Thesis (Ph.D.) - Cairo University - Faculty of Engineering - Department of Civil Engineering

Several techniques have been introduced to detect cracks and damages in concrete elements. Current practices of evaluating damages in concrete elements are costly and time consuming. This research presents a framework that utilizes artificial intelligence techniques to recognize cracks in reinforced concrete beams. The framework consists of three main components; Image Processing tool, Neural Network models, and Expert System model. Image processing tool utilizes percolation to identify the presence of the structure element and crack map. Then, Red-Green-Blue (RGB) to grayscale and to binary image conversion and filtering algorithms are applied to get a topological crack map.Many aspects are acquired such as coordinates, angels, diagonal, and Total Area of Crack Percentage (TACP) in order to identify geometric properties for both beam element and crack map. Graphical properties including length and orientation are extracted and mapped on the beam element to produce relative measurements and then to crack type recognition. Crack types are predicted using back propagation neural network model. Neural Network model receives geometric properties as an input and produces crack type identification as an output. The expert system model enhances ways of maintenance and rehabilitation. It utilizes the crack type (generated from neural network model) and TACP in order to provide the suitable repair method. Real images for two defected beams are used to validate the proposed framework and to compare its output to manually identified cracks and applied repair method. The results reveal the framework recommended solutions are in compliance with these that have been applied in reality

Issued also as CD