Predictive Maintenance Is Only as Good as the System Behind It

A Case for Managed SAP in Canadian Mining

Mine fleets throughout Canada have invested much of the past decade in outfitting their fleets. Mill and conveyor drive condition monitoring, haul truck telematics, and oil analysis that forecasts bearing failures a few weeks in advance. For all practical purposes, detection is no longer an issue. The challenge yet to be addressed at most facilities is what happens in the few hours following an alarm fire – is it an open work order, a reserved component, a scheduled crew, or just an e-mail opened on Monday?

This disconnect goes unnoticed because it does not relate to the kind of technology challenges that the industry tends to focus on. It relates to a systems challenge, and one of the reasons why predictive maintenance implementations are successful in a few pilot units but do not show any improvement in a reliability metric on a full scale.

Prediction is just a trigger; it does not result in anything.

The benefit comes when the planner has enough time to source the parts, arrange for the team, and move the equipment to carry out his plan at his own pace. All of the processes in between are enterprise processes: notification, planning, procurement, scheduling, execution, and historical recording.

This process lives in the ERP – for most Canadian mining companies, in SAP Plant Maintenance or SAP S/4HANA Asset Management, with an increasing use of SAP Asset Performance Management. The level of sophistication of the analytics process depends on the condition of the system it is being handed over to.

Where the chain breaks

The pain points occur at the same points across sites.

• Asset master data. Hierarchies for functional locations were established many years ago and have evolved organically as equipment has been added. If the prediction has to go to the right team, the asset needs to be described singly.
• External alerts. When the condition monitoring platform manages its own dashboard, someone monitoring the system will act.
• Manual handoffs. The prediction should lead to notification, work order generation, and allocation of spare parts from inventory. This process is manual and works as fast as the person managing it.
• Planning parts ignoring predictions. Reordering points based on historical part consumption will not allow planning for an item that failed to meet the prediction.
•  Scheduling does not consider the workforce. The availability of the workforce, certifications, and contracts influences how work can be executed. If planning ignores those factors, a work order is created and postponed.
• Lack of feedback. Confirmed failures and parts used become an input for tuning predictions. Work orders with minimal information make analytics less effective.
• Integration fragility. Historian/IoT-to-SAP connection is often the least monitored in the entire integration flow.

Why does the Canadian context raise the cost?

These weaknesses affect the situation in different ways depending on the asset’s geographical location. A shortage of parts causing a plant located close to a highway to lose half a shift will cost a fly-in outfit working in the territories one whole week and one charter. Outfits operating in remote sites, including northern Ontario, the Abitibi belt, Saskatchewan potash country, and northern British Columbia, plan their logistics schedules. The standard maintenance guidelines are never intended for such circumstances.

Climate makes matters worse, as failure modes vary with seasonal changes in lubricant viscosity, hydraulic reactions, and material characteristics. Skilled trades are not always available either, due to competition for personnel engaged elsewhere. Furthermore, any catastrophic failures are safety reportable incidents under provincial mines health and safety legislation, not just lost production.

What a managed model change

The critical difference, however, isn’t between supported and unsupported systems, but between a system that is set up once and one that is maintained over time as the operation evolves around it.

• Mastery of data governance, as a continuous practice, not an audit-time fix-up.
• Maintained integrations so that a failed feed between condition monitoring and SAP is discovered within hours.
• Maintained change management through S/4HANA and APM releases so that the chain from notification to work order stays intact through them.
• Alignment of coverage by shift rather than by office hours, and by the actual time zone where the operation operates.

Questions worth asking

• From the point when a prediction alarm goes off, what number of manual actions are involved before a work order is generated?
• Does the planner know at that very second when the alarm goes off if the spare needed is available on-site?
• What is the consistency of the same class of assets being tracked at all operational locations?
• How many closed work orders have sufficient failure information for the model to be improved?
• Has anything broken within the maintenance cycle as a result of the last big upgrade of SAP, and how long did it take to realize this fact?

The system behind the signal

While predictive maintenance has now surpassed the stage where accuracy is the limiting factor, what dictates relevance is the condition of the enterprise system to which it speaks and how well that system is maintained. For an operation with equipment operating in tough conditions, far from assistance, that becomes the limit of what can be gained through investments in reliability.

Through the Digital Asset Reliability Management service from YASH Technologies, SAP Asset Performance Management, SAP Service & Asset Manager, and managed application support services are integrated, such that predictive failures are translated into planned maintenance work. If you would like to evaluate your organization’s weaknesses for this purpose, contact the YASH SAP Services team at info@yash.com

• Predictive Maintenance
• SAP
• SAP Managed Services