Thermal Systems and HVAC Engineering Case Studies
Technical5 Case StudiesUpdated Sep 2026
Thermal systems work rarely fails on the drawings. It fails in the gap between what a design assumed about load, airflow, and water temperature and what the building does on the hottest afternoon of the season. The five cases below come from consulting, plant operation, and equipment manufacturing. Each starts with a real operating problem, then asks what the design constraints, operating data, quality tolerances, and maintenance records can reveal before anyone orders new hardware. Treat them as working notes. Figures belong to the fictional organizations and show how decisions were framed, not what any building should achieve.
Case Study 1
Repeated drive failures in a clinic chilled water plant
Cedarline Mechanical Group, a design-build contractor that commissions and services chilled water plants for outpatient clinics and office campuses.
The situation
The primary chilled water loop for a six-story outpatient clinic was commissioned with two variable speed pumps, each driven by a dedicated frequency drive mounted in a shared mechanical room on the lowest level. Within eleven months, the drive serving the lead pump had failed three times, each time with an overcurrent trip followed by a burned capacitor bank. The room was designed for a maximum ambient of 35 degrees Celsius, and the drive cabinet had no independent cooling, relying on the room exhaust fan that ran on a thermostat set for the general space. The pump selection assumed design flow at full load, and the clinic had not yet reached that load in its first year.
The challenge
The owner wanted a written root cause before the drive warranty ran out, and the vendor and contractor disagreed. The drive vendor pointed to input harmonics from the building's other loads and suggested a line reactor. The contractor's service lead believed heat was the real problem. The trend logs covered only the last six weeks because the building automation system had been reset during handover, and nobody had recorded the room temperature history before that. The owner also asked whether the other pump drive would fail next.
Analysis
Teams grounded in Thermal Systems and HVAC Engineering usually begin with the nameplate and the commissioning sheet, because design constraints are written there before anyone opens a panel. Here the drive held full output only up to 40 degrees Celsius ambient, and the room was designed to stay near 35. The cabinet had no local cooling and depended on an exhaust fan wired to a general space thermostat. Six weeks of drive current and heatsink temperature showed the cabinet sitting close to its derating threshold on warm afternoons, which points to thermal stress more clearly than harmonic distortion, since distortion tends to track load steps rather than time of day.
The maintenance file told a second story. The drive had been cleaned twice in its first year, but the exhaust fan filter was missing from the service schedule, so airflow had been dropping for months. Failure modes matching this pattern include electrolytic capacitor dryout, control board stress from repeated trips, and bearing wear on the pump, which can raise motor current and mimic a drive fault. Corrective options ranged from a closed-loop cabinet cooler, which carried the highest capital cost, to a line reactor that would address harmonics but leave the heat problem untouched. Relocating the drive and scheduling the filter addressed the cause at lowest cost, though it needed a planned outage.
Actions taken
- Pull the full trend history from the controller and the room sensor, then overlay drive temperature against outdoor and mechanical room readings to confirm the thermal pattern.
- Relocate the drive cabinet to a wall with its own supply air path, and confirm the new location holds ambient below the manufacturer's full output limit during a simulated fan failure.
- Add the exhaust fan filter and cabinet airflow check to the monthly service schedule, with a recorded differential pressure reading that triggers a work order above the set limit.
- Hold the line reactor as a contingency only, and fund it if post-move logs still show distortion beyond the drive's input tolerance after a full season of operation.
Results
The drive was relocated during a planned weekend outage. Over the following hot weeks, cabinet temperature stayed about eight degrees below the derating threshold, and the lead pump ran through a full cycle of peak afternoons without a trip. The line reactor was not purchased. The owner accepted the root cause report, and the warranty claim for the failed modules was approved on the basis of the logs showing sustained heat stress. The other pump drive was moved in the same outage as a precaution.
Questions for discussion
- Which of the logged readings would you need before accepting a harmonic explanation for repeated drive failures?
- How would you weigh the cost of a cooled cabinet against a relocation that requires a planned outage?
- What maintenance items should be tied to warranty conditions for equipment installed in shared mechanical rooms?
Case Study 2
Brazing furnace bottleneck and leak rate drift
Brightwater Coil Works, a plant that fabricates finned tube heat exchanger coils for rooftop and packaged air handling units.
The situation
Brightwater Coil Works runs one brazing furnace that sets the pace for the whole plant. Upstream, tube expansion and fin stacking can feed about twice the furnace's throughput, so coils queue on the line for most of the shift. Over several weeks, output per shift fell from around 240 coils to about 190, while the leak rate at braze joints, measured by pressure decay after brazing, rose from under one percent to just over two and a half percent. Scrap and rework took up a growing share of furnace time, which made the bottleneck worse. The furnace's atmosphere dew point and belt speed are logged, but the logs had never been compared with leak results.
The challenge
The plant manager needed output back above 230 coils per shift before the seasonal order book started in earnest, and the quality manager would not release coils with a leak rate above one percent to the packaged unit line. Adding a second furnace was a capital request that would take most of a year to approve and install. The operations team wanted a quick fix, such as running the furnace faster, but nobody could say whether faster belt speed would raise leaks further.
Analysis
Start with the tolerances, because they set what counts as a defect before any process change. The braze joint specification allowed a leak rate up to one percent at the release test, and the fin collar fit tolerance was tight enough that a few hundredths of a millimeter of gap variation changed capillary flow. Operating data from the furnace showed two sources of variation that operators had treated as noise. Dew point at the furnace entrance swung widely between startup and mid-shift, and the two shifts set the atmosphere differently at startup, so the same belt speed produced different joint quality. Stacking fixtures also drifted as they wore, which changed fin spacing across the batch.
The maintenance records explained why the drift had gone unnoticed. The gas analyzer was calibrated on the quarterly schedule, but the dew point sensor had been replaced once with an unlabeled spare, and its offset had never been checked. The belt drive had also been adjusted twice in a single month, suggesting belt speed was being used to chase leaks. Sensor repair and a fixed startup recipe cost little and could be tested within one week; fixture replacement needed a longer trial. Each change was given an acceptance criterion: leak rate below one percent on three consecutive shifts, and furnace output at or above 230 coils per shift.
- Measure Log dew point, belt speed, and leak rate every shift, using one time source.
- Diagnose Compare sensor offsets and shift setups against the leak results.
- Adjust Apply the locked recipe and test one belt speed at a time.
- Verify Accept the change only after three shifts below one percent leak rate.
Actions taken
- Replace the dew point sensor, verify it against a reference hygrometer across the full startup range, and log both readings for five consecutive production days.
- Write one locked startup recipe for atmosphere and belt speed, signed off by both shift leads, and remove the manual override from the furnace panel for the trial period.
- Run a designed trial of three belt speeds at two dew point setpoints, measuring leak rate on every coil in each run, so the faster speed is accepted only if joints still pass.
- Replace the worn stacking fixture in the second week, check fin spacing across the batch with a go or no-go gauge, and reject any batch that fails the spacing check.
Results
The sensor fault explained most of the drift, and the locked recipe brought leak rate from about two and a half percent to under one within two weeks. The designed trial showed that the faster belt speed was acceptable only at the tighter dew point setpoint, so the plant settled on that combination. Output recovered to about 232 coils per shift once the queue at the furnace cleared. Fixture replacement reduced fin spacing variation, and the rework share fell enough to free furnace time that had been lost to repeat brazing. The second furnace request was withdrawn.
Questions for discussion
- Which measurement would you trust most to separate a sensor offset from a real change in furnace atmosphere, and why?
- How should a plant decide between a short trial of faster belt speed and a longer capital case for a second furnace?
- What acceptance criteria would you set for a fixture change when the defect appears only after brazing?
Case Study 3
Refrigerant change review for a campus chiller plant
Ashgrove Building Systems, an engineering consultancy that reviews mechanical designs for owners and public agencies.
The situation
Ashgrove was asked to review a design package for replacing the refrigerant in a campus chiller plant. The existing machines used a refrigerant that was being phased down, and the replacement was a mildly flammable option that the owner's design team had chosen on lifetime cost. The plant room sat beneath an occupied clinic wing, with one door to a loading corridor and a ventilation fan rated for normal operation, not for a leak event. The package had been prepared by a team that had moved on to another job, and its equipment schedules showed a refrigerant charge that differed between the drawings and the submittal.
The challenge
The review had to decide whether the package could go to the authority having jurisdiction for approval, or whether it needed more evidence first. Missing items included a leak detection and alarm sequence, a room ventilation calculation based on the actual charge, a refrigerant detector placement plan, and proof that the machine room door and corridor arrangement met the required egress and separation. The owner's schedule assumed submittal in the next month, and the contractor had already ordered the equipment.
Analysis
The first question was whether the design stayed inside the charge limit for the room. The drawings listed one charge, the submittal another, and the difference of several kilograms was enough to change the room size required for a leak to disperse below the lower flammability limit. Operating data from the existing plant helped only partly, because it recorded run hours and pressures but not leak history. The tolerance on the detector's response had to be matched to the lower flammability limit with a margin, and the calibration interval was not in the package. Each gap was logged as missing evidence rather than design error.
The maintenance records for the existing plant were useful for a different reason. Leak repairs were concentrated at two flanged joints and a sight glass, and the new design had kept all three. That history supported a control at source: fewer joints, brazed rather than flanged where service access allowed, and a leak test before charging. For approval, Ashgrove set three gates: a complete evidence set before submittal, a commissioning witness test of the alarm sequence, and a handover record that listed detector calibration and ventilation fan checks on a fixed schedule. Each gate had a named owner, and none could be passed by note alone.
Actions taken
- Obtain a corrected charge schedule that matches the drawings and submittal, then recalculate the room ventilation requirement and the lower flammability limit margin from that figure.
- Specify a leak detection and alarm sequence with two sensor heights, a fan interlock that starts on detection, and a shutoff of the compressor, and document each interlock on the control drawings.
- Revise the plant room door and corridor separation so that the discharge path and the egress route are shown on the plan, and confirm the arrangement with the fire marshal before submittal.
- Hold the submittal until the evidence set is complete, and record the reason for any deferred item in the design log with the name of the person accepting the risk.
Results
The owner accepted a delay of about six weeks rather than submitting an incomplete package. The corrected charge was higher than the drawings showed, which required a larger ventilation rate and a revised detector count. The redesign kept the sight glass but removed two joints from the liquid line. The authority having jurisdiction reviewed the complete set without a return comment on the leak sequence. The commissioning witness test found one interlock wired to the wrong fan, which was corrected before handover.
Questions for discussion
- Which missing item in this package would you have asked for first, and what would have changed if it arrived after submittal?
- How would you set the threshold for deferring an item at a design gate without weakening the review?
- When fewer joints reduce leak risk, what trade-offs in service access or installation cost should the design team weigh?
Case Study 4
Late air handler and duct clash on a research lab
Larkspur Owner Services, an owner's representative firm that tracks mechanical packages on laboratory and clinic builds.
The situation
Larkspur Owner Services managed the mechanical package for a three-story research laboratory with a central plant serving two air handling units, a variable flow chilled water loop, and exhaust systems for fume hoods. The milestone that mattered most was plant energization, because the lab's commissioning schedule, tenant move-in, and the owner's lease start all depended on it. The primary air handling unit had been ordered on a lead time that was confirmed in the second month, but the supplier then moved its ship date twice. Separately, a coordination review found that the main supply duct on the third floor clashed with a post-tensioned beam, forcing a reroute through a corridor ceiling already committed to electrical trays.
The challenge
The energization milestone was now about eleven weeks late, and the general contractor's recovery schedule assumed the replacement unit would arrive with a shorter lead time than the supplier had quoted. The team had to choose between paying for expedited freight and a temporary chiller rental, rerouting the duct at the cost of a redesign, or accepting a phased energization that left one laboratory wing without conditioned air for longer. Each option traded cost against schedule differently, and the owner wanted one recommendation for the budget meeting.
Analysis
The design constraints decided most of the options. The air handling unit's coil and fan were sized for a specific static pressure, and rerouting the duct added length and two elbows that would push the fan above its selection point unless the duct was resized. The chilled water loop tolerated a temporary chiller only if the rental unit matched the building's supply temperature band and flow, and the controls contractor needed three days to tune a temporary sequence. Load data from the design model showed the plant reached full load only on the hottest days, so a phased start would limit the wing only during peak afternoons.
Quantifying the trade-offs was the next step. Each path had to be priced against the schedule it bought. Expedited freight was quoted at 64,000 dollars and would recover about five weeks if the unit cleared the factory on time, which the supplier's recent shipment record did not support. The temporary chiller rental cost 118,000 dollars over ten weeks and added controls risk, while the duct reroute cost 22,000 dollars plus a six-day redesign that the coordination team could run alongside procurement. Phased energization had no direct cost but carried a late occupancy exposure of 40,000 dollars for each month of delay. Factory test records from the last order were incomplete, so schedule risk was scored conservatively.
Actions taken
- Request a firm ship date with a written penalty clause from the supplier, and confirm that the factory test sequence for the unit is scheduled before the ship date is accepted.
- Reroute the supply duct within the existing ceiling depth by resizing the third floor branch, and send the revised static pressure to the supplier so the fan selection is confirmed.
- Keep the temporary chiller rental as a three week contingency, with a cancellation date tied to the permanent unit arrival and a controls tuning window.
- Approve the phased start for the two lab wings with the lowest load, and schedule the remaining wing after the permanent unit passes its airflow test, so the owner can take partial occupancy sooner.
Results
The team chose the reroute and phased start, and skipped both freight and the temporary chiller. The permanent unit arrived about four weeks later than the supplier's first promise, but it still passed its airflow test in time. The two lab wings with the lowest load opened on schedule, and the third wing followed about five weeks after plant energization. Total cost growth came to about 31,000 dollars, including the reroute, compared with the 182,000 dollars that freight and rental together would have added. Lease exposure was limited to the third wing's late occupancy, roughly one month.
Questions for discussion
- How would you test whether a supplier's revised ship date deserves more weight than its earlier record when choosing a recovery path?
- Which trade-offs should be priced in dollars and which should be scored as schedule risk when the owner must choose?
- What would change your recommendation if the duct reroute required a structural review that added a month?
Case Study 5
Cutting plant energy and water without losing performance
Tallow Creek Foods, a regional food packaging plant with a chilled water plant and air handling for cold packing rooms.
The situation
Tallow Creek Foods ran a chilled water plant with two water-cooled chillers, two cooling towers, and air handling for cold packing rooms that had to hold product temperature within a narrow band during every shift. The plant's energy use per ton-hour had drifted upward over the last eighteen months, and the towers were drawing more makeup water than the production schedule explained. The site had a target to cut energy and water use without changing product holding conditions. Operators had been running the chilled water supply at a fixed setpoint because it was simpler than resetting it, and the tower fans ran at full speed in most conditions.
The challenge
The site's target was a 15 percent cut in plant energy and a 20 percent cut in makeup water, and any measure had to pass a holding test before it went live. The quality team would not accept a supply setpoint change that pushed room humidity above the limit for packaged product. Capital for new equipment was limited, and the sustainability lead wanted the options ranked by lifecycle cost, including the refrigerant and treatment chemicals consumed over the equipment's life, not just the first-year savings.
Analysis
The design constraints set the envelope. The cold rooms were designed for a relative humidity ceiling that limited how high the chilled water supply temperature could move, because the coils dehumidify only when the surface runs cold enough. Operating data from the building automation system gave the plant's load profile, which showed that the rooms rarely needed their full cooling capacity outside a few hours of peak production. That was the opening for a supply reset. The tolerance on product temperature, however, was tight, so any change had to be proved on a holding test, not on the energy meter alone.
The maintenance records shaped the comparison. Tower fill had been inspected twice in the past year and showed scale, and blowdown ran at a high cycle rate. Each option was scored on first-year energy and water savings, on the lifecycle cost of equipment and consumables, and on operating measures such as kilowatt-hours per ton-hour and makeup water per unit of cooling delivered. Variable speed drives on the tower fans scored high on energy, but their replacement cost at the end of life was large. Blowdown recovery cut water purchase at moderate capital cost, and leak repair was cheapest with a small refrigerant saving.
- 1Reset chilled waterRaise supply setpoint when loads allow, after checking dehumidification limits.
- 2Trim condenser flowMatch tower fan and pump speed to wet bulb, cutting water and fan energy together.
- 3Recover blowdown waterReuse filtered blowdown for washdown, reducing treated water purchase and discharge.
- 4Fix leak hotspotsTarget joints with repeated top-ups, using the charge log as the trigger for repair.
Actions taken
- Set a supply temperature reset that raises the chilled water setpoint only when room humidity and dew point margins stay inside the packaging limit, logged every five minutes.
- Install variable speed control on both tower fans and link pump speed to the wet bulb reading, then compare kilowatt-hours per ton-hour against the baseline for one full production cycle.
- Route filtered blowdown to the washdown supply after a two-week water quality check, and set a limit on conductivity and chlorides that closes the valve automatically if breached.
- Run a leak survey on the refrigerant circuit, repair the joints with repeated top-ups first, and enter every charge addition in the log with a reason code so repairs can be ranked.
Results
The reset was adopted in stages, starting with the lower load hours, and the holding test passed on every product line during the first trial. Plant energy use per ton-hour fell by about 11 percent over the following quarter, and makeup water fell by about 18 percent after the blowdown recovery line was commissioned. Tower fan drives were installed later, after the lifecycle review confirmed the payback under the site's threshold. The leak survey found three joints responsible for most top-ups, and repairing them cut refrigerant purchases by roughly a third.
Questions for discussion
- What humidity or dew point evidence would you need before accepting a higher supply temperature in a packaging room?
- How would you weigh a measure with strong energy savings but a large end of life replacement cost against a cheaper measure with smaller savings?
- Which operating measures would you publish alongside a single headline energy figure, and why?
Learning Resources
Technical CASE STUDIES
China, India, United States, Indonesia, Pakistan, Nigeria, Brazil, Bangladesh, Russia, Japan, Mexico, Philippines, Vietnam, Egypt, Turkey, Germany, United Kingdom, France, Italy, Canada, South Korea, Australia, Spain.