Storage, sensors, and processing
The cloud's slowly growing acceptance has been driven outside of industrial manufacturing by all kinds of cloud services providing more mobility in everyone's life.
Table 1. Typical business needs that can be addressed with cloud-hosted solutions
|Collaboration and work task execution||• Condition-based maintenance (CBM)
• Workflows and certain types of procedural enforcement
• Shared development and collaborative testing
• Virtual reality simulation and training environments
|Reporting and analytics||• Process analytics, especially for post-production optimization and root-cause analysis
• Remote diagnostics and system health monitoring
• Long-term process historians
• Manufacturing business intelligence
• Process or batch summary reports
• Energy management
• Mobile summary reports, alerts, and notifications
• Dashboards, KPI monitoring, and other web portal-based solutions
• Manufacturing Execution System (MES) reports
Source: "The Cloud for Manufacturing." Invensys and Microsoft, 2014.
"With companies becoming more and more involved in international/global business, using cloud services is the only answer to many of the challenging questions like providing production, process, and equipment transparency beyond the limits of a single factory," says Martin Brucherseifer, consultant product engineer at Siemens.
“The exponential growth of information accumulated from various sources by organizations nowadays demands a highly scalable infrastructure for collection, processing, and analyses of a huge quantity of data in real time,” says Anand Nayak, marketing solutions manager at Meridium. “Organizations that adopt cloud technologies faster will continue to enhance their competitive advantages.”
Within the industry, large sectors such as oil and gas have relied on their own internal "private cloud" solutions for years, adds Brucherseifer. "They run elaborate condition monitoring programs and collect the data in their company internal servers, and they have the budget to run a highly trained maintenance team including data analysts," he says.
"A lot of cloud-based approaches are provided by vendors like Emerson to deliver services to customers, because users don't have the expertise to do the reliability monitoring that they want to do or need to do to be competitive," adds Boudreaux. "These approaches may be particularly welcome by smaller organizations, which may not have the IT infrastructure and support to do the things that they want to do or need to do."
For organizations such as these, the Software as a Service (SaaS) model is much more attractive and cost-efficient (Figure 1). With the SaaS model, the client pays a subscription fee to use software over the Internet. All software updates are included, and users can access the software and configure it to their needs. All other tasks associated with maintaining the software, including data backups and server maintenance, are handled by the provider.
"When you look at the normal manufacturing areas and facilities management areas, SaaS is very popular," says Kevin Price, product director at Infor EAM. "We can deploy a new image of our environment in a matter of minutes, where it would take a customer several days to get the people involved and then to get it set up and running."
Infor EAM product manager Mike Stone sees the cloud enabling advances in CM measurement and analysis through Internet connectivity with assets that alleviate the need for on-site servers. "Route-based monitoring and analysis are still needed for assets that are not connected," says Stone. "Using the cloud, measurements and analysis can be executed faster and correlated with more data."
Siemens’ Brucherseifer suggests that collecting data at a central point (i.e., the cloud) has several benefits for the transparency and analysis of this data. "The first and obvious benefit is the availability of all data from a central location, including any kind of evaluation and analysis," he says. "Here it does not matter if the data comes from multiple locations within one plant or several plants that may even be distributed around the world; this central access to all the data allows us to run various statistics for a comparison/benchmarking of the individual production lines or plants. It supports furthermore a common data security and backup strategy led by a professional IT team which would otherwise be left to each individual operation team of each plant."
Another factor is the reliability of using cloud-based compute engines as secondary options to help ensure business continuity in the face of unforeseen challenges. "When you have everything centered on a common data center or when you do it on-premise, there's a cost associated with it," says Todd Landry, corporate vice president of product and market strategy for JMA Wireless. "And when you need to distribute the data centers around, there's even a deeper cost of replication. A hybrid approach that uses a combination of on-premise and cloud provides you with an inherent mechanism for replication and for disaster recovery, as well as giving organizations a secure way to begin leveraging the cloud as a redundant or overflow mechanism for their compute engines."
"Hybrid solutions are mostly around the public sector, as we see a lot of these with utilities (i.e., water and wastewater)," says Infor's Price. "The movement is definitely trending upward to people going out of on-premise and single-premise, and going to more of a SaaS environment."
“When it comes to condition monitoring, cloud is poised to have a transformative effect,” says Dan Miklovic, principal analyst with LNS Research. “What we are seeing as early developments are a number of smaller, specialized condition-based maintenance and reliability-centered maintenance (CBM/RCM) predictive analytics, usually by industry, start to enter into the market. At the same time, bigger players like Meridium with their Asset Answers solution are showing the real potential of cloud in drastically changing the way we could think about reliability benchmarking.”
Shawn Lyndon, ABB senior vice president of product management – data analytics, sees the cloud enabling data-centric techniques more so than other CBM approaches, regardless of the specific condition monitoring technique.
"I believe that empirical approaches lend themselves in most cases to being able to build richer models, because the larger the data set, the larger the "n" typically, the more accurate and precise you can be with your algorithms and calculations," says Lyndon. "If you're doing vibration monitoring, or IR, temperature, pressure, oil analysis, with any of those methods, there's several different ways you can use that data. One, if you know that equipment really well, and you know what its physical thresholds are, you can model that equipment in software. The other approach is an empirical 'machine-learning' approach, where you take big data sets and you try to identify patterns that would then correlate with certain indications of failure. There is a third approach used a lot in-plant equipment, which is a bit like what GE Smart Signal, In-Step and others use, where you look at the deviation from normal. It's still a very empirical approach where you look at changes from normal, but that's based on an individual asset."
On the cost of warehousing all that potential data in the cloud, Emerson's Boudreaux says simply, "Storage is extremely inexpensive." For example, the messaging technology that Microsoft has developed is something like 2 cents for every million messages, so "it’s extremely inexpensive to get data into the cloud and to store it there. The real cost is in processing the data. Once you store data on a hard drive, there's not a whole lot of energy consumed to maintain that data on the hard drive. Once you start taking the data and using it to run software and algorithms on it, that's where the main cost is."
"Wi-fi has been nice, because it handles a lot of information," says John Bernet, Vibration Analyst, Training and Reliability Professional for Fluke. "I think we need to think about how to get the diagnosis as close to the machine as we can, because moving data and information around the plant costs resources somehow, whether in storage, or in transfer, or in labor walking around to collect it. If we could get that diagnosis down into a smart machine that is on the plant floor, and have a technician who has to be down on the plant anyway to monitor those smart diagnoses, then you have smaller packets of information you're sending either wirelessly or however you transfer it up to the cloud. When it comes to critical machines being monitored, maybe we just monitor them with a simple overall vibration and then only collect the detailed information once a week, or when we need it."
ABB’s Lyndon sees challenges bringing intelligence too close to the machine. "One, you need to maintain that very complex sensor," he says. "You also increase the costs of your sensors; obviously that's coming down more and more, but it still has an increased cost. The sensors are incredibly sensitive – they're electronic equipment, and you've got them near equipment that's either rotating or heating up, or going through all sorts of changes, which is really tough on sensors. We have one customer where they put really complex sensors on some of their devices, and they found out that the cost of their condition maintenance effort went up and not down, because they were doing more maintenance on the actual (failing) sensors. Storing data is nowhere near as expensive as doing lots more maintenance."
At the moment, adds John Benders, senior vice president of product management – SaaS and cloud for ABB, the industry may not know enough about which data are important to make decisions today on which data get captured. "If you make decisions at the sensor or at a lower level as to what information gets put up in the cloud, what you're potentially missing is the something which may appear to be irrelevant from the perspective of the sensor level. When taken in conjunction with other things that are happening in different parts of the plant, it all of a sudden may become critical or extremely relevant, and you're going to miss the opportunity to draw on those sorts of correlations if you're not collecting all the data."
Fluke's Bernet also identifies another opportunity for cost reduction. "If you have a consultant or expert there, they are probably charging a pretty good fee, so a lot of smaller companies will say, I can only afford for you to come out and do the top 10% of my machines, and I can only afford for you to come in once or twice a year. If you have the staff on hand that has some smart tools that can help them do the data collection and the basic diagnoses themselves, then they do the monitoring, they do the diagnosing, and then the consultant just gets called in for problem machines or remote consulting. This way, you do a lot more machines in the plant, and you do them a lot more frequently."