Application of Cognitive Technologies for Forecasting the Strength of Thin-Walled I-Beams

This paper discusses the application of cognitive technology to predict the strength of thin-walled I-beams. Experimental data were the results of own tests and experiments conducted by other researchers. Universal calculation methods for beams with composite and steel transverse reinforcement were obtained on the basis of artificial neural networks and classical regression analysis. Neural network and regression analysis were performed in the corresponding modules of the multifunctional software product STATISTICA. The neural network architecture is a multilayer perceptron consisting of three layers: input, hidden and output. The sample for creating the artificial neural network is divided into training, test, and validation. Both approaches demonstrated close values of experimental and calculated failure loads. The algorithm by which the neural network and regression models are implemented as a desktop application is written in the Python programming language. The advantages of Python are its simple syntax and wide possibilities of using libraries and tools, including for creating applications with a graphical interface. The application is developed on the basis of the cross-platform graphical toolkit Tkinter. With the ongoing digital transformation and the transition to BIM Level 3, the proposed algorithm can also be conveniently integrated into the corresponding BIM-models.