Gate A Review
How to run a Gate A modelling review for large commercial HPWH projects, covering AI-assisted checks and manual peer review.
Gate A is a modelling consistency review. It runs twice: once after the first complete system is modelled, and again after all systems are complete. The purpose is to catch modelling errors before they propagate across every variant in the application.
Gate A is not a submission check — it is a modelling quality check. Audit file preparation does not begin until Gate A has passed.
When Gate A runs
| Stage | Trigger |
|---|---|
| Gate A1 (initial) | First system fully modelled (CPL found, results reviewed, client approval pending or received) |
| Gate A2 (final) | All remaining systems modelled to the same standard |
If the initial Gate A1 finds issues, fix them before modelling any other systems. Carry-forward errors are much more expensive to fix after all systems are done.
Running the Claude Gate A Review
Use the Large Commercial Gate A Review tool in Claude Projects.
The Claude review checks modelling consistency systematically across the following areas:
Performance map
- EN 14511 test data entered into ComHPdata is complete and matches the raw test report (temperatures, capacities, COPs)
HPdata.txtgenerated correctly; normalisation workings exported asHPdata.xlsx- Performance map coverage figure shows simulation operating hours fall within (or close to) the test matrix
HP parameters (FAM_COM sheet)
- Rated capacity, standby power, and refrigerant charge sourced from the EN 14511 report or data plate, with source reference noted
- Control settings (setpoint, deadband, legionella temperature and frequency) match the installation manual and any client-confirmed overrides on the PIC
- Flow rate and pump power consistent with pump performance data
TRNSYS template
- Template selection (parallel/series, internal coil, fixed/variable flow) matches the system schematic
- Number of HP and tank units matches the PIC
Tank parameters
- Tank volumes, fitting heights, and insulation values match the tank drawings
- Tank heat loss value sourced from the Tank Heat Loss Calculation or from test certificate
Simulation outputs
- MDT ≥ 45 °C and energy savings ≥ 60 % for all three climate zones (Z3, Z4, Z5)
- CPL values in the Com_Results sheet match the Find Loads output
- VEEC/ESC quantities are plausible for the system size
Save the Claude report PDF to the Gate A folder (/Projects/AExxx/03-Gate-A/ClaudeReport_SystemName_YYYYMMDD.pdf) before requesting peer review.
Manual peer review — what AI misses
The peer reviewer reads the Claude report and then checks the things Claude cannot assess from data fields alone. These are the failure modes that recent projects have demonstrated.
Tank drawing provenance
Confirm the tank drawing is the client’s own dimensioned engineering drawing — it should carry a client logo, document number, and date. A reference image, product brochure extract, or unlabelled sketch does not meet this requirement. Claude can note the file is present; it cannot judge whether it is the right kind of document.
Schematic readability and accuracy
Review the system schematic as a human reader would. Does it clearly show the pipe connections, the HP position relative to tanks, the flow direction, the sensor locations? A schematic that is technically present but unreadable or inconsistent with the TRNSYS template is a risk at audit.
Installation manual coherence
Read the installation section of the manual. Does the installation described match the system as modelled? Check specifically:
- Is the manual for this exact model, or a related but different product?
- Does the control section describe setpoints and deadbands consistent with what was modelled?
- Are the flow rates and pump specifications consistent with the FAM_COM entries?
Model-to-system correspondence
Does the TRNSYS configuration actually reflect the physical system in the schematic? This requires holding both in mind simultaneously. Common mismatches: tank connection order in TRNSYS vs. schematic, HP serving a different subset of tanks than drawn, internal coil HX in schematic but not modelled (or vice versa).
EN 14511 test conditions representativeness
Are the test conditions in the report representative of the modelled application? If the test was conducted at a non-standard inlet temperature or with non-standard flow, flag for director review before proceeding.
Cross-document consistency requiring judgement
Some consistency issues only become apparent when documents are read together rather than checked field-by-field. Look for: brand name inconsistencies across schematic, manual, and test report; model designations that differ by one character between documents; document dates that suggest the manual was written for a prior product generation.
Handling failures
If any check fails — whether from Claude or from manual peer review:
- Stop the review and return to the project lead with a written description of the specific issue (file, field, discrepancy).
- Project lead fixes the issue, re-runs the Claude self-review, and confirms the fix is resolved.
- Resubmit to peer review. Do not proceed until the peer reviewer has signed off the corrected version.
Do not accumulate a list of issues and fix them in batch after review. Each issue must be fixed, re-reviewed, and confirmed before the next system is modelled.
What to file
| Item | Location |
|---|---|
| Claude Gate A report PDF | /Projects/AExxx/03-Gate-A/ClaudeReport_SystemName_YYYYMMDD.pdf |
| Gate A Excel spreadsheet (peer reviewer) | /Projects/AExxx/03-Gate-A/GateA_AExxx_YYYYMMDD.xlsx |