Who’s governing the AI in your facility?

Who’s governing the AI in your facility?

Max Smolaks

October 21, 2019

7 Min Read

In the age of Industry 4.0, managing asset health increasingly requires managing digital health to ensure that prescriptive recommendations are unbiased and verifiable

by Jim Stuart, Lloyd’s Register

21 October 2019

Recently, at Maximo World, I delivered a presentation to more than 1,000 asset management professionals on digital transformation. Specifically, I discussed things to consider when adding advanced capabilities such as digital twins, machine learning, and artificial intelligence to a comprehensive asset performance management (APM) program in the pursuit of enhancing failure elimination, identifying and mitigating risk, and improving safety.

In this newdigital world, advanced methods are being used to better understand assethealth in real time, predict failures, and (using AI) prescribe requiredcorrective actions. By using new risk and reliability methodologies, failuredata libraries, modeling tools, and advanced analytics to process vast amountsof inspection and maintenance data, modern facilities are obtaining actionableinsights that pinpoint risk and enhance asset and plant performance andreliability.

It’s aboutresults at the end of the day, and through industry collaborations with theirdigital partners, operators I have observed have realized up to 20% gains inoverall production using these technologies with failure risk reduced by 80%and cost savings of up to 50% achieved. More important, all these projects acrossmultiple industries continually add to the ever-growing knowledge base andlibrary of failure and risk data.

So what’snext? How do we move our organizations to the next level? The next frontier offailure elimination is maturing your asset management program from the advancedanalytics of predictive APM methodologies to the artificial intelligence/Industry 4.0 world of prescriptive recommendations. But it is clear that thespeed of change in the digital age has brought with it a new set of challengesfor business leaders: namely, how to ensure that AI recommendations and modelsimulations aren’t biased and that the results are verifiable.

Governance requirements and decision assurance in the digital age

The concept ofdigital twins is predicated on a real-time data connection; without thisconnection, digital twin technology would not exist. This connectivity iscreated by sensors on the physical asset that obtain data and communicate itback to the system. Digital twin technology strictly depends on monitoring thephysical twin and how the environment and people interact with it – in otherwords, it is theoretically failure-proof from the moment it is built, but onlyif the data integrity has gone through a diligent validation process.

New governancerequirements are emerging around data integrity and decision validation withthese new technologies. As artificial intelligence, machine learning, anddigital twin modeling are introduced in Industry 4.0 applications, we now musthave a process for how these advanced technologies and their technicalintegrity are being assured to deliver the right answers. This is not atomorrow concern. Governance of digital is a current and pressing problem thatdemands immediate and substantive efforts to address.

Operatingplants, especially in the oil and gas and chemical processing industries, forexample, is a dynamic and continuous endeavor where operating conditionscontinually change. For a digital twin model to be a true reflection of thephysical asset, an entirely new set of processes is required, with each newprocess delivering new data, insights, and actions.

Asset health requires digital health management

All of thesenew activities require validation to confirm the accuracy of the models, andthis new area of governance requirements is earning a name: digital healthmanagement, or DHM. This new term encompasses all of the digital technologiesand systems that are used to gather data and insights on an asset’s health,which incorporates digital twin technology.

Furthermore, adigital twin can be defined as a “multiphysics, data-driven representation of aphysical asset, often residing in a cloud-based environment using data streamedfrom the physical asset” with varying applications from designers and operatorsto autonomy. In other words, a digital twin is a dynamic digital representationof a physical piece of equipment or asset. This understanding of the digitallandscape and its complexity drives the requirement for a comprehensivegovernance program to assure accurate output.

On the shopfloor, the result is helping operators improve aspects of their operationalperformance and maintenance regimes through insights generated by the twins aspart of the DHM. Key elements of DHM are ensuring digital model accuracy andanswering questions critical to the success of any digital twin initiative:

•  Whatstandards are applied, and what is the human role, if any, in the validation ofmodels and digital twins?

•  Do thedigital simulation, data structure, and technology take into considerationsafety and license-to-operate standards?

•  How issensitive company data used for digital twins kept secure?

Alarmingly,the data security question isn’t getting the focus required by the risk rankingin most facilities. Best-in-class facilities take this risk seriously anddedicate significant resources to addressing vital digital security questions.In addition to security issues, operators face challenges in ensuring accuratedata streams from sensors and IIoT devices as well as accuracy from remoteinspection technologies. The unintended consequences of a digital programwithout a rigorous governance protocol in place to ensure digital integritycould be bad decisions, lost profits, and, in the worst case, a catastrophicevent.

Beyond the fear
of Skynet

If you’re afan of the Terminator franchise, you know that August 29 was the day thatSkynet became self-aware. Could this happen in real life? Could artificialintelligence technology become self-aware and determine that humans are athreat? Lucky for us, technology isn’t this advanced yet, and Skynet isn’treal. While we can poke fun at AI becoming “self-aware” like in ascience-fiction movie, the reality is that a digital governance programprovides assurance that your digital program has the procedures and protocolsin place to deliver what it was designed to do.

It’s important that digital solutions be well-balanced across the technical solution and its business application. Practitioners and their partners also need to understand the challenges around data-sharing and data ethics in this collaborative digital ecosystem and address the value of the data and its contributions. In this way, owner-operators who are embarking on their digitalization journey will be better able to realize new value, and of more importance, build confidence in these technologies so that they can be trusted to make better, more-informed decisions safely.

Get the newsletter
From automation advancements to policy announcements, stay ahead of the curve with the bi-weekly AI Business newsletter.