Construction and research of bioreactor simulation scale-up model

Abstract

Objective: Building a perfect bioreactor scale-up model provides convenience for enterprise production, and a guidance for reactor manufacturers. Methods: Through the test of 1.5, 5, 20, 200 L scale reactor, the correlation and parameters of the volume dissolved oxygen coefficient kLa and mixing time tm were optimized and explored, the simulation amplification model of bioreactor was established, and the Scaleuper amplification system was built by MATLAB. Results: Compared with the test results, the simulation deviation of kLa was within ±15%. By introducing the definition of critical speed into the correlation of tm, the overall simulation deviation of tm was reduced by 17.4% compared with the original formula. Within ±10%, the simulation deviation of kLa was reduced by embedding BP neural network. The minimum simulation deviation of BP model can reach 0.1% for the training data range. Conclusion: The scale-up model of the reactor based on the empirical correlation has better simulation results after the relevant parameters are measured. It is feasible to use neural network to simulate and amplify the reactor, but it still needs more data training support.

Keywords

• mixing time
• volumetric dissolved oxygen coefficient
• bioreactor scale-up
• neural network
• scale-up model