Purely data-driven models frequently miss failures because they ignore physical reality. This project bridges that gap by combining machine learning with mechanical domain expertise, delivering a high-fidelity asset tracking pipeline that respects physical variables like thermal limits, stress, and vibrational frequencies.