Bioreactor Operation Modeling for Optimal Fermentation Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current microbial fermentation processes lack the ability to consistently determine optimal operation strategies, making it difficult to predict changes in production amounts and adapt to market demands, especially when using genetically improved strains.
Innovation Solution
A bioreactor model is constructed using actual process data, allowing for the determination of optimal operation conditions by receiving and processing experimental data sets, estimating parameter sets, creating candidate models, validating simulation performance, and selecting the best model to minimize objective functions such as yield or productivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional mathematical modeling of strain behavior is used, then the model accurately simulates the behavior of a specific strain, but the predictive power for genetically improved strains is low
Solution Approach 1:
The patent transitions from strain-specific parameter models to a data-driven approach where model parameters are dynamically adjusted based on operational data. The system uses historical process data to train machine learning models that can predict behavior across different strain types, effectively changing the modeling parameters from fixed strain characteristics to adaptive operational patterns.
Solution Approach 2:
The patent creates a virtual replica of the bioreactor system through digital twining, where a computational model mirrors the physical system's behavior. This virtual model is trained on historical data and can predict outcomes for genetically improved strains without requiring physical experiments, thus copying the system's behavior in silico rather than relying on strain-specific mathematical models.
2Ease of operation
If operation strategy is changed based on experience and past results, then the process can be operated, but the effect of change on production amount cannot be predicted
Solution Approach 1:
The patent implements a closed-loop system where operational data is continuously collected, analyzed, and fed back into the predictive model. This feedback mechanism allows the system to learn from past operations and improve its predictions, enabling operators to see the expected outcome of strategy changes before implementing them, thus eliminating the information loss about production effects.
Solution Approach 2:
The system performs preliminary simulations and predictions before actual operation changes are made. By using the trained model to forecast the outcomes of potential operation strategy changes, the system allows operators to evaluate multiple scenarios in advance and select the optimal approach, preventing the loss of predictive information that would otherwise require costly trial-and-error experimentation.
3Adaptability or versatility
If market-driven production target changes are implemented, then adaptability to market demand improves, but the application to actual process becomes slow
Solution Approach 1:
The patent enables preliminary evaluation of market-driven production targets by simulating their impact on the actual process before implementation. The predictive model可以快速评估不同生产目标下的工艺参数调整方案，使企业能够在市场变化发生前就准备好应对策略，从而缩短从市场信号到实际生产调整的响应时间，解决了适应性与实施时间之间的矛盾。
Data Source
AI summary
Disclosed are a method and a device for determining an operating condition of a bioreactor. Each of the method and the device receives N experimental data sets, models the bioreactor using the N experimental data sets to construct a bioreactor model, and determines an optimal operation condition based on an operation scheme of the bioreactor, using the constructed bioreactor model.