Bayesian Parameter Estimation for Scalable Bioprocess Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Bioprocessing systems face challenges in configuring and controlling complex bioprocesses due to numerous variables and interactions, leading to inefficiencies in yield, concentration, and purity, with existing methods being time-consuming, costly, and inaccurate, especially when scaling up.
Innovation Solution
A computer-implemented method using a probabilistic algorithm to determine a-posteriori parameter distributions for bioprocess models, combining numerical simulation, data storage, and physics-informed artificial intelligence to rapidly infer and predict bioprocess parameters, reducing computational time and improving scalability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical experiments are conducted to determine bioprocess parameters, then measurement accuracy is improved, but time consumption and cost increase significantly
Solution Approach 1:
The patent creates virtual copies of physical experiments through computational models. Instead of conducting actual physical experiments to determine bioprocess parameters, the system uses Bayesian inference algorithms that generate virtual experimental data and perform parameter estimation in silico, achieving comparable accuracy without the time and resource costs of physical experimentation.
Solution Approach 2:
The patent replaces the mechanical system of physical experimentation with an information-processing system. Bayesian computational methods substitute for physical measurement apparatus, using probabilistic algorithms to infer parameters from virtual or limited experimental data, thereby eliminating the need for extensive physical testing while maintaining estimation accuracy.
2Reliability
If comprehensive physical experiments are performed to account for all variables, then model accuracy is improved, but computational resources and cost increase
Solution Approach 1:
The patent applies partial action by using Bayesian inference to selectively estimate only the most critical parameters that have the greatest impact on model accuracy. Rather than exhaustively measuring all possible variables through comprehensive experiments, the system identifies and focuses computational resources on key parameters, achieving sufficient model reliability with reduced resource consumption.
Solution Approach 2:
The patent transforms the approach from fixed parameter measurement to dynamic parameter estimation. Bayesian methods allow parameters to be treated as probability distributions rather than fixed values, enabling the system to adapt parameter estimates based on available data and update beliefs as new information becomes available, improving model accuracy without requiring complete parameter knowledge.
3Ease of manufacture
If traditional parameter estimation methods are used, then simplicity is maintained, but scalability and adaptability to different bioprocess conditions deteriorate
Solution Approach 1:
The patent implements a universal Bayesian inference framework that can be applied across different bioprocess types and scales. The same probabilistic algorithmic structure handles diverse scenarios including batch and continuous processes, different reactor configurations, and various operational conditions, making the method highly adaptable while maintaining a consistent implementation approach that preserves simplicity.
4Measurement precision
If extensive experimental data collection is performed to improve parameter estimation, then prediction accuracy is improved, but productivity and process efficiency decrease
Solution Approach 1:
The patent performs preliminary Bayesian parameter estimation using available limited data before proceeding with full-scale bioprocess operations. By establishing initial parameter distributions and uncertainty bounds in advance, the system enables informed decision-making and process optimization without requiring extensive experimental data collection during actual production, thereby improving productivity while maintaining prediction accuracy.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
The application relates to a computer-implemented method for configuring and/or controlling a bioprocessing system that is configured to physically perform and/or simulate a bioprocess, a respective computer program product, a method for performing a bioprocess, a control device for controlling and/or configuring a bioprocessing system and a bioprocess. The method uses a probabilistic algorithm to determine a predictive posterior distribution for a target set of model parameter values and configure and/or control at least one state parameter of the bioprocessing system cand/or bioprocess based on the determined predictive posterior distribution. The probabilistic algorithm employs a combination of evaluation methods based on a common physical model, including a numerical simulation method, a search method for searching in a data storage storing a plurality precomputed solution-parameter sets of the physical model and an artificial intelligence based method using a physics informed artificial intelligence model (such as physics informed neural network model), which is trained based on the solutions of the physical model generated by the simulation method and the physical model.