Rated vs. Actual kW/TR: Using Scatter Plot Analysis for Continuous Chiller Performance Validation

Most chiller plants in India undergo performance testing once a year. The procedure is well-established: a technician connects portable instruments, the chiller runs at a controlled condition for a stabilisation period, and the measured kW/TR is compared against the manufacturer's rated curve. If it passes, the chiller is declared fit. If it fails, maintenance is scheduled.

The problem is timing. Annual testing is a snapshot -- one operating point, at one load, on one day. If the chiller degraded steadily over the nine months since the last test, the snapshot misses the entire decline.

The alternative is continuous validation: every operating point over a rolling period -- seven days, thirty days, or longer -- plotted against the manufacturer's rated kW/TR curve. Each dot is one interval of actual operation. The dashed line is the rated performance at that load. The pattern of dots relative to the rated curve tells you not just that performance has degraded, but why it degraded and when the degradation started.

Reading the scatter plot: What the pattern of deviation tells you about the failure mode

The scatter plot is constructed from two data streams. The X-axis represents cooling load, expressed either in TR or as a percentage of rated capacity. The Y-axis represents the chiller's actual kW/TR -- the power consumed per ton of refrigeration delivered. Each dot on the chart is one operating interval during the observation period. The dashed line is the manufacturer's rated kW/TR at the corresponding load.

An important clarification on the rated curve: the reference line should not be the single-point AHRI 550/590 full-load rating (which assumes 6.7°C leaving CHW and 29.4°C entering CW). In Indian conditions, entering CW routinely reaches 30--34°C during peak summer. A chiller operating at 32°C CW supply instead of AHRI's 29.4°C will show higher kW/TR simply because of higher lift -- that is physics, not degradation. The rated curve should be the manufacturer's multi-condition performance map adjusted for site CW and CHW temperature ranges. Most manufacturers provide part-load curves at multiple CW temperatures for this purpose.

For a chiller performing to specification, the scatter cloud should track the site-adjusted rated curve closely. Some scatter is normal -- variations in entering condenser water temperature, fluctuations in chilled water return temperature, transient conditions during load changes. But the dots should cluster around the rated line, not systematically above it.

The scatter plot's diagnostic value also depends on sensor accuracy. BMS-installed temperature sensors typically have an installed accuracy of ±0.3--0.5°C, compared to ±0.1°C for calibrated portable instruments. The scatter band will be wider than what a controlled annual test produces. But the diagnostic signatures described below depend on systematic patterns across many data points, not individual readings -- hundreds of operating points with ±0.5°C scatter will reveal a 5% systematic deviation that a single measurement at ±0.1°C might catch or miss depending on test-day conditions. For meaningful diagnostics, the power meter should be Class 0.5S or better, and temperature sensors should be calibrated at least annually against a reference standard.

When a chiller is underperforming, the dots shift upward -- the chiller is consuming more kW per TR than its rated specification predicts. But what distinguishes this chart from a simple kW/TR trending line is the load dimension. You are not just seeing that the chiller is underperforming -- you are seeing where across the load range it underperforms. A deviation confined to low loads points to the unloading mechanism. A deviation only at high loads points to the condenser circuit. A uniform shift across all loads points to the refrigeration cycle itself. The diagnosis is different, the investigation is different, and the corrective action is different.

A single kW/TR number averaged over a day or a week obscures these load-dependent patterns. The scatter plot preserves them.

Four diagnostic signatures -- system-level, part-load, high-load, and discrete events

Four distinct scatter patterns emerge from field data, each pointing to a different failure category.

Signature 1: Upward shift across all loads -- system-level degradation

The entire scatter cloud sits above the rated curve, at every load point from 30% to 100%. The gap between actual and rated kW/TR is present across the full load range, though it may be proportionally larger at lower loads -- for example, refrigerant undercharge represents a bigger fraction of active system capacity when the evaporator is partially unloaded.

This pattern indicates a problem affecting the fundamental refrigeration cycle, independent of load. The most common causes: refrigerant undercharge (reduces evaporator effectiveness and increases compressor work per unit of cooling), oil contamination (oil logging in evaporator tubes reduces heat transfer surface area), compressor mechanical wear (worn bearings or seals increase friction losses at every operating point), and evaporator fouling (scale or biological growth on tube surfaces degrades the heat transfer coefficient).

The investigation is a full system check: refrigerant charge verification, oil analysis, compressor current draw at known conditions, and evaporator approach temperature measurement. If evaporator approach has increased by more than 1--2°C relative to commissioning data, tube fouling is the likely contributor.

Signature 2: Deviation at low loads only -- unloading mechanism fault

Points above rated only below 40--50% load. Above that range, the scatter cloud tracks the rated curve normally.

This is an unloading mechanism problem. Centrifugal chillers modulate capacity using inlet guide vanes (VGVs/IGVs) that control refrigerant flow into the impeller. Screw compressors use a slide valve that controls effective compression volume. When either mechanism fails to modulate correctly -- VGVs not closing fully, slide valve sticking at an intermediate position -- the compressor does more compression work than the load requires, but only at the part-load conditions where the unloading mechanism is supposed to be actively modulating. At high loads, VGVs are nearly fully open and the slide valve nearly fully extended, so malfunction is invisible. The scatter plot reveals what a full-load performance test cannot.

The investigation is mechanical: inspect VGV linkage for binding, check slide valve position feedback against the controller command, verify actuator stroke and calibration. On centrifugal machines, worn VGV bearings or a bent linkage rod are common culprits. On screw machines, slide valve position sensor drift can cause the controller to believe the valve is at one position when it is physically at another. Centrifugal chillers with variable-speed drive (VSD) capacity control rather than VGVs can show a similar low-load deviation pattern if the VSD minimum speed limit is set too high or the speed-to-load control curve is miscalibrated.

Signature 3: Deviation at high loads only -- condenser-side constraint

Points above rated only above 70--80% load. Below that, performance tracks the rated curve.

This is a heat-rejection bottleneck. At part load, the condenser has excess capacity, so even a partially compromised condenser circuit maintains reasonable condensing pressure. As load increases, the condenser must reject more heat per unit time, and any restriction in the heat-rejection path becomes the limiting factor.

Causes: condenser tube fouling or scaling (mineral deposits increase condensing pressure at high loads), restricted condenser water flow (partially closed valves, pump impeller wear, strainer fouling), or cooling tower degradation (fill media deterioration, nozzle clogging, air-side fouling). Approach temperature (CW leaving minus wet-bulb) is a direct health indicator of the cooling tower -- rising approach at constant fan speed and load indicates CT fouling.

Each 1°C increase in CW supply temperature increases compressor power by approximately 2.5--3%. If the condenser circuit cannot maintain design CW temperatures at high loads, the kW/TR penalty accumulates precisely where the scatter plot shows it.

The investigation starts at the cooling tower (approach temperature trending, fill inspection, basin check) and works inward through the condenser water piping (strainer DP, pump flow verification) to the condenser tubes (tube-side pressure drop, eddy current testing if age warrants it).

Signature 4: Sudden shift across all points -- discrete event

All points shift upward at a specific date. Before the date, the scatter cloud tracks the rated curve. After the date, it sits uniformly above -- resembling Signature 1, but with a clear temporal boundary rather than a gradual drift.

This is a discrete event, not progressive degradation. Something changed on or near that date. The most common causes: a refrigerant leak (sudden loss of charge shifts every operating point), sensor drift (if a temperature sensor or power meter drifted, the computed kW/TR shifts even though actual thermodynamic performance may be unchanged), or a valve failure (a stuck expansion valve or bypass valve changes the operating envelope across all loads).

The investigation is forensic: identify the shift date from the scatter plot, cross-reference with the maintenance log, alarm history, and work orders. Sensor drift is the most commonly overlooked cause -- a CW temperature sensor drifting by 2°C shifts the calculated cooling load and the computed kW/TR, creating a false degradation signal. Always verify sensor calibration before chasing a mechanical fault.

Continuous validation vs. annual audit: Why periodic testing misses progressive degradation

The annual chiller performance test, typically conducted per AHRI 550/590 protocols, measures kW/TR at one or two load points under controlled conditions. It is a useful calibration exercise. But it has structural limitations that no amount of measurement precision can overcome.

First, the timing problem. If a chiller passes its annual test in January and develops a slow refrigerant leak in March, the degradation runs undetected until the following January. Consider a 500 TR water-cooled screw chiller operating at 70% average load (baseline kW/TR of approximately 0.64), running 6,570 hours per year at ₹9/kWh. A 5% kW/TR degradation adds roughly 16 kW of excess consumption -- approximately ₹9.5 lakh per year. If the degradation runs undetected for ten months between annual tests, that is roughly ₹7.9 lakh in avoidable energy cost from a single chiller.

Second, the condition problem. Annual tests are conducted at a specific load and specific condenser water conditions. A chiller that passes at 80% load under controlled CW conditions may be failing at 40% load (Signature 2) or at 95% load with elevated CW temperatures (Signature 3). The single-point test cannot expose load-dependent failure patterns.

Third, the trend problem. Progressive condenser fouling does not announce itself. It develops over months as mineral scale builds on tube surfaces. On a scatter plot with weekly or monthly resolution, the high-load points drift upward gradually -- as an illustrative progression, a plant might see 2% above rated in month one, 4% by month four, 7% by month eight (the actual trajectory depends on water quality, treatment, and operating hours). This trend is visible on the scatter plot long before the degradation is severe enough to trigger a comfort complaint or an alarm. Annual testing catches it only when the accumulated degradation happens to be present on the test day.

Continuous scatter plot validation changes when and why you schedule maintenance. Instead of cleaning the condenser on a calendar basis (every six months, regardless of condition), you schedule it when the high-load points drift above a threshold -- say, 5% above rated kW/TR. Instead of checking refrigerant charge annually, you check it when a sudden shift appears. You neither service too early (wasting labour on a clean condenser) nor too late (wasting energy on a fouled one).

The cost comparison favours continuous monitoring. A third-party annual performance test -- portable power analyser, calibrated temperature sensors, flow measurement, and a written report -- typically costs ₹50,000--₹1,00,000 per chiller. Continuous validation uses sensors already installed for plant control. The marginal measurement cost is zero; the cost is in the analytical platform, rated curve configuration, and engineering attention to interpret the patterns. Compared to ₹7.9 lakh in undetected degradation from a single chiller running between annual tests, the economics are straightforward.

A note on complementarity: the scatter plot does not replace the annual performance test. It serves a different function. The annual test gives you one high-accuracy measurement at one operating point, validating absolute instrument calibration. The scatter plot gives you many lower-accuracy measurements at every operating point the chiller encounters over weeks, providing load-dependent pattern visibility and temporal continuity that a single-point test structurally cannot. The two approaches are complementary. The annual test validates the measurement system. The scatter plot tells you what happened between tests.

Where continuous chiller performance validation leads

The scatter plot's value extends beyond diagnostics into operational decisions. Efficiency-based chiller sequencing -- loading the most efficient chiller first, unloading the least efficient first -- depends on knowing each chiller's actual kW/TR, not nameplate COP or the value from the last annual test. If CH-1 is running 8% above rated at all loads (Signature 1) while CH-2 tracks the rated curve closely, CH-2 should lead. The scatter plot provides the continuous performance data that makes this sequencing decision defensible.

The Chiller Plant Manager (CPM) constructs this rated-vs-actual scatter plot from live plant data, with the manufacturer's rated curve as the reference baseline. The Energy dashboard presents the scatter plot alongside equipment energy attribution, load profiling by chiller, and an efficiency trend that overlays actual kW/TR against a weather-normalised baseline. The alarms layer flags when plant-level IKW/TR exceeds baseline by a configurable threshold, providing early warning when performance drift crosses from acceptable scatter into actionable degradation.

The rated-vs-actual scatter plot is a diagnostic instrument, not a dashboard widget. It tells you what is wrong, where to look, and when the problem started. Four patterns, four failure categories, four different maintenance responses -- all from a single chart that most chiller plants already have the sensors to construct.

Conclusion

Continuous chiller performance validation turns a once-a-year performance check into an ongoing diagnostic process. By plotting actual kW/TR against the manufacturer's site-adjusted rated curve, facilities teams can identify whether degradation is system-wide, limited to part-load operation, concentrated at high loads, or linked to a discrete event.

The scatter plot does not replace the annual test -- the two serve different functions. The annual test validates measurement accuracy under controlled conditions, while continuous monitoring provides the load-dependent and time-dependent visibility needed to identify problems between tests.

For facilities teams, the practical principle is simple: don't wait for the next annual test to discover that a chiller has been wasting energy for months. Use the data already being generated by the plant to continuously validate performance, identify deviations early, and connect those deviations to specific maintenance actions.

The result is not simply a better dashboard. It is a more measurable, defensible approach to chiller plant performance management.