Artificial Neural Networks for Monitoring pH and Acidity in Yogurt Fermentation

Strategic Insights -

Researchers from the Universidad Nacional del Altiplano have successfully developed artificial neural network (ANN) models to predict pH and titratable acidity in yogurt fermentation using CIELAB color parameters. The study, funded by the Consejo Nacional De Ciencia, TecnologiA E InnovacioN Tecnologica, aimed to improve process control in industrial yogurt production by providing a non-destructive approach for monitoring pH and acidity. The researchers trained 40 models for each output variable, using 90% of the data for training and 10% for validation, and achieved strong correlations between color and physicochemical changes.

Key Takeaways:

  • Researchers developed ANN models to predict pH and titratable acidity in yogurt fermentation using CIELAB color parameters (L, a*, b*).
  • The study used 90% of the data for training and 10% for validation, resulting in strong correlations between color and physicochemical changes.
  • The best pH model had two hidden layers with 28 neurons, achieving a high R[superscript]2 of 0.969 and a low root mean squared error (RMSE) of 0.007.
  • The optimal acidity model had four hidden layers with 32 neurons, achieving a high R[superscript]2 of 0.868 and a low RMSE of 0.002.
  • The study integrated ANN models and colorimetry to offer a practical solution for real-time monitoring of pH and acidity in industrial yogurt production.
  • The research has the potential to improve process control in industrial yogurt production by providing a non-destructive approach for monitoring pH and acidity.
  • The study was conducted by researchers from the Universidad Nacional del Altiplano, including Ulises Alvarado, Jhon Tacuri, Alejandro Coloma, Edgar Gallegos Rojas, Herbert Callo, Cristina Valencia-Sullca, Nancy Curasi Rafael, and Manuel Castillo.

Statistics:

  • 40 models were trained for each output variable (pH and acidity).
  • 90% of the data was used for training, and 10% for validation.
  • The best pH model had 2 hidden layers with 28 neurons.
  • The optimal acidity model had 4 hidden layers with 32 neurons.
  • The R[superscript]2 of the best pH model was 0.969, with an RMSE of 0.007.
  • The R[superscript]2 of the optimal acidity model was 0.868, with an RMSE of 0.002.

Sources:

  • Development of a Hybrid System Based on the CIELAB Colour Space and Artificial Neural Networks for Monitoring pH and Acidity During Yogurt Fermentation. Dairy, 2025,6(4):41.
  • NewsRx. Universidad Nacional del Altiplano Researchers Broaden Understanding of Artificial Neural Networks (Development of a Hybrid System Based on the CIELAB Colour Space and Artificial Neural Networks for Monitoring pH and Acidity During Yogurt ...). Journal of Engineering. September 8, 2025; p 2975.