You are Home   »   News   »   View Article

AspenTech - optimising plant with simulation and data patterns

Tuesday, September 18, 2018

Optimising oil and gas plant using simulation and pattern spotting requires careful effort to make sure you are guiding people, not instructing them, and making sure your software is very easy to use. We talked to AspenTech about how the use of the technology is progressing.

Oil and gas companies have been talking for many years about using data analytics and simulation techniques to work out how to optimise plant operations and maintenance activities.

But actually doing it is very difficult in practise, when you have a range of different equipment of different ages, and with different cost structures and objectives. Companies have staff with different amounts of understanding about how the field operates, and different levels of comfort and competence with technology.

Perhaps the biggest areas computer systems can contribute is in making computer simulations of the actual plant, which can be used to guide decision making, and from scanning large amounts of sensor data, to spot for useful patterns. That's easy to say but very hard to do.

One of the biggest areas of market interest is analytics systems for large compressor systems, which are very expensive, can be very troublesome, and can take out an entire production environment when they fail, and there isn't space or money to keep a spare available, says Ron Beck, energy industry director of oil and gas equipment optimisation company Aspen Technology ('AspenTech').

The analytics systems are also popular for optimising gas lift, reducing slugging, and reducing failures on drilling, he says.

A completely automated situation is many years away, so you will always need staff who understand the equipment operation and typical problems. They will combine the guidance and suggestions from the computer system with their own judgement. 'An oilfield is very complex. You can't just set it all off to run automatically,' he says.

AspenTech is based in Bedford, Massachusetts, and promises to help oil companies keep their equipment running reliability and at its limits, focussing on oil and gas, chemicals and refining sector.

Alaska gas lift

AspenTech was recently involved in a project on the North Slope of Alaska, looking for the best way to optimise oil production using gas lift injection. The project involved selecting the right option from a range of possibilities.

The cost of replacing equipment in Alaska North Slope can be three times as expensive as doing it anywhere else, because of the high transportation costs, so there is a need to try to get more out of old equipment.

The study observed that there was a big problem with inefficient gas compressors. If they don't run at maximum efficiency, they pump less gas into the reservoir.

AspenTech built a simulation model, running on real data, of how the simulators were operating and how they could run more efficiently. The simulation model is now used in Alaska, providing guidance to the (human) compressor operators.

If any of the sensor data is missing, the computer system can often make 'synthetic' data, based on looking at other data from the same period, to fill in the gap.

This project saved $3m in the first month, just by advising staff how to make small adjustments to how the equipment is operating, Mr Beck said.

Scanning data for patterns

In October 2016, AspenTech acquired a San Diego company called Mtelligence Corporation (known as "Mtell"), a company which makes software tools which aim to predict when equipment failures will occur, so companies can be prepared or find ways to avoid the failure, for manufacturing companies. It makes tools which can be used to understand early "failure symptoms" and their root cause.

The Mtell software can 'ingest'' large amounts of data, including perhaps two years of historical data, and all the current data from sensors on equipment.

If you tell the software that certain faults occurred during the past two years, the computer can look for the data patterns associated with that fault, which might include indications of something about to go wrong which could be seen several months before.

The system can then scan current data for the same patterns and indicate if the same fault looks likely to emerge again. 'We call those 'intelligent agents' - an agent is a pattern of data that the system is watching for,' Mr Beck says.

The system understands patterns of abnormal operation, which is called 'abnormality agents'. It also understands the patterns of normal operation.

Often, failures are caused by a change in operating parameters which happened several months before, but it is very difficult to work out what change in operating parameters led to what problem.

For example, operating a piece of equipment at a higher pressure will, over a number of months, lead to a problem of fluid breakthrough into the compressors, but you are not aware of anything until the fluid breakthrough causes equipment breakdown.

With the help of the analytics software, you can recognise the patterns, and then decide to reduce the operating pressure of the equipment.

The system can sometimes have as much as 10 different 'failure modes' or 'abnormality agents' which it is scanning the data for.

Sometimes, the system observes a change in data patterns, but does not recognise the pattern. In this case, the activity can be categorised as 'to be determined'.

Pilot projects

AspenTech has run around 30 pilot projects with different customers, to try to build confidence in this pattern spotting ability.

In the pilot, it asks a client to supply a sample of data which contains some specific equipment faults. Then it asks the client to supply another sample of data, to see if the computer system can spot same faults emerging in the second sample.

'We've done that in 30 pilots over the last year, and been successful in 30 out of 30 pilots,' Mr Beck says.

In one example, a customer was told, that they would see a problem with a compressor in between 80 and 87 days. The customer did not believe it. The compressor then went on to fail after 83 days, Mr Beck says.

The company recently announced a major purchase by one of Italy's largest refineries based on one of these pilots.

Guidance not instruction

It is important that the recommendations made by the software are presented as 'guidance', rather than instruction, Mr Beck says. The computer system can only ever know a small part of the full context of operation.

The equipment operating staff can use the recommendations together with whatever other goals or understanding they have to make a decision about what to do.

For example, the computer system may advise about an emerging risk, but the human operator may decide the risk is worth taking, bearing in mind it may be much cheaper to do maintenance during a scheduled shutdown in a few weeks' time.

'It is making concrete suggestions, 'change the operating pressure of this system in order to avoid this problem,' Mr Beck says.

Also, staff might not be comfortable making a change just on the basis of a recommendation from a computer system. 'You're dealing with large amounts of equipment, operating oilfields where if you make a wrong decision there's an implication,' he said. 'People aren't just going to make whole sale changes without a comfort factor.'

Making it easy to configure

AspenTech has been putting a lot of energy in trying to make its software systems easy to set -up and work with. Whether people can quickly get up to speed with a system is a major factor in the success of the implementation, he says.

Building software tools which subject matter experts and 'end users' can configure themselves is a major part of AspenTech's approach. 'Our strategy is fundamentally to empower companies and empower users,' he says. 'If you empower people, innovation will happen faster.'

Ideally, people should be able to access data without having to be an expert data scientist, or need years of experience working with these pieces of equipment. They should also be able to do it on tablet computers, or when working outside, using gloves.

The software interfaces are designed to be as fluid as possible, like software on mobile phone apps. They can be quickly reconfigured into something else.

The company also tracks how people are using the software, and which parts of the software they are using.

AspenTech uses 'personas' in its product development, an idea of what a typical customer might look like - including what their objectives might be, and their inclination and ability in learning a new piece of software.

To get a better understanding of how younger people work with technology, AspenTech recently hired 200 college graduates, accounting for about 14 per cent of its total workforce, because it wanted people who had grown up with new technologies.

The company also provides a wide range of online training materials, including e-learning systems and YouTube videos.

Associated Companies
» AspenTech
comments powered by Disqus


To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.


Latest Edition Jul-Aug 2022
Aug 2022

Download latest and back issues


Learn more about supporting Digital Energy Journal