Holistic process control framework for smart biomanufacturing: A deep learning driven approach

• Published in: Computers & chemical engineering 2024-08, Vol.187, p.108742, Article 108742
• Main Authors: Nikita, Saxena, Banerjee, Shantanu, Jesubalan, Naveen G., Kulkarni, Amey, Gupta, Keshari, Rathore, Anurag S.
• Format: Article
• Language: English
• Subjects: Acoustic wave separator; Cyber-physical system; Deep neural network; Root cause analysis; Smart controller
• ISSN: 0098-1354; 1873-4375
• DOI: 10.1016/j.compchemeng.2024.108742

•Digitalization, software hardware integration and automation are enabling industry 4.0.•Development of a CPS-DT based holistic process control framework to automate process monitoring and control in manufacturing systems.•Deep learning model employed for accurate prediction of process parameters by identifying process deviation or anomaly in real-time.•Automated process monitoring and control implemented in 10 s, removes 85% of deviation due to induced disturbance.•Model is deployed as a web interface application using FastAPI for facilitating model access and data transfer both automatically (real-time) and manually. This article introduces a cyber-physical system comprising four interactive modules for enabling smart manufacturing in biotherapeutic production. These modules deal with process control, process prognosis, process diagnosis, and act as an external communication platform. An acoustic wave separator, a novel CHO cell clarification device, has been used as a case study with the objective of optimizing and regulating cell clarification process. A deep neural network (DNN)-based model has been proposed for implementing rapid control of bioprocess operations in case of detected anomalies or deviations in real-time. The use of the proposed holistic controller resulted in a 5% mean deviation from the setpoint in the first chamber, with more than 30% mean deviation resulting in the absence of the controller. The intricate layers of the automation framework facilitate real-time monitoring of the critical process parameters at 10-second intervals and ensure robust control action for deviations/anomalies, thus enabling smart manufacturing for biopharmaceutical production.