A pharmaceutical process that has been scaled-up is monitored using the results of in-process controls. These are tests that, for a particular run, either conform or do not conform to a pre-set limit. If the test conforms, the operators proceed to the next instruction in the batch process, but if the result is nonconforming, the result is reported to the manager who decides what action to take next. For an in-process test, the actions to be taken are characteristically thought out in advance. If there is no possible corrective action for an out of specification result, the test is not a proper in-process test but rather just a datum that may be part of the analysis of the result when the final outcome is known.
When a process has been 'optimized', it is taken for granted that it will operate within its control limits and no testing beyond its in-process ones will be required to guide the operators to a successful result.
Pilot Plant Experiments and Forensic Testing.
Although the chemical plant or kilo-lab process can be modelled using a laboratory scale procedure, it cannot be optimized without results from representative samples from the scaled-up process taken at critical decision points in the process.
The pilot plant runs are still experiments, even if the equipment is handled by personnel who are not research chemists. Although the chemist may think that (s)he understands, for practical purposes, the experimental reaction and subsequent purifications steps being scaled-up, this is at least immodest and usually foolish. The experimenter is wisest who anticipates the most potential problems and in the beginning collects samples at every convenient sampling point when the process is executed on scale. That means taking many more samples than simply those which will remain mandatory for in-process control. These extra samples we will call forensic samples because they are very often only analyzed when the result is disappointing in some respect. These samples can determine what went wrong and how to correct it.
Forensic samples are carefully stored so they will not deteriorate. Forensic samples are not analyzed while the process is running, but are for retrospective testing by the process chemists and may never be looked at if the process runs without any deviation.
There is only one downside to the collection of forensic samples. If the process runs perfectly and gives a product of the exact same quality as the laboratory samples but with a lower yield, the question may arises whether these samples could be predominantly responsible for the reduced yield. Most often the size of the samples or the mechanical losses that can occur when taking forensic samples cannot explain a noticeable reduced yield on scale. The samples are typically just not large enough compared to the size of the process. Simple calculations, in most cases, will give a pretty good idea how much yield could have been lost through this extras sampling.
Almost without exception the forensic sampling will provide an improved opportunity to understand what has caused any deviation from the desired result.
Thorough forensic sampling examines not just the physical phases where you expect the product to predominantly reside. Samples of the waste solids and liquids can also be valuable. Filtrates, wash liquids, filter cakes and head space gases may reveal where, for example, the missing product has vanished.
Once the process is set, these forensic tests should be removed from the batch instructions. The only tests that should remain in an optimized batch sheet are the in-process tests that are done at the decision points.
Sampling Problems
Obtaining a truly representative sample from a large reactor presents more difficulties than laboratory sampling. For safety reasons, plant samples must be taken without opening the reactor. Most often a dip-tube that runs many meters down into the reactor is used and the sample is removed by suction. There is no way to verify how effective the cleaning between samples is nor is there any way to guarantee that stratification or gradient has not been created in the sampling line.
Samples that are likely to contain unstable species need to be adequately quenched before labelling and storing them. Samples of distillates, filtrates, washings, and recovered processing solids are easier to handle.
No comments:
Post a Comment