AI

Artificial Intelligence isn’t new to the process industry. In fact, it’s been used since the beginning of process automation

Any machine decision can technically be considered Artificial Intelligence. The relay logic, alarm units and PID controllers process engineers have been using for decades are all forms of what is known as ‘weak’ AI – based on fixed, human-created rules.

As time has passed, systems have become more complex and automation has progressed – decisions have moved onto PLCs and SCADA systems. But these decisions have still been based on those same hard rules.

These rules-based solutions will always be needed, but there’s now the option of making them more adaptable and efficient.

You can take advantage of Machine Learning, also known as Strong AI.

There have been quite a few AI-based buzzwords over the years. Expert Systems and Fuzzy Logic were all the rage – until they vanished completely.

There are major reasons these ideas never reached the production line, but the most significant one of all is that automation systems were already using them.

Expert systems are just large, complex rules-based computer systems – which describes your control system perfectly. Fuzzy logic simply meant analogue control. These were only new ideas for managers and washing-machine manufacturers, but automation systems had embraced them long before.

This time, machine learning is offering something new to the process industry.

The recent explosion in the popularity of Machine Learning has been caused by a few factors…

• Improvements in hardware that allow extremely complex ‘models’ to be created in far less time,
• New techniques that have radically improved the effectiveness of machine-learning systems, and
• Open-source tools that make creating and using AI much simpler.

Machine Learning has been finding its way into a number of solutions, but most of them have been dealing with data at “corporate scale” – looking at issues like machine condition monitoring, where the data is measured over days or weeks.

When customers say “Oh, we’re using AI”, they’re often talking about maintenance or logistics-focused data over long time-frames. It helps reduce downtime, but it doesn’t improve operations moment-to-moment.

There’s still a huge amount of untapped potential in operations – bringing the power of machine learning tools to the high-resolution data in your process control systems.

AI doesn’t take control of your system – but you can use it to drive a number of different improvements…

• Improve the tuning of your system with setpoints that adjust to the current situation rather than the worst-case-scenario.
• Improve alarm thresholds so they dynamically adjust to account for a wide range of factors, allowing you to catch issues faster and when they’re less severe.
• Improve the tuning of your system with setpoints that adjust to the current situation rather than the worst-case-scenario.
• Model equipment so you can spot when specific values are out-of-balance or unexpected.
• Detect novel (ie. unusual) system states to uncover problems or opportunities for improvement.
• Detect and predict specific issues that are too complex for human-written logic to spot.
• Model values that aren’t captured by sensors.

But how do you even begin to use Machine Learning in a process system?

It can seem like a huge task, since often…

• The data you need might be scattered across platforms, computers and networks.
• The data is organised in ways that only a few highly-technical people understand and have access to.
• You may need to mix enterprise and field data, such as staff schedules and planned maintenance times.
• You might need to mix data that belongs to different user groups, such as process, electrical and maintenance people.
• The data is at different time resolutions, formats and time-zones
• Recording, analysing and delivering the information the AI produces can be a huge amount of work.

ARDI is designed to break down the barriers between your information and the value it can provide for your company.

From the person in the field to the manager, ARDI makes data access much simpler across the organisation.

But it’s not just humans who benefit from ARDI – it’s a great platform for analytics and reporting, and a fantastic fit for machine learning applications.

It not only provides the structural and training data you need to build AI, but also helps you run your AI models live, where they can give instant feedback to staff and control systems.