Client case · plant engineering · utility-systems verification
Verifying the utility systems of a GMP plant: where the tender documents diverged from physics
An independent check of the utility systems of a new GMP plant before the tender: recomputing the chiller load and the air balance, and reconciling the proposed equipment against the technical specification — before the contract is signed. The client, site and region are under NDA; here we show the task, the method and the result.
In brief: what we checked and what we found
The client was preparing a tender for the utility systems of a new multi-product pharmaceutical plant (aseptic and solid dosage forms) and had in hand a concept-level load calculation, the technical specification, and suppliers' bids for the chillers and air-handling units. We ran an independent recalculation on the thermophysical library CoolProp and reconciled the proposed equipment against the spec.
- The recalculation exposed a double-count of about 137 kW in the original load calculation and brought the base chiller load down to 773.6 kW.
- The proposed 3 × 473.6 kW arrangement holds N+1 redundancy only at zero process heat gains (a +22.4% margin) and fails once a further +174 kW is added.
- The equipment review produced 22 deviations from the spec, 6 of them critical: the glycol solution freezing at −16.3 °C against a design winter of −28 °C, a refrigerant with a global-warming potential 2.27 times higher than required, unit masses understated by 37%, and the air-change rate in the aseptic zone falling to 44% of the norm.
- The result was negative with respect to the original documents — and precisely that gave the client a negotiating position with the suppliers before the contract.
The starting situation
The client was going to tender for the utility systems of a new GMP plant — a complex with zones of differing cleanliness: Grade B aseptic (syringe filling), Grade C (microbiology), Grade D (general production, dry solid forms, packaging) and a controlled non-classified area. The heating, ventilation and air-conditioning of these zones, plus the central cooling supply, are a critical utility subsystem: both GMP air-cleanliness compliance and the plant's ability to run through winter depend on it. In hand the client had a concept-level chiller-load calculation (the summary report of a previous revision), the technical specification, and suppliers' bids for three chillers and six air-handling units. Our task was an independent check: whether the load calculation adds up, whether the proposed per-unit chiller capacity is justified, and whether the equipment meets the spec's requirements — before the tender is closed and the contract signed.
The principle of the check
Every number is either recomputed (by formula or in CoolProp), taken from the client's documents, or flagged as an assumption or an open question that needs data. Nothing is taken on faith or filled in by guesswork.
What the analysis showed: a double-count in the load calculation
The first thing the recalculation exposed was a methodological error in the previous revision. The load on the AHU cooling coils was added to the internal heat gains in such a way that part of the heat was counted twice. Cooling the whole mixed stream to the coil-outlet parameters (12 °C, 90%) already includes removing the recirculated internal heat gains via the return air; adding those same internal gains again as a separate line is not permissible. The double-count came to about 137 kW. The correct total is obtained by accounting for them separately: the fresh-air load by enthalpy, plus the internal heat gains as a separate line.
| Component | kW | Source |
|---|---|---|
| Load on the AHU cooling coils (fresh air, by enthalpy) | 303.6 | CoolProp: 25 °C → coil outlet 12 °C/90% |
| Internal heat gains | 259.0 | assumption: 120 W/m² × 2,160 m² |
| Building envelope | 28.0 | previous revision |
| Aseptic isolator | 10.0 | previous revision |
| Subtotal of components | 600.6 | sum |
| Reheat after dehumidification + fan heat (+12%) | 72.1 | computed |
| Reserve (+15%) | 100.9 | computed |
| Base load (process = 0) | 773.6 | needs data: process heat gains not included |
N+1 redundancy holds on the edge
The proposed arrangement is three chillers of 473.6 kW each (installed capacity 1,420.8 kW). Under N+1 redundancy (one unit on standby) two machines are available — 947.2 kW. Against the base load of 773.6 kW that is a +22.4% margin. But the base load does not include process heat gains from equipment (reactors, autoclaves, freeze-dryers, compressors) — these are not computed in the input data. As soon as the process adds load, the margin melts away.
| Scenario | Total load, kW | N+1 margin (947.2 kW available) |
|---|---|---|
| Base (process = 0) | 773.6 | +22.4% |
| +100 kW of process | 902.4 | +5.0% |
| +200 kW of process | 1,031.2 | −8.1% (reserve exhausted) |
| N+1 breaking point | +174 kW of total load above base |
Conclusion on the load
The per-unit capacity of 473.6 kW (above the 450 kW in the specification) is a justified choice: it gives a margin over the base. But until the real process heat gains are obtained from the process engineer, one cannot claim that N+1 will hold the plant under full load. This is open question #1, and it bears directly on the final capacity of the chiller plant.
Twenty-two equipment deviations from the spec
Reconciling the suppliers' bids against the technical specification produced 22 recorded deviations: 6 critical, 9 high-priority, 5 medium, 2 low. The critical ones are those that either break the physics of operation or directly contradict the spec.
| Item | Spec requirement / calculation | Proposed / actual | Consequence |
|---|---|---|---|
| Glycol concentration vs climate | protection to the design winter of −28 °C | 35% by mass — freezes at −16.3 °C (CoolProp); the documents claim "to −20 °C" | Risk of the loop rupturing in winter. Raise to 44–45% or justify burst protection with flow and electric heating |
| Refrigerant | R-513A (GWP 631) | R134a (GWP 1430), charge 168 kg/unit | Warming potential 2.27 times higher; ~240 t CO₂-eq/unit vs 106 t; falls under the HFC quotas (2029/2034), risk of a service-gas shortage |
| Mass / foundation | project estimate ~4.5 t/unit | 6.9 t net / 7.19 t in operation — understated by 37% | For three units, 21.6 t vs 13.5 t in the calculation. Recompute the ground slab for 478 kg/m², bearing area 15.1 m²/unit. |
| Air-change rate, aseptic B | 18,250 m³/h (50 h⁻¹) | the AHU delivers 8,000 m³/h → 21.9 h⁻¹ | 44% of the norm. A room-by-room "room → unit" breakdown is required; without it the unit selection is unjustified |
| Process heat gains | include in the load | taken as zero (only the averaged 120 W/m²) | Open question #1: bears directly on N+1 and the final capacity of the chiller plant |
| Power-supply adequacy | confirm the substation capacity and cable cross-sections | peak 608 kW (N+1) … 847 kW (three units at maximum), current 1,054–1,464 A; equipment is 380 V, not 400 V | Needs data: check the transformer and cables against the peak current |
High-priority deviations
Among the high-priority deviations: a seasonal-efficiency figure of 4.826 against the required ≥5.5 (a 12.3% shortfall); a minimum outdoor-air temperature of −25 °C in the base build against the design −28 °C (a −40 °C winter build is available as an option); the absence of the anti-corrosion coating on the heat exchangers that the spec makes mandatory; and a gap in the air balance — the concept assumes 86,000 m³/h (six units), whereas the calculation by air-change rates gives about 103,000 m³/h — a 17% discrepancy, formally explained by offloading part of the load onto split systems, but not closed off by a balance.
The main takeaway
The proposed documentation was not ready to sign. The load calculation contained a double-count; N+1 redundancy rested on an unproven assumption of zero process heat; the proposed equipment deviated from the spec on refrigerant, frost protection, mass, efficiency and air changes in a critical zone. No single deviation was "fatal" on its own — but together they meant the contract had to be reopened with the suppliers, not signed.
The result
The check gave the client three things that were not in the original documentation:
- 00
A reliable load benchmark
A base chiller load of 773.6 kW with a transparent breakdown by component and an explicitly flagged open question (process heat) — instead of the previous revision's figure, overstated by ~137 kW.
- 01
A quantified redundancy limit
N+1 holds the base with a +22.4% margin but fails at +174 kW of process load. This turned uncertainty into a concrete threshold: below that figure the arrangement is reliable, above it it is not.
- 02
A list of 22 required actions for the suppliers
Prioritised — from changing the refrigerant and raising the glycol concentration to recomputing the foundation and a room-by-room breakdown of the air changes. Each item is a negotiating position fixed before the contract is signed, not a complaint after installation.
Risks averted
Direct and material: the hydraulic loop freezing in the very first design winter; the foundation falling short of the bearing capacity for units weighing 7.19 t against the project estimate of ~4.5 t; production stopping when N+1 is exhausted under real process load; and falling under the refrigerant quotas with the ensuing rise in servicing costs across the whole life cycle of the plant.
Why this is value, not criticism
Formally the result is negative: the check did not confirm that the documentation was ready — it found a double-count and two dozen deviations in it. The temptation is to read this as a demolition of someone else's work. In fact the point of the check is the opposite. Each deviation found on paper costs a few weeks of engineering time. The same deviation discovered on site costs incomparably more: a glycol loop burst by frost — a failure and downtime; a slab cracked under underestimated mass — demolition and rebuilding of the foundation; an N+1 shortfall under load — production stoppages in a GMP environment; a refrigerant quota — rising operating costs locked in for decades ahead. The check moved all these scenarios out of the "we'll find out in operation" category and into the "recorded in the negotiation table before the contract" category.
A check doesn't only criticise — it also confirms
The per-unit chiller capacity of 473.6 kW was found justified; the nominal efficiency coefficient (3.105 against the required ≥3.0) compliant; the free-cooling mode exceeding the spec; the noise level within the norm with margin to spare. The client received not a list of complaints but a calibrated map: what in the bid holds up and what needs further work with the supplier.
Caveats and data status
What was computed. The chiller load, the glycol-solution properties (freezing point, density, heat capacity), the coolant flow rate, and the psychrometrics of the AHU cooling coils were calculated on the thermophysical library CoolProp; the glycol-property results matched the data in the input-data package. The arithmetic of the load build-up and the N+1 limits were checked line by line.
- What is an assumption or an estimate. The internal heat gains were taken from an averaged empirical coefficient of 120 W/m² — the process heat gains from equipment were not computed separately (an open question that needs the process engineer's data, and one capable of adding on the order of 100–300 kW). The estimate of savings from free cooling and the annual operating costs are given as an order of magnitude, without an hourly climate profile. A number of hydraulic and electrical parameters (network heads, power factor) are flagged as assumptions.
- What was taken from the client's documents and the suppliers' bids. Unit masses, refrigerant charge and grade, efficiency figures, per-unit capacity, and the air-handling-unit characteristics come from datasheets and bids; they were checked for compliance with the spec but not measured on site. The soil bearing capacity and the site geology require separate data for the final foundation calculation.
- The calculation engine is CoolProp (properties of moist air and of the water–glycol solution). The climate parameters follow the building codes for the design region. The check was performed by a cross-role team (a ventilation engineer, a hydraulics specialist, a power engineer, a mechanical engineer, a process engineer and a fact-checker). The client, the active substance, the product, the site and the region are not disclosed, under a confidentiality agreement.
Need an independent recalculation before your tender?
Message us — we'll show the method and scope the check for your project: chiller load, redundancy, air balance and reconciling the proposed equipment against the spec before the contract is signed.
The case is anonymised. The client, the active substance, the product, the site and the region are under NDA. All numerical parameters (773.6 kW of base load, N+1 failure at +174 kW, 22 deviations from the spec and others) relate to a specific project and are quoted verbatim from the report; the calculations were performed on the CoolProp library, the climate per the building codes of the design region. The material is informational; it is not project documentation, a regulatory opinion or investment advice; applying it to another facility requires a separate engineering calculation.