Why don't the recovery figures add up to the total loss?

Because uptime and yield multiply rather than add. Some parts would have been rejected during time the line was not running anyway, and that overlap cannot be credited to either fix on its own. It arrives only when both are done. Any two-way split that does add up has quietly assigned the overlap to one side, and which side depends on nothing more than the order of the arithmetic.

Which should I fix first?

Compare the two single-fix rows, and then weigh them against what each costs to fix. At these defaults uptime is worth about six times yield, which is typical for a line with decent cure control, but a line with a marginal cure window reverses it, and the whole point of separating the rows is that the ranking is a property of your numbers rather than a rule.

Does this page know whether my parts cured?

No, and it is the most important thing it cannot do. Throughput arithmetic is indifferent to dose: run the belt above the cure window and this page reports a higher number while every part comes off undercured. Validate the speed on the cure-window page first, then treat the throughput as real.

How should I measure uptime?

Over a period long enough to include the bad days. Changeovers, jams, lamp warm-up after a restart and uncovered breaks all belong in it, and a figure drawn from a good week will overstate annual capacity by more than any other input on this page. If a line is planned on that number, the shortfall shows up as overtime rather than as a wrong calculation.

Are uptime and yield really independent?

Not entirely, and the assumption is worth stating because the interaction term depends on it. A line that stops often tends also to run unstable. Lamps restarting, temperatures moving, operators rushing, so poor uptime and poor yield frequently share causes. Where that is true the overlap is larger than the arithmetic here shows, and a combined project is worth more than either estimate suggests.

Not included. Some rejected parts can be re-cured or reprocessed into good output, which raises effective yield at a cost this page does not carry. Use the rework page for that cost, and be careful not to count the recovered parts twice, either they are good output here or they are rejects with a rework cost there.

UV Curing calculator

UV System Throughput Calculator: What Each Fix Actually Returns

Turn belt loading and line speed into net good throughput, and see what each improvement would actually be worth. Enter parts per belt foot, belt speed, line uptime and cure-related yield. The calculator returns gross and net throughput, then splits the shortfall three ways: what fixing uptime alone returns, what fixing yield alone returns, and what can only be recovered by fixing both. Those three are not the earlier version's two, and they do not sum the way its rows did, because the honest split of a shortfall between two multiplicative factors has an overlap that belongs to neither.

What this calculator does

Convert belt loading and speed into net good throughput, and split the shortfall into what fixing uptime returns, what fixing yield returns, and what only fixing both returns.
Use it for ranking an uptime project against a yield project, estimating capacity from belt loading and speed, explaining why two improvement estimates cannot simply be added, setting a realistic hourly rate for scheduling, testing what closing up part spacing is worth.
Convert belt loading and speed into net good throughput, and split the shortfall into what fixing uptime returns, what fixing yield returns, and what only fixing both returns.

Formula used

Gross throughput = parts per belt foot × belt speed × 60
Net good parts/hr = gross × uptime × yield
Recovered by fixing uptime alone = gross × (1 − uptime) × yield
Recovered by fixing yield alone = gross × uptime × (1 − yield)
Recovered only by fixing both = gross × (1 − uptime) × (1 − yield)

Inputs explained

Parts per belt foot: Parts occupying one foot of belt length, at the spacing actually run. Includes whatever gap the fixturing or the loading rate imposes.
Belt speed: Speed the line runs at. It must be a speed that cures. Check it against the cure window before treating this throughput as available.
Line uptime: Fraction of scheduled time the line actually runs, after changeovers, jams, lamp warm-up and breaks.
Cure-related yield: Share of parts that pass on cure. Count only cure defects here; other defect modes belong to their own processes.

How to use the result

Best suited to ranking an uptime project against a yield project, estimating capacity from belt loading and speed, explaining why two improvement estimates cannot simply be added, setting a realistic hourly rate for scheduling, testing what closing up part spacing is worth.
Cannot tell whether the speed cures. A throughput figure for an undercured line is a production rate for scrap. Assumes uptime and yield are independent. In practice they are often correlated. A line that stops frequently also runs unstable, so the interaction term can understate the real coupling. Says nothing about where in the line the constraint sits; the cure station may not be the bottleneck at all. Averages uptime over a period, so a line that runs perfectly then stops for a day reads the same as one that stutters continuously. Excludes rework, which returns some rejected parts to good output at a cost the rework page carries.

Common questions

Why don't the recovery figures add up to the total loss? Because uptime and yield multiply rather than add. Some parts would have been rejected during time the line was not running anyway, and that overlap cannot be credited to either fix on its own. It arrives only when both are done. Any two-way split that does add up has quietly assigned the overlap to one side, and which side depends on nothing more than the order of the arithmetic.
Which should I fix first? Compare the two single-fix rows, and then weigh them against what each costs to fix. At these defaults uptime is worth about six times yield, which is typical for a line with decent cure control, but a line with a marginal cure window reverses it, and the whole point of separating the rows is that the ranking is a property of your numbers rather than a rule.
Does this page know whether my parts cured? No, and it is the most important thing it cannot do. Throughput arithmetic is indifferent to dose: run the belt above the cure window and this page reports a higher number while every part comes off undercured. Validate the speed on the cure-window page first, then treat the throughput as real.
How should I measure uptime? Over a period long enough to include the bad days. Changeovers, jams, lamp warm-up after a restart and uncovered breaks all belong in it, and a figure drawn from a good week will overstate annual capacity by more than any other input on this page. If a line is planned on that number, the shortfall shows up as overtime rather than as a wrong calculation.
Are uptime and yield really independent? Not entirely, and the assumption is worth stating because the interaction term depends on it. A line that stops often tends also to run unstable. Lamps restarting, temperatures moving, operators rushing, so poor uptime and poor yield frequently share causes. Where that is true the overlap is larger than the arithmetic here shows, and a combined project is worth more than either estimate suggests.

Last reviewed 2026-08-25.