher depth even on a fully corroded sample of mild steel. Experiments were carried out with different kinds of samples like mild steel and glass fiber reinforced plastic (GFRP). Various commonly used modern post-processing techniques are applied alongside the proposed techniques for detecting subsurface anomalies. Subsurface anomalies visualized using the proposed deep learning method give better visualization and results when compared to that of other approaches. In addition to it, region-based active contour segmentation-based detection is also proposed for the GFRP sample. This study results in a high signal-to-noise ratio (SNR) of value 108 dB; the least error in defect size is nearly 0.01% using full width at half maximum (FWHM), and the aspect ratio is nearly 1 for the proposed convolutional neural network (CNN)-based processing approach. [ABSTRACT FROM AUTHOR]

The capacity of additive manufacturing (AM) to create products with intricate characteristics has produced its rapid expansion in the manufacturing business and widespread interest from a wide range of other sectors. To ensure that parts made for functional use meet specified standards, especially in terms of quality and dependability, quality management of the AM components has drawn a lot of attention from the industries. Unnecessary porosities have a substantial impact on how well the products manufactured via AM operate mechanically. In order to continue to enhance the standard of the AM components by outlier detection, the quality control procedures are thus necessary. In order to efficiently identify outliers in AM data, this work introduces the binary-weighted sparrow search fine-tuned random forest (BSS-RF) method. The binary-weighted sparrow search optimization (BSSO) method is proposed to optimize the RF’s performance in order to increase the detection accuracy. To remove the noisy data, raw data samples are enhanced using the normalization procedure. The important features are extracted using quadratic discriminant analysis (QDA). Finally, the suggested strategy is used to identify abnormal occurrences. The efficacy of the suggested approach is supported by experimental data. The suggested approach can be used in various industrial sectors, including electrical engineering, mechanical engineering, and medical engineering. We intend to create an interactive outlier identification system that is based on the suggested technique for real-time outlier identification and fault forecasting in the future. [ABSTRACT FROM AUTHOR]