Inspecting Buildings Using Drones and Computer Vision: A Machine Learning Approach to Detect Cracks and Damages

Manual inspection of infrastructure damages such as building cracks is difficult due to the objectivity and reliability of assessment and high demands of time and costs. This can be automated using unmanned aerial vehicles (UAVs) for aerial imagery of damages. Numerous computer vision-based approaches have been applied to address the limitations of crack detection but they have their limitations that can be overcome by using various hybrid approaches based on artificial intelligence (AI) and machine learning (ML) techniques. The convolutional neural networks (CNNs), an application of the deep learning (DL) method, display remarkable potential for automatically detecting image features such as damages and are less sensitive to image noise. A modified deep hierarchical CNN architecture has been used in this study for crack detection and damage assessment in civil infrastructures. The proposed architecture is based on 16 convolution layers and a cycle generative adversarial network (CycleGAN). For this study, the crack images were collected using UAVs and open-source images of mid to high rise buildings (five stories and above) constructed during 2000 in Sydney, Australia. Conventionally, a CNN network only utilizes the last layer of convolution. However, our proposed network is based on the utility of multiple layers. Another important component of the proposed CNN architecture is the application of guided filtering (GF) and conditional random fields (CRFs) to refine the predicted outputs to get reliable results. Benchmarking data (600 images) of Sydney-based buildings damages was used to test the proposed architecture. The proposed deep hierarchical CNN architecture produced superior performance when evaluated using five methods: GF method, Baseline (BN) method, Deep-Crack BN, Deep-Crack GF, and SegNet. Overall, the GF method outperformed all other methods as indicated by the global accuracy (0.990), class average accuracy (0.939), mean intersection of the union overall classes (IoU) (0.879), precision (0.838), recall (0.879), and F-score (0.8581) values. Overall, the proposed CNN architecture provides the advantages of reduced noise, highly integrated supervision of features, adequate learning, and aggregation of both multi-scale and multilevel features during the training procedure along with the refinement of the overall output predictions.

Lifetime measurement of a particular deep crack failure on a flat-type divertor mockup under cyclic high heat flux loading conditions

Plasma facing components (PFCs) are key to enduring high heat flux (HHF) loading from high-temperature plasma in nuclear fusion reactors. Understanding their thermal-mechanical behavior and cracking failure mechanisms related to structural designs and fabrication technologies during high heat flux loading is of great significance for improving their servicing performance and R&D (Research and Development) levels. In this study, a particular deep cracking failure process on the tungsten layer of a flat-type divertor mockup during 1800 cycles of 10 MW m-2 HHF loadings is completely monitored and measured with a special improved digital image correlation (DIC) technique. It is found that the DIC measurement under the HHF loading environment is improved successfully to capture fine deformation and strain fields with a spatial resolution less than 0.35 mm so that field strain on a 1 mm thick copper interlayer and deep crack initiation at several microns scale on the tungsten layer are measured out. Based on both full field and local strain and displacement measurements of the target divertor mockup, the thermal mechanical behaviors from deformation to crack initiation and propagation are successfully measured and traced. It is revealed that for the baseline copper interlayer design of a flat-type divertor mockup, the accumulation of plastic strain in the copper interlayer during ratcheting damage induces enough tensile stress on the tungsten layer during HHF cycles, leading to cracking and fracture failures even in its elastic state earlier than the copper LCF lifetime. Current SDC-IC rules fail to cover this kind of ratcheting cracking failure mode in the design stage. New design models or mechanical validation rules to resolve this design blind spot should be established in the future.

Вихретоковый датчик с двухчастотным возбуждением для обнаружения глубоких трещин

The skin effect is a common phenomenon in the eddy current testing (ECT), which can make the eddy currents concentrated near the specimen surface and can’t flow into the material. In order to improve the ability of the eddy current probe to detect the deep defects, a new probe with dual-frequency excitation is proposed, which can suppress the eddy current density near the surface of the specimen. The purpose of identifying deep crack defects in stainless steel specimen is achieved with the probe. In this paper, the ANSYS software is used to determine the frequency, excitation current and coil arrangement of the probe. The ability of the new probe to detect deep cracks is verified by experiments. The results show that the detection signal of the cracks with depth of 10mm and 15mm is 4 times and 4.47 times than that of the crack with 5mm depth. It is more sensitive to the depth change of deep cracks.

A new eddy current probe with deep penetrating field trajectories for the inspection of deep cracks in metal materials

The detectability of deep cracks in metal materials is an important performance index of eddy current probes. However, because of the limitations of the skin effect of eddy currents, it is difficult to obtain deep crack information in materials using an ordinary probe. This paper proposes a new probe with deep penetrating field trajectories for the inspection of deep cracks. To optimise its performance, contributions of the coil radius, the pick-up position and the excitation frequency to penetration depth of eddy currents are studied. The results show that the capability of the new probe in the inspection of deep cracks is greatly improved when compared to traditional pancake probes.

Development of New Design Fatigue Curves in Japan: Discussion of Crack Growth Behavior in Large-Scale Fatigue Tests of Carbon and Low-Alloy Steel Plates

In Japan, the Design Fatigue Curve (DFC) Phase 1 and Phase 2 subcommittees, which are a part of the Atomic Energy Research Committee of the Japan Welding Engineering Society, have proposed new design fatigue curves and fatigue analysis methods for carbon, low-alloy, and austenitic stainless steels. To confirm the validity of the proposed design fatigue curves, a Japanese utility collaborative project was launched, and the authors conducted fully reversed four-point bending fatigue tests for large-scale specimens of carbon steel and low-alloy steel plates. Subsequently, in a previous paper (PVP2018-84456), the authors reported that the fatigue lives determined by the best-fit curve proposed by the DFC subcommittee corresponded to those of approximately 1.5–7.0-mm-deep crack initiation in large-scale specimens. In this study, the fatigue crack initiation and propagation behavior observed in large-scale specimens was investigated by using a plastic replica and beach mark method. Similar to the case of small-sized specimens, in the large-scale specimens, multiple fatigue cracks initiated at an early stage of testing, and propagated with coalescence to penetrate the specimen width. However, no fatigue cracks were detected at the design fatigue life. Approximately 100-μm-long cracks were observed, albeit only after the specimen was subjected to a number of cycles that corresponded to approximately 3.5 times the design fatigue life. According to NUREG/CR-6909 Rev.1, the crack depths in small-sized round bar specimens at the fatigue lives, which are defined by 25%-stress-drop cycles, are reported to be approximately 3 mm. The results of the large-scale tests indicated that regardless of the specimen size, nearly the same phenomenon occurred on the specimen surface until approximately 3–4-mm-deep crack initiated. The size effect was mainly caused by the stress gradient. The finite element analysis indicated that the stress gradient in the large-scale specimen was gentle owing to the large thickness of the specimen, and the stress in the vicinity of the surface was considered to be uniform. In conclusion, the size effect was not apparent. As the same conclusion can be applied to considerably larger actual components, designers do not need to consider the size effect when designing pressure vessels or piping by using the design fatigue curve determined based on small-sized specimens.

Development of New Design Fatigue Curves in Japan: Discussion of Best-Fit Curves Based on Large-Scale Fatigue Tests of Carbon and Low-Alloy Steel Plates

In Japan, the Design Fatigue Curve (DFC) Phase 1 and Phase 2 subcommittees were organized under the Atomic Energy Research Committee in the Japan Welding Engineering Society and have proposed new design fatigue curves for carbon, low-alloy, and austenitic stainless steels. To confirm the validity of the proposed design fatigue curves, a Japanese utility collaborative project was launched. In this project, fatigue tests were conducted on large-scale and small-sized specimens, and the test data were provided to the DFC Phase 2 subcommittee. This paper discusses the best-fit curves proposed by the DFC Phase 1 subcommittee, focusing on the results of large-scale fatigue tests for carbon steel and low-alloy steel plates. The fatigue test results for large-scale specimens were compared with the best-fit curve proposed by the DFC Phase 1 subcommittee. This comparison revealed that the fatigue lives given by the proposed curves correspond to those of approximately 1.5–4.0-mm-deep crack initiation in large-scale specimens. In this program, fatigue tests with a mean strain were also carried out on large-scale specimens. These tests found that the fatigue lives were almost equivalent to those of approximately 4.4–7.0-mm-deep crack initiation in large-scale specimens. In determining a design fatigue curve, strain-controlled tests are usually performed on small-sized specimens, and the fatigue life is then defined by the 25% load drop. It is reported that the cracks reach nearly 3–4-mm depth under those 25% drop cycles. The test results confirm that the fatigue lives of large-scale specimens agree with those given by the best-fit curve for carbon and low-alloy steels, and no remarkable size effects exist for the crack depths compared in this study.