IoT forging new pathways to innovation
The value in using IoT for knowledge management and decision support is found in the analytics conducted on the data that comes in from the sensors. The SAS Analytics for IoT is built on the well-established SAS analytics platform and uses rich visualizations provided through SAS Viya. It can analyze data that is collected centrally but can also go from the Azure Cloud to the edge, where sensors are located, to analyze data generated there.
Some IoT systems produce readings every few seconds, in which case, it’s important to separate the relevant from the routine data in order to spot trends every 5 seconds, but only a small portion needs to be sent through the network,” said Jason Mann, VP of IoT at SAS. “Many customers have a 10-minute evaluation window. If an event takes place that is outside the norm, data would be captured within a time frame of 10 minutes to detect the precursors or able to distribute analytics to the point where decisions are required.”
Transportation has been an early adopter of IoT technology. “Many trucking companies service and support the vehicles after sales,” noted Mann. Each truck might have dozens or hundreds of sensors, which are collecting information on performance. “If readings are not within normal parameters, fleet managers can plan for maintenance, and by analyzing the driver’s route and schedule, they can determine the best time to bring in that truck. Uptime is a critical KPI.”
Partners play a big role in ensuring that an IoT system functions as intended. “The components in the IoT ecosystem have to work together, and no industry-specific platform has all the data sources,” Mann observed. Typical components include weather data, GIS information, 2D and 3D mapping, hydrology data, and, depending on the industry, a wide variety of sensor inputs. “We make sure that all the systems we work in are compatible with the Azure Cloud, but we also need to be able to work outside that environment and bring the data together.”
SAS Grid Guardian AI was developed in partnership with Exacter, a company that makes sensors to detect problems in the electric grid. It automatically associates sensor data with the related electric system asset while the sensors are on the move. To collect this data at low cost, municipal garbage trucks are outfitted with the sensors and collect data while making their rounds. Once the data is analyzed, areas of weakness can be identified, which helps direct capital investments and avoid electricity outages.
Streamlining operations
One of the reasons the use of IoT is so pervasive is that it is effective in allowing many industries “do more with less,” a mandate that applies across the board. “It is no longer feasible to send engineers around large buildings checking on the status of different systems,” said Rob Mesirow, partner, Connected Solutions, at PwC. “This type of activity needs to be automated and the data converted to actionable information.” Unlike some technologies, IoT is not one for which a case needs to be made; the question is more about how to implement it.
According to Mesirow, PwC provides an end-to-end solution that includes working with its clients from the start to develop a strategy, from the physical installation of a network to delivery of analyzed data. The company looks at the client’s policies and procedures, conducts interviews, and establishes baselines and desired outcomes. When it comes to implementing the technology, PwC has an array of products and partners. “Our overall approach is to build on what the client had, to make ‘dumb things smart,’ and help clients to understand the data on the back end.”
PwC manufactures its own sensors because early in its IoT initiative, the products it needed to implement the systems were not available. In other cases, they are able to enhance existing technology. “HEPA filters have a set of needles that move around and are hard to read,” continued Mesirow. “We added digital cameras that can read the temperature gauge perfectly almost every time. Over time, the sensor gets trained to read that particular meter and is more efficient than a human.”
Once the data is collected, it is analyzed in Signal Graph, PwC’s unified data platform. The data ingested into this graph database includes on-premise historic data, third-party data such as weather data, and the IoT dataset. “The IoT data is the most important because it is live and immediately actionable, and it also is temporal,” Mesirow commented. “Signal Graph is constantly looking for anomalies and patterns of data in unsupervised mode, and then pushes it out to us and the client.”
In order to make the information output actionable, the presentation options need to be flexible. “As an example, a manager in the food industry might want to get alerts if the temperature in a refrigeration unit went out of range, since a great deal of food is lost due to such failures,” explained Mesirow. “However, they also might want to ask a more general question, such as whether all the food in storage was safe.” In this case, the system could report that everything was within the proper range, but that there had been some unusual fluctuations that might point toward an imminent failure in a unit. In that case, remediation measures could be taken in advance, preventing the loss.
Many clients rely solely on the PwC analyses and visualizations, but PwC can also provide APIs to enterprise systems such as SAP or to building management systems. “Our intent is to present the data in such a way that the client does not need to rely on internal IT or data scientists but can access it directly,” noted Mesirow. Given that graph database output can be complex, PwC recently introduced a way to view it via augmented and virtual reality. “Walking around a model and looking at the nodes is an engaging way to understand and discuss the data,” he added.
Sensors have been around for a long time, and so have analytic solutions, cloud storage, and the internet. What’s different now is that the technology is available for them all to work together to bring actionable information to users in real time.