PERKEMBANGAN TEKNIK MACHINE LEARNING DALAM PEMELIHARAAN KESEHATAN STRUKTUR BANGUNAN

Abstract

The structural health of buildings is a critical aspect of infrastructure management, ensuring safety, resilience, and longevity against both natural and environmental threats. In recent years, the integration of machine learning (ML) into structural health monitoring has gained momentum, offering predictive and proactive approaches to building maintenance. This review article examines the latest developments in the application of ML for structural health management in Indonesia, focusing on methods, datasets, and practical implementations. Literature published between 2019 and 2024 was analyzed to highlight trends in crack detection, post-disaster damage prediction, and sensor-based monitoring. Prominent examples include the CRACKSAFE web application employing the YOLOv8 algorithm for wall crack detection, deep neural network models for post-earthquake damage assessment, and hybrid algorithms combining correlation analysis with partial least squares to optimize sensor usage. The findings reveal that ML enhances efficiency by reducing costs, enabling real-time monitoring, and supporting rapid decision-making for risk mitigation. However, several challenges remain, including limited availability of high-quality datasets, insufficient generalization of models under diverse environmental conditions, and minimal translation of research outcomes into practical policy or long-term field applications. This review underscores the need for multimodal data integration, real-time monitoring systems, and collaborative strategies among researchers, practitioners, and policymakers. Future directions should also include the adoption of digital twin and Building Information Modeling (BIM) concepts to advance intelligent, adaptive, and sustainable infrastructure maintenance in Indonesia.