Condenser condition. Flat & low is good — 16 °C means clean.
Suction & freeze margin — keep above 4.5 bar g.
High side. Rises with sea temperature; fine while approach stays low.
Delivered temperature against the 10 °C setpoint — whole plant, both systems.
The main driver of high-side pressure. Both systems draw the same water.
Approach = saturated condensing temp − sea-water outlet. The cleanest measure of how well this system's condenser sheds heat, with sea temperature taken out. Flat and low is healthy; a sustained climb toward 16 °C is fouling.
The shaded band is the gap between saturated condensing temp (top) and sea-water outlet (bottom) — that gap is the approach.
Sea-water temperature entering and leaving this system's condenser. The spread between them is how much heat the seawater is carrying away.
Low-side suction pressure. It sets evaporator temperature and the freeze margin. Below 4.5 bar g the plate evaporator is at freezing risk.
Degrees of subcooling at the liquid line. A few degrees confirms a solid liquid seal and adequate charge.
Temperature rise of the sea water across this system's condenser — the heat actually being rejected overboard.
Chilled-water supply temperature — what the plant delivers to the ship. Shared by both systems, so this chart is not filtered by the system selector.
Chilled-water return minus supply — the load the ship puts on the plant. Historically flat around 2–2.6 °C with one system running.
When both systems are logged within an hour of each other they are seeing the same sea water at the same moment. The difference in condensing approach between them is then a direct fouling comparison — sea temperature, load and the R449A glide convention all cancel out, because they are common to both.
Each dot is one reading: sea-water inlet (x) against approach (y), coloured by system. A clean condenser holds a flat low band whatever the inlet temperature.
Sea-water inlet (x) against HP (y), coloured by system. Healthy, this is a tight rising line — HP simply following sea temperature.
| bar g | °C | bar g | °C | bar g | °C |
|---|
Corrected Honeywell basis (45 °C ≈ 19 bar g). Mean saturated condensing temperature; R449A glide ~5 K gives ±~2.5 °C. High side only — this is what the approach calculation uses.
Approach = saturated condensing temp (from HP) − sea-water outlet. Same method as the live sheet, so the number matches the log to the decimal.
| LP bar g | mean evap. °C | coldest point °C | state |
|---|
The low-side table was correct from the start and is unchanged. Coldest point is the bubble-point end of the glide at the evaporator inlet — roughly the mean minus half of R449A's ~5 K glide. That is the coldest metal in the plate pack and the number that matters for freezing. See the Explained tab for the full reasoning.
Margin = LP − 4.5 bar g. The action below the line is always the same: raise the CHW setpoint. That lifts suction. Reducing chilled-water flow or forcing the compressor to unload makes the margin worse, not better.
Everything on this dashboard, explained from the beginning — what each reading is, how every number is worked out, what the limits mean and why they sit where they do. Written for someone seeing the plant for the first time. Open a section to read it.
The plant has two chiller systems. Pick one with the System 1 / System 2 switch at the top right. Almost everything you then see belongs to that system alone. The few things shared by both are marked PLANT; everything else is marked SYS.
The coloured bar at the top of the Status tab is the whole verdict in one line. Green means nothing needs doing. Amber means something is drifting and should be watched. Red means take action now. Below it, the panel headed What's going on writes out the reasoning in plain sentences, most urgent first.
If you read nothing else, read two numbers:
The other tabs are for looking deeper: Trends shows how each value has moved over time, Relationships strips sea temperature out so you can see the condenser's true condition, and Reference holds the refrigerant tables and two calculators.
Nothing here is generated by an AI at the moment you look at it. Every sentence comes from a fixed set of rules applied to the latest logged numbers, so the same reading always produces the same words.
The ship makes chilled water at about 10 °C and circulates it to the air-handling units. Two independent refrigeration systems produce it. Normally one runs and the other is on standby; both can run together when the load is high.
Each system has its own compressor (Bitzer HSK5353-35Y-40P screw compressor with slide-valve capacity control), its own sea-water-cooled condenser (Bitzer K1973TB), its own sea-water pump, its own expansion device, and its own refrigerant charge of R449A. Everything measured on the refrigerant side — LP, HP, subcooling, condensing approach — therefore belongs to one system only.
Two consequences worth holding on to.
First, because both condensers see identical sea water, comparing the two systems' condensing approach at the same moment is the sharpest fouling test this plant can produce. Sea temperature, load and refrigerant glide all cancel out. That is what the comparison panel on the Relationships tab does.
Second, because the evaporator is shared, a freezing incident caused by either system destroys equipment that both depend on. There is no redundancy against that one failure. Section 5 covers it in full.
Filtering CHW supply or CHW ΔT by system would suggest System 2 has its own chilled water. It does not. Those panels therefore draw on every reading regardless of which system it was logged against, and carry a PLANT tag.
This matters for one figure in particular. The chilled-water ΔT has sat flat at 2.0–2.6 °C throughout the investigation — but every one of those readings was taken with a single system running. When both run together, more capacity is applied to the same water and the ΔT should be expected to change. That change is not a finding and not a fault. It is two systems doing what one was doing before.
Everything on this dashboard is built from a handful of readings taken by hand during a round and entered on the Smartsheet mobile form. Nothing is measured automatically. If a reading is wrong, everything derived from it is wrong, so it is worth knowing what each one is.
"bar g" means gauge pressure — what the dial shows, measured relative to the atmosphere around it. Absolute pressure is gauge plus about 1.013 bar. This distinction only matters in one place, the compression ratio, and it is why that figure is treated with care.
The Chiller field records which system the reading belongs to — 1 or 2. If both are running, the round is logged twice, once for each. The dashboard reads the first digit it finds in that field, so 1, "1" and "nr 1" all work. Anything it cannot recognise is counted and reported in the status line at the top rather than being quietly filed under the wrong system.
Date and Time are the authoritative timestamp for readings. Submitted (auto) is a fallback stamp applied by Smartsheet at the moment of submission, used only when the manual date is missing. Around 34 historical rows were entered in one batch, so their automatic stamps all share a single date — which is precisely why the manual date takes priority.
Six values are worked out from each reading. Two are simple subtractions; the rest depend on a refrigerant property table. Every one of them is calculated in Smartsheet as the reading is submitted, so the dashboard displays them rather than recomputing them — except in the calculators, which use exactly the same method so the answers match to the decimal.
A refrigerant at a given pressure condenses at a specific temperature. So a discharge pressure reading can be converted into the temperature at which the refrigerant is turning back into liquid inside the condenser. You cannot easily measure that temperature; you can always read the pressure. That is why it is done this way.
The conversion uses a published pressure-temperature table for R449A. Readings between the listed pressures are interpolated in a straight line.
The table was corrected part-way through this investigation, and it matters. The original high-side figures were skewed too high, which made saturated condensing temperature, subcooling and condensing approach all read several degrees warmer than reality. Every figure on this dashboard uses the corrected Honeywell values, anchored at 45 °C ≈ 19 bar g — which matches the compressor nameplate rating point. Any older report, card or spreadsheet built on the previous table is superseded. The low-side table was always correct and is unchanged.
This is the most useful single number on the dashboard. It answers: how much hotter than the sea water does the refrigerant have to get before the heat will move?
A clean condenser transfers heat easily, so a small temperature difference is enough and the approach is low. As the sea-water side fouls — marine growth, silt, scale on the tubes — heat transfers less readily, so the refrigerant must run hotter to push the same heat across, and the approach rises.
The reason it is more useful than watching HP alone is that it takes sea temperature out of the picture. On a hot day HP rises, but that is the sea being warm, not the condenser being dirty. Approach subtracts the sea water away and leaves only the condenser's own condition behind.
Either is defensible and both are used in industry. This project settled on the outlet and has used it consistently in every report, sheet and figure since the start. Switching to the inlet would silently shift every historical number and break comparability with everything already written — so it stays as it is. When comparing against manufacturer data, check which convention that data uses first.
Once refrigerant has fully condensed, any further cooling drops it below its condensing temperature. Those extra degrees are the subcooling. A few degrees confirms the condenser outlet is carrying solid liquid rather than a liquid-and-vapour mixture, which is what the expansion device needs to work properly. Persistent low subcooling suggests the charge is short — a slow leak, for instance.
This field is deliberately left blank when the liquid-line temperature was not taken. A blank subcooling is not an error; it means the reading was skipped, which is fine because it only needs sampling about once a day.
CHW ΔT is the load the ship is putting on the plant — how much heat the chilled water picked up on its way round. SW ΔT is the heat actually being carried away overboard, and it doubles as a sea-water flow check: if flow drops, the same heat is dumped into less water, so the rise across the condenser gets larger.
How many times the compressor multiplies the pressure. It is calculated and logged, but it is deliberately excluded from the health assessment. The reason is a mismatch that has never been resolved: the figure computed here from gauge readings runs 2.76–3.51, while the "healthy 2.77–2.97" band quoted in the analysis was derived on a saturated-pressure basis. Two different bases, not reconciled. Gating a health verdict on that would have produced warnings that meant nothing, so it stays out until the basis is settled.
Of everything on this dashboard, this is the limit worth understanding properly. Fouling costs efficiency and can be washed off in an afternoon. Freezing the evaporator destroys a component that cannot be repaired, and on this ship it is the one component both chiller systems depend on.
The short version. Suction pressure sets how cold the refrigerant gets. Cold refrigerant is what makes chilled water, but if the water touching the plates ever reaches 0 °C it turns to ice, and ice inside a brazed plate pack splits it open. 4.5 bar g is the suction pressure at which the safety reserve between "normally cold" and "cold enough to freeze during an upset" has been used up. Below it, raise the chilled-water setpoint.
Water is unusual: it expands as it freezes, by about 9%. In an open container that is harmless. Inside a sealed heat-exchanger channel there is nowhere for that expansion to go, so it goes into the metal instead. The forces involved are far beyond anything the plate pack is designed to carry.
This assumes the chilled water is plain water. If glycol or another antifreeze were dosed into the circuit the freezing point would drop and these figures would shift — worth confirming before treating the numbers as exact.
Heat only flows from warm to cold. For the refrigerant to take heat out of the chilled water, it must be colder than that water — normally by around 4 to 5 degrees at working load. So producing 10 °C chilled water means running refrigerant somewhere in the low single figures or below. An evaporator running below 0 °C is normal, expected, and not in itself a problem. What matters is how far below, and what is protecting the water.
You cannot get a thermometer inside the plate pack. But refrigerant boiling at a given pressure boils at a specific temperature, so the suction gauge tells you the evaporator temperature directly:
| LP bar g | mean evap. °C | coldest point °C | state |
|---|
Lower suction pressure means colder refrigerant. That is the whole reason the freeze rule is written as a pressure rather than a temperature: the pressure is the number you can actually read.
R449A is a zeotropic blend. Unlike a single-substance refrigerant, it does not boil at one temperature — it boils across a range of about 5 degrees, called the glide. It enters the evaporator at its coldest, the bubble point, and warms as it evaporates, leaving at the dew point roughly 5 degrees higher.
The figure quoted in a pressure-temperature table is normally the mean of those two. So the coldest metal in the plate pack — the inlet end — is around 2.5 degrees colder than the number you look up. That is the third column in the table above, and it is the one that matters for freezing.
At 4.5 bar g: the mean is about −4.6 °C, so the coldest surface is around −7 °C. That is seven degrees below the point at which the water against it turns to ice.
If the plates are at −7 °C and water freezes at 0 °C, why is there no ice in normal running? Because the water is moving. Flowing water continuously brings fresh heat to the plate surface, faster than the refrigerant can remove it. The thin film of water actually touching the metal stays above freezing even though the metal is well below it.
So the protection is not the refrigerant being warm enough. The protection is water flow. The freeze margin is a margin against losing that protection, and the things that remove it are:
That last mechanism is why falling suction pressure is the warning signal. It is not just a symptom; it is the plant telling you it is removing more heat than the water is delivering, which is exactly the condition that ends in ice.
There is no pressure at which ice suddenly appears. 4.5 bar g is the point at which the reserve runs out. Count the degrees between the plate surface and the ice point:
At 6.0 bar g the coldest surface sits about a degree below zero: comfortable. At 5.0 bar g it is about five degrees below: normal running with real reserve. At 4.5 bar g it is about seven degrees below, and the two mandatory allowances above have consumed essentially all of it. A stall in one water channel now reaches 0 °C faster than anything can react.
4.8 bar g is set as a watch line — roughly 0.3 bar of warning before the action line, enough to notice a downward trend and respond before it becomes urgent.
The compressor has its own low-pressure cut-out, set lower. That is a machine protection of last resort. The 4.5 bar g rule is an operator action line that sits deliberately above it, so that a person acts before a safety device has to.
This is what makes the limit non-negotiable rather than a guideline.
The evaporator is an Alfa Laval AC230DQ brazed plate heat exchanger. Stainless steel plates are joined by copper brazing at every contact point. There are no gaskets and no bolts. It cannot be opened.
When water freezes inside it, the plates deform and the brazed joints split. What follows is not a trip and a restart:
On a vessel that means procurement lead time, a riding crew or a yard visit to make the change, and no air conditioning at all until it is done.
And on this ship there is only one of them. Both chiller systems evaporate into the same AC230DQ. Freezing it does not take out the system that caused the freeze — it takes out the whole plant. Standby redundancy on the compressors and condensers protects against almost every other failure. It gives you nothing at all against this one.
Alfa Laval build the AC230DQ as a True dual-circuit unit and specifically state that this gives higher freezing resistance than a back-to-back arrangement. The difference is how the two refrigerant circuits are arranged in the pack. Back-to-back gives each circuit its own half of the plates, which creates a cold half and a warm half with a boundary between them where a stagnant, over-cooled corner can develop. True dual-circuit interleaves both circuits through the whole pack, so every water channel has refrigerant on both sides and the cooling is even end to end.
That is a real advantage and it is worth knowing this unit has it. But it is resistance, not immunity — freezing is still a destructive failure and the limit still applies.
When only one system runs, only half the refrigerant channels are active while the water still flows through all of them. That active circuit has to do its entire duty through half the available surface. For the same heat load it will therefore evaporate colder — and sit at a lower suction pressure — than it would if both circuits were working.
The practical reading: single-system operation at high load runs a thinner freeze margin than two-system operation at the same total duty. Running both systems is not only more capacity, it is also gentler on the evaporator. This follows from how the exchanger is built rather than from measurement on this plant, so treat it as reasoning to watch for in the data rather than an established figure — but it is a good reason to prefer two systems over one when the load is high.
The action is always the same: raise the chilled-water setpoint. A warmer setpoint means warmer water returning to the evaporator, which means the refrigerant settles at a higher evaporating temperature, which lifts suction pressure and restores the margin. The cost is slightly warmer accommodation for a while. That is a cheap price.
Two things that feel helpful and are not:
And if suction has fallen without any change in setpoint or load, the cause is somewhere else — check chilled-water flow first: pump status, strainer, valve positions, and whether air-handling valves have closed down.
Freeze risk is not a hot-day problem. It is highest when the plant looks to be having the easiest time of it: early morning, cool weather, low load, most of the ship asleep and the air-handling valves shut.
That is exactly the condition where the compressor cannot unload far enough to match a load that has almost disappeared, and so it draws the evaporator down. This plant already shows the pattern — early-morning suction dips to around 4.8 bar g, which is the watch line, leaving roughly 0.3 bar of margin at the coldest, quietest part of the day.
So when the dashboard is calm and the sea is cool and nothing seems to be under strain, that is the moment to check the freeze margin tile rather than assume all is well.
The plant holds its 10 °C setpoint comfortably in cool conditions and drifts to 14–17 °C on hot afternoons. That is a chain of causes, and it is worth following because the obvious explanations are the wrong ones.
The compressor is not failing and the circuit is not degraded — the compression ratio has stayed healthy throughout. What happens is that the compressor tries to load up as the afternoon load rises, discharge pressure rises with it, and its own protection logic pulls the slide valve back before the high-pressure trip. So it limits itself, repeatedly: load, unload, load, unload.
That cycling matters twice over. The obvious cost is the capacity shortfall and the warm supply water. The less obvious cost is mechanical wear from the constant load-unload movement, which is a maintenance concern in its own right regardless of the temperatures.
Two things this is not:
The two genuine levers are keeping the condenser clean, so that heat rejection is as good as it can be, and running both systems when the load is high.
| Limit | Value | Why there |
|---|---|---|
| Approach — watch | 14 °C | Early warning band, comfortably above normal running |
| Approach — clean | 16 °C | Heat-rejection loss now costs real capacity |
| Freeze — watch | 4.8 bar g | ~0.3 bar of warning before the action line |
| Freeze — act | 4.5 bar g | Safety reserve above the ice point is spent |
| Subcooling — watch | 4 °C | Below this the liquid seal and charge margin thin out |
| CHW setpoint | 10 °C | Plant design supply temperature |
| Reading age | 12 h | Beyond this the assessment may be out of date |
| Standby | 24 h | No reading across a day of rounds means not in service |
All of these live in one place in the dashboard code, so the tiles, the chart lines, the written assessment and the calculators cannot drift apart from each other.
Note that this was corrected from an earlier figure of 13 °C. That older number came from the skewed pressure-temperature table described in section 4. On the corrected basis, approach ran about 13.8 °C before the July condenser clean and 12.5 °C after — a difference of only 1.3 °C.
Which leads to the interpretation rule that governs every conclusion drawn here:
R449A's glide means differences smaller than about 2.5 °C are noise, not signal. A single reading 1.5 °C above yesterday's proves nothing. Only a sustained trend across days, at comparable sea temperature, is evidence. This is why the fouling before the July clean is described as mild rather than severe — the improvement, though real, sat inside the uncertainty band.
The condenser was brushed on 7 July. In the ten days that followed, the condensing approach held flat at about 12.2 °C while the sea-water inlet climbed from 23.5 °C to 28 °C and HP rose from 17 to 19 bar g. That separation is the useful result: the pressure rise was entirely the sea getting warmer, with no fouling contribution at all. It is also the clearest demonstration on this plant of why approach is watched rather than HP.
Five things on this dashboard reliably get misread. Each has been investigated and each has a settled explanation.
Well below the 5–6 °C design figure, and rock steady across every condition tested — ambient, time of day, pump count, condenser state. It is a characteristic of the load connected to the plant, not evidence of refrigeration degradation, and it is partly an artefact of the compressor never holding full load long enough to build a wider differential. It must never be cited as a chiller fault. As section 2 notes, expect it to shift once both systems run together — that too is not a fault.
Entirely expected. HP follows sea temperature. As long as the condensing approach stays low, the condenser is doing its job and the high side is simply reflecting the water it has to reject heat into.
The calculation is deliberately suppressed when the liquid-line temperature was not taken. Since that reading only needs sampling about once a day, most rows are legitimately blank. Not a fault, not missing data.
Covered in section 4 — the logged figure and the quoted band are on two different pressure bases that have not been reconciled. It is excluded from the verdict on purpose. Do not read a number outside 2.77–2.97 as a problem.
A system that has been out of service for weeks has weeks-old numbers. The dashboard recognises this and suspends its verdicts rather than alarming on stale data. Section 9 explains how.
One standing rule for this investigation: the chiller is assessed purely on its own refrigerant circuit. Chilled-water ΔT is treated as a load presented to the plant, never as a diagnostic of what is downstream. The condition of the air-handling units is a separate matter and is never used to explain chiller behaviour — doing so leads straight into circular reasoning.
Normally one system runs and the other sits idle, sometimes for months. Only the running system is logged. So the dashboard has to work out which is which, and it does so from the pattern of readings — no extra field to fill in and nothing to remember.
Each system is compared against the plant's most recent reading, not against the clock:
Comparing systems against each other rather than against the current time is what makes this safe. If nobody does a round for three days, both systems fall behind equally and neither is wrongly marked standby. Instead the separate 12-hour freshness warning fires across the whole dashboard, which is the correct message: the data is old, not the plant is idle.
Its health light goes grey and its verdicts are suspended, not turned red. A stale reading is not a fault, and colouring it red would train people to ignore red. The last known reading stays visible, marked with its age, and all history and charts remain fully available — a standby system's charts are anchored to its own last operating period rather than to today, so you see its final weeks of running rather than an empty axis.
If the last reading before it was stopped had been above a limit, that is reported as a past-tense note rather than a live alarm — for example, a system stopped while its approach was above 16 °C is flagged as one that was already due a condenser clean when it went to standby. That is worth knowing before starting it again.
Why the first reading after a long standby deserves attention. An idle system has its sea-water pump stopped, which leaves stagnant sea water sitting inside a copper-nickel condenser. Stagnant seawater is a well-known accelerator of biological growth and under-deposit corrosion — a condenser can foul while doing nothing at all. So the first approach reading after a long stop is one of the more informative measurements you can take. Treat a high value as a real result rather than dismissing it as a consequence of having been off.