Intelligent structural health detection method for adhesive bonding composite materials based on machine learning
Kierownik projektu: Yang Zhang
Instytut Maszyn Przepływowych PAN w Gdańsku
Data otwarcia: 2021-11-24
Damage failure of these materials is due to undetected and unmonitored damage mechanisms happening in them, because of an external event such as aging, environmental conditions, impact damage such as bird strike, hail damage, tool damage during maintenance, debris attacks and furthermore. Hence diagnostics of such damage and prognostics of the remaining strength and life of the material due to the damage is vital, and it needs to be done in real-time. Non-destructive testing (NDT) is the process of inspecting the composite material from time to time, without damaging the structure. Structural health monitoring (SHM) is an advanced process of NDT, where the sensors are implemented to identify deviations from normal to a healthy state. Prognostics health management (PHM) is the method of using SHM techniques to predict the damage development and the remaining strength and life of the composite material. Based on the recent advancements of Machine learning in other scientific domains, Machine learning has been integrated into these fields of composite materials for significant improvements.
The data learning and analysis methods represented by machine learning are one of the current research hotspots. This project uses machine learning as the main method to learn fault characteristics of bonded composite materials. This research mainly focuse the following aspects. First, analyze the detection signal (using PZT or AE detection structure signal), extract the fault characteristics, analyze the fault characteristics preliminary, calculate the feature matrix, then use the supervised machine learning algorithm, and finally perform the aggregation Class performance evaluation. At the same time, the project is also preparing to study the more popular structural detection algorithms in recent years and try to design a fault recognition algorithm applied in the field of structural health detection. The expected result is to achieve a high fault recognition rate, and the algorithm has a high universality.