Predictive Analytics Vs Prescriptive Analytics - BI maturity

We have been doing product research on statistical analysis techniques such as regression analysis to examine Asset downtime for the PEMAC Assets products.

This examines historical events of Equipment downtime and spots trends and patterns to predict a likely failure before it happens. This can be using IoT feeds from equipment such as runtime hours and diagnostic codes as well as previous breakdown history and mean time to failure (MTTF).

The use of open source tools such as the R language for statistical modelling has a steep learning curve but can produce some interesting insights.

This article from Industrial IoT was interesting because it extends the concept of not just surfacing a potential issue but also "prescriptively" telling you what to do about it.




So, the predictive analytics would identify an issue such as potential Downtime event for an Asset.

Rather than just highlight the issue, "Prescriptive Analytics" tell you what the action should be too… i.e.
·         Create a work order for an engineer to proactively inspect the Asset
·         Move the next planned job for that asset to be sooner to Service it before a failure occurs
·         Shorten future preventative maintenance plans (Routines) intervals and criteria based on the new information

This Prescriptive element could be harder to do reliably but would build more operational intelligence into CMMS and EAM systems by prompting the Maintenance manager with a menu of potential actions and even better, when the choice is clear-cut and unambiguous, automate the actions to improve decision times.

This could extend the concept we have already built in the PEMAC Assets products for "real-time" maintenance where fault codes and error codes from IoT sensors (such as thermal imaging cameras and vibration sensors on production equipment)  can trigger Routines to automatically generate a "Work order" record with pre-defined Job Specification of actions and checks