Work Sampling in Wind Turbine Maintenance
Design observations for wind maintenance with attention to access, safety, tasks and study coverage.
Updated:
What a wind-maintenance study can measure
Work sampling can estimate how team time is distributed between intervention, travel, preparation, waiting and safety. It is useful when the question concerns that distribution rather than the exact duration of a repair. Feasibility depends on observing the selected activities safely without favouring the most accessible wind farms.
It is not the only possible method: work orders, fault logs and continuous measurement can provide complementary information. The proportion of observations involving direct intervention is an activity share, not “maintenance OEE”. OEE combines availability, performance and quality for a production process; it cannot be calculated by dividing interventions by observations of people.
1. Define the scope before counting
Specify the wind farms, technicians, shifts, activities and period represented. Decide whether the objective is a particular campaign or an annual estimate: observing one week of preventive work does not automatically represent corrective work, standby periods or another season.
An observation records the state of a selected unit at a selected instant. If a visit captures several technicians in the same crew, their states may be correlated. Do not treat every row as an independent observation without examining that dependence.
The schedule must account for routes, permissions and actual access times. One observer cannot simultaneously record thirty technicians across remote wind farms. Selecting convenient locations first and completing the planned count afterwards leaves bias that a large sample cannot remove.
2. Sample size and precision
For initial planning of a proportion under simple random sampling and approximate independence:
n₀ = z² × p₀ × (1 − p₀) / e²
With z = 1.96, p₀ = 0.5 and absolute margin e:
| Planning margin | Observations, rounded upwards |
|---|---|
| ±5 percentage points | 385 |
| ±3 percentage points | 1,068 |
| ±2 percentage points | 2,401 |
These are references under the stated assumptions, not guarantees for every design. Dependence within crews, sites or moments requires design-aware analysis and may reduce effective information. Precision for each wind farm depends on its own sample; distributing 1,068 records across five farms does not provide ±3 points for each.
Finite population correction applies only to sampling without replacement from a relevant finite frame. Two hundred technicians are not two hundred time units: do not use headcount to reduce observations of their activity. Report an interval appropriate to the findings; NIST explains the Wilson interval for proportions.
3. Operational categories without treating safety as waste
This is a proposal to adapt and test, not a mandatory taxonomy:
| Category | Recording criterion |
|---|---|
| Intervention | Work on the asset within a defined scope |
| Preparation and logistics | Tools, spares and work preparation |
| Travel and access | Travel and authorised access |
| Waiting | Pending resource or condition, with a reason |
| Documentation | Reports and technical administration |
| Safety and training | Briefings, checks and training included in scope |
Define priorities for ambiguous situations and test agreement between observers. A necessary safety check does not become waste because it is indirect. Include initial and refresher training if they belong to the target time; exclude them if the justified scope differs, explaining the exclusion.
Access training and authorisation depend on the role and operator. The GWO BST standard describes basic safety competencies; it does not itself establish a universal access permit or weather limits for every turbine. Coordinate observation, prevention and manufacturer procedures before entering.
4. Teaching example and decisions
Assume 1,000 valid records within the defined scope: 600 intervention, 100 preparation, 100 access, 80 waiting, 70 documentation and 50 safety. Intervention accounts for 60%. This does not establish a “good” sector performance, recoverable savings or turbine availability.
Check whether waits cluster around spare parts or access and investigate causes against work orders. A later intervention needs another comparable period and uncertainty analysis. Do not convert a hypothetical reduction in waiting into confirmed hours saved, extra output or return on investment.
5. Checks before reporting
Record missed selections and their causes without automatically replacing an unobservable person with an accessible one. Distinguish lack of connectivity from lack of visibility. Report coverage, denominators, dependence and exclusions alongside percentages.
WorkSamp supports capture planning and state recording; validity depends on study design and execution. For fault duration and asset availability, use suitable start-and-end records. The useful study result is a decision supported by data and explicit limitations, not a universal productivity percentage.