Prediction of tensile capacity of adhesive anchors including edge and group effects using neural networks

Abstract Adhesive anchors are widely used in seismic strengthening applications to add new structural members or sections to existing concrete members due to their high tensile and compressive strengths, low cost, and easy and fast installation. To safely design such anchors, it is very important to know their pullout capacity under axial tensile forces. This paper explores the pullout capacity of both single and groups of adhesive anchors loaded in tension in uncracked concrete. Quadruple anchor groups were considered for group effect. To this end, 142 single anchor tests including edge effect (located near a concrete edge) and 175 quadruple group anchor tests (totally 317 tests) were obtained from literature. The formulated three-layered artificial neural network method (ANN) was trained using 75% of the data set by using different learning algorithms. The methods were tested with the remaining 25%. The variables taken into account in this study are anchor diameter, embedment length, concrete compressive strength, concrete body height, edge distance (for single anchors), and anchor spacing (for group anchors). It was determined that experimental data can be estimated to a notably close extent via the ANN model.