Quality or quantity? Traditionally, food manufacturers have had to face this stark tradeoff, making tough decisions between quality improvements and volume and profit considerations. With new machine learning technology, however, it’s possible to maximize production volume and yield without any reduction in quality.
Two main categories of machine learning solutions are typically deployed in the manufacturing sector: black box tools, which yield simple predictions without explanations (think about how Netflix recommends shows), and explainable tools such as Fero Labs, which provide users with more information and context.
You can think of these as similar to Six Sigma’s "Analyze," continuously learning from the production process and improving in real time.
But machine learning’s power goes way beyond simple analyses, extending to predictions, recommendations and even sustainability improvements.
Here are 3 reasons why food manufacturers should employ this valuable technology:
1. To predict—and prevent—quality issues.
Quality issues are one of the biggest risks for food manufacturers, with the potential to cause huge reputational and financial losses. In the US, the average food recall costs $10 million.
Machine learning software can predict quality issues hours ahead of time. Moreover, it can recommend the optimal settings to prevent future quality issues.
How does this work? First, the software analyzes all the data produced on the factory floor and “learns” how each factor, such as temperature or length of a certain process, affects the final quality. Using these learnings, the software can then predict quality violations in real time and tell engineers and operators how to prevent them, whether the solution is increasing the temperature or adding more of a specific ingredient.
These benefits aren’t just limited to one plant. A manufacturer may see quality varying across plants, but not know the reason. A machine learning solution can be scaled across all those plants to reveal the factors underlying quality variation, then optimize settings to standardize the final product across the manufacturer’s entire footprint.
2. To maximize production volume, without diminishing quality.
Reducing quality issues and waste is one way that machine learning can help food manufacturers boost volume and yield. Another way is to reduce cycle time. If machine learning software tells engineers they can produce the same quality item in half the frying time, they’ll be able to churn out more faster, without affecting quality.
Additionally, machine learning helps food producers deal with raw material variation, which can be another cause of low production volume. In the chemicals sector, a faulty batch of raw ingredients can be returned to the supplier for a refund; in food, however, the perishable nature of many food ingredients means that they must be used, regardless of any flaws. That makes it imperative to get the most out of each ingredient. A good machine learning solution will note those quality differences and recommend new parameters to deal with them.
3. To minimize energy consumption, without raising costs.
With climate change a growing C-suite priority, sustainability is top of mind for many manufacturers. Explainable machine learning software can reveal where sustainability improvements—such as reducing heat, or minimizing water consumption—can be made without any effect on quality or throughput. By tapping into these recommendations, factories can produce more food with the same amount of energy.
If this sounds like the earlier example about producing the same item in half the frying time, that’s exactly the point. We live in a world where efficiency, cost savings, and sustainability goals are interconnected. No longer do manufacturers have to juggle multiple priorities and make tough tradeoffs between quality and quantity; rather, they can make one change that optimizes all of these variables at once. That’s the value that machine learning brings to the table.
