Damage detection in composites using non-destructive testing aided by ANN technique: A review

Abstract

Damages are inevitable in structures and effective damage detection techniques are important for maintaining their health. Many weight-sensitive engineering applications use composite materials, especially fiber-reinforced laminates. Common damages of these materials include delamination, fiber breakage, fiber pull-out, etc. Various non-destructive testing (NDT) techniques are reported in the literature for damage detection in composites, such as ultrasonic testing, vibration-based techniques, acoustic emission technique, optical NDT and imagining techniques. However, due to the complex properties of composite materials, conventional techniques for analyzing NDT data are difficult to implement. In this context, artificial neural network (ANN) technique is a promising alternative for analyzing NDT data for damage detection. In this study, an attempt is made to explore the state-of-the-art of damage detection in composites using NDT aided by ANN. The work discusses the pros and cons of different methods and is expected to help in identifying the appropriate method for damage detection in composites.

Keywords

Composites Non-Destructive Testing (NDT)Artificial Neural Network (ANN)Health monitoring Acoustic Emission technique

Introduction

Structures are prone to damage. A precise definition of damage can be stated as an imperfection, or defect that hampers or impairs engineering structures' healthy functioning. In the context of system analysis, structures can be modelled as input excitations and output measurable signals. With this reference, damages are these external excitations that result in energy flow and modification of system characteristics, and finally acquiring the output signals. Important engineering structures are often subjected to different types and sources of damage. A few simple and common types of damage are chemical attack, overloading, dynamic loads, fire, fatigue, cracking, scaling, explosive spalling, and temperature gradients, which are discussed in detail by Kovler and Chernov.¹ These damages, if overlooked, can cause alteration of mechanical properties, resulting in the structure’s overall failure. At the same time, it will result in significant economic and life losses, will hamper intended serviceability, and cause functional disruption. Early detection of the damage and appropriate corrective measures are required to keep the structure in operational condition throughout its design life.

In this context, it may be mentioned that many engineering applications involve the use of composite materials, which offers additional challenge to the engineering and research community in terms of detection and characterization of damages. A composite material is constructed with the combinations of two or more constituents, with enhanced properties than those of its individual constituents.² The different types of composites as reported in the literature are presented in Figure 1. Some of the special features of composites, that are advantageous over conventional materials, are lower weight to strength ratio, high stiffness, enhanced fatigue life, superior corrosion resistance and wear resistance. As a result, composite materials, especially fibre reinforced composites (FRC), are traditionally in high demand for applications such as aerospace, naval and automobile industry.³ In recent times, the acceptability of such composites has expanded to newer field of applications like civil engineering (e.g., modular buildings) also.⁴

Figure 1.

Classification of composites.

It may be mentioned here that some of the engineering applications as above, require composite materials possessing good moldability and recyclability in addition to being light weight.⁵ Thermoplastic composites are useful for this purpose. It is a ductile material that becomes mouldable at a certain elevated temperature and solidify upon cooling. Hence, thermoplastics can be heat-moulded and reshaped again and again, making them recyclable and environmentally friendly.⁶ Some major applications of thermoplastic composites are found in aerospace industry, e.g., in aircraft wings and fuselage⁷ and civil engineering applications e.g., storage tanks, door and window frames.⁴

Composite structures are often subjected to impact loads, fatigue, chemical attack, fire, and lightning strikes⁸ depending on the relevant operating conditions. As a result, serious damages may occur which significantly reduce the structural integrity and mechanical properties. The common types of damage in composite structures are delamination, fatigue failure, matrix-cracking, fiber pull-out, fiber breakage, fiber debonding, fiber waviness etc.^9,10 Timely detection of such damages is of utmost importance for safety and prolonged service life of the concerned structures.

In general, detection of damage is classified as global damage detection and local damage detection.¹¹ While global damage detection techniques are useful for identifying presence of damages in a structure, local damage detection techniques help in detecting the damage locations more precisely.¹¹ However, it must be ensured that the detection procedures should not contribute to any further deterioration of the existing structures. Non-destructive testing (NDT) techniques are extremely relevant and useful for this purpose. Some of the very commonly available NDT techniques for detection of damages in composites are vibration-based techniques, ultrasonic techniques, acoustic emission (AE) technique, infrared thermography (IRT), and endoscopic NDT. All the above techniques help in obtaining relevant test data from the damaged structure, mostly in the form of electronic signal obtained from sensors and images captured through electronic devices. However, a successful detection of the damages depends on effective processing of these signals/images; this procedure is commonly known as post processing. It may be mentioned here that the quality of the obtained signals/images as well as the methods used for extracting the necessary features of the signals/images play a crucial role in successfully predicting the structural damages. Some of the limitations with reference to signals/images and their processing by conventional methods are mentioned below.

Vibration-based NDTs use vibration signals captured by vibration sensors/accelerometers placed at critical locations.¹² However due to presence of noise in the operational environment, these signals are often corrupted. Accordingly, prediction of the presence/extent of damages is prone to be erroneous especially if the damage is small compared to the size of the structure.¹³ The ultrasonic and acoustic emission-based NDTs depend on the velocity of wave propagation inside the material of the structure and time of arrival of ultrasonic/AE waves at sensors. Wave velocity inside a structure, especially when it is operational, is difficult to obtain. The problem worsens when material properties change from one place to another inside the structure as wave velocity is dependent on the material properties. On the other hand, precisely measuring the time of arrival of waves at sensors is also a challenging task.¹² Image processing-based NDTs involve capturing images of the structure at selected locations. However, noise, low contrast and variation in pixel intensity of the images are some of the factors that tend to hamper the accuracy of damage detection/localization using these methods.^14,15

All the above issues, associated with processing the NDT data using the conventional signal/image processing techniques, are liable to become even more complex when the structure under study is made of composite materials. It is because material properties vary from one place to another in composites. In addition, processing of NDT data using conventional methods is often time consuming and it requires advanced knowledge of signal processing.

In light of the above, an alternative approach for processing the NDT data for detection of damages in composite structures is of great necessity. In this context, ANN is a promising technique in the above direction. Influenced by the biological nervous system, the ANN technique has been developed for solving scientific and engineering problems.¹⁵ Additionally, because of the network’s self-organizing abilities, it can be built directly from experimental data. A typical ANN model consists of input layers, hidden layers and output layer. The processing is done inside the hidden layer(s) which utilizes the assigned weight functions and bias. A representative ANN model is presented diagrammatically in Figure 2.¹⁶

Figure 2.

A typical ANN model.¹⁶

ANN technique is capable of processing large dataset within a very short period of time. Moreover, a developed ANN model can be used by practicing engineers/technicians and no background of signal processing approaches is required for the purpose. Accordingly, the ANN method has a good potential to be used for processing NDT data related to detection of damages in composites. Hence, the present study makes an attempt to find out the state-of-the-art of damage detection in composites using NDT aided by ANN.

Types of damages in composites

The damage mechanism of composite structures is complex, and it is yet a subject of concern for the research community. Composites are widely used for various engineering purposes due to various advantages, like good corrosive resistance, high strength-to-weight ratio, and easier fabrication. Damages in these structures might happen during the manufacturing process or in their service life. Common types of damage in composites are delamination, fiber breakage, and fiber pull-out. Several online health monitoring techniques have been developed from time to time for damage detection in composites.

Manufacturing defects are inevitable and have a great potential of causing damage. These defects can be controlled to not jeopardize the safe performance but can’t be eliminated completely. Voids are defects that drastically degrade the mechanical properties of composite structures. Improper curing during the process of prepreg gives rise to void formation. Some research work has suggested that ineffective moisture curing leads to void formation.¹⁷ The effect of moisture was examined to clarify the cause of voids in composite structures.¹⁸ A further reason for damage during the production process is the presence of foreign substances.

When exposed to service conditions, the most common cause of damage is due to impact loadings. Delamination is an important damage mechanism because of interlaminar stresses due to impact loadings which are very common in fiber-reinforced composites. The delamination may not degrade in-plane tensile behaviour but steadily degrades the compressive behaviour as demonstrated.¹⁹ The root cause is poor interlaminar stiffness. The other associated damage mechanisms because of impact loading or severe dynamic loadings are matrix cracking, fiber pull-out, and fiber breakage. Matrix cracking is an important cause of composite structure failure. In composite systems when matrix cracks, the fiber strength is large enough to support loads, and as a result, a crack develops separating out the fibers while they are intact. The stress generated during matrix cracking is an important parameter to detect the associated damage mechanism.²⁰ Pre-existing defects often cause matrix cracking as reported in the work.²¹ The fracture mechanics approach in conjugation with approximate stress analysis was implied to state a theory of progressive matrix cracking in orthotropic composite laminates.²²

Composite structures are often subjected to temperature variations and high stresses, and this causes tensile failure of the composite resulting in fiber breakage, as reported.²³ Fiber breakage and fiber pull-out induce a loss in stiffness, crack closure, hysteresis loop, and anelastic strains.²⁴

Fiber waviness is a material defect very common in thermoplastic composites which result due to improper fabrication and severely effects the mechanical properties.^25,29 When a thermoplastic composite structure is loaded under compression, fiber waviness can cause micro-buckling failure, also known as the development of kink bands, which ultimately results in an early failure.²⁶ Therefore, NDT technique is extremely important for detecting such damages for the sustainable performance of the structure in the long run. Inspite of the lower strength of thermoplastic matrix, aviation companies are more willing to use them. The main advantages are their almost infinite shelf life and multiple processing of the same material into different elements.²⁷ Defects occurring during its manufacturing process significantly reduce their efficiency. Non-destructive testing is extremely important to evaluate their quality and suitability. Although thermoplastic materials have low volume of defects, but monitoring these defects are also vital. The very common defects are void content, porosity and blister. Ultrasonic testing technique is often used to detect the defects trapped in the material.²⁸ Hassen et al., studied the various defects in thermoplastic composite, namely glass fiber/PP subjected to various foreign object inclusion (FOI).²⁹ The study used two different NDT technique, namely ultrasonic testing, and radiography. The ultrasonic testing detected all the FOI in the thermoplastic composite specimen.

Fatigue, high temperature and lightning strikes cause severe sudden unpredictable damage to composite structures. The growing use of CFRP composites in aircraft industries has sought attention for studies on damage detection due to lightning strikes. The low conductivity of CFRP composites adds to the major disadvantage. Aircraft are regularly exposed to lightning strikes which are of two dissimilar magnitudes, a concentrated catastrophic high current followed by a low voltage direct current that spreads along. This low voltage direct current causes material degradation, as reported.⁸ Aktas et al.³⁰ studied the impact behaviour of glass/epoxy laminated composite plates subjected to elevated temperatures. The effect of carbon nanotubes for protection against lightning strikes was studied by.³¹ Damage due to lightning strikes was reported by.³² The study involved artificially simulating relatively low-magnitude current impulses which were discharged through the composite coupons. Low-velocity impact damages for FRP composites with and without open holes were studied by.^33,34 Post-impact evaluation using x-ray computed tomography results showed severe damage around the impact area and matrix cracking and delamination are the principal associated damage mechanisms. The consequent effect of damage development can be mitigated through effective inspection and monitoring through different NDTS. Alternatively, continuous monitoring can be achieved through in-situ SHM systems. These advanced in-situ monitoring systems are coupled with highly sensitive sensing devices and post-processing methodologies for the purpose of continuous real-time monitoring. Damage detection in composite structures is done using active and passive approaches. Active approaches are those NDTs that include external excitations and then measure the excitations, whereas passive techniques involve no actuators, and instead require sensors to sense hidden damages.³⁵

Experimental techniques to acquire damage related data

For the detection of damage in composite structures, a variety of experimental procedures are available in literature and practice. Each of these procedures generates huge data using sensors and data acquisition systems. Important and relevant features are extracted from these experimental data for further analysis to detect the damage. The frequently used experimental techniques are briefly discussed below.

In this section the NDTs that are been reported in the literature for the purpose of damage detection in composites structures are briefly discussed. The discussion includes the brief description of the typical types of equipment for each of the techniques, brief description of the methodology and a few examples of their applications in the various field of engineering.

Vibration-based NDT

The application of the vibration-based damage detection method was initiated in the late 1970s in the sectors of aerospace and offshore engineering. Since then, it has been rapidly expanding its applications in other fields of engineering. The fundamental of the method is that the physical properties of the structure (e.g., mass, stiffness, and damping) influence the modal properties of the structure (e.g., frequencies, modal shapes, modal damping), which means that the changes in these physical parameters are indicative of the consequent changes in the modal parameters. Thus, tracking these variations is important for the purpose of damage detection. Although the vibration-based damage detection method is an emergent technique but can be classified as a traditional vibration-based damage detection method and a modern vibration-based damage detection method. The traditional type utilizes mechanical properties like modal frequencies, modal damping, modal shapes, or strain energy and is not suitable for online monitoring. However, the modern method uses intelligent approach like ANN, genetic algorithm (GA), and signal processing approach like wavelet transform (WT) for the purpose of damage diagnostic studies. Various review works on vibration-based damage detection are published in the literature.

Doebling et al.³⁶ overviewed the methods for the detection of damages and their location by measuring the changes in the vibration responses. The review paper limited the study to the global vibration-based damage detection techniques, that use the modal properties to infer the changes in the mechanical parameters. The study included both methods that are based on the changes in the measured data and the method that uses a finite element model (FEM) for the purpose of damage detection. All the above reviews are based on damage detection, localization, and severity and did not cover the review on the prediction of the remaining useful life of structures.

Yan et al.³⁷ presented a review of vibration-based structural damage detection based on dynamic characteristic parameters. Various intelligent damage detection methods using modern signal-processing methods and artificial intelligence have been presented.

Wang and Chan³⁸ presented a review on vibration-based damage detection (VBDD) in bridge structures. The present review on VBDD methods is based on the changes in natural frequencies, strain modes, modal strain energy (MSE), ANN and signal processing approach like wavelet transform/9WT), Hilbert spectrum methods, and empirical mode decomposition (EMD).

Moughty and Casas^39,40 reviewed the developments in vibration-based damage detection methods in bridges with a focus on the utilization of advanced computational methods. Das et al.⁴¹ presented a comparative study on the available vibration-based damage detection methods; modal examination, local diagnostic method, non-probabilistic methodology, and time-series methods. The authors concluded that out of all the available methods, the time-series method is more effective for damage detection studies. Kong et al.⁴² reviewed the vibration-based damage detection method including the prediction study for the remaining life of a structure. The framework of the paper consisted of various sections which include, damage detection using response-based methods, selecting damage parameters, and constructing objecting functions, adopting optimization approaches, prediction of remaining useful life, and decision-making for the next actions.

Vibration-based damage detection technique involves the study of the modal parameters, primarily natural frequencies, mode shapes, damping, strain energy, etc. Advanced composites offer several advantages for structures that require the combination of high strength and low weight.

Pardoen⁴³ presented the effect on the natural frequencies due to delamination in composite laminates. A substantial reduction in frequency values only in the higher frequency modes was observed in each laminate after the impact. The conclusion drawn reported that in the area of high shear force, the maximum degradation of frequency is observed. Cawley and Adams⁴⁴ presented a vibration-based study on unidirectional and multi-directional plates and on honeycomb composite structures similar to those used in the aerospace industry. The authors implied a variety of damages like holes, local heating, saw cuts, and impact, and subsequently, the variations of results in the damping and frequency values were noted. It was concluded that changes in dynamic stiffness were easier to measure than changes in damping.

Accelerometers are used for the purpose of condition-based monitoring, which allows the monitoring of vibrations produced in various engineering structures. The common method of acquiring vibration data involves connecting accelerometers to the structures and configuring them to a data acquisition system that can deliver high-resolution data. When mounted on structures, accelerometers accordingly convert mechanical energy to electrical energy. There are mainly three types of accelerometers that are used for vibration tests:

1. Piezoelectric accelerometers

i. Charge mode piezoelectric accelerometers

ii. Voltage mode internal electronic piezoelectric accelerometers

2. Variable capacitance MEMS (micro-electro-mechanical systems) accelerometers

3. Piezoresistive accelerometers

Overall, the vibration-based damage detection technique is an efficient technique for the purpose of structural health monitoring. Additionally, more work should be focused on testing real-life complex structures in real-time, rather than representative laboratory tests.

Based on the above discussions it is found that the conventional methodologies are time-consuming and less cost-effective. The following sections cover the overview of ANN-based damage detection techniques available for composite structures.

Ultrasonic Testing (UT)

Ultrasonic testing (UT) involves the application of ultrasound, which is a too high pitch sound to be detected by the human ear, i.e., frequencies above 18 kHz. UT has a wide range of application in the aerospace and nuclear industries. The method of UT involves the propagation of mechanical low-amplitude stress waves through the structure under examination and then measuring the time of travel and, if any, change in the intensity of these waves for a given distance. Application of UT involves distance gauging, damage detection and measuring parameters (such as elastic moduli and grain size) that are related to material structural characteristics.

The typical UT equipment consists of the following functional units:

1. The pulser or the receiver: Produces high voltage electrical pulses within the material under study.

2. The transducer: With the help of the pulser, the transducer generates high frequency ultrasonic sound energy. In the presence of any flaws, this energy is reflected from the damaged surface. The transducer helps to convert the reflected wave signal into an electrical signal.

3. Display device: The display device displays the converted reflected electrical signal.

In late 1880, the application of UT as an NDT was first made by the discovery of the piezoelectric effect by the brothers Pierre and Jacques Curie. Later Firestone, in the USA, and Sproule in the UK, working independently developed a pulse-echo ultrasonic flaw detector, which is still one of the most frequently used UT methods.⁴⁵ The method involves detecting echoes produced as the result of the reflection of an ultrasonic pulse generated by a transducer, from a discontinuity or a flaw inside the medium.

Ultrasonics is probably the most frequently used technique for health monitoring of composite structures.⁴⁶ Researchers have used UT for detecting damages, such as delamination, impact damage, etc., in composites. In the late 1990s, a study was conducted that examined delamination and matrix cracking in composites caused due to low-energy impacts.⁴⁷ The approach considered different means of pulse-echo techniques, namely time-of-flight, amplitude C-scans and backscattering C-scans. Another promising approach is the ultrasonic harmonic eave method, where wave phase characteristics is applied and is suitable for determining changes in composite materials with irregular structures.⁴⁸ Wooh and Wei⁴⁹ used an ultrasonic pulse-echo scheme for detecting delamination. Scarponi and Briotti⁵⁰ evaluated delamination-based damage on different composite materials using UT. Harizi et al.⁵¹ proposed a method for simultaneous measurement of thickness and density for GFRP plates with UT C-scan mode in the form of maps, which is useful for assessing and evaluating damage in composite structures.

Acoustic Emission (AE) Technique

Acoustic Emission (AE) is a naturally occurring phenomenon within a wide range of structures because of built-up stresses due to cracks and deformations. This phenomenon causes the release of transient elastic stress waves in the ultrasound band. These stress waves are referred to as Acoustic Emissions (AE). Acoustic emissions are natural phenomena.

In the 1930s and 1940s, in the United States of America, Obert and Duvall attempted to predict rock bursts from the sub-audible noises due to microseismic activity.⁵² The first significant work in this field was carried out by Kaiser ⁵³ where he obtained the emissions by loading polycrystalline copper, zinc, copper, lead, steel, and aluminium. Kaiser formulated an important conclusion that the AE is an irreversible process. Once a material is loaded to a certain level, acoustic emission will not be generated on the subsequent reloading at that stress level [Kaiser, 1950]. This is lately formulated as the “Kaiser effect”. However, if sufficient time lapse is given for the material recovery, then AE activity reoccurs due to the application of stress.

Greene et al.⁵⁴ reported the first major successful application of AE technology at Aerojet for the inspection of Polaris missile chambers in the late 1960s. Additionally, they implemented AE technique on pressure vessels, tanks and nuclear plants. Spanner and McElroy⁵⁵ and Shiotani⁵⁶ reviewed the applications of secondary AE activity for damage assessments in infrastructures. In a work published in Sandia report by Rumsey et al.,⁵⁷ they compared AE technique with other NDTs for health monitoring of wind turbine blades.

Acoustic emission signals are of two types: continuous type and burst type.⁵⁸ Leakages and defects arising from frictions results in continuous emissions. Whereas sudden discrete acoustic emissions resulting from cracks, jumps and fracture falls under burst type emissions. Acoustic emission waves are broadly categorized as surface and body waves. Further classifications of surface waves are Rayleigh waves and Love waves, and that of body waves: P-waves and S-waves.

AE technique involves the recording and analysing the AE signals captured from the mounted AE sensors on the surface of the structure. To conduct AE monitoring, following functional units are required:

4. Network of sensors: Capture the indiscernible AE waves of the structure under the supervision and hence convert them to analog signals.

5. Preamplifiers: Magnify the analogue signals.

6. Recording and acquisition section: Acquisitioning the amplified signals and converting from analogue to digital signals.

7. Data processing unit: Analysing the recorded AE signals. The system should be able to distinguish and filter the noises (friction, impact, electromagnetic interferences) from the required signals.

The AE inspection technology allows the detection of continuous emissions or detectable burst-type emissions. The signals are thereby analysed using certain signal parameters extracted from AE waveforms. The most common and widely used signal features are:

8. Amplitude: The peak voltage attained by the AE waveform. It is a vital feature which determines the magnitude of the source event. The measured unit is in decibels (dB).

9. Threshold: The AE signals exceeding the threshold value are the only ones that is to be reported. Measured units are in dB.

10. Duration: The interval of time that passes between the first and last threshold crossings. It can be useful for noise filtering. Measured unit is in microsecond (µsecs).

11. Rise Time: The time elapsed since the peak signal crosses the first threshold. The measuring unit in microseconds. (µsecs).

12. Counts: Within the waveform’s duration, the number of times the waveform crosses the threshold in a rising direction.

13. Energy: The area below the squared waveform within the duration of the waveform. The unit is Attojoule (aJ), (1aI = 10⁻¹⁸ Joule).

The AE technique offers an enhanced and efficient methodology for the purpose of health monitoring of an entire system of structure effectively. Moreover, in complex structures, like composites where the wave speed varies greatly with the direction, errors are more evident. The AE technique, therefore, is a suitable option for the purpose of monitoring these complex structures. On the other hand, one of its major advantages lies in the fact that the technique is independent of the size and shape of the defect to be detected, and hence it is possible to detect even the minute kinks and flaws. To make practical use of this technique and to detect minute displacements, sensors are placed to record the AE signal data which are later stored and post-processed in a data acquisition system. The obtained AE data has to be postprocessed, which evidently is achieved using the ANN, to develop a network that can capture and process complex input information without using complex analytical functions.

In general, the SHM has two broad aspects: diagnosis part and prognosis part. With the help of the AE technique one can perform multiple aspects of the diagnosis part, starting with the first step i.e., damage detection, then the second step i.e., localization^59,61–66 of the damage source, and finally the classification of the damage.⁶⁰ The process of locating the acoustic sources, by recording the signals and analysing them is termed as acoustic source localization. Complex structures made of composites need to be thoroughly monitored not only for impact damages, matrix cracking, delamination and fiber breakage but also for locating these sources. The prognosis part deals with the prediction study of the remaining useful life of a structure,⁶⁶ which too can be achieved using AE data effectively. Inspite of the several advantages that the technique yield, Nesvijski⁶⁷ in his work reported the disadvantages the AE technique for 3D source localization. Focus was laid for analysing the developed mathematical algorithm of 3D acoustic source locations using direct geometric arrays algorithm approach.

The AE sensors which are used to capture the AE waves are very expensive and instead of using multiple numbers of sensors, the use of an optimal number of sensors would be more efficient and efficacious. Lately, multiple works are being conducted on the optimal location of AE sensors⁶⁸ to simplify the entire methodology. It can be further noted that future studies related to the optimal placement of sensors may be of much importance considering the risen economic constraints.

Optical NDT

Optical NDT has gained increased attention in the past few years, due to its non-destructive imaging properties with high intensity and precision. The recent development of optical NDT eases signal multiplexing and resistance to electromagnetic interference. The main types of optical NDT are fiber optics, electronic speckle, infrared thermography, endoscopic, and terahertz technology.

Fiber Optics

Fiber optic NDT uses optical fiber to collect and sense light from the object under test (OUT). This test is resistant to electromagnetic interference and corrosion; hence it can be applied to monitor structures in extreme conditions. The optical fibers are lightweight and can be placed and attached to the surface or within the structure under observation without hampering its working conditions. Multiple types of sensor designs have been reported in the literature, including fiber Bragg grating sensor, IFPI (intrinsic Fabry Perot Interferometer), and EFPI (Extrinsic Fabry Perot Interferometer).⁶⁹ Fiber Bragg gratings have been conveniently used to measure dynamic and static loads on bridge decks, and also for the rehabilitation purpose of composite repairs. Optical NDT has been successfully used in the field of civil engineering over the last 20 years. Tennyson et al.⁷⁰ studied the efficiency of the incorporation of fiber optic sensors to monitor the health of bridges in Canada. Lin et al.⁷¹ studied the effectiveness of fiber Bragg grating (FBG) sensors to monitor the response of highway bridges during the process of their construction and service life. Chan et al.⁷² investigated the feasibility of FBG sensors for the purpose of SHM via monitoring the strain data. These FBG sensors are installed on the Tsing Ma bridge (TMB), Hong Kong, which is the world’s longest (1377m) suspension bridge that carry both traffic and railway loads.

Similarly, enough research work has been carried out on optical fiber NDT for composite materials, where the fibers are either attached on the surface or within to measure the strain variations and damage detection. Thursby et al.⁷³ employed optical fiber polarization sensors, implanted within the structure and attached to the surface, to identify holes using Lamb waves.

Optical fiber NDT is used for the purpose of long-time health monitoring. Real-time parameters like temperature and strain are obtained by these sensors and transferred to the data acquisition system through an optical transmission network. Hence, optical fiber sensing technology is a popular emergent research area for the purpose of damage detection.

Electronic speckle

Laser speckle interferometry is an efficient NDT that is widely used in many industries. The major advantages of this technology include, high sensitivity, higher detection rate, and the surface of illumination need not be uniform. The technology, as reported, accurately measures surface stresses, which infers its efficient application in the field of aerospace, automotive, marine, civil, and high-tech engineering applications.

The two most important technologies in the field of speckle interferometry are ESPI (Electronic Speckle Pattern Interferometry) and ESPSI (Electronic Speckle pattern Shearing Interferometry). ESPSI also known as digital shearography, is a non-destructive optical technique that is used to measure field displacement derivatives. Hung et al.⁷⁴ established a variety of practical speckle shearing interferometry instruments. Huang et al.⁷⁵ proposed a novel impulsive thermal stressing method using a shearographic setup. The proposed technique offered both qualitative and quantitative measurements. Multiple experiments were conducted on samples with cracks and debonds.

In recent years, a combination of speckle pattern interferometry and Digital Image Correlation (DIC), is an emerging research topic. The DIC method measures the grey values of the structure under maintenance before and after strain or deformation. With the help of the iteration method, it thereby calculates the correlation coefficient, the maximum correlation coefficient measures the corresponding displacement and strain.⁷⁶

Digital speckle image measurement has become a useful new technique in the field of modern photo mechanics. It will play an important role in the field of materials’ mechanical properties for the purpose of health monitoring of structures.

Infrared thermography

Infrared thermography (IRT) is a mature discipline in the field of NDT.⁷⁷ It has several advantages, including the real-time and non-contact monitoring of large areas.⁷⁸ Many countries have implemented the IRT technique to detect and characterize internal defects and for the purpose of quality performance monitoring.⁷⁹

For a long time, there has been interest in employing IRT to inspect composite materials. For instance, Katunin et al. conducted research to identify deterioration in composite aircraft components.⁸⁰ Fatigue behaviour of composite using infrared thermographic approach have shown promising results.⁸¹ Porosity, a very common manufacturing defect, often degrades the performance of composite structures. IR thermography has been successfully used to study these underlying defects.⁸² The technique involves of heating one part of the sample with high power laser and obtaining the thermal response on the other part as the heat is transferred through the sample. Samples with high porosity temperature rise showed lesser rise in temperature with longer time to reach the other fact.⁸³ Impact damage is a frequent source of defect that aircraft composite components are often subjected to. Boccardi et al. conducted an experimental study on impacted carbon/epoxy specimens to present the suitability of IRT as an efficient NDT.⁸⁴

IRT is an infrared imaging system that, thanks to the help of external apparatus, converts surface dissimilarities in infrared radiation from the surface into a 2D image.⁸⁵ The thermal surface variations are unveiled in a range of false colors linked to a temperature scale.⁸⁶ The thermal image is the output obtained by an infrared camera, that after image processing takes the name of thermogram.⁸⁷ Figure 3 shows various approaches of IR thermography, which is classified according to the modes, configuration of the experimental procedure, the origin of the source of excitation, and the temperature difference of the constituents.⁸⁸ Active thermography and passive thermography are the two basic classification on the basis of the experimental procedure. Active thermography where an external excitation additionally is needed can be further classified as, PT, VT, SHT and LIT based on the source of the excitation.⁸⁹ Péronnet et al. explored the use of infrared thermography (IRT) to discover flaws. Three infrared thermographic techniques—IRT, Lock-in IRT, and Pulse IRT—were employed in this experiment to examine various composite materials used in the aviation sector.⁹⁰ Furthermore, Montanini used pulse phase and lock-in infrared thermography to measure subsurface flaws in a Plexiglas reference specimen. After post-processing, thermal pictures taken in the frequency domain at various frequencies, it was possible to directly assess the thermal diffusivity of the material.⁹¹ For the NDT&E of aircraft composites, Findeis et al. conducted a comparative results examination between IRT and Electronic Speckle Pattern Interferometry (ESPI).⁹² Another comparison study employing pulsed thermography, lock-in thermography, and vibrothermography was conducted by Castanedo et al. for the inspection of aerospace materials.⁹³ Bendada et al. also used the pulsed thermography approach to assess the composites used in aircraft.⁹⁴ Non uniformity of non-planar object inspection like carbon fiber plastic makes their analysis challenging. Tao et al. in their work evaluated the use of recurrent neural network and ANN in pulsed thermography for an automated inspection of carbon fiber plastic samples.⁹⁵

Figure 3.

Classification of various approaches of infrared thermography.

Rantala et al.⁹⁶ in their study conducted the heating of the carbon fiber composite board sample using modulated ultrasonic excitation and thermography; by analysing the phase of the thermal wave, material information from combined heat and mechanical stimulation is obtained. This method can detect smaller collision and lamination faults than individual methods because it has a stronger signal and a deeper detection depth. A carbon fiber composite board [(45/0/90) s] of 0.125 mm thickness was put to the test in eight distinct orientations. Seven of these influences visual damage. Dattoma et al.⁹⁷ applied the thermographic technique for the purpose of damage detection in a wind-turbine blade made of a sandwich composite structure. The latest developments in IR NDT techniques involve the incorporation of infrared and other available NDT testing methods.

Endoscopic NDT

The endoscopic NDT technique involves optical measurement through a hole to evaluate the internal structural properties of a sealed object. The notable advantage of the technique is that it allows visual inspection, with the flexibility to change the direction of sight for better detection through visual inspection. Since it is an NDT, there is little damage caused to the object under inspection. Schuth et al.⁹⁸ presented a study on the development of a shearographic endoscope for the purpose of inspection of both external and internal surfaces of the object under inspection. Macedo et al.⁹⁹ studied the development of endoscopic shearography for the purpose of damage inspection of composite materials in inner cylindrical surfaces, like pipelines.

The three main approaches of development in endoscopic NDT, are in-situ endoscopic NDT, 3D, and automated. Endoscope automation can be divided into two categories: wired probes with auto-tracking capabilities and remote control via wireless image transmission. The most recent 3D endoscopic display technology was created in the US and is called Real-Depth. It doesn’t need any additional display equipment. In-situ endoscopic NDT has a very high application value since it can save a lot of time and money, particularly in situations like real-time aircraft diagnostics.

Terahertz technology

The electromagnetic waves with frequencies between 0.1 THz and 10 THz are known as terahertz (THz) waves. The methodology involves using the THz radiation to illuminate the object, and then capturing the radiation after its interaction with the object. Detection of internal flaws and damages are assessed by inspecting the changes in the THz signal before and after the interaction with the discontinuities within the object.

The THz technology nowadays has found its application in the field of aerospace engineering. Karpowicz et al.¹⁰⁰ reported the success of the THz technique in damage detection, such as debonding, in the tile adhesive layer in the thermal protection layer of the Orbiter spacecraft. Anatasi and Madaras¹⁰¹ used the THz technique to identify corrosion under painted surfaces.

Although the THz technique has not been studied thoroughly unlike the commonly established NDTs, it has a high application potential. The object, modeling, and analysis of the THz technique will have efficient practical applications for the purpose of damage detection.

Damage detection in composites using ANN

Composites are versatile engineering material having their consistent application in various fields of engineering. Owing to their high stiffness-to-weight ratio, durability, corrosion resistance, and low life cycle maintenance cost, composites are used in the automobile industry, the construction of aircraft structures and turbojet aircraft engines, and both for the purpose of rehabilitation and new construction of civil engineering infrastructure. However, when subjected to overloading, environmental impacts and even due to manufacturing anomalies, composite material undergoes degradation. To improve the serviceability and durability of composite structures, localization and detection of damages are essential. The ANN based damage detection approach is an efficient health monitoring tool.

As already mentioned earlier that ANN is a highly potential approach for the purpose of damage detection in composite structures, the present section discusses existing works in this area as available in the literature. One of the earlier works in damage detection of the composite structure is found in the work of Islam and Craig.¹⁰² The study reported damage detection in composite structures using the ANN approach applying modal analysis, whereby both piezoceramic sensors and actuators were used for exciting and measuring responses respectively. The piezoceramic patches were used as both sensors and actuators to carry out the modal analysis. The NN developed was trained with the frequency data as the input. Although the study concentrated on delamination but can be applied to other damages.

Fiber pull-out, fiber breakage, matrix-crack, and delamination are very common damages in composites consequently degrading structural stiffness and reliability which may lead to catastrophic failures. Muir et al.¹⁰³ in their work studied the damage identification and classification using the acoustic emission technique and then applied the machine learning framework in a SiC/SiC ceramic matrix microcomposites (CMC) subjected to uniaxial tension. Frequency information was the signal feature considered to classify matrix cracking and fiber failures. There are many significant works in the area of damage detection using the ANN methodology. Table 1 represents the major works reported in the literature, whereby the focus is laid on the types of composite materials used and the details of the ANN model(s) in the research works. The following sections are presented with the available literature survey of the common types of sources leading to defects in composite structures.

Table 1.

A representative table on different major type of work that has been found in literature is placed below.

Author(s)	Year	Composite specimen	Type of NDT	Experimentation/Numerical Simulation	Modelling details of artificial neural network
Author(s)	Year	Composite specimen	Type of NDT	Experimentation/Numerical Simulation	Training algorithm	Activation function	Input layer	Hidden layer	Output layer
Ramu & Johnson ¹⁴⁴	1994	Glass-epoxy composite plate	Vibration based	Numerical simulation	Back Propagation Neural network	Sigmoid Function	1 input layer with 11 nodes.	2 hidden layers with 11 nodes in each.	1 output layer
James L. Walker & Eric V.K. Hill ¹⁴⁵	1996	Graphite/epoxy tensile specimen	Graphite/epoxy lam + B9:K19inated plate	ASTM D3039 tensile testing	Back Propagation Neural network	TanH Function	1 input layer with 23 nodes.	2 hidden layers with 11 nodes in each.	1 output layer
Seo & Lee ¹⁴⁶	1999	Carbon Fibre Reinforced Plastic composite laminates	Electrical resistance measurement method	Fatigue tests- servo hydraulic testing system	Backpropagation neural network	Information not available	1 input layer with 1 node.	2 hidden layers with 15 nodes in each.	1 output layer
Dua et al. ¹¹⁴	2001	Cross ply Glass/epoxy composite plate	Vibration based method	Drop weight impact test and Finite element modelling	Backpropagation neural network	LogSig	1 input layer with 503 nodes.	1 hidden layer with 10 nodes.	1 output layer with 3 nodes.
Bhat et al. ¹⁴⁷	2003	Carbon Fibre Reinforced Polymer	AE technique	Fatigue spectrum loading	Hybrid NN- Kohomnen Self Organizing Feature Map (KSOM) AND Multilayer Perceptron (MLP) backpropagation algorithm	Information not available
D. Chakraborty ¹⁴⁸	2004	Graphite/epoxy laminated plate	Vibration based method	3D- finite element analysis- ANSYS 5.4 A	Backpropagation neural network	Sigmoidal	1 input layer with 10 nodes.	1 hidden layer with 9 nodes.	1 output layer with 3 nodes.
Graham et al. ¹¹⁵	2004	Carbon Fibre Reinforced Plastic composite sheet	Optical NDE	SQUID NDE scanning system	Feed Forward Perceptron network- via back propagation algorithm	Sigmoid, Tanh	1 input layer with 100 nodes.	1 hidden layer with 201 nodes.	1 output layer with 12 nodes.
Ativitavas et al. ¹⁴⁹	2005	Glass Fibre Reinforced Polymer T-sections	AE technique	Transverse tension, Longitudinal tension, Short-beam shear, Bond test	Backpropagation and probabilistic neural network (PNN)	Binary sigmoid function	Information not available
Leone et al. ⁶⁶	2006	Glass-epoxy prepeg laminae	Four-point bending- AE technique	Fatigue testing	Back Propagation Neural network	Sigmoid Function	1 input layer with 2 nodes.	1 hidden layer with 4 nodes.	1 output layer
Watkins et al. ¹¹⁶	2007	Glass/epoxy composite plate	Optical NDT- Visual inspection and Shearography	Drop weight Impact test	Backpropagation neural network	Sigmoid	1 input layer with 503 nodes.	1 hidden layer with 10 nodes.	1 output layer with 3 nodes.
Rajendraboopathy et al. ¹⁰⁸	2008	Carbon/epoxy tensile specimens	AE technique	INSTRON 5582 UTM	Back Propagation Neural network	Hyperbolic tanH	1 input layer with 66 nodes.	1 hidden layer with 45 nodes.	1 output layer
Arumugam et al. ¹⁰⁹	2010	Carbon/epoxy laminates	AE technique	Drop weight impact- CEAST Fractovis drop impact tower	Feed-forward backpropagation network	TANSIG	1 input layer with 1 node.	5 hidden layers with 10 nodes in each.	1 output layer
Albuquerque et al. ¹⁵	2010	Carbon/epoxy plate	Image processing- Radiography	Drilling experiments- OKUMA MC-40VA machining center	Backpropagation neural network	Information not available	1 input layer with 3 nodes.	1 hidden layer with 2 nodes.	1 output layer with 3 nodes.
Sudha et al. ¹³³	2011	Drilled Glass fibre reinforced Plastic composite plate	AE technique	Drilling experiments	A multi-layer feed-forward ANN using gradient descent with momentum and adaptive learning rate back-propagation algorithm	Traingdx	1 input layer with 4 nodes.	1 hidden layer with 6 nodes.	1 output layer
Zhang et al. ¹³¹	2013	Carbon Fibre Reinforced Plastic composite laminate	Vibration based method	Numerical simulation- FEM modelling	Backpropagation neural network	Information not available	1 input layer with 7 nodes.	2 hidden layers with 20 nodes in each.	1 output layer with 3 nodes.
Ramasamy & Sampathkumar ¹¹⁰	2014	Woven Glass Fibre Reinforced Polymer (Woven) composite laminate	AE Technique	Drop Impact Test	Back Propagation Neural network- gradient descent algorithm, purlin adaptive learning function- Feed forward backpropagation network	Traingdx	1 input layer with 4 nodes.	2 hidden layer with 20 nodes in each.	1 output layer
Kumar et al. ¹⁵⁰	2017	Glass/epoxy composite laminates	AE Technique	PLB and Three-point bending using 100 kN Tinius Olsen UTM	Radial basis function neural network (RBFNN) and generalized regression neural network (GRNN)	Gaussian basis function	1 input layer with 5 nodes.	1 hidden layer with 23 and 39 nodes for RBFNN and GRNN respectively.	1 output layer
Okafor et al. ¹²⁵	2017	Boron/epoxy composite pactch	AE Technique	Tension-tension fatigue loading	Cascade feed forward back-propagation neural network with Levenberg-Marquardt Training Algorithm	Tan Sigmoid function in input layer, a log-sigmoid function in hidden layer and a linear function in output layer	I input layer with 1 node	1 hidden layer with 4 nodes	1 output layer
Li et al. ¹⁵¹	2020	Glass fiber reinforced plastic specimens	AE technique	UTM- tensile test	Backpropagation neural network	Log-sigmoid and Tan-sigmoid	1 input layer with 32 nodes.	1 hidden layer with 8 nodes.	1 output layer with 2 nodes.
Sathiyamurthy et al. ¹⁵²	2020	Carbon Fiber Reinforced Plastic composite sheet	AE Technique	Instron-5582 Universal Testing Machine	Levenberg-Marquardt backpropagation algorithm	TANSIG	1 input layer with 5 nodes.	2 hidden layers with 50 and 25 nodes respectively.	1 output layer with 2 nodes.
Califano et al. ¹⁵³	2020	Carbon Fiber Reinforced Plastic composite plate	Strain-based Damage detection method	Four-point bending test- INSTRON 5969	Feed Forward Perceptron network- via Levenberg-Marquardt back propagation algorithm	Information not available	1 input layer with 1 node.	1 hidden layer with 1 node.	1 output layer
Oliver et al. ¹⁵⁴	2020	Laminated composite plate	Vibration based method- Modal analysis	FE modelling	Feed Forward Perceptron network- via Levenberg-Marquardt back propagation algorithm	Log-Sigmoid	1 input layer with 6 nodes.	2 hidden layers with 8 and 4 nodes respectively	1 output layer with 2 nodes.
Zenzen et al. ¹⁵⁵	2020	Cross ply Laminated composite	Vibration based method	FEM of MATLAB	Backpropagation neural network	Sigmoid	1 input layer with 1 node.	1 hidden layer with 1 node.	1 output layer
Tran-Ngoc et al. ¹⁵⁶	2021	Laminated composite plate	Vibration based method	FE modelling	Artificial neural network based on a hybrid metaheuristic optimization algorithm	Sigmoid function	1 input layer with 1 node.	1 hidden layer with 20 nodes.	1 output layer
Maurya et al. ¹⁴³	2022	Carbon Fiber Reinforced Plastic composite laminate	Vibration Analysis	Numerical simulation- FEM modelling	Cascade Forward Back Propagation Neural Network	Non-Linear Activation Function	1 input layer with 3 nodes.	1 hidden layer with 9 nodes.	1 output layer with 2 nodes.

Damage due to impact

Fiber-reinforced composites are widely being deployed in the wide field of engineering. One of the major areas of concern is the effects of these structures due to impact loadings. Fatigue life prediction study have been always under viewpoint, but most of the studies were based on mathematical relationship. In absence of a well-defined failure criterion for fatigue life prediction, experiments must be performed thoroughly. Lee et al.¹⁰⁴ in their study trained the ANNs to model the constant-stress fatigue behaviour of fiber composites. It was found that the trained network could provide accurate representation of stress ratio for carbon-fiber composites. Assaf and Kadi¹⁰⁵ studies fatigue failure prediction based on feedforward NN for a glass fiber/epoxy laminate composite under tension-tension and tension-compression loading. Reduction of tensile and compression strength has been noted as a result impact loading. These modes of damage lead to premature catastrophic failures. In the early 1970s and 1980s, Cantwell and Morton¹⁰⁶ and Dorey¹⁰⁷ studied the fracture growth process and the residual strength in composite structures due to high-velocity impact loading. It is necessary to study and find out effective means of quantifying and assessing the resulting damage modes due to the effects of both high and low-velocity impact loadings.

Fibers are the primary load bearers in fiber/matrix composites, structural integrity is affected as soon as these fibers begin to break off. As of now it is yet not clear whether proof load test lowers the actual failure load. Backpropagation was used to develop a neural network to predict the failure load of composite tensile specimen.¹⁰⁸ Fiber-reinforced composites have the potential to reduce the weight of high-performing structures, like those used in aerospace engineering. However, they are often subjected to low energy impact loadings which may cause serious structural failures. These damages are termed as “barely visible impact damage” (BVID) and is a potential threat to sensitive structures like those use in aerospace, naval and nuclear industries. Arumugam et al.¹⁰⁹ used counts, energy, duration, rise time and amplitude as the AE signal parameters to carry out ultimate strength prediction in carbon/epoxy composite specimens from AE data by developing ANN. Ramasamy and Sampathkumar¹¹⁰ studied the effect of drop impact damages which also give rise to BVID in composite materials by using the AE technique. the study used significant AR parameters, root mean square value, signal strength and counts to peak. An ANN model was developed by training the network using AE signal parameters as an input and impact damage tolerance as output. High velocity impact in comparison with low velocity impact results in damages which makes the structures often irreplaceable, whereas low-velocity impacts results in damages which in the long run may cause catastrophic failures. Raut et al.¹¹¹ discussed methods of detect low-velocity impact-based damages in composite structures. The paper dealt with the SHM techniques based upon pattern recognition, signal processing, vibration-based methods, ANN, optimization methods, and inverse problems.

Impact damages are unpredictable in nature unlike damages due to corrosion or fatigue and are a subject of concern specifically for fiber composite structures, which are widely used in the aircraft industry. Liu et al.¹¹² studied the drop-weight impact test of Carbon fiber reinforced Polymer (CFRP) composite, whereby back propagation NN model was developed. The model was used to interpret huge dataset of AE transient waveform signals, which was thereafter used as an impact damage indicator. Sung et al.¹¹³ conducted similar study for impact-based damage assessment using time frequency analysis like Wavelet Transform to measure the changing frequencies of the AE waves and improved neural network concepts. To improve the reliability of the NN-based approach, the author used the Levenberg-Marquardt algorithm. Dua et al.¹¹⁴ further studied for the detection and classification of impact-induced damages on composite plates employing ANN with the backpropagation algorithm. The authors additionally implied finite element analysis (FEA) to simulate the impact-induced strain profile due to the impact on the composite plates. The study was carried out experimentally, whereby strain sensors were placed on the composite structure. The developed ANN was thereby applied to characterize, detect, and localize the damage using the strain data. Graham et al.¹¹⁵ proposed a feed-forward NN- based NDT setup using HTS SQUID gradiometers and double-D excitation coils to evaluate the impact of damaged CFRP composites.

Watkins et al.¹¹⁶ coupled ANN with other damage detection mechanisms to classify impact-induced damages on thirteen glass/epoxy laminated plates, where backpropagation NN is based on the kinetic energy of the impact and uses limited strain signatures for damage classification.

To meet the targeted engineering application, composites are designed using the appropriate manufacturing process and the correct material composition.

Damage due to fatigue

Owing to the superior mechanical properties of composites, they are being widely used in the wide field of engineering. It is imperative to detect and monitor these structures in real time. Fatigue damage in composite structures is a complex phenomenon involving different complicated mechanisms. The fatigue damage can be represented by the changes in the mechanical and material properties of the structure, and reduction in stiffness and strength. From the phenomenological viewpoint, fatigue damage can be monitored, in the global sense, by the degradation of strength and stiffness in composites and thermoplastic nanocomposites.^117,118

The carbon fibers in CFRP laminates are electrical conductors, hence electrical resistance has been used as an indicative parameter for the purpose of damage detection. Lee et al.¹⁰⁴ established a relationship between electrical resistance with fatigue life and stiffness reduction with the help of the NN model.

A study based on ANN for predicting fatigue behaviour of unidirectional glass fiber/epoxy composite laminae with different geometrical orientation and stress ratios have been reported by Kadi and Assaf.¹¹⁹ Strain energy-based fatigue has also been used for fatigue life prediction in the context that the strain energy variation is directly proportional to damage growths. Strain energy when considered as a factor gives a complete context of the stress-strain behaviour due to fatigue failure. Kadi and Assaf¹²⁰ attempted to successfully imply strain energy as the only input parameter to train an ANN for fatigue life prediction for fiberglass/epoxy composite. Vassilopoulos et al.^121,122 conducted an experimental investigation on multidirectional glass Fiber Reinforced Plastic (GFRP) composite for modeling fatigue life based on the developed ANN model under constant amplitude loading patterns. Applying the ANN, the authors developed constant life diagrams (CLDs) which can be useful for the design of structures loaded under variable amplitudes. A noteworthy conclusion was made by the authors i.e., efficient modeling of fatigue life can be achieved from a small portion of the entire collected dataset.

Fatigue is a complex problem for composites and their failure mechanism is still under rigorous study. Mathur et al.¹²³ applied the intelligent health monitoring tool, ANN to study the fatigue life of carbon fiber reinforced composites. A pre-processing program, the damage relativity assessment technique (DRAT) was coupled with ANN for determining the damage location and its extent, Kesavan et al.¹²⁴ Okafor et al., in their work, studied fatigue crack propagation in elliptical boron/epoxy composite subjected to tension-tension fatigue loading. the study show that the AE signals can can be distinguished for different damage type such as matrix cracking, fiber breakage, and shear of the composite patch.¹²⁵ The authors used three backpropagation feed-forward neural network for post processing the output signal data to predict fatigue crack propagation. Fatigue cycle and AE events as the input parameter showed promising results to predict fatigue growth.

The stress ratio was taken as a parameter by Assadi et al.¹²⁶ in their study of fatigue life prediction using different NNs in FRP composites. The study included rigorous training of the NNs on a set of composites, whereas the prediction study was validated using another set of composite samples. The results of the conducted study were within acceptable limits. Fatigue damage results in various classes of damage mechanisms, such as fiber breakage, matrix cracking, and delamination.

Damage due to delamination

Advanced composites are being extensively used in aerospace, naval, civil, and mechanical engineering constructions. They have the advantages of high strength, longer durability, and flexible design. The damage mechanism in these advanced composites is complex and difficult to detect. Delamination is potentially a serious damage mechanism that has been investigated thoroughly over the years. These defects cause structural failures at loads below the design load. A significant reduction of natural frequencies has been observed as a result of delamination growth.^127,128 The reduction of natural frequencies leads to a reduction in the stiffness values which thereafter affects the structural integrity. Delamination behaviour study is vital for composite design and application. Impact loads, imperfect bonding, cracks, fiber debonding, broken fibers, and fatigue loading promotes delamination.

Okafor et al.¹²⁹ studied to predict the delamination size in composite laminates with the help of back propagation neural network models. The NN model was developed and trained based on the data of the modal frequencies. A similar study was carried out by Jiang,¹³⁰ whereby AE signatures were used to detect delamination and thereafter back propagation neural network was used to predict the size of the damage using strain signals for a composite plate. Zhang et al.,¹³¹ in their study presented a quantitative prediction of the delamination-based damage growth in FRP composite structure. Three different inverse algorithms were used to predict the location and extent of delamination: a graphical method, ANN, and surrogate-based optimization. Delamination significantly affects strain signatures within the composite structures. Therefore, the measurement of strain distribution is an important parameter for damage detection. Kesavan al.¹³² utilized strain distribution as a parameter to study damage detection in composites. A program that uses multiple ANNs is developed-the Global Neural network Algorithm for Sequential Processing of Internal Sub Networks (GNAISPIN) which can detect delamination.

In the drilling of composite structures, AET can be used to understand the associated damage mechanisms. Delamination is one of the distinct damages that occurs during the drilling process. Sudha et al.¹³³ concentrated on developing an ANN model to study one of the key damages in composites, delamination using AET (acoustic emission technique). Four AE signal parameters were used to carry out delamination damage prediction, RMS voltage, AE count, peak amplitude and energy and was then used to develop feedforward ANN.

Fotouhi et al., investigated different time-to-failure delamination mechanism in glass/epoxy composite laminates. The specimens were subjected to the double cantilever beam, end notch flexure, and mixed mode bending tests. AE source classification were performed using wavelet packet transform (WPT) and fuzzy clustering method (FCM). Additionally, scanning electron microscopic (SEM) examination was used to examine the damaged mechanisms. It was found that the presented methodologies can be used to enhance the characterization and distinction of damage mechanisms in the actual occurring modes of delamination in composite structures.¹³⁴

Su and Ye^135,136 proposed a lamb-wave based quantitative delamination identification study for CF/EP composite structure by using an Intelligent Signal Processing and Pattern Recognition (ISPPR) package. The approach included a multi-layer ANN supervised by error backpropagation which was trained from the spectrographic dataset obtained from the Lamb wave signals. The later work by the authors consisted of a damage identification scheme using the ANN methodology, where Wavelet Transform was performed to extract the signal characteristics with Signal Processing and Interpretation package (SPIP). A damage parameter database (DPD) was constructed to train the NN. The composite plate was the same as reported in their work in the year 2004. In the year 2005, another work was reported in,¹³⁷ where they introduced the concept of “digital damage fingerprints” which is used to construct a damage parameters database (DPD) for the purpose of offline training of a multilayer feedforward NN for Carbon-fiber Epoxy (CF-EP) composites. Su and Ye,¹³⁸ in their work studied the efficiency of genetic algorithms (GA) and ANN for delamination-based damage detection. Crivelli et al.¹³⁹ used an unsupervised neural network to detect, locate and classify two damage mechanisms, matrix cracking and delamination, in a carbon fiber panel.

Albuquerque et al.^15,140 in their article studied delamination as a result of drilling mechanisms in carbon/epoxy composite structures from radiographic images. To obtain their objective, a novel methodology based on ANN was employed for analysing the radiographic images. Hein and Feklistova^141,142 presented an integrated vibration-based method for delamination detection in a fiber-reinforced composite beam. The methodology used Haar wavelets and the ANN approach. A set of 68 databases was constructed using Haar wavelet using Chen-Hsiao method (CHM) and frequency-based approaches and was thereby used to train the neural network model. Maurya et al.¹⁴³ investigated delamination prediction in CFRP composite structures using both vibration-based NDT and then applied ANN to postprocess the obtained vibration data. To train the neural network, natural frequencies and mode shapes obtained from finite element analysis (FEA) were used.

Summary and conclusion

Studies on structural damage detection are an active area of research among the structural engineering community. The present work discusses the importance of damage detection in composites along with various NDTs used for this purpose where the NDT data are processed through ANN approach.

Composites are important structural components that are being lately used in the wide field of engineering applications. The properties that have aided their wide acceptance are low weight-to-strength ratio, light weight, resistance to corrosion, flexibility, mouldability and higher durability. However, when exposed to working condition, composite structures are often subjected to adverse loadings and harsh environmental conditions. As a result, damages like delamination, fiber breakage, matrix cracking, fiber-debonding, fiber waviness etc occur. Hence, detecting damages early in composite structures is extremely important. Various NDT techniques for this purpose are reviewed. Specific attention is given to studies that use ANN approach for processing NDT test data.

The ANN methodology is particularly preferred because it neither involves computationally challenging mathematical/numerical modelling of composite structures nor mathematically tedious signal processing procedure. Instead, this technique utilizes the experimental/test data, and it is found to be fast and highly efficient tool to identify damages. Accordingly, different NDT techniques aided by ANN are studied under a few broad categories as follows:

• Ultrasonic testing for damage detection has been studied for many years and it is potentially an efficient method that has high penetrating power for the purpose of detection of internal flaws. Compared to other NDTs it has a superior advantage for determining the depth of internal flaws. However, most of the reported work in the literature using ultrasonic testing involves conventional structures and a very few works on composite structures are available. It could be due to the complex geometry, anisotropy and inhomogeneity that is involved with composite structures. Hence, it’s a challenge to the research community to address these issues. In this context, ANN is expected to be very helpful in capturing the effect of variation in geometry and material property inside the structure as the method is based on extensive data collected from all parts of the sample.

• Vibration-based damage detection technique has been studied for the last few decades and many studies, aided by ANN, are also available for composites. Most of these works are based on plate-like structures having laminated/sandwich construction. Investigations are mainly focused on finding out whether damage(s) are present in the structure; however, studies are rare where location and characterization of such damages are involved.

• Optical NDT is a comparatively new technique in comparison to the conventional ones, used for the purpose of damage detection in engineering structures. This method is extremely useful to monitor the inaccessible areas of a structure. Further, it provides a wide range of advantages such as its resistance to electromagnetic interference, non-contact, and real-time results. However, only a limited work has been reported on its application on composite structures. Moreover, the application of ANN on the data obtained by this method is rare.

• AE technique is found to be an effective NDT method for damage detection. Many studies are available for damage detection in composites using AE technique aided by ANN. These works are mainly carried out on laminated FRP plates and sandwich plates including those having honeycomb cores. Many such works have investigated the presence of damages and/or their possible location. Some of the studies have shown that the appropriate analysis of AE signal using ANN helps in differentiating/classifying the damages. However, there is scope for further research in this area so as to identify, localize and characterize damages based on acquired AE signals. Moreover, investigating the above for an entire composite structure, rather than only structural components, is a further challenge.

• It may further be added that most of the reported work has used a supervised training algorithm for ANN. It would be interesting to further carry out research work using an unsupervised training algorithm, which can operate on minimal data obtained from various NDTs.

Future Scope

Based on the findings of the present work, it may be stated that there is adequate scope for further research in the area of damage detection in composites using NDT where the testing data are processed through ANN. Some of the possible areas in this context are mentioned below:

While many work is available on detection and characterisation of damages in composites, there is scope for more rigorous studies on localisation of damages using NDTs of all types. Potential for extensive research exists in large scale engineering structures involving composites like modular buildings and thermoplastic applications like pipes, tanks, etc. Terahertz NDT, optical NDT and endoscopic NDT are comparatively new in the area of damage detection, hence there is a lot of scope for use of these techniques for detection of damages in composite structures through ANN. It may further be added that most of the reported works have used a supervised training algorithm for ANN. It would be interesting to further carry out research work using an unsupervised training algorithm, which can operate on minimal data obtained from various NDTs. This will be particularly useful in cases where many parts of the structure may not be easily accessible for carrying out NDTs. On a final note, it is worthwhile to mention that detection, localisation and characterisation of all the possible damages in an entire structure, rather than in certain components of it, is a challenge to the engineering community. A combination of more than one NDT techniques may be suitable for the above purpose. To process the huge dataset, thus generated, ANN technique would be extremely helpful. This entire exercise, especially when automated, would be tremendously conducive from the view point of maintenance and management of important and strategic structures.

Footnotes

Acknowledgements

The authors are thankful to the Department of Civil Engineering, National Institute of Technology, Durgapur for providing the infrastructure to carry out the present study.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Neetika Saha

Parikshit Roy

Pijush Topdar

References

Kovler

Chernov

. Types of damage in concrete structures. Failure, distress and repair of concrete structures. Amsterdam, Netherlands: Elsevier, 2009, pp. 32–56.

Jones

. Mechanics of composite materials. Boca Raton, FL: CRC Press, 2018.

Offringa

. Thermoplastic composites—rapid processing applications. Composites Part A: Applied Science and Manufacturing 1996; 27(4): 329–336.

Uddin

(ed). Developments in fiber-reinforced polymer (FRP) composites for civil engineering. Amsterdam, Netherlands: Elsevier, 2013.

Vaidya

Chawla

. Processing of fibre reinforced thermoplastic composites. International Materials Reviews 2008; 53(4): 185–218.

Biron

. Thermoplastic composites. Thermoplastics and thermoplastic composites. 3rd ed. London, UK: Elsevier, 2018, pp. 821–882.

Okolie

Latto

Faisal

, et al. Manufacturing defects in thermoplastic composite pipes and their effect on the in-situ performance of thermoplastic composite pipes in oil and gas applications. Appl Compos Mater 2023; 30: 231–306. DOI: 10.1007/s10443-022-10066-9.

Kotikalapudi

Singh

. Mechanical strength degradation of carbon fiber polymer matrix composites exposed to constant low-density direct current. In: ASME International Mechanical Engineering Congress and Exposition. New York: American Society of Mechanical Engineers, 2019.

Talreja

Singh

. Damage in composite materials. In: Damage and Failure of Composite Materials. Cambridge: Cambridge University Press, 2012, pp. 36–56. DOI: 10.1017/CBO9781139016063.004.

10.

Piggott

. The effect of fibre waviness on the mechanical properties of unidirectional fibre composites: a review. Composites Science and Technology 1995; 53(2): 201–205. https://www-sciencedirect-com-443.web.bisu.edu.cn/science/article/pii/S0263822321006280#b0030

11.

Yoon

Heider

Gillespie

Jr , et al. Local damage detection using the two-dimensional gapped smoothing method. Journal of Sound and Vibration 2005; 279(1–2): 119–139.

12.

Palma

Steiger

. Structural health monitoring of timber structures–Review of available methods and case studies. Construction and Building Materials 2020; 248: 118528.

13.

Katunin

. Nondestructive damage assessment of composite structures based on wavelet analysis of modal curvatures: state-of-the-art review and description of wavelet-based damage assessment benchmark. Shock Vib. Dig 2015; 2015: 1–19.

14.

Dura˜ o

Magalha˜ es

Tavares

JMRS

, et al. Analyzing objects in images for estimating the delamination influence on load carrying capacity of composite laminates. Electronic Letters on Computer Vision and Image Analysis 2008; 7(2): 11–21.

15.

De Albuquerque

VHC

Tavares

JMR

Durão

. Evaluation of delamination damage on composite plates using an artificial neural network for the radiographic image analysis. Journal of Composite Materials 2010; 44(9): 1139–1159.

16.

https://in.mathworks.com/help/deeplearning/ref/feedforwardnet.html

17.

Kardos

Duduković

Dave

. Void growth and resin transport during processing of thermosetting—Matrix composites. Epoxy Resins and Composites 1986; IV: 101–123.

18.

Grunenfelder

Nutt

. Void formation in composite prepregs–Effect of dissolved moisture. Composites Science and Technology 2010; 70(16): 2304–2309.

19.

Garg

. Delamination—a damage mode in composite structures. Engineering Fracture Mechanics 1988; 29(5): 557–584.

20.

Marshall

Cox

Evans

. The mechanics of matrix cracking in brittle-matrix fiber composites. Acta Metallurgica 1985; 33(11): 2013–2021.

21.

McCartney

. Mechanics of matrix cracking in brittle-matrix fibre-reinforced composites. Proceedings of the Royal Society of London A Mathematical and Physical Sciences 1987; 409(1837): 329–350.

22.

Tan

Nuismer

. A theory for progressive matrix cracking in composite laminates. Journal of Composite Materials 1989; 23(10): 1029–1047.

23.

Argon

. Fracture of composites. Treatise on Materials Science and Technology 2013; 1: 79–114.

24.

Hild

Burr

Leckie

. Fiber breakage and fiber pull-out of fiber-reinforced ceramic-matrix composites. European Journal of Mechanics-A/Solids 1994; 13(6): 731–749.

25.

Sitohang

RDR

Grouve

WJB

Warnet

, et al. Effect of in-plane fiber waviness defects on the compressive properties of quasi-isotropic thermoplastic composites. Composite Structures 2021; 272: 114166.

26.

Piggott

. The effect of fibre waviness on the mechanical properties of unidirectional fibre composites: a review. Composites Science and Technology 1995; 53(2): 201–205.

27.

Erber

Spitko

. Expanded role for thermoplastic composites. Reinforced Plastics 2014; 58(4): 29–33.

28.

Szymański

. Non-destructive testing of thermoplastic carbon composite structures. Transactions on Aerospace Research 2020; 2020(1): 34–52.

29.

Hassen

Taheri

Vaidya

. Non-destructive investigation of thermoplastic reinforced composites. Composites Part B: Engineering 2016; 97: 244–254.

30.

Aktaş

Karakuzu

İçten

. Thermal impact behavior of glass-epoxy-laminated composite plates. Journal of Thermoplastic Composite Materials 2011; 24(4): 535–553. DOI: 10.1177/0892705710393113.

31.

Zhang

, et al. Lightning strike damage on the composite laminates with carbon nanotube films: Protection effect and damage mechanism. Cancer Cell Int 2019; 19: 342–352.

32.

Gharghabi

Lee

Mazzola

, et al. Development of an experimental setup to analyze carbon/epoxy composite subjected to current impulses. arXiv preprint arXiv:161207204. 2016.

33.

Rubio-González

José-Trujillo

Chávez

, et al. Low velocity impact response of composites and fiber metal laminates with open holes. Journal of Composite Materials 2017; 51(6): 797–810.

34.

Rubio-González

Chávez

José-Trujillo

, et al. Impact behavior of carbon fiber/epoxy composites and fiber metal laminates with open holes. Fibers Polym 2021; 22(3): 772–785.

35.

Staszewski

Mahzan

Traynor

. Health monitoring of aerospace composite structures–Active and passive approach. Composites Science and Technology 2009; 69(11–12): 1678–1685.

36.

Doebling

Farrar

Prime

. A summary review of vibration-based damage identification methods. Shock and Vibration Digest 1998; 30(2): 91–105.

37.

Yan

Cheng

, et al. Development in vibration-based structural damage detection technique. Mechanical Systems and Signal Processing 2007; 21(5): 2198–2211.

38.

Wang

Chan

. Review of vibration-based damage detection and condition assessment of bridge structures using structural health monitoring. In: Proceedings of the Second Infrastructure Theme Postgraduate Conference: Rethinking Sustainable Development-Planning, Infrastructure Engineering, Design and Managing Urban Infrastructure. Queensland: Queensland University of Technology, 2009.

39.

Moughty

Casas

. Vibration based damage detection techniques for small to medium span bridges: A review and case study. In: Proceedings of the 8th European Workshop on Structural Health Monitoring (EWSHM 2016), Bilbao, Spain, 2016.

40.

Moughty

Casas

. Advancements of vibration based damage detection techniques for small to medium span bridges. In: Civil Engineering Research in Ireland 2016, Ireland, 2016.

41.

Das

Saha

Patro

. Vibration-based damage detection techniques used for health monitoring of structures: a review. J Civ Struct Health Monit 2016; 6(3): 477–507.

42.

Kong

Cai

C-S

. The state-of-the-art on framework of vibration-based structural damage identification for decision making. Applied Sciences 2017; 7(5): 497.

43.

Pardoen

. Effect of delamination on the natural frequencies of composite laminates. Journal of Composite Materials 1989; 23(12): 1200–1215.

44.

Cawley

Adams

. A vibration technique for non-destructive testing of fibre composite structures. Journal of Composite Materials 1979; 13(2): 161–175.

45.

Graff

. A history of ultrasonics. Physical Acoustics V15 2012.

46.

Ambu

Aymerich

Ginesu

, et al. Assessment of NDT interferometric techniques for impact damage detection in composite laminates. Composites Science and Technology 2006; 66(2): 199–205.

47.

Aymerich

Meili

. Ultrasonic evaluation of matrix damage in impacted composite laminates. Composites Part B: Engineering 2000; 31(1): 1–6.

48.

Nesvijski

. Phase ultrasonic testing of joints in multilayered composite materials. Journal of Thermoplastic Composite Materials 1999; 12(2): 154–162. DOI: 10.1177/089270579901200206.

49.

Wooh

S-C

Wei

. A high-fidelity ultrasonic pulse-echo scheme for detecting delaminations in composite laminates. Composites Part B: Engineering 1999; 30(5): 433–441.

50.

Scarponi

Briotti

. Ultrasonic technique for the evaluation of delaminations on CFRP, GFRP, KFRP composite materials. Composites Part B: Engineering 2000; 31(3): 237–243.

51.

Harizi

Chaki

Bourse

, et al. Mechanical damage characterization of glass fiber-reinforced polymer laminates by ultrasonic maps. Composites Part B: Engineering 2015; 70: 131–137.

52.

Obert

. Use of subaudible noises for prediction of rock bursts. US Department of the Interior, Bureau of Mines, 1941.

53.

Kaiser

Untersuchungen über das Auftreten von Geräuschen beim Zugversuch (An investigation into the occurence of noises in tensile tests or a study of acoustic phenomena in tensile tests). Ph. D. Thesis (in German). Munich, Germany: Technische Hochschule München. 1950.

54.

Green

Lockman

Steele

. Acoustic verification of structural integrity of polaris chambers. Modern Plastics 1964; 41(11): 137–139.

55.

Spanner

McElroy

. Monitoring structural integrity by acoustic emission: a symposium presented at ft. Lauderdale, FL: American Society for Testing and Materials: ASTM International, 1974.

56.

Shiotani

. Recent advances of ae technology for damage assessment of infrastructures. Journal of Acoustic Emission 2012; 30: 21.

57.

Rumsey

Paquette

White

, et al. Experimental results of structural health monitoring of wind turbine blades. Reston, VA: AIAA. AIAA Paper AIAA-2008-1348. 2008.

58.

Schofield

. Acoustic emission under applied stress: air force materials laboratory. Aeronautical Systems Division, Air Force, 1964.

59.

Tobias

. Acoustic-emission source location in two dimensions by an array of three sensors. Non-destructive Testing 1976; 9(1): 9–12.

60.

Sachse

Sancar

. Acoustic emission source location on plate-like structures using a small array of transducers. Google Patents, 1986.

61.

Kundu

. Acoustic source localization. Ultrasonics 2014; 54(1): 25–38.

62.

Sengupta

Datta

Topdar

. Structural damage localisation by acoustic emission technique: A state of the art review. Lat Am J Solids Struct 2015; 12: 1565–1582.

63.

Selva

Cherrier

Budinger

, et al. Smart monitoring of aeronautical composites plates based on electromechanical impedance measurements and artificial neural networks. Engineering Structures 2013; 56: 794–804.

64.

Sengupta

Topdar

Datta

(eds). On the Applicability of Wavelet Transform in Localising Defect in a Small Plate Using AE Technique: An Experimental Study. In: International Conference on Advances in Structural Mechanics and Applications. New York: Springer, 2023.

65.

Kalafat

Sause

. Acoustic emission source localization by artificial neural networks. Structural Health Monitoring 2015; 14(6): 633–647.

66.

Leone

Caprino

De Iorio

. Interpreting acoustic emission signals by artificial neural networks to predict the residual strength of pre-fatigued GFRP laminates. Composites Science and Technology 2006; 66(2): 233–239.

67.

Nesvijski

. Problems of Acoustic Sources Location in 3D Medium. Journal of Thermoplastic Composite Materials 2005; 18(4): 351–362. DOI: 10.1177/0892705705049562.

68.

Kundu

Datta

Topdar

, et al. Optimal location of acoustic emission sensors for detecting rail damage. In: Proceedings of the Institution of Civil Engineers-Structures and Buildings. ICE. 2022: 1–10.

69.

Ansari

. State-of-the-art in the applications of fiber-optic sensors to cementitious composites. Cement and Concrete Composites 1997; 19(1): 3–19.

70.

Tennyson

Mufti

Rizkalla

, et al. Structural health monitoring of innovative bridges in Canada with fiber optic sensors. Smart Mater Struct 2001; 10(3): 560–573.

71.

Lin

Pan

Kuo

, et al. Online monitoring of highway bridge construction using fiber Bragg grating sensors. Smart Mater Struct 2005; 14(5): 1075–1082.

72.

Chan

Tam

H-Y

, et al. Fiber Bragg grating sensors for structural health monitoring of Tsing Ma bridge: Background and experimental observation. Engineering Structures 2006; 28(5): 648–659.

73.

Sorazu

Dong

Betz

, et al. (eds). Damage detection in structural materials using a polarimetric fiber optic sensor. In: Smart Structures and Materials, Smart Sensor Technology and Measurement Systems; 2003: SPIE, 2003.

74.

Hung

Shang

Yang

. Unified approach for holography and shearography in surface deformation measurement and nondestructive testing. Optical Engineering 2003; 42(5): 1197–1207.

75.

Huang

Liu

, et al. NDT&E using shearography with impulsive thermal stressing and clustering phase extraction. Optics and Lasers in Engineering 2009; 47(7–8): 774–781.

76.

Sutton

Wolters

Peters

, et al. Determination of displacements using an improved digital correlation method. Image and Vision Computing 1983; 1(3): 133–139.

77.

Müller

Netzelmann

Valeske

. Defect shape detection and defect reconstruction in active thermography by means of two-dimensional convolutional neural network as well as spatiotemporal convolutional LSTM network. Quantitative InfraRed Thermography Journal 2022; 19(2): 126–144.

78.

Kim

Park

Jang

, et al. Automated classification of thermal defects in the building envelope using thermal and visible images. Quantitative InfraRed Thermography Journal 2022: 1–17.

79.

Fleuret

Ebrahimi

Castanedo

, et al. On the use of pulsed thermography signal reconstruction based on linear support vector regression for carbon fiber reinforced polymer inspection. Quantitative InfraRed Thermography Journal 2022; 20: 39–61.

80.

Katunin

Dragan

Dziendzikowski

. Damage identification in aircraft composite structures: A case study using various non-destructive testing techniques. Composite Structures 2015; 127: 1–9.

81.

Montesano

Fawaz

Bougherara

. Non-destructive assessment of the fatigue strength and damage progression of satin woven fiber reinforced polymer matrix composites. Composites Part B: Engineering 2015; 71: 122–130.

82.

Alhammad

Avdelidis

Ibarra Castanedo

, et al. Multi-label classification algorithms for composite materials under infrared thermography testing. Quantitative InfraRed Thermography Journal 2022: 1–27.

83.

Connolly

. The measurement of porosity in composite materials using infrared thermography. Journal of Reinforced Plastics and Composites 1992; 11(12): 1367–1375.

84.

Boccardi

Boffa

Carlomagno

, et al. Infrared thermography and ultrasonics to evaluate composite materials for aeronautical applications. Journal of Physics: Conference Series 2015; 658(No. 1): 012007.

85.

Yoon

Park

Cho

. An experimental study on the evaluation of temperature uniformity on the surface of a blackbody using infrared cameras. Quantitative InfraRed Thermography Journal 2022; 19(3): 172–186.

86.

Minkina

. Theoretical basics of radiant heat transfer–practical examples of calculation for the infrared (IR) used in infrared thermography measurements. Quantitative InfraRed Thermography Journal 2021; 18(4): 269–282.

87.

Grys

Dudzik

. Investigation on dual-domain data processing algorithm used in thermal non-destructive evaluation. Quantitative InfraRed Thermography Journal 2022; 19(3): 196–219.

88.

Sfarra

Ibarra-Castanedo

Santulli

, et al. Monitoring of jute/hemp fiber hybrid laminates by nondestructive testing techniques. Science and Engineering of Composite Materials 2016; 23(3): 283–300.

89.

Chung

Kim

. Latest advances in common signal processing of pulsed thermography for enhanced detectability: A review. Sci Rep 2021; 11(24): 12168.

90.

Péronnet

Pastor

Huillery

, et al. Non destructive investigation of defects in composite structures by three infrared thermographic techniques. In: International conference on experimental mechanics ICEM 15, Porto, Portugal, 2012.

91.

Montanini

. Quantitative determination of subsurface defects in a reference specimen made of Plexiglas by means of lock-in and pulse phase infrared thermography. Infrared Phys Technol 2010; 53(5): 363–371.

92.

Findeis

Gryzagoridis

Lombe

. Comparing infrared thermography and ESPI for NDE of aircraft composites. Insight - Non-destructive Test Condition Monit 2010; 52: 244–247.

93.

Ibarra-Castanedo

Genest

Guibert

, et al Inspection of aerospace materials by pulsed thermography, lock-in thermography, and vibrothermography: a comparative study. SPIE 2007; 6541: 321–329.

94.

Bendada

Erchiqui

Lamontagne

. Pulsed thermography in the evaluation of an aircraft composite using 3D thermal quadrupoles and mathematical perturbations. Inverse Problems 2005; 21(3): 857–877.

95.

Tao

Zhang

, et al. Automated Defect Detection in Non-planar Objects Using Deep Learning Algorithms. J Nondestr Eval 2022; 41(1): 14.

96.

Rantala

Busse

. NDT of polymer materials using lock-in thermography with water-coupled ultrasonic excitation. NDT E Int 1998; 31(1): 43–49.

97.

Dattoma

Marcuccio

Pappalettere

, et al. Thermographic investigation of sandwich structure made of composite material. NDT E Int 2001; 34(8): 515–520.

98.

Schuth

Vössing

Yang

. A shearographic endoscope for nondestructive test. Jnl of Holography and Speckle 2004; 1(1): 46–52.

99.

Macedo

Benedet

Fantin

, et al. Inspection of defects of composite materials in inner cylindrical surfaces using endoscopic shearography. Optics and Lasers in Engineering 2018; 104: 100–108.

100.

Karpowicz

Zhong

Zhang

, et al. Compact continuous-wave subterahertz system for inspection applications. Appl Phys Lett 2005; 86(5): 054105.

101.

Anastasi

Madaras

(eds). Terahertz NDE for under paint corrosion detection and evaluation. In: AIP Conference Proceedings. College Park, MD: American Institute of Physics. 2006.

102.

Islam

Craig

. Damage detection in composite structures using piezoelectric materials (and neural net). Smart Mater Struct 1994; 3(3): 318–328.

103.

Muir

Swaminathan

Almansour

, et al. Damage mechanism identification in composites via machine learning and acoustic emission. Npj Comput Mater 2021; 7(1): 95.

104.

Lee

Almond

Harris

. The use of neural networks for the prediction of fatigue lives of composite materials. Composites Part A: Applied Science and Manufacturing 1999; 30(10): 1159–1169.

105.

Al-Assaf

El Kadi

. Fatigue life prediction of unidirectional glass fiber/epoxy composite laminae using neural networks. Composite Structures 2001; 53(1): 65–71.

106.

Cantwell

Morton

. Detection of impact damage in CFRP laminates. Composite Structures 1985; 3(3–4): 241–257.

107.

Dorey

. Fracture behaviour and residual strength of carbon fibre composites subjected to impact loads. Aeronautical Research Council, 1975.

108.

Rajendraboopathy

Sasikumar

Usha

, et al. Artificial neural network a tool for predicting failure strength of composite tensile coupons using acoustic emission technique. Int J Adv Manuf Technol 2009; 44(3): 399–404.

109.

Arumugam

Shankar

Sridhar

, et al. Ultimate strength prediction of carbon/epoxy tensile specimens from acoustic emission data. Journal of Materials Science & Technology 2010; 26(8): 725–729.

110.

Ramasamy

Sampathkumar

. Prediction of impact damage tolerance of drop impacted WGFRP composite by artificial neural network using acoustic emission parameters. Composites Part B: Engineering 2014; 60: 457–462.

111.

Raut

Kolekar

Gombi

. Methods of damage detection on composites under low velocity impact: Review. Materials Today: Proceedings 2020; 27: 2823–2827.

112.

Zhu

Wei

Dykes

, et al. (eds). Impact damage detection in carbon fibre composites using neural networks an acoustic emission. In: Key Engineering Materials. Trans Tech Publ, 1999.

113.

Sung

D-U

J-H

Kim

C-G

, et al. Impact monitoring of smart composite laminates using neural network and wavelet analysis. Journal of Intelligent Material Systems and Structures 2000; 11(3): 180–190.

114.

Watkins

Wunsch

Chandrashekhara

, et al. (eds). Detection and classification of impact-induced damage in composite plates using neural networks. In: IJCNN'01 International Joint Conference on Neural Networks Proceedings (Cat No 01CH37222). IEEE, 2001.

115.

Graham

Maas

Donaldson

, et al. Impact damage detection in carbon fibre composites using HTS SQUIDs and neural networks. NDT E Int 2004; 37(7): 565–570.

116.

Watkins

Akhavan

Dua

, et al. Impact-induced damage characterization of composite plates using neural networks. Smart Mater Struct 2007; 16(2): 515–524.

117.

Yang

Jones

Yang

, et al. A stiffness degradation model for graphite/epoxy laminates. Journal of Composite Materials 1990; 24(7): 753–769.

118.

Shokrieh

Esmkhani

Taheri-Behrooz

. A novel model to predict the fatigue life of thermoplastic nanocomposites. Journal of Thermoplastic Composite Materials 2015; 28(11): 1496–1506. DOI: 10.1177/0892705713513289.

119.

El Kadi

Al-Assaf

. Prediction of the fatigue life of unidirectional glass fiber/epoxy composite laminae using different neural network paradigms. Composite Structures 2002; 55(2): 239–246.

120.

El Kadi

Al-Assaf

. Energy-based fatigue life prediction of fiberglass/epoxy composites using modular neural networks. Composite Structures 2002; 57(1–4): 85–89.

121.

Vassilopoulos

Georgopoulos

Dionysopoulos

. Modelling fatigue life of multidirectional GFRP laminates under constant amplitude loading with artificial neural networks. Advanced Composites Letters 2006; 15(2): 096369350601500.

122.

Vassilopoulos

Georgopoulos

Dionysopoulos

. Artificial neural networks in spectrum fatigue life prediction of composite materials. International Journal of Fatigue 2007; 29(1): 20–29.

123.

Mathur

Gope

, et al. Prediction of fatigue lives of composites material by artificial neural network. Society for Experimental Mechanics, Inc, 2007.

124.

Kesavan

John

Herszberg

. Structural health monitoring of composite structures using artificial intelligence protocols. Journal of Intelligent Material Systems and Structures 2008; 19(1): 63–72.

125.

Okafor

Singh

, et al. Acoustic emission detection and prediction of fatigue crack propagation in composite patch repairs using neural network. Journal of Thermoplastic Composite Materials 2017; 30(1): 3–29. DOI: 10.1177/0892705715573649.

126.

Al-Assadi

El Kadi

Deiab

. Using artificial neural networks to predict the fatigue life of different composite materials including the stress ratio effect. Appl Compos Mater (Dordr) 2011; 18(4): 297–309.

127.

Fotouhi

Najafabadi

. Acoustic emission-based study to characterize the initiation of delamination in composite materials. Journal of Thermoplastic Composite Materials 2016; 29(4): 519–537. DOI: 10.1177/0892705713519811.

128.

Tracy

Pardoen

. Effect of delamination on the flexural stiffness of composite laminates. Thin-Walled Structures 1988; 6(5): 371–383.

129.

Okafor

Chandrashekhara

Jiang

. Delamination prediction in composite beams with built-in piezoelectric devices using modal analysis and neural network. Smart Mater Struct 1996; 5(3): 338–347.

130.

Jiang

. Damage detection in composite structures using multi-sensing, actuation, and neural network systems. University of Missouri-Rolla, 1997.

131.

Zhang

Shankar

Ray

, et al. Vibration-based inverse algorithms for detection of delamination in composites. Composite Structures 2013; 102: 226–236.

132.

Kesavan

John

Herszberg

. Strain-based structural health monitoring of complex composite structures. Structural Health Monitoring 2008; 7(3): 203–213.

133.

Sudha

Sampathkumar

Kumar

. Condition monitoring of delamination during drilling of GFRP composites using the acoustic emission technique a neural network model. Insight-Non-Destructive Testing and Condition Monitoring 2011; 53(8): 445–449.

134.

Fotouhi

Sadeghi

Jalalvand

, et al. Analysis of the damage mechanisms in mixed-mode delamination of laminated composites using acoustic emission data clustering. Journal of Thermoplastic Composite Materials 2017; 30(3): 318–340. DOI: 10.1177/0892705715598362.

135.

. Lamb wave-based quantitative identification of delamination in CF/EP composite structures using artificial neural algorithm. Composite Structures 2004; 66(1–4): 627–637.

136.

. Lamb wave propagation-based damage identification for quasi-isotropic CF/EP composite laminates using artificial neural algorithm: Part I-methodology and database development. Journal of Intelligent Material Systems and Structures 2005; 16(2): 97–111.

137.

. Quantitative damage prediction for composite laminates based on wave propagation and artificial neural networks. Structural Health Monitoring 2005; 4(1): 57–66.

138.

. A fast damage locating approach using digital damage fingerprints extracted from Lamb wave signals. Smart Mater Struct 2005; 14(5): 1047–1054.

139.

Crivelli

Guagliano

Eaton

, et al. Localisation and identification of fatigue matrix cracking and delamination in a carbon fibre panel by acoustic emission. Composites Part B: Engineering 2015; 74: 1–12.

140.

Tavares

JMR

Durão

(eds). Evaluation of delamination damages on composite plates using techniques of image processing and analysis and a backpropagation artificial neural network. EngOpt 2008-International Conference on Engineering Optimization, 2008.

141.

Hein

Feklistova

. Computationally efficient delamination detection in composite beams using Haar wavelets. Mechanical Systems and Signal Processing 2011; 25(6): 2257–2270.

142.

Feklistova

Hein

. Delamination identification using machine learning methods and Haar wavelets. Computer Assisted Methods in Engineering and Science 2017; 19(4): 351–360.

143.

Maurya

Sadarang

Panigrahi

, et al. Detection of delamination in carbon fibre reinforced composite using vibration analysis and artificial neural network. Materials Today: Proceedings 2022; 49: 517–522.

144.

Ramu

Johnson

. Damage assessment of composite structures—a fuzzy logic integrated neural network approach. Computers & Structures 1995; 57(3): 491–502.

145.

Walker

Hill

. Backpropagation neural networks for predicting ultimate strengths of unidirectional graphite/epoxy tensile specimens. Adv Perform Mater 1996; 3: 75–83.

146.

Seo

Lee

. Damage detection of CFRP laminates using electrical resistance measurement and neural network. Composite Structures 1999; 47(1–4): 525–530.

147.

Bhat

Murthy

CRL

. Acoustic emission characterization of failure modes in composites with ANN. Composite Structures 2003; 61(3): 213–220.

148.

Chakraborty

. Artificial neural network based delamination prediction in laminated composites. Materials & Design 2005; 26(1): 1–7.

149.

Ativitavas

Pothisiri

Fowler

. Identification of fiber-reinforced plastic failure mechanisms from acoustic emission data using neural networks. Journal of Composite Materials 2006; 40(3): 193–226.

150.

Suresh Kumar

Arumugam

Sengottuvelusamy

, et al. Failure strength prediction of glass/epoxy composite laminates from acoustic emission parameters using artificial neural network. Applied Acoustics 2017; 115: 32–41.

151.

Shi

, et al. Intelligent recognition of acoustic emission signals from damage of glass fiber-reinforced plastics. Advanced Composites Letters 2020; 29: 2633366X2097468.

152.

Sathiyamurthy

Duraiselvam

Sevvel

. Acoustic emission based deep learning technique to predict adhesive bond strength of laser processed CFRP composites. FME Transactions 2020; 48(3): 611–619.

153.

Califano

Chandarana

Grassia

, et al. Damage detection in composites by artificial neural networks trained by using in situ distributed strains. Appl Compos Mater (Dordr) 2020; 27: 657–671.

154.

Oliver

Ancelotti

Gomes

. Neural network-based damage identification in composite laminated plates using frequency shifts. Neural Comput Appl 2021; 33: 3183–3194.

155.

Zenzen

Khatir

Belaidi

, et al. A modified transmissibility indicator and Artificial Neural Network for damage identification and quantification in laminated composite structures. Composite Structures 2020; 248: 112497.

156.

Tran-Ngoc

Khatir

Ho-Khac

, et al. Efficient Artificial neural networks based on a hybrid metaheuristic optimization algorithm for damage detection in laminated composite structures. Composite Structures 2021; 262: 113339.