Operationalizing AI for Predictive Maintenance: Best Practices Today’s Reliability and Maintenance Practitioners Need to Know

How are leading industrial companies developing and prioritizing strategies that are proven to enhance equipment reliability and maintenance with greater predictability? Here are the top three lessons you should learn.

The world of reliability and maintenance is rapidly changing. But it’s evolving in ways you might not expect.

A lot is written about how new technologies like AI will fundamentally change the future of work for industrial sectors. But the reality is that those innovations are already making an impact in very practical ways on front-line operations. People are becoming increasingly empowered to do their best work.

I recently attended the 2019 Reliability Conference (TRC) where around 800 industry professionals gathered to share the latest breakthroughs in asset management, reliability engineering, maintenance and condition monitoring.

Over the course of five days, many best practices were shared. Here are the key takeaways you need to know right now.

The Top Three Reliability and Maintenance Lessons You Should Learn Right Now:

1) Sure, AI is hot. But it’s the people — specifically subject matter experts (SMEs) — who are the catalyst of successful AI initiatives. They are the ones who add the necessary context to operationalize AI for the front lines. They’re more important to the technology process now than ever before.

The insights generated by AI are only as good as the data we feed it with. For that reason, it’s crucial to train AI models with relevant, high-quality data.

Not only do SMEs need to be involved with identifying the right data sets for AI, but the institutional wisdom they possess should ideally be digitized and entered into a database that can then be used to train and hone AI models.

A digital library of SME knowledge is the secret sauce that lets machine learning work its magic by producing insights of actual value that can be operationalized on the front lines.

2) The move to automation is happening, but it’s an evolution. Many industrial operations still find value in creating asset registries, assessing the criticality of potential failure modes and fine-tuning their maintenance strategies to balance cost and risk.

Though news stories about AI make it seem like science fiction, AI is really manifesting itself at the operational level today. Professionals are already using it to tackle an age-old balancing act: Achieving the ideal level of preventive maintenance (PM) while avoiding overspending and without putting production at risk.

By using data-backed insights from AI, teams can better understand their operations and maintenance (O&M) costs and make better PM decisions faster and at scale. They can use AI recommendations to financially optimize their PM strategies and ensure they’re not under- or over-maintaining assets.

The truth is that AI is enabling workers to be even better at their jobs. That includes ensuring critical production assets are ready and able to deliver when they’re needed — all while de-risking and cost-optimizing the maintenance strategies behind them, which increases reliability and reduces unplanned downtime.

3) It’s not enough to just turn data into insights. It is the right starting point. But insights aren’t actionable on their own, nor are they effective in a vacuum. In order for insights to truly be actionable and effective, they must be integrated directly into your workflow.

Here are some words to live by: If it’s not integrated, it’s not actionable.

It’s essential to connect AI model outputs to workflow systems like Computerized Maintenance Management Systems (CMMS) and Enterprise Asset Management (EAM) systems. This is the vital step that makes AI insights both actionable for workers and effective for the daily operations of the business.

Taking that a step further: There also needs to be a feedback loop between those systems — a learning process that continuously hones AI insights based on how effective they actually were in the real world. And that process isn’t just a one-time thing. It’s always evolving and never finished.

Why is this required? A method must be in place to ensure completed actions are linked back to AI models, so that they can learn from the delivered outcomes.

To see our tips for lowering your O&M spend without taking on more risk, check out our post: “Are Your Maintenance Strategies Financially Optimized?

READ MORE