Digital twins are a transformative technology for monitoring complex systems and providing critical insights into inaccessible machine components. Conventional monitoring of internal machine parts can be inefficient, as it often requires machine downtime and a partial disassembly to enable visual inspection. Minimizing machine downtime while avoiding catastrophic failure due to wear or damage of internal components motivates the need for a non-invasive, yet reliable monitoring method. Consequently, a predictive maintenance framework to provide a non-invasive prediction of Remaining Useful Life (RUL) of critical internal machine components is needed. To address this need, we present an adaptable, physics-based simulator as part of a multiphysics digital twin for the case of a scraped surface heat exchanger used in industrial ice cream production. The digital twin integrates real-time process data with physics-based surrogate models, enabling real-time prediction of wear and RUL for critical components. High-fidelity Computational Fluid Dynamics (CFD) and Finite Element Analysis (FEA) simulations were conducted to determine the loads acting on critical machine parts under various operating conditions, forming the basis of this methodology. The simulation results were condensed into lightweight surrogate models in the form of Functional Mock-up Units (FMUs), which allow for efficient, real-time load tracking. The simulator uses the FMUs to incrementally accumulate the wear occurring at every time step, which in turn is used to estimate the RUL. The presented methodology of physics-based predictive maintenance for inaccessible components significantly reduces downtime and maintenance costs, and improves product safety. Its modular design, enabled by the integration of FMUs into the digital twin simulator, ensures high adaptability and scalability, allowing for new, similar machine types to be integrated simply by exchanging FMUs. Due to the modular framework of the approach, it is also transferable to other complex machinery with similar hidden wear mechanisms.