No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Probabilistic assessment of time to cracking of concrete cover due to corrosion using semantic segmentation of imaging probe sensor data
Abstract
This paper presents a framework for segmentation of imaging probe corrosion sensor data using a deep learning algorithm and estimation of the remaining service life of the structure using the segmented data. The sensor consists of a sacrificial metal foil that is imaged using the optical probe and the changes in the images as corrosion develops can be used as a proxy to monitor the condition of the concrete. In this paper, DeepLabV3+ which is a deep learning network architecture is implemented for the segmentation of sensor images. The neural network model trained on labeled corroded and uncorroded images of foil captured under various chloride levels yields a test accuracy of 95%. The mass loss of steel is estimated using a Bayesian curve fitted over the estimated mass loss from the segmented images and the mass loss from the accelerated corrosion test. This is then used for the estimation of the corrosion rate, which is given as the input for the probabilistic estimation of the time at which the concrete cover is expected to crack. A case study is presented to demonstrate how the segmented images from the neural network model can be used for estimating the time to cracking of concretes.
... A plethora of algorithms, such as FCN, PSPNet, U-Net, SegNet, the Deeplab series, Transformer, and notably SegFormer, have been developed to address this intricate task [11][12][13][14][15][16][17][18]. Within the sphere of civil engineering, and more specifically in the identification of structural defects, the application of semantic segmentation algorithms has demonstrated superior performance [19][20][21][22][23][24][25]. The research presented herein utilizes the SegFormer network, which is predicated on the Transformer design paradigm, expressly conceived for the exigent task of pixel-level image segmentation. ...
Within the domain of architectural urban informatization, the automated precision recognition of two-dimensional paper schematics emerges as a pivotal technical challenge. Recognition methods traditionally employed frequently encounter limitations due to the fluctuating quality of architectural drawings and the bounds of current image processing methodologies, inhibiting the realization of high accuracy. The research delineates an innovative framework that synthesizes refined semantic segmentation algorithms with image processing techniques and precise coordinate identification methods, with the objective of enhancing the accuracy and operational efficiency in the identification of architectural elements. A meticulously curated data set, featuring 13 principal categories of building and structural components, facilitated the comprehensive training and assessment of two disparate deep learning models. The empirical findings reveal that these algorithms attained mean intersection over union (MIoU) values of 96.44% and 98.01% on the evaluation data set, marking a substantial enhancement in performance relative to traditional approaches. In conjunction, the framework’s integration of the Hough Transform with SQL Server technology has significantly reduced the coordinate detection error rates for linear and circular elements to below 0.1% and 0.15%, respectively. This investigation not only accomplishes the efficacious transition from analog two-dimensional paper drawings to their digital counterparts, but also assures the precise identification and localization of essential architectural components within the digital image coordinate framework. These developments are of considerable importance in furthering the digital transition within the construction industry and establish a robust foundation for the forthcoming extension of data collections and the refinement of algorithmic efficacy.
Three-dimensional concrete printing (3DCP) faces challenges in determining and ensuring adequate bond strength between reinforcement and printed concrete.
Traditional methods for predicting bond performance are merely deterministic without considering potential uncertainty, which would lead to risks for structural safety.
To address this issue, this study develops a trustworthy machine learning based prediction model for bond strength in reinforced printed concrete (RPC) structures using Natural Gradient Boosting algorithm.
This developed model provides both scalar bond strength predictions and corresponding standard deviations, and in the test, it achieved a 94.5% safety rate and outperformed empirical formulas and deterministic approaches.
Instructive guidance can be offered for structural engineers and designers in determining reinforcement embedment lengths for 3D-printed concrete during constructions.
This probabilistic prediction approach can further enhance the safety and efficiency of digitally fabricated concrete structures, potentially extending its application to other critical parameters in printed concrete.
A strain transfer model is developed to describe the transfer of crack opening displacement (COD) of double cracks in a structural substrate to the distributed fiber optic sensor (DFOS). Global and local coordinate systems are set up to aid in the deduction of the transfer of CODs of the double cracks to the DFOS. The analytical model of strain transfer is first derived in the local coordinate system and then is transformed into the global coordinate system. Analytical results show that the DFOS fiber strain induced by CODs of the double cracks in the structural substrate is affected by both the spacing and the CODs of the double cracks, and the strain peak is suppressed as the crack spacing is less than the characteristic length which is defined as the range of influence of the COD of the individual crack. The influences of the material properties of the DFOS, the fiber coating, the bonding adhesive, and the DFOS-fiber coating interfacial stiffness are discussed. The developed model is also experimentally validated by monitoring CODs of double cracks in aluminum alloy plates with optical frequency domain reflectometer (OFDR) DFOSs. Experimental results show that the strain distribution measured by the OFDR DFOSs agree well with the analytical results of the developed model. The limitations of the developed model and future work anticipated are then discussed.
Structural health monitoring (SHM) is crucial for maintaining concrete infrastructure. The data collected by these sensors are processed and analyzed using various analysis tools under different loadings and exposure to external conditions. Sensor-based investigation on concrete has been carried out for technologies used for designing structural health monitoring sensors. A Sensor-Infused Structural Analysis such as interfacial bond-slip model, corroded steel bar, fiber-optic sensors, carbon black and polypropylene fiber, concrete cracks, concrete carbonation, strain transfer model, and vibrational-based monitor. The compressive strength (CS) and split tensile strength (STS) values of the analyzed material fall within a range from 26 to 36 MPa and from 2 to 3 MPa, respectively. The material being studied has a range of flexural strength (FS) and density values that fall between 4.5 and 7 MPa and between 2250 and 2550 kg/m3. The average squared difference between the predicted and actual compressive strength values was found to be 4.405. With cement ratios of 0.3, 0.4, and 0.5, the shear strength value ranged from 4.4 to 5.6 MPa. The maximum shear strength was observed for a water–cement ratio of 0.4, with 5.5 MPa, followed by a water–cement ratio of 0.3, with 5 MPa. Optimizing the water–cement ratio achieves robust concrete (at 0.50), while a lower ratio may hinder strength (at 0.30). PZT sensors and stress-wave measurements aid in the precise structural monitoring, enhanced by steel fibers and carbon black, for improved sensitivity and mechanical properties. These findings incorporate a wide range of applications, including crack detection; strain and deformation analysis; and monitoring of temperature, moisture, and corrosion. This review pioneers sensor technology for concrete monitoring (Goal 9), urban safety (Goal 11), climate resilience (Goal 13), coastal preservation (Goal 14), and habitat protection (Goal 15) of the United Nations’ Sustainable Development Goals.
The application of deep learning methods in civil engineering has gained significant attention, but its usage in studying chloride penetration in concrete is still in its early stages. This research paper focuses on predicting and analyzing chloride profiles using deep learning methods based on measured data from concrete exposed for 600 days in a coastal environment. The study reveals that Bidirectional Long Short-Term Memory (Bi-LSTM) and Convolutional Neural Network (CNN) models exhibit rapid convergence during the training stage, but fail to achieve satisfactory accuracy when predicting chloride profiles. Additionally, the Gate Recurrent Unit (GRU) model proves to be more efficient than the Long Short-Term Memory (LSTM) model, but its prediction accuracy falls short compared to LSTM for further predictions. However, by optimizing the LSTM model through parameters such as the dropout layer, hidden units, iteration times, and initial learning rate, significant improvements are achieved. The mean absolute error (MAE), determinable coefficient (R2), root mean square error (RMSE), and mean absolute percentage error (MAPE) values are reported as 0.0271, 0.9752, 0.0357, and 5.41%, respectively. Furthermore, the study successfully predicts desirable chloride profiles of concrete specimens at 720 days using the optimized LSTM model.
In the desert region of northwest China, the frequency of wind-sand disasters is high. All types of concrete buildings built in this area face severe wind erosion due to high wind speed, resulting in varying degrees of wind-erosion damage to concrete. To accomplish intelligent identification of concrete wind-erosion damage, a concrete wind erosion experiment was conducted in the laboratory, and a concrete wind-erosion damage dataset was generated under the interference of water stains, scratches, shooting distance, and background noise. This paper combined with transformer theory to improve YOLO-v4 and proposed an object detection algorithm called MHSA-YOLOv4 suitable for wind-erosion damage of concrete. The results demonstrate that MHSA-YOLOv4 exhibits improved object detection performance than YOLO-v3, improved YOLO-v3, and YOLO-v4. On the test set, ACC, Precision, Recall, and mAP of MHSA-YOLOv4 are 91.30%, 91.52%, 92.31%, and 0.89, respectively. MHSA-YOLOv4 can accurately identify wind-erosion damage of concrete images under different test conditions, which reflects strong robustness. The applicability of computer vision technology to the intelligent identification of wind-erosion damage on concrete has been verified.
Metal thin films are used in a wide range of devices. To ensure long-term reliability, damage prediction of the films should be considered. In particular, it is important to perform damage prediction of Cu thin films on flexible circuit boards because of severe fatigue conditions under repetitive, large strains. In this study, the development of multiple, complicated fatigue cracks in Cu thin films was predicted using a deep convolutional neural network (CNN). The CNN was trained to output a microscopic image of the fatigue crack development after 2000 cycles upon receiving a microscopic image of the area prior to crack development as an input. The microscopic images used for training were obtained via an actual fatigue test. In addition, the electrical resistance ratio, which is an important indicator of damage, was predicted by another CNN. The CNN output the electrical resistance ratio upon receiving a microscopic image of fatigue crack development as an input. More than half of the predicted cracks corresponded with actual cracks. Furthermore, the average error of the evaluated electrical resistance ratio was small (6.5%).
Corrosion of reinforcing steel is the leading cause of deterioration in concrete and impacts strongly the safety and durability of civil infrastructures. It occurs in marine environment with the presence of chloride which breaks the thin oxide film passivation layer leading to dissolution. This research proposes an innovative embedded sensor in concrete for the detection of corrosion of steel. The autonomous UHF RFID (Ultra High Frequency Radio Frequency Identification) sensor is based on the coupling between the antenna of a dipolar RFID tag and a layer of steel exposed to chloride in concrete. Variation of the RSSI (Received Signal Strength Indication) measured by the reader localized in air has two origins, namely the degree of moisture of concrete and the presence of the steel layer. By minimizing the impact of seawater ingress on the antenna property, the presence of metallic films of few micrometers thickness can be detected. This authorizes the development of the proposed method for monitoring mass loss of steel in concrete.
In civil engineering, many structures are made of reinforced concrete. Most degradation processes relevant to this material, e.g., corrosion, are related to an increased level of material moisture. Therefore, moisture monitoring in reinforced concrete is regarded as a crucial method for structural health monitoring. In this study, passive radio frequency identification (RFID)-based sensors are embedded into the concrete. They are well suited for long-term operation over decades and are well protected against harsh environmental conditions. The energy supply and the data transfer of the humidity sensors are provided by RFID. The sensor casing materials are optimised to withstand the high alkaline environment in concrete, having pH values of more than 12. Membrane materials are also investigated to identify materials capable of enabling water vapour transport from the porous cement matrix to the embedded humidity sensor. By measuring the corresponding relative humidity with embedded passive RFID-based sensors, the cement hydration is monitored for 170 days. Moreover, long-term moisture monitoring is performed for more than 1000 days. The experiments show that embedded passive RFID-based sensors are highly suitable for long-term structural health monitoring in civil engineering.
The structural behavior of reinforced concrete (RC) beams subjected to non-uniform rebar corrosion is studied here by developing a simplified numerical approach using three-dimensional (3D) nonlinear finite element (FE) analysis and experimental study. In order to validate the proposed numerical model and to understand the influence of nonuniform rebar corrosion on the structural performance of RC beams, seven reinforced concrete beams with identical dimensions and reinforcement layout were tested under static loading. Out of the seven beams, six beams were subjected to a different level of corrosion using an electrochemical method, and another beam was used as a control specimen without corrosion. To further verify the accuracy of the numerical model, another experimental study from the literature was selected. Finally, to investigate the influence of non-uniform rebar corrosion on the global behavior of RC structures, a hypothetical building was modelled based on the experimental result, and non-linear structural analysis was carried out. Results from the experimental and numerical study demonstrated that non-uniform corrosion of steel bars led to a significant decrease in the ductility and load carrying capacity of the RC beams. Findings of this study will be useful in assessing the residual structural performance of the RC structures with non-uniform rebar corrosion. Therefore, this study will facilitate the optimization of the maintenance and repair strategies for the deteriorated RC structures.
The corrosion-induced expansion of pre-stressed concrete cylinder pipes shapes a major determining factor in the durability of most pipeline systems. However, a structure extended in a very large span and accompanied by a slow corrosion-induced structural damage is still a challenge in practice. Here, a low-coherent fiber-optic interferometry is proposed for in situ measuring the corrosion-induced expansion by directly circling the sensing optical fiber on the outer surface of the pipe. Since the accuracy of the absolute deformation measurement by the low-coherent fiber-optic interferometry can reach 3–5 μm, we are able to improve the sensitivity by choosing the length of the sensing fiber. In the experiment, we employ 50-m sensing fiber; therefore, a resolution of about 0.1 microstrains had been approached. We chose 12 places along the pipeline as the corrosion-induced expansion monitoring points. On monitoring for about half a year, a non-uniform interface in structure, such as the repaired places, was found as a key factor of the corrosion-induced expansion development, which could be a reference for the further health care of the pipeline structures.
Herein, we report the first electrochemical sensor based on a screen-printed electrode designed to evaluate the corrosion level in iron-reinforced concrete specimens. The combination of an Ag pseudoreference electrode with a gel polymeric electrolyte allows for fast, stable and cost-effective potentiometric measurements, suitable for evaluating the corrosion of iron bars embedded in concrete samples. The sensor was found to be capable of discriminating between a standard non-corroded sample and samples subject to corrosion due to the presence of chloride or carbonate in the concrete matrix. The potential in concrete-based specimens containing carbonate (pH 9, −0.35 ± 0.03 V) or chloride (4% w/w, −0.52 ± 0.01 V) was found to be more negative than in a standard concrete-based sample (−0.251 ± 0.003 V), in agreement with the ASTM standard C876 method which uses a classical Cu/CuSO4 solid reference electrode. Our results demonstrate that a printed Ag pseudoreference electrode combined with KCl agar provides an efficient and reliable electrochemical system for evaluating the corrosion of iron bars embedded in concrete-based structures.
To monitor the initiation of concrete cracking beyond the service life of the structure, a novel prediction model of time to cracking of concrete cover using artificial neural network (ANN) was developed in this study. Crack mitigation prevents corrosion and crack development to occur in a more rapid phase that is an essential component in performance-based durability design of reinforced concrete structures. Data available in various literatures were used in the development of the ANN model which is a function of compressive strength, tensile strength, concrete cover, rebar diameter, and current density. The neural network model was able to provide reasonable results in time predictions of cracking of concrete protective cover due to formations of corrosion products. The performance of ANN model was also compared to various analytical and empirical models and was found to provide better prediction results. Even with limitations in the available training data, the ANN model performed well in simulating cracking of concrete due to reinforcement corrosion.
Corrosion of steel in reinforced concrete structures has been a major worldwide problem. The time to cover cracking plays a key role in assessment of serviceability of reinforced concrete structures subjected to corrosion. A large number of analytical, numerical, and empirical models have been developed to predict the time to time to cover cracking. In addition, extensive experiments have been conducted in order to verify the developed models. In this paper, an overview of the existing models is presented. A large experimental database of reported data on time to cover cracking is collated. Performance of four analytical models taken from the available literature is then examined using the established experimental database. Sensitivity analysis followed by a probabilistic study is carried out to identify the factors affecting time to cover cracking as a service life indicator by means of the selected models. The results from sensitivity analysis show that the porous zone and the properties of rust are the most influential factors for time to cover cracking. It is also shown that there is high uncertainty in predicting service life of reinforced concrete structures based on time to cover cracking.
During the long history of computer vision, one of the grand challenges has been semantic segmentation which is the ability to segment an unknown image into different parts and objects (e.g., beach, ocean, sun, dog, swimmer). Furthermore, segmentation is even deeper than object recognition because recognition is not necessary for segmentation. Specifically, humans can perform image segmentation without even knowing what the objects are (for example, in satellite imagery or medical X-ray scans, there may be several objects which are unknown, but they can still be segmented within the image typically for further investigation). Performing segmentation without knowing the exact identity of all objects in the scene is an important part of our visual understanding process which can give us a powerful model to understand the world and also be used to improve or augment existing computer vision techniques. Herein this work, we review the field of semantic segmentation as pertaining to deep convolutional neural networks. We provide comprehensive coverage of the top approaches and summarize the strengths, weaknesses and major challenges.
Several methods for corrosion monitoring of reinforced concrete structures (RCS) have been proposed in the last few decades. These systems may be used either in new, existing or repaired structures. The corrosion monitoring can be performed by different methodologies. These may or may not be destructive, use different degrees of complexity and cost, and provide information on the progression and kinetics of the corrosion phenomena. The destructive methods are limited to sampling. Therefore, these may not be representative of the whole structure, which is extremely important in RCS with large heterogeneities both in terms of materials used and in terms of the exposure environment. Within this context, non-destructive methods have been widely developed, which are intended to provide quick information about the entire structure. Ideally, these systems should be able to detect the corrosion state of the steel inside the concrete, the main causes of corrosion and the evolution of corrosion phenomena over time. This manuscript reviews and summarizes the actual state of the art and the main achievements in the field of electrochemical sensors based on non-destructive methods for corrosion monitoring of RCS in the last few years. The challenges and perspectives in this field will also be discussed.
A strain based corrosion sensor for the detection of environmental corrosion of pre-stressed structures was developed and tested on mild steel specimens readily available. Theoretically, a beam shaped specimen under a displacement load exhibits a linear relationship between the strain observed at any point through the thickness of the beam cross-section. This property was exploited to detect thickness changes in pre-stressed mild steel specimens in a double bending configuration under an electrochemically excited corrosion reaction. The reaction was accelerated by supplying a DC current to the cell where the specimens act as anodes of the system, while graphite rods serve as cathodes. The strain was logged using fiber optic Bragg grating technology and conventional electrical strain gages simultaneously. Results show a strong relationship between the corrosion rate observed by back-calculation from supplied current and the time derivative of the measured strain values. Application of this sensor can therefore be extended to a variety of structures under mechanical loading, proving valuable for both its ability to measure corrosion rate in real-time, while maintaining an intrinsically safe nature appropriate by industrial standards.
We present a novel and practical deep fully convolutional neural network
architecture for semantic pixel-wise segmentation termed SegNet. This core
trainable segmentation engine consists of an encoder network, a corresponding
decoder network followed by a pixel-wise classification layer. The architecture
of the encoder network is topologically identical to the 13 convolutional
layers in the VGG16 network . The role of the decoder network is to map the low
resolution encoder feature maps to full input resolution feature maps for
pixel-wise classification. The novelty of SegNet lies is in the manner in which
the decoder upsamples its lower resolution input feature map(s). Specifically,
the decoder uses pooling indices computed in the max-pooling step of the
corresponding encoder to perform non-linear upsampling. This eliminates the
need for learning to upsample. The upsampled maps are sparse and are then
convolved with trainable filters to produce dense feature maps. We compare our
proposed architecture with the fully convolutional network (FCN) architecture
and its variants. This comparison reveals the memory versus accuracy trade-off
involved in achieving good segmentation performance. The design of SegNet was
primarily motivated by road scene understanding applications. Hence, it is
efficient both in terms of memory and computational time during inference. It
is also significantly smaller in the number of trainable parameters than
competing architectures and can be trained end-to-end using stochastic gradient
descent. We also benchmark the performance of SegNet on Pascal VOC12 salient
object segmentation and the recent SUN RGB-D indoor scene understanding
challenge. We show that SegNet provides competitive performance although it is
significantly smaller than other architectures. We also provide a Caffe
implementation of SegNet and a webdemo at
http://mi.eng.cam.ac.uk/projects/segnet/
The reinforcement corrosion detection has received considerable attention in the field of structural health monitoring (SHM) due to the rising awareness about the safety and longevity of reinforced building structures particularly the steel bars. In this study, we propose a corrosion-monitoring device based on a tapered polymer optical fiber sensor (TPFS) for steel bars. The sensing principle of the proposed sensor was based on the corrosion expansion law and the attenuation loss equation to the transmitted light. In the test, TPFS with three different taper diameters were characterized in the separate accelerated corrosion test with a steel bar immersed in the 3.5 wt% NaCl solution. Experimental results show that each TPFS had a unique characteristic response corresponding to the increasing corrosion products that build up around the steel bar. In addition, the sensitivities of three types of TPFS were compared in terms of the two stages, and results indicated that the sensor with 0.6 mm tapered size would be an ideal alternative for corrosion monitoring of steel bars.
Towards the automatic defect detection from images, this research develops a semi-supervised generative adversarial network (SSGAN) with two sub-networks for more precise segmentation results at the pixel level. One is the segmentation network for the defect segmentation from labeled and non-labeled images, which is built on a dual attention mechanism. Specifically, the attention mechanism is employed to extract the rich and global representations of pixels in both the spatial and channel dimension for better feature representation. The other one is the fully convolutional discriminator (FCD) network, which employs two kinds of loss functions (the adversarial loss and the cross-entropy loss) to generate the confidential density maps of unlabeled images in a semi-supervised learning manner. In contrast to most existing methods heavily relying on labeled or weakly-labeled images, the developed SSGAN model can leverage unlabeled images to enhance the segmentation performance and alleviate the data labeling task. The effectiveness of the proposed SSGAN model is demonstrated in a public dataset with four classes of steel defects. In comparison with other state-of-the-art methods, our developed model using 1/8 and 1/4 labeled data can reach promising mean Intersection over Union (IoU) of 79.0% and 81.8%, respectively. Moreover, the proposed SSGAN is robust and flexible in the segmentation under various scenarios.
Corrosion of reinforcement in concrete is one of the major causes for deterioration of concrete structures. This work presents an approach in the monitoring of concrete structures for damage induced by corrosion of steel in concrete using a lens-based plastic optical fiber (LPOF) strain sensor. The optical fiber sensor offers the advantage of being light weight, small in size, low-cost, immune to electromagnetic interference and it does not pose any spark hazard. The intensity-based optical fiber sensor used in this work consists of an emitter fiber, a ball lens, a receiving fiber and a light detector system. The light from the emitter fiber converges to focal point using the ball lens, before it enters the receiver fiber. A change in distance between the ball lens and the receiver fiber would lead to a change in the light intensity transmitted. The intensity change is correlated to the relative distance using a suitable calibration curve. Following the calibration of the sensor, it was used to monitor response of concrete members subjected to flexural loading. Subsequently, the sensor was tested for its ability to monitor corrosion of steel in concrete using an accelerated corrosion test set-up. The sensor shows promising results for detection of corrosion. The test results were used for identification of the corrosion propagation phase, which can be used for predicting the remaining service life of the structures. The optical fiber sensor strain readings were then correlated to the corrosion penetration depth to estimate the extent of damage during the corrosion of rebar. The test results show that it is possible to monitor concrete structures for damage due to flexural loading and corrosion of rebar using an intensity-based LPOF strain sensor.
Corrosion detection on metal constructions is a major challenge in civil engineering for quick, safe and effective inspection. Existing image analysis approaches tend to place bounding boxes around the defected region which is not adequate both for structural analysis and prefabrication, an innovative construction concept which reduces maintenance cost, time and improves safety. In this paper, we apply three semantic segmentation-oriented deep learning models (FCN, U-Net and Mask R-CNN) for corrosion detection, which perform better in terms of accuracy and time and require a smaller number of annotated samples compared to other deep models, e.g. CNN. However, the final images derived are still not sufficiently accurate for structural analysis and prefabrication. Thus, we adopt a novel data projection scheme that fuses the results of color segmentation, yielding accurate but over-segmented contours of a region, with a processed area of the deep masks, resulting in high-confidence corroded pixels.
Corrosion of steel is one of the major causes for the degradation of reinforced concrete (RC) structures. In this regard, the assessment of reinforcement corrosion is critical in evaluating the service life of RC structures. The present study aimed to develop an electromagnetic-based apparatus which can be externally positioned on RC structures, to detect and monitor the corrosion of embedded steels. Experiments and numerical simulations based on finite element method (FEM) were carried out to verify the effectiveness of the external electromagnetic (EM) sensor. Different degrees of corrosion for steel in concrete specimens were obtained by impressing current method. Experimental results showed a linear relationship between the mass loss of corroding steel and the change of magnetic flux density sensed by the external EM sensor. Compared with the numerical results obtained from FEM simulations, a good accuracy was exhibited in estimating the mass loss of corroding steel in concrete by the change of magnetic flux density. Numerical simulations were further conducted to investigate the effects of the dimension and position of steel, the distance between testing surfaces of the EM sensor, and the magnitude of magnetic field intensity. As a non-destructive technique, a promising application of the external EM sensor was demonstrated in terms of evaluating the corrosion degree of steel in concrete, with the feasibility to increase its sensitivity by controlling the driving current and coil turns.
This research develops a novel computer vision approach named a spatial-channel hierarchical network (SCHNet), which is feasible to support the automated and reliable concrete crack segmentation at the pixel level. Specifically, SCHNet with a base net Visual Geometry Group 19 (VGG19) contains a self-attention mechanism, which is realized by three parallel modules, including the feature pyramid attention module, the spatial attention module, and the channel attention module. It can not only consider the semantic interdependencies in spatial and channel dimensions, but also adaptively integrate local features into their global dependencies. The segmentation performance is evaluated by a metric named Mean Intersection over Union (IoU) in a public dataset containing 11,000 cracked and non-cracked images with a unified resolution at 256 × 256 pixels (px). The experimental results confirm the effectiveness of the three attention modules, since they can individually increase Mean IoU by 1.62% (74.16%–72.54%), 5.15% (79.31%–74.16%), and 5.76% (79.92%–74.16%), respectively. With the help of new strategies like the data augmentation and multi-grid method, SCHNet can boost Mean IoU to 85.31%. In a comparison of the state-of-the-art models (i.e. U-net, DeepLab-v2, PSPNet, Ding, Dilated FCN) on the test dataset, SCHNet can outperform others with an improvement of at least 7.51% in Mean IoU. Moreover, SCHNet is robust to noises with a better generalization ability under various conditions, including shadows, roughness surfaces, and holes. Overall, this research contributes to developing SCHNet to integrate spatial and channel information in feature extraction, resulting in a more accurate and efficient crack detection process.
Reinforced bar corrosion is a primary reason for performance deterioration of reinforced concrete (RC) structures. Corrosion monitoring of reinforced bars in concrete is a key point to evaluate the durability and the remaining service life of structures, especially in marine environments. Non-destructive tests (NDTs) have been developed as critical methods to assess damage levels inside concrete without external damage based on physical characterization (e.g., electromagnetic response, acoustic signals, etc.). This paper provides a novel idea where home-made electromagnetic (EM) sensors and an acoustic emission (AE) apparatus are combined in order to monitor the entire corrosion process of reinforced concrete beams; meanwhile, a home-made electromagnetic sensor is also upgraded and introduced in this paper. The result shows that it is an excellent option to jointly utilize two kinds of monitoring apparatus to monitor signals from corrosion initiation and subsequent cracking, respectively, based on their own physical performances. Moreover, the reliability of monitoring signals is also highlighted in this paper.
This paper presents a novel sensor for monitoring concrete surrounding the rebar for chlorides ingress, which is one the main precursors for rebar corrosion. The sensor consists of an optical probe housing a camera and electrical units. It also consists of a sacrificial metal foil, which is representative of the rebar material. The thickness range of the foil is identified such that it would allow detection of chlorides before cracking of concrete cover. Corrosion of this sacrificial metal foil is indicative of the presence of chlorides beyond threshold levels in concrete surrounding the rebar. The optical probe and the inner surface of the foil are housed in a tube and completely sealed with a suitable adhesive. The sensor was tested under different levels of chlorides in concrete. Images of the inner surface of the foil captured using the camera are used to visually identify the corroded and uncorroded regions of the foil. In this work a suitable pixel counting image processing algorithm is also discussed to automate the corrosion detection of the sacrificial metal foil. Some of the factors affecting the performance of the sensor such as the effect of the sensing foil material and the type of adhesive used for sealing the housing are studied and presented here. Based on the tests carried out, a suitable adhesive for the sensor is identified. In conclusion, the key contribution of this paper is the development and demonstration of a novel low-cost imaging sensor for monitoring chloride ingress in the concrete surrounding the rebar. The preliminary test results for sensor development are promising and requires further testing for successful field implementation, which will be undertaken in future work.
Discovering and assessing cracks is widely thought to be critical for maintaining the healthy conditions of asphalt pavement. Unfortunately, the inspection of pavement for cracks is not only labor-intensive, time-consuming, inefficient, and costly, but it is also unable to detect and quantify cracks accurately at the pixel level. To address this problem, we propose an integrated approach based on the convolutional neural network DeepLabv3+ for crack detection, as well as a crack quantification algorithm for crack quantification at the pixel level. The quantification algorithm is used to evaluate five important indicators: crack length, mean width, maximum width, area, and ratio. To fully verify the performance of DeepLabv3+, 50 images were studied; the best image showed a mean intersection of union (MIoU) of 0.8342. For testing, 80 new images (including both asphalt pavement images and concrete pavement images) were used. DeepLabv3+ was found to be reliable and widely applicable for crack detection, and it demonstrated an MIoU of 0.7331. Of the various quantitative indicators, the crack length had the lowest relative error rate of the predicted values and therefore had the highest accuracy (its relative error rate ranged from −25.93% to 14.11%). We also compared our system with four state-of-the-art methods. The results showed our integrated approach to be more effective and more accurate in both the detection and quantification of cracks. The integrated approach could potentially serve as the basis of an automated, cost-effective pavement-condition assessment scheme for the operation and maintenance of pavement.
Corrosion of steel bars compromises the safety and service life of reinforced concrete structures. This study develops an in-situ corrosion monitoring method for reinforced concrete with a distributed fiber optic sensor through experimentation. Reinforced concrete beams instrumented with distributed fiber optic sensors were prepared. A constant current was impressed to the beams immersed in a NaCl solution for accelerated corrosion. The distributed fiber optic sensor was deployed in a helix pattern on the steel bar to measure expansive strains generated by corrosion of the steel bar. The corrosion process of the steel bar was assessed in an electrochemical test. The strain measured from the sensor was utilized to evaluate the volume of the corrosion products surrounding the steel bars and predict the cracking of the concrete cover. To investigate the deterioration process of reinforced concrete, different levels of concrete cover thickness (28 mm, 35 mm, and 43 mm) and water-to-cement ratio (0.4, 0.5, and 0.6) were studied. The relationship between the mass loss of steel bars and the volume of corrosion products is established to provide a method for evaluating the effects of steel corrosion on the deterioration of reinforced concrete.
Graphene oxide-manganese oxide (GO-MnO 2 ) nanomaterial was tried as electrochemical corrosion potential sensor and corrosion rate monitoring sensor in reinforced concrete structures. Graphene oxide-manganese dioxide nanomaterial was prepared by a simple chemical method. The synthesized material was characterized by Fourier transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), Raman spectra and transmission electron microscopy (TEM). An embeddable corrosion potential sensor (ECPS) and embeddable corrosion rate monitoring sensor (ECRMS) was assembled using nanomaterial suitable for measuring rebar potential and corrosion rate of rebar in concrete structures. The reversibility and reliability of the sensors were monitored under simulated concrete environmental conditions. The long-term stability was monitored for a period of 24 months under an active and passive state of reinforcing steel in concrete with respect to ECRMS and the results were compared with the surface mounting techniques. The stability of GO-MnO 2 based ECPS and ECRMS was found to be excellent in an active and passive condition and hence can be put forth as a promising new candidate material for corrosion monitoring in concrete structures.
Reliable predictions of the time to onset of corrosion in reinforced concrete are essential for service life modelling, to ensure sufficient durability, and for holistic sustainability assessments of new materials. All existing models are based on the same concept, that is, predicting the development over time of the chloride content at the level of the steel and comparing this numerical result with the critical chloride content for corrosion initiation, Ccrit. This paper presents example calculations utilizing input data derived from both laboratory specimens and from structures, illustrating the poor predictive power of state-of-the-art models. While it is generally assumed that improving chloride ingress models will improve the prediction of the time-to-corrosion, this paper shows that the bottle neck to more reliable predictions are rather i) the lack of fundamental understanding of corrosion initiation, ii) the use of non-representative laboratory results, and iii) ignoring the size-effect in localized corrosion.
Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task. The former networks are able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective fields-of-view, while the latter networks can capture sharper object boundaries by gradually recovering the spatial information. In this work, we propose to combine the advantages from both methods. Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries. We further explore the Xception model and apply the depthwise separable convolution to both Atrous Spatial Pyramid Pooling and decoder modules, resulting in a faster and stronger encoder-decoder network. We demonstrate the effectiveness of the proposed model on the PASCAL VOC 2012 semantic image segmentation dataset and achieve a performance of 89% on the test set without any post-processing. Our paper is accompanied with a publicly available reference implementation of the proposed models in Tensorflow.
This paper summarizes the grand societal, economic, technological, and educational challenges related to corrosion of steel in concrete, and presents the state-of-the-art of the most relevant issues in the field. The enormous financial impact of infrastructure corrosion seems to be inadequately balanced by educational and research activities. This presents a unique opportunity in many countries for maintaining or improving their competitiveness, given the major technological challenges can be solved. The main technological challenges are (1) the ever-increasing need to cost-effectively maintain existing, ageing reinforced concrete structures, and (2) designing durable, thus sustainable new structures. The first challenge arises mainly in industrialized countries, where there is a need to abandon conservative, experience-based decision taking and instead move to innovative, knowledge-based strategies. The second challenge regards mainly emerging countries expanding their infrastructures and where thus a major beneficial environmental impact can still be made by providing long-lasting solutions. This means to be able to reliably predict the long-term corrosion performance of reinforced concrete structures in their actual environments, particularly for modern materials and in the absence of long-term experience. During the second half of the last century, civil engineers, materials scientists, and chemists have in many countries made considerable attempts towards understanding corrosion of steel in concrete, but many of the approaches got bogged down in empiricism. From reviewing the state-of-the-art one can conclude that transport modeling in concrete is relatively well-advanced, at least in comparison with understanding corrosion initiation and corrosion propagation, where many questions are still open. This presents a number of opportunities in scientific research and technological development that are discussed in this paper.
The corrosion of steel bars in reinforced concrete structures induces concrete cover cracking and leads to reduction in steel–concrete bond strength. In the present study, a series of pull-out tests were carried out to investigate the potential correlations between corrosion level, surface crack width, and bond strength. The main variables are corrosion level, concrete cover depth, and spacing of stirrups. The test results indicate that the surface crack width is closely linked to the corrosion level of the tensile steel bar, and the crack propagation can be divided into three stages. It is further revealed that the bond strength decreases exponentially with the surface crack width. Based on the test results, an empirical model for estimating the bond strength using the surface crack width is formulated, which considers the pronounced limiting effect of stirrups on the bond degradation. The model is further improved by incorporating the adverse effect of the corrosion of stirrups according to the previous investigations. Comparison with existing experimental database in the literature proves that the proposed models are able to realistically predict the bond strength of corroded steel bars.
Can a large convolutional neural network trained for whole-image classification on ImageNet be coaxed into detecting objects in PASCAL? We show that the answer is yes, and that the resulting system is simple, scalable, and boosts mean average precision, relative to the venerable deformable part model, by more than 40% (achieving a final mAP of 48% on VOC 2007). Our framework combines powerful computer vision techniques for generating bottom-up region proposals with recent advances in learning high-capacity convolutional neural networks. We call the resulting system R-CNN: Regions with CNN features. The same framework is also competitive with state-of-the-art semantic segmentation methods, demonstrating its flexibility. Beyond these results, we execute a battery of experiments that provide insight into what the network learns to represent, revealing a rich hierarchy of discriminative and often semantically meaningful features.
Convolutional networks are powerful visual models that yield hierarchies of
features. We show that convolutional networks by themselves, trained
end-to-end, pixels-to-pixels, exceed the state-of-the-art in semantic
segmentation. Our key insight is to build "fully convolutional" networks that
take input of arbitrary size and produce correspondingly-sized output with
efficient inference and learning. We define and detail the space of fully
convolutional networks, explain their application to spatially dense prediction
tasks, and draw connections to prior models. We adapt contemporary
classification networks (AlexNet, the VGG net, and GoogLeNet) into fully
convolutional networks and transfer their learned representations by
fine-tuning to the segmentation task. We then define a novel architecture that
combines semantic information from a deep, coarse layer with appearance
information from a shallow, fine layer to produce accurate and detailed
segmentations. Our fully convolutional network achieves state-of-the-art
segmentation of PASCAL VOC (20% relative improvement to 62.2% mean IU on 2012),
NYUDv2, and SIFT Flow, while inference takes one third of a second for a
typical image.
In this work, we revisit atrous convolution, a powerful tool to explicitly adjust filter's field-of-view as well as control the resolution of feature responses computed by Deep Convolutional Neural Networks, in the application of semantic image segmentation. To handle the problem of segmenting objects at multiple scales, we design modules which employ atrous convolution in cascade or in parallel to capture multi-scale context by adopting multiple atrous rates. Furthermore, we propose to augment our previously proposed Atrous Spatial Pyramid Pooling module, which probes convolutional features at multiple scales, with image-level features encoding global context and further boost performance. We also elaborate on implementation details and share our experience on training our system. The proposed `DeepLabv3' system significantly improves over our previous DeepLab versions without DenseCRF post-processing and attains comparable performance with other state-of-art models on the PASCAL VOC 2012 semantic image segmentation benchmark.
A Major factor that affects the durability of a concrete structure is cracks formation induced by expansion of reinforcement corrosion. Therefore, monitoring and evaluating the corrosion level of the structure is essential for its safety. In order to monitor corrosion, an innovative methodology based on Fiber Bragg Grating (FBG) sensing technique was developed and tested in this paper. The method uses the volume of corrosion products to detect the evolution of corrosion. The corrosion process was accelerated by impressed current technique. A correlation between the FBG wavelength shift and corrosion percentage of reinforcement was found.
To evaluate the damage degree of reinforced concrete due to steel bar corrosion, a damage factor is proposed to quantitatively evaluate the damage degree of concrete before initial cracking and during the development of cracks. Brillouin optical fiber time domain analysis (BOTDA) sensors are fabricated to monitor the expansion strain of steel corrosion. Two concrete specimens embedded with corrosion sensors are cast. An accelerated corrosion experimental program is used to accelerate the process of steel corrosion. The experimental results show that the corrosion sensor can be used to monitor the expansion strain of steel corrosion in real-time. At last, to map the monitoring results with the damage factor, finite element analysis is used to simulate the process of steel corrosion to determine the cracking strain of the interfacial concrete and concrete cover.
We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available.
To investigate the fundamental relationship between corrosion rate and magnetic induction surrounding steel reinforcement, an innovative magnetic-based corrosion evaluation apparatus has been developed, which can be directly embedded inside reinforced concrete structures to monitor the corrosion rate of reinforcement. Preliminary calibration results show that the mass loss of corrosive reinforcement has a linear relationship with the voltage increment detected by Hall-effect sensor due to the variation of magnetic induction surrounding the corroded reinforcement. Subsequently, an experiment on the reinforced concrete beam subjected to chloride solution coupled with external loading has been conducted. Magnetic technique, half-cell potential measurement and acoustic emission detection have been utilized simultaneously in the experiment to monitor the corrosion process of reinforcement. The experimental results from the aforementioned techniques demonstrate good consistency. Especially, the magnetic corrosion monitoring device proposed in this study has the good capacity of quantitative analysis of corrosion rate.
Feature pyramids are a basic component in recognition systems for detecting objects at different scales. But recent deep learning object detectors have avoided pyramid representations, in part because they are compute and memory intensive. In this paper, we exploit the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost. A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales. This architecture, called a Feature Pyramid Network (FPN), shows significant improvement as a generic feature extractor in several applications. Using FPN in a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners. In addition, our method can run at 5 FPS on a GPU and thus is a practical and accurate solution to multi-scale object detection. Code will be made publicly available.
There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at
http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net
.
In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or 'atrous convolution', as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed "DeepLab" system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7% mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.
Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, improve on the previous best result in semantic segmentation. Our key insight is to build "fully convolutional" networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning. We define and detail the space of fully convolutional networks, explain their application to spatially dense prediction tasks, and draw connections to prior models. We adapt contemporary classification networks (AlexNet, the VGG net, and GoogLeNet) into fully convolutional networks and transfer their learned representations by fine-tuning to the segmentation task. We then define a skip architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations. Our fully convolutional network achieves improved segmentation of PASCAL VOC (30% relative improvement to 67.2% mean IU on 2012), NYUDv2, SIFT Flow, and PASCAL-Context, while inference takes one tenth of a second for a typical image.
Corrosion of reinforcing steel is the primary deterioration mechanism for reinforced concrete structures, however its impact on structural behaviour can be difficult to assess using visual inspections. This paper investigates the potential for using distributed fibre optic sensors to monitor the impact of corrosion to supplement visual inspections. Tension tests were conducted on bare reinforcement and reinforced concrete specimens subjected to accelerated corrosion. The distributed fibre optic strain data provided insights into the concrete-reinforcement bond performance of corroded specimens and the detection of pitting corrosion appeared to be possible although challenging.
Compressive strength reduction of concretes due to corrosion of reinforcement is one of the main reasons of failure in reinforced concrete structure which has not been taken into account by researchers yet. Therefore, in this article, for the sake of examination of concrete's compressive strength reduction due to rebar corrosion, reinforced concrete cubic specimens are constructed, and then, accelerated corrosion is applied to them. With fixing all effective parameters in compressive strength except water-to-cement ratio, specimens with different water-to-cement ratios (0.4, 0.45, and 0.5) are constructed with various reinforcements. Finally, compressive strength tests are performed for corroded and non-corroded specimens and reduction in compressive strength is measured under different degrees of corrosions.
Deeper neural networks are more difficult to train. We present a residual
learning framework to ease the training of networks that are substantially
deeper than those used previously. We explicitly reformulate the layers as
learning residual functions with reference to the layer inputs, instead of
learning unreferenced functions. We provide comprehensive empirical evidence
showing that these residual networks are easier to optimize, and can gain
accuracy from considerably increased depth. On the ImageNet dataset we evaluate
residual nets with a depth of up to 152 layers---8x deeper than VGG nets but
still having lower complexity. An ensemble of these residual nets achieves
3.57% error on the ImageNet test set. This result won the 1st place on the
ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100
and 1000 layers.
The depth of representations is of central importance for many visual
recognition tasks. Solely due to our extremely deep representations, we obtain
a 28% relative improvement on the COCO object detection dataset. Deep residual
nets are foundations of our submissions to ILSVRC & COCO 2015 competitions,
where we also won the 1st places on the tasks of ImageNet detection, ImageNet
localization, COCO detection, and COCO segmentation.
A physical-mathematical model is developed for concrete exposed to sea water. The model describes: (1)Diffusion of oxygen, chloride ions, and pore water through the concrete cover of reinforcement; (2)ferrous hydroxide near steel surface; (3)the depassivation of steel due to critical chloride ion concentration; (4)the cathodic and anodic electric potentials depending on oxygen and ferrous hydroxide concentrations according to Nernst equation; (5)the polarization of electrodes due to changes in concentration of oxygen and ferrous hydroxide; (6)the flow of electric current through the electrolyte in pores of concrete; (7)the mass sinks or sources of oxygen, ferrous hydroxide, and hydrated red rust electrodes, based on Faraday law; and (8)the rust production rate, based on reaction kinetics. To enable calculations, numerical values of all coefficients are indicated. The theory is completed by formulating the problem as an initial-boundary value problem.