PWR-330 · Lesson 12 of 12 · 25 min
Boiler operation case studies
▶ Video coming soon A video version of this lesson is on the way.
For the last session of the boiler module, Grazyna did not bring a lecture. She brought a folder of cases: some from Cantwell's own autumn, one from Keswold, a few from the industry's history, and two from her own years at the mill with the names removed. "You have learned the parts," she said. "Now take them apart again in the order things actually go wrong. For each one, tell me what happened, why it happened, what should have caught it sooner, and what was changed. If nothing was changed, tell me what should be."
This lesson works through those cases. Each uses the knowledge of the earlier lessons and of PWR-220, and each ends with the changes that followed or should follow. The cases are arranged from the familiar to the distant, but the pattern in them is the same: small signs, noticed or missed, and a decision.
A method for a case
A case analysis that teaches anything goes beyond the immediate cause. A useful structure, the one PWR-110 used for incident investigation, is:
- What happened, in time order, with the measurements that show it.
- The immediate cause: the physical failure, such as a cracked lens, a pitted tube or a stuck valve.
- The underlying causes: why that failure occurred, and why it was not found sooner; often a missed test, an unclear procedure, a design weakness or a habit.
- The defences: which protections worked, which did not, and which were missing.
- The changes: what was done to prevent a repeat, and whether it addresses the underlying causes or only the immediate one.
Asking "why" several times in a row, each time of the previous answer, usually leads from the immediate cause to something a manager or engineer can change.
Case 1: boiler 1's flame failure
What happened. At 02:40 on October 20, boiler 1 locked out on flame failure while firing steadily. The burner management system closed the safety shutoff valves, ran the post-purge and displayed the first-out alarm: loss of flame signal. Desmond Kirwan confirmed the safe shutdown, recorded the alarm, and called the instrument technician, who found a cracked flame scanner lens. The lens was replaced, the furnace purged, and the burner relit through its normal sequence.
Causes. The immediate cause was the cracked lens, which scattered and weakened the flame's ultraviolet signal until it fell below the relay's threshold. The underlying cause was that the scanner's signal strength was not recorded, so its decline, which probably took days, was invisible until the trip.
Defences. The burner management system worked as designed: it saw the lost signal and shut the fuel within its response time. The procedure worked too: the operator did not reset the burner until the cause was found.
Changes. Lorne added the scanner signal strength for all three boilers to the night log sheet. A falling signal now triggers a work request before it triggers a trip. That change addresses the underlying cause: it turns a hidden decline into a trend someone reads.
Case 2: boiler 3's tube leak and its pitting
What happened. At about 01:30 on November 2, Desmond noticed that boiler 3's feedwater flow exceeded its steam flow and blowdown by about 166 kg/h, roughly 1.5 in 100 of the feed, and that the drum level controller was opening the feed valve slowly to hold level. He took the boiler off line. A pinhole was found next day in a furnace floor tube near the rear wall, the tube was replaced and the boiler hydrotested and returned to service on November 6. The laboratory's report on the removed tube, on November 14, found oxygen pitting dating from the summer wet layup: the sulphite residual had not been tested after June, and the drum had been opened twice during the layup, letting air in.
Causes. The immediate cause was a pit that grew through the wall. The underlying causes were a layup that lost its oxygen scavenger without anyone knowing, because the layup testing stopped, and two drum openings that the layup procedure did not account for.
Defences. The water balance caught the leak while it was a pinhole, which is a credit to the night shift's attention. The layup procedure was the defence that failed.
Changes. A revised layup procedure, with full wet layup, a higher residual, weekly tests and a nitrogen cap, and a dry layup option, was prepared for the November operations meeting. The deepest-pitted floor tubes are to be replaced at a planned 2027 outage, with ultrasonic thickness testing of the rest. Grazyna's comment on the case was about the drum openings: "The procedure has to say what happens when somebody needs to open the drum. Somebody always does."
Case 3: boiler 2's stack temperature
What is known. Between November 3 and 10, boiler 2's stack temperature at a steady load climbed from about 196 to 214 °C (385 to 417 °F), with its oxygen unchanged and its CO low. Lesson 6 put the cost at about 1.1 points of efficiency, or 1.3 in 100 more gas for the same steam. A fire-side inspection was booked for later in November.
What it could be. This is an open case, and the value of working it is in the method. Okonkwo listed the possibilities and what would distinguish them:
| Possible cause | What would point to it | What would rule it out |
|---|---|---|
| Soot on the fire side from a burner fault | CO rising, flame shape changed, trim at the edge of its band | Low and steady CO, normal flame |
| Scale on the water side | Boiler water hardness or a softener problem in the period, rising tube metal temperatures | Normal hardness tests and softener records |
| A failed gas baffle letting gas bypass the tube bank | Sudden rather than gradual rise; changed furnace pressure; hot spots on the casing | Gradual rise with no other change |
| Instrument error in the stack thermocouple | No change in fuel used for the same steam | Fuel per unit of steam rising with the temperature |
| Change in load or excess air | Load or oxygen different between the readings | Same load and oxygen |
Worked problem
What the rise is costing
If boiler 2 burns about 1.8 million m³ of gas over the rest of the heating season, and the rise costs 1.3 in 100 more gas for the same steam, how much gas is that, and what is its cost at a sample price of 0.30 dollars per m³?
Show solution
- \[1{,}800{,}000 \times 0.013 = 23{,}400 \text{ m}^3, \qquad 23{,}400 \times 0.30 = 7{,}020 \text{ dollars}\]
Answer: About seven thousand dollars at the sample price, if nothing changes and nothing gets worse. The money is not the main reason to investigate: a rise of 18 degrees in a week means something inside the boiler has changed, and both soot and a failed baffle can get worse quickly. The case will be closed by the inspection, not by the arithmetic.
Okonkwo's own guess, written in pencil at the bottom of his list, was a baffle, because the rise had been fast and steady with no change in combustion. Grazyna did not say whether she agreed. "Good," she said. "Now you will learn something when they open it."
Case 4: the deposit at Keswold
What happened. During Keswold's fall outage, the steam turbine's Curtis wheel and first impulse stages were found carrying a thin tan deposit, mostly silica with some iron oxide, too thin to affect performance. The first-stage pressure trend had been flat for two years. A review of the HRSG's steam purity with the water treatment supplier was planned for after the outage.
What to look for. Using lessons 7 and 8, the possible routes for silica into the steam are vaporous carryover from a drum running high in silica, mechanical carryover in droplets during level excursions or start-ups, silica in the attemperator spray water, and silica breakthroughs from the mixed-bed polishers. Iron oxide points to corrosion products released in start-ups and carried by the steam or spray. Each route leaves a different pattern in the trends: a drum silica trend that tracks the steam silica; steam silica spikes at start-ups or load swings; steam silica that tracks the feedwater rather than the drum. Ingeborg's three-line trend was designed to separate them.
Lesson. A deposit too thin to matter is a warning that costs nothing to heed. The time to find its source is before the next inspection shows it thicker, and before a valve stem collects enough of it to stick.
Case 5: a pipe that thinned for years
What happened. On December 9, 1986, at the Surry nuclear power station in Virginia, an elbow in the condensate line to a main feedwater pump ruptured without warning, releasing hot water and steam that killed four workers. The pipe wall had been thinned by flow-accelerated corrosion, then called erosion-corrosion, over the years of the plant's operation, at a location that had never been inspected because nobody expected it to thin. In 2004, at the Mihama nuclear plant in Japan, a condensate pipe ruptured for the same reason and five workers died; that location, too, had been left out of the inspection program.
Causes. The immediate cause was wall thinning by FAC in carbon steel carrying fast, hot, oxygen-free water. The underlying cause was an inspection program that did not include the susceptible locations.
Changes. Utilities and industrial plants now run FAC programs: they identify every carbon steel line in the susceptible range of temperature, velocity and chemistry, model the expected wear, measure wall thickness at the most susceptible points by ultrasonic testing, replace thinned components with chromium-bearing steels that resist FAC, and adjust water chemistry, as lesson 7 described for AVT(O). The same mechanism attacks HRSG low-pressure evaporators and economizers, and feedwater piping in any plant.
Worked problem
Remaining life from a thickness survey
A carbon steel feedwater elbow had a nominal wall of 9.5 mm when installed. Twelve years later, an ultrasonic survey finds 7.1 mm at its thinnest point. Its minimum required thickness for its pressure is 5.2 mm. Assuming the wear rate stays the same, what is its remaining life, and when should it next be measured if the plant's rule is to re-inspect at half the remaining life?
Show solution
- \[\text{wear rate} = \frac{9.5 - 7.1}{12} = 0.20 \text{ mm/yr}\]
- \[\text{remaining life} = \frac{7.1 - 5.2}{0.20} = 9.5 \text{ years}, \qquad \text{re-inspect in about } 4.7 \text{ years}\]
Answer: About 9.5 years of remaining life, with the next measurement in under five years. The calculation assumes the nominal wall was the actual wall when new, which is not always true, and that wear continues at a steady rate, which a change in chemistry or flow can alter. Real FAC programs use a measured baseline where they can, and treat any location that is thinning faster than predicted as a priority.
Case 6: the explosion that made the code
What happened. On March 20, 1905, a firetube boiler at the R. B. Grover shoe factory in Brockton, Massachusetts, exploded early in the morning shift. The boiler shot up through the building, which collapsed and caught fire; 58 people died and about 150 were injured. Boiler explosions were common at the time, and this one was the worst of many.
Causes. Accounts of the investigation pointed to a boiler that was old and had been repaired, and to its safety valve, which had been reported as defective or adjusted; boilers then were built and maintained to no common standard, and their safety devices were often neglected. What is certain is the response.
Changes. Massachusetts passed boiler construction rules in 1907, and other states followed. The American Society of Mechanical Engineers formed a committee in 1911 to write a uniform code, and the first edition of its Rules for the Construction of Stationary Boilers, the beginning of the ASME Boiler and Pressure Vessel Code, was published in 1915. The National Board of Boiler and Pressure Vessel Inspectors followed in 1919. Every code rule, inspection and licence that PWR-220 and this course have described descends from cases like this one.
Case 7: two from the mill
Grazyna's two cases from the mill were told without names or dates.
The puff. A gas support burner on the bark boiler would not light on its first trial one winter morning. The operator reset it and tried again, then again, and on the third trial there was a loud bang: a puff that blew out two observation port covers and lifted dust from every ledge in the boiler house. Nobody was hurt. The investigation found a pilot gas regulator frozen by condensate, so that the pilot was weak, and the main gas had flowed for the full trial time on each attempt; three trials had put more gas into the furnace than one. The changes were a heated enclosure for the regulator, a rule that a second failed trial on any burner called the shift supervisor, and a burner management system change that forced a full purge after every failed trial, which the old logic had not required.
The cold feed. A young operator, on a night shift years before Grazyna arrived, found the water out of sight in the gauge glass of a small package boiler whose feed pump had tripped. He restarted the pump and opened the feed wide. The boiler's tubes, which had been exposed, were already hot enough to be weakened; the cold water flashed, and two tubes ruptured. The low water cutoff had failed to trip, because its float chamber was blocked with sludge; it had not been blown down on the schedule. Grazyna had studied the report as a trainee. "The operator was taught that day what this course teaches you now," she said. "And the plant learned to blow down its cutoffs every shift and to test them properly, not just to tick the box."
What the cases share
Set side by side, the cases show the patterns that run through boiler operation:
- Slow changes hide in plain sight: a weakening scanner, a thinning pipe, a rising stack temperature, a silica film. Each became visible only because someone measured it and looked at the trend.
- Defences fail quietly: a layup whose residual faded, a low water cutoff full of sludge, an inspection program with gaps, a safety valve nobody tested. Testing the defences is as important as having them.
- Good catches come from people who believe the numbers: Desmond's water balance, the night operator who did not reset boiler 1 blindly.
- The lasting changes address underlying causes: a log sheet entry, a procedure that covers the drum opening, a code, an inspection program.
Okonkwo's notebook for the module ended with the case method written out on one page and Grazyna's question copied above it: what happened, why, what should have caught it sooner, and what was changed. Below that he had added a line of his own: and what would I have done at three in the morning? He planned to sit the boiler papers in the spring.
Common mistakes
MistakeStopping a case analysis at the immediate cause.
InsteadKeep asking why until you reach something that can be changed: a test, a procedure, a design or a habit.
MistakeTreating a defence that worked as proof that all is well.
InsteadAsk also which defences failed or were missing; the next event may not meet the same defence.
MistakeClosing an open case on a guess.
InsteadRecord the possibilities and what would distinguish them, and close the case on the evidence.
MistakeAssuming a location is safe because it has never been inspected.
InsteadUninspected locations are where FAC and other slow damage go unseen; inspection programs must cover the susceptible points.
MistakeResetting and retrying a burner that will not light.
InsteadEach failed trial adds fuel to the furnace; purge, find the cause, and follow the limit on retries.
Summary and key terms
- A good case analysis sets out what happened, the immediate cause, the underlying causes, the defences and the changes.
- Boiler 1's flame failure led to trending scanner signals; boiler 3's pitting led to a revised layup procedure covering drum openings.
- Boiler 2's stack temperature rise is an open case, worked by listing possible causes and what would distinguish them.
- Keswold's deposit calls for tracing silica through the make-up, drum, spray and steam trends before it thickens.
- FAC ruptures at Surry and Mihama led to FAC programs with thickness surveys and remaining-life calculations.
- The Grover shoe factory explosion of 1905 led to the ASME Boiler and Pressure Vessel Code; a puff and a cold feed show why trials and low water rules are strict.
- Immediate cause
- The physical failure or action that directly produced an event.
- Underlying cause
- The condition or decision that allowed the immediate cause to occur or go undetected.
- Defence
- A device, test, procedure or person that prevents an event or limits its consequences.
- Water balance
- Comparison of feedwater flow with steam and blowdown flows to detect unaccounted water.
- FAC program
- A systematic program of identifying, modelling and measuring locations susceptible to flow-accelerated corrosion.
- Remaining life
- The time until a component reaches its minimum allowable condition at its current rate of damage.
- Open case
- An event under investigation whose cause has not yet been established.
Practice problems
Work each problem on paper first, then reveal the answer to check.
Feedwater flow is 9,800 kg/h, steam 9,300 kg/h and blowdown 400 kg/h. What is unaccounted for, as a fraction of feed?
Show answer
100 kg/h; 100 ÷ 9,800 = 0.010.
A pipe of 8.0 mm nominal wall measures 6.4 mm after 10 years; its minimum is 4.8 mm. What is its remaining life?
Show answer
Rate 0.16 mm/yr; (6.4 − 4.8) ÷ 0.16 = 10 years.
A boiler's stack temperature rise costs 1.0 in 100 more gas on 2.5 million m³ a year. How much gas is that?
Show answer
25,000 m³.
Name three things to compare to tell soot from scale as the cause of a rising stack temperature.
Show answer
CO and flame appearance, water hardness and softener records, tube metal temperatures or the change over time.
Lesson quiz · 5 questions
Score 60% or more to complete this lesson.
Create a free account to save your progress, take quizzes and earn a certificate.