Detection of weak joints and damages for beams using machine learning

Abstract

For maintaining the safe operation of structures, it is necessary to develop SHM methods that can detect not only the presence of cracks in the structure but also any alterations of its fastening conditions. The current paper presents a method for developing an Artificial Intelligent model that can detect if a beam is affected by transverse cracks and at the same time, by improper boundary conditions. To this aim, a cantilever steel beam is considered as the in the current study. The training data for the artificial neural network (ANN) is created using an original analytic method which allows calculating the natural frequency loss caused by the occurrence of transverse cracks even if the beam is improperly fastened. The intelligent model is trained by employing the MATLAB software and tested using data acquired from numerical simulations. The results show very high accuracy in determining the presence of transverse cracks, and the capability of detecting the presence and severity of improper clamping conditions.