In order to ease manual processes for users not to do the guess work of which asset might fail next, email notifications with the list of assets needing attention are distributed. These notifications are based on ML predictions that take into consideration historical information, IoT measures and other asset and maintenance attributes, which users would not have had time to review and analyze across all assets on the floor. The email will focus on a subset of assets that are at risk of failure within the next few days, which can be configured (1 or 2 weeks or a month) based on the business process.

User can navigate directly to the Birst reports in order to see predictions and additional insights generated by ML model. Predictions are color coded for users to be able to differentiate critical assets at the high risk in red from the rest. Failure probabilities output can be configured to longer or lesser timeframe depending on the business process. Other charts include but not limited to correlation between equipment age and failure probability, historical predictions vs. actual failures, top contributing features into ML predictions (examples being asset age, number of days since the last preventative and corrective maintenance, average failure interval, temperature deviation among other IoT features).

Infor OS Homepage displays the list of all assets sorted from the most to least critical. Widget includes details about the asset attributes such as age, manufacture, type, etc. Visual representation of predictions, days since last corrective and preventative maintenance helps users to relate to current asset state. In addition to that, user can see historical performance of the model to compare previous predictions with the actual failures. This helps to develop trust in the ML model to act on future predictions, enabling user to set an action for equipment inspection.

As data gets refreshed on the periodic basis with the latest IoT data or/and maintenance occurrences, asset attributes updates, etc. workflow is configured to run weekly or daily to automatically trigger the Coleman quests. The workflow includes a series of Coleman quests, including data loading and transformation, staging observations along with data cleansing, feature engineering, ML training process, evaluation against benchmark and accuracy computations.