Co-authors​ :

 Ameni RAGOUBI (FRANCE), Jean-Baptiste CASIMIR , Julien BROCAIL (FRANCE), Patrick DEWAILLY (FRANCE), Ronan LE GOFF , Alban AGAZZI  

In recent years, the integration of the digital twin has emerged as a promising strategy to enhance structural health monitoring (SHM) capabilities. By simulating real-time structural behavior, the digital twin strengthens monitoring capabilities by offering more accurate predictions, thereby further optimizing the reliability of composite materials in the aerospace industry. In the context of my research, I employ an innovative approach by integrating the Piezoelectric Macro-Fiber Composite (MFC) into the composite structure to excite the structure through vibrations and record its response. The primary aim of our digital twin is to assess the structure's post-impact lifespan. The methodology relies on two finite element models: the first simulates various damage scenarios, while the second models wave propagation to record the signature of each post-damage scenario. Each damage scenario generates a signature, which is then transformed into a wavelet, constituting a database of time-frequency responses (Figure 1). The physical model will also be excited in the same way as the numerical model. Afterward, the signature of the structure will be transformed into wavelets. With each excitation, a Continuous Wavelet Transform (CWT) is obtained, which will then be compared to the CWTs in the database generated by the digital twin. Machine learning algorithms will handle the comparison of the CWT images. If an image from the physical model matches a scenario in the database, we can trace it back to the associated damage model and determine the location and degree of the impact. To ensure the reliability of the models, they will be validated through experimental tests. As seen in Figure 2, the numerical damage model implemented in Abaqus, based on the Hashin criterion and cohesive contact, has been validated against experimental tests. Figures 3a and 3b show the validation of the numerical model of the wave propagation model against experimental tests. The Structural Similarity Index (SSI) was employed to compare the two images. For the signature of the healthy structure, the SSI is equal to 0.94 between the physical model and the digital model. This value indicates a high degree of similarity between the two images. In future work, we plan to generate various scenarios to create a comprehensive database and link this database with data from sensors integrated into the composite structure using machine learning algorithms. Overall, the integration of the digital twin and machine learning algorithms has great potential to improve SHM capabilities, especially in the aerospace industry, where the reliability and safety of composite materials are of utmost importance