Maintenance is essential for the reliability, availability, and safety of a railway network, which is composed of various infrastructures like tracks, tunnels, stations, switches, overhead wiring, signaling systems, and safety control systems. In this paper we focus on track maintenance, which in general takes up a large portion of the annual maintenance budget of a railway infrastructure network, e.g., 40% for the Dutch railway network [1]. As shown in Fig. 1, a railway track contains different assets, e.g., rails, ballasts, sleepers, fastenings, welds, etc., that are interconnected and work together. These assets suffer from quality degradation over time due to regular usage. For example, the contact between wheel and rail leads to squats, a typical rolling contact fatigue that first appears on the rail surface and might cause rail breakage if not treated properly [2]. Early-stage squats can be effectively treated by grinding, while late-stage squats can only be addressed by rail replacement [3]. Due to the high cost of railway track maintenance interventions (e.g., over EUR 10,000 for one grinding operation), and the limited resource for track maintenance (e.g., limited track possession time for maintenance), how to plan maintenance interventions in a cost-efficient way without sacrificing the safety and reliability of the whole network has become a primary concern for railway infrastructure managers. This explains why most European countries have started a shift from reactive maintenance to proactive maintenance in recent years [4, 5]. Conditionbased maintenance [6, 7], where maintenance interventions are planned based on