AI-enabled configuration translation

I grew up in a frustratingly monolingual household. This was in stark contrast to the primary and secondary schools I attended, both of which had large numbers of first-generation Italian, Greek and Asian students who had acquired sophisticated multilingual abilities.

As someone who is fluent in only one language and loves to travel, I am not surprised that language translation is one area where AI is having such a significant impact. The speed, accuracy and breadth of software-based language translation has greatly improved with the rapid adoption of AI-based software tools.

Automation has long been hampered by its own language translation problems. Configuration software prepared in one platform is not easily transferrable to another. Even systems from the same vendor, but of different generations, typically require complex translation. This problem has been recognised by many, including the Open Process Automation Forum, which defined configuration portability as a core objective. It is perfectly reasonable that end users, nominally operating the same plant in locations with different automation platforms, shouldn’t have to build and maintain unique configurations.

The accurate functional translation of configuration between automation systems is a complex and laborious task. This difficulty has long hindered automation modernisations. Because of the challenges imposed by the need to translate the configuration, end users have too often persevered with platforms — or vendors — that have long ceased providing the value expected from modern automation systems. The core of the translation challenge is that it requires a high level of proficiency in both the source and target systems. The software world is facing similar challenges, such as translating legacy COBOL codebase to modern alternatives.

Control system modernisation is typically a once-in-a-generation activity, rich with opportunity and risk. While the result of a system modernisation should provide a more maintainable, highly featured, expandable, easier-to-use system, there are significant risks. Among these are unplanned production losses, safety issues, equipment damage and operator disorientation.

Therefore, somewhat hardened by my career of modernising automation systems, I’m excited by the possibilities that AI can bring to this challenge. An example of such an initiative is Emerson’s DeltaV Revamp tool, where AI and ML power the assessment and translation of legacy system configuration into the DeltaV platform. If tools such as this can reduce the effort and risk of accurate functional translation, then the barriers to system modernisations can be greatly reduced. This will bring forward a whole host of benefits to end users.

In the same way that native languages of different cultures will always exist, it’s likely that different ‘languages’ of automation platforms will always exist. Perhaps the most efficient way of tackling this is through more effective translators. This will allow system vendors to differentiate and advance their platforms without the burden of potentially restrictive standards.

Some caution will of course be required. Errors in language translation between cultures might result in some embarrassment, but errors in automation configuration may have physical-world impacts. Some level of expert human translation will likely always be necessary to reach the standards required. Therefore, organisations that invest in both AI tools and their engineers will benefit most from the accuracy and speed of this new translation paradigm. As I see it, the coming years will offer great opportunities for the application of these exciting tools.

*With over 25 years of experience delivering automation projects, Stephen Cowhey, Lead Engineer at Emerson, has a keen interest in the adoption of new technology to improve project execution.

Top image credit: iStock.com/Sean Anthony Eddy