Unlocking potential with predictive maintenance

Optimum Energy developed a platform to remotely manage commercial A/C and HVAC installations, which optimizes the energy required to cool a building. For the last seven years, they have been collecting data about customer sites. However, they did not have the in-house knowledge to unlock the AI value found in this data. EXIT83 developed a proof of concept AI model that took two years' worth of telemetry data for a single customer, where the data was unlabeled and included over 300 unique variables. To understand how these variables were related to each other, we first developed a massive data correlation table to understand which unlabeled telemetry characteristics were related to each other. Finally, we trained a model utilizing these correlations to accurately predict when an installation would need servicing—ultimately preventing unnecessary downtime.