Hyperight

Predictive maintenance that unites data analysts and domain experts

predictive maintenance

In predictive maintenance, there is often a gap between “nice to have” and “must-have” solutions. Experts in analytics and customers often disagree on what should be the final outcome of the digitalisation. 

The solution comes in the form of a framework that will enable both sides to voice their points and views. Julian Zec, Global Manager Performance & Maintenance Optimisation at Cameron, will explore how to achieve such collaboration in his presentation How to Develop a Predictive Maintenance Program Aligned to the PoV (Proof of Value) Process, at the virtual Maintenance Analytics Summit.

Julian talked to Hyperight Read on this topic to lay the groundwork for our readers, as well as shared his thoughts on AI developments in the asset management field and his prognostics about fully automated maintenance.

Hyperight: Hi Julian, we are very glad to have you with us again, but this time as a speaker at the virtual edition of the Maintenance Analytics Summit. To begin, please tell us a bit about yourself and your role at Cameron.

Julian Zec - Global Manager Performance & Maintenance Optimization at Cameron

Julian Zec: It is my pleasure to work with you again. It seems that there is not so big difference between going to work through cyberspace or by visiting the office. Me and many others who have the opportunity to work from home, see it as an efficient way of working and the technology behind as reliable.

 This change is less drastic than I thought it would be. I do believe that virtual Maintenance Analytics Summit will fulfil expectations visitors have and remove some prejudices people hold about the virtual workplace.

I have recently joined Cameron, a Schlumberger company. My responsibility is to develop the concept of ‘Digital Equipment’. It’s a new solution which will join and utilise available information mainly from IoT streams, inspections, analytics with residual knowledge we have about equipment and oil & gas operations and help our clients to improve their financial bottom-line. We support field users. They shall be advised on how to improve the performance of equipment, maintain equipment or manage assets over a longer time period/ using global resources effectively.

A predictive maintenance framework that unites data analysts and domain experts

Hyperight: You are going to discuss How to Develop a Predictive Maintenance Program Aligned to the PoV (Proof of Value) Process, in which you will introduce the challenge in digital maintenance efforts related to a mismatch between produced analytics and their value / practical usability in the field. Why does this mismatch occur? And what are the implications of it in digital maintenance?

Julian Zec: We as a company are investing a lot into new product development, fundamental R&D and explore several paths towards the solution. But we are also a commercial actor, not a nonprofit one. It means that our research and development efforts have to be focused on meeting our customers’ needs and requirements. Our solutions have to optimise their drilling, maintenance or asset management processes, so the total cost of ownership is lower. 

And over the years, I have observed a significant gap between “nice to have” and “must-have”. There is a difference between what technical experts prefer and find interesting versus what our customer really needs and wants to achieve by introducing digitalisation. Investments are not easy to approve today, and we must be sure that we are not providing gadgets but value generators. Teams working on advanced analytics must be closely followed by domain experts (maintenance, drilling) and end-users to unlock success. 

Due to the difference in the type of knowledge and experience between these experts, a framework is needed that will enable them to get the possibility and time to express their points and views. Development of analytics is a truly collaborative process, not limited to IT.

A predictive maintenance framework that unites data analysts and domain experts

Hyperight: I’m sure our readers are familiar with your previous presentation included in an article on our Read channel, where you shared lessons learned providing condition-based maintenance in the offshore oil & gas industry. As one of the points, you advised companies to look beyond equipment failures, and instead, focus on digitalisation and building comprehensive AI-driven asset management strategy, if they want to fully utilise AI into their operations. One year later, where do you see organisations in their journey to AI implementation, with a focus on the oil and gas industry?

Julian Zec: Good things are already happening, and I have noticed a growing amount of publications on the topic. But still, I feel there is a lot to be done on developing AI on a higher asset management level. And there is an emerging need because digital maintenance is being more widely adopted. As the number of monitoring assets is growing, humans are falling behind, and the need for AI contribution to systematise strategic decision-making is increasing.

The major difference is that focus on Asset Management AI requires a shift of smart/predictive analytics from being a solely engineering discipline dealing with the equipment condition and failures to becoming a business initiative. Intelligent asset management combines technical, financial, efficiency, organisational and other information and is trying to answer how do we do business out of it and how do we maximise our values created.

oil and gas industry

Hyperight: Your talks and workshops frequently include condition-based maintenance. How is CBM different from predictive maintenance or other types of maintenance?

Julian Zec: That is a good question. There is a lot of confusion around terminology. What we all probably seek is value for investment relative to the level of risk we are prepared to take. So what most actors want is not a hi-tech solution but the most appropriate approach within our risk and budget constraints. This, by definition, is reliability-centred maintenance. The terms condition-based and predictive come into the picture when we start using data as feedback to adjust our maintenance philosophy. Condition-based maintenance uses predictive tools at scheduled intervals to advise on the action.

A predictive maintenance framework that unites data analysts and domain experts

Hyperight: And lastly, considering the current situation with COVID-19, experts predict greater robotics involvement in asset maintenance and operations. The final stage is to have dark factories that are fully automated and don’t require any human presence. Do you see it happening any soon, as companies are urged to fast track their innovation efforts?

Julian Zec: For sure. At several of my recent speeches and lectures, I am emphasising that I see condition-based maintenance and predictive maintenance as transitional stages between human executed and fully automated maintenance. For offshore installations, due to long distances and remote areas or even subsea environments, this will probably be a combination of monitoring systems, communication links and dedicated repair robots. Spares shall probably be deployed as standard containers and replaced annually or so. 

A predictive maintenance framework that unites data analysts and domain experts

It is difficult to put a number to how soon this will happen, but looking at the pace of how quickly things develop if there is business validation for it, I am sure that it will be sooner than we think. Of course, that will not happen suddenly and in all sectors at once. We are already seeing first steps in that direction, by mapping together information about the condition, spares status and location, geographic availability of the repair workforce, status of their certifications and so on. 

Gradually more and more components of that autonomous maintenance chain will fall in place. Advanced analytics is rapidly used to combine so many variables and produce the most efficient outcomes. That kind of AI is closely related to your previous question about building analytics on asset management level.

But full automated maintenance will require much more than analytics and machine reasoning. It will require machines capable of repairing machines and very important – change in how we design equipment. We will need to rethink equipment design and make it easily repairable by AI. It can be modular designs, less compact assemblies, standard connectors and may be more. I am not sure if it will happen in a year or ten years, but will for sure happen.


Add comment