Prediction of the shear strength of reinforced masonry walls using a large experimental database and artificial neural networks

Abstract

Sandoval, C.; Adam Martínez, JM.; Garzón-Roca, J.; Valdebenito, G. (2016). Prediction of the shear strength of reinforced masonry walls using a large experimental database and artificial neural networks.

Behavior of Partially Grouted Concrete Masonry Walls under Quasi-Static Cyclic Lateral Loading

“…Much of the most recent research effort has been devoted to understanding the role of different design variables on the seismic behavior of PG-RM walls, as well as proposing shear strength equations for design purposes. In this context, several authors (e.g., [3][4][5][6][7][8][9]) have developed different expressions and methodologies for estimating the shear resistance of HCB PG-RM shear walls. However, most of those studies have focused on shear walls with bond-beam reinforcement.…”

“…As mentioned, the available experimental evidence on PG-RM shear walls made of HCB and with bed-joint reinforcement (BJR) is more limited and less heterogeneous in comparison with the available for PG-RM shear walls with BBR [6,7]. This fact illustrates the need to enlarge the experimental database on BJR-PG-RM shear walls built with HCB.…”

A computational framework for infrastructure performance predictions based on data, mechanics, and machine learning

“…The back-propagation algorithm was chosen to train a multilayer network with onward connections. Regarding to the shear expression used in the algorithm, several expressions had been tested and the one from ACI Committee showed more accurate results [5]. All in all, the proposed method predicted experiment results in a more accurate and less conservative way.…”

“…Besides shear strength of reinforced concrete structures, Aguilar [5] estimated in-plane shear strength of reinforced masonry (RM) using artificial neural network (ANN). The wall aspect ratio, the compressive strength of masonry, as well as details about reinforcement were taken into consideration as input variables.…”

“…The best test and partition will be chosen if its heuristic value is larger than initialized heuristic value of the best test in line [15]. As long as best test and partition are returned by function BT in line [2], invoke PCT function recursively to obtain tree and corresponding node in step k in line [5] and [6]. However, if best test is not found in line [7], namely, no test significantly reducing variance, then the algorithm will return a leaf and labeled as repetition of the training instances in line [8].…”