Factory Digitalization Ver13. “How to Incorporate Seasonal Factors into AI Learning”
- Shigenori Tanaka
- 5月27日
- 読了時間: 2分
May 27, 2026
Thank you for reading.
Today’s topic is “How to incorporate seasonal factors into AI learning.”
Products manufactured in clean‑room environments—such as food, pharmaceuticals, and semiconductors—are almost unaffected by outside air conditions.
However, in iron foundries, which operate in open environments directly connected to outside air, product quality is significantly influenced by season, weather, and even time of day.
To give an extreme example, temperature and humidity can be completely different between morning and midday on the same day. In summer, high humidity changes the moisture content of sand. In winter, dry air affects sand compaction. On rainy days, mold drying slows down; on windy days, drying accelerates.
In short, many “external disturbance factors” directly affect casting quality.
■ Seasonal factors should ideally be reflected in AI prescriptions
AI prescriptions generate optimal operating conditions based on past production and quality data. However, if the learning data does not sufficiently include seasonal, temperature, humidity, or weather variations, AI cannot produce prescriptions that account for seasonal effects.
Ideally, AI should learn from:
Temperature and humidity captured by external sensors
Seasonal sweet spots
Time‑of‑day variations in optimal conditions
And then adjust prescriptions according to the season.
■ My own attempt — but it was hindered by “rare timing”
In the past, I worked with an AI partner company to incorporate seasonal factors into AI learning.
However, at that time, the factory I supported happened to be in a rare timing: the customer’s products were replaced with new ones roughly every six months.
Because the product cycle was unusually short during that period, it was difficult to accumulate production and quality data for the same product across different seasons, and the amount of data was insufficient for AI to learn seasonal effects.
The initiative ended halfway, but I believe that if sufficient data had been available, the frequent casting defects seen in summer could have been reduced even further.
■ For factories producing the same product for several years, seasonal AI is highly effective
On the other hand, factories that manufacture the same product continuously for several years are in a very different situation.
They can accumulate data across multiple seasons
They can observe correlations between temperature/humidity and quality
They can identify optimal conditions for each season
In such factories, incorporating seasonal factors into AI can be extremely effective.
In many foundries, defects tend to increase during summer. If AI can understand seasonal patterns, it can provide more accurate prescriptions to address this recurring issue.
🟦 Summary
Foundries are strongly affected by outside air
Season, temperature, humidity, and weather directly influence quality
AI prescriptions should ideally change based on seasonal factors
Short product cycles make seasonal learning difficult
Long‑term production of the same product enables effective seasonal AI
Seasonal AI could further reduce summer defect spikes
🟦 Contact
If you need practical support with AI prescription design that incorporates seasonal factors, external sensor utilization, or data‑accumulation frameworks, please feel free to contact us.
🟦 TAGS
コメント