Industrial Maintenance and Reliability Engineering Case Studies
Industrial5 Case StudiesUpdated Sep 2026
Reliability work sits where mechanical design, operating practice, and maintenance history meet. The cases come from plants that learned, sometimes late, that a failure is rarely one bad part. Each begins with a symptom someone had to explain: a cracked housing, a line that slowed while scrap climbed, a guard that passed inspection but failed a test, a milestone that moved, an energy bill that would not fall. Every analysis draws on design constraints, operating data, quality tolerances, and maintenance records. Watch the first questions each team asked, and the places where a reasonable answer proved incomplete.
Case Study 1
Cracked pump housings in high-ambient field units
Calder Vane Hydraulics, a mid-sized maker of gear pumps for mobile equipment, with two plants and a field service group.
The situation
Calder Vane had shipped its gear pump to mobile equipment builders for six years with few field returns on the housing. Then Marisol, who led the service group, began logging cracks in aluminum housings near the mounting boss on the drive side. The first reports came from units working in hot yards, where ambient temperature stayed high into the afternoon. Roughly 2,400 units were in service, and the return rate for cracked housings had climbed past one percent over two months. Most failures appeared after a few thousand operating hours, and none had caused a leak large enough to trigger the pressure alarm.
The challenge
The team had to separate three explanations: design margin, casting quality, and installation practice in the field. Engineering had already signed off the wall thickness against the rated pressure, and suppliers pointed to torque practice at the service level. Replacement housings were expensive, and the warranty group wanted a corrective action before the next build. Whatever the team proposed had to survive a question about which evidence supported it, not just a plausible story.
Analysis
The first useful step was to put the design constraints beside the operating data. The housing was designed for a rated pressure with a margin that looked comfortable on paper, but the drawing did not reflect how often the pump saw relief events. Telemetry from 310 instrumented units showed pressure spikes well above rated pressure at cold start, with the largest spikes in units that had been idle overnight. Quality tolerances added another layer. The casting drawing allowed local wall thickness down to a limit the foundry met on average but not at the boss radius, where sections cut from returned parts showed thinning and small shrinkage pores.
Maintenance records told a different story. Service logs showed that bolts at the boss were torqued to specification at build but rarely rechecked after the first hundred hours, and several field technicians had used a different thread lubricant than the one assumed in the torque table. Teams grounded in Industrial Maintenance and Reliability Engineering usually start by asking which evidence each candidate cause would predict. Preload loss predicts cracks that start at the bolt hole and follow the thread line. Porosity predicts cracks that begin inside the boss wall. The fracture faces from returned housings showed both patterns, which pointed to a combination rather than a single culprit, and changed the list of corrective options.
- Pressure spikes Cold starts push relief events well above rated pressure.
- Boss fatigue Thin radius sections accumulate cycles faster than predicted.
- Preload loss Bolts relax after early service, changing load paths.
- Porosity sites Internal voids start cracks and reduce effective thickness.
Actions taken
- Section twenty returned housings at the boss radius and map wall thickness and pore size against the casting lot numbers before any design change is released.
- Add a radius increase and a local rib to the boss, sized with a fatigue check using measured pressure spike histograms rather than the rated pressure alone.
- Issue a revised torque procedure with a specified thread lubricant, a second torque check at the first service interval, and a torque label on the housing.
- Tighten the foundry acceptance criteria at the boss, with a radiograph sample from each lot and a reject rule for porosity above the agreed size.
Results
The corrective package went into the next production build. The boss radius was enlarged, a local rib was added, and the revised housing passed cyclic testing at rated pressure with the measured spike profile applied on top. Field returns for cracked housings dropped steadily over the following months, and the remaining cases traced to units serviced outside the new torque procedure. The service group addressed those through dealer training and a torque label on the housing. The telemetry link stayed in place so that any rise in spike frequency would show before the next crack.
Questions for discussion
- Which field evidence would have to appear before you accepted the design as the primary cause rather than a contributor?
- How would you set a foundry acceptance limit for porosity when the casting drawing and the failure mode disagree on what matters most?
- What telemetry would you keep after the fix so that a return of the problem shows up early?
Case Study 2
Deburr bottleneck behind rising bracket scrap
Northfield Precision Stamping, a supplier of formed steel brackets for vehicle subassemblies, running two press lines and one finishing cell.
The situation
Northfield had run its bracket line for several years with steady output, but over the last quarter the finished-part rate fell while the scrap rate rose from under one percent to a little over three percent. The line had three presses, a washing station, a deburr cell with two robotic brushes and one manual bench, and a final gauge station. Shift reports showed that presses ran near capacity most of the day, yet the deburr cell was the only place where parts queued. Operators attributed the scrap to die wear. Maintenance pointed to brush replacement intervals that had been stretched to save consumables.
The challenge
The team had one quarter to bring finished output back to demand and to bring scrap under the customer limit of 1.5 percent. Capital for a fourth press was not available, and the deburr cell could not simply run faster without changing the burr profile that the gauge checked downstream. Every change had to show up in measured data within the same shift, because the plant could not afford a long trial on customer orders.
Analysis
Tomas mapped the process from coil to gauge, and the constraint was plain. Press cycle time averaged 9.2 seconds with a narrow spread, so the presses were not the source of variation. Deburr cycle time ranged from 14 to 31 seconds depending on brush condition and on how often the manual bench stopped for adjustment. Burr height at the gauge followed the same pattern. The mounting edge allowed 0.08 millimeters of burr, and parts from the manual bench exceeded that limit far more often than parts from the robotic cells. The main variation sources were therefore brush wear, uneven fixture loading, and rework that sent parts back through the queue.
The brush log and the die log added a second layer. Brush replacement had shifted from a count of parts to a visual check made once at the start of each shift, so worn brushes ran for hours unnoticed. Die sharpening intervals had also been stretched during the same period. The team ranked candidate fixes by two measures: how much each would raise deburr throughput, and how much each would narrow burr height at the gauge. Fixed-count brush replacement scored high on both, while a fourth press scored well on throughput alone and would have left the variation untouched.
Actions taken
- Add a parts-processed counter to each deburr brush set, with replacement at a set count and a confirmation step at the start of each shift.
- Pair the manual bench with a fixture that locates the mounting edge, then rebalance part loading so the two robotic cells and the bench finish within 2 seconds of each other.
- Move die sharpening back to the original interval, and log each sharpening against burr height readings from the first ten parts after the change.
- Set the acceptance criteria before the trial starts: deburr cycle at or below 16 seconds, burr height inside 0.08 millimeters for ninety-five percent of parts, and scrap under 1.5 percent over five consecutive shifts.
Results
Within the trial window, the deburr cycle settled near 15 seconds and its spread narrowed to a band the bench operators could hold without adjustment. Burr height stayed inside the limit for a little over ninety-five percent of parts, and scrap fell to about 1.2 percent. Finished output rose to around 51,000 brackets a week, close to demand, without a new press. The brush counter stayed in place as a standing control, and a monthly review now compares burr height trend against sharpening records so that drift shows before it reaches the gauge.
Questions for discussion
- Which measurement would you trust most to show that the deburr cell, not the presses, is limiting output, and why?
- How would you decide whether the manual bench should be automated or simply fixtured when volume is still uncertain?
- What would change your priority ranking if the gauge began rejecting parts for a reason unrelated to burr height?
Case Study 3
Guarding and interlock review for a new palletizer cell
Brightwater Food Packaging, a regional maker of sealed pouches and jars with two filling lines and a shared warehouse.
The situation
Brightwater planned to add a robotic palletizer to the end of its second filling line, with capacity near 1,200 cases per hour at full speed. The project team had designed the cell with a fenced perimeter, a light curtain at the infeed, and a pallet exit door with a mechanical stop. Commissioning was scheduled for the third week after drawing approval. Two weeks before the review, Dana, the plant safety lead, asked for the full risk assessment, control validation records, and the training file for operators and maintenance staff. Only part of that package existed. The fence drawing had been revised twice, and the interlock test sheets were blank for the exit door.
The challenge
The review had to decide whether the cell could be approved as designed, approved with conditions, or held. Gaps existed in hazard identification, in the evidence that the control circuit met its required performance level, and in the documentation that would let maintenance enter the robot envelope safely during jams. Schedule pressure was real, and the project manager argued that the light curtain already covered the main hazard. The review had to separate what was proven from what was assumed.
Analysis
The hazard list was built around robot reach and the stored energy in the gripper and vacuum system, but it treated the pallet exit as a fixed-position hazard when a pallet could be removed while the robot was still moving. The risk estimate for that exit scored low because the stop was assumed to hold. Standards covering machinery guarding and control reliability require that assumed risk reduction be demonstrated, not drawn. The review therefore asked for a validated control path, measured stopping distance from the light curtain, and proof that the exit interlock removed motion before a person could reach the envelope.
Documentation was the second concern. The risk assessment had no record of the maintenance tasks that would occur inside the envelope, and the lockout procedure listed the main disconnect but not the pneumatic supply, which could hold pressure after power was removed. The plant's own approval gates required a signed risk assessment before design freeze and validation evidence before first production. Neither gate had been passed, and the schedule had assumed both would be waived on the strength of supplier certificates. Those documents described the robot, not the installed cell, and the review said so directly.
Actions taken
- Hold design freeze until the hazard list includes the exit door bypass, the pneumatic residual energy, and the jam recovery task, each with a named mitigation and owner.
- Commission a validation test that measures stopping distance from the light curtain and confirms the exit interlock stops all motion, with results recorded on signed test sheets.
- Write a lockout and residual energy procedure covering the pneumatic supply, and add a bleed valve with a visible gauge at the envelope access point.
- Schedule a second review gate before first production, requiring the signed risk assessment, validation sheets, and training file, with no waiver based on supplier certificates alone.
Results
The review approved the cell with conditions. The exit door interlock was rewired to remove power from the robot drive before the door could open, and the validation run recorded a stopping distance inside the figure the design had assumed, with margin. The pneumatic bleed valve was added, and the jam recovery method was written with the maintenance lead and tested on a mock-up. First production started in the fifth week, two weeks behind the original plan, and the project closed with a complete file. The plant took away a clear lesson: supplier paperwork answers a different question than a site review does.
Questions for discussion
- Which evidence would you require before accepting a supplier's performance level claim for a safety circuit you did not design?
- How would you weigh schedule pressure against an open hazard that has a low likelihood but a severe consequence?
- What belongs in a jam recovery procedure that the robot supplier's manual is unlikely to cover?
Case Study 4
Compressor skid delayed by a substituted valve
Ostrander Cold Storage Engineering, a design and build firm for refrigerated warehouses and processing rooms.
The situation
Ostrander was building a refrigeration upgrade for a food distribution warehouse. The scope included a new ammonia compressor skid, a set of oil separators, and control panels that tied into an existing plant room. The milestone that mattered most was the handover date for the cold room expansion, which was tied to a customer's inventory move. Four weeks before the milestone, the compressor vendor reported that a specific discharge check valve would ship late because of a casting shortage at its own supplier. The valve was rated for the design pressure and temperature, and the original delivery had been planned for the fourth month of the project.
The challenge
The milestone carried a penalty for each week of delay past the handover date, and the customer had already scheduled product moves. Ostrander had to choose between a recovery that held the date with a substitute valve, a partial handover that kept a rented compressor on temporary service, or a slip that accepted the penalty. Each option carried a different cost, a different engineering review burden, and a different risk to the design basis. The decision had to be made before the vendor's revised date arrived.
Analysis
The first question was whether the substitute valve could meet the same constraints. The design basis fixed the pressure class, the seat material compatibility with ammonia, the oil carryover limit, and the flow coefficient that kept discharge pressure drop inside the vendor's envelope. A valve from another manufacturer with the same nominal rating looked suitable, but its seat material and body casting differed. The coordination gap was that the controls team had programmed a position signal tied to the original part. A substitute would need a revised signal scheme, an updated control narrative, and a review of the start sequence before operating data could confirm it.
Quantifying the options came next. Holding the date with the substitute meant roughly 310 hours of engineering across design review, control changes, and commissioning, plus a premium for expedited freight. The partial handover cut penalty exposure considerably but needed a rented compressor for about eight weeks, and that rental cost came close to the penalty it avoided. A full slip cost nothing in engineering but accepted the entire penalty and pushed back the customer's inventory move. Maintenance records from the existing plant room added one more input: past oil return problems at low load meant the new separator behavior needed a dedicated test, which added a day to commissioning in the chosen plan.
- 1Vendor delay noticeShipment slips past the milestone with no firm replacement date.
- 2Design basis checkSubstitute tested against pressure class, seat material, and flow.
- 3Control signal reviewPosition feedback logic and start sequence need reprogramming.
- 4Trade-off comparisonCost, penalty exposure, and commissioning risk compared side by side.
- 5Approved recovery planEngineering sign-off, customer notice, and revised gate dates issued.
Actions taken
- Obtain a written substitute valve datasheet and matching material certificates, then run the design basis check against the original calculation package.
- Reprogram the check valve position feedback with a dual-sensor arrangement during site preparation, so the substitute can be tested before the compressor starts.
- Keep the rented compressor for the bridging period only, with a fixed return date written into the rental contract and a daily cost review by the project controller.
- Notify the customer in writing with the revised handover date, the decision basis, and the commissioning tests that will be witnessed, so the penalty discussion starts from evidence.
Results
Ostrander chose the substitute valve with a partial handover. The cold room expansion was handed over six days after the original milestone, which triggered a penalty for that period but avoided the larger exposure of a full slip. The substitute passed commissioning with discharge pressure drop inside the vendor envelope, and the reprogrammed position signal matched the physical valve through the first 60 operating hours. Ben, the project controller, returned the rental compressor on its scheduled date. Total recovery cost ran above the planned contingency but below the penalty a full slip would have produced. The customer accepted the revised plan because the commissioning evidence was specific.
Questions for discussion
- How would you decide when a substitute part needs a full design review rather than a documented equivalence check?
- What would you need from the vendor before accepting a revised delivery date as firm?
- At what point does a bridging rental become more expensive than the penalty it was meant to avoid?
Case Study 5
Compressed air waste cut without losing line pressure
Pell Valley Bottling, a beverage plant with three canning lines and a shared utilities building.
The situation
Pell Valley's compressed air system served filling, capping, and pneumatic conveying across the plant. Two compressors ran most of the day, with a third on standby, and the air share of the utility bill had grown as lines added valves and conveying branches. Jun, the utilities engineer, measured weekday air system draw at about 1,800 kilowatt hours, and a flow survey suggested a large share of that air was lost to leaks and to open blowoffs on idle equipment during shift changes. The sustainability goal was to cut compressed air energy by one fifth without lowering line pressure at the capper, which had a documented minimum for seal integrity.
The challenge
Several options were on the table: a leak program, controls to shut off blowoffs when a line stopped, a variable speed drive on the main compressor, and heat recovery from compressor cooling to the building. Each promised savings, but the capital, labor, and operational risk varied widely. The team had to choose a sequence, prove each step against pressure and uptime data, and avoid any change that caused a capping failure during a high-volume run.
Analysis
The analysis started with the system's own records. Maintenance logs showed that leak tags had been raised for years but seldom closed, and the flow survey found that one headering branch accounted for a large share of measured demand during idle periods. Pressure data from the capper line showed a minimum of 6.1 bar during high-speed runs, just above the seal integrity threshold. That margin meant any reduction had to be proven at the line, not only at the compressor room. A lifecycle comparison was also needed, because a drive on a compressor that ran at low load for much of the day would pay back differently than one on a compressor holding a steady base load.
The options were compared on three measures: daily energy saved, capital and labor cost, and effect on pressure stability. A leak program and idle blowoff controls ranked high on savings per dollar and left line pressure untouched. A variable speed drive reduced energy at part load but added a control interaction that would need validation across every shift pattern, and its total cost was roughly four times the leak work. Heat recovery suited the building but did not reduce compressor energy, so it was deferred. The lifecycle view also changed the ranking, since the drive carried spare parts and service contract commitments that the leak work did not.
- First two weeksTag and repair the largest headering leaks found in the survey.
- Week threeInstall timed shutoff valves on idle blowoffs at all three lines.
- By week sixMeasure capper pressure and compressor load across two full shift patterns.
- Second monthDecide on the variable speed drive using the measured part load profile.
- Quarter reviewCompare realized savings against the target and set the next phase.
Actions taken
- Close the open leak tags in survey order, starting with headering branches, and record the flow reduction measured after each repair on the leak log.
- Fit timed shutoff valves on blowoffs at the three lines, with a manual override and an interlock that restores air when a line restarts after a fault.
- Log capper line pressure at one-second intervals through two full shift patterns before and after the changes, and set a floor of 6.0 bar that stops any further reduction.
- Hold the drive decision until the part load profile is measured, and require a lifecycle cost that includes spare parts, control validation, and service contract terms.
Results
Leak repairs and blowoff controls together cut compressed air demand by about one fifth, which met the target. The compressors ran fewer hours at base load, and weekday electricity use for air fell from about 1,800 kilowatt hours to a figure near 1,400. Capper line pressure held above the 6.0 bar floor through every measured run. The drive was deferred after the part load profile showed that base load stayed high enough to limit its payback. Jun kept the leak program as a standing task, with monthly tag reviews.
Questions for discussion
- How would you test whether a reduction in compressed air demand has reached the point where an end user notices a loss of performance?
- Which lifecycle costs are easiest to leave out of a drive proposal, and how would you bring them into the comparison?
- What would you expect to happen to the leak program's effect over a longer horizon, and how would you keep it from fading?
Learning Resources
Industrial 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.