Integrity of in-service engineering structures is prone to fatigue damage over their lifespan. Majority of the currently existing elastic-wave-based damage identification techniques have been developed and validated for damage at macroscopic levels, by canvassing linear properties of elastic waves such as attenuation, transmission, reflection and mode conversion. However the real damage in engineering structures often initiates from fatigue crack, presenting highly nonlinear characteristics under cyclic loads. It is of great significance but vast challenge to detect fatigue damage of small dimension at its initial stage. In this study, traditional elastic-wave-based damage identification techniques were first employed with an attempt to detect fatigue crack initiated from a notch in an aluminium plate with the assistance of a signal correlation analysis, to observe the deficiency of the approach. Then the higher-order harmonic wave generation was used to exploit the nonlinear characteristics of acousto-ultrasonic waves (Lamb waves), whereby the fatigue damage was characterised. Results show that nonlinear characteristics of acousto-ultrasonic waves can facilitate more effective detection of fatigue damage than linear signal features such as wave reflection, transmission or mode conversion.© (2011) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.