Machine learning–assisted <i>in situ</i> corrosion monitoring: a review

Abstract As the critical first step in structural health management and fault diagnosis, corrosion monitoring is inherently multidisciplinary in nature. While conventional in situ techniques capture real-time electrical, vibrational, and thermal signatures, their effectiveness is constrained by limited detection precision, inefficient data analysis, and unreliable predictive modeling. The convergence of artificial intelligence (AI) and big data analytics has fundamentally transformed this field, generating considerable academic interest over the past decade. Machine learning (ML) – serving as the cornerstone of this revolution – excels not only in extracting nonlinear features from nonstationary processes but also employs probabilistic inference frameworks to quantify predictive uncertainty, thereby substantially augmenting in situ monitoring capabilities. This review systematically examines advancements in ML-assisted corrosion monitoring throughout the preceding decade, categorizing prevalent algorithms according to domain-specific implementations while evaluating enhanced in situ techniques through empirical case studies demonstrating superior data processing efficacy. Finally, we project future trajectories for intelligent monitoring technology in light of persistent challenges and emergent innovations.