Condition monitoring is central to understanding cogeneration plant performance and so is at the heart of any real chance of optimising potential economic advantages. Now though, the digital revolution is seeing the volume of data emerging from heat and power installations increase exponentially. And, as David Appleyard discovers, with the right know-how operators have an opportunity to truly maximise asset performance.
Tools delivering ‘live’ data on key operational parameters – such as boiler pressure or the temperature of inlet and exhaust gases – have been in place since the earliest days of the industrial revolution. But a century and a half on and the world is changing again. This time though, it’s a digital revolution.
With bearing temperatures, vibration and stress, exhaust emissions quality, valve positions and flow rates now available, for example – condition monitoring systems continue to accumulate variables to consider.
Indeed, a recent report from US-based industrial analysis firm Gartner says that by 2020 more than half of all major new business and processes will incorporate elements of connectivity or the Internet of Things. A key challenge for modern condition monitoring systems is therefore not a matter of acquiring data, it’s processing and presenting that wealth of data such that decision makers can act upon it in an effective and timely way.
Andrew Hubbard, Switzerland-based Product Management & Customer Support Director at Meggitt Sensing Systems, outlines the issue: “Combustion on a gas turbine is somewhat of a high frequency, high density data, and you could easily generate 3 or 4 GB in a few minutes. How are you going to store and manage that data? You could collect all the data in the world and unless you’re looking at it in some sensible way, it’s useless to you.”
This is a point echoed by Nikhil Kumar, General Manager, Asset Integrity Management (AIM) Engineering Director, Risk & Economics for Intertek: “There’s a large amount of all these different types of variables. You have the ability to add all these sensors in a plant, but you probably only need 3% or 4% of those in key locations. It costs very little to put sensor on right now, at some point you’ll have too much information that’s useless, especially on anything that runs like a power plant or any process centre.”
Kumar continues: “One of the problems is that if you have say 500 data points and you put that in the system then it becomes extremely difficult to pinpoint which one of those 500 individual items is of value to you. Especially from a plant operations perspective, he just wants to know that he’s not destroying the machine and he’s being profitable.”
Establishing a baseline for conditions
Effective condition monitoring is certainly key to achieving the twin goals of profitable and safe operation, as Petr Němec, Maintenance and Service Director at Czech Republic CHP project developers TEDOM makes clear: “Condition monitoring is important in two ways, one way is recording regular plant operations and maintenance, the other thing we can say that is important is if something happens [potential problems can be identified and addressed].”
Hubbard expands on this theme and its potential impact on the economics of plant operations: “Often when plants are operating it’s not a single person’s decision to operate or shut down. Often it will be a ‘committee decision’, it will be someone from the financial area of the business, the maintenance, the operations. A question often asked of condition monitoring is; ‘How long can I operate for before I shut down?’”
“Understanding the conditions, that establish a baseline and finding out what’s different enables you to make those decisions.
“What the industry is trying to avoid is forced outages, because planning resources and planning to get parts is a huge business cost, so if you can avoid forced outages it’s very financially attractive. That’s one of the attractive areas of condition monitoring.”
He continues: “If you’ve got this amount of data and you understand the controlling parameters, and understand the failure modes of a machine, then you’ll put the appropriate measurements on, then you’ll be able to say, for example, ‘let’s reduce the load or the duty cycle, let’s increase the oil flow’, so you’ll find a way to managing to the next scheduled outage.”
At its best then, condition monitoring can reduce downtime and reduce service costs – a significant economic benefit.
Beefing up the power of condition monitoring tools
As suggested by Hubbard, one of the key ways the new generation of condition monitoring systems is achieving operational benefits is through more sophisticated and timely failure analysis. More frequently today this is coupled with a more rapid reaction and analysis based on remote, internet-based access.
Němec highlights some of the advantages of a remote connection: “We can immediately analyse the information features and other things recorded. With internet access we can sometimes fix the problem immediately.
He continues: “The main advantage of the connectivity is that it’s possible to solve a problem without travelling to the site. For a long time, it was necessary to go to the site find out what happened and change it. Travel expenses are sometimes higher than repair costs.”
“For example, if there is the necessity to modify some set points we are able to change them remotely and the plant is up and running in a short time. Another example is that with the help of a monitoring system, we can analyse a failure and send the operator to change the spark plug and the result is perhaps four hours or more hours saved in the operation.”
Němec concludes that by monitoring the system online it is possible to observe non-standard operation of the whole CHP system and thus predict up-coming failures.
Kumar emphasises that one of the key enablers of this revolution has been advances in computer processing and power. “The software has been in development for a very long time.”
He explains: “The computing power required and the quality of data we were getting was insufficient. We have gained from the fact that our personal computers themselves have tons of memory we can calculate things much faster than we could eight years ago. The quality of information in terms of real time data is a much better quality than we had eight years ago.
“We have a lot more data to work with, that can be a good and bad thing.”
In the Intertek example he says: “We feed our algorithm, it’s a neural network, about one to two years’ worth of one minute data or 15 minute data and we teach the algorithm what the optimal performing parameters are for each of those regions of the power plant. The idea is to feed it all this information and then using some fancy statistics and engineering knowledge we know that in some cases there was spurious data or the plant was forced to operate in a way that wasn’t optimal and we come up with what in simplest terms would be a coefficient for a regression for each of those variables.”
Kumar also notes opportunities for fleet-wide analysis: “From our perspective we do relative stresses. That gives us the ability to use what our algorithm learns from all these different inputs to transport it to another unit. It also gives us a slight advantage from the fact that no two power plants are the same.”
Outlook for future condition monitoring
The rapid infiltration of data harvesting tools and sensors is set to have a profound effect on the operations and maintenance of combined heat and power installations. It’s also expected to see a significant change in industry’s approach to condition monitoring: “Companies are developing data reduction techniques, if you wanted to coin the phrase ‘The big data approach’,” says Hubbard.
“People are open to applying techniques from other parts of industry to deal with things such as data that can then be minimised, aggregated. I think it’s more open mindedness that’s coming these days, people are exposed to techniques from other industries, such as finance or weather forecasting, and they try to apply them and use them.”
Hubbard also notes the role of academic research in the development of future condition monitoring systems: “Often it will include a university or a research area to prove a concept or to start some kind of feasibility study to establish whether a technique will work or can be applied.”
He adds: “You’re looking for that edge, something that gives you a differentiator, I think there’s a few key areas in the next few years that are going to stand out.
“Generally, it’s going to be around automated processes that help with decision making, anomaly detection and also cause and effect analysis in an automated way.”
Kumar also envisages the continued advance of IT platforms in condition monitoring: “Cloud based operations, I can see that happening really soon. Already the Mitsubishi and GE’s of the world can get information from the gas turbines worldwide into their data systems in different locations. A lot of these algorithms will become control algorithms. Already a lot of the operation in the plant has become automated, algorithms are running power plants already.
“In my mind they’ll first start controlling by staying within parameters, where the designer of the system has set certain limits, the algorithm will make sure that you don’t go out of balance.”
I think step two, will be you can go out of balance in certain situations, if it’s profitable to do so, and then step three would be the optimisation of the whole plant, not only making profit but; ‘Can I reduce the damage [wear] on certain parts of the plant?’”.
However, he also sounds a note of caution: “If these algorithms can control power plants at this point, they’re only monitoring, the moment it can control the plant then security will be a huge issue.”
Hubbard also emphasises the importance of safety and security: “I think safety is also going to drive this. There will be more and more regulation that comes in and says ‘you must monitor’. The more measurement points we have through condition monitoring then the industry may be driven to ensure that everything that’s installed is underpinned by a specific regulation or standard to drive it.”
Němec, too, foresees greater emphasis on digital security, but sums up: “At the end, advanced monitoring is a great ‘added value’ for our customers. It gives us much a better condition for providing superior maintenance, which results in improved reliability.”
Reduced downtime and more effective maintenance equates directly to cost management, increased efficiency and economic effectiveness. Perhaps more significant though, advanced condition monitoring creates an opportunity to apply the most sophisticated analytical techniques and therefore the ability to establish a competitive edge.
David Appleyard is a contributing editor for E2.