A multiclassifier convolutional neural network to identify defect type and severity in roofing elements

Conventional damage localisation algorithms used in ultrasonic guided wave-based structural health monitoring (GW-SHM) rely on physics-defined features of GW signals. In addition to requiring domain knowledge of the interaction of various GW modes with various types of damages, they also suffer from errors due to variations in environmental and operating conditions (EOCs) in practical use cases. While several machine learning tools have been reported for EOC compensation, they need to be custom-designed for each combination of damage and structure due to their dependence on physics-defined feature extraction. In this work, we propose a convolutional neural network (CNN)-based automated feature extraction framework coupled with Gaussian mixture model (GMM) based temperature compensation and damage classification and localisation method. Features learnt by the CNNs are used for damage classification and localisation of damage by modelling the probability distribution of the features using GMMs. The Kullback–Leibler (KL) divergence of these GMMs with respect to corresponding baseline GMMs are used as signal difference coefficients to compute damage indices (DIs) along various GW sensor paths, and thus for damage localisation. The efficacy of the proposed method is demonstrated using FE generated GW-data for an aluminium plate with a network of six lead zirconate titanate (PZT) sensors, for three different types of damages (rivet hole, added mass, notch) at various temperatures (from 0 ∘C to 100 ∘C), with added white noise and pink noise to incorporate errors due to EOCs. We also present experimental validation of the method through characterisation of notch damage in an aluminium panel under varying and non-uniform temperature profiles, using a portable custom-designed field programmable gate array based signal transduction and data acquisition system. We demonstrate that the method outperforms conventional temperature compensation method using GMM with physics-defined features for damage localisation in GW-SHM systems prone to EOC variations.

This article addresses the general problem of built heritage protection against both deterioration and loss. To continuously monitor and update the structural health status, a crowd-sensing solution based on powerful and automatic deep learning technique is proposed. The aim of this solution is to get rid of the limitations of manual and visual damage detection methods that are costly and time-consuming. Instead, automatic visual inspection for damage detection on walls is efficiently and effectively performed using an embedded Convolutional Neural Network (CNN). This CNN detects the most frequent types of surface damage on wall photos. The study has been conducted in the Kasbah of Algiers where the four following types of damages have been considered: Efflorescence, spall, crack, and mold. The CNN is designed and trained to be integrated into a mobile application for a participatory crowd-sensing solution. The application should be widely and freely deployed, so any user can take a picture of a suspected damaged wall and get an instant and automatic diagnosis through the embedded CNN. In this context, we have chosen MobileNetV2 with a transfer learning approach. A set of real images have been collected and manually annotated and have been used for training, validation, and test. Extensive experiments have been conducted to assess the efficiency and the effectiveness of the proposed solution, using a 5-fold cross-validation procedure. Obtained results show in particular a mean weighted average precision of 0.868 ± 0.00862 (with a 99% of confidence level) and a mean weighted average recall of 0.84 ± 0.00729 (with a 99% of confidence level). To evaluate the performance of MobileNetV2 as a feature extractor, we conducted a comparative study with other small backbones. Further analysis of CNN activation using Grad-Cam has also been done. Obtained results show that our method remains effective even when using a small network and medium- to low-resolution images. MobileNetV2-based CNN size is smaller, and computational cost better, compared to the other CNNs, with similar performance results. Finally, detected surface damages have also been plotted on a geographic map, giving a global view of their distribution.

So far, the assessment or measurement of tree damage has only been done using the Forest Health Monitoring (FHM) method. This study aims to determine the types of tree damage using Forest Health Monitoring (FHM) and Convolutional Neural Network (CNN) methods. The research was conducted at the TAHURA WAR Utilization Block and the Computer Science Laboratory at FMIPA Lampung University. Measuring the type of tree damage using the FHM method is carried out on trees that are in the FHM cluster. Identification of tree damage types with the CNN algorithm using the MobileNet architecture. The results showed that there were 13 types of tree damage found, with five types of tree damage that were commonly found (> 60 cases): open wounds (218 cases), cancer (94 cases), Broken / Cracks and stems (87 cases), broken or dead branches (73 cases), and loss of dominant shoots (69 cases). As for the identification results with the CNN method, there were nine out of 13 types of damage that obtained precision, recall, and F1 scores of 100%. Thus, five types of dominant tree damage were found, one of which was open wounds (218 cases), and nine types of tree damage obtained high accuracy values.

Recognition, location, measurement, and 3D reconstruction of concealed cracks using convolutional neural networks

*† Composites present additional challenges for inspection due to their anisotropy, the conductivity of the fibers, the insulating properties of the matrix, and the fact that damage often occurs beneath the visible surface. This paper addresses the characterization of damage within composite materials, specifically for structural health monitoring (SHM). Fundamentally, one would like to distinguish between a pristine and damaged structure, however taking a micromechanics view, materials are inherently damaged. Microscopic flaws grow over time, and can be greatly accelerated by events such as overloads or impacts, until a critical damage size is achieved. Therefore a threshold must be introduced, where at some level of detectable flaw size, the structure must be labeled as “damaged”. Using one or two recorded features, such as time or frequency domain measurements, to characterize damage may not be feasible, as they may not be linearly separable. While it may be possible to differentiate between “pristine” and “damaged”, or between 2 discrete damage modes with limited features, it is not possible to separate the entire mode space. Therefore it is necessary to extract several feature sets to allow multi-dimensional classification of damage modes. The presented research utilized Lamb wave testing coupled with principal component analysis and pattern recognition methods, with the goal of providing the presence, type, and severity of damage with a high degree of confidence. Experiments were performed using quasi-isotropic graphite/epoxy laminates with 2 bonded actuator/sensor pairs. Three types of damage were investigated, each at 4 levels of severity: impact, hole and delamination. A total of 9000 datasets were collected in pulse-echo mode at 100kHz. Training data was collected from 1 plate and testing data from the other plates for each damage type. Subsequently, pattern recognition (PR) algorithms were developed to determine presence of damage, as well as to predict the type and severity of damage. These results have shown that PR methods can be used to successfully characterize damage in composites for SHM, with results that would only improve with additional training data.

Objective/scope The U.S. fuels infrastructure has undergone significant changes in the last several years. These changes are in response to shifts in domestic production, imports and exports, processing and distribution. The objective of this analysis is to define and model the U.S. fuels infrastructure, from production, transportation, and processing of crude oil to the distribution, storage, and consumption of refined products, and assess resiliency to natural disasters. This has become a pressing concern in the wake of Hurricane Harvey's impacts on U.S. infrastructure and fuels markets. Methods, procedures, process The analysis was conducted through a detailed assessment of the petroleum infrastructure, including wells, pipelines, refineries, crude and product terminals, import and export facilities, natural gas storage and processing facilities, and other key components. Interdependencies between petroleum and other types of infrastructure were also evaluated. As part of this analysis, a wide range of natural disasters, including earthquakes, hurricanes, tsunamis, and wildfires, were considered. For each of these natural threats, the geographic risks, the severity and type of damage, the likelihood of that damage, and the time required for recovery were all established using the latest available research. To complete the analysis several tools were developed and applied. These include a comprehensive geospatial system used to identify major markets and submarkets and to determine dependencies and vulnerabilities; models of the market supply/demand requirements; a national disruption model which details the major crude and product markets; and disruption models to predict the impact on crude and refined products from hurricanes, earthquakes, and tsunamis. Results, observations, conclusions The analysis showed that the U.S. fuels infrastructure is vulnerable to being impaired by natural disasters. Furthermore, the interdependencies of the infrastructure, in terms of both type of infrastructure and geographic location, leave the U.S. vulnerable to cascading risks. As part of the study, potential mitigation efforts were determined to address regional and national vulnerabilities. The proposed mitigation efforts have been incorporated into the Department of Energy's recent Quadrennial Energy Review. This study provides an update and extension of work performed earlier. Novel/additive information The paper provides a detailed assessment of the U.S. fuels infrastructure and its vulnerabilities to natural threats. These include earthquakes, hurricanes, and other disasters. For each of these threats, the geographic risks, as well as the type, probability, and severity of damage, are discussed. Finally, case studies of recent disasters, including the impact of Harvey on the Gulf Coast, Midcontinent, and East Coast, are used to illustrate the fuels infrastructure vulnerabilities and the need for mitigation.

Structural damage classification in composite materials using the Wigner-Ville distribution and convolutional neural networks

Road infrastructure is critical in transportation systems because it ensures the safe and efficient movement of people and goods. However, the deterioration of roads over time as a result of various factors such as weather and heavy traffic poses significant maintenance and safety challenges. Early and accurate detection of road damage is critical for timely repairs and accident prevention. This paper proposes a novel approach to detecting road damage using Convolutional Neural Networks (CNNs). CNNs have demonstrated remarkable success in a variety of computer vision tasks, making them an appealing option for automated road damage detection. The goal of this research is to use deep learning and computer vision techniques to create an efficient and accurate system for detecting road damage from images. Our methodology entails gathering a diverse dataset of road images with various types of damage, such as potholes, cracks, and road surface degradation. The dataset is pre-processed to improve image quality and annotated for training and evaluation. Using this dataset, a custom CNN architecture is designed and trained to recognize and classify various types of road damage. A separate validation dataset is used to evaluate the trained model's performance in terms of accuracy, precision, recall, and F1 score. Furthermore, we investigate the model's ability to generalize to previously unseen road damage scenarios by testing it on real-world images captured under varying conditions. Our CNN-based road damage detection system achieves high accuracy in identifying and classifying road damage types, according to the results. This system can be integrated into existing infrastructure management systems, allowing for cost-effective and timely road maintenance. Furthermore, it helps to improve road safety by identifying potential hazards before they cause accidents.

We propose a model assisted method to identify damage types and severity based on mode converted wave strength. Machine learning techniques are employed to develop classification models complemented by the finite element simulation models. Finite element simulation models provide the training data for various cases of damage and severity involving common types of damages in composites. Damage classification models are based on mode conversion strength versus frequency curves of participating four wave modes. For damage recognition and classification, a multi-layer Convoluted Neural Network (CNN) has been trained using the back-propagation paradigm on the generated dataset.

Post-earthquake seismic assessment of residential buildings following Sarpol-e Zahab (Iran) earthquake (Mw7.3) part 1: Damage types and damage states

Artificial intelligence-based hull structural plate corrosion damage detection and recognition using convolutional neural network

Concrete is known for its a strength and durability as a building material. It is heavily utilized in almost all infrastructures, from pipes, building structures to roads and dams. However, due to external factors or internal compositions, concrete can be damaged and hence affects the quality of the constructions. The type of damage that appeared on concrete is often the first a clue as to how it occurred. Therefore proper diagnosing of the problem can help engineers determine how quickly and how best to fix it. The application of information technology, especially artificial intelligence, to automatically classify the damage types can help tremendously in this aspect. There have been some studies in using computer vision to examine the surfaces of concrete for damages. This study attempts a more challenging task of classifying the five common types of concrete damage. A new dataset is built and the Convolutional Neural Network (CNN) architecture is used for classification. The results obtained have an accuracy of 95 and 93% on the training set and the test set respectively.

Ground penetrating radar (GPR) is a mature and important research method in the field of structural non-destructive testing. However, when the detection target scale is small and the amount of data collected is limited, it poses a serious challenge for this research method. In order to verify the applicability of typical one-dimensional radar signals combined with convolutional neural networks (CNN) in the non-destructive testing of concrete structures, this study created concrete specimens with embedded defects (voids, non-dense solids, and cracks) commonly found in concrete structures in a laboratory setting. High-frequency GPR equipment is used for data acquisition, A-scan data corresponding to different defects is extracted as a training set, and appropriate labeling is carried out. The extracted original radar signals were taken as the input of the CNN model. At the same time, in order to improve the sensitivity of the CNN models to specific damage types, the spectrums of A-scan are also used as part of the training datasets of the CNN models. In this paper, two CNN models with different dimensions are used to train the datasets and evaluate the classification results; one is the traditional one-dimensional CNN model, and the other is the classical two-dimensional CNN architecture AlexNet. In addition, the finite difference time domain (FDTD) model of three-dimensional complex media is established by gprMax, and the propagation characteristics of GPR in concrete media are simulated. The results of applying this method to both simulated and experimental data show that combining the A-scan data of ground penetrating radar and their spectrums as input with the CNN model can effectively identify different types of damage and defects inside the concrete structure. Compared with the one-dimensional CNN model, AlexNet has obvious advantages in extracting complex signal features and processing high-dimensional data. The feasibility of this method in the research field of damage detection of concrete structures has been verified.

An accurate assessment of the type and extent of sewer damage is an important prerequisite for maintenance and repair. At present, distinguishing drainage pipe defect types in the engineering field mainly relies on the human eye, which is time consuming, labor intensive, and subjective. Some studies have used deep learning to classify the types of pipe defects, but this method can only identify one main pipe defect. However, sometimes a combination of defects, such as corrosion and precipitation on a section of pipe wall, can be classified as one category by picture classification, which is significantly different from the reality. Furthermore, the deep learning method for defect classification is unable to pinpoint the precise location and severity of a defect or estimate the number of flaws and the cost of maintenance and repair. Therefore, an image segmentation method based on deep convolutional neural networks is proposed to achieve pixel-level image segmentation of defect regions while classifying pipe defects. Compared with the deep learning network for defect classification, it can segment a variety of defects and reduce the number of samples, which is convenient for defect measurement. First, the image defect locations of seven typical defects were manually labeled to create the dataset. Then, a model based on the SegNet network was used to label defect areas automatically in an image. The pipeline image dataset was used to test the previously trained model using the CamVid dataset. Finally, the model was applied to drainage pipe network images that were provided by periscope and closed-circuit television inspection cameras, and the pixel accuracy of image segmentation reached 80%. From the results, it can be concluded that image segmentation and annotation technology based on deep learning is applicable to sewer defect detection. The identification results of pipeline defects were accurate. The SegNet model is a reliable method for image analysis of pipeline defects, which can accurately evaluate the type and degree of sewer damage.

Conventional inspections in car damage assessments depend on visual judgments by human inspectors, which are labor-intensive and prone to fraudulent practices through manipulating damages. Recent advancements in artificial intelligence have given rise to a state-of-the-art object detection algorithm, the You Only Look Once algorithm (YOLO), that sets a new standard in smart and automated damage assessment. This study proposes an enhanced YOLOv9 network tailored to detect six types of car damage. The enhancements include the convolutional block attention module (CBAM), applied to the backbone layer to enhance the model’s ability to focus on key damaged regions, and the SCYLLA-IoU (SIoU) loss function, introduced for bounding box regression. To be able to assess the damage severity comprehensively, we propose a novel formula named damage severity index (DSI) for quantifying damage severity directly from images, integrating multiple factors such as the number of detected damages, the ratio of damage to the image size, object detection confidence, and the type of damage. Experimental results on the CarDD dataset show that the proposed model outperforms state-of-the-art YOLO algorithms by 1.75% and that the proposed DSI demonstrates intuitive assessment of damage severity with numbers, aiding repair decisions.