Supplementary Materials: Lifecycle-Oriented Hybrid RL-Physics-Based Modelling for Predictive Monitoring in Fluid Machinery Systems (2)

This dataset represents the first systematic evidence synthesis examining the integration of physics-based models with reinforcement learning for predictive monitoring in fluid machinery systems. Through rigorous screening of 127 peer-reviewed papers published between 2020 and 2025, we have constructed a comprehensive evidence matrix that maps the landscape of hybrid RL-physics approaches across their complete lifecycle—from initial design constraints through training optimization to deployment validation. Our analysis reveals a concerning methodological gap: while the field has developed sophisticated training-phase integration mechanisms, including multi-fidelity surrogates, physics-informed neural networks, and differentiable physics operators, deployment-ready systems with calibrated uncertainty quantification remain remarkably sparse (only 6 of 127 papers report deployment-level validation). The dataset includes six meticulously curated supplementary tables documenting paper classifications, lifecycle categorizations, methodological coupling patterns, domain distributions, and evidence quality assessments, alongside validated analysis scripts that implement our novel integration scoring rubric. Our sensitivity analysis demonstrates robustness through ±20% weight perturbations, with 84% of top-tier papers maintaining their classifications. Each paper's assignment to tiers A (physics-monitoring coupling with deployment considerations), B (partial integration), or C (limited integration) is fully justified with explicit criteria and detailed justifications. Beyond mere cataloging, this collection provides actionable insights for practitioners through standardized reporting frameworks, identifies critical research bottlenecks in sim-to-real transfer protocols, and establishes baseline metrics for future meta-analyses. The reproducible screening protocols, inter-rater coding sheets, and complete decision audit trails ensure full transparency and enable community reuse, making this dataset an essential foundation for advancing deployable physics-informed monitoring systems from laboratory concepts to industrial reality.