An improved impedance-based damage classification using Self-Organizing Maps

The identification and severity of structural damages, especially in the early stage, is critical in structural health monitoring (SHM) systems. Among several approaches used to accomplish this goal, the electromechanical impedance (EMI) technique has taken place within non-destructive evaluation (NDE) methods. On the other hand, neural networks (NN) based on self-organizing maps (SOM) has been a promising tool in many engineering classification problems. However, there is a gap of application regarding the combination of the EMI technique and SOM NN. To encourage this, an enhanced EMI-based damage classification method using self-organizing features is proposed in the present research paper. A SOM NN architecture was implemented whose inputs were derived from representative features of the impedance signatures. As a result, self-organizing maps can be used as an effective tool to enhance the damage classification in EMI-based SHM applications. For the present application, the results indicated a promising and useful contribution to the grinding field.