Precision intelligence
in high-hazard environments.

Why are drawings the weakest link in a high-hazard facility?

Most process facilities are older than the drawings that describe them are accurate. A unit gets a tie-in during one turnaround, a line is rerouted in the next, a pump skid is replaced with a different footprint, a temporary scaffold becomes semi-permanent. Each change is small. Twenty years of small changes produce a plant that no longer matches the P&IDs, the isometrics, or the general arrangement drawings in the document control system. The gap between the record and the plant is called as-built drift, and everyone who has walked a unit with a drawing in hand knows exactly how wide it gets.

That gap is tolerable during normal operations because the people running the unit carry the corrections in their heads. It stops being tolerable the moment someone who does not have that knowledge needs to act. A mutual-aid engine from the next county, a contract crew on their first day of a turnaround, a hazmat team staging at the gate: none of them have twenty years of memory to fill in what the drawing left out.

There is a second problem that no amount of drawing accuracy solves. A congested pipe rack is a three-dimensional object. So is a five-level process structure with grating, headers, exchangers and cable tray stacked into the same vertical envelope. A plan view flattens all of it. Two lines that a drawing shows a foot apart on paper can be twelve feet apart vertically, and the person reading the drawing cannot tell. Elevation is where the risk lives, and a 2D drawing is the one format that removes it.

What does a measured 3D capture of a process facility give you?

Ark captures the facility as it exists, not as it was drawn. Ground-level LiDAR captures the dense areas: process units, pipe racks, compressor buildings, pump rows, control rooms, tank farm perimeters. Drone capture handles what a tripod cannot reach, including elevated structures, flare stacks, tank tops, cooling towers and roof-level equipment. The two datasets are processed together through Ark’s proprietary point-cloud processing into cloud meshes that hold measured geometry rather than an impression of it.

What comes back is the plant. Piping and its routing. Valves and their positions. Vessels, exchangers, columns. Structural steel, platforms, stairs and ladders. Electrical and mechanical equipment in the locations they actually occupy. Because the geometry is measured, a dimension taken in the model is a dimension taken from the plant, and it does not depend on whether anyone updated a drawing after the last modification.

Positioning deserves a precise statement. RTK latitude and longitude is applied to the exterior: the site footprint, the perimeter, the gates and the approach roads. Inside the structures, location is handled by an alphanumeric grid overlay and true-north orientation rather than coordinates, which is the reference system people can actually use out loud. Ark does not claim interior coordinates.

How does pre-incident planning change when responders can walk the site without entering it?

A pre-incident plan built from drawings asks a responder to imagine the site. A pre-incident plan built on a measured twin lets them look at it. An incoming crew can open the model in a browser from the station, from the truck, or from a laptop at the staging area, and move through the plant before they ever cross the fence line. Approach routes, gate access, hydrant and monitor locations, staging areas clear of the vapor path: all of it can be worked out in advance rather than negotiated on arrival.

Inside the fence, the value is in the specifics. Isolation points and shut-offs can be identified, labeled and found in the model before an event, so the question of where the block valve for a line sits has an answer that does not depend on who is on shift. Routing through the model gives a crew the path to a location, including how they get up a structure rather than only where the destination sits on a plan.

The grid is what makes this work under radio discipline. An alphanumeric grid overlay plus true-north orientation gives incident command and the field team one shared vocabulary for location. "North side of the unit" means one thing to a person facing the flare and another to a person facing the gate. A grid square and a true-north bearing mean the same thing to both of them, and they mean the same thing to the mutual-aid crew who has never been on the site.

The honest way to state the safety benefit is this: it removes reasons to enter a hazardous area to gather information. Entries made to confirm a valve location, to check whether a route is passable, or to see what a structure looks like are entries that a measured model can answer from outside the fence. Ark does not claim to eliminate entry. It reduces the number of entries whose only purpose is to find something out.

Can it connect to the control and monitoring systems we already run?

Yes, through API integration into the systems a facility already operates. The twin is not a replacement for a control system and is not positioned as one. It is the spatial layer those systems have never had.

SCADA integration is supported on that basis: an API connection into the systems already in place, with operational and sensor data surfaced against the model so a reading has a location attached to it. A gas detector reading is a number on a screen. The same reading placed on the correct structure, at the correct elevation, in a model a responder can navigate, is a piece of situational awareness. Equipment condition data behaves the same way. The point of the integration is not to move data somewhere new. It is to give data a place.

What does it do for turnaround and capital work?

Turnaround planning runs on assumptions about what is physically there. When those assumptions come from drawings that predate the last three modifications, the corrections happen in the field, during the window, at field rates. Planning against measured as-built geometry moves those discoveries earlier.

Clash and access checking is the clearest case. A replacement exchanger, a new skid, a crane pick over live equipment: each can be checked against measured geometry before the equipment arrives, including the questions that only elevation answers, such as whether the lift path clears the rack and whether the removal route is wide enough with scaffold in place.

Contractor briefing improves for the same reason. A crew can walk the scope in a browser before mobilizing, from wherever they are. Engineering, procurement and the turnaround team can look at the same model in the same meeting without anyone flying to site or standing in a unit. For remote midstream assets and unmanned facilities, that is often the difference between a scope reviewed by the people who need to see it and a scope reviewed by whoever happened to be nearby.

Does the model stay current as the plant changes?

A one-time capture becomes another stale record, which is the problem Ark exists to fix rather than repeat. Ark builds a living lifecycle twin: the model is maintained across the life of the asset instead of delivered once and abandoned.

In practice that means rescanning after events that change the plant. A turnaround that adds or moves equipment. A capital project that changes a unit’s arrangement. A modification that reroutes piping or alters access. The rescan updates the affected areas so the record and the plant stay aligned, and so the pre-incident plan built on that record stays true.

The test is simple. If a responder opens the model two years after delivery and finds a structure that no longer exists, the model has failed at the only moment that mattered. Keeping it current is not an upsell attached to the deliverable. It is the deliverable.