Making cream cheese is an intricate process—one that requires a mix of artisanal insight and production knowledge.
Manufacturers must cultivate the cheese like artisans but do so in enormous batches—thus the need for production know-how. Mass producing cheese requires machinery and elaborate processing plants that are difficult to manage. So to better understand their operations, dairy companies are increasingly turning to data analytics. Sight Machine recently helped one of the world’s largest dairy producers optimize its processing plants.
The separation process
The recipe for cream cheese is fairly standard. Lactic-acid bacteria are added to milk, which lowers the pH levels and causes it to coagulate into curds.
Put simply, machines separate the curd from the whey—the watery part of milk that remains after the formation of curds. The whey is then drained off, the curds get heated, and cream cheese is formed.
When ramped up, however, this separation process is far from simple. Raw milk varies from cow to cow, and the proportion of fat in milk shifts with the seasons. That variation, in turn, affects how long each vat of cream must ferment. If you let it ferment too long, the cheese will emerge with high viscosity and a sour flavor. If you remove it too soon, you end up with watery cheese. Because of these variables, chemists and engineers at the dairy plant were running into bottlenecks, especially in terms of:
- The blend of protein and fat in the raw materials
- The temperature of the batches coming out of the upstream cookers
- The pH of output from the upstream fermentation process
To help them, Sight Machine’s experts designed a digital model of the dairy plant’s entire production process. They placed sensors on the machines and collected data on every aspect of the processing lines. The experts trained the model to analyze cooker and fermentation data, from which they could make recommendations for improving the lines. For instance, they determined optimal temperatures and release times for the fermenters; optimum speed and run times for the centrifuges used in the separation process; and accounted for daily production schedules and variations in the arrival of raw materials (milk and additives) to the plant.
Data models, AI and optimized production
Sight Machine was delighted to find that integrating process and batch data allowed the company to optimize production settings for its manufacturing process, both upstream and downstream. We maximized the quality and throughput of its production process—the mixing, blending, and packaging of cheese. Combining chemistry and statistics, we also calculated the optimal amount of water to add to the vats; the best time to clean the vats to avoid waste (of curds and water); and ways to decrease the dairy company’s energy use, thereby enhancing sustainability. Our models, moreover, interpreted data from raw materials, also derived from sensors, to forecast completion times for the vats of cheese.
Working with chemists at the plant, the data science team also designed a physics-based digital twin of the machinery that produces the cheese. The digital twin modelled essential information such as how milk acts inside the separator, how long it needs to ferment, and when to drain the tanks.
To get technical for a moment, Sight Machine used the dairy company’s chemistry-based models and statistics to predict optimal fermentation times for the cheese. Sight Machine’s data scientists used Monte Carlo simulations (models used to predict probability when random variables are present), and general linear models (GLM) to optimize performance. They used the GLM’s to account for variables that have error models beyond normal distribution. The team also used algorithms and AI to analyze and combine daily production schedules, real-time plant data, and variable raw material arrival. This helped the diary company dynamically optimize asset utilization and minimize energy use.
Paradise by the dashboard lights
Additionally, Sight Machine’s experts designed a dashboard that generated automated recommendations for the employees on the production lines. The dashboard recommended optimal times to run and stop the machine’s lines and alerted workers when to begin and end the fermentation times. Those recommendations enabled a continuous flow of production.
Manufacturing cream cheese is indeed an intricate and tricky process. But now, along with artisanal insight and production know-how, dairy and food processing companies have the pinpoint accuracy of data analytics. In the end, Sight Machine helped this global dairy company increase the yield of its product line by 5 percent. Our experts empowered the company to optimize production and enhance profits—squeezing margins from every vat of delicious cream cheese.
