Scalable physical reactor characterization

Abstract

The present invention relates to a method for manufacturing a model reactor for reflecting specifically defined hydrodynamics of a target reactor of different sizes than the model reactor, to a model reactor manufactured by this method, and to the use of a model reactor in process scaling-up or scaling-down.

Description

Technical Field

[0001] The present invention relates to the field of biotechnological production plants for producing valuable products, to reactor modelling and its use in scaling up or down production processes. Background Art

[0002] Perfect DownScale Reactor Representation (pDS) attempts to address the scale-up problem in biotechnology manufacturing. In biotechnology process development, small pilot reactors are initially used, typically with a volume of about 1 liter or less. Subsequently, the processes established in these small pilot reactors are gradually transferred to large production reactors, which can typically have a volume of several cubic meters. The fluid physics of small-scale reactors can differ significantly from those of large production reactors, such as the distribution of supplied nutrients or the required mixing time. All of these parameters depend on the volume and shape of the reactor used and are inherent physical properties of reactors of different sizes. Transferring a custom process established for a specific reactor size to a reactor of a different size is time-consuming, resource-intensive, and costly. The process needs to be essentially redeveloped at each reactor scale.

[0003] US2021230532A1 discloses a scalable manufacturing system and method for treating cells in a bioreactor based on fluid dynamics considerations. This method attempts to optimize culture conditions without changing the reactor geometry. WO2020205611A1 describes a computational method for modeling bioreactors by combining a mechanistic model of metabolic flux kinetics and flux balance analysis to predict cell culture performance in the reactor. Due to the lack of a universal and effective cell metabolism model that is independent of the culture and product, the predictive power of these methods is limited.

[0004] CN11283644A discloses a method for optimizing a dry anaerobic biogas mixing system based on computational fluid dynamics (CFD). This method optimizes CFD modeling by varying various factors within the CFD model itself to achieve a more accurate representation of the CFD modeling of the dry anaerobic biogas mixing system. The method further describes the use of optimized, high-performance CFD modeling to simplify the computational model and enable rapid evaluation of predefined structures. The "optimized model" mentioned in CN11283644A relates to the optimization of the CFD modeling process itself, rather than the internal geometry of the reactor.

[0005] In addition, some previous studies on non-stirred vessels and static mixers have used CFD as a research tool but without automated random design. For example, US2019 / 0388859A1 describes a CFD evaluation of a static mixer for a continuous flow catalytic reactor. Static mixers can be configured for continuous flow chemical reactors, such as tubular continuous flow chemical reactors for heterogeneous catalytic reactions.

[0006] CN112100944A describes the use of CFD to visualize multiphase systems and evaluate different predefined reactor geometries. It relates to the visualization of reactor conditions, specifically the study of flow fields at different scales using simulation tools (CFD). The use of particle image velocimetry is shown to validate and optimize CFD simulations at different scales.

[0007] Current approaches to address process transfer issues are inadequate, time-consuming, and technically complex. For example, adapting small-scale mixing parameters to large-scale production reactors may result in suitable mixing characteristics but, on the other hand, may lead to poor cell growth.

[0008] In the past, attempts to directly transfer data have been made by either adopting data from conventional small-scale cultures, or attempting to replicate small-scale conditions at scale, or using models to predict behavior in poorly mixed systems by using data from well-mixed small-scale cultures. These attempts have a high probability of failure, primarily due to the unavoidable scale-dependent physics of fluids. All of these approaches have had limited success until comprehensive and robust cell models became available.

[0009] An alternative approach involves placing cells in an environment that mimics large-scale production conditions. This approach uses a two-vessel, small-scale system, where one reactor performs mixing and nutrient supply, while the other, which can be a mixing vessel or tubular reactor, does not. As a complete system, this two-vessel system has comparable mixing times to those of large-scale vessels, but the nutrient and oxygen distributions are still insufficient to fully represent large-scale conditions, and its implementation is technically complex. Another approach uses a two-vessel-within-a-single-vessel system, where the culture volume is divided into two by a partition. This achieves the same effect as a two-reactor system in terms of mixing time, but still fails to accurately mimic large-scale nutrient and oxygen distributions.

[0010] Therefore, there is a need for a method that allows direct transfer of process parameters from small laboratory scale to large production scale without introducing any complications in the established bioprocess. Summary of the Invention

[0011] The object of the present invention is to provide a method that enables a bioprocess established in a model reactor to be smoothly transferred to a corresponding production reactor. This object is achieved by the subject matter of the present invention.

[0012] According to the present invention, there is provided a method for generating a reactor geometry having specifically defined fluid dynamic properties, comprising:

[0013] a. Define the objective function;

[0014] b. applying an algorithm to generate at least three reactor geometries, wherein the inner wall of the reactor has a randomized structure;

[0015] c. determining the fluid dynamics characteristics of the reactor geometry having the randomized structure generated in step b) using automated computational fluid dynamics;

[0016] d. Rank all generated reactor geometries according to the objective function of step a);

[0017] e. Selecting the best-fitting randomized structure, where the number of structures selected is less than the number of structures generated;

[0018] f. applying an automated algorithm to randomly revoke the best-fit structure selected in step e) and add further randomized structures to generate at least three additional reactor geometries having randomized structures;

[0019] g. iteratively repeating steps c to f until a target criterion of similarity between the objective function value and the randomly generated reactor geometry characteristics is found; and

[0020] h. Optionally provide a digital file for 3D printing the generated reactor.

[0021] According to one embodiment of the present invention, the randomized structure is derived from a library of structural elements. The structural elements are voxel-based additives or spacers.

[0022] Another embodiment relates to the methods described herein, wherein the outer diameter, height, and minimum wall thickness of the target reactor are used as inputs.

[0023] Another embodiment relates to the method described herein, wherein the generation of completely closed spaces during the random structure generation in steps c and e is avoided.

[0024] Another embodiment relates to the method described herein, wherein the generation of structures in the predetermined spaces of components inserted into the reactor after the reactor is manufactured is avoided. Components are sensors, stirrers, addition ports, etc.

[0025] Another embodiment relates to the methods described herein, wherein the generated reactor geometry does not include overhanging structures according to the intended 3D printing orientation.

[0026] Another embodiment relates to the methods described herein, wherein the resulting reactor geometry has a minimum wall thickness.

[0027] Another embodiment relates to the methods described herein, wherein the generated reactor geometry comprises a minimum opening diameter of any structure.

[0028] Another embodiment relates to the method described herein, wherein the objective function is a target value derived from an existing reactor.

[0029] Another embodiment is directed to the methods described herein, wherein the single vessel mixing reactor is generated based on defined fluid dynamics derived from a large scale target reactor.

[0030] Another embodiment relates to the methods described herein, wherein the resulting reactor is a small-scale reactor.

[0031] Another embodiment relates to the method described herein, wherein the computational method is selected from computational fluid dynamics, engineering equations, and deep learning AI derived CFD equivalent methods.

[0032] Another embodiment is directed to the method described herein, wherein the target value is one or more of kLa, mixing time, nutrient distribution, and the like.

[0033] Another embodiment relates to the methods described herein wherein the objective function is the optimal value of any fluid dynamics property such as kLa, mixing time, and nutrient distribution.

[0034] Another embodiment is directed to the method described herein, wherein for the second and any subsequent iterations, the objective function defined in step a and / or the method of adding elements in step b are adjusted.

[0035] One embodiment of the present invention is directed to a reactor that reflects the same mixing characteristics as target reactors of different sizes, wherein the reactor has a randomized structure on the inner wall.

[0036] The reactor can be a small-scale reactor that reflects the same fluid flow distribution as the target reactor. The reactor has a specially designed internal wall structure that can reflect, for example, the mixing characteristics of an existing target reactor, which can be of the same or different size.

[0037] Another embodiment relates to the methods described herein, wherein the reactor is made by the methods described herein.

[0038] Another embodiment relates to the method described herein, wherein the reactor is manufactured by 3D printing.

[0039] One embodiment of the present invention relates to the use of the reactor according to the present invention.

[0040] Another embodiment relates to the use of a reactor as described herein, wherein the reactor is used for scale-up or scale-down of a bioprocess.

[0041] Another embodiment relates to the use of a reactor as described herein, wherein the reactor is used to transfer a bioprocess to a large-scale bioreactor.

[0042] Another embodiment relates to the use of a reactor as described herein, wherein the reactor is used to transfer a bioprocess from large scale to large scale.

[0043] One embodiment of the present invention relates to a bioprocess scale-up method comprising the following steps:

[0044] a. determining at least one objective function derived from large-scale reactor characteristics;

[0045] b. generating a small-scale model reactor according to the method of any one of claims 1 to 17;

[0046] c. Fabricate a small-scale model reactor by 3D printing;

[0047] d. Setting up the bioprocess in a small-scale model reactor;

[0048] e. Determine process parameters; and

[0049] f. Transfer of process parameters to large-scale reactors.

[0050] Another embodiment relates to the method described herein, wherein the determined objective function is selected from one or more of mixing time, nutrient feed mixing, shear rate, oxygen distribution, nutrient distribution, mass transfer, and the like.

[0051] As used herein, the term "objective function" refers to any function that describes an objective. The objective function can have an optimization goal of minimization, maximization, a range, or a specific value, and can be composed of one or more constraints / equations that can be combined into a single equation. The objective function can be as simple as a maximization / minimization goal of a single parameter (e.g., mixing time), or can contain any combination of any number of terms, such as a combination of minimum mixing time and a specific kLa.

[0052] The specifically defined fluid dynamics characteristics can be determined by evaluating the fluid dynamics characteristics of an existing reactor or by defining optimization parameters. The fluid dynamics characteristics of an existing reactor can be determined by computational fluid dynamics (CFD) software.

[0053] According to one embodiment of the present invention, specifically defined fluid dynamics of an existing large-scale reactor are determined and used, or specifically defined optimization parameters are used. The specifically defined optimization parameters are selected from a group of process-related parameters, such as mixing time, power input, etc., or from a group of reactor-related parameters, such as volume, surface area, etc.

[0054] According to one embodiment of the present invention, the specific shape of the reactor structure is generated based on the results obtained from CFD software, according to the mixing time or component distribution results calculated by the CFD software. According to one example, the mixing time or the component distribution of dissolved oxygen and / or nutrients is calculated using CFD software.

[0055] One embodiment of the present invention relates to a model reactor that reflects the same mixing characteristics as an existing reactor, wherein the model reactor has a structure that simulates the mixing characteristics of the existing reactor.The model reactor includes a predetermined wall structure.

[0056] The model reactor may be a small scale reactor or a pilot scale reactor, or even a large scale reactor.

[0057] Model reactors can be manufactured by 3D printing.

[0058] One embodiment of the present invention relates to the use of the model reactor described herein in the scale-up or scale-down of a bioprocess.

[0059] A manufacturing process amplification method comprises the following steps:

[0060] a. Determine the process conditions for large-scale production reactors;

[0061] b. transferring the process conditions determined in step a) to a small-scale reactor;

[0062] c. Based on the transfer evolution model reactor structure of step b);

[0063] d. generating a 3D model of the reactor having the structure derived from step c);

[0064] e. establishing a manufacturing process in the model reactor of step d); and

[0065] f. Transfer the established process to large-scale production reactors.

[0066] The transformation and evolution of the model reactor can be performed using computational fluid dynamics (CFD) software or the like.

[0067] According to one embodiment of the present invention, the determined process conditions are one or more of mixing time, nutrient feed mixing, shear rate and mass transfer. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1A : Traditionally used step-by-step process from laboratory scale to large scale.

[0069] Figure 1B : Scaling-up scheme of the pDS reactor technology according to the present invention.

[0070] Figure 2 : The difference between traditional numerical model-based scale-up methods and the generation of dedicated physical small-scale reactors in this application and their application in process scale-up.

[0071] Figure 3A : A detailed decision parameter and iterative approach using random structure generation / reduction and selection to generate new reactor designs. In each generation (iteration) of this process, the reactor is expected to meet the desired target value (objective function) more and more.

[0072] Figure 3B : Selected generational maps showing the interrelationships between different generations and their respective parent and offspring structures.

[0073] Figure 4 The mixing time for each generation from generation 1 to generation 5 is shown to achieve a mixing time of at least 40 seconds to characterize the large-scale mixing situation.

[0074] Figure 5 shows the tracer distribution results after 8 seconds (5A) and 21 seconds (5B) after top addition of the tracer in a small-scale bioreactor vessel (1 liter).

[0075] Figure 6 Mixing curves for different reactor geometries are shown, which provide different mixing times and different mixing behaviors compared to exponential washout behavior.

[0076] Figure 7 The workflow is demonstrated, starting from a large-scale target reactor, determining the fluid dynamics using CFD software, and ultimately obtaining a small-scale laboratory model reactor with the same fluid dynamics as the large-scale target reactor.

[0077] Figure 8 A process transfer solution from scale-up to large-scale is demonstrated. DETAILED DESCRIPTION

[0078] The present invention provides a method for smoothly transferring a bioprocess established in a reactor to a target reactor of different size.

[0079] One embodiment of the present invention relates to a method for designing a small-scale reactor with a specific mixing time or as high a mixing time as possible. One of the most important differences between a small-scale reactor and a large-scale reactor is the mixing time, which is significantly longer for a large-scale reactor. This unavoidable basic fluid dynamics principle is one of the main reasons for the performance differences when the same bioprocess is run in a small-scale reactor and a large-scale reactor (and vice versa). For example, a process can first be developed in a small-scale reactor by providing a small-scale reactor with the same long mixing time as a large-scale reactor to enable it to exhibit the performance of a large-scale reactor.

[0080] An often overlooked issue is the system's bulk fluid flow and how it varies with scale. Fluid displacement and velocity are not linear at different scales, so the mixing curve of liquid motion varies not only based on volume, tip velocity, and diameter, but also on the spatial distribution of power within the reactor and the presence of large and small vortices. In practice, the design should adjust the local mixing performance of the small-volume system to match the bulk mixing and mass transfer performance of the large-scale system. Good designs will account for the effects of baffles or flow disruptors.

[0081] Typical process development in biotechnology involves first determining the optimal parameters at a small scale and then transferring the optimized process to a larger scale. Because mixing times in large-scale reactors are significantly longer than in small-scale reactors and process performance is expected to change during scale-up, process scale-up is currently done by increasing reactor size stepwise (see Figure 1A ) to reduce costs and risks when large-scale process performance declines. This process is time-consuming and costly, and it is not possible to achieve the optimal process parameters for large-scale production in this way because the cost of developing a process that has reached production scale is too high. To reduce costs and risks, pDS technology is used to scale up the process (see Figure 1B ).

[0082] In addition to saving costs and time, efficient optimization and scale-up of biotechnology processes are crucial to achieving multiple sustainable development goals. This is particularly evident considering the impact of microbial biotechnology in the agri-food, environmental, biopharmaceutical and chemical industries. Many of these applications require scale-up after proof-of-concept. However, the behavior of such biological culture systems remains unpredictable when moving from small-scale laboratory to industrial conditions. To this end, a complete computational framework for scale-up or scale-down is required. Although model-based scale-up exists, as described in this article, the currently available models are for specific products and hosts. This means that large-scale production data are required for each process to recalibrate the model, resulting in very limited usefulness in terms of saving time and cost ( Figure 2The present invention provides a method for direct scale-up from laboratory to production scale. This method is a host- and product-independent scale-up or scale-down method. The present invention also provides a model reactor with a long mixing time.

[0083] Scaling up fermentation processes in the biotechnology industry is a key step in bringing bioprocess and bioproduct innovations to commercialization, and life sciences will become the cornerstone of the future circular economy. This includes carbon dioxide capture technology, bioplastics, novel foods, and biopharmaceuticals such as cancer treatments, vaccines, and gene therapy. As mentioned above, the scale-up step is known for performance losses and development delays. Scaling down methods are considered a wise way to reduce these problems, but current technologies are either technically complex (dual-container systems) or limited to manual rather than automated design. The method proposed in this invention overcomes existing shortcomings through the following interrelated steps:

[0084] Detailed analysis of conditions in large-scale target reactors;

[0085] Computational characterization of dynamic fluid properties;

[0086] Using the proposed automated structure generation method for small-scale bioreactors, dynamic fluid properties were transferred to a laboratory-scale model reactor (see Figure 2 );

[0087] Fabricate small-scale model reactors with complex internal geometries, for example, by 3D printing;

[0088] Develop customized processes at the bench-scale under conditions representative of the target large-scale reactor;

[0089] Finally, successful results are directly transferred to large-scale target reactors without any intermediate steps ( Figure 1B ).

[0090] The method provided by the present invention avoids the time-consuming, resource-intensive and costly scale-up approach currently used in biotechnology manufacturing processes - that is, first establishing the process in a small pilot reactor and then transferring it to larger reactors in a series of steps with increasing volume until a final production volume of several cubic meters is reached.

[0091] The target reactor can be a large-scale production reactor in a factory for producing various desired biotechnology products. The volume of a large-scale production reactor can be about 100 liters, 500 liters, 1000 liters or up to several cubic meters.

[0092] The model reactor can be a small-scale laboratory reactor. The volume of the small-scale model reactor can be 0.1 liter, 0.5 liter, 1.0 liter, 1.5 liter, 2.0 liter, 5.0 liter or up to 10.0 liters.

[0093] Specifically, the fluid physics of reactors of varying sizes and shapes vary significantly. For example, the distribution and mixing times of nutrients are crucially dependent on the size and shape of the reactor being used. Whenever an existing, established process is transferred to a reactor of a different size, significant time and resources are required to redevelop the process in the target reactor to achieve the same results, but this is not always possible.

[0094] The goal of one embodiment is to achieve a mixing time that represents a large-scale reactor. For example, a 20 cubic meter reactor typically has a mixing time of approximately 40-50 seconds. The small-scale model reactor generated should reflect a mixing time of approximately 40-50 seconds. The overall decision-making scheme for generating a random structure that generates a specific fluid dynamics behavior in a small-scale model reactor is shown in Figure 3. The cycle of structure generation and deletion will continue as needed to achieve a specific goal. In this case, the algorithm will automatically generate a new reactor geometry based on the last iteration and automatically evaluate (determine its mixing time) by CFD until generating a reactor structure with a required mixing time of approximately 40 seconds.

[0095] Using this method, mixing time will inevitably increase with each iteration if parameters are chosen that are appropriate for the number and types of structures generated and deleted in each iteration. According to an exemplary embodiment of the present invention, a 1-liter bioreactor was designed using the described method, with a target mixing time of at least 40 seconds. Typical mixing times for 1-liter bioreactors are only 2-5 seconds, far from representative of the mixing times exceeding 40 seconds typically seen in large-scale reactors. Figure 4 The mixing times of various generations of bioreactors randomly generated using the described method are shown, showing the successful increase in mixing time after each iteration. These progressively longer mixing times are achieved by randomly adding and removing structures from the bioreactor's 3D architecture. Each generation, the best performing structure is selected based on an objective function and used as the basis for the next generation of additions / removals.

[0096] The increase in mixing time depends on the addition of internal wall structures, which hinder the mixing of any added substances (such as nutrients) because they disrupt the efficient mixing of small-scale reactors. In this example, the nature and design of the internal wall structures were chosen to be purely random. The mixing behavior can be easily visualized by adding scalars to the CFD simulation from the top of the reactor, simulating the addition of nutrients through the feed line at the top of the reactor. Figure 5A Figure 2. B shows the tracer distribution in one of the reactor configurations generated in this example, with a mixing time exceeding 40 seconds. This behavior is impossible in traditional bioreactors, where the contents are homogenized within 2–5 seconds at small scale. Only iterative random configuration methods can achieve this tracer distribution, representative of a large-scale reactor.

[0097] The difference between a traditional single-vessel or dual-vessel system separated by a single partition can also be seen in the standard deviation of the tracer variation over time in the bioreactor. For dual-vessel systems and bioreactors with a perforated partition dividing the vessel into two compartments, the resulting curve consistently resembles an exponential washout curve. This exponential washout curve is not observed in large bioreactors because they contain different zones that mix differently with the bulk of the bioreactor liquid, resulting in more complex mixing of nutrients over time ( Figure 6 The mixing curves shown, represented by the standard deviation of the tracer over time, provide a range of behaviors from quasi-exponential washout behavior (lower curve) to more complex mixing behaviors that may be more suitable for scaled-down bioreactors for process development and more representative of mixing curves in large-scale bioreactors. The present method can use mixing time as an objective function or Figure 6 The mixing curve shown serves as an objective function to generate a physical small-scale reactor that more accurately reflects the conditions in the large-scale bioreactor.

[0098] In one embodiment of the present invention, the reactor is evaluated using M-Star CFD, although any other CFD modeling software, such as Siemens StarCCM+ or open source solutions (such as OpenFoam), may also be used. The basic idea of ​​CFD is to solve the equations of motion (i.e., the Navier–Stokes equations) for the fluid in the reactor. These equations describe the conservation of mass and momentum balance of the fluid. In particular, momentum balance has nonlinear characteristics that almost prohibit analytical solutions for all actual flows. This requires discretization of the mathematical equations and solving them numerically. During the amplification process, the flow becomes more turbulent, and a characteristic of flow turbulence is that there is a wide range of scales in space and time, resulting in significant mixing differences between small-scale and large-scale reactors.

[0099] In this example, the histogram of well-mixed / poorly mixed regions, along with any other values ​​or sequences calculated by CFD, can be used as objective functions to determine whether the large-scale and small-scale reactors have identical flow characteristics. According to the present invention, any target attribute can be used as long as it can be calculated using appropriate software (e.g., CFD). Even combinations of nominal attributes can be used, such as a mixing time > 40 seconds while targeting a specific kLa gas transmission coefficient.

[0100] The shape of the small-scale and large-scale reactors can be chosen to any desired form, as long as they can be modeled in 3D. This can be a simple cylindrical reactor with a known volume, equipped with an agitator, optionally including feed lines, or any other layout, size, or configuration required for the reactor's use.

[0101] According to one embodiment of the present invention, a simple cylindrical reactor with a volume of 1 liter can be used as a model reactor, or a reactor with any other geometric structure can be used.

[0102] The reactor provided by the present invention is a physically scaled-down model of large-scale fluid dynamics behavior and is therefore applicable to scenarios independent of the host and product. The lack of a detailed model for cell behavior is an obstacle to the predictive power of digital models of traditional methods (see Figure 2 ), the present invention does not involve cell modeling and is not limited thereto. The cells used were transferred to a physical environment representative of a large-scale model production reactor, allowing the process parameters developed at a small scale to be directly scaled up to large-scale production without performance loss (Figure 1, right).

[0103] Finally, after digitally generating the geometry of the small-scale reactor that represents the mixing behavior of the large-scale reactor, the small model reactor can be manufactured by 3D printing, and the method of the present invention has provided a digital file for 3D printing. 3D printing is also the only manufacturing technology that can accurately and quickly manufacture the complex geometries generated by the method. The production of model reactors can be easily achieved by 3D printing. Since the algorithm ensures that the reactor is suitable for 3D printing, the physical reduction of the reactor can be achieved by any 3D printing technology. Suitable 3D printing processes are, for example, laser sintering, FDM, DLP, LCA or other technologies. Suitable 3D printing materials are, for example, thermoplastics (such as PLA or ABS), resins, medical resins, glass and ceramic resins, as well as any sintering material from plastic to metal, without any restrictions. The method of generating the reactor geometry and the subsequent manufacture of the reactor are related as follows Figure 7 shown.

[0104] The mixing characteristics of the large-scale reactor were evaluated, and the small-scale reactors were iteratively developed to achieve the same flow characteristics by randomly changing the small-scale reactor structure and selecting the best performing structure for the next generation.

[0105] Different methods can be used to select model reactors for each generation (e.g., using the best x-rank and also using some outliers). In addition, various mutation methods can be used only for certain generations (e.g., making large changes to the reactor structure in the first iteration and smaller changes in subsequent iterations).

[0106] The process from the model reactor to the large-scale production reactor is an evolutionary approach (random variation and selection of the best performers) using the same previously determined objective function.

[0107] The optimization cycle mainly includes the following steps:

[0108] 1. Select an appropriate model reactor for the cycle;

[0109] - If it is the first cycle, use the model reactor provided;

[0110] - Otherwise, the best performing reactor (or reactors) is selected;

[0111] 2. Crossbreed or randomly mutate the model reactor with other well-performing samples from the previous generation to create multiple reactor variants (approximately 100 samples, this number can be freely selected);

[0112] 3. All reactor variants were run through CFD software;

[0113] 4. Calculate the grade based on CFD results in combination with key parameters of a large-scale model production reactor;

[0114] 5. Repeat this cycle as needed until the calculated CFD results are close enough to the pre-defined key parameters. This may take up to 50 cycles (this number can vary depending on the specific simulation).

[0115] Generate a model mutation reactor based on the following steps:

[0116] a. Create random structures until the target volume is reached;

[0117] b. Delete random structures until the target volume is reached;

[0118] c. Create weighted random structures with seed points to achieve larger structures and generate more variations in each generation;

[0119] d. Create structures from predefined elements (such as disks on walls, spheres, etc.);

[0120] e. Variation of the size and position of predefined component structures;

[0121] f. Use predefined elements as starting points for random structures.

[0122] Model reactors can be generated by hybrid approaches, such as first creating a predefined structure, then applying a weighted random structure approach, and finally mutating the predefined structure accordingly.

[0123] Finally, the corresponding algorithm is used to generate a 3D printable file of the model reactor. The quality of the model reactor is checked to ensure that its structure is valid (for example, the model reactor wall has no holes and the printable structure does not require any supports).

[0124] The optimization cycle ends when the key CFD parameters are close enough to the predefined parameters of the large-scale production reactor.

[0125] Another embodiment of the present invention uses a small-scale representative reactor as described herein to transfer a large-scale process operating in one facility to another facility, where the processes in both facilities utilize large-scale reactors.

Claims

1. A method for generating a reactor geometry having specifically defined fluid dynamic properties, comprising: a. Define the objective function; b. applying an algorithm to generate at least three reactor geometries having a randomized structure on the inner wall of the reactor; c. determining the fluid dynamics characteristics of the reactor geometry having the randomized structure generated in step b) using automated computational fluid dynamics; d. ranking all generated reactor geometries according to the objective function of step a); e. Selecting the best-fitting randomized structure, where the number of selected structures is less than the number of generated structures; f. applying an automated algorithm to randomly revoke the best-fit structure selected in step e) and add further randomized structures to generate at least three additional reactor geometries having randomized structures; g. iteratively repeating steps c to f until a target criterion of similarity between the objective function value and the randomly generated reactor geometry characteristics is found; and h. Optionally provide a digital file for 3D printing the generated reactor. The method according to claim 1 , wherein the randomized structure is derived from a library of structural elements. The method of claim 2 , wherein the structural element is a voxel-based additive or spacer.

4. The method according to claim 1, wherein the outer diameter, height and minimum wall thickness of the target reactor are used as input.

5. The method according to any one of claims 1 to 4, wherein completely closed spaces are avoided during the random structure generation process in steps c and e.

6. The method according to any one of claims 1 to 4, wherein the generation of structures in predetermined spaces of components inserted into the reactor after the reactor is manufactured is avoided.

7. The method of claim 6, wherein the component is a sensor, a stirrer, an addition port, or the like.

8. The method according to any one of claims 1 to 7, wherein the generated reactor geometry does not include overhanging structures according to the intended 3D printing direction.

9. The method according to any one of claims 1 to 8, wherein the resulting reactor geometry has a minimum wall thickness.

10. The method of any one of claims 1 to 9, wherein the resulting reactor geometry comprises a minimum opening diameter of any structure.

11. The method according to any one of claims 1 to 10, wherein the objective function is a target value derived from an existing reactor.

12. The process according to any one of claims 1 to 11, wherein the single vessel mixing reactor is generated based on defined fluid dynamics derived from a large scale target reactor.

13. The process according to any one of claims 1 to 11, wherein the resulting reactor is a small-scale reactor.

14. The method according to any one of claims 1 to 13, wherein the computational method is selected from computational fluid dynamics, engineering equations, and deep learning AI derived CFD equivalent methods.

15. The method according to claim 11 or 14, wherein the target value is one or more of kLa, mixing time, nutrient distribution, etc.

16. The method according to any one of claims 1 to 15, wherein the objective function is the optimal value of any fluid dynamics property such as kLa, mixing time and nutrient distribution.

17. The method according to any one of claims 1 to 16, wherein for the second and any subsequent iterations, the objective function defined in step a and / or the method of adding elements in step b are adjusted.

18. A reactor that reflects the same mixing characteristics as a target reactor of different size, wherein the reactor has a randomized structure on the inner wall.

19. The reactor according to claim 18, which is produced by the method according to any one of claims 1 to 17.

20. The reactor according to claim 18 or 19, wherein the reactor is manufactured by 3D printing.

21. Use of a reactor according to any one of claims 18 to 20, wherein the reactor is used for scale-up or scale-down of a bioprocess.

22. Use according to claim 21, wherein the reactor is used to transfer a bioprocess to a large-scale bioreactor.

23. The use according to claim 20, wherein the reactor is used to transfer a bioprocess from large scale to large scale.

24. A bioprocess scale-up method comprising the following steps: a. determining at least one objective function derived from large-scale reactor characteristics; b. generating a small-scale model reactor according to the method of any one of claims 1 to 17; c. Manufacturing the small-scale model reactor by 3D printing; d. setting up the bioprocess in the small-scale model reactor; e. Determine the process parameters; and f. Transferring the process parameters to a large-scale reactor.

25. The method of claim 24, wherein the determined objective function is selected from one or more of mixing time, nutrient feed mixing, shear rate, oxygen distribution, nutrient distribution, mass transfer, etc.

Images