---
title: "Overlapping Model Boxes Do Not Prove the Equipment Shapes Collide"
description: "Leviathan Systems — Technical content Published 2026-10-20; updated 2026-10-20."
canonical: "https://www.leviathansystems.co/articles/cad-bounding-box-overlap-versus-shape-intersection"
last-updated: "2026-10-20"
---

# Overlapping Model Boxes Do Not Prove the Equipment Shapes Collide

Leviathan Systems — Technical content
Published 2026-10-20; updated 2026-10-20.

Overlapping bounding boxes establish that enclosing volumes overlap. They do not establish that the represented equipment bodies share space: part of each enclosure can be empty. Treat a box-only result as a candidate for an appropriate shape-aware assessment before calling it a collision. Conversely, correctly enclosing boxes that are disjoint prove that their enclosed shapes are disjoint. Both conclusions concern the represented geometry, rather than the completeness of the equipment model or its suitability for installation.

## An enclosing box includes space the object does not occupy

An axis-aligned bounding box describes an object's extent along three chosen perpendicular axes. For a tight box, its opposite faces sit at the minimum and maximum coordinates of the represented shape. The box contains every point of that shape, but it also contains every combination of coordinates between those limits. Those combinations need not all belong to the object.

That loss of information matters for equipment with openings, recessed regions or separated components. A frame can extend to each face of its enclosing box while leaving most of the middle empty. The enclosure records how far the frame reaches; it does not record where the frame's material is absent.

The geometric logic works in one direction. If two represented bodies share a point, their correct enclosing boxes must also share that point. If the boxes are disjoint, no point can belong to both enclosed bodies. But when the boxes overlap, the shared box region might contain material from both objects, one object or neither. The screen has retained a possibility rather than proved a collision.

This explains why the same box can be useful for quickly excluding a distant object and insufficient for deciding a nearby object's relationship. The limitation follows from enclosure itself, even when all dimensions and calculations are correct.

## Identify whether the reported result is a screen or a shape assessment

Autodesk's Revit 2026 API guide describes BoundingBoxIntersectsFilter alongside other bounding-box filters. Their Outline input is a rectangular prism aligned with Revit's world axes. The guide warns that an element's box can be much larger than its geometry, producing false positives, and points to the actual element geometry for resolving the criterion. This is documentation of a named API filter, rather than evidence that every Revit interference check or every clash product stops at boxes.

A staged distinction also appears in NVIDIA's PhysX 5.5.0 documentation: broad-phase box overlaps identify potential collisions, while the following narrow phase detects actual collisions in that physics pipeline. That example explains the vocabulary; it does not identify the implementation of a building-coordination tool.

Nor does the presence of AABB in a tool's name establish that its final answer is box-only. CGAL's AABB Tree user manual, displayed as version 6.2.1 when checked, describes a box hierarchy used to accelerate queries against underlying geometric primitives. Its documented intersection tests address those primitives. A bounding structure can organize a shape-aware query without replacing its geometric test.

For the buyer receiving a result, the consequential distinction is therefore the claim actually returned. An overlapping enclosure identifies a candidate. An assessment of the represented bodies can resolve whether those bodies intersect. A result whose test basis is unspecified leaves that distinction unanswered.

References: [A1](https://help.autodesk.com/cloudhelp/2026/ENU/Revit-API/files/Revit_API_Developers_Guide/Basic_Interaction_with_Revit_Elements/Filtering/Revit_API_Revit_API_Developers_Guide_Basic_Interaction_with_Revit_Elements_Filtering_Bounding_Box_filters_html.html), [N1](https://nvidia-omniverse.github.io/PhysX/physx/5.5.0/docs/RigidBodyCollision.html), [C1](https://doc.cgal.org/latest/AABB_tree/)

## BB-284: overlapping boxes around two disjoint bodies

BB-284 is constructed geometry in a common Cartesian frame. All coordinates and radii use one arbitrary length unit, u. Body R is a ring-shaped prism: an annulus with outer radius 5u and inner radius 3u, centered on the z axis and extruded from z = 0 to z = 1u. Body S is a solid square prism with x and y each from −1u to 1u and the same z interval. These definitions describe ideal bodies, not equipment dimensions or a software execution.

R's tight axis-aligned box extends from −5u to 5u in both x and y, and from 0 to 1u in z. S fills its own box, extending from −1u to 1u in x and y and from 0 to 1u in z. S's entire box lies inside R's box. Their boxes therefore overlap throughout S's volume.

Now examine the bodies. At every point of S, x² + y² is at most 1u² + 1u² = 2u². Its greatest radial distance from the z axis occurs at a square corner and equals √2u, approximately 1.4142u. R contains material only at radial distances from 3u to 5u. Because √2u is less than 3u, every point of S lies within the empty central hole, throughout the shared height. The two bodies have no common point.

This is a box-overlap positive and a body-intersection negative for exactly the stipulated geometry. There is no contradiction: the box around R includes its hole, although R itself does not. If R were replaced by a solid cylinder with the same outer radius and height, its box would be unchanged, but it would contain S. Identical outer boxes can therefore accompany opposite body-intersection answers.

## Orientation changes how tightly a world-axis box encloses a shape

The hole in BB-284 is one source of unused space inside a box. Orientation supplies another. Consider a separate ideal cuboid with horizontal dimensions 8u by 2u and height 1u. With its long and short sides parallel to the chosen x and y axes, its tight axis-aligned box coincides with the cuboid.

Rotate that cuboid 45 degrees about the vertical axis while keeping the reference axes fixed. Each horizontal span of its new tight box is 8u/√2 + 2u/√2 = 5√2u, approximately 7.0711u. The box's horizontal area is now 50u², although the cuboid's horizontal area remains 16u². The additional enclosed area is empty space outside the rotated rectangular footprint. The cuboid itself has not changed size.

More generally, rigidly rotating a complete arrangement preserves which bodies intersect each other. Their axis-aligned enclosures can nevertheless change relative to the fixed axes, so candidate box overlaps can change. A different number of box candidates does not, by itself, prove that the underlying arrangement gained or lost body intersections.

A box allowed to follow an elongated object's orientation can enclose it more tightly. That does not solve every enclosure problem: a single rectangular box around a ring still includes its central hole. The useful question is how closely the chosen enclosure follows occupied geometry for the particular screen, rather than whether boxes are universally accurate or inaccurate.

## The represented shape is still separate from the physical requirement

Resolving a box candidate against geometry answers a narrower question than whether equipment fits its intended use. First identify what geometry the assessment treats as the body. A simplified solid that fills a real recess can yield a different answer from a representation preserving that recess. A missing projection can omit a conflict. Changing from boxes to a more detailed test cannot restore geometry that was never supplied.

Geometry fidelity and coordinate placement are separate dependencies. A tessellated approximation raises a question about the shape represented; a misplaced object raises a question about where that shape is being compared. Neither question is resolved merely by establishing that the reported result came from a shape-aware test. This article assumes the compared geometry has already been placed in a common frame and does not supply a mesh-accuracy or placement method.

Also distinguish a solid-body question from a boundary-only question. One solid can sit wholly inside another without their boundary surfaces crossing, while their occupied volumes still overlap. A returned surface-intersection result should therefore retain its stated meaning rather than silently become a conclusion about solid-volume exclusion.

Finally, required space can extend beyond modeled material. A separately defined service or movement envelope expresses a requirement that an equipment body's geometry need not contain. A body-intersection negative cannot establish that an absent envelope is satisfied. The responsible equipment and project sources must define that requirement and the appropriate review. No tolerance, clearance distance, route acceptance or work authorization follows from the enclosure mathematics.

## Use a candidate result to direct review before changing the layout

Suppose a coordination return flags two equipment objects because their enclosing boxes overlap. The supported immediate decision is to examine that pair further. Redesigning the layout solely from the flag would treat possible intersection as confirmed intersection. Dismissing it solely because boxes sometimes produce false positives would make the opposite mistake: the candidate could represent a real conflict.

A useful resolution distinguishes three outcomes in ordinary language: the pair remains a box candidate; the relevant represented bodies intersect; or an appropriate assessment found those bodies disjoint. These describe different evidence states. The finding should identify which comparison it resolves so the recipient can understand why the candidate was retained or closed.

BB-284 can be closed as disjoint by its explicit geometric proof. An actual equipment pair needs evidence applicable to its own geometry and the commissioned question; the ring example supplies no shortcut for dismissing it. Equally, a shape test that resolves one pair does not decide an unexamined service envelope or another omitted object.

The existing clash-review versus cable-length article addresses the separate purchase decision between coordination and complete cable requirements. Here the decision comes earlier: establish what a geometric positive means before using it to change the arrangement or describe the equipment as colliding.

## Carry the resolved geometry question into physical planning

Leviathan's Physical Infrastructure Planning & CAD service is a starting point for discussing agreed rack layouts, cable pathways and installation documents. Use the related service and project inquiry to describe the physical planning decision, the nonconfidential source information available and the outputs needed.

The authoring software, delivery resources, reviewer and output formats are confirmed for the engagement. A discussion of this article does not establish a Revit clash-analysis service or licensed design authority. Where specialist geometric review is needed, identify its responsible reviewer and agreed result before that result becomes an input to the physical planning package.

References: [L1](https://www.leviathansystems.co/services/engineering-design)

## Primary sources

- [A1 — Autodesk Revit 2026 API Developer's Guide: Bounding Box Filters](https://help.autodesk.com/cloudhelp/2026/ENU/Revit-API/files/Revit_API_Developers_Guide/Basic_Interaction_with_Revit_Elements/Filtering/Revit_API_Revit_API_Developers_Guide_Basic_Interaction_with_Revit_Elements_Filtering_Bounding_Box_filters_html.html). Checked 2026-10-20. Documents the named filters, world-axis-aligned Outline input and false-positive limitation. Useful for interpreting that API screen; it does not describe every clash tool's final test.
- [N1 — NVIDIA PhysX 5.5.0: Rigid Body Collision — Broad-phase Collision Detection](https://nvidia-omniverse.github.io/PhysX/physx/5.5.0/docs/RigidBodyCollision.html). Checked 2026-10-20. Provides a named example of potential box collisions followed by narrow-phase collision detection. Used to explain staged assessment, without equating a physics pipeline with CAD coordination.
- [C1 — CGAL AABB Tree User Manual, displayed version 6.2.1](https://doc.cgal.org/latest/AABB_tree/). Checked 2026-10-20. Introduction and Interface distinguish an accelerating box hierarchy from queries against geometric primitives. The latest URL can change version; this reference records the version displayed when checked.
- [L1 — Leviathan Systems: Physical Infrastructure Planning & CAD](https://www.leviathansystems.co/services/engineering-design). Checked 2026-10-20. Defines the physical planning and documentation offer, with authoring capability, resources, reviewer and outputs confirmed for each engagement.

## Related Leviathan resources

- [Clash review or cable-length engineering: choose the deliverable](/articles/gpu-cabling-clash-review-vs-length-engineering)
- [Physical Infrastructure Planning & CAD: discuss the package](/services/engineering-design)
- [Discuss a physical planning project](/contact?service=engineering-design&from=%2Farticles%2Fcad-bounding-box-overlap-versus-shape-intersection)
- [A Smooth CAD View Can Hide a Faceted Equipment Model](/articles/cad-brep-surfaces-versus-tessellated-mesh)
- [A GPU Depreciation Charge Is Not Another Equipment Payment](/articles/ai-hardware-depreciation-expense-purchase-cash-flow)

## Discuss the Physical Planning Decision and Required Outputs

Share the nonconfidential equipment context, available plans, unresolved geometry question and requested documents so we can discuss scope, review responsibilities and availability.

[Discuss the scope](/contact?service=engineering-design&from=/articles/cad-bounding-box-overlap-versus-shape-intersection)