Relief Valve Rating, an Expired Gauge, and a Certificate Due Today
You are the mechanical engineer of record for a 1.2 MW packaged glycol chiller supplied to a hospital. The evaporator shell has an MAWP of 10 bar g at 120 °C. The supplier's data sheet lists the relief valve as suitable for a maximum set pressure of 8.6 bar g at 120 °C, but the valve on the skid is tagged and set at 9.0 bar g. The hydrostatic test record shows 13.0 bar g, which is 1.3 × MAWP, but the pressure gauge's calibration sticker expired four months before the test was run. The supplier promises a revised data sheet after shipment. Liquidated damages are $18,000 per day, your firm's fee is tied to the job, and the project manager has asked you to sign the mechanical completion certificate today.
1. Separate the technical non-compliances from the professional-practice problems and rank them by severity.
Read the worked answer
Technical non-compliances
- Overpressure protection. The set pressure of 9.0 bar g is below the 10 bar g MAWP, so that part is fine. The problem is that the valve itself is rated only to 8.6 bar g at the coincident temperature of 120 °C. A relief valve set above its own rated limit may not reseat, may not pass its certified capacity, and may not function at all - so the shell's protection is not demonstrated, whatever the set pressure says. The certified relieving capacity versus the maximum possible heat input also has to be checked, and nothing in the pack shows it.
- Unverified pressure test. 13.0 bar g is the right number for a 1.3 × MAWP hydrotest with a stress ratio of 1.0. But the gauge that read it was out of calibration, so there is no traceable evidence the vessel actually saw 13.0 bar g. A test is not a test without a calibrated instrument.
- Retroactive documentation. A data sheet issued after shipment cannot make an installed, non-compliant valve compliant.
Professional-practice problems
Signing a certificate that attests to successful completion when you know two of the underlying records are unsound is a misrepresentation, regardless of who else is pressing. The fee dependency is a conflict of interest you must disclose and manage. Schedule and liquidated damages are commercial risks that do not change the engineering facts. Ranked: the protection gap first (it is a live safety issue), the unverified test second, the paperwork third - but the signature question outranks all three, because it is the point at which you would personally convert an open issue into a false statement.
2. State the disposition you would issue for each item before any certificate is signed.
Read the worked answer
Relief valve - Reject as installed. Either fit a valve rated for a set pressure of at least 9.0 bar g at 120 °C, or reset the existing valve to 8.6 bar g or lower and produce a certified capacity calculation showing it covers the maximum heat input with the condenser at the design condition. New tag, new set-pressure record, witnessed setting and sealing.
Hydrotest - Do not accept the existing record. Require a re-test at 13.0 bar g with a gauge calibrated within the last 12 months and with a known uncertainty (typically ≤1% of full scale, and reading in the middle third of its range), witnessed and signed by a second party. Confirm the test was run with the relief valve removed or blanked, that the metal temperature and water temperature met the code's minimum, and that the holding period was observed. Re-testing is cheap; arguing about it is not.
Documentation - Issue a written non-conformance to the supplier and copy the client. Hold the certificate until closure items are signed off.
Escalation - If the project manager instructs you to sign anyway, put the refusal in writing, state the technical basis, offer the compliant path and its schedule impact, and refer the matter to the engineering manager or the regulator if the instruction is repeated. Keep the record; it protects the client, the supplier and you.
3. The project manager says: "The valve will never see 9 bar anyway, so sign it and we can swap it later." How do you answer, and why?
Read the worked answer
The answer is no, and the reason is structural rather than personal.
A relief valve is a device that operates only in the failure case - loss of cooling water, a control valve failing open, a fire, a blocked discharge. The entire justification for its existence is that the normal control that keeps the vessel below MAWP has already failed. So "it will never see 9 bar" is not an argument that the protection works; it is an argument that the protection has never been called on. The one day it is called on is the day you cannot inspect it.
There is also a mechanical point: the valve's rating at coincident temperature is a property of its spring, seat and body. Exceeding it does not just reduce margin; it changes the failure mode from "lifts and reseats" to "lifts, chatters, or fails to reseat", which can leave the shell venting or over-pressurised. Neither outcome is recoverable by a later swap.
The commercial framing is worth answering directly: $18,000 per day of liquidated damages is a known, bounded number. An uninsured overpressure event on a hospital chiller is neither bounded nor insurable if the file shows the engineer of record knew about the rating gap and certified the skid anyway. Offer the practical alternative instead - a temporary valve of the correct rating can often be sourced in 48–72 hours, and the re-test can be scheduled in parallel - and let the manager own the schedule decision with the technical facts on the table.