With the emergence of Industry 4.0, smart systems, machine learning (ML) within artificial intelligence (AI), predictive maintenance approaches have been extensively applied in industries for handling the health status of industrial assets. Due to digital transformation towards Industry 4.0, computerized control, and communication networks, it is possible to collect massive amounts of operational and processes conditions data generated from several pieces of equipment and harvest data for making an automated fault detection and diagnosis with the aim to minimize downtime and increase utilization rate of the components and increase their remaining useful lives. Machine learning (ML) techniques have emerged as a promising tool in Predictive Maintenance applications for smart manufacturing. In the proposed work, we test the recent advancements of ML techniques widely applied to predictive maintenance to forecast natural gas regulators pressure deviations and to predict failures. The data represents about a hundred of gas regulators with monitored output pressure, temperature and flow, plus the observed failure mode ``pressure regulation out of specification'' (too high or too low), dates and durations of preventive maintenance over the last three years. ML and neural networks models are tested to forecast the output pressure of the regulators and to predict the passing over the failure thresholds. Defining the parameters that optimize the prediction of failure while limiting the spurious detections is a challenge that is investigated in the proposed paper. The deployment of such methods is expected to reduce the preventive and field maintenance operations costs by providing early warning notification and diagnosis of gas regulators issues.