Machine learning-based defect characterization in anisotropic materials with IR-thermography synthetic data

Daghigh, Vahid; Naraghi, Mohammad

doi:10.1016/j.compscitech.2022.109882

Cited by 14 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Faced with curvilinear surface defects, Ma et al [21] proposed a surface defect detection method based on improved Gabor filters and achieved good detection accuracy. Daghigh et al [22] employed the k-nearest neighbor (k-NN) algorithm to provide a model for predicting the size, thickness, and location of penny-shaped defects in composite laminates. Aiming at the surface defects of industrial materials, Liu et al [23] proposed a Haar-Weibull-variance model for steel surface defect detection in an unsupervised manner.…”

Section: A Machine Learning-based Defect Detection Methodsmentioning

confidence: 99%

PDDD-Net: Defect Detection Network Based on Parallel Attention Mechanism and Dual-Channel Spatial Pyramid Pooling

Sui,

Wang

2023

IEEE Access

View full text Add to dashboard Cite

Owing to the small size of the defect pixel area and poor defect-background contrast issues in industrial images, noise and missed detection can easily occur. Therefore, automated defect detection is both necessary and challenging. To address these issues, with parallel attention mechanism (PAM) and dual-channel spatial pyramid pooling-fast block (DC_SPPF), a novel defect detection network, namely, PDDD-Net, is proposed in this paper. First, to make the detection network emphasize small defect areas better, the PAM block is proposed to be embedded into YOLOv5 to obtain more low-level visual features and improve the detection accuracy of microdefects. Meanwhile, by fusing multi-channel features, the DC_SPPF block is proposed to replace the raw spatial pyramid pooling-fast block to acquire richer discriminative features of the defect areas. Finally, The soft non-maximum suppression (Soft-NMS) module is used to undertake the feature candidate box filtering task in YOLOv5 to reduce missed detection. Two public datasets are adopted for the model evaluation: the Tianchi aluminum profile defects dataset (APDDD) and the power line insulator dataset (CPLID). The experimental results indicate that the proposed PDDD-Net network exhibits remarkable defect detection performance compared with other related detection methods.

show abstract

Section: A Machine Learning-based Defect Detection Methodsmentioning

confidence: 99%

PDDD-Net: Defect Detection Network Based on Parallel Attention Mechanism and Dual-Channel Spatial Pyramid Pooling

Sui,

Wang

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Another alternative, presented in [10,20], proposes the use of two-dimensional median and Gaussian smoothing filters to suppress the defective regions of the material and obtain a smoothed image representing the background; however, the filters require parameters to be defined for their implementation. Most recent works reported in the literature use machine learning [21], deep learning models [22,23], convolutional neuronal networks [24], and actual datasets [25] or synthetic datasets [21] to train and validate the proposed approaches. Regarding the methods mentioned above, the reported methods must establish reference regions that suppose prior sample knowledge, apply smoothing filters and define their parameters, perform high-order polynomial regression, or make a complete or partial analysis of the temporal/frequency evolution of the temperature profile or its characteristics.…”

Section: Introductionmentioning

confidence: 99%

Defect Detection and Depth Estimation in Composite Materials for Pulsed Thermography Images by Nonuniform Heating Correction and Oriented Gradient Information

et al. 2023

View full text Add to dashboard Cite

Pulsed thermography is a nondestructive method commonly used to explore anomalies in composite materials. This paper presents a procedure for the automated detection of defects in thermal images of composite materials obtained with pulsed thermography experiments. The proposed methodology is simple and novel as it is reliable in low-contrast and nonuniform heating conditions and does not require data preprocessing. Nonuniform heating correction and the gradient direction information combined with a local and global segmentation phase are used to analyze carbon fiber-reinforced plastic (CFRP) thermal images with Teflon inserts with different length/depth ratios. Additionally, a comparison between the actual depths and estimated depths of detected defects is performed. The performance of the nonuniform heating correction proposed method is superior to that obtained on the same CFRP sample analyzed with a deep learning algorithm and the background thermal compensation by filtering strategy.

show abstract

A review on recent applications of machine learning in mechanical properties of composites

Liang,

Wei,

Peng

et al. 2024

Polymer Composites

View full text Add to dashboard Cite

Composites are undergoing extensive research and utilization due to their excellent mechanical properties, driven by human needs. Traditionally, the research methods in materials science predominantly rely on empirical theory or experimental trial and error approaches. However, the increased complexity of composite materials results in a greater intricacy in their mechanical behavior. Consequently, the utilization of traditional research methods may not achieve sufficient efficiency. Materials science is rapidly transitioning into a data‐driven era, with machine learning (ML) emerging as a potent tool to expedite materials development and enhance properties prediction. Significant advancements have been achieved in the application of ML to the study of composite mechanics. In this review article, we elucidate various ML methods employed in the construction of constitutive models for isotropic and anisotropic composites, and delve into the research on construction ML models that leverage input data derived from composite processes, structures, and environmental conditions to predict material mechanical properties. Additionally, we summarize recent noteworthy ML applications in composite design and optimization. Finally, possible prospective viewpoints are proposed for future development, with the aim of providing essential scientific guidance for advancing material science and technology through ML.Highlights Machine learning can address complexity in constitutive model of the anisotropic composites. Machine learning predicts mechanical properties of composites well by process and structure. Machine learning enhances efficiency in inverse design to optimize composites. Limitations, challenges, development trends of ML in composites.

show abstract

Machine learning-based defect characterization in anisotropic materials with IR-thermography synthetic data

Cited by 14 publications

References 29 publications

PDDD-Net: Defect Detection Network Based on Parallel Attention Mechanism and Dual-Channel Spatial Pyramid Pooling

PDDD-Net: Defect Detection Network Based on Parallel Attention Mechanism and Dual-Channel Spatial Pyramid Pooling

Defect Detection and Depth Estimation in Composite Materials for Pulsed Thermography Images by Nonuniform Heating Correction and Oriented Gradient Information

A review on recent applications of machine learning in mechanical properties of composites

Contact Info

Product

Resources

About