Showing papers in "Journal of Nondestructive Evaluation in 2021"

Augmented Ultrasonic Data for Machine Learning

Abstract: Flaw detection in non-destructive testing, especially for complex signals like ultrasonic data, has thus far relied heavily on the expertise and judgement of trained human inspectors. While automated systems have been used for a long time, these have mostly been limited to using simple decision automation, such as signal amplitude threshold. The recent advances in various machine learning algorithms have solved many similarly difficult classification problems, that have previously been considered intractable. For non-destructive testing, encouraging results have already been reported in the open literature, but the use of machine learning is still very limited in NDT applications in the field. Key issue hindering their use, is the limited availability of representative flawed data-sets to be used for training. In the present paper, we develop modern, deep convolutional network to detect flaws from phased-array ultrasonic data. We make extensive use of data augmentation to enhance the initially limited raw data and to aid learning. The data augmentation utilizes virtual flaws—a technique, that has successfully been used in training human inspectors and is soon to be used in nuclear inspection qualification. The results from the machine learning classifier are compared to human performance. We show, that using sophisticated data augmentation, modern deep learning networks can be trained to achieve human-level performance.

NDE 4.0—A Design Thinking Perspective

Abstract: Cyber technologies are offering new horizons for quality control in manufacturing and safety assurance in-service of physical assets. The line between non-destructive evaluation (NDE) and Industry 4.0 is getting blurred since both are sensory data-driven domains. This multidisciplinary approach has led to the emergence of a new capability: NDE 4.0. The NDT community is coming together once again to define the purpose, chart the process, and address the adoption of emerging technologies. In this paper, the authors have taken a design thinking approach to spotlight proper objectives for research on this subject. It begins with qualitative research on twenty different perceptions of stakeholders and misconceptions around the current state of NDE. The interpretation is used to define ten value propositions or use cases under ‘NDE for Industry 4.0’ and ‘Industry 4.0 for NDE’ leading up to the clarity of purpose for NDE 4.0—enhanced safety and economic value for stakeholders. To pursue this worthy cause, the paper delves into some of the top adoption challenges, and proposes a journey of managed innovation, conscious skills development, and a new form of leadership required to succeed in the cyber-physical world.

The Core of the Fourth Revolutions: Industrial Internet of Things, Digital Twin, and Cyber-Physical Loops

Abstract: Like with the previous revolutions the goal of the fourth revolution is to make manufacturing, design, logistics, maintenance, and other related fields faster, more efficient, and more customer centric. This holds for classical industries, for civil engineering, and for NDE and goes along with new business opportunities and models. Core components to enable those fourth revolutions are semantic interoperability, converting data into information, the Industrial Internet of Things (IIoT) offering the possibility for every device, asset, or thing to communicate with each other using standard open interfaces, and the digital twin converting all the available information into knowledge and closing the cyber-physical loop. For NDE this concept can be used #1 to design, improve, and tailor the inspection system hard- and software and #2 to choose and adapt to best inspection solution for the customer or to enhance the inspection performance. Enabling better quality, speed, and cost at the same time. On a broader view, the integration of NDE into IIoT and Digital Twin is the chance for the NDE industry for the overdue change from a cost center to a value center. In most cases, data gathered by NDE is used for a quality assurance assessment resulting in a binary decision. But the information content of NDE goes way deeper and is of major interest for additional groups: engineering and management. Some of those groups might currently not be aware of the benefits of NDE data and the NDE industry makes the access unnecessarily difficult by proprietary interfaces and data formats. Both those challenges need to be taken on now by the NDE industry. The big IT players are not waiting and, if not available on the market, they will develop and offer additional data sources including ultrasonics, X-ray or eddy current.

Automated Flaw Detection in Multi-channel Phased Array Ultrasonic Data Using Machine Learning

Abstract: Modern ultrasonic inspections utilize ever-richer data-sets made possible by phased array equipment. A typical inspection may include tens of channels with different refraction angle, that are acquired at high speed. These rich data sets allow highly reliable and efficient inspection in complex cases, such as dissimilar metal or austenitic stainless steel welds. The rich data sets allow human inspectors to detect cracks with low signal-to-noise ratio from the wider signal patterns. There’s a clear trend in the industry to even richer data sets with full matrix capture (FMC) and related techniques. Convolutional neural networks have recently shown capability to detect flaws with human level accuracy in ultrasonic signals at the B-scan level. To enable automated flaw detection at human-level accuracy for critical applications, these neural networks need be developed to take advantage of today’s rich phased array data-sets. In the present paper, we extend previous work and develop convolutional neural networks that perform highly reliable flaw detection on typical multi-channel phased array data on austenitic welds. The results show, that the modern neural networks can accommodate the rich ultrasonic data and display high flaw detection performance.

Automated Defect Recognition for Welds Using Simulation Assisted TFM Imaging with Artificial Intelligence

Abstract: In this paper, Artificial Intelligence (AI) algorithms are employed for first, automating the process of creating a large synthetic Total Focusing Method (TFM) imaging dataset using a small set of Finite Element (FE) simulation datasets, and second for the automated defect-recognition (ADR) in butt-welds. In this paper, six types of imaging datasets are created with three approaches. In the first approach, weld TFM images are constructed using ultrasonic A-scan signals obtained from Full Matrix Capture (FMC) performed using FE analysis on models with weld defects (porosity and slag). The second approach generates near real-time weld TFM images by implementing fast deep convolution generative adversarial networks (DCGAN). This second technique permits simulations that are several orders faster when compared to the FE method. In the third approach, noise is extracted from FMC-TFM experimental measurements using the sliding kernel approach, and this noise is supplemented to individual simulated datasets for creating near to realistic scenarios. The first dataset is created using the first approach. The second dataset is created using the second approach, and the third hybrid dataset is a combination of FE and DCGAN weld TFM imaging. The fourth dataset is noise supplemented to FE based dataset. The fifth dataset is generated by adding noise to DCGAN images. The sixth hybrid dataset with noise is a combination of FE and DCGAN weld TFM noise images. AI plays a significant role in object detection and classification through robust feature extraction, reducing human intervention. In this work, for automated weld defect recognition, a convolutional neural network (CNN) is trained using six types of simulation-assisted weld TFM imaging datasets, which improves the reliability and efficiency of welds quality assurance. The mAP value is 85% for the ADR model trained using the hybrid weld TFM dataset with noise. The model prediction on classification on the hybrid dataset for porosity is 0.86 F1-score, and for slag is 0.80 F1-score.

Automated Defect Recognition on X-ray Radiographs of Solid Propellant Using Deep Learning Based on Convolutional Neural Networks

Abstract: For defense applications, rapid X-ray inspection of propellant samples is essential for the identification and assessment of defects Automation of this process using artificial intelligence is possible by properly training a neural network model Convolution Neural Networks (CNNs) have recently demonstrated excellent success in both the tasks of image recognition and localisation using an adequate amount of data In real-world, it’s not an easy task to produce the correct amount of experimental data required for the deep neural network to operate In this work, we propose a method for producing synthetic radiographic data that is supported by ray tracing based radiographic simulations for the deep learning algorithms to automatically detect anomaly in X-ray images The simulation results, which are then supplemented by noise extracted from the experimental data, show a good comparison with the measurements This Simulation assisted Automatic Defect Recognition (Sim-ADR) system simultaneously perform defect detection and defect instance segmentation The accuracy of the defect detection system is more than 87% on a testing set included 416 images

Convolutional Neural Networks for Semantic Segmentation as a Tool for Multiclass Face Analysis in Thermal Infrared

Abstract: Convolutional neural networks were used for multiclass segmentation in thermal infrared face analysis. The principle is based on existing image-to-image translation approaches, where each pixel in an image is assigned to a class label. We show that established networks architectures can be trained for the task of multiclass face analysis in thermal infrared. Created class annotations consisted of pixel-accurate locations of different face classes. Subsequently, the trained network can segment an acquired unknown infrared face image into the defined classes. Furthermore, face classification in live image acquisition is shown, in order to be able to display the relative temperature in real-time from the learned areas. This allows a pixel-accurate temperature face analysis e.g. for infection detection like Covid-19. At the same time our approach offers the advantage of concentrating on the relevant areas of the face. Areas of the face irrelevant for the relative temperature calculation or accessories such as glasses, masks and jewelry are not considered. A custom database was created to train the network. The results were quantitatively evaluated with the intersection over union (IoU) metric. The methodology shown can be transferred to similar problems for more quantitative thermography tasks like in materials characterization or quality control in production.

Active Thermography for Non-invasive Inspection of an Artwork on Poplar Panel: Novel Approach Using Principal Component Thermography and Absolute Thermal Contrast

Abstract: The conservation of the works of art represent a topic of global interest. The development of effective tools based on advanced technology for analysing and monitoring their health-state is essential to assuring their preservation. In fact, detecting and preventing the formation of defective areas or assessing for an accurate pre-restoration analysis are the main objectives of non-destructive inspection. Active thermography is a well-known non-invasive imaging technique and reliable tool for providing a fast and low-cost analysis of a work of art. In this study we combine the potential of Principal Component Thermography and of Absolute Thermal Contrast to analyse thermal images acquired in-situ on a poplar panel painting representing an original artwork dating in the end of XVI century. We first optimized the thermal stimulation parameters in the laboratory using special phantom samples. These samples were specially made by reproducing in high fidelity the structural properties and materials of the artwork in order to perform effectively the preliminary tests. Then we moved the equipment in-situ by performing the non-destructive inspection directly on the real artwork. We have developed a specific experimental protocol that combines active thermography with two specific and appropriate image processing modalities that allowed us the effective detection of various types of defects in the painting layer. We report a complete analysis and deep discussion concerning the detection and characterization of the defects. Results show that our diagnostic protocol is a powerful tool in assessing the pre-restoration health-state and suitable for in situ analysis of wood artworks.

Training AI-Based Feature Extraction Algorithms, for Micro CT Images, Using Synthesized Data

Abstract: X-ray computed tomography (CT) is a powerful technique for non-destructive volumetric inspection of objects and is widely used for studying internal structures of a large variety of sample types. The raw data obtained through an X-ray CT practice is a gray-scale 3D array of voxels. This data must undergo a geometric feature extraction process before it can be used for interpretation purposes. Such feature extraction process is conventionally done manually, but with the ever-increasing trend of image data sizes and the interest in identifying more miniature features, automated feature extraction methods are sought. Given the fact that conventional computer-vision-based methods, which attempt to segment images into partitions using techniques such as thresholding, are often only useful for aiding the manual feature extraction process, machine-learning based algorithms are becoming popular to develop fully automated feature extraction processes. Nevertheless, the machine-learning algorithms require a huge pool of labeled data for proper training, which is often unavailable. We propose to address this shortage, through a data synthesis procedure. We will do so by fabricating miniature features, with known geometry, position and orientation on thin silicon wafer layers using a femtosecond laser machining system, followed by stacking these layers to construct a 3D object with internal features, and finally obtaining the X-ray CT image of the resulting 3D object. Given that the exact geometry, position and orientation of the fabricated features are known, the X-ray CT image is inherently labeled and is ready to be used for training the machine learning algorithms for automated feature extraction. Through several examples, we will showcase: (1) the capability of synthesizing features of arbitrary geometries and their corresponding labeled images; and (2) use of the synthesized data for training machine-learning based shape classifiers and features parameter extractors.

Application of Pulsed Thermography and Post-processing Techniques for CFRP Industrial Components

Abstract: Several studies demonstrate the effectiveness of pulsed thermography for detection and visualization of sub-superficial flaws in composites. Continuous improvement of thermal data manipulation makes active thermography an attractive and powerful inspection method for industrial process control and maintenance aims. Therefore, temperature image-processing is the major ongoing challenge in the thermographic research field. However, the particular interest for thermographic inspections is to be more addressed to its simple and relatively fast industrial application; an appropriate image processing tool should be implemented and verified on industrial components, containing manufacturing and in-service defects. In the proposed research, well-established and previously proposed methods were analysed and compared for different defect typology inside three CFRP components. The main goal is not solely focused on establishing the suitable data processing approach, providing detection limits of processed data in terms of damage type, size and distribution. The aim of proposed work is to present detailed examples of thermal imaging methods applied on similar critical defects, evaluating different results among methods in terms of defects mapping capabilities and Tanimoto evaluation criterion, coupled also with the signal-to-noise ratio as assessment of defect detectability.

Cyber-Physical Loops as Drivers of Value Creation in NDE 4.0

Abstract: Across so many industries, non-destructive evaluation has proven its worth time and again through quality and safety assurance of valuable assets. Yet, over time, it became underappreciated in business decisions. In most cases, the data gathered by NDT is used for quality assurance assessments resulting in binary decisions. And we seem to miss out on value of the information content of NDE which goes way deeper and can help other stakeholders: such as engineering, management, inspectors, service providers, and even regulators. Some of those groups might not even be aware of the benefits of NDE data and its digitalization. Unfortunately, the NDE industry typically makes the data access unnecessarily difficult by proprietary interfaces and data formats. Both those challenges need to be addressed now by the NDE industry. The confluence of NDE and Industry 4.0, dubbed as NDE 4.0, provides a unique opportunity for the NDE/NDT Industry to not only readjust the value perception but to gain new customer groups through a broad set of value creation activities across the ecosystem. The integration of NDE into the Cyber-Physical Loop (including IIoT and Digital Twin) is the chance for the NDE industry to now shift the perception from a cost center to a value center. This paper provides an overview of the NDE ecosystem, key value streams, cyber-physical loops that create value, and a number of use cases for various stakeholders in the ecosystem.

Vibration-Based Damage Identification in Steel Girder Bridges Using Artificial Neural Network Under Noisy Conditions

Abstract: The vibration-based damage identification techniques use changes in modal properties of structures to detect damages. However, the results of these methods are not reliable under noise. Therefore, it is essential to clarify which method performs vigorous under noisy conditions. In this study, three damage detection methods, called modal strain energy-based damage index, modal flexibility, and modal curvature, are considered to detect damage with and without the presence of noise. The feasibility of these methods is demonstrated by applying a range of damage scenarios in the validated FE model of the I-40 Bridge. The info of the only first three bending mode shapes of the bridge is used to calculate damage indices. The outcome showed while all three methods were capable of detecting damage in the absence of noise, only the modal flexibility method could locate damages in the presence of noise. Thus, an approach is proposed to eliminate noise and quantify damage magnitude using an artificial neural network (ANN) and modal flexibility method. The modal flexibility damage index of different damage severities was contaminated with various noise levels used as input parameters to train the ANN. Results indicate the adequate performance of the trained ANN in noise-canceling and damage magnitude estimation.

Enhancement of Thickness Measurement in Eddy Current Testing Using a Log–Log Method

Abstract: Eddy current testing for thickness measurement has great advantages, such as non-contact, low cost, and high efficiency. It is reported that there is a linear relationship between the tangent of the phase angle of impedance change and the thickness, termed as the approximate linear method (ALM). However, the accuracy of ALM is not very good, especially when the thickness of a specimen is very thin compared with standard penetration depth. The relationship between tangent of phase angle and thickness is simulated by Dodd-Deed model. The first and second derivatives of tangent of phase angle to thickness is consistent with the power function. Thus, the log–log method (LLM) is obtained by taking logarithm of power fitting equation. And, it is found that the change of excitation frequencies and lift-offs hardly affect the slope and linearity of LLM. The correctness and feasibility of LLM are verified by numerical simulation and experiments.

X-ray CT-Based Defect Evaluation of Continuous CFRP Additive Manufacturing

Abstract: The spread of additive technologies from prototyping to manufacturing has made the development of new products possible, but still needs effective methods in order to allow their characterization. In particular, porosity is considered a crucial aspect of AM products. A prototype system for the deposition of continuous carbon fiber-reinforced polymers with a thermoplastic matrix has been recently developed at Mechanical Engineering Department of Politecnico di Milano. This application is of interest, as it would avoid the expensive development and manufacturing of specific molds. The mechanical performance of the manufactured components depends mainly on porosity and on non-correct adhesion among filaments, even in the case of conventional manufacturing processes. The additive deposition shows even more relevant issues of this kind. Hence the need for a characterization of the process. The conventional approach considers a destructive test to characterize the composite mechanical properties or porosity. The aim of this paper is proposing original approaches to evaluate both porosity and non-correct adhesion by means of X-Ray computed tomography. The method is validated by comparing the porosity with the reference destructive method defined in the ASTM D3171 standard. It is also shown that the amount of defects is correlated to the mechanical properties of the obtained components, thus the approach can be used for a non-destructive evaluation of the manufactured parts.

The Effect of Different Flaw Data to Machine Learning Powered Ultrasonic Inspection

Abstract: Previous research (Li et al., Understanding the disharmony between dropout and batch normalization by variance shift. CoRR abs/1801.05134 (2018). http://arxiv.org/abs/1801.05134 arXiv:1801.05134 ) has shown the plausibility of using a modern deep convolutional neural network to detect flaws from phased-array ultrasonic data. This brings the repeatability and effectiveness of automated systems to complex ultrasonic signal evaluation, previously done exclusively by human inspectors. The major breakthrough was to use virtual flaws to generate ample flaw data for the teaching of the algorithm. This enabled the use of raw ultrasonic scan data for detection and to leverage some of the approaches used in machine learning for image recognition. Unlike traditional image recognition, training data for ultrasonic inspection is scarce. While virtual flaws allow us to broaden the data considerably, original flaws with proper flaw-size distribution are still required. This is of course the same for training human inspectors. The training of human inspectors is usually done with easily manufacturable flaws such as side-drilled holes and EDM notches. While the difference between these easily manufactured artificial flaws and real flaws is obvious, human inspectors still manage to train with them and perform well in real inspection scenarios. In the present work, we use a modern, deep convolutional neural network to detect flaws from phased-array ultrasonic data and compare the results achieved from different training data obtained from various artificial flaws. The model demonstrated good generalization capability toward flaw sizes larger than the original training data, and the effect of the minimum flaw size in the data set affects the $$a_{90/95}$$ value. This work also demonstrates how different artificial flaws, solidification cracks, EDM notch and simple simulated flaws generalize differently.

Inspection of Welding Defect Based on Multi-feature Fusion and a Convolutional Network

Abstract: Robot welding is a basic but indispensable technology for many industries in modern manufacturing. However, many welding parameters affect welding quality. During the real welding process, welding defects are inevitably generated that affect the structural strengths and comprehensive performances of different welding products. Therefore, an accurate welding defect recognition algorithm is necessary for automatic robot welding to assess the effects of defects on structural properties and system maintenance. Much work has been devoted to welding defect recognition. It can be mainly divided into two categories: feature-based and deep learning-based methods. The detection performances of feature-based methods rely on effective image features and strong classifiers. However, faced with weak-textured and weak-contrast welding images, the realization of strong image feature expression still faces a certain challenge. Deep learning-based methods can provide end-to-end detection schemes for welding robots. Nevertheless, an effective deep network model relies on much training data that are not easily collected during real manufacturing. To address the above issues regarding defect detection, a novel welding defect recognition algorithm is proposed based on multi-feature fusion for accurate defect detection based on X-ray images. To improve network training, an effective data augmentation process is proposed to construct the dataset. Combined with transfer learning, the multi-scale features of welding images are acquired for effective feature expression with the pre-trained AlexNet network. On this basis, based on multi-feature fusion, a welding defect recognition algorithm fused to a support vector machine with Dempster–Shafer evidence theory is proposed for multi-scale defect detection. Experiments show that the proposed method achieves a better recognition performance in terms of detecting welding defects than those of other related recognition algorithms.

Nondestructive Testing of Defects in Polymer–Matrix Composite Materials for Marine Applications Using Terahertz Waves

Abstract: Fibreglass components employed in the marine environment are susceptible to moisture ingress, resulting in a degradation of mechanical properties. The nondestructive testing and evaluation of such materials using acoustic methods is possible under certain conditions, but the detection of internal voids and delaminations is masked by the presence of water in such internal flaws. Herein, we present an investigation into the use of terahertz technology to overcome these limitations, for the detection of damage and water ingress in thick woven glass-fibre composites. This investigation is facilitated by terahertz time-domain spectroscopy, coupled with physical raster scanning to realize object-penetrating imaging. Air-filled defects within the fibreglass volume are clearly identified using this technique. A spectroscopic investigation on the alteration of the terahertz-range dielectric properties of the fibreglass material due to water ingress is performed and a small change is measurable, although likely to be obfuscated by the interaction of terahertz-frequency radiation with the internal fibreglass structure. Moisture-ridden laminates do not impede propagation of terahertz radiation, and therefore wetted materials may be inspected for volumetric defects. A simulation of water-filled volumetric defects shows promise for practical application.

Rail Surface Defect Detection and Analysis Using Multi-Channel Eddy Current Method Based Algorithm for Defect Evaluation

Abstract: The railroad rail support trains and contributes to their operation. Internal and surface defects occur on the rail due to various combinations of causes including fatigue loading and cyclic tension and compression among others from the deterioration of the rail along with the temperature differences of seasonal changes. Surface defects such as head check, shelling, and squats start out in the rail head and become internal defects due to poor maintenance, ultimately resulting in rail failure. In order to prevent rail failure, it is important that defects are identified through nondestructive evaluation (NDE) in advance and to carry out maintenance techniques including grinding. NDE methods include MFL, EMAT, and ECT, and among these, the ECT method is a representative method with excellent detection sensitivity that nondestructively inspects metal surfaces such as rails and pipes using an electromagnetic field. Also, since the defect signal is obtained as an electrical signal, the depth, length, and width of defects can be assessed using a defect evaluation algorithm. This study investigated the field applicability and future practical use of the 16 channel eddy current testing equipment and defect evaluation algorithm developed in this study. Therefore, the field applicability of the equipment and defect evaluation algorithm was investigated through the detection of artificial defects with varying size and depth. Afterwards, future practical use was evaluated by inspection of areas of rail that are in use and with naturally occurring surface defects and analysis of their size (length, width), depth, and phenomena.

Noise Reduction for Improvement of Ultrasonic Monitoring Using Coda Wave Interferometry on a Real Bridge

Abstract: Reinforced concrete bridges are iconic parts of modern infrastructure. They are designed for a minimum service life of 100 years. However, environmental factors and/or inappropriate use might cause overload and accelerate the deterioration of bridges. In extreme cases, bridges could collapse when necessary maintenance lacks. Thus, the permanent monitoring for structure health assessment has been proposed, which is the aim of structural health monitoring (SHM). Studies in laboratories have shown that ultrasonic (US) coda wave interferometry (CWI) using diffuse waves has high sensitivity and reliability to detect subtle changes in concrete structures. The creation of micro-cracks might be recognized at an early stage. Moreover, large-volume structures can be monitored with a relatively small number of US transducers. However, it is still a challenge to implement the CWI method in real SHM practical applications in an outdoor environment because of the complex external factors, such as various noise sources that interfere with the recorded signals. In this paper, monitoring data from a 36-m long bridge girder in Gliwice, Poland, instrumented with embedded US transducers, thermistors, and vibrating wire strain gauges, is presented. Noise estimation and reduction methods are discussed, and the influence of traffic, as well as temperature variation, are studied. As a result, the relative velocity variation of US waves following the temperature change with a very high precision of $$10^{-4} \%$$ is shown, and a good bridge health condition is inferred. The influence of lightweight real traffic is negligible. The study verified the feasibility of the implementation of the CWI method on real bridge structures.

The Detection and Characterization of Defects in Metal/Non-metal Sandwich Structures by Thermal NDT, and a Comparison of Areal Heating and Scanned Linear Heating by Optical and Inductive Methods

Abstract: It is common on space vehicles to have thermal insulation adhesively bonded to a metal structure. A typical defect in such structures is an interlayer disbond, which may occur either between the insulation and the metal substructure or between the layers of multilayer thermal insulation. One-sided thermal nondestructive testing (TNDT) using surface optical heating, such as Xenon flash or quartz tube, may detect disbonds if the thermal insulation thickness does not exceed a few millimeters and disbonds are not very small. In thicker insulation, the effectiveness of the inspection can be improved by using electrical induction to heat the metal base. In both cases, thermal excitation can be areal heating, which is heat projected over an area by a stationary heat source, or scanned linear heating (SLH), which is a linear heater scanned across the test subject. In the latter, either the linear heater is moved across a stationary test subject, or the linear heater is stationary and the test subject is moved. The SLH method usually provides a higher inspection rate (inspected area unit time). In this research, both the theoretical and experimental features of both optical and induction heating have been investigated and compared in the application to non-metallic insulation adhesively bonded to a metal structure. The effectiveness of using neural networks (NN) for characterizing defects has also been studied to demonstrate that optimal NN training should involve 4–5 points selected in defect areas close to non-defect areas, and the NN input data should be prepared by applying the known technique of Thermographic Signal Reconstruction (TSR). Since SLH provides more uniform heating, it provides higher quality IR thermograms than those obtained from areal (flash) heating and this improves the detectability of defects in thermal insulation to a depth of 4–6 mm. Other advantages of SLH for TNDT testing are (1) an inspection rate that is twice as high as an area heating technique and (2) a better potential for fully automated (robotic) testing.

Structural Health Assessment of Timber Utility Poles Using Stress Wave Propagation and Artificial Neural Network Techniques

Abstract: Timber utility poles represent a significant part of the power distribution and telecommunication infrastructure. Weathering, decay induced by fungus, and termite attacks deteriorate the condition of timber poles, causing a loss in their strength and toughness. Routine inspections are carried out to assess the condition of poles using conventional inspection techniques. However, the reliability of these techniques is in question. This paper proposes the stress wave propagation (SWP) technique for condition assessment of timber poles and the direct application of the artificial neural network (ANN) pattern recognition algorithm for signal classification. A Fourier-based signal decomposition method has been adopted and the frequency domain statistical features of the decomposed subcomponents were extracted to be used as the inputs for the ANN. Experiments were conducted using both intact and defective timber poles subjected to stress wave propagation. Different ANN models have been developed to classify the signals, and several controlling parameters were evaluated to obtain the best performance model. Further, the support vector machine (SVM) and k-means clustering algorithms were employed to classify the stress wave signals from intact and defective poles. Finally, the results of developed ANN models, SVM classifiers and k-means clustering models for pole classification were compared. The obtained success rates of the ANN model, the SVM classifier and the k-means clustering algorithm were 92%, 87% and 81%, respectively. Further, the trained ANN model was used to predict the health status of in-service poles, which were uprooted and subjected to full-scale bending tests after performing the stress wave propagation.

Transfer Component Analysis for Compensation of Temperature Effects on the Impedance-Based Structural Health Monitoring

Abstract: The effects of temperature fluctuations in the impedance measurements’ spectral estimates confuse the procedures to distinguish actual states’ classification, demanding compensation. The present paper demonstrates a new method to achieve temperature compensation based on a Transfer Component Analysis (TCA), a subtype of transfer learning, of the features from a source domain (in a well-known labeled condition) to another target domain (in an unknown condition). This procedure assumes only the labeled features data in the healthy condition (baseline) and damaged state in a specific known temperature as source data. The features computed are the Root Mean Square Deviation (RMSD) indices of the real and imaginary impedance signals. A machine-learning algorithm based on Mahalanobis squared distance ( $$\mathcal D^{2}$$ ) is trained using the features computed from the baseline condition in the reference temperature. Also, the other temperature and structural conditions data are assumed as testing data of the target condition. TCA’s main idea is mapping the features from the original features space to a new subspace where the detection becomes possible using the same training data in the source domain. The results performed in a testbench with a piezoelectric element (PZT) bonded under a set of temperatures monitored, and simulated damage confirmed that the proposed method could recognize the real states correctly by transferring the knowledge from the features of the source domain into the target domain, assuming different temperatures.

Analysis of Progressive Tensile Damage of Multi-walled Carbon Nanotube Reinforced Carbon Fiber Composites by Using Acoustic Emission and Micro-CT

Abstract: The potential to provide improved performance for advanced composites through the addition of multi-walled carbon nanotubes (MWCNTs) to carbon fiber composites is of interest in several applications. To investigate performance four types of composite specimens with different off-axis angles were subjected to progressive tensile loading. The results show that MWCNTs can improve the bearing capacity of the composite and the off-axis orientation angle can enhance the toughness of the composite. During loading acoustic emission (AE) signals were collected and they were post-processed using cluster analysis based on a Fuzzy C-Means algorithm. The analysis of the AE signals shows that data can be divided into categories which correlate with three damage modes: matrix cracking, fiber debonding and fiber breakage. The AE peak frequency characteristics of each damage mode were identified. Additional characterization was provided by using micro-computed tomography (Micro-CT) during the progressive tensile loading process. The CT images visualize damage location and evolution in the composites and data exhibit good correlations with the AE data for defects predication. The combination of AE and micro-CT technology were shown to effectively characterize damage evolution of the composites, and such data can potentially serve as a reference for the structural health monitoring of these composites when used in structures.

Application of a Particle Filter to Flaw Identification for Ultrasonic Nondestructive Evaluation: Assimilation of Simulated and Measured Data

Abstract: The assimilation of simulated and measured data is essential for advancing technology in NDE 4.0. In this study, a particle filter (PF) was applied to identify the geometry of flaws for ultrasonic nondestructive testing. A PF based on a probabilistic approach that allows errors in measurement and simulation models may be of great assistance for the data assimilation. In the PF, state variables are expressed by random data samples called particles, together with their associated weights. The PF estimates the probabilistic density function of the state variables by merging simulation data with measured data. Data types must be physically identical in the simulation and measurement to enhance the accuracy and to accelerate the convergence speed in the PF. Here, the scattering component, which is specific information related to the flaw geometry, was used for the likelihood evaluation in the PF. The simulation, which needed many particles, was conducted using the elastodynamic finite integration technique accelerated by parallel computing with graphics processing units. The proposed PF approach was demonstrated in ultrasonic measurement, and the geometries of artificial flaws in aluminum specimens were identified using only one pulse-echo signal at a single transducer.

Fruit Defect Prediction Model (FDPM) based on Three-Level Validation

Abstract: It is a known fact that infrared radiation is produced by all objects with a temperature above absolute zero. In fruits, IR light sensor senses the invisible areas and exposes obscure objects in the image. In normal RGB images, it is very difficult to predict the internal defect of the fruit accurately without chopping it into pieces. In this paper, a thermal IR imaging-based fruit defect detection technique is proposed to identify and estimate the internal defect in the pome fruits. The technique is non-invasive and non-destructive which helps in minimizing the fruit wastage during the quality check. To achieve high accuracy, a three-level validation process is adopted. First level involves the prediction made by the proposed Deep learning-based expert system using RGB and thermal images of apple respectively. In second level, the validation of results is done using Fourier Transform Infrared spectroscopy technique. And finally, in third level, an invasive destructive method is used for inspection of the fruit quality by cutting them into pieces. The apple defect detection accuracy by the proposed Naive Bayes classifier is observed to be 97.6% for thermal IR imaging samples whereas true color based achieved only 59% for the same sample. An apple seems to be healthy externally but internally there is a probability of a defect. Thermal IR imaging detects the heat from the surface of the fruit. Due to defective tissues of fruit pulp, the non-uniform temperature difference is observed and sensed on the surface of the fruit.

POD of NDT Techniques Using High Temperature Oxidized Fatigue Cracks in an Aero Engine Alloy

Abstract: Establishing probability of detection (POD) or reliability of various nondestructive testing (NDT) techniques is essential for implementing damage tolerant (DT) methodology for aero-engines. This POD is usually established with the help of a large number of service expired aero-engine components containing several fatigue cracks. In the absence of such components, artificial defects such as electrical discharge machining (EDM) notches or starter cracks were explored. However, such artificial defects would not meet the key features such as tightness of the fatigue cracks and the possible oxidation in the crack opening and thus, limiting their usage. Therefore, in the current study, an innovative approach of generating fatigue cracks at 650 °C (~ typical aero-engine service temperatures) with key high temperature service degradation aspects of oxidation and fatigue cracking is demonstrated for the first time using Gleeble® test system. Further, POD is estimated by inspecting these laboratory generated fatigue cracks using fluorescent liquid penetrant technique (FLPT) and eddy current technique (ECT) under HIT (defect detected) vs. MISS (defect not detected) and â (signal response) vs. a (crack size) methodologies. The current study also discusses a statistical approach of random generation of crack sizes for use in NDT reliability analysis. In addition, an attempt has been made to understand the effect of a90/95 values on remnant life calculations. It is concluded that the eddy current response of oxidized fatigue cracks results in better (high sensitive) a90/95 values compared to the eddy current response obtained from non-oxidized fatigue cracks.

Exploring Flood Filling Networks for Instance Segmentation of XXL-Volumetric and Bulk Material CT Data

Abstract: XXL-Computed Tomography (XXL-CT) is able to produce large scale volume datasets of scanned objects such as crash tested cars, sea and aircraft containers or cultural heritage objects. The acquired image data consists of volumes of up to and above $$\hbox {10,000}^{3}$$ voxels which can relate up to many terabytes in file size and can contain multiple 10,000 of different entities of depicted objects. In order to extract specific information about these entities from the scanned objects in such vast datasets, segmentation or delineation of these parts is necessary. Due to unknown and varying properties (shapes, densities, materials, compositions) of these objects, as well as interfering acquisition artefacts, classical (automatic) segmentation is usually not feasible. Contrarily, a complete manual delineation is error-prone and time-consuming, and can only be performed by trained and experienced personnel. Hence, an interactive and partial segmentation of so-called “chunks” into tightly coupled assemblies or sub-assemblies may help the assessment, exploration and understanding of such large scale volume data. In order to assist users with such an (possibly interactive) instance segmentation for the data exploration process, we propose to utilize delineation algorithms with an approach derived from flood filling networks. We present primary results of a flood filling network implementation adapted to non-destructive testing applications based on large scale CT from various test objects, as well as real data of an airplane and describe the adaptions to this domain. Furthermore, we address and discuss segmentation challenges due to acquisition artefacts such as scattered radiation or beam hardening resulting in reduced data quality, which can severely impair the interactive segmentation results.

Nondestructive Evaluation of Drilled Shaft Construction Anomalies Using Full Waveform Tomography of Simulated Crosshole Measurements

Abstract: Construction of drilled shafts for transportation infrastructure presents quality assurance challenges related to structural anomalies, including voids and degraded concrete. Efforts to evaluate anomalies often use stress-wave non-destructive testing (NDT) techniques such as the crosshole sonic logging (CSL) method. CSL relies on the propagation of stress waves between access tubes inserted alongside the reinforcement cage. Differences in travel time and amplitudes can be used to estimate concrete quality within the source-receiver ray paths. Crosshole tomography (CT) advances this approach further and models ray paths to solve an inverse problem based on first arrival times. However, both stress-wave methods cannot typically evaluate anomalies smaller than 10% to 15% of the shaft cross sectional area. Recent advances in wave propagation modeling have led to the development of full waveform inversion (FWI) approaches. FWI attempts to generate a more detailed tomographic model that matches the entire waveform recordings. This paper presents findings from a numerical study where full waveform tomography was performed on a portion of a drilled shaft with a structural anomaly. Crosshole stress-wave propagation was simulated using a spectral element method (SEM) forward modeler. The resulting waveforms were then used as inputs into a CT travel-time inversion algorithm using ray-path modeling and into a FWI algorithm using the entire recorded signals. The results highlighted that FWI estimated concrete velocities and the geometry of the anomaly more accurately than CT in the idealized numerical model. This was the case even though information from more sources and receiver pairs was used for the CT inversion.

Muon tomography of the interior of a reinforced concrete block: first experimental proof of concept

Abstract: Quality assurance and condition assessment of concrete structures is an important topic world-wide due to the aging infrastructure and increasing traffic demands. Common topics include, but are not limited to, localisation of rebar or tendon ducts, geometrical irregularities, cracks, voids, honeycombing or other flaws. Non-destructive techniques such as ultrasound or radar have found regular, successful practical application but sometimes suffer from limited resolution and accuracy, imaging artefacts or restrictions in detecting certain features. Until the 1980s X-ray transmission was used in case of special demands and showed a much better resolution than other NDT techniques. However, due to safety concerns and cost issues, this method is almost never used anymore. Muon tomography has received much attention recently. Novel detectors for cosmic muons and tomographic imaging algorithms have opened up new fields of application, such as the investigation of freight containers. Muon imaging also has the potential to fill some of the gaps currently existing in concrete NDT. As a first step towards practical use and as a proof of concept we used an existing system to image the interior of a reference reinforced 600 kg concrete block. Even with a yet not optimized setup for this kind of investigation, the muon imaging results are at least of similar quality compared to ultrasonic and radar imaging, potentially even better. The data acquisition takes more time and signals contain more noise, but the images allowed to detect the same important features that are visible in conventional high energy X-ray tomography. In our experiment, we have shown that muon imaging has potential for concrete inspection. The next steps include the development of mobile detectors and optimising acquisition and imaging parameters.

Machine Learning Based Preprocessing to Ensure Validity of Cross-Correlated Ultrasound Signals for Time-of-Flight Measurements

Abstract: High precision ultrasonic time-of-flight measurement is a well known part of non-destructive evaluation used in many scientific and industrial applications, for example stress evaluation or defect detection Although ultrasonic time-of-flight measurements are widely used there are some limitations where high noise and distorted ultrasonic signals are conflicting with the demand for high precision measurements Cross-correlation based time-of-flight measurement is one strategy to increase reliability but also exhibits some ambiguous correlation states yielding to wrong time-of-flight results To improve the reliability of these measurements a new machine learning based approach is presented based on experimental data collected on tightened bolts Due to the complex structure of the bolts the ultrasonic signal is influenced by boundary conditions of the geometry which lead to high number of the ambiguous cross-correlation results in practice In this particular application, bolts are in practice evaluated discontinuously and without knowledge of the time-of-flight in the unloaded condition which prevents the use of all other available comparative preprocessing techniques to detect time-of-flight shifts Three different preprocessing strategies were investigated based on variations in the bolting configurations to ensure a machine learning based model capable of predicting the state of the cross-correlation function for different bolting parameters With this approach, we achieve up to 100% classification accuracy for both longitudinal and transversal ultrasonic signals under laboratory conditions In the future the method should be extended to become more robust and be applicable in real-time for industrial applications