Edge blurring suppressing in ultrasonic reflection coefficient measurements for contact state assessments with a PSO-BP neural network approach

Abstract The reflection coefficient of ultrasonic waves serves as an essential parameter for assessing the contact condition of mechanical interfaces. However, edge blurring effects that may occur during the detection process can significantly increase measurement errors of contact state (contact stress, contact stiffness, etc). This study addresses the issue of edge blurring in ultrasonic measurement by developing a signal prediction and separation technique that integrates particle swarm optimization with a back propagation neural network model. The model’s prediction precision was verified through finite element simulation, and interference signals at different measurement positions were effectively predicted. The effective signals were accurately extracted to determine the reflection coefficient and evaluate the model’s effectiveness in minimizing edge blurring. The simulation results indicated that this approach effectively lowered the discrepancy in the reflection coefficient, reducing it from 213% to under 6.5%. Further experimental validation confirmed that, after suppression, the discrepancy was reduced from 50% to below 13%, effectively proving the theoretical and practical feasibility of this technique in edge blurring suppression under various conditions. The method has the potential to enhance the accuracy and reliability of ultrasonic detection at boundaries while improving the measurement quality and accuracy of ultrasonic imaging technology in various fields.