--- Skip to content
Guide 2026-10-20 Guide

The 41-Rule Electrical Design Check: Catch Errors Before Construction

Most electrical errors are cheap on screen and expensive in a wall. A 16 A breaker on a 1.5 mm² cable costs nothing to fix while the drawing is open. Once the cable is pulled, the fix involves a chase, a plasterer and a conversation with the client about who pays.

Rule-based checking exists to move that discovery as early as possible. Not to replace the engineer — to make sure the boring, mechanical contradictions get caught by software that never gets tired, so the engineer's attention goes to the decisions software cannot make.

Author: Guide

What a rule check actually is

A rule is a condition over structured project data. It has three parts: a scope (which objects it applies to), a predicate (what must be true), and a message (what to fix).

"Every circuit protected by a breaker whose rating does not exceed the current-carrying capacity of its conductor" is a rule. It applies to circuits, it compares two stored values, and it fails deterministically. Run it a thousand times and you get the same answer. That is the property that makes it trustworthy: there is no model guessing, no plausible-sounding prose, no temperature.

This is the opposite of an AI review. A language model asked to "check this drawing" returns a confident list, some real and some invented, and you cannot tell which without checking every one. A rule engine returns fewer findings, each tied to an object, each reproducible. GorkyCAD keeps the two separate: 41 deterministic rules over the project model, with AI used for reading inputs and explaining results.

The families of rules

The 41 checks are not 41 variations of the same test. They cluster into families that follow the chain of an installation.

Family What it protects against Typical failure
Supply and protection Wrong breaker type or rating for the supply Breaker rated above the conductor it protects
Conductor sizing Undersized cable for load and installation method 2.5 mm² on a 32 A circuit
Voltage drop Excessive drop on long runs A garden circuit 60 m from the board
Residual current protection Missing or misrated RCD/RCBO Wet-area circuit without 30 mA protection
Earthing and bonding Missing protective conductor or bonding Metal bath or riser not bonded
Panel capacity and balance Overfilled board, unbalanced phases 14 single-pole devices in a 12-way enclosure
Circuits and loads Impossible or contradictory circuit data Load total exceeding the declared supply
Documentation Drawings that cannot be built from Circuit with no description, cable or destination

The set is written around ПУЭ and ГОСТ 21.613, the norms the product is built for. If your project is governed by IEC 60364, BS 7671 or the NEC, treat the findings as engineering prompts rather than a compliance verdict — the underlying principles overlap heavily, the clause numbers and specific limits do not.

Five findings worth recognising

These are the kinds of results a first pass produces on a real apartment or small commercial project.

Breaker above cable capacity. The most common finding by a wide margin, and usually a copy-paste artefact: a circuit was duplicated, the load was increased, the cable was not. The check compares the protective device rating against the conductor's capacity for the declared installation method.

Wet-area circuit without 30 mA protection. Bathrooms, shower rooms and outdoor sockets. Either the RCD is missing entirely or the circuit is on the wrong side of the split.

Voltage drop over the limit on the longest run. Frequently the last circuit added — a pump, a gate, a shed — routed around the building rather than through it. The rule uses the actual route length from the plan, not a straight line.

Panel over capacity. More ways assigned than the enclosure has, or no spare ways left for the client's future work. Both are cheaper to fix on screen.

Circuit without a complete description. No cable type, no destination, no load. This is the finding that most often gets dismissed as cosmetic and most often causes a site question three weeks later.

What the check cannot see

Honesty is more useful than a longer feature list.

It cannot see the site. Ambient temperature, grouping in a conduit, thermal insulation, the actual route above a suspended ceiling — these are inputs, and the rules can only reason about what you entered. A wrong input produces a confidently wrong pass.

It cannot judge intent. A 32 A circuit may be entirely correct. The rule flags the relationship between rating and conductor; it does not know whether the load is a welder or a lighting circuit with a design margin.

It cannot certify compliance. No software certifies an installation. The rules implement a subset of requirements in a form a machine can evaluate; the standard contains far more, including requirements about workmanship, inspection and testing that no drawing can satisfy.

It cannot cover what is not modelled. Control schematics, PLC I/O, fire alarm loops, lightning protection design and complex discrimination studies sit outside the model. If it is not an object in the project, no rule applies to it.

It cannot replace the reviewer. A design that passes all 41 checks is a design with no detected contradictions in the modelled data. That is a weaker and more useful statement than "compliant".

Running a check in the right order

The order matters, because later findings are often artefacts of earlier ones.

  1. Finish the input data first. Rooms detected, scale correct, nameplates known, installation methods set. Half of a noisy report is incomplete data.
  2. Run the full check once, then stop reading. Fix the structural findings — panel capacity, missing protection, wrong conductor — before touching anything cosmetic.
  3. Re-run. Some findings disappear as a consequence of the first fixes; chasing them individually wastes time.
  4. Read the rest with judgement. Decide, for each remaining item, whether it is a real defect, an accepted design decision, or a false positive caused by a modelling shortcut.
  5. Record the decisions. A one-line note on why a finding was accepted is worth having when the project is reviewed in six months.
  6. Export and archive the report with the drawing set. It is evidence of the review, not just a to-do list.

The whole cycle on a small project is minutes, not hours, which is the point: it is cheap enough to run at every revision rather than once before issue.

You can see the report and the objects behind each finding in the live demo. It is worth running on a project whose answer you know — that is the only reliable way to judge whether the rule set matches the way you design.

Reading the report without panic

Two failure modes appear when a team first adopts rule checking.

Treating every finding as a defect. A rule engine does not know about your conventions. If you model a sub-board as a single object, rules about circuits inside it will report oddly. Learn which findings are modelling artefacts in your template and fix the template, not the project.

Ignoring the report because it is noisy. The opposite mistake, and the more dangerous one. If the same rule fires on every project for a reason nobody remembers, the rule should be understood or the data fixed — not muted.

The healthy middle: a short, shared list of accepted exceptions, reviewed once a year, and a rule set that stays switched on.

Where this fits in the workflow

Rule checking is most valuable at two moments: after circuits are grouped and before the panel schedule is finalised, and again after any load change. Both are points where one edit propagates into several documents.

It is least valuable as a final gate five minutes before submission. At that point you are looking for a reason to delay, and the report will supply one whether or not the design is sound.

The tool is a review aid with a fixed, inspectable rule set. That is a modest claim, and it is why the results can be trusted: you can see which object failed which condition, and disagree with it in writing.

Ready to test it on your own project? Open the demo, import a plan and run the check. Pricing shows what the free tier includes and what Pro adds, and the documentation explains the rule set and the export formats. If you are comparing tool categories first, the electrical CAD software guide covers the wider market.

Sources

  • ПУЭ-7, Правила устройства электроустановок, 7th edition — full text (verify the edition in force in your jurisdiction): ruscable.ru
  • ГОСТ 21.613-2014, Правила выполнения рабочей документации силового электрооборудования: standartgost.ru
  • IEC 60364 series, Low-voltage electrical installations: webstore.iec.ch
  • NFPA 70, National Electrical Code (comparison reference): nfpa.org

This describes software behaviour, not a compliance certificate. Verify every requirement against the edition enforced in your jurisdiction.

What are the 41 rules in GorkyCAD?
They are deterministic checks over the project model, grouped around supply and protection, conductor sizing, voltage drop, residual current protection, earthing and bonding, panel capacity, circuit data and documentation. Each finding names the object it applies to, so you can inspect the data rather than trust a sentence.
Does passing all 41 checks mean the design is compliant?
No. It means no contradiction was detected in the modelled data. Compliance depends on the site, the installation method, the equipment actually installed, local amendments and inspection — none of which software can certify.
Are the checks based on NEC, IEC or ПУЭ?
The rule set is built around ПУЭ and ГОСТ 21.613. Many principles overlap with IEC 60364, but clause references and specific limits differ. For projects under another standard, use the findings as engineering prompts and verify against the enforced edition.
Can the check find errors in a scanned PDF plan?
Only after the plan has been imported and reviewed. If a room was not detected, no rule can reason about it. Check the imported geometry and the scale before running the rules — most false passes originate upstream.
Does the check replace a peer review?
No. It replaces the mechanical part of review: comparing ratings, capacities, counts and completeness. Judgement about intent, site conditions, discrimination and future load stays with a person who signs the design.
How long does a check take?
Seconds on a typical apartment or small commercial project, because it runs over stored data rather than rendering. The time is spent fixing findings and deciding which ones to accept, not waiting.

Design it in the browser

Turn the plan into circuits, a panel schedule, a single-line diagram and a calculated bill of materials. Start free or open the demo.

Open demo See pricing