Feature Selection for Automated Monitoring of Hybrid Roof Structures

The problem of selecting the important features of hybrid roof structures is considered. The hybrid roof consists of a light-weight framework and a flexible membrane shell. The features are used for estimation of the adverse effects, which act on the roof, by the Machine Learning model in form of Artificial Neural Network. Statistical methods are used for finding the meaningful features and for detecting redundant ones. The present work allows to cheapen the procedures of structural health monitoring by means of reducing the number of gauging devices which must be installed on the structure. It simplifies the Machine Learning model by reducing the input vector. Thus the model, which requires fewer resources to be trained and stored, can be deployed on mobile devices.