There is a lot of hype lately around artificial intelligence, edge computing, and the Internet of Things, but manufacturers need solutions to help them be competitive. Connecting numerous devices such as robots and sensors generates vast amounts of data, making streamlining industrial operations a challenge. How can automation and big data analytics help?
Robotics Business Review recently asked John Crupi, vice president of IoT analytics at Greenwave Systems, about how smart manufacturing can take advantage of analytics and the emerging Industrial Internet of Things (IIoT). He discussed how edge computing and predictive maintenance can help build production value.
Q: What are the most pressing challenges facing modern manufacturers, and how can automation and IoT help?
Crupi: Knowing where to start is the biggest challenge. IoT needs an end-to-end infrastructure, and it’s overwhelming for manufacturers, which may not have the team or expertise to create a connected analytics architecture that addresses the following infrastructure:
- Security at the edge
- Machine connectivity to industrial gateway
- Cloud provider, cloud services, and cloud scalability
- Secure connectivity to the cloud and back
- Edge computing and analytics
- Real-time and historical cloud analytics
- Device management and software updates
And that’s just the infrastructure and architecture needed to support solving their business problems. Luckily, trusted mega cloud providers, Amazon and Microsoft, are doing a great job helping Industrial IoT customers get connected and acclimated to their cloud and, most importantly, trusting their end-to-end security.
Also, Amazon and Microsoft are making it even easier for customers to “extend” the cloud to the edge by introducing edge software that provides cloud connectivity and security, plus device management.
Q: We’ve heard a lot about IIoT, but how much is it really used today? How is the market evolving?
Crupi: Although Industrial IoT has been around for quite a while and was called M2M — machine-to-machine — in its prior life. It has many stages of maturity that span from connectivity to remote access and management to monitoring and analyzing connected data.
Many organizations are only at the beginning of their journey to full-blown IoT solutions. Adoption is finally growing rapidly, but it’s not something that will happen overnight.
The good news is secure connectivity to trusted cloud providers is open for all to “try before you buy” and then flip a switch to move to production. It’s not that simple, but the tools (and great marketing) are easing the traditional integration pain, and fast-tracking end-to-end solutions.
But, there’s more. Customers are very interested in machine learning and artificial intelligence being applied to industrial predictive maintenance. At the same time, Microsoft and Amazon are pushing out tools and accelerators so you don’t have to be a data scientist to get moving with machine learning. However, you should have access to your own data scientists.
The big payoff is when all the machines are connected and being analyzed in real time, and machine learning models are predicting anomalies that solve real business problems in new and exciting ways. Continued advancements in machine learning and AI make it possible for companies to require these tools as part of their IIoT strategy.
Q: We’re aware of edge computing for functions such as perception and navigation for robots, but what else can it do?
Crupi: What may be lost in the midst of more popular functions like perception and navigation is edge computing’s ability to better manage and reduce costly equipment malfunctions.
The ability to perform predictive maintenance, which alerts you to what is likely to happen before a critical malfunction occurs, is extremely valuable in asset-intensive industries like robotics. It can give companies a more accurate view of equipment health.
Edge analytics vs. the cloud
Q: What are the advantages of edge computing versus big data analytics and management in the cloud?
Crupi: “Analytics at the edge” is a newer generation of analytics that runs on the device or on a gateway to which one or more devices are connected. More companies are adopting this for increased processing power on the edge device, an ever-increasing amount of real-time data from devices, and the need for on-device pattern recognition and anomaly detection.
Analytics in the cloud is typically associated with “batch processing” unbelievable amounts of big data using thousands of processors to split loads. The data is then processed to obtain results.
Edge analytics takes a similar approach but leverages all of the processing power to analyze data as it flows in. Then, it passes the analytics, patterns, and detected anomaly results to the cloud for second stage real-time and historical processing.
By processing at the edge, you not only get real-time results for visualizations and insights, but you also dramatically reduce the amount of data that travels to the cloud. We sometimes call this “lean IoT communication.” This is important for costly cellular and satellite connections and overall bandwidth concerns.
It is also highly valuable to real-time and efficient real-time IoT analytics problems. Use cases best suited for edge analytics are ones that require time-sensitive solutions to take action within seconds or minutes, not days or weeks.
Securing IIoT a priority
Q: What are the biggest challenges for companies looking to adopt and use IIoT?
Crupi: Security, connectivity standards, analytics architecture, and market fragmentation are a few issues holding back companies from adopting IIoT.
IIoT cybersecurity issues are real and can be extremely devastating. Everything from DDoS [distributed denial-of-service] attacks to hijacking an industrial system is of grave concern.
Many of these industrial systems were never designed to be on a network outside of their environment, let alone the Internet. Therefore, all industrial systems in the IIoT architecture must be protected and hardened from outside attacks and infiltration concerns. Once connected, we must determine how to communicate and handle machine data; this is where connectivity standards play a crucial role.
If IIoT connectivity is the means to delivering sensor data, the biggest challenge then becomes how to architect for analytics. Analytics can happen at many points in an IIoT system because there are many architectural tiers from which machine data travels from device to cloud.
The success of IIoT analytics won’t be in dumping data to the cloud and analyzing it. Instead, it will be the combination of real-time edge and cloud analytics working in harmony.
A fourth impediment holding back the IIoT relates to fragmentation of the market. There are too few vendors offering end-to-end solutions, so customers must go to multiple vendors for hardware, software, integration/implementation services, and training.
Analytics in real time
Q: How is Greenwave’s AXON Predict different from other products on the market? How competitive is your space?
Crupi: AXON Predict is a true, real-time edge analytics platform. It was built from the ground up to take advantage of the growing compute power at the edge as well as time-critical monitoring, analytics, and action needs. AXON Predict’s edge analytics engine is platform-agnostic. It’s small enough to run and perform analytics on almost any size edge device.
The AXON Predict visual pattern development tools ensure users with minimal technical expertise can create advanced analytics queries with ease. The pattern library provides step-by-step templates that remove the difficulties often experienced when developing and deploying queries with other analytics products.
Q: In addition to manufacturing, what other verticals can benefit from this sort of edge processing?
Crupi: Whether you’re a telco looking to optimize internet speed, in the oil and gas industry seeking to prevent a spill, or trying to efficiently manage a renewable energy source, edge analytics platforms can be applied to a wide range of industrial verticals.
The smart city is also an interesting example of edge analytics at work. Thousands of cameras, road sensors and traffic lights take in massive amounts of data. Edge analytics is especially valuable here, not only because of the amount of analysis required at the device, but also due to the amount of analytics required between devices, which make up the larger analytics.
For real-time security or preventative maintenance regarding infrastructure, analytics are highly dynamic and have to happen in a split second. Therefore, the edge is essential for these kinds of scenarios.
Q: What technical advancements do you want or anticipate for IIoT technology in the near future?
Crupi: Point-and-click machine learning for machine operators and analysts. The ability for users to move through a series of questions and selections begins the learning process and sets the foundation for learning behavior patterns for classification, anomaly detection, and predictive analytics.
Models can automatically be trained and tested for accuracy. The system will continue to train until accuracy reaches accepted levels, and then the analytics and models will be seamlessly deployed and inferenced at the edge.
Analytics capabilities and predictive models will be included as part of newly-manufactured devices and retrofitted to older devices. As new devices get connected, they leverage and integrate with analytics from existing devices and then begin to feed their own performance data into the system. The nirvana is for machines to self-analyze, self-tune, and self-react to predicted issues — whether that means notifying people or systems or initiating repair automatically.