Abstract
Active infrared thermography operates without contact, covers the full field from a single side, and can operate at a limited temperature rise, which makes it well suited to the inspection of polymer–matrix composites. However, its diagnostic performance is governed less by the detector hardware than by the mechanism through which heat is generated at, or transported to, damage. This review compares optically and mechanically excited thermography from that standpoint. The underlying material physics is summarized first: diffusive transport and the loss of contrast with damage depth for externally heated configurations, coupled thermoviscoelastic dissipation for internally heated ones, and the glass-transition constraint that limits admissible excitation power in polymers. Pulsed, lock-in, step-heating, pulsed-phase, and frequency-modulated modalities are then reviewed with the depth inversion relations they support. Mechanical excitation is treated in three branches: friction dominated ultrasonic thermography, local defect resonance as an excitation strategy, and self-heating based vibrothermography (SHVT) as a viscoelastic heat-generation mechanism. The reported detection capability of SHVT is then analyzed. Two limits emerge: one set by the size of the damage, the other by its position through the thickness. Finding damage and measuring its size also prove to be different capabilities with different limits. Signal processing is reviewed in three groups, namely classical sequence-based transforms, procedures for background suppression and quantitative characterization, and machine learning-based methods. Open issues in quantification, benchmarking, and data availability are noted, along with the limited use of machine learning-based vibrothermography processing.