CASE STUDY · ILLUSTRATIVE RECONSTRUCTION

A Real Case Anatomy

500 MW combined-cycle gas plant, Southeastern US, restored to full output. This is the analysis CondenserSentinel™ automates.

ILLUSTRATIVE RECONSTRUCTION OF A RECORDED CASE: TIMELINE AND INSTRUMENT VALUES ARE REPRESENTATIVE.

DAY 0

A hairline crack opens

Thermal cycling on a unit that starts and stops with the market fatigues a prior weld repair on the HP/LP crossover expansion-joint bellows. Air begins bleeding into the vacuum space. Baseline air in-leakage reads 40 SCFM — comfortably inside the air-removal system's capacity. Nothing alarms, because nothing is out of limits.

AIR IN-LEAK 40 SCFMBACK-PRESSURE NOMINALCONDENSATE DO <10 PPB

1/9

VIEW THE CHART VALUES AS A TABLE

AS IT HAPPENED · ILLUSTRATIVE RECONSTRUCTION OF A RECORDED CASE: TIMELINE AND INSTRUMENT VALUES ARE REPRESENTATIVE.
DAY	EVENT	READINGS
DAY 0	A hairline crack opens	AIR IN-LEAK 40 SCFM · BACK-PRESSURE NOMINAL · CONDENSATE DO <10 PPB
DAY 1–27	The creep nobody can see	AIR IN-LEAK 40→70 SCFM · ΔBP +0.05–0.1 inHg VS EXPECTED · DO TRENDING UP
DAY 28–40	Three symptoms, three logbooks	AIR IN-LEAK 115 SCFM · HEAT RATE ≈ +1% · DO EXCURSIONS LOGGED AS DRIFT
DAY 41–44	Air removal saturates	AIR IN-LEAK >120 SCFM · EJECTORS SATURATED · BP CLIMBING DAILY
DAY 45	First alarm — and a forced derate	HIGH BACK-PRESSURE ALARM · OUTPUT 500→150 MW · ~8,400 MWh/DAY LOST · $250K–340K/DAY
DAY 46–64	Firefighting in the dark	AIR IN-LEAK 240 SCFM · BP +0.8–1.0 inHg · OUTPUT HELD AT 150 MW
DAY 65	Found — helium doesn't lie	HELIUM TRACER POSITIVE · CROSSOVER BELLOWS · PRIOR WELD
DAY 66–74	Nine days to repair	EMERGENCY REPAIR · 9 DAYS · DERATE CONTINUES
DAY 75–80	Recovery	AIR IN-LEAK 10 SCFM · BP NOMINAL · OUTPUT 500 MW

WITH CONDENSERSENTINEL™ · MODELLED EARLIER-INTERVENTION SCENARIO: WHAT EARLIER DETECTION WOULD HAVE CHANGED, NOT A RECORDED OUTCOME.
DAY	EVENT	READINGS
DAY 0–26	Same crack, same physics	AIR IN-LEAK 40→70 SCFM · TWIN RESIDUAL ACCUMULATING
DAY 27	Drift flagged — 18 days before any alarm	DRIFT ALERT · ΔBP +0.1 inHg VS TWIN · 18 DAYS BEFORE FIRST ALARM
DAY 27–28	Likely cause ranked for the engineers	LIKELY CAUSE: AIR IN-LEAKAGE · TOP CANDIDATE: CROSSOVER BELLOWS
DAY 30	Recommendation, costed	RECOMMENDED: He TRACER ≤10 DAYS · CURRENT LOSS ≈ $5K/DAY
DAY 36	Confirmed at full load	HELIUM TRACER POSITIVE · AIR IN-LEAK ~80 SCFM · OUTPUT 500 MW
DAY 43–45	Repaired in a planned window	PLANNED 3-DAY WINDOW · AIR IN-LEAK → 10 SCFM
DAY 46	Closed loop — the day the alarm would have fired	VERIFIED RECOVERY · BP = TWIN EXPECTED · DO IN SPEC THROUGHOUT

Over roughly three months, condenser air in-leakage climbed from 40 to 240 SCFM. Vacuum degraded, heat rate climbed, and output was forced from 500 MW down to 150 MW, about 8,400 MWh lost every day, at an industry-estimate cost of $250K–$340K per day.Bionetics isolated the leak region from historical RheoVac data, reading the shift in air offtake from the HP to the LP shell. AES confirmed the location on-site with helium tracer-gas testing: a cracked expansion-joint bellows at the HP/LP crossover duct, at a prior weld repair.The repair brought air in-leakage from 240 down to roughly 10 SCFM. Condenser pressure returned to nominal from 0.8–1.0 inHg(a) above normal, and the unit came back to its full 500 MW.The manual forensics behind this recovery took weeks of expert time. CondenserSentinel™ predicts and flags the drift weeks earlier.

EVENT RECONSTRUCTION

How it unfolded, day by day

The chart above, replayed as the control room lived it — reconstructed from the historian record, air-removal data, and field reports.

READ THE FULL DAY-BY-DAY RECONSTRUCTION

DAY 0
A hairline crack opens
Thermal cycling on a unit that starts and stops with the market fatigues a prior weld repair on the HP/LP crossover expansion-joint bellows. Air begins bleeding into the vacuum space. Baseline air in-leakage reads 40 SCFM — comfortably inside the air-removal system's capacity. Nothing alarms, because nothing is out of limits.
AIR IN-LEAK 40 SCFMBACK-PRESSURE NOMINALCONDENSATE DO <10 PPB
DAY 1–27
The creep nobody can see
Leakage climbs toward 70 SCFM. The steam-jet air ejectors keep up, so condenser pressure runs only 0.05–0.1 inHg above what the circulating-water temperature says it should be — a real loss, but invisible without a model of expected performance. Condensate dissolved oxygen starts drifting upward as more air reaches the hotwell.
AIR IN-LEAK 40→70 SCFMΔBP +0.05–0.1 inHg VS EXPECTEDDO TRENDING UP
DAY 28–40
Three symptoms, three logbooks
Now 70→115 SCFM. Air starts blanketing the coldest tube bundles, shrinking effective heat-transfer surface. Heat rate is up about 1%, and peak output runs a few MW soft on warm afternoons — written off as summer circulating-water temperatures. Chemistry logs the DO excursions as probable instrument drift. Each symptom lands in a different system; nobody correlates them.
AIR IN-LEAK 115 SCFMHEAT RATE ≈ +1%DO EXCURSIONS LOGGED AS DRIFT
DAY 41–44
Air removal saturates
Leakage passes the ejector system's holding capacity (~120 SCFM on this unit). Past that point, free air accumulates in the shell faster than it can be pumped out and insulates tube surface directly. Back-pressure stops creeping in hundredths per week and starts climbing day over day.
AIR IN-LEAK >120 SCFMEJECTORS SATURATEDBP CLIMBING DAILY
DAY 45
First alarm — and a forced derate
Condenser pressure crosses the DCS high alarm: the first hard alarm of the entire event, 45 days after the crack opened. With a growing fraction of the bundle air-blanketed, the only way to hold exhaust pressure under the LP-turbine limit is to cut steam flow. Over the following days the unit is walked down to 150 MW.
HIGH BACK-PRESSURE ALARMOUTPUT 500→150 MW~8,400 MWh/DAY LOST$250K–340K/DAY
DAY 46–64
Firefighting in the dark
The obvious suspects go first: tube fouling and circulating-water flow. Meanwhile the leak keeps growing toward 240 SCFM and condenser pressure holds 0.8–1.0 inHg above normal. Bionetics pulls the historical RheoVac record and reads the tell — air offtake shifting from the HP to the LP shell. That signature means an in-leak near the crossover, not fouling.
AIR IN-LEAK 240 SCFMBP +0.8–1.0 inHgOUTPUT HELD AT 150 MW
DAY 65
Found — helium doesn't lie
An AES crew tracer-tests the suspect joints: helium sprayed at the HP/LP crossover expansion joint appears immediately at the air-removal exhaust. The bellows is cracked through at the old weld repair. Confirming the diagnosis took 20 days from the first alarm — and nine weeks from the day the crack opened.
HELIUM TRACER POSITIVECROSSOVER BELLOWS · PRIOR WELD
DAY 66–74
Nine days to repair
Scaffold the crossover, cut out the cracked bellows section, weld in the replacement, leak-check under vacuum. Nine days of around-the-clock work — with the derate running the entire time.
EMERGENCY REPAIR · 9 DAYSDERATE CONTINUES
DAY 75–80
Recovery
Air in-leakage falls from 240 to roughly 10 SCFM. Condenser pressure returns to design, and the unit ramps back to its full 500 MW by day 80.
AIR IN-LEAK 10 SCFMBP NOMINALOUTPUT 500 MW

DIRECT PRODUCTION LOSS

≈ $6M–8.5M

Roughly 25 days of deep derate at $250K–$340K/day — plus six weeks of creeping heat-rate loss before the alarm, emergency crew mobilization, and eleven weeks of elevated dissolved oxygen working on the condensate and feedwater train.

Timeline reconstruction is illustrative, based on the recorded case and standard condenser engineering practice; instrument values are representative.

THE COUNTERFACTUAL

The same three months, with CondenserSentinel™ in place

Same crack, same physics, same data sources. The only change: a digital twin computing expected condenser performance every few minutes — and acting on the difference.

READ THE FULL COUNTERFACTUAL, DAY BY DAY

DAY 0–26
Same crack, same physics
The bellows fatigues and the creep begins exactly as before: 40 SCFM drifting upward. From day one, CondenserSentinel™'s digital twin computes what condenser pressure should be at the current load and circulating-water temperature, and compares it against the measured value — continuously, on data the plant already collects.
AIR IN-LEAK 40→70 SCFMTWIN RESIDUAL ACCUMULATING
DAY 27
Drift flagged — 18 days before any alarm
Measured back-pressure has now run outside the twin's confidence band for 72 hours, and the trend classifier fires. This is the moment marked CONDENSERSENTINEL WOULD HAVE FLAGGED HERE on the chart above — 18 days before the DCS threshold alarm, while the fix is still cheap and schedulable.
DRIFT ALERTΔBP +0.1 inHg VS TWIN18 DAYS BEFORE FIRST ALARM
DAY 27–28
Likely cause ranked for the engineers
Rising air offtake with an HP→LP shell shift, dissolved oxygen trending up, and a growing back-pressure residual while the cleanliness factor holds steady: that signature points to air in-leakage rather than tube fouling or circulating water, for the engineers to confirm. Inspection history ranks the crossover bellows' prior weld repair as the most probable source.
LIKELY CAUSE: AIR IN-LEAKAGETOP CANDIDATE: CROSSOVER BELLOWS
DAY 30
Recommendation, costed
CondenserSentinel™ recommends an on-line helium tracer test within ten days: the drift is costing about $4–6K/day now, versus $250K–340K/day if it runs to a forced derate. The intervention is an inspection at full load — not an outage.
RECOMMENDED: He TRACER ≤10 DAYSCURRENT LOSS ≈ $5K/DAY
DAY 36
Confirmed at full load
An AES crew, coordinated in the shared CondenserSentinel™ workspace, tracer-tests the crossover during normal operation and confirms the cracked bellows at about 80 SCFM total in-leak. The unit hasn't lost a megawatt-hour it wasn't scheduled to lose.
HELIUM TRACER POSITIVEAIR IN-LEAK ~80 SCFMOUTPUT 500 MW
DAY 43–45
Repaired in a planned window
The bellows is repaired during a scheduled weekend low-dispatch window — bid into the market in advance, crew and materials staged, scaffolding up before the unit comes down. Three days at reduced load instead of a month at 150 MW.
PLANNED 3-DAY WINDOWAIR IN-LEAK → 10 SCFM
DAY 46
Closed loop — the day the alarm would have fired
The twin residual returns to zero and CondenserSentinel™ logs a verified recovery. On the day the reactive timeline was getting its first alarm, this unit is already repaired, at full load, with condensate chemistry that never left spec.
VERIFIED RECOVERYBP = TWIN EXPECTEDDO IN SPEC THROUGHOUT

TOTAL EVENT COST

≈ $150K–300K

A planned inspection, a bellows repair, and one scheduled reduced-load weekend — versus roughly $6M–8.5M on the reactive path.

It also avoids what never shows up on the production ledger: eleven weeks of elevated dissolved oxygen driving oxygen pitting through the condensate and feedwater train, and the accumulated damage of running the unit hard against a degrading condenser.

Same crack. Same plant. The only variable is when you find out.

REACTIVE · FOUND AT THE ALARM, DAY 45$6M–8.5M

PROACTIVE · FLAGGED BY CONDENSERSENTINEL™, DAY 27$150K–300K

Roughly 95% of the event cost sits in the gap between day 27 and day 45.

The cost of not seeing it

~$27.6M

At a different 500 MW gas plant, undetected condenser air leakage forced the unit down for 92 days at $250K–$340K per day of lost production.

The data existed. Nobody was acting on it. That is exactly the gap CondenserSentinel™ closes.

“E-Signal allows our dispatches to be far more efficient, identify root causes faster, and lower maintenance costs for our clients.”
BOB MECHEM · AMERICAN EFFICIENCY SERVICES

“Remote access to plant data allows us to engage earlier, stop firefighting, and help customers prevent outages before performance is impacted.”
JASON REYNOLDS · BIONETICS

“Efficiency Signal's platform and services promote accessibility, and would enhance our sensors' use. We are very confident that it will have an immediate and long-term effect on the industry.”
COLLIN ECKEL · SITE MANAGER, BIONETICS CESG

EPRI'S OPEN POWER AI CONSORTIUM NAMED USING DATA ANALYTICS TO INFORM CONDENSER MAINTENANCE A TOP-4 AI USE CASE

Your plant's data already holds this story.

Request an assessment