Our TUBERS consortium is excited to announce its first project publication! The paper, “Enhancing Material Thickness Measurement: Ultrasonic Sensor Data Analysis and Thickness Prediction Using Neural Networks,” was published in the “Applications of Modelling and Simulation” journal in 024. Congratulations to the team from the AI Innovation Centre of the University of Essex, specifically to the authors Vahid Hassani, Antonis Porichis, Farhan Mahmood, and Panagiotis Chatzakos.

In this study, neural networks were used to enhance the accuracy of thickness measurements for immersed steel samples. Training data was gathered by conducting experiments on immersed wedge samples with varying thicknesses using the A-scan method. This data was then used to train a single-layer neural network. The performance of the trained neural network was evaluated using test data from different samples with various thicknesses. The study demonstrated a promising methodology for accurate thickness prediction using neural networks, with outcomes showing good agreement when predicting void locations in similar materials. The method achieved an error of less than 3% for thickness prediction and less than 7% for void detection.

The publication of our research findings marks a significant milestone for the TUBERS consortium. Our team is dedicated to ongoing collaboration and knowledge sharing within the scientific and technological community, and we are eager to see the potential impact of our research in the field of ultrasonic sensors and material analysis.

Access the full publication here.