Filippo G. Praticò, Rosario Fedele
Sensing systems and Artificial Intelligence (AI)-based data treatment are becoming increasingly important, fostering a new paradigm of predictive maintenance that enables timely maintenance to prevent costly breakdowns and downtime. This poses new challenges when authorities need to choose the optimal maintenance/renewal (M/R) strategy. Consequently, the objectives of this study were to investigate the optimal maintenance strategy under various hypothetical scenarios, including degradation curves and the cost of implementing predictive maintenance through sensing systems. To this end, this study analysed actual M/R costs, established theoretical degradation curves, and inferred predictive maintenance economics. The methodological novelty lies in setting up an objective function (present value) that has an intrinsic and quantitative relationship with the mortality curve (i.e. cumulative defect occurrence), implying that the solution (maintenance time and type) is tailored based on the characteristics of the given family of degradation curves (convexity). This allows for a rationale-based choice of the maintenance strategy, where ‘more frequently’ is not necessarily ‘better’. Notably, the quantitative and parametric model allows for assessing when Sensing/AI costs are convenient, resulting in the outcome that this occurs when the cost of the sensing system per year and km is about 3%–9% of the construction cost per km.