Partial discharge (PD) in electrical equipment, such as air switchgear, can lead to insulation degradation and, ultimately, equipment failure. Monitoring PD is crucial for preventing failures and ensuring the reliability of power systems. This study proposes an online monitoring method for PD in air switchgear based on characteristic gases generated by insulation defects. A novel gas detection system is developed that identifies gases released during partial discharge events. The system employs gas sensors to detect specific gases like acetylene (C2H2), methane (CH4), and ethylene (C2H4), which are associated with PD. The system's performance was evaluated in both laboratory conditions and in-field testing, showing that it successfully detected PD events with high accuracy. The research demonstrates the potential for real-time, non-invasive monitoring of PD to improve the reliability and safety of electrical switchgear.

The rapid integration of machine learning into autonomous driving systems has fundamentally altered long-established assumptions about safety assurance, regulatory compliance, and public trust in automotive technologies. Unlike traditional rule-based automotive control systems, machine learning-enabled autonomous driving systems are adaptive, probabilistic, and context-sensitive, challenging both technical validation practices and regulatory frameworks that were designed for deterministic behavior. This research article offers an in-depth qualitative and normative analysis of how safety assurance, standards, and regulatory approaches interact in the governance of adaptive autonomous driving systems. Drawing strictly on established international standards, regulatory theory, and peer-reviewed legal and safety research, the study explores the structural tension between innovation and accountability, the role of standards as trust-building instruments, and the evolving relationship between rules-based and goals-based regulation. The methodology adopts a comprehensive interpretive analysis of regulatory typologies, international standardization frameworks, and qualitative insights from prior empirical studies, synthesizing them into an integrated conceptual model of autonomous vehicle governance. The results reveal that existing standards and regulatory approaches provide partial but insufficient mechanisms to address learning-enabled behavior, particularly in post-deployment adaptation and system evolution. The discussion elaborates on the implications for institutional trust, liability allocation, and safety culture, while identifying critical limitations in current assurance practices. The article concludes that a hybrid governance approach—combining enforceable standards, adaptive assurance arguments, and continuous oversight—is essential for the safe and socially legitimate deployment of machine learning-based autonomous driving systems.