Verification of Solenoid-Enabled Percussion to Monitor Bolted Structures with Neural Network and Support Vector Machine Classifiers

Percussion-based methods of structural health monitoring have been tested on various structures. Through verification of their viability in more specific situations, it could prove to be useful in giving a non-invasive and non-destructive method monitoring structures that are hard to reach. After setting up a solenoid to test the percussion method on two different flanges and a set of metal plates, it was found that in more specific scenarios as this, it can be difficult for the machine learning algorithms to sort specific levels of torque. It is proposed that more refinement can be applied to the method, both through refining the algorithms it uses and through changing how the algorithms sort the data.