Feasible is not the question
Put a small domestic water pipe through a large supply duct and ask a coordinator which one moves. The answer comes back before you finish the sentence: the pipe. Not because the duct cannot move. In most ceiling voids there is room to drop the duct too. The pipe moves because moving it is cheap and moving the duct is not.
That instinct is easy for a person and invisible to geometry. A router that only asks "is there a path?" finds two answers here, and both clear the clash. The part that matters is what each change costs and who has to re-check something afterwards.
For years the industry answer has been a give-way list by category: conduit and cable tray give way to pipe, pipe to duct, small duct to large duct, everything to gravity drainage. ClashFree used one too. It gets the common case right and fails in exactly the cases that cost money. It cannot tell a 300 mm chilled-water main from a 50 mm pipe, a sprinkler branch from a domestic one, or a 900 mm cable tray from a 100 mm one. And it says nothing about what the change will cost once you make it.
Every run has a system class
So ClashFree replaced the list with a table of system classes. Every pipe, duct, conduit and tray in your model resolves to one, from its category, its system classification and its size, and, where Revit has no classification for it, from words in the system name (a "smoke" duct, a "steam" pipe). Each class carries three kinds of knowledge:
- Change costs. What a typical change on this kind of run disrupts, and multipliers on the cost of each fitting, each millimetre of added length and each new high or low point.
- Sensitivity tags. Which changes need an engineer's re-check:
pressure_losson ducts,hydraulic_calcon sprinklers,head_losson water,fallon sloped lines,ratedon life-safety systems,bend_radiuson cable tray. - Hands off. Whether ClashFree may move it at all. In the base table only gravity drainage is hands off; a project can add more.
| Class | Change cost | Fittings cost | Re-check |
|---|---|---|---|
| Gravity drainage | hands off | never moved | fall |
| Smoke extract duct | 3 | 2.5 × | pressure_loss, rated |
| Kitchen exhaust duct | 3 | 2.5 ×, high/low points 3 × | pressure_loss, grease, rated |
| Medical gas pipe | 3 | 1 × | rated |
| Sprinkler pipe | 2.5 | 2.5 ×, high/low points 2 × | hydraulic_calc |
| Air duct (supply, return, exhaust) | 2 | 2 × | pressure_loss |
| Condensate, steam | 2 | 1 × | fall |
| Hydronic pipe (heating, chilled water) | 1.5 | 1 ×, high/low points 1.5 × | head_loss |
| Domestic water, other pipe | 1 | 1 × | head_loss |
| Cable tray | 1 | 1.5 × | bend_radius |
| Conduit | 0.5 | 0.5 × | bends |
These are start values from general practice, shipped as such and calibrated against real fixes. They are data, not prompt text: you can read them, a project can override them, and the evaluation suite guards them.
Size scales the cost
A class says what kind of run it is. Size says how much of it you are moving. ClashFree multiplies the class's change cost by a size factor: the run's size divided by 200 mm, held between 0.5 and 3. A 50 mm pipe sits on the floor at 0.5; a 600 mm duct or larger sits on the ceiling at 3.
disruption = class change cost × clamp(size ÷ 200 mm, 0.5, 3)
When two runs clash and both may move, the one with the lower disruption gives way. On a tie, the smaller run moves; then the class that costs less at any size; then a stable key from the model. The order of rows in your clash report never decides.
Who gives way, in six pairs
Here is the rule at work on pairs a category list gets wrong as often as right. The costs are ClashFree's base values, computed with the formula above.
| Clash | A costs | B costs | Moves |
|---|---|---|---|
| DN50 domestic water × 800 mm supply duct | 0.5 | 6 | the pipe, by 12 × |
| DN300 chilled-water main × 200 mm supply duct | 2.25 | 2 | the duct: a big main outweighs a small duct |
| DN150 sprinkler main × DN65 domestic water | 1.88 | 0.5 | the domestic pipe: no hydraulic re-calculation |
| 900 mm cable tray × 250 mm supply duct | 3 | 2.5 | the duct: a wide tray is costly to bend |
| 25 mm conduit × DN40 heating pipe | 0.25 | 0.75 | the conduit, almost always |
| DN100 sanitary drain × anything | hands off | — | never the drain; the other side, or a flag |
The second and fourth rows are the interesting ones. A category list says "pipe gives way to duct" and "tray gives way to duct", and both answers are wrong for these sizes. Rerouting a 300 mm insulated chilled-water main means new supports, more insulation and a pump-head question; dropping a 200 mm branch duct is two fittings. The table knows that because size is in the arithmetic.
The reason is written down
Every group in the ClashFree list carries its reason in plain text, from the same numbers:
cheaper change: 50 mm domestic water pipe costs 0.5, 800 mm air duct costs 6 (pressure_loss)
The AI receives the class facts of both runs (id, tags, costs and where each value came from) in its context, so it does not have to remember HVAC practice or invent it. Its job is to choose and explain, citing what decided: "pipe moved: a duct reroute adds a pressure_loss re-check". It may not claim that the airflow is fine afterwards; it may only repeat the tag. ClashFree simulates nothing.
At import, each clash pair is classified and the cheaper change is chosen as the mover, with the reason shown on the group. When you open the clash, the real options for that mover are listed and scored with the class's factors, so a duct's fittings cost twice a pipe's in the score too. Gravity drainage and any class your project marks hands off are never selected to move.
When the cheaper side is another trade's
ClashFree only edits runs of your trade in the host model. Linked models and other trades' elements are read-only obstacles. So when the cheaper change is on the other side, for example you are HVAC and the cheap mover is a Plumbing pipe, the group does not land in your To fix list. It goes to Coordination, with the reason, and the AI queue will not send it for your trade.
That is the right outcome: the plumber should drop the pipe, not you the duct. If you decide otherwise, you can edit the group and pick the other mover by hand. What ClashFree will not do is swap the mover silently because one side has no room; that case is flagged, naming the other side.
Tuning it for your project
Projects differ. On one hospital every sprinkler change needs the fire engineer's sign-off; on another the client names smoke-extract systems in their own language. The coordination profile has a classes section for that, in three layers: the base table, your project's file, and your own preferences.
- A project class with a built-in id changes only the members it gives:
"sprinkler_pipe": {"hands_off": "flag_owner"}makes sprinklers hands off and keeps everything else. - A new class needs a match that narrows (a kind, a classification, a name word or a minimum size); one that would take every run is refused when the file is read.
- Your user preferences can change costs, never hands off. A preference never loosens what the project made hard.
Every value the AI sees carries its source, so its explanation can say "sprinkler pipe is hands off (project)".
What it does not do
The estimate at import is class-based; it does not run the router for both sides of every clash. When you open a clash, the real options and their scores replace the estimate for the moving side. If the estimate picked wrongly, say the pipe has no room at all, you see a flag naming the other side and its cost, never a silent wrong move.
And it is not an engineering calculation. A pressure_loss tag means "an HVAC engineer should look at this", not "the fan can take it". That is the line between coordination knowledge, which belongs in the tool, and engineering judgement, which stays with the engineer.
Read next: how the mover's options are scored, term by term, in Five valid fixes, one right answer, or try it on your own model in the free beta.