Mechanical Systems and Equipment Manufacturing Case Studies
Industrial5 Case StudiesUpdated Sep 2026
Mechanical equipment makers learn most of what they know after a machine leaves the building. The five studies here come from fictional shops in mechanical systems and equipment manufacturing, each with its own customers, tolerances, and maintenance habits. Every case draws on the same four kinds of evidence: the design constraints that were promised, the operating data returned from the field, the quality tolerances parts actually held, and the maintenance records showing what happened after handover. The figures belong to the fictional organizations, so treat them as examples of method rather than benchmarks.
Case Study 1
Gearbox output shaft fractures on a bottling line
Corvane Drive Systems, a fictional maker of right-angle gearboxes and drive modules for bottling and packaging lines.
The situation
Corvane builds right-angle gearboxes and drive modules for bottling and packaging lines. One model, rated for a continuous output shaft load of 1,850 newton meters, had passed internal fatigue testing before release. In its first season in service, field returns reached eleven units out of about 2,400 installed. Every returned shaft showed a fatigue crack starting at the keyway root, a location the original finite element model had marked as low stress. Two customers had added a second bottle size, changing the start-up torque profile, and several ran longer shifts than the specification assumed. The service group had no failure data tied to serial numbers, so the first job was rebuilding what had happened in each field unit.
The challenge
Engineering had to decide whether the fault lay in shaft geometry, the material lot, hub assembly torque, or the load assumptions, and it had to decide before the next production run was scheduled. Field replacement was expensive and would damage the product's standing with the three customers involved. Any redesign also had to keep the existing mounting envelope, because customer frames could not change.
Analysis
The first pass set the design assumptions against field evidence. The rated torque came from a steady-state case, but controller logs showed start-up peaks about 38 percent above it, several times per shift. Measured fillet radii on returned shafts sat at the low end of the drawing band, and stress concentration rises fast as that radius shrinks. Hardness and case depth were within tolerance, which pointed away from the heat treatment lot. The shaft met its own drawing while failing the load it actually carried.
Engineers trained in Mechanical Systems and Equipment Manufacturing would treat the maintenance records as measured evidence, not background. Service logs showed two affected sites had replaced couplings and skipped the hub re-torque after the first month. Loosening at the hub moves load onto the shaft shoulder and could explain part of the crack initiation. Corrective options were a larger fillet with a modified keyway, a shrink-fit hub with no keyway, or a limited start ramp in the drive controller. Each carried different costs in mounting envelope, tooling, and retrofit time, so the team compared them side by side.
- Season one startField returns begin at lines running the new bottle size.
- Controller log reviewStart-up peaks run about 38 percent above rated torque.
- Fillet inspectionReturned shaft radii sit at the low edge of the drawing band.
- Hub torque auditTwo sites skipped re-torque after the first month.
- Rig validationRevised ramp passes 2,000,000 cycles without crack initiation.
Actions taken
- Pull serial-level service histories for all 2,400 installed units and match each return to its bottle size, install record, and hub torque log.
- Instrument two lines with strain gauges on the output shaft for one full production cycle, and map captured peaks against controller logs.
- Revise the drive controller start ramp to cap peak torque, then validate on a fatigue rig at 1.4 times the measured peak for 2,000,000 cycles.
- Issue a field service bulletin requiring hub re-torque at the first and third months, and add a torque check to shaft fit-up on the assembly line.
Results
The strain gauge data on two instrumented lines confirmed start-up peaks reaching 40 percent above rated load with the new bottle size. A fatigue rig run at the revised ramp limit passed 2,000,000 cycles without crack initiation. Of the eleven returned shafts, nine showed loosened hubs or fillet radii at the low limit, which made the combined cause credible. Service calls on the installed base fell to three in the following quarter, all traced to a retrofit that had been missed. The geometry change first proposed would have added tooling cost without touching the peak, which became a useful point at the next design review.
Questions for discussion
- Which assumption in the original finite element model would you have checked first, and what field data would have exposed it earliest?
- When is a service bulletin an acceptable fix, and when does it simply move the risk onto the customer's maintenance crew?
- What evidence would have justified changing the controller ramp instead of redesigning the shaft?
Case Study 2
Valve body line loses throughput as defects climb
Pelham Valve Works, a fictional maker of cast iron valve bodies for water and HVAC systems, with one foundry and a machining hall of six cells.
The situation
Pelham turns rough castings into pressure-rated valve bodies in a machining hall that feeds water and HVAC customers. Over eleven weeks, output from the bottleneck cell, a five-axis mill that finishes the seat bore, fell from about 210 bodies per shift to about 165, while scrap and rework rose from under 2 percent to 6.5 percent. The seat bore carries a diameter tolerance of plus or minus 0.02 millimeters and a concentricity requirement tied to the sealing face. In the same period the foundry changed its sand mix and incoming casting supplier. Downstream, the inspection room was seeing more bodies fail the bore check, and the cell supervisor had begun holding parts back for manual gauging.
The challenge
The plant manager wanted the bottleneck recovered without a second shift, which the customer contracts could not absorb. The quality group had three suspects: casting variation, tool wear on the boring bar, and a fixture clamp that had been reset during a spindle service. Each suspect pointed to a different fix, and each fix had its own lead time.
Analysis
Mapping the process began with the part's own constraints. The seat bore references the machined datum face, so any shift in the datum appears as concentricity error rather than size error. Cell controller data showed bore operation cycle time up about 22 percent, concentrated in the rough boring pass, where the tool ran longer between offset corrections. Casting hardness held within specification, but incoming porosity ratings drifted toward the upper limit, which increases interrupted cuts. Cycle drift and tolerance drift arrived together, which suggested one shared cause rather than three separate ones.
Teams trained in Mechanical Systems and Equipment Manufacturing treat the maintenance log as a measured variable. The log showed the boring bar swapped twice in the period and the fixture clamp reset during a spindle service with no torque recorded. Fixture clamping ranked first because it explained both the concentricity drift and the longer cycle. Tool wear came second, since inserts showed normal flank wear. Casting variation came third, handled by tighter porosity acceptance at the gate rather than a new foundry contract. Each fix then carried a criterion: concentricity within 0.015 millimeters over 30 consecutive parts, cycle time within 3 percent of baseline, and rework below 2.5 percent over two weeks.
Actions taken
- Reset the fixture datum with a calibrated locator set, and record clamp torque on the setup sheet for every changeover from now on.
- Replace the boring bar and inserts when measured bore size drift crosses a set limit, not on elapsed hours, and log each change against the cell.
- Add incoming porosity acceptance at the foundry gate, sampling 12 castings per lot before release to the machining hall.
- Run a two-week trial with control charts on bore diameter and concentricity, fixing the release criteria before the trial begins so the result cannot be argued afterward.
Results
After the datum was reset and clamp torque recorded, concentricity on the next 30 consecutive parts stayed within 0.015 millimeters, meeting the first criterion. Bore operation cycle time came back within 3 percent of baseline by the end of the second week. Rework fell to 2.4 percent over the trial window, and output rose to about 205 bodies per shift. Porosity screening at the gate rejected roughly 4 percent of lots in the first month, a result the foundry disputed until the sample data were shared. The plant kept its single shift, and customer delivery commitments held throughout.
Questions for discussion
- How would you set acceptance criteria for a bottleneck cell when the contract limits added shift time?
- When two suspected causes explain the same drift, which test separates them in the fewest production days?
- Who should own a supplier gate that relies on sampling done at the receiving plant?
Case Study 3
Hydraulic press review before customer acceptance
Tolland Industrial Presses, a fictional builder of hydraulic press lines for metal forming, with a design office and customers in three countries.
The situation
Tolland builds hydraulic press lines for forming steel brackets and housings. Its newest 600-ton model was scheduled to ship to a customer that planned two shifts on it. The design office had completed most of the technical file, but the review panel found gaps before the approval gate. The light curtain muting logic for the loading station had been tested on a bench, not on the installed machine. The relief valve setting on the drawing differed from the value on the commissioning sheet. The risk assessment credited a two-hand control for the ram crush hazard, yet the wiring at one station used a single-channel input. Two of three guard interlocks had no documented test records.
The challenge
The panel had to decide whether the machine could ship on its date, which depended on a final acceptance visit, and whether each gap was a paperwork problem or a real hazard. It also had to confirm which standards the machine was declared against, since the customer's site inspector would check against requirements the design team had not used.
Analysis
The control architecture set the first constraint: a single-channel input cannot support the performance level the assessment assumed, since one fault could defeat the stop unseen. Bench data showed light curtain response times inside sensor specification, but the muting window was wider than the stroke-based calculation allowed. The relief setting carried a tolerance of plus or minus 5 bar, and the commissioning value sat at the top edge of that band. Together they showed the file described a machine that had not yet been tested as built.
Reviewers who read the declared standards first find the weakest link quickly, since every other control is judged against them. The file declared conformity with EN 693 and ISO 13849-1, which set the performance level target for the guarding functions. Service records from a sister site showed interlock nuisance trips had led to bypass keys in service, a maintenance hazard to design against. The panel set three approval gates: documentation complete, installed-machine testing passed, and a signed declaration before release. Each gate had an owner and a record the customer's inspector could request.
- Hazard assessment Confirm hazards, controls, and assumed performance levels before any build.
- Design verification Check wiring, settings, and drawings against the assessed risks.
- Installed testing Witness muting, stopping, and relief tests on the real machine.
- Dossier release Declare standards and release the file only after every test passes.
Actions taken
- Rewire the station input to a dual-channel safety module with cross monitoring, then recalculate the achieved performance level for the ram crush hazard.
- Repeat light curtain muting and stopping time tests on the installed machine, recording response at the boundary of each muting window.
- Reconcile the relief valve setting across drawing and commissioning sheet, issuing one signed value with the test certificate attached to the revision.
- Hold shipment until the technical file holds validated test records for all three interlocks and a signed declaration against EN 693 and ISO 13849-1.
Results
The rewired station reached the performance level the assessment had assumed, and the cross-monitored input caught a simulated contact fault during the witness test. On the installed machine the muting window was narrowed to match the stroke calculation, and stopping distance stayed inside the guard clearance with margin. The relief valve was confirmed at a single value inside the drawn tolerance. Shipment moved about three weeks later than planned. The customer's inspector accepted the file at the first visit, and the panel added a standing gate requiring installed-machine testing before any future approval.
Questions for discussion
- Which gap in this review would you have escalated before the on-site test, and what would have triggered that choice?
- How much weight should a declared standard carry when the site inspector applies a different set of requirements?
- What belongs in a dossier to show that a control works on the installed machine rather than only on the bench?
Case Study 4
Tubesheet delay and recovery options for the exchanger
Quellan Heat Exchangers, a fictional maker of shell-and-tube heat exchangers for process plants, with a fabrication yard and a design group of about 60.
The situation
Quellan was building a large shell-and-tube exchanger for a refinery customer, with a factory acceptance test serving as the milestone that released a major payment. The tube bundle depended on two forged tubesheets from one supplier. The first forging arrived with hole pattern deviations, and the second sat in heat treatment queuing at the forge. The design group had also revised nozzle reinforcement after a customer comment, which moved two welds into the sequence the yard had planned for the bundle. The test date was now at risk by about six weeks, and the contract penalty started accruing from the first day past the grace period.
The challenge
Project management needed a recovery path quickly, because every week of delay carried a known penalty and risked pushing the test into the customer's plant outage window. Expediting the forging, resequencing the yard, adding overtime, or holding the plan each traded cost against schedule in a different proportion, and none of them addressed the drawing tolerance on its own.
Analysis
The hole pattern carried a position tolerance tight enough that the rejected forging could not be reworked in the yard without losing flatness on the gasket face. Yard data put bundle assembly on a 19-day critical path, with tube expansion the longest step, and both moved welds sat on it. Inspection records showed hole pattern deviation of up to 0.4 millimeters in one quadrant, against a drawing tolerance of 0.15. Supplier correction was therefore unavoidable, and the open question was how much of the schedule could be recovered around it.
Recovery options should be compared on expected cost per week saved, not on total cost alone. Maintenance records for the expansion rig showed two unplanned stops last quarter, both from seal wear, so overtime on it risked losing the time it was meant to buy. The team modeled four options: premium forging, resequenced welds, overtime across both shifts, and holding the plan. Resequencing recovered about nine working days at low cost. The premium forging recovered the most time at the highest cost, while overtime alone recovered little because the rig was the constraint.
Actions taken
- Resequence the nozzle reinforcement welds to follow tube expansion, and confirm the new order against the welding procedure qualification records before work starts.
- Order a replacement tubesheet forging from a second approved supplier and accept an 18 percent premium on that line item to protect the test date.
- Service the expansion rig seals in a planned two-day window before overtime begins, so extra shifts run on a machine in known condition.
- Issue a drawing revision that fixes the hole pattern tolerance, and send it to the forge in writing so the corrected pattern is checked on every lot.
Results
The resequenced welds recovered nine working days on the critical path, and the expansion rig ran the planned overtime without an unplanned stop. The replacement forging arrived in the sixth week after the decision and passed dimensional check on receipt. The factory acceptance test slipped four working days, well short of the six weeks first projected. The forging premium and overtime added about 7 percent to the remaining labor and materials budget, partly covered by the customer's change order. The penalty applied only to those four days.
Questions for discussion
- How would you compare a premium supplier with a resequenced schedule when both recover similar time at different risk?
- Which penalty terms would change your recovery choice, and when would you negotiate them?
- Where should a drawing revision sit in the recovery plan so the same tolerance problem does not return on the next lot?
Case Study 5
Cutting test bay energy and casting scrap
Morrow Fluid Equipment, a fictional maker of oil-free scroll compressors for food and beverage plants, with one assembly plant and a six-stand test bay.
The situation
Morrow assembles oil-free scroll compressors for food processing lines. Every unit runs on a test stand for about 40 minutes before shipment, and that energy is not recovered. The test bay accounted for roughly 28 percent of plant electrical use in the last period, and its share grew as order volume rose. Separately, machined aluminum scroll housings were scrapped at a rate of 3.6 percent, mostly from porosity found during leak testing, and each scrapped housing carried its full casting and machining cost. Management asked engineering to cut both energy and scrap without lengthening the test or relaxing the leak limit.
The challenge
The customer's leak and efficiency limits were fixed, and the leak test could not be relaxed. Any change to test duration or load profile also had to keep the same data format, because the customer's audit files compared results across batches. The team compared three options: regenerative drives on the test stands, a shorter load profile validated against full runs, and tighter casting control upstream.
Analysis
Options should be scored on several axes, because an energy saving can still add scrap or downtime. The leak limit corresponds to a small fraction of rated flow, so the test cannot be shortened without losing sensitivity. Stand controller data showed about 60 percent of energy per unit went to holding the motor at test load for the full run, with the rest going to ancillary heating and cooling. The average run took 41 minutes, and the performance section accounted for most of that time. Porosity leaks usually showed in the first few minutes of the leak section, so stand time went to housings bound for scrap.
Drive unit records showed a steady replacement interval, and a regenerative unit would add a service item with its own history. The team scored each option from 1 to 10 across energy, scrap, validation effort, and lifecycle burden. Regenerative drives led on energy but needed the longest validation. The shortened profile scored well on energy and validation but risked the audit comparison. Upstream casting control scored lower on energy but highest on scrap, using a foundry relationship already in place.
Actions taken
- Add an early pressure decay screen at the start of the leak section so housings with gross porosity are rejected within the first three minutes.
- Validate a shortened load profile against twenty full-length runs, confirming efficiency and leak results match within audit tolerance before changing the standard procedure.
- Tighten melt degassing and inspection at the foundry for the scroll housing lot, sectioning a sample of castings each week to track porosity.
- Install a regenerative drive on one test stand as an eight-week pilot, logging energy and service events for a lifecycle comparison.
Results
The early pressure decay screen rejected porous housings in the first three minutes on most affected lots, returning roughly 30 minutes of stand time for each scrapped housing. Scrap fell from 3.6 percent to 2.1 percent over the trial, with foundry degassing changes accounting for most of the gain. The shortened load profile matched full-length results within audit tolerance across twenty runs, and the customer accepted the revised procedure. The regenerative pilot cut energy per tested unit by about a fifth but added a service item the team tracked separately. Test bay share of plant electrical use fell over the following period.
Questions for discussion
- Which measure would break a tie between an energy option and a scrap option that score the same?
- How would you keep an audit comparison valid after the test procedure changes?
- What service history would you need before committing to a lifecycle cost for a regenerative drive?
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.