Every reliability manager evaluating a digital twin platform for a refinery or petrochemical plant eventually asks some version of \"how long until this is actually usable.\" It's a fair question, and it deserves a real answer rather than a marketing timeline that assumes every plant has clean P&IDs, an available laser scan crew, and an IT department ready to approve OT network access on day one. None of those assumptions hold universally. What follows is a realistic breakdown of what a 90-day rollout for a mid-size facility — a single process unit or a cluster of related units, not an entire multi-thousand-acre refinery in one pass — actually looks like phase by phase, including where delays typically happen and why. This is a description of typical phases, not a fixed contractual schedule; every plant's starting data quality and internal approval process shifts the timeline in one direction or the other.
\n\nThe rollout starts with a scoping exercise that decides what actually goes into the model, because \"digitize the plant\" is not an achievable first deliverable and treating it as one is the single most common way a digital twin project stalls before it starts. A realistic scope for a first 90-day phase is a defined unit or asset group — a crude distillation unit, a specific tank farm, a piperack corridor — selected based on where integrity risk, inspection burden, or turnaround planning pain is highest, not simply where it's easiest to scan.
\nDuring these two weeks, the implementation team meets with inspection, reliability, and operations stakeholders to build an asset inventory: which vessels, piping systems, and rotating equipment are in scope, what inspection history already exists for them (UT, RT, MT, PT records), and what condition monitoring locations are already established under existing API 570 or API 653 programs. This is also when data sources get inventoried rather than assumed — existing P&IDs, isometrics, prior 3D scans if any exist, CMMS/EAM records, and the format each of those is actually in. A plant that assumes its P&ID set is current often discovers during this phase that field walk-downs over the past decade were never reflected back into the drawings, which becomes a scope item rather than a surprise later. Getting this inventory wrong — starting data capture before scope and data-source reality are both confirmed — is the single most common cause of rework in weeks 6 through 8.
\n\nWhat commonly slows this phase down: stakeholder availability is the most frequent early bottleneck. Inspection and reliability leads are also running the plant's actual turnaround and inspection schedule, and getting sustained time from them in week one is genuinely difficult at a facility mid-campaign. The second common delay is data quality discovery: finding out during discovery, rather than during a later phase, that as-built drawings diverge meaningfully from field reality reshapes the data capture plan and should be surfaced now rather than three weeks into scanning.
\n\nWith scope locked, the physical data capture phase begins. For process units, terrestrial laser scanning is typically the backbone method — a scanning crew moves through the unit with a survey-grade scanner, capturing dense point cloud data of piping, vessels, structural steel, and equipment at a resolution fine enough to extract accurate dimensions and clash detection later. A mid-size process unit generally takes several days to a week or more of active scanning time depending on complexity and how much of it is congested piperack versus open equipment area, plus registration time to stitch individual scan positions into one coherent point cloud.
\nPhotogrammetry — deriving 3D geometry from overlapping photographs — supplements laser scanning in areas where a scanner has poor line of sight or where visual/photographic detail (coating condition, nameplate data, corrosion staining) matters as much as dimensional accuracy. Drone survey adds coverage for elevated structure, flare stacks, tank tops, and pipe racks that are impractical or unsafe to access with ground-based scanning equipment, and is frequently the only practical way to capture condition data on structures that would otherwise require scaffold or rope access just to photograph.
\nRunning in parallel with physical capture, the existing document set — P&IDs, isometrics, equipment data sheets, prior mechanical integrity records — gets pulled and organized for the model-construction phase. This parallel track matters because document gathering from engineering archives and historical CMMS exports routinely takes longer than the scan itself, particularly at facilities where records span multiple ownership changes or system migrations over the plant's history.
\n\nWhat commonly slows this phase down: access and turnaround scheduling is the dominant real-world delay here. Scanning a live process unit means working around operations — confined space entries, permit requirements, area isolations for hot work nearby, and simply not being able to access certain equipment while it's in service. A scan crew planning five days of unit access can lose two of those days to permit delays or unexpected outages elsewhere in the plant competing for the same operations support. The second recurring delay is discovering mid-capture that a supposedly complete document set is missing key isometrics or that an as-built P&ID revision was never issued after a documented plant modification — which pushes reconciliation work into the next phase rather than resolving cleanly here.
\n\nPoint cloud and photogrammetry data get processed into a structured 3D model — piping runs, vessels, and structural elements identified and, where the platform supports it, tied to equipment tags matching the plant's existing asset numbering. Georeferencing anchors the model to real-world coordinates, which matters more than it might initially seem to: a model that isn't accurately georeferenced can't be reliably cross-referenced against site GIS data, underground utility records, or drone survey imagery captured on a different flight, and precise coordinates matter directly for subsea or buried asset work where GPS drift is a known problem.
\nThis is also where the model gets reconciled against the P&IDs and isometrics gathered in the prior phase — checking that what the scan actually captured in the field matches what the drawings say should be there, and flagging discrepancies rather than silently trusting either source. Discrepancies are common and not a sign of a failed scan; they're usually evidence of an undocumented field modification, and resolving them (confirming which source reflects reality) is exactly the kind of work that justifies having done the scoping and document-gathering carefully in the earlier phases.
\n\nWhat commonly slows this phase down: reconciling drawing discrepancies against field-verified scan data takes real engineering judgment and, more often than not, a phone call or site visit to confirm which source is correct — this can't be fully automated away, and plants that budget zero time for it in their internal planning are consistently surprised when it takes longer than expected.
\n\nWith the geometric model built, the phase that actually makes it an integrity tool rather than a pretty visualization is linking historical inspection data — UT thickness readings and CMLs, RT film or digital radiography records, MT and PT findings on welds and nozzles — to their exact locations on the model. This phase runs in parallel with model construction where possible, because pulling inspection history out of whatever system currently holds it (a CMMS, a document management system, or in less mature programs, a mix of spreadsheets and scanned PDFs) is its own significant task independent of the 3D work.
\nCMMS/EAM integration is the other half of this phase: connecting the digital twin to the plant's existing maintenance management system so that work orders, equipment criticality rankings, and maintenance history are visible against the same asset the inspection data lives on. This is frequently where IT/OT integration approval becomes the critical path rather than any technical task — connecting a new platform to systems that touch operational data, even read-only, typically requires a plant's IT security review, and that approval process runs on the plant's timeline, not the implementation team's. Facilities that flag this requirement during week 1 scoping rather than discovering it in week 7 avoid a genuinely common source of multi-week delay.
\n\nWhat commonly slows this phase down: IT/OT integration approval is, in practical experience, the single most underestimated timeline risk in the entire 90 days. A plant's cybersecurity review process for a new system touching CMMS or historian data can run on its own multi-week cycle independent of how ready the technical integration is, and that review often can't be meaningfully accelerated by the implementation team — it has to be started as early as institutionally possible, ideally in parallel with the very first scoping conversations, not queued until the model is otherwise ready.
\n\nBefore anyone treats the model as authoritative, inspection and reliability teams walk it against the physical plant — comparing what the digital twin shows to what they know from working the unit directly. This isn't a formality; it's where subtle errors get caught that automated processing can't reliably flag on its own: a tag mismatched to the wrong vessel, a CML that landed on the model a few feet off from its actual field location, a piping isometric that was reconciled to the wrong revision. Inspectors who've worked a unit for years will spot these discrepancies immediately in a walkthrough in a way that's much harder to catch by reviewing the model in isolation.
\nThis phase typically surfaces a punch list of corrections rather than a clean pass/fail result, and that's the expected, healthy outcome — a validation phase that finds nothing wrong is more often a sign the walkthrough wasn't rigorous than a sign the model was perfect. Corrections get routed back into the model, and for larger discrepancies, back to the data capture or reconciliation phases.
\n\nWith the model validated, the final phase shifts from building to adoption. This is also when role-based access gets finalized, since inspection, reliability, and operations users typically need different views and edit permissions into the same underlying model.
\nRollout at the end of week 12 is rarely a single on/off switch — it's typically a soft launch to the core user group from the validation phase, expanding access as confidence builds, rather than a plant-wide announcement on day one.
\n\nA realistic 90-day timeline gets a scoped unit or asset group to a validated, adopted digital twin — it does not mean the work is finished. The feedback loop that starts in week 12 and continues indefinitely afterward is where the platform actually earns its long-term value: new inspection findings get tagged as they're collected rather than batch-loaded later, CMMS work orders stay synced as maintenance happens, and each subsequent turnaround adds another data point to the corrosion trends and inspection history the model holds. Plants that treat day 90 as the finish line rather than the point where the twin starts accumulating real operational value tend to see the platform's usefulness plateau; plants that build the tagging habit into standard turnaround and inspection workflow see it compound.
\n\nThe phases above assume reasonably cooperative conditions — an engaged stakeholder group, a document set that's mostly current, and an IT security review started early rather than late. Real projects deviate from that in both directions: a facility with excellent existing documentation and a fast-moving IT approval process can compress this timeline meaningfully, while a facility discovering major as-built drawing gaps or facing a slow multi-week cybersecurity review should expect the same 90-day scope to stretch, sometimes considerably, without that being evidence anything went wrong. The single highest-leverage thing a plant can do to protect its own timeline is starting the IT/OT approval conversation and the inspection-history data pull in week 1, in parallel with scoping, rather than sequencing them after the model itself is built.
\nPrograms that pair the digital twin rollout with structured inspection data management through Atlantis NDT ERP and consistent field reporting via NDT reporting software tend to move through the inspection-history integration phase faster, simply because the data arriving from ongoing inspection work is already structured rather than needing retroactive cleanup. For plants weighing whether their internal data and approval processes are twin-ready, an ASNT Level III consulting review of existing inspection records before kickoff is often the single best use of the two weeks before a formal project start.
\n\n\nAtlantis NDT pairs field expertise with software: NDT inspection management software — Atlantis ERP, a digital twin platform for asset integrity, and NDT reporting software. Build your team with NDT training & certification (ASNT SNT-TC-1A) and ASNT certification pathways, or bring in ASNT Level III consulting. Affordable, accessible, fully customizable — book a free consultation.
\n