That day, when the Cpk report came out, the whole room fell silent for three seconds, and then I decided to fall in love with DOE
I remember a time when our new process reactor immediately encountered problems upon startup. We initially followed the supplier's recommended parameters, but the yield was only 70%, and the critical impurities in the product were excessively high, far exceeding the specifications. Everyone's faces turned green looking at the Cpk report of only 0.82. When the boss asked what to do, I thought, in this situation, should we try parameters one by one using brute force? By the time the test results came out, it would be too late. The only way was to call upon our old friend—DOE.
Where the problem lies, simply put, is "not knowing who the boss is"
To be honest, the most troublesome aspect of chemical processes is the sheer number of variables. Temperature, pressure, reaction time, catalyst concentration, stirring speed… just picking a few parameters can lead to hundreds of permutations and combinations. In the past, we often used "One Factor at a Time" (OFAT), for example, adjusting temperature first, and then adjusting pressure once we felt it was optimal. Frankly, this approach is extremely inefficient.
Consider this: if you adjust both temperature and pressure simultaneously, and the yield increases, how do you know if it's solely due to temperature? Or solely due to pressure? Or perhaps it's only after their "combination" that a better effect is produced? Simply put, DOE is a systematic method that helps you identify which of these parameters is the true "boss" influencing the outcome, and even uncover if there's any "collaboration" between them, leading to unexpected impacts.
Therefore, the key point is that the core idea of DOE is to enable you to find the critical factors (Main Effect) influencing the process with the fewest experimental runs, and even identify any interaction effects (Interaction Effect) between factors.
How is it actually done? The numbers speak for themselves
Taking the case of excessively high impurities, we selected three factors for DOE: reaction temperature, catalyst concentration, and reaction time. For each factor, we set two levels: high and low.
- Factor Settings:
* Catalyst Concentration: 1.0% (Low) / 1.5% (High)
* Reaction Time: 3 hours (Low) / 4 hours (High)
- Experimental Design: We ran a 2^3 full factorial design. You don't need to understand too much mathematics; just know that this means we performed 2 to the power of 3, which is 8 experimental runs. Compared to testing parameters one by one, where each parameter might need to be tested at high, medium, and low levels, this already saves a tremendous amount of time.
- Data Analysis: After completing the 8 experimental runs, we analyzed the data using statistical software. The results showed that the interaction between catalyst concentration and reaction time was extremely significant! Individually increasing catalyst concentration slightly reduced impurities, and individually increasing reaction time also did, but when both were set at high levels simultaneously, the impurities dramatically dropped below 0.05%, and Cpk soared to 1.58! This is a treasure that OFAT alone would never find.
In other words, the magic of DOE lies not only in telling you which parameters are effective but also in revealing what kind of "chemical reaction" occurs when these parameters "hold hands."
The most common pitfall is stubbornly trying brute-force methods
The most common pitfall I've encountered is when some new engineers, or those transferred from older departments, see DOE as "unnecessary trouble." They would rather rely on experience to "guess" which parameters are effective and then test them one by one. The result? Either a waste of valuable raw materials or a delay in the entire project's progress.
Another pitfall is selecting too many factors or setting levels too narrowly when designing experiments. This leads to an explosion in the number of experimental runs, taking an eternity to complete. The essence of DOE is to "minimize experimental runs while maximizing information extraction." You must learn to discard some less critical factors or set the levels a bit further apart to obtain the most valuable results within limited time and resources.
So, simply put, DOE is an investment—an investment of your valuable time and raw materials to achieve faster process optimization.
One thing you can do today
Think about whether you have any processes currently being adjusted by "gut feeling." Try to apply the spirit of DOE: list at least 2-3 parameters you believe have the most significant impact, and then simply plan how to thoroughly understand these parameters with the fewest experimental runs. You might just uncover some unexpected secrets!