That Day, After Conveyor Belt PM, We Found the Yield Dropped from 99.8% to 99.5%
I recall a few years ago, on one of our production lines, the wafer conveyor belt would experience a slight decrease in yield after every PM (preventive maintenance), typically dropping from around 99.8% to 99.5%. While 0.3% might not seem like much, in our industry, it directly impacts DPMO, skyrocketing it from 2000 to 5000, which greatly displeased management. Each time, equipment technicians would report no issues, and the PM checklist was duly marked, yet wafers would occasionally shift off-center or experience subtle vibrations during transport, ultimately resulting in one or two defects on the Wafer Map. During that period, the production line manager and I were constantly at odds; he insisted it was an equipment issue, while the equipment team maintained it was a process issue. In essence, neither knew "what the problem was."
Where Was the Problem? It Was About Missing the PM 'Nuances'
In reality, the PM plan for conveyor belt systems seems straightforward on the surface, involving just three items: tension, alignment, and bearings. However, truthfully, the 'nuances' of these three are the key. Many people think simply tightening screws and straightening the belt is sufficient, but the devil is in the details. Consider this: as wafers travel on the belt, even slight deviations accumulate into significant impacts. This isn't assembling LEGOs; it's about ensuring every single wafer travels 'as if on level ground.'
How to Actually Do It? Let the Numbers Speak
So the crucial question is, how do these three items need to be executed 'precisely'?
- Tension: You cannot rely solely on touch. We later purchased a digital tension meter that provides a direct kilopond-force (kgf) reading. Assuming the manufacturer's specification is 100 ± 5 kgf, your PM target is to keep it within this range. However, a more advanced approach is to measure at different speeds, as belt tension can change slightly during high-speed operation. We once found that some equipment had 102 kgf at low speed but only 90 kgf at high speed – that's where the problem lay!
- Alignment: This, even more so, cannot be done by feel. We required the equipment team to use a laser alignment tool to scan both sides of the conveyor belt from end to end. If a deviation exceeding 0.5 mm was found, adjustments were mandatory. Many people overlook the parallelism of the belt's two sides, only checking for skew in the middle. But frankly, if the two sides are not parallel, the belt experiences uneven stress, leading to deformation over time, and consequently, wafers will inevitably travel crookedly.
- Bearings: This is the most easily overlooked item. You cannot solely listen for abnormal noises; some bearings, when they first start to fail, don't produce obvious sounds, but their internal rolling elements are already wearing down. This causes subtle vibrations in the conveyor belt, too small to be detected by the naked eye. We later introduced a vibration analyzer, inspecting the bearing's spectrum every six months. If abnormal peaks at specific frequencies were detected, a proactive replacement was carried out even without abnormal noise. I recall one instance where a bearing's Cpk dropped to 1.08, and we replaced it immediately, preventing potential future production line downtime.
The Most Common Pitfall: PM Report Checked, But Not 'How It Was Checked'
The most common pitfall I've encountered is when the PM report is checked, but there's no indication of 'how' it was checked. Equipment engineers, pressured by time, might just feel the belt with their hand and deem tension acceptable, or visually inspect it for straightness and consider alignment acceptable. Frankly, it's not that they don't want to do a good job, but they haven't been required to verify with 'numbers' and 'tools.' Another point is that many people forget 'environmental factors.' If there's a large machine vibrating nearby the conveyor belt, or significant temperature and humidity fluctuations, these will affect the belt's stability, yet they are often not considered in the PM plan.
One Thing You Can Do Today
Add the requirement for 'quantitative tools' to your conveyor belt PM process; stop relying solely on feel!