Automated system for concrete damage classification identification using various classification techniques in machine learning / Nur Haziqah Mat ... [et al.]

Reinforced concrete is the most widely used material for Malaysian building construction. However, the significant disadvantage of this material is it is prone to the material damage, which causes a decrease in the durability of the concrete and causes structural damage. To determine the suitable repair technique on this material, proper identification procedure on damage classification must be executed. Currently, manual inspection performed by a qualified inspector is the primary inspection method to determine the concrete damage. The manual inspection is a process that is subjective and scarcely effective since it depends heavily on the personal experience and expertise of the inspector to interpret
the damage classification. Besides its subjective nature, manual inspection is also to be a time consuming approach, dangerous, inconsistent, costly, and a laborious task. The demand of experienced inspectors also presents a challenge for the pressing lack of highly skilled and experienced construction inspectors. To overcome the issues, datasets of reinforced concrete damage images are intelligently trained and classified by selected Machine Learning algorithms such as Naïve- Bayesian, Discriminant Analysis, K-Nearest Neighbor, and Support Vector Machine. This invention can recognize a certain damage while the classification of defects is classified according to the features extracted from the images by using GLCM algorithm. The performance of these algorithms is evaluated by dividing the dataset into two sections: testing and training. Cost and time usage can be minimized by using this invention which can help the engineers or construction inspectors. This invention is a significant tool that can predict types of reinforced concrete damage accurately.