Data-driven structural health monitoring system for sheet metal assemblies

Sheet metal or polymer structures are used in engineering applications, as they provide an adequate weight-to-strength ratio. Monitoring the health and integrity of the sheet metal becomes crucial. Thin-walled structures such as aircraft wings, automobile bodies, and manufacturing machines can transfer vibrational waves through them, depending on the defect or damage in the material the vibrational wave characteristics change. Analysis of the vibrational wave using machine learning methods can help infer the type of damage as well as localize the damage. Data from experiments conducted on sheet metal specimens is used to train the model, the same machine learning model is applied to the Finite Element Analysis model of the sheet metal structure. As it is established that the machine learning based SHM works for the sheet metal without joints or discontinuity, the method is expanded on an FEA model of an assembly of sheet metal. Getting the data for this requires many samples of the targeted assembly, in this research Finite Element Analysis simulation techniques are used to create samples of the various types of sheet metal and composite assembly methods such as rivets, screws, welds, and adhesive to train the machine learning models for the Sheet Metal Structural Health Monitoring application. The developed method shows an effective and economical way forward for the machine learning based acoustic SHM application trained on Finite Element Models that utilize k-nearest neighbor and custom neural network.