Advancing Glucose Content Prediction: A Multivariate Fuzzy Time Series Methodology
Zhi Liu, Zhiyi Zhang · 2024
Throughout the dual-enzyme synthesis of zinc gluconate, the glucose concentration within the reaction solution serves as the pivotal metric for tracking the reactions advancement. In the practical production scenario, the constraints of existing detection mechanisms necessitate the use of manual sampling and chemical titration as the predominant methods for glucose content determination. The sampling interval of this method can affect the precision of glucose content measurement, and the significant time delay it introduces prevents the realization of immediate control and adjustment in the production process. This paper proposed an innovative online prediction approach for glucose content, leveraging a multivariate fuzzy time series model. Initially, principal component analysis was employed to identify the primary factors influencing glucose content. Subsequently, the Hermite interpolation algorithm was employed to establish the optimal time interval and to fill in any missing data. Next, an enhanced clustering analysis algorithm was utilized to define and partition the domain. The fuzzy set for predicting glucose content was derived by constructing a mapping from multivariate to univariate fuzzy relationships. Ultimately, the glucose content was computed using the inverse fuzzy number formula. By implementing the method outlined in this article for predicting glucose content in production engineering and comparing its results with actual values and those from other prediction methods, we found that the accuracy of the predictions here was superior. Furthermore, this method reduces the need for manual labor and eliminates errors associated with manual measurement processes.