Today’s guest blogger, Anodot's Amir Kupervas, is an expert in the world of data analytics. His company provides real time anomaly detection and analytics that discovers outliers in vast amounts of data and turns them into valuable business insights.
Despite all the collecting and analyzing of data from devices and systems connected to the Internet of Things — and the inherent cost savings that many use cases point to —quite a bit of that information is underutilized or is even unused.
A new McKinsey Global Institute report verifies the huge economic potential of collecting and analyzing Internet of Things (IoT) data. The report, "The Internet of Things: Mapping the Value Beyond the Hype," suggests that IoT could provide a potential economic impact of $4 trillion to $11 trillion in 2025.
Sixty percent of the potential value of IoT data, according to McKinsey, requires the ability to integrate and analyze data from an amalgamation of IoT systems. Amazingly, most IoT data is not currently being used, squandering immense economic benefits, McKinsey reports. For example, only 1 percent of data from an oil rig with 30,000 sensors are exploited. Leveraging all of this sensor data would deliver huge benefits.
[You may be also interested in Internet of Things Applications For Industries.]
One of the most promising areas in which IoT can be put to work benefits is that of predictive maintenance. Sensor monitoring of machine health will shift maintenance to a condition-based model. Data will obviate the need for regularly scheduled maintenance calls and reduce the incidence of equipment breakdowns. According to McKinsey, predictive maintenance has the potential of saving companies more than $630 billion by 2025.
McKinsey calculates that predictive maintenance could reduce factory upkeep cost by 10 percent to 40 percent. It will also foster better maintenance, resulting in a 50 percent reduction in downtime, simultaneously shrinking equipment and capital investment by 3 percent to 5 percent by extending machine life.
Predictive maintenance will reduce maintenance calls, summoning maintenance only when the need is demonstrated rather than it being based on schedule. Machine reliability will be enhanced and downtime reduced, as data-driven identification sends pre-failure alerts.
In order to realize the hefty benefits of predictive maintenance, IoT data must be analyzed. Traditional business intelligence (BI) solutions were not designed to analyze IoT data and deliver its full value for several reasons. First, BI does not analyze data in real time, causing latency in detecting issues.
In addition, these systems are only capable of deploying static alert thresholds. Placing a static threshold very low causes many false alarms while placing them too high may identify the next glitch later than necessary.
BI platforms were also not designed to spot correlations in real time. Typically, when a problem arises, detection requires studying several data points. For example, an engine that is about to fail may rotate faster than usual, have an unusual temperature reading and low oil level. Linking all of these sensor readings exponentially increases the likelihood of detection.
The shortcomings of traditional BI are where automated anomaly detection excels. The new technology:
- Analyzes data in real time, and using machine learning will learn the normal behavior of a metric stream.
- Detects issues and generate alerts before breakdowns happen. It can take into account the seasonality of metric streams – for example, temperatures that are naturally higher during the day than at night – and expose outliers that do not adhere to the normal behavior. That type of dynamic solution minimizes false alerts.
- Connects the dots among an almost limitless number of sensor data feeds to foresee problems in the making. Correlations offer a more robust and accurate analysis, one far superior to a single metric.
The result is data-driven predictive maintenance that provides early detection, prevents breakdowns and machine downtime. That, in turn, saves companies money by reducing production delays, downtime and emergency maintenance calls.
The McKinsey report suggests that IoT offers a tremendous economic value for the world economy, with a significant portion realized in factories and worksites. Predictive maintenance uses data and analytics to shift maintenance calls from time-based to need-based. To profit from that promise, companies must put an automatic anomaly detection platform to work to analyze data in real-time, use dynamic thresholds and drive powerful correlations.