Estimating the compressive strength of concrete with automatic production tecknology by machine learning and deep learning

Due to the automatic building production technology, which has become widespread with the developing technology, fast and robust structures can be obtained. Automatic building production technology is based on the principle of breaking down the designed building model with computers and producing the disintegrated models in layers using concrete mortars. Compressive strength, which is one of the most important mechanical properties of concrete, is an important research topic. The methods used to measure the compressive strength of concrete can be examined under two main headings as destructive and non-destructive. Within the scope of this thesis, a new approach to non-destructive methods is proposed in order to estimate the compressive strength of concretes suitable for automated building production technology. In the thesis study, 192 cube concrete samples with 24 different mixing ratios were produced in accordance with automatic construction production technology. In this study, 192 cube concrete samples with 24 different mixing ratios in accordance with automatic construction production technology were produced. Aggregate particle diameter, cement ratio and fiber were used as variables in mixing ratios. Images were taken to estimate the compressive strength of the produced concrete samples with image processing methods. After the imaging process, the estimation and actual values are presented comparatively by looking at the compressive strengths in order to obtain the actual compressive strength results of concretes. In order to obtain textural and color information of concrete images, their features were extracted using Gray Level Co-occurrence Matrix and Histogram techniques. The created features were tested with the machine learning algorithms K Nearest Neighbor (KNN) and Support Vector Machines (DVM) and also with the developed deep learning network model. As a result of the studies, the accuracy rates of the KNN algorithm and SVM algorithm, which are among the machine learning methods, were observed as 88.44% and 88.19%, respectively. The compressive strength of concretes suitable for automatic production technology reached the highest estimation rate of 90.12% with the proposed deep learning algorithm.