Factory Digitalization Ver10. “AI Prescriptions Must Distinguish Between Controllable and Uncontrollable Parameters to Function Properly”
- shigenoritanaka3
- 4月23日
- 読了時間: 3分
Apr 23, 2026
Hello everyone,
In this article, I would like to discuss an essential point when introducing AI for the purpose of reducing defects:
AI prescriptions must be operated by clearly separating Controllable (directly adjustable) and Uncontrollable (not directly adjustable) parameters.
Although there are countless parameters that influence product quality, the number of truly critical parameters can generally be narrowed down to around 20. (This concept was introduced previously through a case study from a large foundry in South Africa.)
However, here is the important fact:
Many of these 20 critical parameters are Uncontrollable — they cannot be directly adjusted by operators.
In other words, even though these parameters have the greatest impact on quality, they are mostly result parameters that the shop floor cannot directly manipulate.
If this structure is not understood, AI prescriptions will not work. No matter how accurate the AI's recommendations are, operators will not be able to execute them, and quality will not stabilize.
1. Most critical parameters are “results” and cannot be directly adjusted
In a sand-casting foundry, for example, the following parameters are commonly recognized as critical to product quality:
Compactability (CB)
Sand temperature before molding
Product chemical composition (C, Si, Mg, S, P, Cu)
Compression strength
Permeability
Sand moisture
Pouring temperature
Sand to metal ratio (SMR)
Pouring weight / pouring cycle time
These are result parameters, emerging from upstream conditions. They are not values that operators can directly manipulate.
2. To influence Uncontrollable parameters, Controllable parameters must be identified
As noted above, critical parameters cannot be directly controlled. However, by identifying the Controllable parameters that influence them, operators can adjust these factors to bring the critical parameters into their target ranges.
3. Every critical parameter has corresponding Controllable parameters
✔ Compactability (CB)
Sand-bin retention time
Moisture level
✔ Pouring temperature
Furnace melting temperature
Ladle transfer time
Pouring cycle time
✔ Product chemical composition (Si)
Furnace Si level
FeSi inoculation amount
Thus, even if a critical parameter is an Uncontrollable result, there are always Controllable causes that influence it.
4. What AI should output is only the “target ranges” of the 20 critical parameters
The AI's role is to output target ranges for the critical parameters:
CB: target range
Pouring temperature: target range
Product chemical composition (C, Si, Mg, S, P, Cu): target range
Compression strength: target value
Permeability: target value
Moisture: target value
SMR: target value
Pouring weight / time: target value
AI does not need to instruct operators with commands such as: “Increase sand-bin retention time by 20 seconds.”
5. Aligning critical parameters to their targets is the responsibility of human judgment, not AI
To bring critical parameters into their target ranges, operators must monitor the relevant Controllable parameters in real time and adjust them accordingly.
Since the presence of AI implies that factory data is already visualized, operators and managers must determine:
“Which Controllable parameter should be adjusted to influence the critical parameter?”
Rather than blindly following AI outputs, operators must think for themselves and choose the optimal adjustments to influence Uncontrollable parameters.
This human judgment is essential for stabilizing quality.
6. Without an operational culture, AI prescriptions will not function
If operators do not observe data, think independently, adjust conditions, and verify results, the adjustments required to align critical parameters will not be executed. And without these adjustments, AI prescriptions cannot function.
Conclusion:
AI provides the target ranges for the critical parameters. Operators select and adjust the Controllable parameters, then confirm the results. At the center of this cycle lies human judgment.
Human growth × AI support = Organizational growth
Only when this structure is established can quality stabilize and digitalization produce real results.
🟦 Contact
If you have questions about this article, or if you would like to discuss digital transformation in manufacturing or the practical operation of AI prescriptions, please feel free to contact me at:
🟦 TAGS
コメント