The context
At an Enerjisa power plant in Turkey, a Six Sigma team investigated lost steam-turbine electricity output by adjusting heat-recovery boiler operating conditions within safe limits, as documented in the published case study.
This is an independent analysis of that project. The publication describes a single-plant application. The measurement-system assessment and multi-site extension below are proposed Black Belt additions, clearly separated from reported work. There is no claimed client relationship with Lean Six Sigma International.
The problem and its financial stakes
The published target was an additional 30 kWh in average hourly production and $25,000 in annual benefit, according to the project description.
The practical question was whether different operating settings could recover economically useful output. A Black Belt reviewing this charter would also ask what happened to fuel consumption, auxiliary electricity demand and equipment condition. Those measures belong beside the output target in the business case. Finance would need an agreed valuation method before the team selected its preferred operating regime.
How the Black Belt scope would be framed
For a comparable assignment, I would scope the process from heat-recovery boiler controls through steam delivery to net electrical output. Operations, instrumentation, maintenance, engineering and finance would share responsibility, with the plant manager sponsoring decisions that cross departmental boundaries.
The initial charter would cover an approved operating envelope, explicit stop conditions and a pilot on one turbine. Fleet deployment would be a subsequent decision requiring evidence from other sites. This keeps the first experiment manageable while making transferability part of the project from the outset.
The tools, phase by phase
Define: agree on the decision
A SIPOC and process map would identify where operating authority changes hands. The critical decision would be whether the proposed regime produces a repeatable increase in net output under comparable conditions, at an acceptable incremental cost and risk.
My proposed responsibility matrix would make the release decision explicit:
| Function | Evidence contributed | Decision supported |
|---|---|---|
| Operations | Stable operating windows and shift feedback | Whether the trial is executable |
| Instrumentation | Measurement uncertainty and signal checks | Whether the gain is distinguishable |
| Engineering and maintenance | Equipment limits and failure risks | Whether settings can be sustained |
| Finance | Incremental revenue and cost reconciliation | Whether the benefit is realizable |
Measure: establish whether the instruments can resolve the gain
The case team rechecked calibrated transmitters but did not conduct a separate measurement study, according to the measurement section. That leaves a useful review question: how small a production change could the complete measurement chain reliably detect?
For this assignment, I would assess the electrical meter and relevant pressure, temperature and flow signals across the trial range. Reference comparisons, repeat readings and checks across days would address bias, repeatability and stability. These are distinct elements of measurement characterization covered by NIST’s measurement guidance.
The data plan would also reconcile timestamps and aggregation intervals. Trial acceptance would depend on the uncertainty of the estimated gain being small enough to support the operating decision.
Analyze: separate controllable effects from operating conditions
The authors reported response-surface regression involving seven operating variables in their statistical analysis.
My review would examine the relationship between output and candidate settings alongside load, ambient conditions and condenser conditions. Residual plots would be central: NIST identifies them as a principal model-validation tool and cautions that a high R-squared alone does not establish adequate fit in its regression guidance.
A fleet extension would retain site and turbine identifiers. I would test whether a setting’s effect differs by site, then examine those differences with local engineers. A pooled average would be insufficient evidence for approving identical settings everywhere. Confirmation on another turbine would remain a separate acceptance step.
Improve: design a safe comparison
The published team changed high-pressure steam temperature and high- and low-pressure steam pressures, then compared 50 matched operating pairs under similar atmospheric and vacuum conditions in the improvement phase.
For a new project, the experimental plan would specify factor combinations, independent repeat runs, stabilization periods and the smallest economically useful effect. A factorial design could investigate interactions; evidence of curvature could justify a response-surface design. NIST discusses these different objectives in its experimental-design guidance.
I would agree on feasible randomization with operations and document any restrictions. Confirmation trials would compare the proposed settings with the existing regime while tracking safety and equipment constraints. At another site, the same protocol would test local applicability before release.
Control: make the change executable across departments
The case included operator calibration training and scheduled equipment checks, recorded in the improvement actions.
My proposed control plan would assign ownership of settings, alarms, instrument checks and benefit reporting. Shift supervisors would participate in drafting the operating procedure and trial handover. Maintenance would own inspection triggers; engineering would approve exceptions.
For multi-site deployment, each plant would receive a local acceptance record and rollback procedure. A persistent adverse shift in adjusted output would trigger a joint review of process conditions, instrument performance and adherence to the approved regime.
What the result was
The authors reported an additional 114 kWh in average hourly production and an expected annual return of $100,000 to $110,000, dependent on electricity prices, in their results section.
For financial approval, I would retain that distinction between reported production improvement and expected annual return. The benefit ledger would reconcile operating hours, saleable incremental energy, realized prices and additional costs. Any extrapolation to other plants would stay outside booked benefits until locally demonstrated.
What a trainee should take from it
A useful training exercise is to prepare the release recommendation for a second plant. The trainee would present the measurement uncertainty, experimental evidence, site-specific prediction and departmental sign-offs, then identify what remains unresolved. If the second plant responds differently, the next assignment is to explain that difference and propose a bounded follow-up trial.
