Autonomous damage recognition in visual inspection of laminated composite structures using deep learning

Fotouhi, Sakineh; Pashmforoush, Farzad; Bodaghi, Mahdi; Fotouhi, Mohammad

doi:10.1016/j.compstruct.2021.113960

Cited by 71 publications

(19 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…With the latest advances in neurosciences and high-capability computing devices, machine learning (ML) algorithms based on Artificial Neural Networks (ANNs) have rapidly emerged as a promising tool to solve damage/defects identification and localization problem [8,10]. Various techniques, including mode shapes, natural frequencies, strain history and many others [11][12][13][14][15], have been used for damage identification during the years for different application fields. This was achieved by utilizing different supervised or unsupervised machine learning techniques for damage recognition [16][17][18].…”

Section: Introductionmentioning

confidence: 99%

Damage Detection in Flat Panels by Guided Waves Based Artificial Neural Network Trained through Finite Element Method

et al. 2021

View full text Add to dashboard Cite

Artificial Neural Networks (ANNs) have rapidly emerged as a promising tool to solve damage identification and localization problem, according to a Structural Health Monitoring approach. Finite Element (FE) Analysis can be extremely helpful, especially for reducing the laborious experimental campaign costs for the ANN development and training phases. The aim of the present work is to propose a guided wave-based ANN, developed through the use of the Finite Element Method, to determine the position of damages. The paper first addresses the development and assessment of the modeling technique. The FE model accuracy was proven through the comparison of the predicted results with experimental and analytical data. Then, the ANN was developed and trained on an aluminum plate and subsequently verified in a composite plate, as well as under different damage configurations. According to the results herein proposed, the ANN allowed to detect and localize damages with a high level of accuracy in all cases of study.

show abstract

Section: Introductionmentioning

confidence: 99%

Damage Detection in Flat Panels by Guided Waves Based Artificial Neural Network Trained through Finite Element Method

et al. 2021

View full text Add to dashboard Cite

show abstract

“…When comparing SEM micrograph shots of the pure epoxy sample before and after edge trimming ( Figure 8 ), it can be seen that the epoxy fresh cut area manifests much more irregular topography than that captured at its initial state [ 46 ]. The inspections reveal that the surface integrity of the machined pure epoxy edge is affected by the presence of several micro-cracks spread around the tridimensional profiles.…”

Section: Resultsmentioning

confidence: 99%

Thermal Aspects in Edge Trimming of Bio-Filled GFRP: Influence of Fiber Orientation and Silica Sand Filler in Heat Generation

et al. 2022

View full text Add to dashboard Cite

The present work aims to determine the influence of Glass Fiber-Reinforced Polymer (GFRP) laminating configuration in heat generation during the dry edge trimming process. Temperature measurement experiments were conducted on pure epoxy matrix, 15% unidirectional glass fiber reinforced epoxy, and 28% silica sand-filled GFRP specimens through eight type-K thermocouples evenly distributed along the trim plans and connected to a data acquisition system. Infrared thermographic measurements were also conducted to investigate the tool temperature evolution while processing. It was found that perpendicular fiber edge milling induces a sharp increase with peak temperature measurements reaching 119 °C, while machining parallel to fiber leads to a maximum temperature history of 41 °C, which is very close to that obtained from the pure epoxy test. It was also found that the addition of silica sand grains in the GFRP matrix reduces both tool and specimen temperature magnitudes up to 67% for 90° plies and 14% for 0° plies compared to silica sand-free composite initial values. The heat partition was calculated from the measured (electric) and estimated energies for the tool, the workpiece, and chips, respectively. It appears from predictions that the addition of silica sand grains increases the heat conductivity of the GFRP materials (with rates of 20% for 0° fiber orientation and 10% for 90° fiber direction), while it reduces that conducted to the milling tool. Scanning Electron Microscopy (SEM) inspections helped detect the dominating machining defects relative to each GFRP configuration and explained the heat generation and dissipation effects in light of peak temperature measurements.

show abstract

“…Stone and Krishnamurthy 144 proposed a thrust force controller using an ANN to reduce the development of delaminations caused by the entry and removal of a drill bit to an FRP. Machine learning was employed to classify the Visual inspection [85][86][87][88][89][90] Ultrasonic inspection [91][92][93][94] Eddy current [95][96][97][98][99] Radiography 100-105 (e.g., x-ray) Thermography 97,[106][107][108][109][110] Acoustic emission [111][112][113][114][115] Fiber optic sensors [116][117][118][119][120][121] (e.g., fiber Bragg grating) Piezoelectric transducers [122][123][124][125] Laser vibrometry [126][127][128][129][130][131][132] failure methods of composite plates bolted together. 145 A beneficial application of ML is prediction making.…”

Section: Composite Applications With Machine Learningmentioning

confidence: 99%

The intersection of damage evaluation of fiber-reinforced composite materials with machine learning: A review

Nelon

Myers

Hall

2022

Journal of Composite Materials

View full text Add to dashboard Cite

Machine learning (ML) has emerged as a useful predictive tool based on mathematical and statistical relationships for various engineering problems. The pairing of structural health monitoring (SHM) and nondestructive evaluation (NDE) methods with ML algorithms has yielded beneficial results in addressing the damage state of a material or system. Damage state descriptions addressed with ML include detecting a damage mechanism, locating a mechanism, identifying the type of mechanism, assessing the extent of the damage mechanism, and estimating the useful remaining life of a material or system. Damage evaluation research of composite materials has progressed with the increased usage of composite structural elements in the aerospace industry. NDE methods are a viable candidate for pairing with ML algorithms to improve damage state monitoring of composite materials due to the complexity associated with the structure of composites. Fiber-reinforced polymers (FRP), for example, contain at least two constituent materials a fiber and matrix material whose mechanical behavior and interactions contribute to the performance of an FRP. Unlike conventional composite analytical models that require explicit information about the constituents and microstructure of a laminate, an ML algorithm can construct damage evaluation predictions when employing exclusively past operational performance or data from an SHM or NDE method. A researcher determines the type of data selected when applying an ML model for trend analysis, anomaly detection, or prediction making. However, no one specific input feature is required for utilizing an ML model, and examples of possible data features include material properties, physical dimensions, and collected evaluation data. In the present review, applications of ML combined with the damage state evaluation of composite materials, particularly examining FRPs, are discussed to demonstrate the predictive capabilities of ML and its viability for future applications, especially in industrial environments, to minimize costs and improve damage detection rates.

show abstract

Autonomous damage recognition in visual inspection of laminated composite structures using deep learning

Cited by 71 publications

References 64 publications

Damage Detection in Flat Panels by Guided Waves Based Artificial Neural Network Trained through Finite Element Method

Damage Detection in Flat Panels by Guided Waves Based Artificial Neural Network Trained through Finite Element Method

Thermal Aspects in Edge Trimming of Bio-Filled GFRP: Influence of Fiber Orientation and Silica Sand Filler in Heat Generation

The intersection of damage evaluation of fiber-reinforced composite materials with machine learning: A review

Contact Info

Product

Resources

About