A method for rapid development of microplate fermentation process based on CFD simulation and response surface methodology

Abstract

The application discloses a kind of based on CFD simulation and response surface method's microplate fermentation process rapid development method, comprising: S1, establish microplate CFD model, obtain the quantitative relationship of fluid mechanics parameters and operating condition;S2, with fluid mechanics parameters as constraint, construct response surface model and carry out multi-objective optimization, obtain optimal process parameter;S3, adopt 3D printing to manufacture microplate prototype;S4, carry out high-throughput fermentation verification, correct model;S5, scale-up verification.The application is coupled with CFD simulation and response surface method, realizes accurate prediction and control to fluid mechanics condition while maintaining high-throughput advantage, solves the problem of long cycle, low optimization efficiency, scale-up difficulty in traditional process development.The application can shorten process development cycle by more than 50%, product concentration is higher than traditional method by 10-20%, scale-up success rate is increased to more than 80%, and is suitable for process rapid development of a variety of microbial fermentation system.

Description

Technical Field

[0001] This invention belongs to the field of microbial fermentation process development technology, specifically relating to a rapid development method for microplate fermentation processes based on CFD simulation and response surface methodology. Background Technology

[0002] Fermentation process development is a key link in the biomanufacturing industry, encompassing multiple tasks such as strain screening, culture medium optimization, and culture condition control. It is a large-scale, complex, and extremely time-consuming process.

[0003] Traditional fermentation process development methods are mainly based on shake flasks or small-scale fermenters. However, traditional methods have significant bottlenecks: shake flasks have low automation, cannot dynamically monitor the fermentation process, have poor reproducibility, and are prone to scale-up failures; small-scale fermenters have low throughput and also face the problems of poor reproducibility and high cost. Based on a week's workload, an experimenter can only conduct 10-20 shake flask cultures or 4-8 small-scale fermenter cultures, far from meeting the demands of modern biomanufacturing for high-throughput process development.

[0004] Microplates (MTPs) have gradually become the preferred container for microbial screening and process development as a high-throughput culture system, with common forms including 24-well, 48-well, and 96-well plates. However, microplate systems face the following core technical challenges in fermentation process development:

[0005] 1. Fluid dynamics conditions cannot be accurately predicted: Due to the small volume (0.2-2mL), complex geometry, and special operation mode (orbital oscillation) inside the microplate, key engineering parameters such as fluid flow, gas-liquid mass transfer, and shear distribution inside are difficult to measure directly through experiments, resulting in the optimization of process conditions relying on a large number of trial and error experiments.

[0006] 2. The contradiction between high throughput and controllability of conditions: High throughput necessarily means small volume, while small volume means that the fluid dynamics conditions (such as volumetric gas-liquid mass transfer coefficient kLa, energy dissipation rate, mixing time, etc.) are extremely sensitive to small changes in operating conditions and are difficult to control precisely.

[0007] 3. Low process development efficiency: Even with the use of microplate systems, optimization directions still need to be explored through a large number of single-factor experiments or orthogonal experiments, which cannot fully leverage the parallel advantages of high-throughput platforms.

[0008] 4. Difficulty in scale-up: There are significant differences in hydrodynamic conditions between microplate culture conditions and shake flask and fermenter conditions, making it difficult for the optimized conditions selected by microplate screening to directly guide subsequent scale-up.

[0009] In existing technologies, the application of CFD simulation in microplate systems mainly focuses on flow field analysis and geometric optimization. For example, He et al. (2023) used CFD to develop a hexagonal six-baffle microplate, optimizing the oscillation frequency and liquid volume. However, this method is only applicable to specific microplate geometries and has not formed a universal process development methodology. Response surface methodology is widely used in fermentation optimization, but most of it is based directly on statistical analysis of experimental data and is not combined with fluid dynamics prediction models.

[0010] The present invention aims to solve the following technical problem: how to achieve accurate prediction and directional optimization of the internal hydrodynamic conditions of microplates while maintaining the advantage of high throughput, so as to establish a systematic, efficient and scalable method for rapid development of fermentation processes. Summary of the Invention

[0011] To address the shortcomings of existing technologies, this invention provides a rapid development method for microplate fermentation processes based on CFD simulation and response surface methodology. By establishing a CFD hydrodynamic parameter prediction model, combined with multi-objective optimization using response surface methodology and rapid prototyping via 3D printing, the method enables rapid development and precise scale-up of fermentation process conditions, shortening the process development cycle by more than 50%.

[0012] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0013] This invention provides a rapid development method for microplate fermentation processes based on CFD simulation and response surface methodology, comprising the following steps:

[0014] S1: Establish a three-dimensional CFD model of the target microporous plate, perform mesh generation and mesh independence verification, set boundary conditions and solution model, obtain fluid dynamic parameters under different operating conditions through CFD simulation calculation, and establish a quantitative mapping relationship between operating conditions and fluid dynamic parameters.

[0015] S2: Using fluid dynamics parameters as constraints and fermentation performance indicators as response variables, a response surface model is constructed and multi-objective optimization is performed to obtain the Pareto optimal solution set. The optimal process parameters are then selected as the optimization result.

[0016] S3: Based on the optimization results, a microplate prototype was manufactured using 3D printing technology for verification in fermentation experiments;

[0017] S4: Conduct high-throughput fermentation verification experiments under optimized conditions, measure fermentation performance indicators, compare experimental results with model predictions, and if the deviation exceeds the preset threshold, feed the experimental data back to steps S1 and S2 to correct the model parameters and perform iterative optimization.

[0018] S5: Establish the scale-up criteria between the microplate and the scale-up system, and convert the optimal process parameters of the microplate obtained in step S2 into the operating parameters of the scale-up system for scale-up verification.

[0019] Preferably, in step S1, the CFD simulation uses a VOF model to track the gas-liquid two-phase interface and combines an RNGk-ε turbulence model to simulate the turbulent flow of the liquid phase in the micropores.

[0020] The boundary conditions are set as follows: the orifice wall is set as a non-slip wall boundary condition; the gas-liquid interface adopts a pressure outlet boundary condition; and the orbital oscillation trajectory is defined by a user-defined function (UDF).

[0021] The fluid dynamic parameters include the volumetric gas-liquid mass transfer coefficient kLa, energy dissipation rate ε, and maximum shear stress τ. max Mean shear stress τ avg Mixing time t mix At least one of the following: gas holdup α.

[0022] Preferably, in step S2, the response surface model formula is as follows:

[0023] Y=β0+Σβ i .x i +Σβ ii .x i ²+ΣΣβ ij .x i .x j

[0024] Where Y is the response variable, representing the fermentation performance index to be optimized, which includes the target product concentration C. P Biomass X, Product yield to substrate Y P At least one of / S;

[0025] x i x j Let V be the independent variable, representing the operating conditions affecting fermentation performance. These operating conditions include the liquid volume V. L At least one of the following: oscillation frequency f, oscillation diameter d0, and temperature T; i and j are used to distinguish different independent variables;

[0026] β0 is the constant number, representing the baseline value of the response variable when all independent variables are at zero levels;

[0027] β i The coefficient of the linear term represents the independent variable x. i The linear main effects on Y;

[0028] β ii The coefficient of the quadratic term represents the independent variable x. iThe square of the bending effect of Y;

[0029] β ij The coefficients of the interaction term represent the interaction coefficients between two distinct independent variables x. i With x j The impact of the interaction between them on Y.

[0030] Preferably, in step S2, the multi-objective optimization employs a multi-objective genetic algorithm (MOGA) or an artificial neural network-genetic algorithm (ANN-GA) to globally optimize the response surface model. The objective function is set to maximize the target product concentration, maximize the product yield, and minimize the by-product concentration. Constraints include kLa ≥ minimum mass transfer requirement and maximum shear stress τ. max ≤Strain tolerance threshold. When multiple optimization objectives exist, plot the Pareto front and select the optimal process parameters from the Pareto solution set using the ideal point method (TOPSIS) or the satisfaction function method.

[0031] Preferably, the 3D printing technology in step S3 includes SLA, DLP, or FDM technology, and the printing material is biocompatible photosensitive resin or PLA. Post-processing is performed after printing, including removal of the support structure, surface polishing, and sterilization (ultraviolet irradiation or ethylene oxide sterilization).

[0032] Preferably, in step S4, the high-throughput fermentation verification experiment is conducted online using a high-throughput microbioreactor, and the monitoring parameters include at least one of biomass, pH, dissolved oxygen, and fluorescence signal.

[0033] Preferably, in step S5, the amplification criterion is the equal kLa criterion, the equal P / V criterion, or the equal τ criterion. max At least one of the following criteria: The equal kLa criterion maintains the volumetric gas-liquid mass transfer coefficient kLa equal; the equal P / V criterion maintains the power input per unit volume equal; the equal τ criterion... max The criterion is to keep the maximum shear stress equal.

[0034] Preferably, the method is applicable to at least one of the following microbial systems: bacteria, yeast, filamentous fungi, and actinomycetes.

[0035] Preferably, the microplate is a 24-hole, 48-hole, or 96-hole plate, and the hole shape is circular, hexagonal, or square.

[0036] Preferably, the microporous plate is further provided with a baffle structure, which is at least one of a spiral baffle, a straight baffle, an arc baffle, and a grid baffle.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. Significantly shortened process development cycle: By using CFD simulation to predict fluid dynamic conditions in advance, the number of experimental trials is greatly reduced. The traditional shake-flask method typically takes 6-8 weeks from strain screening to process condition determination, while the method of this invention can shorten this to 3-4 weeks, a reduction of approximately 50%.

[0039] 2. High-throughput parallel optimization: More than 30 sets of experimental data can be obtained in a single experiment, increasing the throughput by 5-10 times compared to the traditional shake-flask method. The microplate system, combined with DOE experimental design, allows for the systematic examination of the interaction effects of multiple factors in a single experiment.

[0040] 3. Precise and controllable fluid dynamics conditions: The quantitative distribution of key engineering parameters such as kLa, shear stress, and energy dissipation rate is obtained through CFD simulation, enabling process optimization to shift from experience-based trial and error to mechanism-guided optimization. The deviation between the predicted and measured values ​​of kLa can be controlled within 10%.

[0041] 4. High scale-up success rate: A parameter transfer method based on the hydrodynamic similarity criterion was established, which solved the parameter mismatch problem between the traditional microplate system and the shake flask / fermenter, and the scale-up success rate can be increased to more than 80%.

[0042] 5. Wide applicability: This method is applicable to a variety of microbial systems (bacteria, yeast, filamentous fungi, actinomycetes) and various microplate formats (24 / 48 / 96-well plates), and has broad applicability.

[0043] 6. Automatable integration: This method can be seamlessly integrated with automated workstations, high-throughput microbioreactors and other equipment to achieve fully automated operation. Attached Figure Description

[0044] Figure 1 This is a flowchart of the rapid development method for microplate fermentation process based on CFD simulation and response surface methodology of the present invention.

[0045] Figure 2 a is a schematic diagram of the structure of the hexagonal microporous plate (96 holes) with six baffles in this invention;

[0046] Figure 2 b is a schematic diagram of the microporous structure of the hexagonal microporous plate with six baffles in this invention;

[0047] Figure 2 c is a top view of the micropores in the hexagonal microporous plate with six baffles in this invention;

[0048] Figure 3 a is the integral cloud diagram of the liquid phase in the micropore obtained by CFD simulation in this invention;

[0049] Figure 3 b is the velocity vector and streamline diagram inside the micropore obtained by CFD simulation in this invention;

[0050] Figure 4 This is a comparison chart of the CFD predicted values ​​and experimental measured values ​​of kLa under different operating conditions in this invention;

[0051] Figure 5 This is the three-dimensional surface plot of the response surface in this invention;

[0052] Figure 6 This is the Pareto front plot for multi-objective optimization in this invention;

[0053] Figure 7 This is a prototype diagram of the 3D printed microporous plate in this invention;

[0054] Figure 8 This is a diagram showing the correlation between the microporous plate and the scaled-up parameters of the shake flask / fermenter in this invention.

[0055] Figure 9 This is a schematic diagram of the automated liquid handling system of the present invention. Detailed Implementation

[0056] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0057] like Figure 1 As shown, this embodiment of the invention provides a rapid development method for microplate fermentation processes based on CFD simulation and response surface methodology, comprising the following steps:

[0058] S1: Establish a three-dimensional CFD model of the target microporous plate, perform mesh generation and mesh independence verification, set boundary conditions and solution model, obtain fluid dynamic parameters under different operating conditions through CFD simulation calculation, and establish a quantitative mapping relationship between operating conditions and fluid dynamic parameters.

[0059] In this embodiment, the three-dimensional CFD model of the target microplate can be established using one of the following two methods:

[0060] ① A three-dimensional model was created using the actual geometric dimensions of a standard commercial microplate (24 / 48 / 96 wells);

[0061] ② Customized microplate geometry design for specific fermentation systems, including pore shape (circular / hexagonal / square), pore spacing, pore depth / pore diameter ratio, and baffle structure (number, shape, and arrangement of baffles). Hexagonal plates with six baffles exhibited optimal mass transfer performance, achieving a 0.61s mass transfer efficiency. -1The kLa value is 4-5 times that of traditional 96-well plates. The baffles can be designed in one of the following forms: spiral baffles, straight baffles, curved baffles, grid baffles, or combinations thereof. Culture flasks using grid baffles achieved a 30.4% higher final cell concentration compared to traditional baffle-free culture flasks.

[0062] In this embodiment, the mesh generation adopts a strategy combining structured and unstructured meshes, with finer meshing near the wall. The number of meshes is determined based on mesh independence verification. Figure 2 As shown, the microplate (96 wells) contains a hexagonal structure with six baffles, a spiral baffle structure, a gas-liquid layer grid, and a boundary layer grid, achieving a speed of 0.62 s. -1 oxygen transfer efficiency and 2364 W / m 3 The mixed level.

[0063] In this embodiment, the CFD simulation uses a VOF model to track the gas-liquid interface and combines it with an RNG k-ε turbulence model to simulate the turbulent flow of the liquid phase within the micropores. The gas-liquid two-phase distribution cloud map and velocity vector map obtained from the CFD simulation are shown below. Figure 3 .

[0064] In this embodiment, the boundary conditions are set as follows: the orifice wall is set as a non-slip wall boundary condition; the gas-liquid interface adopts a pressure outlet boundary condition; and the orbital oscillation trajectory is defined by a user-defined function (UDF).

[0065] In this embodiment, a transient solution is performed using a CFD solver such as Fluent, with the time step set to 1 / 200-1 / 500 of the oscillation period. After reaching a quasi-steady state, fluid dynamic parameters are extracted. These parameters include the volumetric gas-liquid mass transfer coefficient kLa, the energy dissipation rate ε, and the maximum shear stress τ. max Mean shear stress τ avg Mixing time t mix α is the gas content.

[0066] Volumetric gas-liquid mass transfer coefficient kLa: The volumetric gas-liquid mass transfer coefficient kLa was calculated by using the Higbie permeation model and obtaining the contact time tc through CFD simulation.

[0067] Energy dissipation rate ε: The average energy dissipation rate is obtained by integrating the local energy dissipation rate, and then the power input per unit volume P / V is calculated.

[0068] Maximum shear stress τ max and mean shear stress τ avg Extracted from the wall shear stress distribution;

[0069] Mixing time t mix The assessment was based on the time required for the tracer concentration to reach 95% uniformity.

[0070] Gas holdup α: the ratio of the volume fractions of the gas and liquid phases.

[0071] In this embodiment, the following experiments are used to verify the accuracy of the CFD model in predicting the volumetric gas-liquid mass transfer coefficient kLa within the microporous plate, demonstrating that the simulation results can serve as a reliable input for subsequent response surface optimization.

[0072] The volumetric gas-liquid mass transfer coefficient kLa within a microplate was measured using the sulfite oxidation method and the dynamic method with an active dissolved oxygen electrode.

[0073] Measurements were performed in 48-well microplates at different oscillation frequencies (1500-300 rpm) and fixed liquid volumes (e.g., 1.2 mL). At least three parallel wells were set up for each condition, and the average value and standard deviation were calculated as the error bar.

[0074] Based on the Fluent simulation in step S1, the Higbie permeation model was used, and the gas-liquid contact time t was calculated by CFD. c kLa = 2 was calculated using the oxygen diffusion coefficient Do2. This allows us to obtain the theoretical prediction values ​​under the corresponding conditions.

[0075] like Figure 4 As shown, solid squares represent CFD predicted values, and hollow circles represent experimental measured values. Error bars are also provided. The deviation between the predicted and measured values ​​of kLa can be controlled within 10%. The two curves in the figure basically overlap, indicating that CFD has high prediction accuracy and can provide accurate mass transfer parameters for subsequent process optimization.

[0076] S2: Using fluid dynamics parameters as constraints and fermentation performance indicators as response variables, a response surface model is constructed and multi-objective optimization is performed to obtain the Pareto optimal solution set. The optimal process parameters are then selected as the optimization result.

[0077] In this embodiment, the response surface model employs a Box-Behnken design or a central composite design, with operating conditions as independent variables and 3-5 levels for each factor. The response surface model is used to fit experimental data in response surface methodology, thereby describing the quantitative relationship between operating conditions and fermentation results, and is used for subsequent multi-objective optimization. Figure 5 As shown, the liquid volume V is displayed on the three-dimensional surface plot of the response surface. L The interaction between the oscillation frequency f and the target product concentration C P The impact.

[0078] The formula for the response surface model is as follows:

[0079] Y=β0+Σβ i .x i +Σβii .x i ²+ΣΣβ ij .x i .x j

[0080] Where Y is the response variable, representing the fermentation performance index to be optimized, which includes the target product concentration C. P Biomass X, Product yield to substrate Y P / S;

[0081] x i x j Let V be the independent variable, representing the operating conditions affecting fermentation performance. These operating conditions include the liquid volume V. L The oscillation frequency f, oscillation diameter d0, and temperature T; i and j are used to distinguish different independent variables;

[0082] β0 is the constant number, representing the baseline value of the response variable when all independent variables are at zero levels;

[0083] β i The coefficient of the linear term represents the independent variable x. i The linear main effects on Y;

[0084] β ii The coefficient of the quadratic term represents the independent variable x. i The square of the bending effect of Y;

[0085] β ij The coefficients of the interaction term represent the interaction coefficients between two distinct independent variables x. i With x j The impact of the interaction between them on Y.

[0086] In this embodiment, multi-objective optimization employs a multi-objective genetic algorithm (MOGA) or an artificial neural network-genetic algorithm (ANN-GA) to globally optimize the response surface model. The objective function is set to maximize the target product concentration, maximize the product yield, and minimize the by-product concentration. Constraints include the volumetric gas-liquid mass transfer coefficient kLa ≥ the minimum mass transfer requirement and the maximum shear stress τ. max ≤ Strain tolerance threshold.

[0087] When there are multiple optimization objectives, plot the Pareto front ( Figure 6 The optimal process parameters are selected from the Pareto solution set using the ideal point method (TOPSIS) or the satisfaction function method.

[0088] S3: Based on the optimization results, a microporous plate prototype was manufactured using 3D printing technology. Figure 7The culture flasks were used for fermentation experiments for verification. Different types of baffles were fabricated using 3D printing technology, and the effects of grid-type, arc-shaped, and spiral-shaped baffles on mass transfer performance were evaluated.

[0089] The 3D printing technology includes SLA, DLP, or FDM, and the printing material is biocompatible photosensitive resin or PLA. Post-printing processing is performed after printing, including removal of the support structure, surface polishing, and sterilization (ultraviolet irradiation or ethylene oxide sterilization).

[0090] Stereoscopic laser curing (SLA): Using photosensitive resin materials, the precision can reach 25-100μm, with good surface smoothness, and it is suitable for making transparent microporous plates;

[0091] Digital light processing (DLP): Precision up to 10-50μm, fast forming speed;

[0092] Fused deposition modeling (FDM): Lower cost, suitable for materials with high biocompatibility requirements (such as PLA).

[0093] S4: Conduct high-throughput fermentation verification experiments under optimized conditions, measure fermentation performance indicators, compare experimental results with model predictions, and if the deviation exceeds a preset threshold, feed the experimental data back to steps S1 and S2 to correct model parameters and perform iterative optimization.

[0094] In this embodiment, the high-throughput fermentation validation experiment steps are as follows: The target strain is inoculated into the optimized culture medium and cultured in microplates according to the optimized conditions, including liquid volume and shaking frequency. Online or offline monitoring is performed using a microplate reader or a high-throughput microbioreactor (such as BioLectorXT) to determine the biomass (OD). 600 ), target product concentration, substrate consumption, pH, dissolved oxygen, and fluorescence signal.

[0095] S5: Establish the scale-up criteria between the microplate and the scale-up system, and convert the optimal process parameters of the microplate obtained in step S2 into the operating parameters of the scale-up system for scale-up verification.

[0096] In this embodiment, the scale-up system is a shake flask / fermenter, and the scale-up criteria are the equal kLa criterion, the equal P / V criterion, or the equal τ criterion. max At least one of the following criteria: The equal kLa criterion maintains the volumetric gas-liquid mass transfer coefficient kLa equal; the equal P / V criterion maintains the power input per unit volume equal; the equal τ criterion... max The criterion is to keep the maximum shear stress equal.

[0097] like Figure 8 As shown, taking the equal kLa criterion as an example, the oscillation frequency f of the microporous plate is established. MTP With the oscillation frequency f of the shaker SFThe formula establishes a quantitative relationship between the two, using the similarity criterion of equal gas-liquid mass transfer capacity as the core. It quantitatively transforms the hydrodynamic conditions optimized from microplates into shake-flask operating conditions, thereby ensuring that the cell culture environment after scale-up is consistent with that of the microplate, thus improving the scale-up success rate. The specific formula is as follows:

[0098] f SF =f MTP ×(kLa SF (f) / kLa MTP (f)) -1

[0099] Among them, f MTP The microplate oscillation frequency, i.e., the rotation speed of the shaker during microplate culture, is the optimal operating condition obtained through step S2 optimization.

[0100] f SF The shaking oscillation frequency, which is the rotational speed that the shaker should be set when scaled up to the shaking flask system, is the target operating parameter obtained by conversion through similarity criteria.

[0101] kLa MTP (f) is the volumetric gas-liquid mass transfer coefficient function of the microporous plate system, which represents the quantitative relationship between kLa and the oscillation frequency f in the microporous plate. This relationship is obtained by fitting a large number of data points from the CFD simulation in step S1.

[0102] kLa SF (f) is the volumetric gas-liquid mass transfer coefficient function of the shake flask system, which represents the quantitative relationship between kLa and the oscillation frequency f in the shake flask. It is obtained in advance through the classical shake flask mass transfer correlation or shake flask CFD simulation.

[0103] (kLa SF (f) / kLa MTP (f)) -1 The reciprocal of the ratio of the mass transfer coefficients of the two systems is essentially a frequency conversion factor, and its derivation logic is as follows:

[0104] 1. Determine the mass transfer coefficient kLa*= kLa under optimal conditions for the microplate. MTP (f MTP );

[0105] 2. According to the equal kLa criterion, the same mass transfer coefficient, i.e., kLa, must be achieved in the shake-flask system. SF (f SF =kLa*;

[0106] 3. Obtain f through the inverse function SF = If the functional relationship is a simple power law, then the proportional conversion relationship shown in the above formula can be derived.

[0107] Scale-up fermentation experiments were conducted in shake flasks (250mL / 500mL) and / or small fermenters (1-10L), and the deviations of fermentation performance indicators (product concentration, yield, production intensity, etc.) from the microplate optimization results were compared. If the deviations were within acceptable limits, the final process scheme was output; otherwise, the scale-up criteria were adjusted and iterative optimization was performed.

[0108] like Figure 9 As shown, the above technical solution is integrated with an automated liquid handling system to achieve full-process automation: robots and automated workstations are used to complete operations such as culture medium dispensing, inoculation, and sampling. Combined with the online monitoring function of the microplate culture system, process data is fed back in real time, and optimization strategies are automatically adjusted. The integrated automated system can operate unattended 24 hours a day, significantly improving R&D efficiency.

[0109] Example 1: Rapid development of a microplate fermentation process for the production of succinic acid from recombinant Escherichia coli (flow process see...) Figure 1 ).

[0110] Step S1: CFD modeling and prediction of fluid dynamics parameters.

[0111] A three-dimensional geometric model was established using a standard 48-well circular microplate as the object. The diameter of each well is 10.2 mm, the depth is 17.5 mm, and the working volume is 1.0-2.0 mL. CFD simulation was performed using ANSYS Fluent 2024, with the following specific settings:

[0112] Mesh generation: Unstructured tetrahedral meshes were generated using ICEMCFD, with a 5-layer prismatic mesh boundary layer, resulting in a total mesh count of approximately 850,000. Mesh independence was verified (with mesh counts set to 450,000, 850,000, and 1,500,000 respectively, and kLa prediction deviation <2%), to determine the 850,000 mesh scheme.

[0113] Multiphase flow model: VOF model, with air as the main phase and fermentation broth (density 1050 kg / m³) as the secondary phase. 3 (Viscosity 1.2 mPa·s).

[0114] Turbulence model: RNGk-ε model, with standard wall functions used near the wall.

[0115] Boundary conditions: Define orbital oscillation motion via UDF, with an oscillation diameter of 25 mm, an oscillation frequency of 150-300 rpm (5 levels), and liquid volume of 1.0, 1.5, and 2.0 mL (3 levels).

[0116] Simulation calculations yielded kLa, energy dissipation rate ε, and maximum shear stress τ under different conditions. maxThe results show that kLa increases exponentially with increasing oscillation frequency (kLa = 0.018 s at 150 rpm). -1 At 300 rpm, kLa = 0.052 s -1 The shear stress decreased linearly with increasing liquid volume. The maximum shear stress reached 3.2 Pa at 300 rpm and 1.0 mL, which is lower than the shear tolerance threshold of E. coli (approximately 5 Pa).

[0117] Step S2: Response surface methodology for multi-objective optimization.

[0118] The oscillation frequency (X1, 150-300 rpm), liquid volume (X2, 1.0-2.0 mL), and initial glucose concentration (X3, 10-30 g / L) were selected as independent variables. A Box-Behnken design was used, and a total of 17 experiments were conducted.

[0119] Fermentation experiments were conducted in a BioLectorXT high-throughput microbioreactor. The strain was recombinant Escherichia coli BL21(DE3) / pTrc99A-pncB, with an inoculum size of 5% (v / v), a culture temperature of 37℃, and the pH maintained at 6.8-7.0 via a microvalve control system for a culture time of 24 h.

[0120] A response surface model was established using succinic acid concentration (Y1, g / L) and succinic acid yield (Y2, g / g glucose) as response variables:

[0121] Y1=12.35+1.87X1-0.95X2+2.34X3-0.42X1X2+0.61X1X3-0.38X2X3-1.21X1 2 -0.78X2 2 -1.56X3 2 (R) 2 =0.967, P<0.001)

[0122] Y2=0.42+0.05X1-0.03X2+0.08X3-0.02X1X2+0.02X1X3-0.01X2X3-0.05X1 2 -0.02X2 2 -0.06X3 2 (R) 2 =0.943, P<0.001)

[0123] The multi-objective genetic algorithm (MOGA) was used for optimization, with the constraint kLa ≥ 0.03s. -1 (Minimum oxygen demand determined by CFD simulation), maximum shear stress τ max≤4Pa. Optimal conditions were obtained: oscillation frequency 285 rpm, liquid volume 1.2 mL, initial glucose concentration 28 g / L. Predicted succinic acid concentration 14.8 g / L, yield 0.49 g / g.

[0124] Step S3: 3D printing manufacturing.

[0125] Based on the optimization results, a novel 48-well microplate with six baffles was designed and manufactured using a DLP 3D printer (50μm precision) with a biocompatible photosensitive resin as the material. After printing, it was sterilized by ultraviolet irradiation for 30 minutes.

[0126] Step S4: Validation experiment and model correction.

[0127] Three independent and repeated validation experiments were conducted under optimized conditions. The measured succinic acid concentration was 14.5 ± 0.3 g / L, and the yield was 0.48 ± 0.01 g / g. The deviations between the predicted and measured values ​​were 2.0% and 2.1%, respectively, both less than the preset threshold (10%), indicating that the model validation was successful.

[0128] Step S5: Magnify and verify.

[0129] Using the iso-kLa value as the scale-up criterion, the optimized conditions for the microplate were transformed into operating conditions for a 500mL shake flask: for the microplate, 285rpm and a 25mm oscillation diameter, kLa = 0.048s⁻¹, which translates to an oscillation frequency of 220rpm for a 100mL shake flask. Scale-up verification was performed in a 500mL shake flask. The succinic acid concentration was 14.1g / L, and the yield was 0.47g / g. The deviation from the microplate results was within 5%, indicating successful scale-up verification.

[0130] Example 2: Rapid development of a microplate fermentation process for ethanol production from brewer's yeast.

[0131] Step S1: CFD modeling and prediction of fluid dynamics parameters.

[0132] A standard 96-well circular microplate (approximately 360 μL per well) was used, with a liquid volume ranging from 100 to 200 μL. The oscillation frequency was set to 400-1000 rpm (existing research indicates that sufficient mixing can only be achieved with 96-well plates at higher oscillation frequencies). The CFD simulation settings were the same as in Example 1.

[0133] Step S2: Response surface methodology for multi-objective optimization.

[0134] A central composite design (CCD) was used to investigate the effects of oscillation frequency (X1, 500-900 rpm), liquid volume (X2, 100-200 μL), and initial glucose concentration (X3, 50-150 g / L) on ethanol yield and efficiency. The strain used was *Saccharomyces cerevisiae* S288C, with an inoculum size of 2%, and the culture temperature was 30 °C for 48 h.

[0135] After establishing the response surface methodology, multi-objective optimization was performed using the ANN-GA method. Previous studies have shown that ANN-GA outperforms the traditional RSM in handling nonlinear and complex interactive effects during fermentation. The optimal conditions were obtained: oscillation frequency 820 rpm, liquid volume 130 μL, and initial glucose concentration 125 g / L. The predicted ethanol concentration was 52.3 g / L, and the yield was 0.45 g / g.

[0136] Steps S3-S5.

[0137] 3D printing was used to manufacture a 96-well microplate with an arc-shaped baffle. The ethanol concentration in the verification experiment was 51.8 ± 0.7 g / L (predictive deviation 1.0%). Shake-flask scale-up verification was performed using equal P / V as the scale-up criterion, with an ethanol concentration of 50.2 g / L and a yield of 0.44 g / g.

[0138] Example 3: Rapid development of a microplate fermentation process for cellulase production by the filamentous fungus Trichoderma reesei.

[0139] Fermentation by filamentous fungi differs significantly from that of bacteria and yeast. Their mycelial morphology results in non-Newtonian fluid properties (pseudoplastic fluids), and the mycelium is more sensitive to shear stress. This invention provides adaptive improvements to the filamentous fungal system:

[0140] The CFD model uses a power law model to describe the non-Newtonian rheology of the fermentation broth: η = K·γ^(n-1), where the consistency coefficient K and the flow behavior index n are measured by a rheometer.

[0141] Enhanced simulation of shear stress distribution was achieved to obtain spatial distribution cloud maps of shear stress within the borehole.

[0142] In response surface optimization, the mean shear stress τ is added. avg As a constraint, ensure that the optimization conditions do not exceed the strain's shear tolerance threshold.

[0143] Key points for implementing steps S1-S5

[0144] A standard 24-well microplate was used, with a liquid volume of 2.0-3.0 mL. In the CFD simulation, the fermentation broth density was 1030 kg / m³, and the power-law model parameters were K = 0.15 Pa·s^n, n = 0.65. The oscillation frequency was set to 150-250 rpm.

[0145] The CCD experimental design consisted of 20 experiments. The strain used was *Trichoderma reesei* Rut-C30, and the spore inoculum size was 10. 5 The culture temperature was 28℃, and the culture time was 120 h. Filter paper enzyme activity (FPase, U / mL) and endoglucanase activity (CMCase, U / mL) were used as response variables.

[0146] The optimal conditions obtained through ANN-GA optimization were: oscillation frequency 210 rpm and liquid volume 2.4 mL. Under these conditions, the maximum shear stress τ was... max =2.8 Pa, lower than the shear tolerance threshold of Trichoderma reesei (approximately 4 Pa). The verification experiment showed filter paper enzyme activity of 2.35 ± 0.12 U / mL, a deviation of 2.5% from the predicted value of 2.41 U / mL. Shake flask amplification verified filter paper enzyme activity at 2.18 U / mL, with a deviation of 7.2%.

[0147] Comparative Example 1: Traditional shake-flask single-factor experimental method (compared with Example 1).

[0148] Methods: The traditional shake-flask method (500mL shake flasks, 100mL liquid volume) was used. The oscillation frequency (150, 180, 210, 240, 270, 300 rpm), liquid volume (50, 75, 100, 125, 150mL), and initial glucose concentration (10, 15, 20, 25, 30g / L) were optimized sequentially through single-factor experiments. Each condition was set up in triplicate, requiring a total of (6+5+5)×3=48 shake-flask experiments. Assuming a maximum of 20 shake-flasks per week, completing all single-factor optimizations would take 2-3 weeks, and single-factor experiments cannot examine the interaction effects between factors.

[0149] Results: The optimal conditions obtained from the single-factor experiment were an oscillation frequency of 270 rpm, a liquid volume of 100 mL, and an initial glucose concentration of 25 g / L. Under these conditions, the succinic acid concentration was 12.8 ± 0.6 g / L, and the yield was 0.41 ± 0.02 g / g.

[0150] Comparative analysis: The method of this invention obtains optimized conditions through only one response surface methodology experiment (17 groups of microplate parallel experiments), with an experimental cycle of 3-4 days, a succinic acid concentration of 14.5±0.3g / L (13.3% higher than the traditional method), and a yield of 0.48±0.01g / g (17.1% higher than the traditional method).

[0151] Comparative Example 2: Simple Response Surface Methodology (not coupled with CFD) (compared with Example 1).

[0152] Methods: Seventeen microplate experiments were designed directly using the response surface methodology, without CFD simulation. The experimental design was the same as in Example 1, but the optimization was based solely on experimental data, without using CFD-predicted fluid dynamic parameters (kLa, shear stress, etc.) as constraints.

[0153] Results: Response surface methodology optimization yielded an oscillation frequency of 290 rpm and a solution volume of 1.1 mL. Under these conditions, the succinic acid concentration was 14.0 ± 0.4 g / L, and the yield was 0.46 ± 0.02 g / g. However, in subsequent experiments, dissolved oxygen limitation was observed in some culture wells (measured kLa value 0.022 s). -1 This is 0.03s lower than the theoretical requirement. -1 This leads to a decrease in repeatability (between batch coefficient of variation CV = 8.5% vs. CV = 2.1% for this invention).

[0154] Comparative analysis: While response surface methodology (RSM) without coupled CFD can also achieve good optimization results, the lack of constraints and guidance from fluid dynamic parameters means that the optimization conditions are located in the critical region of kLa, resulting in poor process robustness. This invention establishes a quantitative relationship between operating conditions and kLa beforehand using CFD, ensuring that the optimization conditions are within a safe range, thus significantly improving the robustness and repeatability of the process.

[0155] Comparative Example 3: Simple CFD simulation method (not coupled with response surface methodology) (compared with Example 1).

[0156] Method: CFD simulation was used to predict the kLa value under different operating conditions. The operating conditions were selected with the maximum kLa as the single objective, and then a verification experiment was conducted.

[0157] Results: CFD simulations showed that kLa increased monotonically with increasing oscillation frequency and increased with decreasing liquid volume. Based on this, the conditions of highest oscillation frequency + minimum liquid volume (300 rpm, 1.0 mL) were selected. Under these conditions, the measured value of kLa was 0.056 s. -1 The verification experiment showed that the succinic acid concentration was 11.2 ± 0.5 g / L and the yield was 0.36 ± 0.02 g / g, which was lower than the 14.5 g / L and 0.48 g / g in Example 1.

[0158] Comparative analysis: Simple CFD simulation can only optimize fluid dynamic parameters and cannot comprehensively consider biological reaction kinetics (such as the effect of substrate concentration on product synthesis). In this comparative example, although kLa was the largest, the insufficient liquid volume resulted in insufficient total substrate, limiting product accumulation. This invention, through RSM multi-objective optimization, simultaneously considers mass transfer efficiency (provided by CFD) and substrate supply / product synthesis kinetics (determined experimentally), achieving true global optimization.

[0159] Comparative Example 4: Direct Amplification Method (without establishing a parameter transfer model) (compared with Example 1).

[0160] Method: The optimized operating conditions of the microplate (oscillation frequency 285 rpm, liquid volume 1.2 mL) were directly scaled up to the shake flask in a linear ratio without performing parameter conversion based on the hydrodynamic similarity criterion.

[0161] Results: Direct scaling up based on geometric similarity (well depth 17.5 mm → liquid level in shake flask approximately 40 mm, proportionally equivalent to an oscillation frequency of approximately 650 rpm, far exceeding the maximum speed of a shaker) is impractical. Using the same oscillation frequency of 285 rpm directly, the succinic acid concentration in the shake flask (100 mL volume) is only 8.6 ± 0.7 g / L, with a yield of 0.31 ± 0.03 g / g, significantly lower than the results from microplate.

[0162] Comparative analysis: The hydrodynamic environments of microplates and shake flasks are fundamentally different (geometric scale, operation method, etc.). Direct scaling up results in severely inadequate mass transfer conditions (measured kLa value 0.012 s⁻¹ vs. microplate 0.048 s). -1 This invention effectively solves the parameter matching problem in cross-scale scaling by establishing a parameter transfer model based on the equality of kLa.

[0163] Comparative Summary Table

[0164]

[0165] Comparative Example 5: Verification of universality with other strain systems.

[0166] To verify the universality of the present invention, comparative experiments with existing methods were conducted in the following strain systems:

[0167] (1) Production of L-lysine by Corynebacterium glutamicum

[0168] A 96-well microplate system was used. The method of this invention: CFD simulation of kLa in the range of 0.02-0.08s. -1 Optimal conditions (oscillation frequency 850 rpm, liquid volume 150 μL) were obtained through RSM optimization, resulting in a lysine yield of 18.2 g / L. The comparative example (traditional shake-flask method) yielded 16.5 g / L of lysine. The method of this invention increases yield by 10.3% and shortens the cycle time from 6 weeks to 3 weeks.

[0169] (2) Production of amylase by Bacillus subtilis

[0170] A 48-well microplate system was used. The method of this invention: using equal P / V as the scale-up criterion, the enzyme activity was scaled up from microplate (210 rpm, 2.0 mL) to shake flask (180 rpm, 100 mL), reaching 325 U / mL. Comparative example (empirical scale-up method): directly using the same shaking frequency of 210 rpm, the enzyme activity was only 215 U / mL. The method of this invention increases enzyme activity by 51.2%.

[0171] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid development method for microplate fermentation processes based on CFD simulation and response surface methodology, characterized in that, Includes the following steps: S1: Establish a three-dimensional CFD model of the target microporous plate, perform mesh generation and mesh independence verification, set boundary conditions and solution model, obtain fluid dynamic parameters under different operating conditions through CFD simulation calculation, and establish a quantitative mapping relationship between operating conditions and fluid dynamic parameters. S2: Using fluid dynamics parameters as constraints and fermentation performance indicators as response variables, a response surface model is constructed and multi-objective optimization is performed to obtain the Pareto optimal solution set. The optimal process parameters are then selected as the optimization result. S3: Based on the optimization results, a microplate prototype was manufactured using 3D printing technology for verification in fermentation experiments; S4: Conduct high-throughput fermentation verification experiments under optimized conditions, measure fermentation performance indicators, compare experimental results with model predictions, and if the deviation exceeds the preset threshold, feed the experimental data back to steps S1 and S2 to correct the model parameters and perform iterative optimization. S5: Establish the scale-up criteria between the microplate and the scale-up system, and convert the optimal process parameters of the microplate obtained in step S2 into the operating parameters of the scale-up system for scale-up verification.

2. The method according to claim 1, characterized in that, In step S1, the CFD simulation uses the VOF model to track the gas-liquid two-phase interface and combines it with the RNGk-ε turbulence model to simulate the turbulent flow of the liquid phase in the micropores. The boundary conditions are set as follows: the orifice wall is set as a non-slip wall boundary condition; the gas-liquid interface adopts a pressure outlet boundary condition; and the orbital oscillation trajectory is defined by a user-defined function (UDF). The fluid dynamic parameters include the volumetric mass transfer coefficient kLa, energy dissipation rate ε, and maximum shear stress τ. max Mean shear stress τ avg Mixing time t mix At least one of the following: gas holdup α.

3. The method according to claim 1, characterized in that, In step S2, the response surface model formula is as follows: Y=β0+Sβ i .x i +Sv ii .x i ²+SSv ij .x i .x j Where Y is the response variable, representing the fermentation performance index to be optimized, which includes the target product concentration C. P Biomass X, Product yield to substrate Y P At least one of / S; x i x j Let V be the independent variable, representing the operating conditions affecting fermentation performance. These operating conditions include the liquid volume V. L At least one of the following: oscillation frequency f, oscillation diameter d0, and temperature T; i and j are used to distinguish different independent variables; β0 is the constant number, representing the baseline value of the response variable when all independent variables are at zero levels; β i The coefficient of the linear term represents the independent variable x. i The linear main effects on Y; β ii The coefficient of the quadratic term represents the independent variable x. i The square of the bending effect of Y; β ij The coefficients of the interaction term represent the interaction coefficients between two distinct independent variables x. i With x j The impact of the interaction between them on Y.

4. The method according to claim 3, characterized in that, In step S2, the multi-objective optimization employs a multi-objective genetic algorithm or an artificial neural network-genetic algorithm to globally optimize the response surface model.

5. The method according to claim 1, characterized in that, The 3D printing technology mentioned in step S3 includes SLA, DLP or FDM technology, and the printing material is biocompatible photosensitive resin or PLA.

6. The method according to claim 1, characterized in that, In step S4, the high-throughput fermentation verification experiment is conducted online using a high-throughput microbioreactor, and the monitored parameters include at least one of biomass, pH, dissolved oxygen, and fluorescence signal.

7. The method according to claim 1, characterized in that, In step S5, the amplification criterion is the equal kLa criterion, the equal P / V criterion, or the equal τ criterion. max At least one of the criteria.

8. The method according to claim 1, characterized in that, The method is applicable to at least one of the following microbial systems: bacteria, yeast, filamentous fungi, and actinomycetes.

9. The method according to claim 1, characterized in that, The microplate is a 24-well, 48-well, or 96-well plate, with the holes being circular, hexagonal, or square in shape.

10. The method according to claim 9, characterized in that, The microporous plate is also provided with a baffle structure, which is at least one of a spiral baffle, a straight baffle, an arc baffle, and a grid baffle.

Images