The ability to predict the future is a powerful form of knowledge for any business. If you know an event that could cause harm to your business lies ahead, like equipment failure, you can take steps to address it and avoid negative consequences, like unnecessary costs or downtime. If you know a beneficial event is ahead, such as an increase in demand, you can harness it to increase revenue, customer satisfaction and more. Use cases like predictive failure, predictive (or forecasting) demand and predictive maintenance are quickly gaining popularity across a range of industries for these reasons. Individually, these IoT-enabled capabilities can pack a strong ROI punch by helping businesses predict the future using real-time data. When layered together, their benefit potential grows exponentially.Content Continues Below
Maximizing the uptime of revenue-generating equipment is critical for all industrial companies, which has made predictive failure one of the more desirable predictive capabilities. Failures can cost thousands of dollars or more in unplanned downtime and present the additional threat of expensive collateral damage. Additionally, false positive alerts from equipment sensors pose an altogether different problem by straining service resources and driving up expenses when technicians are deployed to find that no problem exists. Fortunately, sophisticated data analytics make it possible to predict and resolve these problems before they occur, virtually eliminating unplanned downtime and reducing false positives.
Take, for example, a pump failure at an oil refinery. Typically, there’s not a backup option, so the facility is forced into an unplanned shutdown. These sorts of environments are generally highly utilized — often running nonstop, 24/7. Such elevated levels of plant activity mean any production time lost to down equipment is gone forever, due to a shortage of extra capacity. And that’s in addition to the outsized costs associated with emergency repairs. Failures of this nature can also introduce contaminants into the production pipeline, which run the risk of corrupting product quality and damaging equipment downstream. Furthermore, these sorts of breakdowns come with the ever-present danger of causing a line breach, which can carry a steep cleanup price tag all on its own.
With data insights into the pump’s health, as well as like assets across a population, the refinery could have foreseen the upcoming failure. Knowing an asset is headed for trouble, management could have proactively scheduled a repair to minimize the impact on production and prevent unfavorable side effects, such as unnecessary cleanup costs and equipment damages.
Predictive (or forecasting) demand is another powerful capability for businesses. Itron, a leading provider of utility and smart city technologies, uses IoT to gauge demand for electricity. For the longest time, electricity demand patterns were static. Basic forecasting was possible by analyzing the number and density of homes against historical records. However, the introduction of electric cars, solar panels and more has complicated demand forecasting. This added complexity creates an issue as regulations dictate utilities must generate a certain amount of electricity to ensure they can meet consumer demand. Without predictive capabilities, utilities struggle to right-size energy generation, leading to wasted excess and unnecessary costs. Using IoT capabilities for edge devices, such as energy meters, Itron is able to harness real-time data to predict demand more accurately.
That said, demand data isn’t necessarily limited to information coming from a device. It’s easy to focus on device data because that’s the new variable, but there are more pieces to the puzzle. Companies experiencing the most predictive success typically incorporate additional variables, such as enterprise system data and public data sources (e.g., weather, geospatial, etc.), with device data. This sort of supplementary information helps build context to create a more complete picture of what’s actually happening with a device.
Whether it’s a wet versus dry environment or a cold versus warm climate, there are countless external factors that can affect different assets in a variety of ways, with varying degrees of severity. In fact, this secondary information is significant enough on its own that, in some cases, it can provide a general sense of what’s wrong with a device or when an error is likely to occur. But it takes an IoT system that integrates real-time device data with these additional data sources to pinpoint root causes and predict failures with accuracy.
Predictive maintenance is another IoT use case with exciting potential applications across the industrial sector. Traditional methods of equipment maintenance are generally reactive — servicing equipment once it fails — or preventative based on static time intervals or amount of use. Reactive maintenance is often expensive due to compounding costs associated with unplanned downtime and emergency repairs. Preventative measures carry the risk of under- or over-servicing equipment, which can inflate maintenance costs and reduce asset longevity.
Condition-based maintenance (CBM) is an important step in the right direction, as it uses real-time operational data, as well as environmental and historical information, to identify when service is truly needed based on actual device performance. While this does help industrial companies improve maintenance and repair processes across all of their connected equipment, CBM is still reactive in that it’s a response to current conditions. Predictive maintenance takes CBM and makes it proactive by flagging service needs in advance, so that companies can take action before those needs arise. In addition to significantly improving asset reliability and longevity, this ability to better plan maintenance can significantly lower service costs and help defer capital expenditures.
Predictive analytics is transforming many of today’s connected industrial organizations by proactively identifying asset efficiency, performance issues and other key factors before they impact operations. However, a lack of understanding surrounding a key enabling technology — machine learning — has led some businesses down a path of disappointment. While machine learning can recognize patterns in large volumes of data to help identify the different states of equipment operation, it cannot incorporate related factors or time values with those potential state changes — a key factor in predicting future events. This is why digital twins are so valuable. By combining the states identified through machine learning with other variables (time, environment, etc.) influencing those changes, digital twins create a model that makes predictive capabilities possible.
Keep in mind that even with the most advanced technology, the ability to carry out any sort of predictive analysis hinges primarily on having the right data set. So be sure to perform a thorough review of what data you have versus what’s required before beginning any initiative. I also typically recommend starting with the predictive capability that shows the most potential for profitability or cost reduction. Stakeholders deserve a return on investment and shouldn’t have to bear the burden of a work in progress.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.