Prescriptive maintenance: An AI method to solve faults – By Aditya Abeysinghe

Maintenance of systems is a method to avoid issues and make systems up to date and effective. Maintenance is carried out in digital systems using digital components that notify when maintenance is needed and where maintenance is needed. Different methods of maintenance in digital systems are used. AI (Artificial Intelligence) models are used in maintenance methods to predict and suggest when maintenance is required and what needs to be performed to avoid issues.

Predictive Maintenance

Planned maintenance of a system includes maintenance after a certain milestone or event. For example, maintenance of a machine may be done after a certain period since the last maintenance. This type includes a condition on which a system should be maintained. With this maintenance method, systems are repaired when the event to be maintained is planned even when there is no issue. Therefore, this type of maintenance is often not effective for most systems which have large volume of users.

Predictive maintenance uses certain rules to maintain a system. It is more effective compared to planned maintenance as maintenance is not performed after a certain time or a condition and a system is repaired before a failure. It is performed only when an event for maintenance is detected. An alert that is displayed to a driver on a vehicle which has overheated is an event to prevent heating. Maintenance in prediction to a certain rule will also often not include downtime of systems in contrast to planned maintenance.

Prescriptive Maintenance

Prescriptive maintenance is maintenance which uses predictive method for maintenance and suggests which action to be used to avoid issues in a system. Prescriptive method is similar to advices or ideas to follow to avoid issues when using the predictive method. Prescriptive maintenance may be used when a system is processing sensitive and large quantities of user data.  

Data used

Planned maintenance uses data which calculate the level of completion of a precondition. Therefore, data are periodically captured and then compared against this condition and if the condition has been met the condition is notified to the user. AI is rarely used in this type of maintenance except for maintenance of systems which use heavy data processing.

Predictive maintenance and prescriptive maintenance also use historical data. Also, data from various sensors, datasets and other sources are used for analyses. Data obtained are used in AI-based processes and predictive processes to make insights. Predictive maintenance uses AI models only to predict when maintenance is needed and prescriptive maintenance uses AI models to provide recommendations on analyzed data on how to perform maintenance.

Image Courtesy: https://www.plantservices.com/