Work Sampling in Solar Module Manufacturing
Study solar module manufacturing activities without confusing observed productivity with product testing or certification.
Updated:
A module line combines interconnection soldering, lay-up, lamination, assembly, testing and packing. Sampling describes observable state proportions; process times, thermal parameters and controls must come from validated procedures and equipment, not a generic productivity rule.
People, equipment and product
Separate the observed unit. During lamination, equipment may transform product while a person prepares the next batch. “Machine cycling” and “person waiting” are not mutually exclusive states of the same resource.
For people, define operation, inspection, preparation, movement, coordination, confirmed waiting, breaks and other. For machines, use production, changeover, input starvation, blocking, confirmed failure and planned stop. Rework and planned finishing need distinct rules; returning to a station does not always prove a defect.
Required inspections are not removed merely because they do not transform a module. Training, calibration and breaks are not automatically waste either.
Cadence and power differ
Teaching example: one departure every 25 seconds over eight continuous hours gives 1,152 modules before losses. At 600 Wp each, they sum to 0.6912 MWp nominal capacity per shift, or 2.0736 MWp over three equivalent shifts. These are not MWh of generated energy. Verify module specifications and operating conditions before extrapolating.
Size and design
For a proportion under independent Bernoulli records:
n = ceiling[z²p₀(1−p₀)/e²]
With z=1.96, p₀=0.08 and e=0.02, plan 707 records. For p₀=0.05 and e=0.01, plan 1,825. Using z=2 gives 736 and 1,900 respectively at approximately 95.45% confidence; this is not the same exact level.
Margins are absolute percentage points. Rare categories need a different sample for small relative error. Simultaneous intervals for multiple categories require explicit joint coverage; simply multiplying n by category count is not generally correct.
Five daily rounds × two observers × five days ×28 workstations produce up to 1,400 participations. They are not 1,400 visits or necessarily independent: stations may share failures or supply. Review three-shift coverage, route order and actual times. Do not exclude holidays or events merely because they are atypical if part of the target population.
Simulated 1,456-record case
| Defined state | Count |
|---|---|
| Direct operation | 698 |
| Inspection | 187 |
| Waiting during cycle | 96 |
| Preparation | 142 |
| Movement | 67 |
| Confirmed failure | 204 |
| Logistics waiting | 38 |
| Calibration | 17 |
| Breaks/other | 7 |
| Total | 1,456 |
The table assumes one unit and compatible rules; real studies must not mix machine and person states. Failure is 204/1,456=14.01%, with an approximate normal 95% margin of 1.78 points under independence. Instantaneous observation alone cannot establish that individual stops last under five minutes.
The 47.94% operation share is not OEE. To verify a reported 82% OEE, collect planned time, run time, ideal cycle and total/good units with reconciled definitions. Another OEE cannot be deduced from the table.
Quality and improvement evaluation
IEC TS 60904-13:2018 addresses electroluminescence image capture, processing and interpretation. It does not establish a required inspection percentage per shift here. Quality staff must check applicable standards, editions and control plans; sampling does not certify modules or permit omission of test requirements.
A scenario assuming OEE increases from 56% to 71% describes fifteen points, not demonstrated WorkSamp results. A €245,000 assumed investment needs costs and realizable flows for payback; four months requires more than €61,250 net monthly. Capacity is not cash.
WorkSamp supports planning and recording. Check synchronisation before deleting local data; offline capture does not mean everything is server-confirmed.