This Engineering Research Initiation (ERI) grant will fund research that seeks to develop adaptive and explainable monitoring methods supporting real-time operational decision-making under uncertainty. Modern engineering systems, including manufacturing, energy infrastructure, and aerospace platforms, rely on continuous monitoring to operate safely and efficiently. Although these systems generate large volumes of data, identifying abnormal behavior remains difficult because operating conditions change over time. As a result, decision-makers face uncertainty when responding to potential system failures, leading to increased downtime and reduced operational efficiency. This project seeks to improve system reliability, reduce maintenance costs, and strengthen the resilience of critical infrastructure. By enabling more informed and timely decisions, this project contributes to national priorities in enhancing industrial productivity, infrastructure reliability, and workforce development. The project will also support student training in artificial intelligence and data-driven engineering, as well as disseminate open tools and educational materials to broaden societal impact. This project develops a unified framework for anomaly detection in multivariate time series data, dissolving the traditional boundaries between supervised and unsupervised learning, under changing operating conditions. The approach formulates anomaly detection as a reinforcement learning-based sequential decision-making process, where a learning agent identifies abnormal system behavior from streaming data. A hybrid reward structure enables consistent operation across different levels of supervision, reducing reliance on labeled data while maintaining detection performance. The project advances explainable artificial intelligence by formalizing Gradient-weighted Class Activation Mapping (Grad-CAM) for spatio-temporal attribution within a policy network, providing a new method to interpret sequent