Bad batches aren’t inevitable. They happen because of inattention to any one of just four possible areas. Four categories. That’s not too many to keep in mind. Let’s look at them one by one.
Recipes. For routine batches, only one thing is expected of the recipe: that it has previously made a satisfactory batch. That original batch met all the quality analysis metrics and was packaged and shipped successfully. Flawless is nothing more than a dependable copy of this original result. But what about scale-up? There we need a modification of the recipe done at formulation; an adjusted recipe needs to be created, one that enables the desired product to be manufactured at market-needed volume. Not an easy task, but an essential one. Indeed, I have an introductory blog on that subject.
The outcome of successful scale-up is, in fact, the recipe able to make that satisfactory original batch. Once done, the recipe category can be considered a given and the cause of bad batches directed to other areas.
Ingredients. The two key tasks here are (a) are you adding the named ingredient directed by the recipe? and (b) is that ingredient really what you think it is? With the attention on inventory control these days, the use of bar coding is commonplace. Everything is traceable from the loading dock to the stock room to the scale to the factory floor. These routine procedures address “a” very well. On the other hand, sometimes it is important to look beyond the name of the product – or even the name and purity. Sure, for a neat ingredient, say, ethylene glycol, a purity number (95%) and disclosure of the impurity (water, 5%) may be enough. But polymeric materials, e.g. surfactants, need more detailed specification, such as the size distribution and the degree of branching. Thus, a simple staple ingredient may be described in the same way when re-ordered: but is it really the same product? Commonly, the vendor will certify purity, though some ingredients may be preferably re-checked by in-house technicians. For critical ingredients used to manufacture high-value-added products, at-line analytical instruments such as fiber-based optical spectrometers can be do the last-minute quality check. I’ve done that on all incoming ingredients primarily to assure that the precisely correct one was admitted to the tank in precisely the correct amount.
With careful attention to the criteria above – correct selection of a duplicate ingredient – this category will not be a source of troubled batches.
Hardware. The ingredients need to be exposed to the same forces and energy every time the batch is made. If the thermal probe reads the wrong temperature or the agitators are wobbling; if the steam or cold-water valves are stuck or the vacuum has a leak, the batch is likely to fail. The standard way to assure operability is preventive maintenance. The precise procedures are highly site-dependent. Most factories are diligent about maintaining their equipment, realizing that the net cost of such maintenance is likely to be significantly less than the cost of lost batches.
To improve machine dependability even further, batch manufacturers can now adapt predictive maintenance. One especially cost-effective way to do so is by collecting real-time data during batch manufacture and earmarking specific algorithms to evaluate the status of the electro-mechanical components.
Overall, the frequency of bad batches due to defective hardware can be brought down very low.
Control. Proper management of recipe, ingredients and hardware require a disciplined approach, but can be straightforwardly and objectively addressed. Bad batches due to these can be driven way down. The most challenging problem of all is control: the method of executing the recipe steps.
The optimum solution is automation: performing each step in the procedure with absolute computerized precision. This is rarely done in batch manufacturing because conventional ways of writing recipes are complex, though my group has made it simple for dedicated batch manufacturers to write recipes in minutes.
The HMI solution, increasingly seen these days, masks the fact that control is not really automated; it is the compounder who pushes the icons on the PLC screen that is really in control. All the errors attributable to humans – including being called upon mid-batch to address problems elsewhere in the factory – are present. True, HMI replaces dials and switches located in multiple places, with a central command screen, but the setting of the parameters and the timing of the steps are subject to the same human variability they always were.
The advent of data collection and creative algorithms to go along with this manual procedure enhance accountability. The quality analyst reviewing the data can see exactly what happened to the batch. Furthermore, the compounder, knowing that his work is auditable, is newly incentivized to assure compliance. A third benefit of data collection is the opportunity to do predictive maintenance, as mentioned earlier.
Absence of automation leaves control the most vulnerable of the four categories, though data collection and processing can go a long way toward reducing risk.