Experience, not data, drives predictive maintenance & operational efficiency

Sketch of predictive maintenance keywords drawn by a hand on white

Predictive maintenance models will automate facilities experience

The Smart Building says…the pump is not running.  Is this a problem?  Your Smart Building should know and act if required.

It’s easy to put sensors on pumps, compressors, fans, air handling units, boilers or chillers, but no amount of data will allow the building (or operator) to make the right decision, without suitable “experience”.

Experienced facilities operators  analyze incoming  data based on their own knowledge, expertise, and most importantly, past experiences.  Smart Buildings need to capture this same experience.

Today, few Smart Building CVs simply can match the skills of an experienced operator.

Transform business processes to capture better experiential data

Current building automation systems fail to capture it.   CMMS systems, even if diligently maintained, don’t get it either.

Experiential data can only be captured at the point-of-service, where people do the work, to change buildings.  As these folks observe and act on what they see, digital service processes will capture the details of this service intervention, including asset condition, costs, and action details.

True predictive operations and maintenance need a lot of experiential data, generated by knowledgeable people, with equipment or system specific condition, and outcome details.    This is the critical data required to train a predictive model.

When ingested by machine learning, the industry rule of thumb is 10,000 records about a specific issue to allow machines to find patterns.   That’s 10,000 fan service outcomes, 10,000 pump service outcomes, and 10,000 boiler outcomes.  Minimum.

Note: Microsoft Azure AI provides a good primer on AI & predictive maintenance found here.  The Microsoft authors go far deeper into business and uses cases, and the related data science.  I’ve highlighted some of their key points below. 

Predictive maintenance needs to capture both good and bad outcomes

Microsoft, in describing the data needs for predictive maintenance training,  puts that very simply:

The problem should have a record of the operational history of the equipment that contains both good and bad outcomes.  (Microsoft)

Capture financial outcomes with integrated digital business processes

Microsoft then goes on to say that the relevant data sources for predictive maintenance include, but are not limited to:

  • Failure history
  • Maintenance/repair history
  • Machine operating conditions
  • Equipment metadata

In buildings, people matter more

Buildings ultimately provide services to people…people are also the primary source of operating  costs (in both what they ask for, and what services they provide).  If the goal is to manage operating costs, you need to measure operating costs in granular detail.

Machine learning will optimize Smart Building services

Machine learning algorithms will use this data to develop a predictive model that, if successful, will anticipate the need for operational changes, with a goal to reduce operating costs, and risk.

The success of any learning depends on (a) the quality of what is being taught, and (b) the ability of the learner. Predictive models learn patterns from historical data, and predict future outcomes with certain probability based on these observed patterns. A model’s predictive accuracy depends on the relevancy, sufficiency, and quality of the training and test data. The new data that is ‘scored’ using this model should have the same features and schema as the training/test data. The feature characteristics (type, density, distribution, and so on) of new data should match that of the training and test data sets. The focus of this section is on such data requirements. (Microsoft)

Machine learning will find patterns and determine probabilities of different outcomes, enhancing visibility to potential issues, effectively drawing on past experience to optimize future service and maintenance costs.

Replace your CMMS with digital twins hosted integrated service processes

If reactive CMMS maintenance data is insufficient for predictive analytics, how can you collect better data?

By now you may have heard about digital twins.   This is the data technology evolving to allow the capture of  knowledge and expertise of people, through integrated business processes.  The digital twin, a simple digital model of each physical asset, allows facilities staff and service contractors to interact with the Smart Building to efficiently provide services, and capture each new service change to the equipment or building.    The digital twin completely replaces the CMMS, with a far more efficient service process that reduces truck rolls, tool time at the equipment, and administrative costs in the back office.

If you are planning a Smart Buildings strategy, it’s time to rethink your traditional work order and maintenance management processes.     Here’s a preview of what that will look like.

What do your HVAC contractors really see on the rooftop?

Stay in touch.  Join our newsletter here.