Identification of Single and Multiple Damage in Beams Using Natural Frequencies and Bayesian Data Fusion

Resumo

Monitoring the changes in geometric or structural properties of systems during their operation to evaluate the presence of damage is a well-established process entitled Structural Health Monitoring (SHM). It is one of the most widely discussed topics in mechanical, aerospace, and civil engineering, and exhibits an increasing importance in many other fields. SHM has a crucial relevance in determining the remaining useful life of a structure, which is essential in avoiding catastrophic failure. However, these analyses can be complicated and sometimes require destructive methods in order to achieve accurate results. Nonetheless, vibration-based methods have shown promising results as non-destructive approaches, relying primarily on variations of the modal parameters. In addition, previous studies demonstrated that natural frequencies are easily determined and provide great accuracy of results. The present work proposes the identification of single and multiple damages in beam-like structures using measures of natural frequencies. First, the beam is modeled using three different methods with increasing complexity. The first model is built in Ansys® APDL module with 100 elements using BEAM188 element type, and damage is introduced as a stiffness reduction in a particular element; the second model is constructed as a beam with solid elements, and damage is presented as a local reduction in stiffness; the third one is also built with solid elements, but with the damage mechanically introduced in the beam. The results for the natural frequencies are obtained via simulation. Then, an experimental setup is assembled in order to acquire results for the beam's natural frequencies and compare them to the values obtained via numerical analysis. These results are further processed by employing a Bayesian data fusion algorithm to predict the location of damage. Three possible damage locations are examined with varying intensities, ranging from a decrease in local stiffness from 10% to 50%. Different types of beam supports are also investigated. The results demonstrate that the algorithm successfully detects single and multiple damage and locates it with great accuracy under various types of supports.