In the context of industry 4.0, the predictive maintenance (PdM) approach plays an important role in the optimization of maintenance costs, as breakdowns and over-maintenance costs. Health supervision of production behaviors is one of the critical steps to implement PdM. To do this, specific Maintenance KPIs, called also Health Key Performance indicators (HKPI) must be identified. Furthermore, dealing with real data, finding the right HKPIs, and validating the final proposed solution by the industrial actors, are less seen in theoretical studies. In this paper, based on available historical and real-time data and information, some suitable HKPIs have been proposed, allowing to build the dashboards for monitoring the health state at machine and production system levels. Regarding the variety of real-time data to present in the dashboard, Principal Component Analysis (PCA) method has been applied, allowing to find the most appropriate features. At the production system level, HKPI aggregation and identification of decision levels are rarely studied. We propose also for each HKPI, a suitable decision-level, and an appropriated aggregation methodology. The proposed methodology has been applied to production equipment of the FPT Industrial Company and the results were appreciated by managers.