<p>Unexpected failures of automatic mechanical presses can lead to severe production disruptions, particularly when critical transmission subcomponents such as clutches degrade unpredictably. This paper proposes a Maintenance 4.0 oriented hybrid Remaining Useful Life estimation framework that integrates physics based reliability modeling with vibration driven machine learning to support predictive maintenance of automatic press clutches. Unlike conventional system level prognostic studies, this work focuses on a rarely investigated yet critical subcomponent and formulates a subcomponent level hybrid prognostic strategy grounded in both mechanical degradation principles and real time sensor analytics. The physics based model characterizes clutch degradation through reliability and failure rate evolution, while the data driven model exploits vibration features extracted from IoT enabled sensor and processed using an ensemble learning strategy. The hybridization of both approaches enables adaptive and physically consistent RUL estimation under variable operating conditions. Experimental investigations illustrate the complementarity of both models and highlight the benefit of hybrid prognostics for connected, predictive, and proactive maintenance strategies. The proposed framework supports Maintenance 4.0 by enabling real time health assessment and intelligent decision support for industrial press systems. Quantitative evaluation shows that the proposed hybrid model reduces RUL prediction error by approximately 50% compared to the data driven model alone, achieving an RMSE of 52 cycles over the evaluated degradation horizon.</p>