That Day, UPH on the MES Was Like a Rollercoaster, and My Heart Nearly Stopped
“Ah Qiang, come look at this MES report. What’s going on with the UPH for yesterday’s batch of 12-inch wafers? It’s like a sauna, surging to 300 one moment, then plummeting to 50? Weren't the preceding processes supposed to be smooth?” Xiao Lin, our new process engineer, pointed at the screen's fluctuating, ECG-like curve, his face pale. To be honest, my face wasn’t much better when I first saw this monstrosity. Back then, we frequently encountered such situations: one day the boss would be praising soaring yields, only to receive a customer complaint email the next, reporting another delivery delay. Equipment utilization followed suit, sometimes running at full capacity, other times idling, waiting for material. Not only was the production line chaotic, but the quality of the engineers' work also resembled a rollercoaster; you never knew when the next headache would strike.
Where's the Problem? Is Your Production Line Playing 'Musical Chairs'?
Simply put, this is the typical mess caused by 'Batch Production'. We used to operate this way, habitually producing hundreds of pieces of the same product (e.g., Type A wafers) in one go, solely to 'push for volume,' and then changing over the line to produce Type B. Just think, how much time is wasted on line changeovers, machine parameter adjustments, and recalibrations?
The point is, when you focus on producing large quantities of a single product at once, it might seem highly efficient, but it actually leads to downstream processes being either 'starved' or 'stuffed.' Starving means no material to process, leading to idle machines; stuffing means too much material, overwhelming capacity, and creating piles of work-in-progress. This production model is like playing musical chairs: all chairs are removed at once, then brought back at once, causing everyone to scramble for a spot, which is utterly inefficient.
How Is It Actually Done? Cut the Cake into Small Pieces and Share It
The solution is 'Heijunka' (Production Leveling), a concept frequently discussed in Lean production. In simpler terms, it's a 'scheduling strategy for mixed-model production.' It doesn't ask you to produce all products at once, but rather to 'level' the production volume of different products across each day and each shift.
How to do it:
- Calculate Average Demand: Suppose you need to ship 1000 pieces of Type A, 500 pieces of Type B, and 200 pieces of Type C in a month. If you work 20 days a month, then on average, you need to produce 50 pieces of Type A, 25 pieces of Type B, and 10 pieces of Type C daily.
- Plan for Smallest Production Batches: Don't produce 50 pieces of A all at once, then 25 pieces of B. Instead, break down these 50 pieces of A further; for example, intersperse 5 pieces of B after every 10 pieces of A, then 2 pieces of C, and so on.
- Establish a Visual Board: This is crucial! Back then, we used a 'Heijunka Box' on our visual board to clearly indicate which product and quantity needed to be produced during each time slot. You see, many companies now use electronic display boards, where even equipment engineers know which material is next to be processed, allowing them to prepare in advance.
In other words, you're not aiming for 'fastest production,' but rather 'most stable production.' After we implemented this strategy, the UPH curve on the MES genuinely resembled a smooth, flat line. Equipment utilization also surged from the original 60% to 85% and was sustained at that level.
The Most Common Pitfall: Chasing 'Maximized Single-Product Efficiency'
Frankly speaking, when we first started implementing it, we encountered the most resistance. Many process engineers would complain, "Won't the increased frequency of changeovers take more time? Our Type A product already has a Cpk of 1.08, which is good enough, why bother switching back and forth?"
This is the blind spot. You only see the Cpk of a single product, but you miss that the 'flexibility' of the entire production line is being sacrificed. Frequent changeovers do indeed increase some setup time, but these losses are offset by greater benefits such as 'reduced inventory,' 'shortened lead times,' and 'reduced idle costs.' When production batches are smaller, if a problem occurs, the loss is also smaller, and the response speed is faster. On one occasion, due to this improvement, the DPMO for a certain batch dropped from 6210 to 230, because the problem was quickly detected when it happened and did not escalate further.
One Thing You Can Do Today
Go back and look at your production line's MES report. If your UPH is like a rollercoaster, try 'cutting your schedule into smaller pieces' for production!
Article Category: Lean Production