Determination of Steel Area in Reinforced Concrete Beams Using Data Mining Techniques

Abstract

This study aimed to determine the area of reinforcing steel in rectangular reinforced concrete beams, a critical concern given that a significant proportion of residential structures in Peru do not conform to technical design regulations. Data were collected through a structured form encompassing various design variables, yielding a comprehensive data matrix. The methodology adopted involved Knowledge Discovery in Databases (KDD), executed in several stages: (1) selection, where the InfoGainAttributeEval algorithm was utilized to identify variables influencing the reinforcing steel area; (2) preprocessing, during which anomalous and duplicate data were purged using the Python libraries, Pandas and Numpy; (3) reduction, and (4) data mining. For the latter, Decision Stump, Hoeffding Tree, J48, Logistic Model Trees (LMT), and Random Tree classification algorithms were employed, facilitated by Weka 3.9.4. Accuracy rates of these algorithms were found to be 25, 51.70, 77.97, 78.39, and 88.98% respectively. The Random Tree algorithm, in conjunction with the GP_04 model, provided estimations of the steel area in the beams with a reliability exceeding 88%. The application of this model could enable the optimization of beam design, facilitate cost savings on materials, and enhance structural safety. This research thus presents a significant contribution to the field of structural engineering, particularly in regions where compliance with technical design standards is suboptimal.