What is meant by a concrete crack?

Concrete materials are meant for supporting structures, building foundations and they act as several load-bearing structures. Over time, due to environmental conditions, and based on the constituent materials used, concrete materials are subjected to a great degree of distress. Corrosion of the reinforcement bars due to chemical influences, uneven base, earthquake, drying shrinkages, and alkali-aggregate reactions are some of the reasons for concrete distress.

As a result of distress, the concrete develops cracks over time. If the cracks are left unrepaired, it might lead to crack propagation and ultimately, failure of the concrete material. A crack may be defined as a complete or incomplete separation of concrete material that is initiated due to fracture or material failure. Build cracks are the major issues related to a building, or any civil engineering material, which affect the aesthetics as well as the structural integrity of the building. Cracks reduce the durability of the structure.

It is therefore an important task of the civil engineers to early detect building cracks on the verge of crack initiation and they should undertake preventive measures to stop the crack propagation. Crack detection includes various techniques and procedures used by engineers and technicians during the building maintenance as well as construction phases. With the onset of technology and technology-based modernized equipment, it has become an easy task for engineers to carry on the crack detection methods and undertake necessary steps to ensure the longevity of the building structure.

Crack detection using image processing techniques

The figure below outlines the basic steps involved in crack detection using the image processing technique.

The image acquisition phase is an image-based method where high-resolution images of the component to be analyzed are acquired either through high definition camera or other sources. In the pre-processing stage, the input image is further processed through software and algorithms to remove distortions involved due to noise and shadows. The method requires the use of filters and segmentations. If required, the image is sometimes converted into greyscale and binary form, based on applications. In the detection stage, different methods of crack detection, such as edge detection, segmentation, and pixel analysis have been applied that highlight the affected part. In the crack feature extraction, the crack parameter estimations such as the crack depth, length, width, and density are determined.

The machine learning method of crack detection

It is a learning-based method of crack detection, which involves computer vision applications. The image below shows the architecture of a typical crack detection method using machine learning.

The first step involves creating a datasheet model shown on the surface cracks of a concrete structure. A surface crack can be identified on the concrete surface. Surface cracks result when the rate of evaporation is more than the water from the water ground. The dried surface results in segregation of the concrete materials, and eventually develops cracks. Pre-processing of the images is done to remove image noises and shadows. Minor alterations of images are done to adjust the image brightness and size, this increases the image clarity. The labeling algorithm annotates defected pixels. In the model training section, an appropriate crack detection model is selected, such as convolution neural networks, artificial neural networks, support vector machines, decision trees, and so on. The obtained new images are clarified and validated if they accurately match the crack characteristics.

Convolutional neural networks make use of deep learning algorithms executed by computer processors. Convolution neural networks take in the input of various images and assign importance to that image based on different aspects and features present in that image. It is used to differentiate between multiple images. Filters used in image processing are usually programmed in machine languages by engineers. With training, these filters can be executed by convolution neural networks. The architecture of this algorithm follows the connection mechanism of different neurons in a human brain in the visual cortex.

Magnetic particle testing for crack detection

This is a non-destructive method of crack detection, which detects surface and thin cracks of ferromagnetic materials. This forms a part of the visual inspection technique. The specimen or the concrete structure to be tested is magnetized locally depending on the presence of the crack. The magnetization induces a magnetic flux throughout the area. Due to the presence of a crack, the uniformity of the magnetic flux lines is distorted, causing leakage of flux lines. This leakage can visually be seen by laying out iron particles over the surface which may be dry or suspended on a liquid base. The particles get lined up near the regions of flux leakage, which are easily evaluated through visual inspection. Hence, the accumulation of iron particles near the area indicates cracks in the concrete structure.

Liquid penetrant testing

This is a low-cost non-destructive visual inspection testing to determine hairline surface flaws and cracks. The principle of capillary action is used to determine the crack characteristics of the part to be analyzed. The penetrating liquid is drawn into the material to be tested through capillary action. The phenomenon of capillary action occurs by breaking the surface discontinuities. Once the complete penetration is achieved, the access liquid is then removed. The process is followed by the application of a developer, which is typically in powdered form. This powder is applied at the site of the liquid paths which draws out the liquid from the material. This produces an indication by highlighting the crack positions inside the material.

For more knowledge and reading regarding crack detection techniques, one can refer to the works of Abdel-Qader who did extensive works in the field of mathematical modeling of crack detection.

Context and Applications

This topic is primarily taught in various undergraduate and postgraduate degree courses of:

• Bachelors of Technology in Mechanical engineering
• Bachelors of Technology in Civil engineering
• Masters of Technology in Civil engineering
• Masters of Technology in Mechanical engineering

Practice Problems

Q1) Which of the following is a non-destructive testing method of crack detection?

1. Liquid penetration technique
2. Magnetic particle testing
3. Both a and b
4. Image-processing technique

Answer: Option c

Explanation: Both liquid penetration technique and magnetic particle testing are non-destructive methods of crack detection techniques.

Q2) Which of the following techniques uses a developer?

1. Image processing technique
2. Liquid penetration technique
3. Magnetic particle testing
4. Machine learning technique

Answer: Option b

Explanation: The liquid penetration technique of crack detection makes use of the developer in the form of a powder.

Q3) The term, "edge detection" is associated with which of the following technique?

1. Image processing
2. Machine learning
3. Visual inspection
4. All of these

Answer: Option a

Explanation: The term edge detection is associated with image processing.

Q4) In the magnetic particle testing method, which of the following material is used to visually inspect the presence of a crack?

1. Iron particles
2. Steel particles
3. Aluminum powders
4. Magnetic field detectors

Answer: Option a

Explanation: Fine iron particles suspended either freely or in liquid are used to detect the altered magnetic fluxes. These fluxes directly signify the presence of a crack.

Q5) In which of the following steps in the machine learning technique, convolution neural networks are used?

1. Labeling step
2. Datasheet collection
3. Pre-processing
4. Model training

Answer: Option d

Explanation: In the model training section crack detection algorithm, convolution neural network is used.