Automatic Tablets Inspection Machine

Our company buy a tablet inspection machine. sensitivity 70um,
What should I check in OQ, PQ?
For example
To prepare 500 tablets(preparation a good tablets of 400 and bad tablets of 100)

  1. reject all bad tablet?
  2. How should the permitted range for good tablets? 400tablets +, - ?
    Please tell me what is check OQ, PQ?

For OQ - you somehow need to show that bad tablets are rejected at 100% - you can do this by introducing bad tablets (somehow if possible) or causing the machine to generate bad tablets. You should check that all defects which can be detected and rejected, are actually detected and rejected as expected. This is probably a very manipulated type run (artificial settings are put in which are not typical). This OQ doesn’t show the machine operating normally - it shows that rejects are appropraitely detected and rejected.

For PQ - The best practice is to use your acceptance criteria for your commercial release of the product, and increase the sampling to a tighter confidence level. This ideally is based on statistical acceptance levels. If using ANSI (for example) you can go from a level I inspection (for commercial manufacturing) to a level II inspection level (for PQ). If your criteria is a “serious” failure, then leave it as a serious failure, just increase from level I to a level II. Overall, this PQ should be done on an extended run (full operation, at full speed, at at least 1/10 size,) but you might also wanted to run a full production lot as an “endurance run” if the machine has not alraedy been run extensively during FAT/SAT. There is no exact rule here, just that your sample size can be considered representative, statistically.

Make sure you have the following questions ask the following, and in the following order
1 - What are the defects that could happen and which we want to inspect for?
2 - Are they cosmetic, or serious, or critical (and could hurt someone)?
3 - How big is our OQ and PQ lot size going to be?

With the information above, you should be able to use published statistical tools to determine
1- sample size
2 - accept/reject allowable.


Dear sir
thank you very much. i have one,
How it does evaluate the performance of vision camera?
vision camera have sensitivity 70um pixel…
I think check the OQ.
wait your comments…

Gotcha. Sorry about the misunderstanding. My comments were regarding the tableting press OQ/PQ. For the vision system, I would definitely tighten the accept/reject rate considerably.

I think the OQ would test the vision system somewhat in isolation of the tablet press, and the PQ would test the vision system operating in conjunction with the tablet press (like an endurance run or full commercial run done in triplicate, done concurrently, and done under quarantine until 3 lots are fully released).

My comments above regarding sample size are still valid. For a vision system, I would expect all attributes to be considered “critical”, and would increase the sample size to Level II or even tighter (if using AQL statistics). The vision system is the last line of defense to capture any faulty tablets, and a high level of assurance should be put on the inspection system.

Doing an FMEA, three criteria for risk are (1) probability to fail (2) impact of failure and (3) probability of detection. A well designed in implemented vision system increases the probability of detection and reduces the overal risk for many failure modes. Overall, I would require the vision system to be as reliable and trustworthy as possible, and this is done (in part) by a very strict qualification acceptance criteria.

1 Like