Fatigue cracks are critical types of damage in steel structures due to repeated loads and distortion effects. Fatigue crack growth may lead to further structural failure and even induce collapse. Efficient and timely fatigue crack detection and segmentation can support condition assessment, asset maintenance, and management of existing structures and prevent the early permit post and improve life cycles. In current research and engineering practices, visual inspection is the most widely implemented approach for fatigue crack inspection. However, the inspection accuracy of this method highly relies on the subjective judgment of the inspectors. Furthermore, it needs large amounts of cost, time, and labor force. Non-destructive testing methods can provide accurate detection results, but the cost is very high. To overcome the limitations of current fatigue crack detection methods, this study presents a pixel-level fatigue crack segmentation framework for large-scale images with complicated backgrounds taken from steel structures by using an encoder-decoder network, which is modified from the U-net structure. To effectively train and test the images with large resolutions such as 4928 × 3264 pixels or larger, the large images were cropped into small images for training and testing. The final segmentation results of the original images are obtained by assembling the segment results in the small images. Additionally, image post-processing including opening and closing operations were implemented to reduce the noises in the segmentation maps. The proposed method achieved an acceptable accuracy of automatic fatigue crack segmentation in terms of average intersection over union (mIOU). A comparative study with an FCN model that implements ResNet34 as backbone indicates that the proposed method using U-net could give better fatigue crack segmentation performance with fewer training epochs and simpler model structure. Furthermore, this study also provides helpful considerations and recommendations for researchers and practitioners in civil infrastructure engineering to apply image-based fatigue crack detection.

The US Department of Transportation suggests a minimum thickness of 7 in (178 mm) to 8 in (203 mm) for unbonded concrete overlay. Thinner overlays driven by economic advantages have attracted worldwide attention. To prove the feasibility of thinner overlay, especially ultra-thin unbonded concrete overlays, robust overlay assessment techniques are demanded. In this article, glass fiber–reinforced polymer–packaged fiber Bragg grating sensors were developed and deployed to evaluate the early performance of fiber-reinforced unbonded concrete overlay with a thickness of 3 in (76 mm) placed on a 7.5-in (191-mm) concrete pavement, using a fabric interlayer as bond breaker. The early performance of the ultra-thin concrete overlay within a year was monitored and investigated, indicating the occurrence of transverse cracks located adjacent to the transverse cracks in the existing pavements. The numerical analysis and field experiments confirmed the accuracy and robustness of the fiber optic sensors, which showed their potential for further applications in concrete pavement monitoring.

The international roughness index (IRI) is a characterization of road roughness or ride quality that transportation agencies most often report. The prevalent method of acquiring IRI data requires instrumented vehicles and technicians with specialized training to interpret the results. The extensive labor and high cost requirements associated with the existing approaches limit data collection to at most once per year for portions of the national highway system. Agencies characterize roughness only for some secondary roads but much less frequently, such as once every five years, resulting in outdated roughness information. This research developed a real-time roughness evaluation approach that links the output of durable in-pavement strain sensors to prevailing indices that summarize road roughness. Field experiments validated the high consistency of the approach by showing that it is within 3.3% of relative IRI estimates. After their installation and calibration during road construction, the ruggedized strain sensors will report road roughness continuously. Thus, the solution will provide agencies a real-time roughness monitoring solution over the remaining service life of road assets.

Response-type methods have been widely used for road roughness evaluation. However, an important limitation is that they require calibration to account for response and speed variations among instrumented vehicles. The findings of this research obviate the need for calibration by applying a machine learning technique to estimate a roughness category and a roughness index from inertial sensors aboard at least two connected vehicles. The method leverages the future availability of inertial sensor data feeds from connected vehicles. The approach offers an alternative to specifically instrumented vehicles and specially trained technicians. In lieu of data from actual connected vehicles, the authors validated the method by numerical simulations using a model of the vehicle suspension system and a mathematical representation of the road roughness profile. Solving the dynamic response model as a function of various levels of roughness excitation and suspension parameters produced vertical acceleration signals. Subsequently, speed normalization and a convolution of the vertical acceleration responses from at least two simulated vehicles produced a common signal for feature extraction. Finally, a feature selection algorithm provided the most impactful features for a machine learning algorithm to train, test, and classify into a roughness category, and to estimate the international roughness index. Results show that the classification and estimation accuracy exceed 90%.