An Open Conformance Suite for Pipe, Cable, and HVAC Routing
ISO 15649, AS50881, ASHRAE — every routing standard is cited by every vendor and tested by no one in public. NeuroCAD proposes a 10-test open conformance suite to close the gap.
ISO 15649 and AS50881 are the rules. There is no open conformance suite to test against them. Until now.
The standards everyone cites and nobody tests in public
Walk into any process plant design office and ask which routing standard their CAD setup honours. You will hear the same three letters and numbers back: ISO 15649 for piping in petroleum and natural gas industries, SAE AS50881 for cable harness installations on aerospace vehicles, ASHRAE Handbook — Fundamentals for HVAC duct sizing. Ask a follow-up: how do you verify that the CAD seat is actually applying the rule? The answers shrink to three patterns: “the vendor says so”, “we trust the routing module”, “we eyeball the result and call it good”.
That is not conformance. That is a marketing claim wrapped in goodwill.
This is not a complaint about any specific vendor. AutoCAD Plant 3D, Bentley OpenPlant, AVEVA E3D, CADprofi, CATIA Electrical, Mentor Capital — every commercial routing tool in the market ships with rule sets that are tied to the standards by name. The gap is structural: none of these tools come with a published, reproducible test pack that demonstrates rule conformance on documented inputs. The supplier writes “ISO 15649 compliant” on a feature page and the buyer takes it on faith.
We can do better than faith. This post is about the open conformance suite NeuroCAD is proposing for routing standards — what it tests, why each test matters, and why open-sourcing the suite, not just the kernel, is the part that moves the industry.
Why the gap exists
Three forces converge to keep routing-rule conformance opaque:
1. The standards themselves are paywalled. ISO 15649 costs roughly CHF 200 from the ISO store. SAE AS50881 is behind SAE’s subscription. ASHRAE Handbook chapters cost about USD 200 each. The communities that could write open test suites — universities, hobbyist EDA users, open-source CAD contributors — face a cost wall before they even read the spec. Vendors who can afford the paywall keep their interpretation of the spec inside the product.
2. Routing is “configuration”, not “geometry”. Booleans, fillets, tessellation — these are core kernel features and benchmarks exist for them. Routing rule application is treated as a configurable knowledge base on top of generic 3D modelling. When the rule engine ships as a database of bend radii and clearances, vendors treat the database as a competitive moat, not a specification under test.
3. Routing failures are silent. A bend radius below spec doesn’t crash the CAD seat. It produces a deliverable that looks right on screen and reveals itself only at the kink in a copper tube during commissioning, or at the broken conductor after the first thermal cycle on the harness. The test that would have caught it would have run in milliseconds at design time. It just didn’t run.
The combination — paywalled spec, vendor lock, silent failure — is exactly the kind of conformance gap that an open-source community can close.
What the open suite covers
The proposed suite is ten tests. They are derived from the relevant standards and from the routing scenarios that recur across process-plant, aerospace harness, and HVAC design. Each test takes a documented input, applies the standard’s rule, and asserts a numeric or set-membership outcome that anyone can reproduce.
Here is the full list, with the standard each one anchors:
- Pipe minimum bend radius enforcement (ASME B31.3 process piping; ISO 15649 process plant context). Asserts the router never emits a segment with bend radius below the per-pipe-class minimum — for example 1.5 × OD for steel, 3 × OD for copper soft annealed.
- Pipe-pipe / pipe-equipment clearance (ISO 15649). For a routed network, asserts no segment passes within
clearance_minof any other segment or equipment. NeuroCAD’s SDF representation makes this a single min-distance query against the union-of-pipes field; in B-rep CAD it is a quadratic-time pairwise interference check. - P&ID ↔ 3D routing closure (ISO 10628-2). Every P&ID line tag must be present in the 3D model with matching pipe spec; orphaned tags fail the test.
- HVAC duct sizing — ASHRAE equal-friction method (ASHRAE Handbook — Fundamentals 2021, ch. 21). Verifies the duct cross-section the tool selects matches the table value for given CFM and target friction loss per 100 ft.
- Cable harness AS50881 minimum bend radius (SAE AS50881). Asserts harness bend radius ≥ 10 × OD per AS50881, with the documented exceptions for coax (6 × OD) and small-gauge wire (3 × OD), at every routing node.
- Cable bundle slack tolerance zone (industry harness practice). Verifies the routed harness has spec-compliant slack at clamp transitions to absorb thermal expansion.
- Pipe routing algorithm shortest-path optimality (CAD literature on automatic harness layout). For a benchmark obstacle field, asserts the router finds a path within 5 % of the shortest feasible path.
- Variable-topology routing under suppress / unsuppress (NeuroCAD architectural pin). Suppressing one of three parallel pipes preserves the SDF lawfulness and the segment IDs of the remaining two — a property that B-rep routing tools cannot satisfy without extra topology repair.
- Routing → NeuroFEM thermal-flow handoff (ASME B31.3 thermal expansion). The routed network exports a 1-D thermo-fluid mesh with conserved cross-sections at junctions.
- P&ID → AsBuilt redline traceability (ISO 10628). Edits to the 3-D route surface as redline edges in the P&ID’s NeuroGraph view, so commissioning data flows back into the design record.
The suite is biased toward the rules that fail silently and bite later: bend radius (#1, #5), clearance (#2), sizing tables (#4). Tests #3, #8, #10 cover the data-integrity rules that make the routing model trustworthy as a single source of truth.
What “open” means here
There are three things that are typically called “open” in CAD, and they are not the same thing.
Open source kernel. FreeCAD’s piping workbench is open source. So is OpenSCAD. So is Solvespace. The source is readable; the rule database, where it exists, is also readable. This is the form of openness that has existed for a decade.
Open standard. ISO 10628 is an open international standard in the sense that any vendor may implement it; the document itself is paywalled but the contents are stable and citable.
Open conformance suite. This is the missing piece. A conformance suite is the set of input artefacts, expected outputs, and pass/fail criteria that determine whether a given implementation honours a given rule. For routing, no such suite exists in public. The Open Source Hardware Association (OSHWA) has published whitepapers on the broader question of certifying open-hardware processes, but none of the routing standards has an OSHWA-style public conformance pack.
NeuroCAD’s proposal is to publish the third one. The kernel and the routing workbench remain part of the commercial NeuroCAD platform. The suite — input fixtures, expected results, pass/fail asserts, and a CI runner — is to be released under a permissive open-source licence on a separate repository, with semantic versioning and a governance process that lets non-NeuroCAD CAD tools register conformance results.
Comparison with what the market ships
We don’t have access to the proprietary internals of AutoCAD Plant 3D, Bentley OpenPlant, AVEVA E3D, or CADprofi. What we can compare is what each tool publishes about its routing-rule coverage and whether a third party can reproduce the claim.
| Capability | NeuroCAD (proposed) | AutoCAD Plant 3D | Bentley OpenPlant | CADprofi | FreeCAD Piping |
|---|---|---|---|---|---|
| Routing rules tied to ISO 15649 | Yes | Yes (claimed) | Yes (claimed) | Partial | No |
| AS50881 harness bend-radius rule | Yes | No | No | No | No |
| ASHRAE equal-friction sizing | Yes | Add-on | Add-on | Yes | No |
| Public test inputs | Yes | No | No | No | No |
| Public expected outputs | Yes | No | No | No | No |
| Reproducible CI pass/fail | Yes | No | No | No | No |
| Third-party conformance registration | Planned | No | No | No | No |
The columns most of the industry sells on are columns 1 to 3 — claims of standard coverage. The columns nobody currently sells on are 4 to 7 — evidence of standard coverage. That is the column the open suite fills.
We should be honest about what this comparison doesn’t say. AutoCAD Plant 3D is a mature, deployed-everywhere tool with a vast library of pipe specs and equipment templates that NeuroCAD does not have today. Bentley OpenPlant has integrations with plant operations and lifecycle systems that are years of work to match. The point is not that an open conformance suite makes those tools obsolete. The point is that the suite gives any of those tools — and any new entrant — a way to publicly demonstrate the rule conformance the marketing pages claim.
Why this matters to the engineer at the desk
If you are the process engineer signing off a piping isometric, the open suite gives you something you currently cannot get: a way to prove the deliverable was checked against the rule, and to inspect the check itself. If the design was routed in NeuroCAD and the suite ran in CI, the conformance result is a versioned artefact. If it was routed in another tool, you can run the suite against the exported model — pipe centrelines, bend radii, clearance buffers — and get the same conformance verdict. The suite is the rule made executable; it does not depend on which tool the design came from.
If you are the harness engineer working to AS50881, the same applies, with one extra benefit: AS50881’s bend-radius rule has more exceptions than ISO 15649’s — coax versus normal versus small-gauge versus shielded — and getting any one of them wrong produces a part that passes visual inspection and fails in service. A test that runs at design time costs nothing. A failure at integration costs schedule.
If you are buying CAD seats, the suite is a procurement instrument. “Demonstrate routing-standard conformance against the open suite” becomes a contract clause you can write and verify.
Status
The Routing workbench in NeuroCAD is currently a placeholder. The kernel can represent pipe centrelines, the SDF representation supports the union-of-pipes geometry the clearance test needs, and the DCG can re-evaluate the routing graph under variable topology. The ten tests above are documented in the engineering test catalogue but are not yet implemented; the conformance suite repository does not yet exist. We are publishing the design now to invite review before the implementation lands, because a conformance suite is a community artefact and the worst outcome is one vendor shipping a “conformance” pack that nobody else can sign off on.
When the suite ships, it will be open-sourced under a permissive licence on a separate repository, with public CI runs and a registration process for third-party tools. Until then, this post is a public design document.
Comparison numbers for AutoCAD Plant 3D, Bentley OpenPlant, CADprofi, and FreeCAD in the table above reflect publicly published feature claims; we have not run any of those tools against the proposed suite.