Condition Assessment of Dissimilar Friction-Stir-Welded Joints Using Ultrasonic Lamb Waves

Abstract

Condition monitoring is increasingly being considered a necessity to ensure the safe use of our structures and to extend their lifetime. Lamb waves (LWs) are ultrasonic guided waves (GWs) that are widely used for damage detection, towards integration in structural health monitoring of mechanical, aerospace, and civil structures.

The growing mechanical and economical demands in modern systems and structures are forcing an inevitable need for joining dissimilar materials, thus creating the challenge of establishing a process to inspect and monitor dissimilar joints. Friction stir welding (FSW) has emerged as one of the most promising and successful methods for producing sound dissimilar-metallic joints. This work is an extensive investigation of LWs’ potential for the nondestructive evaluation of dissimilar FSW.

As a prior step, the propagation behavior of the fundamental symmetric (S0) and anti-symmetric (A0) LW modes, upon interaction with welded joints of dissimilar materials, was investigated. A plate with an intact AA6061-T6/AZ31B dissimilar joint was employed, and the interaction of the propagating wave with the material interface was scrutinized numerically and validated experimentally. Plane-wave approximation (PWA) analytical solution was also adopted to investigate the behavior of the symmetric modes, and its performance was compared to the numerical and experimental results. The effect of the angle of incidence on the reflection, transmission, and mode conversion of the incident modes was analyzed. The transmission coefficients of the S0 and A0 modes were found to be almost constant until reaching very steep incidence angles (i > 78°). Further, the fundamental shear-horizontal (SH0) GW mode has evolved upon the interaction of the obliquely-incident S0 mode with the interface. The experimental results from an intact AA6061-T6/AZ31B FSW joint showed very good agreement with both the analytical and numerical ones. PWA was shown to be a very good approximation to determine the transmission and reflection coefficients of the in-plane symmetric modes for a certain frequency range. The importance of the obtained findings for the implementation of LW sensor networks in structures containing dissimilar-material joints was finally explained.

In the next step, a novel Bayesian-based framework for full damage identification was proposed. Simulated damage within a dissimilar-material joint was identified, in six parameters of damage extent and location (length, width, thickness, and x-, y-, and z- positions), using only one actuator and one sensor. Surrogate models that can predict LW sensor measurements, given the damage size and position, were developed. The surrogate models were based on artificial neural networks (ANNs) trained using finite element (FE) simulations of the monitored plate, with a pre-allocated sensor network. The ANNs were utilized to perform a statistical damage inference based on Approximate Bayesian Computation by Subset Simulation (ABC-SubSim). Data fusion for ABC-SubSim inference using multiple sensor measurements was successfully employed. The results showed that damages of different sizes and locations were identified with a high level of resolution and with quantified uncertainty. A precise and robust damage inference was achieved using a minimal sensing set-up based on one PZT actuator and two sensors. These findings are very promising for damage detection and assessment and form a step forward towards online/onboard monitoring applications.

Furthermore, the potential of LWs for the detection and evaluation of micro-scaled intermetallic compounds (IMCs) at the weld interface was examined. Intermetallic regions are common in welded joints of dissimilar materials, and their presence leads to weakness of the joint (due to their brittle behavior). FSW lap joints, between AA5052-H32 aluminum and ASTM 516-70 steel, with various intermetallic conditions, were simulated using the Murnaghan nonlinear-elastic model. The synchronism and non-zero power flux conditions between higher-order symmetric LW modes (S1 and S2) were used to determine the needed LW excitation frequency. Symmetric LWs were selectively generated and the collected signals were analyzed in the time, frequency, and time-frequency domains. It was found that the relative acoustic nonlinearity parameter varies linearly with the thickness of the IMC layer, where variations down to 2 μm are easily differentiated. The attained results prove the sensitivity of LW nonlinear features to microstructural variations, within dissimilar FSW joints, and demonstrates the capability of LWs in accurately scrutinizing their strength. Such novel findings would open the way towards the quantitative nondestructive assessment of intermetallic compounds using LW-based techniques.