A multivariate statistical process control (MSPC) system for red wine fermentations combining forecasting, process boundaries and monitoring with infrared spectroscopy calibrations

Abstract

The wine industry continuously seeks innovative approaches to improve process monitoring and quality control, particularly during the fermentation process, where multivariate statistical process monitoring (MSPM) has emerged as a valuable tool [1]. Traditional monitoring methods relying on a few parameters have evolved with advancements in spectroscopy and multivariate analysis, allowing for the quantification of a wide range of compounds [2,3]. However, managing high-dimensional data remains a challenge, necessitating the development of more effective statistical approaches such as principal component analysis (PCA) and its extension, multiblock PCA. This study aimed to establish a novel monitoring system for red wine fermentation using multiblock PCA to define process boundaries and identify deviations with enhanced interpretability. To achieve this, a stepwise PCA approach was implemented, separating phenolic extraction and alcoholic fermentation into distinct analytical blocks. A single graphical representation was generated, enabling real-time assessment of fermentation progression and early detection of anomalies. Four test cases were analyzed, demonstrating the method’s capability to identify process deviations and potential faults based on their relative position within the established boundaries. Additionally, a probabilistic approach utilizing probability density functions (PDF) was integrated to define process boundaries and refine fault detection [4]. A forecasting model, based on first principle mathematical equations, was developed to predict fermentation trajectories using initial grape and juice composition data, enabling proactive process control [5] (Lambrecht, 2024). Sensitivity analysis revealed that while the system had over 99% accuracy in detecting anomalies, false alarms remained a challenge. However, its adaptability allowed for the customization of the process boundaries based on specific wine styles or varietals, addressing industry needs. Case studies using historical fermentation data illustrated the flexibility of the proposed system, as wineries could adjust process limits to balance precision and variability. The findings suggest that this novel MSPM approach enhances real-time decision-making in winemaking, providing a streamlined and intuitive monitoring solution compared to traditional PCA methods, which often require supplementary contribution plots and extensive post-analysis. Future improvements include integrating historical data for predictive analytics and automated corrective recommendations. This study presents a significant advancement in fermentation monitoring, bridging the gap between complex statistical models and practical winery applications, ultimately contributing to more efficient and data-driven winemaking practices.

References

[1] Reis, M. S., & Gins, G. (2017). Processes, 5(3), 35.

[2] Aleixandre-Tudo, J. L., Nieuwoudt, H., Aleixandre, J. L., & du Toit, W. (2018). Talanta, 176, 526-536.

[3] Lambrecht, K., Nieuwoudt, H., du Toit, W., & Aleixandre-Tudo, J. L. (2022). Journal of Food Composition and Analysis, 110, 104542.

[4] Lee, W.J., Mendis, G.P., Triebe, M.J. & Sutherland, J.W. (2020). Journal of Intelligent Manufacturing. 31(5):1175–1189.

[5] Lambrecht, K. (2024). PhD dissertation.