Development of an ultrasonic NDE&T tool for yield detection in steel structures

Nondestructive Evaluation and Testing (NDE&T) is a commonly used and rapidly growing field that offers successful solutions for health assessment of structures. NDE&T methods have gained increasing attention in the last few decades especially with the contribution of the advancements in computer and instrumentation technologies. The applications of numerous NDE&T methods in civil engineering mostly focus on material characterization and defect detection. Techniques for nondestructively identifying the stress state in materials, on the other hand, mostly rely on the Theory of Acoustoelasticity. However, the sensitivity and the accuracy of acoustoelasticity are affected by several factors such as the microstructure of the material, temperature conditions, and the type, propagation and polarization directions of the signals used. This dissertation presents the results of an experimental study that investigates the changes in the characteristics of ultrasonic signals due to the applied stresses. Using a specially built testing system, ultrasonic signals were acquired from four different groups of steel specimens subjected to uniaxial tension below and above the yield stress of the material. The experimental database was first analyzed in terms of the acoustoelastic theory. Then, well known Digital Signal Processing (DSP) methods were used to calculate a total of seven time and frequency domain characteristics of the first three echoes of the acquired signals. The investigated time domain parameters were the peak positive amplitudes and the signal energies of the echoes, while the peak amplitude of the Fast Fourier and Chirp-Z Transforms, peak and peak-to-peak amplitudes and the root mean square of the Wavelet coefficients were used for the spectral analyses. Even though the acoustoelastic effects can be very small for certain measurement cases and they can be influenced by several other factors, clear distinctions between prior to and post yielding were observed for all investigated time and frequency domain parameters. The results were further analyzed with statistical methods and Receiver Operating Characteristics (ROC) curves in order to investigate the potential of the presented study for being used as a nondestructive testing tool for yield detection in steel structures.