Beyond the buzz: How IIoT helps manage reliability
The Industrial Internet of Things (IIoT) is not just a buzzword anymore; it’s fully in action, with a wide range of benefits across so many different industries. Nowhere are these benefits clearer than on the maintenance side of the business. The focus had been preventive maintenance with a reliance on a calendar to conduct an equipment check. Advanced analytics, and in particular machine learning, is spinning that concept on its head, turning maintenance into an actual value creator.
Why is that so powerful? When a piece of equipment or an asset malfunctions or breaks down, even the smallest of such scenarios can put an entire plant into an unplanned production shutdown. Shutdowns can cost organizations millions of dollars in lost revenue over the course of a year. Preventive maintenance, while important, unfortunately cannot accurately predict the future, so unexpected equipment breakdowns occur no matter how persistent the scheduled maintenance.
According to ARC Advisory Group, 82% of all failures cannot be detected or avoided with traditional maintenance approaches. Additionally, a Boeing study found that 85% of equipment failed at random, regardless of how much that equipment was inspected and serviced. Traditional maintenance can fail to achieve what IIoT technologies such as machine learning are poised to do: create a completely reliable plant environment and a world that doesn’t break down.
Why predictive is the future
Unlike preventive maintenance, predictive maintenance powered by machine learning can recognize degradation patterns and predict impending failures. Such information improves equipment availability, limiting downtime and producing the most prolific improvements in overall equipment effectiveness (OEE), assuring manufacturers get the highest lifetime return on capital expended. Industrialized machine learning provides real-time insight into the health of plant assets, determining which assets run the risk of malfunction or degradation.
Predictive analytics and machine learning software achieves this using digital software agents that run continuously. They distinguish between normal and abnormal equipment and process behavior by recognizing complex data patterns that reveal the precise signatures of degradation and failure.
The tasks digital agents perform are two-fold. One type works to detect anomalies, deviations from an otherwise normal behavior pattern. The second detects actual failure patterns, identifying behaviors that result in specific failures. Upon detecting unusual behavior or failure patterns, machine learning-powered agents not only alert weeks, sometimes even months, in advance of failure, but also prescribe specific corrective action.
The outcomes evade downtime, fix small problems that could cause larger problems, lessen the blow of the costs of downtime and provide operators and maintenance staff with trust in failure prediction, giving them the time to remove the root causes of asset failure. Such an automated approach, embedding and abstracting the details of machine learning techniques, is significantly more effective and accurate than other systems that exist today in maintenance. Furthermore, machine learning alone will not get organizations across the finish line. How machine learning is used and how it’s applied is what really makes or breaks digital transformation initiatives in maintenance.
At Saras, a European oil refinery, machine learning-powered analytics predicted equipment failures 45 days in advance of an impending failure. Another company, Borealis, a polymers producer, saw four weeks advanced warning of hyper-compressor failures with the predictive approach. Both sites showed quite significant leads ahead of potentially major unplanned downtime events, allowing time to avoid the issues or to safely plan the restoration.
The biggest value-add of such advanced warnings is that industrial companies operating in plant environments are no longer victims to production losses, which can, as mentioned, lead to millions in dollars of lost revenue and profits. Both Saras and Borealis are examples of the IIoT at work, where early alerts can ensure unplanned downtime doesn’t occur when operators and maintenance professionals appropriately respond to alerts and prescribed action.
Such reliability assurance within a plant environment saves revenue losses and allows plant operators to run machines to their limits of performance. In turn, this increases profitability without a higher risk of equipment breakdown. Predictive maintenance is the superior form of maintenance that truly helps companies make a profit.
Making it work
Here’s a tough pill to swallow: where 50% or more of all data warehouse projects failed, history may repeat itself with failures of IIoT projects. Why? Many companies embark on digital transformation initiatives with great enthusiasm for the technology, but technology alone is never enough to move the needle. So, the platform arrives, and nobody know what they are going to do with it.
The one thing lacking to really find technology success is a clear business objective, a pain point needing a solution. To be fruitful in their efforts to transform, businesses must first identify what they want to solve before applying technology to their operations. It’s important not to get swept up in the excitement over a new technology, such as machine learning, only to be left clueless about how to use it to the best business advantage.
Only when an organization identifies a key challenge can it move on to determine the appropriate product that will best solve the problem quickly, analyzing data and automating action. Regarding maintenance, plenty of contemporary tools offer help, but if the business issue at hand is breakdowns, then predictive, autonomous solutions that are scalable and supportable are the best bet because of the superior plant reliability they offer organizations. However, industrial data is no simple equation with an easy answer. Upon determining the problem needing a solution, organizations must ask themselves what data they need, the data rate requirements, how the data will be collected and authenticated and, finally, determine any inaccuracies or false signals that must be removed.
Beyond identifying a specific use case for IIoT in maintenance, the intrinsic corporate culture must drive IIoT initiatives forward.
Breaking down existing business unit silos, especially between maintenance and production operations will encourage these teams to work alongside each other with one larger, common goal of operational excellence. Common shared goals can champion the cause and drive better business outcomes including: machines stop breaking and last longer, net product output increases and maintenance costs go down. Another great organization-centric initiative is to identify change agents in the employee base. Who is someone who can drive and see-through a predictive maintenance overhaul? How will this person get organization-wide interest in this approach? How does he or she get different teams to understand the value-add of new IIoT technologies? Such are questions that leaders of change are responsible for answering.
To do maintenance “the right way” and to actually glean its hidden value, a number of factors ultimately must align. In part, an aspirational belief in new, truly data-driven IIoT technologies is required. Also, there must be an organization-wide approach on how to best implement the technology. Only when those two things come together will organizations be on the right path toward digital transformation in maintenance.