If you are into pharmaceutical manufacturing then you must have faced ups and downs in the manufacturing process.
For instance one batch of drugs is manufactured perfectly with brilliant purity.
While in another batch you may face problems with lower quality that cannot be dispatched for usage.
According to the operators they had prepared both the batches with the same and exact guidelines they have been instructed with.
Usually each batch is of thousands of dollars, and process failure of any of the lot means loss of thousands of dollars.
Thus the plant manager and supervisor have to be on their toes round the clock.
Every time a batch failed, the supervisors would examine thoroughly the manufacturing records and find out what went wrong.
They are aware that during second batch manufacturing something has been changed but they are not able to identify it.
Many times they think that they have located the problem and fixed up the issue with process control, but when the work retains same problem persists.
If the process variability is not explained clearly then the 'out of control' processes would be impossible to repair, and the real underlying reasons cannot be found by only comparison of the batch records and observing the process control charts and trending data.
The main reason of process variability can be complex and complicated interactions among the various process parameters that do not have individual effect on the procedure.
For instance, process variability of an active pharmaceutical ingredient creating procedure may be output of mixture of residual water, temperature and concentration.
The overall effect would not be observed if any two of the factors were limited to the range that had caused variability.
There are certain steps that can be followed to increase your chances of finding out the main causes of the faulty process variability and figure out proper ways to bring the problem under control.
Here the quantitative statistical analysis and data mining are combined with the help of qualitative process expertise and proper evaluation and find out the real cause and effect relationships that had had alterations in process variability.
Firstly bring your attention towards the process variability that needs the investigation.
Jolt down the history of process parameters gained from many sources like raw materials testing, incident reports and batch records.
Prepare a robustness assessment report which would have data about in-depth processes and parameters.
Conduct statistical analysis of the database with the help of appropriate software to find out variations with relation of the process variability.
Create hypotheses and expand a predictive model.