Measurement data on the lookout for artificial intelligence
Networked intelligence is the key to taking the process industries to the next level.
Industry 4.0 is the solution for everything — this is the general impression promoted today in relation to the efficient running of process plants. But there are grounds for doubt: Is it all just hype? Just something for the industry giants with a lot of IT manpower? This is not quite right either.
Every company that makes use of largely modern field and control engineering has it: large amounts of data that could be utilised to monitor things better and improve efficiency. It is therefore worth taking a detailed look at which solution suits which application. Implementation is easier than some might think — even when big data analysis and artificial intelligence come into play. New or modernised process plants should be fun places to work: they are equipped with reliable instrumentation; the latest control engineering makes operation easier; and ideally, they feature marking and wireless solutions that mean maintenance and inspection rounds can be documented electronically.
If you stop improving…
Now it’s time to take the next step. A functioning, safe plant is the compulsory component. As for the additional element, over the years this has turned out to be a plant that offers high availability, where the equipment never has to undergo unscheduled maintenance, and which never fails at short notice. This may still satisfy the judges today, but in the global race for competitive product prices, it is the most efficient processes that will sooner or later win the race. And process owners cannot afford to ignore, in addition to the basic features already mentioned, commodities that they already have access to tons of: data, information and knowledge. Knowledge, if used correctly, can lead to even greater plant efficiency, even better safety and even higher reliability. So utilising the additional data becomes compulsory — in the medium term even in plants where, until now, highly qualified personnel with a great deal of skill have been teasing out every last drop of efficiency and thus creating a competitive advantage. But due to skills shortages and demographic changes, this advantage will not last.
It’s something that every plant operator is familiar with: they have been installing expensive fieldbus or HART-compatible measuring devices for years, which constantly provide more data in addition to the actual measured values. But, with the exception of some basic diagnostic data, the vast majority of this information is never used. It could do so much more; but where to start?
Start where it counts! A diligent analysis of any weaknesses simply cannot be avoided if an operator wants to improve things and use the available resources wisely and effectively. The good news is that some things can be set in motion with relatively little effort and expense. The analysis could start with the emergency showers and end with the rotating equipment, or start with the neglected manual valves, whose position is still recorded on a piece of paper… the list goes on. The team will definitely know where the recurring snags are, where additional monitoring would be appropriate or which variables that are already being recorded just need to be analysed properly. In fact, plant users, maintenance engineers, automation technicians, logistics specialists and process experts all have their own particular view of things — and their own particular needs when it comes to supplementary information.
Artificial intelligence and mass data: a fertile partnership
Many plants have a huge variety of diagnostic information available, but it is not all useful to the various user groups. Many monitoring solutions simply serve to notify the maintenance team of a malfunction, but there are also some application areas, such as motor monitoring, where the data from just one or a few machines does not provide the maintenance engineer with enough information to draw up a maintenance schedule.
Operating pumps or compressors, in particular, generate a huge amount of data, which must then be analysed in a time-consuming process. The operator gets the most benefit from this if the data is not just sent to the on-site notification system but is made available to a cloud platform, together with data from thousands of other pumps. Patterns can then be identified and investigated. Ultimately, artificial intelligence can turn a heap of measured data relating to vibrations, performance and other parameters into reliable predictions: when will which pump sustain what damage? When is maintenance the appropriate course of action, and when a replacement? Data from the many becomes knowledge and advantages for all. Many pump providers have picked up on the potential here and are now offering corresponding online services, often linked to maintenance services.
The endless computing capacity that the internet is able to harness through outsourced cloud services enables it to provide predictive maintenance better than any on-site control system or application could. But it is by no means always necessary to get the internet and artificial intelligence on board in order to make process plants more user-friendly and efficient to run. For other plant components, such as trace heating, you might simply want to know: does it still work? And for that, you don’t need to analyse massive amounts of data. Monitoring — by measuring current, for instance — would be easy; whereas a failure noticed too late could be catastrophic.
Emergency shower alerts paramedics: data analysis can save lives
An emergency shower that is in good condition can save lives; no-one in need of emergency assistance would want to find out that it just happened to have failed right after the last regular inspection. What’s more, this piece of first aid equipment is a good example of how not all information is equally important to everyone. The maintenance team needs to know whether or not the equipment is working; but if an emergency shower is activated, first aiders and paramedics need to get that information immediately.
So many different monitoring applications means that there is a need for one common thread: all roads lead to OPC UA. This open interface standard is a vendor-independent solution for data exchange — no matter whether it is HART data from the field device, the water pressure of the emergency shower or mass data from pumps. With OPC UA, the data is made available to whatever system needs it — be that the cloud, a local maintenance tool or a separate alarm system.
NOA and OPA need time. Brownfield plants cannot wait
Of course, one can throw money at the problem and assemble a large workforce to carry out additional monitoring and optimisation tasks. Structures such as those put forward by Namur (NOA, Namur Open Architecture) or Exxon Mobile (OPA, Open Process Automation) can be helpful here. Both approaches enable extra sensor signals to be retrieved via an open interface such as OPC UA, and several large corporations are currently in the process of laying the foundations to do just that. But for countless smaller firms, these approaches still seem a long way off.
But that doesn’t have to be the case: every plant operator can do something for the availability of their plant and improved maintenance processes — and not necessarily just by shifting everything to the cloud. Where it is appropriate, plant operators could indeed do this with modest means, by dedicating a portion of their maintenance budget to it. If a plant already has full-coverage WLAN and performs paperless inspection rounds, it is not far away from taking the next step of introducing augmented reality to assist its fitters. And if this can be combined with a predictive maintenance solution, great! Then the maintenance engineer will not only be provided with the best possible support when servicing a pump, he will also be automatically guided to a unit which would fail sooner or later without his attention.
Monitoring wells: the maintenance team has better things to do
Big data analysis makes sense when it comes to pumps, because generally speaking, one should pay particular attention to the rotating equipment that is responsible for 50% to 60% of all unscheduled plant failures.
This is true for many plants, but not all however. For example, lots of maintenance teams in oil and gas spend a great deal of time regularly inspecting all the wells and shafts. But there is a simple solution that offers an alternative means of continuously monitoring the shaft, and which reports both unauthorised opening and the level within the shaft. All it takes is one measuring device with a modem and a battery — the staff just have to pop by every six months to change that battery. Even better: personnel will be notified when it is time to carry out this task.
It is just as easy to monitor whether hand valves are in the open or closed position. Simple technical solutions are all that are needed to record the end position, based on a NAMUR limit switch, for example. The signal can be transmitted into the control system wirelessly, so errors made by production workers, whether from actuating the valve incorrectly or simply neglecting to enter information in the software, can be avoided.
Sending diagnostic data to the cloud! Untapped HART data is yesterday’s news
Things become much more complex if a plant operator wants to squeeze maximum availability from valves and positioners, since these provide hundreds of items of diagnostic information that cannot be so readily assigned to an ageing process. But there are software tools that can make reliable predictions based on mass data analysis. If the provider of such a tool is granted access to the HART signals of all positioners, the software can warn of leaks or a deteriorating valve lift in good time. An app, which each user pays just a small fee for, facilitates the scheduling of maintenance work, thus optimising system availability.
The majority of field devices make diagnostic information available via HART, but this data is hardly ever utilised. Now this could finally change, thanks to digitisation and Industry 4.0. The computing capacity of control systems or a maintenance tool is necessarily limited, and systems are often not up to date — and there could be years to wait until the next migration. But in the cloud, things are different — even in a public cloud.
Network intelligence for columns! When APC reaches its limits
Networked intelligence could revolutionise how distillation columns are controlled, for example — an issue that, despite all the progress made with Advanced Process Control, has still not been solved satisfactorily. If technical specialists had access to all available variables and disturbance variables, they would be able to create the ideal closed-loop control. All they need are these variables and enough computing capacity. An IIoT gateway gives them access to both. Any online services the IT experts like can be used on the internet or an intranet. Some new setpoints are returned, eg, as Profinet variables or as a recommended action, depending on the application. Taking pump analyses as an example, diagnostic data is fed into the network as big data, and what comes back is this: please replace pump 27 within the next three months.
Even weather events can be accessed by the automation or control system, no matter how old it is. Gone are the days when personnel at the control board always had to keep one eye on the weather radar and adjust the system in response to every unusual weather event. In future, the weather forecast can be used as an online service; the plant will adapt automatically to strong rain from the west or a cold front from the north-east. Weather events become process variables. Other potential applications will be a welcome playground for IT and automation graduates joining the industry today.
Digital technology is helping mining companies across the world realise the true potential of...
Commissioning a new system for a food manufacturing plant requires software assistance to plan...
Online condition monitoring involves a number of challenges aside from data collection: data...