Experimental and analytical investigation of the chemical anchors behaviors under the axial pulling effect

Significant portion of the existing building stock in Turkey is consisting of buildings made of reinforced concrete. Despite the heavy sanctions in developing technology and new earthquake codes, it is known that a significant portion of the existing building stock cannot perform sufficiently in terms of earthquake and needs to be strengthened urgently. In conventional reinforcement processes, such as reinforced concrete jacketing and adding shear wall methods, the connection between the existing structure and the newly added elements is done via chemical anchors with epoxy. The individual performance of the anchored anchors (whether or not they peel off from the existing structure) directly affects the global earthquake performance to be achieved by strengthening the structure. Therefore, the implementation of anchors is the most important phase of reinforcement. Anchors that do not have this qualification will peel off from the concrete. The fact that the anchors are stripped off from the concrete depends on many factors. Factors such as anchor depth, anchor diameter, hole diameter, opening hole cleaning and concrete class are the factors affecting the performance of anchors under pulling effect. In this study, to observe behavior of anchors under pure tension, 108 different samples were prepared and pull-out test was conducted. According to the different concrete class, reinforcement bar diameter, anchor depth, the size of the opening holes and the cleaning of the drilled holes, 108 pull out tests were carried out and the axial load capacities and the failure conditions due to these parameters were examined experimentally and analytically. In addition, the data obtained from the experimental results were compared by using artificial neural networks (ANN), support vector machine (DVM), stochastics gradient descent (SGD), trees and random forest (RF) algorithms commonly used in classification and estimation. According to the results of the study, it is observed that the increase of anchorage depth linearly increases their load capacity. The increase in concrete class and reinforcement diameter increases the load capacities similarly. Specimens that are sown in water-cleaned holes reach higher axial load capacities than air-cleaned. While the axial load capacities of non-cleaned holes were decreased significantly compared to the other cases. In the analytical study, it is observed that the ANN algorithm has a better estimation success (R2 = 0.783) than other algorithms. The success of the emprical formulas did not give satisfactory results.