The methodology for prescriptive maintenance of complex technical systems is presented. The proposed methodology is based on a hybrid physics-based and data-driven modelling of complex systems. This approach integrates traditional physics-based simulation techniques such as finite-element modelling, finite-volume modelling, bond-graph modelling and data-driven models, with machine learning algorithms. Combined implementation of the both approaches results in the development of a set of reliable, fast and continuously updating models of technical systems applicable for predictive and prescriptive analytics. The methodology is demonstrated on the jet-engine power plant preventive maintenance case-study.