How Digital Operations Analytics Will Predict The Future

The last thing any manufacturing engineer wants to hear is that production has shut down to await an equipment repair. It doesn’t happen often, thanks to the quality of modern machinery, but when it does it can cost thousands of dollars per second. That’s why manufacturers are investing heavily in assembly-line sensors that can keep tabs on vibration and other parameters to monitor shop-floor well being.

Manufacturers aren’t alone in the quest for technology uptime. As business becomes more digitized and connected, operations and other processes are more susceptible to shutdowns as well. The immediate cost penalties may not be as severe, but a breakdown in a key operations process can also have a devastating impact on anything from supply-chain planning to customer service.

The Internet of Things and big data, with their dual promise to transmit up-to-the-minute device and process performance to enterprise managers, appear to offer the solution. Yet the solution also raises a larger problem: how to make sense of all this new data in time to enact the fastest possible repair.

Enter Analytics

Enter real-time analytics, which, for good reason, sits today at the top of most every organization’s technology shopping list.

Today’s analytics toolsets are varied and wide-ranging. They can ingest massive data volumes in real time, combine new data with contextual models, extract patterns of anomalies and add to their knowledge as they go. The best ones can do all those things, and they can also learn to predict outcomes, to perform what’s called dynamic failure prediction.

Dynamic failure prediction, or predictive analytics, is a fast-growing phenomenon across many industries.

In manufacturing, for instance, predictive analytics can trigger preventive maintenance to catch problems before they occur, alerting maintenance, or even the original manufacturers, that service is needed. In logistics, predictive analytics can help supply chain planners route around unfavorable weather conditions. In customer support, predictive analytics can help determine potential product failures in time to alert customers.

Building Blocks

Dynamic failure prediction doesn’t happen overnight. It requires a specific set of analytics building blocks to set the foundation.

  • Operations-wide visualization – This creates a dashboard-accessible visual plane over processes, machines, customers and other aspects of the enterprise ecosystem. This is the first task for analytics – basically, connecting the parts to the whole;
  • Advanced anomaly detection – Once the parts are visible, analytics can find patterns of anomalies, even when they encompass multiple machines, processes, or levels of operation within processes;
  • Change management – The next step is to determine what processes, or customers, for instance, may be affected by these anomalies. Change management analytics can be used to identify target populations so they can be serviced proactively, in some cases before, rather than after, a major failure has occurred;
  • Incident life-cycle automation – Once the likely impact of a failure is determined, incident life-cycle automation takes change management to the next logical step by automating the path to incident resolution.

Assessing Risk

With these functions in place, and the relevant models and analytics algorithms populated and tuned, the infrastructure is prepared to assess risk, that is, the likelihood of failure. This is where the analysis gets tricky, because it has to apply what it knows already to future usage patterns, whether by smartphone customers, machines further down the assembly line or new network requirements – 5G, anyone?

By adding this information the analytics can predict the likelihood of failure. But this is simpler said than done, since it requires major processing horsepower to analyze all possible new behavior patterns. And the more complex the technologies analyzed, the more complex and various the behavior patterns.

The good news is that, being based on machine-learning models, the analytics will get smarter over time. As Deloitte says in Making Maintenance Smarter: Proactive Maintenance And The Digital Supply Network,

“…while PdM (predictive maintenance) depends on the accuracy of failure thresholds determined in a pilot program or review cycle, machine learning technologies improve these thresholds iteratively over time by analyzing the outcomes of each prediction and adjusting the thresholds accordingly.”

Next: Predictive Maintenance

Once the likeliness of a failure is known, the operations manager or production engineer can build an appropriate predictive maintenance schedule, using look-ahead analytics to determine if inventory is sufficient. In some cases, the analytics might kick off automated routines to head off process failures before they occur.

The impact of dynamic failure prediction can be immediate. In manufacturing or field service it can greatly reduce the costs of dispatching resources for repairs. In one example, a manufacturer was able to reduce process downtime by 70 percent by generating preventive maintenance an average of eight days before likely failures.

In customer service it can reduce the less tangible costs of lost or dissatisfied customers, a critical advantage in today’s highly competitive consumer markets.

Longer-term benefits are at least as compelling, especially for product manufacturers and distributors. With today’s constant flow of newer features and functions, product quality is becoming ever more important as a way to maintain lasting customer loyalty. Dynamic failure prediction can go a long way in helping to keep quality high, and customers loyal, in the face of rapid-fire feature announcements.

There’s a benefit, too, to those organizations now undergoing digital transformation. Long-term success will require top efficiency in business and operations processes. These processes are more connected than ever before, so a breakdown in one can affect multiple others. Dynamic failure prediction is a way to head off those breakdowns before they ever happen.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>