Two-step approach for fatigue crack detection in steel bridges using convolutional neural networks

Abstract

The advent of parallel computing capabilities, further boosted through the exploitation of graphics processing units, has resulted in the surge of new, previously infeasible, algorithmic schemes for structural health monitoring (SHM) tasks, such as the use of convolutional neural networks (CNNs) for vision-based SHM. This work proposes a novel approach for crack recognition in digital images based on coupling of CNNs and suited image processing techniques. The proposed method is applied on a dataset comprising images of the welding joints of a long-span steel bridge, collected via high-resolution consumer-grade digital cameras. The studied dataset includes photos taken in sub-optimal light and exposure conditions, with several noise contamination sources such as handwriting scripts, varying material textures, and, in some cases, under presence of external objects. The reference pixels representing the cracks, together with the crack width and length, are available and used for training and validating the proposed model. Although the proposed framework requires some knowledge of the "damaged areas", it alleviates the need for precise labeling of the cracks in the training dataset. Validation of the model by means of application on an unlabeled image set reveals promising results in terms of accuracy and robustness to noise sources.

Low saliency crack detection based on improved multimodal object detection network: an example of wind turbine blade inner surface

“…In the future, reducing alignment error for low saliency objects will be an issue worth investigating. (2) In this paper, we take cracks in carbon fiber composites of WTBs as a low saliency cracks example. In addition, it can be further expanded to other materials, especially metallic materials.…”

“…Timely detection and repair of cracks is important for the safe maintenance. Crack detection by visible (Vis) images and object detection algorithms is a low-cost method that can effectively identify cracks in metal, 1 – 3 concrete, 4 – 6 composite materials, 7 – 9 and food surfaces 10 , 11 . However, in some complex scenarios, the low saliency of the cracks causes poor detection results.…”

Fusing damage-sensitive features and domain adaptation towards robust damage classification in real buildings

“…Provided extensive labeled datasets for training and validation, neural networks have demonstrated very high performance in damage classification tasks, by combining information from multiple features [31,34]. Typically, CNNs are deployed directly on time-series measurements [84] or picture data [85]. In this work a CNN architecture is deployed on the pre-computed features (see Section 2.4), as illustrated in Fig.…”

A convolutional neural network deep learning method for model class selection

“…Importantly, the convolutional neural networks have already shown an impressive performance on selecting the class of visual imagery data 26 via an ability to recognize patterns. Here, a one‐dimensional version is examined for the vibration signals, which has shown a great potentially for damage detection in one or more dimensions 27–44 . The ability to provide the model class selection using a unique degree of freedom (DOF) response measurement, without system identification, and by using a neural network classification approach makes this approach distinctive from the current methodologies.…”