Wireless technology standards, networks evolve and start to mesh
The idea of wireless technologies aren’t new, as panelists at the Manufacturing Summit noted. But Hesh Kagan of Invensys and Ian McPherson of Apprion noted during a panel discussion moderated by managing editor Jack Smith that new uses for old technology and new standards for an emerging market mean wireless will continue to grow as a plant-floor solution:
SMITH: Four years ago, NASA was doing telemetry and putting a lot of measurements back and forth using wireless. Now, all of a sudden, here it is in industry. What’s changed? Why now? What’s new? What’s the difference?
KAGAN: Wireless has been around a long time, and wireless for controls has been around a long time. Remote control is not at all unusual
What is new and what is causing all of the excitement is the fact that we’re getting two technologies: One, a mesh networking technology, is being applied to radio, what’s called an 802.15.4 radio that’s now available as a piece of silicon. So we have silicon radios that can be made extraordinarily inexpensively; and we have a mesh networking technology. Mesh in itself isn’t necessarily new. It comes out of telephony from years ago. But when you apply mesh over a low-power radio that’s relatively short-range, you can now have one radio talk to this guy, who talks to that guy, who talks to that guy, and gets from here to there over quite a distance.
So the combination of mesh networking, low-powered inexpensive radios is enabling a whole new class of sensors, and that’s what’s really happening. That’s what’s causing all the excitement.
McPHERSON: I think that gets right to the point. The cost of bringing new measurement into the plant now has dropped considerably with wireless technology, and the way in which that measurement now comes into the plant is standardized enough so that you have replicability of performance across different vendors releasing different technologies.
So it really comes down to the price/performance metric, and that’s now gotten to the point where you really have to consider wireless systems in the context of the overall cost of capital projects if it has those remote collection capabilities.
KAGAN: There are a class of guys — companies who started out of Berkeley and Stanford and MIT that spun off a number of companies like Dust, Sensicast, Millennial Net, Ember, just a few, that are all based around this 15.4 mesh networking technology, and they’re creating a new system.
SMITH: The different frequencies within the proposed standards, the 900 megahertz versus the 2.4 gigahertz, what does the pie chart show? Is there a lot more 2.4 out there?
McPHERSON: The vast preponderance of devices are in 2.4, and that’s a good thing, and that’s a challenge as well. I think it’s a good thing in that, as we talked about, 802.11 is very interchangeable, and there are so many devices now that are wireless enabled over an 802.11.
But, at the same time, for a lot of these communication systems, it would probably be preferable to use an unlicensed band rate if you’ve got access to the band because it is an unlicensed band. Just as soon as you spend three-quarters of a million dollars on a Wi-Fi plant network, you find out that 7-11 is going to put Wi-Fi all around your plant, or McDonald’s, or whoever. It changes the way it works, and your polyethylene is held up by Big Macs and fries.
KAGAN: If I can just kind of go down the security path a little bit, there really is a lot of folklore or urban culture folklore around security. There’s the belief that somebody is going to hack up their microwave oven and bring down the power plant next door. A little bit of that is true. Like anything that’s folklore, it’s based on truth, but it’s extraordinarily difficult to do.
From a technology point of view, the technologies associated with encryption and authentication are really good. They’re very, very difficult to break if you use them correctly. The question becomes whether or not they’re used correctly, whether they’re applied correctly, whether they’re even available in this stuff that you happen to buy, which is what I was alluding to before. Security is really a function of dollars and common sense, and the security is available today. It’s just not necessarily applied or used or maintained.
If it is applied correctly and used correctly, for somebody to try to hop onto a wireless network and spoof it, pretend to be somebody they’re not, it is extraordinarily difficult.
SMITH: You hear all of these problems that still exist with the proposed standards and some of the solutions, and then you hear the sense or the advertisements or the claims of vendors that say, “Oh, our system works end to end, no problems, just buy our system.” So where is that happy medium? Where is the truth? Where is the reality?
McPHERSON: I think that the reality lies in the plant itself. If I was running a plant of any size, I wouldn’t buy a single thing until the vendor came up and showed me that it worked in my plant with the configuration I expected because that’s fine to have a site reference that says we got 10 miles and 10 megabits per second, and that might have been sitting on top of a mountain somewhere, looking down, a clear line of sight down to some plant 10 miles away.
KAGAN: It’s really interesting what somebody means also when they say “it works.” Well, exactly what does that mean?
There is no standard way of providing specifications for something. A customer always wants to know, “Well, how far will it go? How much distance can I get out of this radio?” I don’t know. No idea. “Will this radio that I bought from you go further than this radio that you bought from me?”
There’s no way a customer, an end user, can get a spec sheet from me and a spec sheet from you, two different vendors, and compare in a rational engineering way what it is. They don’t know what the antenna technology is, the power technology. Somebody says “five-year battery life guaranteed.” Well, what does that mean? What they may be telling you is that the radio has been proven to be powered by five years, but the same battery is also needing to power the electronics for the sensor. They didn’t quite tell you that.
So there is no Good Housekeeping Seal of Approval in the way in which you can compare different vendors with different products. That’s something that WINA is trying to address, but it’s an extraordinarily difficult problem, as you can imagine, and it’s not a problem with the standard. It’s yet another story.
What we like to try to do and what we really insist upon doing before starting a wireless implementation is a site assessment study. Some vendors will say, “No site assessment study needed, just take our stuff, stick it in there, and it will work…
SMITH: And let it configure itself?
KAGAN: Yeah, it will find its own way, and it will work it out. I love the way you say “it will work out” to three-year-olds playing in a sandbox, “They will work out.” The wireless doesn’t work out on its own.
So we do a site assessment study. The site assessment study, again, isn’t something that’s written in concrete because the characteristics of the site will change over time, and the 7-11 with their lousy microwave will be built at some point in the future across the street. But the site assessment study at least gives you a baseline that reduces the risk of implementation, reduces the cost of implementation, gives you a good starting point.
SMITH: You anticipated about the necessity of the site survey. Is it always necessary, or does it depends on the application?
KAGAN: I think it’s always necessary, but to some degree. I mean, there are very intensive site assessment studies where you’re really taking a lot of measurements, and there are some based on the application requirements that could be, certainly, a little more casual. But you’ve got to put some thought into this. This is an engineering problem.
McPHERSON: I think it’s also important to look at that in the context of how some things can change for the better in terms of the way wireless can deliver information. One of the things I think we alluded to was the issue of location and the notion of location, of moving things, of people. So we talked to plenty of plants who lose everything from, well, people for some parts of the day to railcars full of product and certainly, in the warehouse, tools and equipment. So location is one of those very sexy things everybody seems to want and nobody seems to really want to pay for because it’s very hard to find that justification for a location tracking system.
But if it’s done well, a lot of these technologies can infer location. It may not be down to the foot or the meter. It might be a 20-meter range. But seven times out of 10, that’s good enough. If you’ve really got to know where that one person is, a man down, and you want X, Y, and Z coordinates, if it’s serious enough, you will pay for it, and you’ll choose something like ultra-wideband that has much more specific location characteristics.
But when you look at the value of understanding signal strength and relative positioning of clients to access points in an engineered, designed architecture, you have some tangential benefits that can come from that that can be inherent with the
technology, depending on the technology selection.
KAGAN: One of my favorite concepts is strategic measurements, the measurements that you would like to have if you could only get them. Now, some of these measurements, the physics might not allow you to get them directly, the texture or the smell or the color of something. But if you had that measurement, it would have an effect on the process. So to get that measurement, you would need to do it in an inferential way. You would have to surround the process, surround whatever it is with enough sensors, with enough other measurements so that when it was just right, you knew all the conditions around it to infer the measurement that you really want; you know, a soft sensor, so to speak. Of course, with wireless now, you can begin to do that because the cost of measurements is so inexpensive. In the past, it would have been really prohibitive to run enough wires and enough sensors to do that kind of inferential measurement.
So the question out to you guys, are there strategic measurements that you’ve been thinking about that you would like to get, but you just couldn’t have gotten or you haven’t been really able to get to date?
AUDIENCE: We have a manufacturing site in Kentucky, about 8-1/2 million square feet, so it’s all under one roof.
So it leads very much to that which you are talking about, getting some of those strategic and difficult points. What we’re looking for is something along the line where you can get down in the $50 a point range to connect those things out to bring back environmental monitors and some of the other things that are more of indicator-type alarms so that we understand when a process is starting to go up because capital is definitely one of the things that’s hard to come by. And those are the first things to go, are the fancies and the control schemes as the equipment becomes installed.
So as operations, we have to come back and retrofit that. So where do you see the costs going associated with some of this wireless technology, and how can you integrate it a whole lot more effectively and efficiently?
KAGAN: To talk about the piece parts, the radio, you can manufacture a subassembly that contains the radio, a little computer, a little bit of I/O on a one-inch-by-one-inch card, so to speak, for probably, with the software associated to drive it, a stack and everything else, in the ballpark of 20 bucks. So if you buy that from a vendor, you might be paying $40 or something. But if you’re in the manufacturing business, that’s all the piece parts are going to cost you and burdened overhead and all that stuff.
Sensors have always been rather inexpensive. Sensors, for the most part, are silicon-based for temperature, vibration, pressure, and they’re literally pennies or a dollar or something. Packaging, once again, becomes the expensive piece. Batteries can be expensive, or not, depending upon what you want. So it’s kind of an engineering estimate, again. You have to ask all the right questions to find out what’s really going to happen.
In the process world for condition monitoring sensors that we’re looking at for process environments, we think that you’re not going to find process condition monitoring sensors to the end user for under $300 for a while. A wireless sensor that’s not requiring the process level packaging — the Class 1, Div. 1 or Class 1, Div. 2 packaging, that can cost the end user $50 to $100, tops. Then there could be another class of sensors that certainly get down to the $50 level.
McPHERSON: That same notion is already being applied in terms of some of these inferential measurements and things like vibration measurements, looking at the health and state of rotating machines and bearings by looking at and applying algorithms to the measurement and then an outgoing collection of that data.
I’m always interested in the ways in which that same discipline can be extended to look at the inferential measurement for either behaviors or other parts of the process that would be enabled by these kinds of measurements and could be in proximity, could be security-related and just who should be where, at what times or what places. But there’s certainly a lot of core manufacturing processes that could change and improve if you can start building those inferences, whether they be about sort of dynamics in terms of the quality of the processing and supplies on hand or other things that are more specific to materials handling.
KAGAN: Condition monitoring is probably in the process world, where I come from, the low-hanging fruit from an application point of view. If you folks are involved in asset performance management and you’re working in that type of world where you’re constantly dealing with maintenance, depending upon the industry, whether it’s a continuous industry or a batching industry, there are different maintenance philosophies. But the worst case is a process industry where you’re non-stop process. The standard way of doing maintenance in these industries is affectionately known as break/fix. You wait for something to break, and then you fix it. It seems kind of bizarre, but that is the least expensive way of doing maintenance.
Preventative maintenance is extraordinarily expensive. If you have to go out every 50,000 miles and change the bearings, by definition, it’s labor-intensive and, by definition, you’re throwing away bearings that are still good. You’re only changing them because the manual told you to change them. So preventative maintenance is just too expensive to do on a large scale. We have seen customers with fairly good-sized plants measured in the number of I/O points like 30,000; 40,000; 50,000 I/O point facilities that could spend $70 million a year on maintenance.
Now, with condition monitoring, low-cost sensors to monitor simply temperature and vibration, you can create a profile for some asset, for some motor, pump, valve, whatever, and then capture that profile as a normal profile. You don’t need to really understand what the root cause analysis is of the problem but all you want to know is that the profile has changed over time and has degraded, and now you can begin to schedule maintenance at a time that’s convenient to do and a time that’s convenient to the process.
That type of model-based predictive maintenance is extraordinarily valuable. If it saves a plant 5% or 10% of their maintenance budget, which it easily could do, you’re talking many, many millions of dollars for a couple of thousand sensors across the plant at maybe $200 or $300 a sensor. It’s a got a fabulous ROI.
McPHERSON: When you apply that to things like other disciplines around quality, it may not change your life a whole lot when you’re making bulk chlorine; but if you’re a pharmaceutical and you’re manufacturing drugs at a very high precision and quality is a big component, those types of predictors and things that %%MDASSML%% assurances that the process is working, the subsystems are working properly, can have a direct impact on revenues’ bottom line.
AUDIENCE: Whether it’s electricity coming in or internal utilities produce compressed air or chilled water and that kind of thing, I would assume there’s a lot of applications there or things you’ve done in those areas?
McPHERSON: Absolutely. Power generation is one of the places we found the most interesting, partly because most of those plants don’t have so many of the Class A-1 types of ratings for hazards. The other is, as Hesh mentioned, it’s a continuous process. They are looking at improvements for outputs and efficiencies and maintenance, or any of their other budget. They can directly attribute that to the bottom line.
KAGAN: The low-hanging fruit that we’re finding, the applications that are of most interest, are condition monitoring because it’s so valuable, but there are very few vendors out there with condition monitoring — wireless condition monitoring sensors. But you’ll see all the biggies, you know, Invensys, Honeywell, Emerson and others, coming to market with these low-cost wireless
condition monitoring sensors very quickly.
AUDIENCE: Okay. How do you integrate that whole thing into a cohesive plant? Because you can spend $10 million on each one of those things.
KAGAN: We do sit down with maintenance, operations, safety, security, IT and start a plant strategy and a road map.
AUDIENCE: Can I give you an example? We’ve got our engineering system and programming system for our machine tools. It’s all wireless. The CAD system downloads to the programmers, the programmers unfold it, they write the program, it downloads to the horizontal machining center or whatever.
How do you put a plan together that already ponies on that wireless system and that wireless network and integrates, say, condition monitors at that horizontal machining center and takes advantage of that system, uploads it. Now, how do you get it out at the other end, because you’ve got data in another system.
KAGAN: Well, to that point, from the wireless stuff, the wireless is coming in from a lot of different technologies, but it comes into an access point somewhere. In our case, what we do is, we have a class of access points that can download — you can think of it like device drivers. You have got lots of device drivers. But the access point can deal with the technologies associated from multiple stuff, multiple vendors’ stuff. We bring that into the access point, where it’s normalized. We have a server that sits in the enterprise that can manage all these different access points. But once the data is in the access point, it can be normalized. So it can be engineered, but it does need to be engineered.
McPHERSON: The other aspect of this, and I kind of went back to it, I don’t think that you can escape having IT be part of those conversations because you’ve got a plant network, you’ve got a business network, and you could have a process network. They could be physically separated, and never the twain shall meet.
So the sensitivities around designing a system that supports those independently or in a unified manner, but partitioned so that they are segmented and separated, are going to be something that you take into a design point but has to be continued in terms of a methodology policy, and IT has to buy into that.
KAGAN: The IT guys win. Get used to it.
AUDIENCE: I was wondering what happens to the early adopters that had these legacy wireless systems once SP100 is fully out there and in place?
KAGAN: Good question. From the access point up, you can preserve the IT; from the intellectual property, IP, the intellectual property, you can preserve the application from the access point up. Between the access point and the device, you may need to change things.
From our perspective, we are architecting access points that have modular interfaces to radios because we know the radios will constantly be changing. So there will be stuff that’s throwaway. The thing you have to look for when you’re buying something is architecturally how isolated is the device from the radio. If we at Invensys are architecting a field transmitter, that field transmitter needs to be able to have a simple interface between the expensive instrument and the relatively inexpensive radio. So they do need to have a design in architecture that allows modularity. Otherwise, you’re going to have some stuff that’s throwaway.
McPHERSON: I think to follow onto that, to me, the fact that you’re using a proprietary protocol in a radio that’s an unlicensed band isn’t so much the problem with SP100 compliance. It’s going to be whether or not that gateway or access point exposed the information about its state and performance so that at least you can understand what that system is doing because if it’s a black box and its behavior changes, there’s not a lot the standard can do for you in terms of ensuring coexistence or manageability.
But if the vendor has provided you those interfaces and you can get the diagnostics and you can change the configurations and you can manage the device, you have a lot greater likelihood that you’ll be able to adapt around the standard, coexist with the standard, rather than just two dumb radios that yell at each other.
AUDIENCE: Just when I thought I was hearing was that you could have, say, a proprietary over-the-air network that runs, but if you’ve got a gateway with some level of capabilities, that could be used to bridge within the context the SP100 is putting forward?
McPHERSON: Yes. So the SP100 definition is that you will use this type of radio, and there will be a specific protocol stack that’s, if less than approved, that allows you to innervate in terms of your configuration and performance, the physical characteristics, but ensures that the way you operate and you speak to the gateway is commonly understood.
Now, what you can do in those instances is, you might have something that’s an outlier, but it operates in a specific way that you can characterize and basically manage around it, encapsulate it, and have it remain within the SP100 domain, but it wouldn’t be SP100-compliant.
SMITH: So that the information gets from one end to the other reliably, I think is the issue, right?
KAGAN: Yes. There’s an SP100 radio and there’s an SP100 gateway, and the SP100 gateway will have to have some capacity for dealing with what they might call foreign devices. So once it’s in the gateway, it becomes okay.
AUDIENCE: Will that be all of the black boxes that we will have to deal with?
McPHERSON: Well, I don’t want to overstate the abstraction that SP100 provides because we think of interoperability, for most of us who especially fly around the country and take our laptops, is, I have a simple Wi-Fi card, and I can speak to a Cisco access point, and it acts just like magic. SP100 isn’t saying that out of the box, you’ll take a Siemens transmitter and talk to an Invensys gateway, and immediately, those things will work. It says that I can use Siemens’ product A and product B next to Invensys’ product A and product B and they coexist, and they work in concert.
AUDIENCE: But in the end, how do you integrate them together?
KAGAN: That’s real tough. There’s a big difference between coexistence and interoperability. And most people are more concerned about coexistence than interoperability. But when you walk out into the field with a wireless tablet PC, you don’t want your RFID reader to stop working.
AUDIENCE: Right. But eventually, you want all your data to come together so you don’t have islands of automation, which is what we have today.
KAGAN: But it will come together. In this case, it comes together at a gateway or it comes together at the access point. Then it’s an application problem. It doesn’t have to happen at the field instrument level.
McPHERSON: The last thing is, I think everybody needs to be conscious of not repeating the mistakes of the past. The most recent past was with the Wal-Mart RFID mandate, dictating that suppliers shall all do this in terms of the goodness for all, and it was pretty much goodness for none, I think, from the early returns I’ve seen there.
So we’re not saying go, both feet, hook, line and sinker, but know where you’re going. Have a plan, have a road map and an architecture you’re building to because you will hit the wall before too long.