Font Size: 

The present research focusses on estimating the elastic constants of orthotropic laminates using ultrasonic guided waves excited through piezo-electric transducer (PZT) and sensed using one-dimensional laser sensor. The elastic constants of a material are crucial for understanding its mechanical behavior and are typically determined through experimental testing. However, this process can be time-consuming and expensive. Thus, in this study, a machine learning model is trained using a time-series dataset pertaining to different elastic constants generated using the simulations. The model is trained using deep regression networks consisting of convolution and linear layers coded in python using Pytorch and designing a loss function based on mean absolute percentage error (MAPE). In the simulation model, PZT is excited using a tone-burst signal at a suitable excitation frequency where the fundamental lamb modes are seen well separated and dispersion effect is minimized. A total of 135 unique sets of simulations are conducted and fed to train the model that comprises of 70% for training and remaining for testing. Next, a random experimental test was conducted, and the corresponding elastic constants were estimated with a mean absolute percentage error (MAPE) of 11.54% and standard deviation of 7.226%. The results demonstrate that the machine learning models are capable of accurately predicting the elastic constants of a material, with a high level of accuracy. This approach has the potential to significantly reduce the time and cost associated with experimental testing and could have wide-ranging applications in materials science and engineering.

Guided waves, Non-destructive Testing, Ultrasonic Technology, Machine Learning, Composite elastic constants