As part of our compromise to boost innovation in our sector through the best technology, we are happy to share the first part of a series of articles, regarding the use of AI in turnaround planning. First, through definition and schedule development.
Of course, if you want to review our case studies, reach out to us at www.frontlinec.com/contact or visit our site: frontlinec.com
This is the first part of a series of blog posts on the topic of turnaround planning. The success or failure of industrial facility turnarounds or shutdowns is very much determined by the scope of work to be carried out and the control the team has on delivering it. Defining the scope of work to be executed during the outage and controlling it to avoid last-minute changes or additions sets the foundation for the success of the event.
A detailed, thorough, and collaborative process involving engineering, reliability, operations, and turnaround teams is required to define the TA scope. First, a comprehensive evaluation of the unit or facility is required looking into maintenance history, operational needs or improvements to be introduced. The outcome of this evaluation then goes into definition mode, where the turnaround/project planning and execution teams need to define how to execute the work. The main objective of any event is to balance required maintenance with improvements to be made and at the same time, prevent last-minute changes, additions, or including wish list items that can jeopardize the shutdown.
Once the scope of the event is more or less known, planners need to start defining the sequence of tasks to deliver each piece of work. Ideally, this work starts 18-24 months in advance of the event. The main challenge with turnaround planning is that the schedule is made of hundreds, if not thousands, of “small” individual and independent units of work. The work that needs to be carried out on one heat exchanger is very similar to the work to be carried out on heat exchanger number 100, but each needs to be looked at in detail and then assembled together into a coherent schedule. This, together with electrical, instrumentation, piping, rotating equipment, etc scope. Each individual unit of work is fairly easy to plan (isolate → decon → open → repair/maintain → inspect → test → despade), the enormous challenge is to align everything in one plan.
To develop a first-pass schedule, planners typically look at:
-
Same unit past shutdown or turnaround schedules (typically in P6/MSP, sometimes even Excel)
-
Past maintenance plans (typically Excel)
-
Company or industry benchmarks (if available)
-
Scope to be executed on the turnaround
The easiest way to start inputting the information for each item on the scope list is if there is a past TA where the same work was done if progress was properly followed, and if it was correctly defined on the schedule. If that is the case, it's a matter of copy and paste. There are a lot of “ifs” there, and anyone who has used project planning software such as P6 or MSP knows that the Excel/Word “copy-paste” functionality is a lot of work in project scheduling software. The naming convention, activity code, WBS location and structure, resources, calendars, etc will be different from TA to TA and need to be manually updated.
A simple copy-paste of work done on one pump from a past turnaround schedule to the new schedule could be 15-30 minutes of work. This is only if all the information from the past schedule is correct (sequence & durations) and is easily identifiable.
Assuming conservatively that there are 1000 units of work for an event, and all are properly recorded on past TA schedules, it would take one planner 3.5 months to develop the new TA schedule, just “copy-pasting”.
1000 units 30 min/unit = 30,000 minutes / (60 min 40 h per week) → 12.5 weeks
This situation is rarely the case and unfortunately, it takes much longer to develop a 1st pass schedule. This may be for numerous reasons:
-
Not every TA does the same scope, some equipment is opened every 2-3 TA cycles (8 or 12 years)
-
Schedule structure and logic are rarely maintained from one TA cycle to the next as people move and change roles
-
A scope may have been under-defined in past TA schedules
-
Accurate progress updates were not properly reflected, plan vs actual durations may have differed but this was not recorded
-
New scope, new equipment, new work methods, etc change the game
When this happens, planners are left with their experience and benchmarks they may find from similar work carried out elsewhere or from company/industry references. And this is a lot of work.
Why? Very simple, one first has to understand the nature of the work to be carried out and then look for similar work from other sources. Trying to find a similar scope on past TA schedules in P6 or MSP is like looking for a needle in a haystack. You may find some relevant information here and there that then needs to be stitched together.
Even using company/industry references, schedule of norms, etc may not be accurate. The typical tables with productivity per flange size and rating, or by type of heat exchanger, etc may be inaccurate. Some questions that arise:
-
With what data were they created?
-
When were they last updated?
-
Is that data accurate for your location?
-
Is it accurate for the contractor that will carry out the work?
-
Is it a like-for-like comparison?
These references serve to develop a draft schedule that needs to be discussed with the execution team for its feasibility. All this accounts for a lot of time to develop the first pass shutdown schedule. This is why ideally this work starts 18 months in advance. Turnaround scheduling is a difficult and complex activity that currently takes a lot of time and we believe may be disrupted in the coming years.
Artificial intelligence (AI) technology thrives with unstructured and disconnected but large amounts of data. And if turnaround events do one thing it is to generate large amounts of data. At Frontline, we have developed AI-based technologies to accelerate the schedule creation and development process. Conceptually, it's relatively simple. Our AI does what planners are already doing today but at a 10000x scale.
Frontline Trender Software Process
AI algorithms are great at understanding activities from natural text to then group and sort them. For turnaround planners this is great. It enables them to have a live repository of productivity data from past projects in just a matter of seconds.
Not only this, if the planner plugs in the high-level scope of work for the future shutdown event, AI can:
-
Generate a detailed schedule with a breakdown of activities and durations based on past shutdown data and current plan
-
Highlight significant differences between the scope from past turnarounds and the current plan, for the planner to check
-
Serve as a chat-GPT-like that explains the why and the data behind activity definition
We have tested with several clients the latest AI technologies developed in-house. The reduction in time required to develop a first past schedule goes from 2-3 months to 1-2 weeks. Why? Most of the repetitive, “copy-paste” work is eliminated and the planners only need to focus on the unclear, challenging, and value-added parts of the scope where their expertise will help them shine.
This post has covered the schedule development phase for an upcoming turnaround event. Keep posted for the next post on Schedule Optimization.