Role of Artificial Neural Network in Process Control and Monitoring in Bioprocessing

Bioprocessing plays an essential role in the large-scale production of biological products, where accurate monitoring and control are key for both yield and quality. This work aims to develop and assess a predictive framework based on Artificial Neural Networks (ANN) for estimating product yield in bioprocess operations. A multi-phase approach was implemented, beginning with data collection from online sensors and laboratory analyses, followed by preprocessing steps that included normalization, outlier removal, noise filtering, and feature engineering, utilizing dimensionality reduction through Principal Component Analysis. A hybrid ANN model was created, integrating Feed-Forward Neural Networks (FNN) for steady-state predictions, Long Short-Term Memory (LSTM) networks for learning temporal sequences, and Convolutional Neural Networks (CNN) for interpreting spectroscopic data.The model, trained using supervised learning and cross-validation, achieved strong predictive performance with a Mean Squared Error (MSE) of 1.0139 and a coefficient of determination (R²) of 0.9756, capturing 97.6% of yield variance. Predicted versus actual values showed high consistency, confirming robustness for real-time monitoring. Minor overfitting was observed at extreme values, highlighting the need for dataset expansion and regularization. Overall, the results demonstrate that ANN-based modeling effectively captures nonlinear dynamics in bioprocessing, supporting proactive optimization, disturbance detection, and integration into industrial-scale monitoring systems.