Ink-jet printing parameter optimization method and system based on numerical simulation and machine learning

Abstract

The invention discloses an ink-jet printing parameter optimization method and system based on numerical simulation and machine learning. The method comprises the following steps: establishing a CFD-based two-dimensional axial symmetry model to simulate the processes of ink droplet generation, movement and deposition; generating droplet sizes under different conditions by utilizing a two-dimensional axial symmetry model, forming a data set, performing comparative analysis by adopting multiple regression algorithms, and selecting an optimal model as a prediction model; and the genetic algorithm is used for optimizing ink-jet printing parameters, a printing parameter combination for achieving the target droplet size is found, the prediction model is used for calculating the droplet size corresponding to the printing parameter combination, ink-jet printing is carried out, and it is further ensured that the generated droplet size meets the requirement. Through combination of numerical simulation and machine learning, ink-jet printing parameters are optimized, the number of experiments is reduced, and the printing efficiency and the printing quality are improved.

Description

Technical Field

[0001] The present invention relates to the field of inkjet printing and machine learning, and in particular to an inkjet printing parameter optimization method and system based on numerical simulation and machine learning. Background Art

[0002] Additive manufacturing (AM), also known as three-dimensional (3D) printing technology, is a technique for manufacturing objects by depositing materials layer by layer. Inkjet printing is a promising AM technology. It is a non-contact printing method that can accurately deposit micro / nano ink droplets on a substrate, suitable for small-scale 3D manufacturing. It has low equipment cost, simple operation, versatility in printing on different materials, and the ability to achieve high-resolution printing. However, one of its disadvantages is that the printing speed is relatively slow, which makes it less suitable for large-scale production. Recent research focuses on how to improve printing speed and printing efficiency without affecting printing quality. This technology has brought certain changes to many industries, including but not limited to 3D printing, flexible electronics, sensors, biomaterials, and optoelectronic devices.

[0003] At the same time, machine learning (ML) is becoming increasingly popular in additive manufacturing research. For machine learning in additive manufacturing, in addition to the typical application of prediction through data fitting, researchers are also exploring new ways to integrate machine learning and artificial intelligence (AI) methods into additive manufacturing. Machine learning algorithms are helping additive manufacturing practitioners improve product quality, optimize manufacturing processes, and reduce costs. A major challenge in the current field of additive manufacturing is the inconsistency of printed product quality, which largely depends on the irrationality of many printing parameters. Therefore, there is an urgent need to reasonably optimize the printing parameters to achieve better printing results.

[0004] With the growing demand for high-resolution printing in various fields today, it is particularly important to further optimize inkjet printing technology. However, there are still many challenges in achieving precise printing, such as droplet control, material selection and printing parameter optimization during the inkjet process. These challenges not only affect the printing quality, but also limit the scope of application of inkjet printing technology. In order to achieve high resolution of printing targets, traditional inkjet printing parameter adjustment methods usually rely on a large number of experiments, so they are time-consuming and labor-intensive. Therefore, the present invention proposes a method that combines numerical simulation and machine learning to optimize the parameters of inkjet printing, reduce the number of experiments, and thus improve printing efficiency and printing quality. Summary of the invention

[0005] The purpose of the present invention is to solve the problems in the prior art.

[0006] The technical solution adopted by the present invention to solve the technical problem is: to provide an inkjet printing parameter optimization method based on numerical simulation and machine learning, comprising the following steps:

[0007] The simulation model establishment step is to establish a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition;

[0008] A step of generating a data set, using a two-dimensional axisymmetric model to generate droplet sizes under different conditions to form a data set;

[0009] The step of establishing a prediction model is to use a variety of regression algorithms for comparative analysis based on the generated data set and select the optimal model as the prediction model;

[0010] Select parameter step, use genetic algorithm to optimize inkjet printing parameters, and find the best parameter combination to achieve target droplet size printing;

[0011] In the parameter verification step, a predictive model is used to verify whether the optimal parameter combination can achieve the printing of the target droplet size.

[0012] Preferably, the different conditions refer to inkjet printing parameters, including: voltage amplitude, voltage bias, pulse width of the printer and physical properties of the ink; wherein the physical properties of the ink include viscosity, surface tension and density.

[0013] Preferably, the step of establishing a simulation model uses computational fluid dynamics methods to simulate droplet generation, movement and deposition during inkjet printing, generating a two-dimensional axisymmetric model based on a leaky dielectric, which is used to analyze the interaction between the electric field, flow field and droplet formation, and capture the dynamic behavior of droplet generation, injection, movement and deposition.

[0014] Preferably, the step of generating a data set comprises the following steps:

[0015] Setting key input parameters for inkjet printing, including physical properties of ink, voltage amplitude, voltage bias, and pulse width;

[0016] According to the set key input parameters, the two-dimensional axisymmetric model is used to simulate the process of droplet generation, movement and deposition. After the simulation, the diameter of the deposited droplet is accurately measured, and the droplet diameter is used as the output variable, combined with the key input parameters to form structured data points.

[0017] Through multiple simulations, the system changes various input parameters to generate diverse data points to form a data set.

[0018] Preferably, the step of establishing a prediction model comprises the following steps:

[0019] Data preprocessing: dividing the dataset into training and test sets, then extracting input features for model training and normalizing or standardizing them;

[0020] Model selection, training and comparing multiple regression algorithms, wherein the training uses grid search or Bayesian optimization method to adjust hyperparameters, and the generalization ability of the model is evaluated by k-fold cross validation; the comparison comprehensively evaluates the model according to multiple indicators, and selects the model with the best performance as the prediction model for predicting droplet size; the indicators include mean square error (MSE), root mean square error (RMSE) and determination coefficient (R2).

[0021] Preferably, the step of selecting parameters comprises the following steps:

[0022] In the initialization step, a set of initial parameter combinations containing key parameters of inkjet printing are randomly generated; the droplet diameter corresponding to the number combination is calculated, and the fitness value is calculated by the absolute value of the droplet size error. The smaller the fitness value, the closer the parameter combination is to the target size.

[0023] Selection step, selecting two individuals with higher fitness in the population as the parent generation and the mother generation according to the fitness value;

[0024] The crossover step partially exchanges the key parameters of the parent and parent generations through the crossover operation to generate new offspring;

[0025] Mutation step: Use Gaussian mutation to randomly fine-tune the parameters of some new offspring, and use the fine-tuned offspring as the new population;

[0026] Repeat the steps, repeat the selection step to the variation step until the termination condition is reached, and the parameter combination obtained at this time is the optimal printing parameter; the termination condition includes reaching a set error threshold or running to a set number of generations.

[0027] Preferably, the parameter verification step comprises the following steps:

[0028] Select appropriate inks and measure their physical properties;

[0029] The target droplet size is input into the prediction model to calculate the corresponding key parameters of inkjet printing, and an inkjet printing experiment is performed to measure the actual droplet size;

[0030] Compare the actual droplet size with the target size to evaluate printing accuracy.

[0031] Preferably, the parameter verification step further uses a two-dimensional axisymmetric model to verify whether the printing parameter combination can achieve printing of the target droplet size, specifically comprising the following steps:

[0032] The optimal parameter combination is input into the 2D axisymmetric model;

[0033] The two-dimensional axisymmetric model completely simulates the droplet generation, ejection, movement and deposition process according to the optimal parameter combination to obtain the simulated droplet size;

[0034] The obtained simulated droplet size is compared with the target size, and the error between the two is calculated to determine whether the optimal parameter combination obtained by genetic algorithm optimization meets the design requirements.

[0035] The present invention also provides an inkjet printing parameter optimization system based on numerical simulation and machine learning, comprising:

[0036] Establish a simulation model module and build a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition;

[0037] Generate dataset module, using two-dimensional axisymmetric model to generate droplet sizes under different conditions to form dataset;

[0038] Establish a prediction model module, use multiple regression algorithms to conduct comparative analysis based on the generated data set, and select the optimal model as the prediction model;

[0039] Select the parameter module and use the genetic algorithm to optimize the inkjet printing parameters to find the optimal parameter combination to achieve the target droplet size printing;

[0040] The verification parameter module uses a predictive model to verify whether the optimal parameter combination can achieve the printing of the target droplet size.

[0041] The present invention has the following beneficial effects:

[0042] (1) The present invention significantly reduces the reliance on experimental data by combining numerical simulation and machine learning algorithms, avoids repeated manual adjustment of printing parameters, greatly reduces the number of experiments, and saves time and resources;

[0043] (2) The present invention can accurately predict and control the droplet size under different conditions through optimized printing parameters, improve the accuracy and consistency of inkjet printing, ensure the final printing quality, and is suitable for application scenarios under high-precision conditions such as micro-nano manufacturing;

[0044] (3) The data set generation, machine learning model selection and genetic algorithm optimization methods in the present invention are suitable for different types of printing inks and printing equipment, have strong adaptability, and can be promoted and applied in multiple fields such as electronic manufacturing and biosensors.

[0045] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, but the present invention is not limited to the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1A method step diagram of an embodiment of the present invention;

[0047] Figure 2 It is a structural schematic diagram of an inkjet printing device;

[0048] Figure 3 It is a schematic diagram of a two-dimensional axisymmetric CFD numerical model;

[0049] Figure 4 This is a numerical simulation diagram of the droplet generation process;

[0050] Figure 5 Generate a flow chart for the dataset;

[0051] Figure 6 Flowchart for machine learning model selection and training;

[0052] Figure 7 It is the flow chart of genetic algorithm;

[0053] Figure 8 Print droplets for some experiments;

[0054] Fig. 9 4 is a system structure diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0055] See also Figure 1 FIG. 1 is a diagram showing steps of a method according to an embodiment of the present invention, comprising:

[0056] S101, a simulation model establishment step, establishing a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition;

[0057] S102, a step of generating a data set, using a two-dimensional axisymmetric model to generate droplet sizes under different conditions to form a data set;

[0058] S103, a prediction model establishment step, based on the generated data set, using multiple regression algorithms for comparative analysis, and selecting the optimal model as the prediction model;

[0059] S104, a parameter selection step, using a genetic algorithm to optimize inkjet printing parameters, and finding an optimal parameter combination for printing a target droplet size;

[0060] S105, a parameter verification step, uses a prediction model to verify whether the optimal parameter combination can achieve printing of a target droplet size.

[0061] The following is a detailed and clear description of the embodiments of the present invention. It should be clearly stated that the embodiments only represent a part of the present invention, not all of it.

[0062] (1) Numerical modeling: The structure of the inkjet printing device is as follows: Figure 2 As shown in Figure 1, computational fluid dynamics (CFD) was used to simulate the droplet generation, movement and deposition during inkjet printing. Figure 3 The 2D axisymmetric model based on a leaky dielectric is shown in Figure 2 to analyze the interaction between the electric field, flow field and droplet formation. Figure 4 The entire process from Taylor cone formation to droplet deposition is shown, capturing the dynamic behavior of droplet generation, ejection, movement, and deposition. This provides a basis for subsequent parameter optimization and accurately reflects the key physical mechanisms of inkjet printing.

[0063] (2) Dataset generation: Figure 5 As shown in Figure 2, the process of dataset generation is divided into four steps.

[0064] a) Input variable determination: Users need to set the key input parameters of inkjet printing, including the physical properties of the ink (such as viscosity, surface tension, etc.), voltage amplitude, voltage bias and pulse width, etc. These parameters provide the necessary boundary conditions for numerical simulation and cover various droplet generation situations.

[0065] b) Numerical simulation: According to the set input parameters, the process of droplet generation, movement and deposition is simulated.

[0066] c) Droplet size measurement: After each simulation, the diameter of the deposited droplets is accurately measured to ensure the accuracy of the data.

[0067] d) Dataset generation: The droplet diameter is used as the output variable and combined with input parameters (such as ink properties, voltage amplitude, bias, pulse width, etc.) to form a structured data set. Through multiple simulations, the system changes each input parameter to generate a variety of data points, which provides a basis for the subsequent training of machine learning models.

[0068] (3) The present invention uses a machine learning model to predict the droplet size under different inkjet parameter combinations. Based on the generated data set, a variety of regression algorithms are used for comparative analysis, and the optimal model is selected for subsequent parameter optimization. The model selection process is as follows: Figure 6 As shown, the main steps are as follows:

[0069] a) Data preprocessing: First, the dataset is divided into a training set (80%) and a test set (20%). Then, the input features for model training (such as ink properties, voltage amplitude, bias, pulse width and other parameters) are extracted. Finally, the input features are normalized or standardized to eliminate dimensional differences and improve training effects.

[0070] b) Model selection: Train and compare multiple regression algorithms and select the most suitable model.

[0071] c) Model training and optimization: Use grid search or Bayesian optimization methods to adjust hyperparameters to ensure optimal model performance.

[0072] d) Cross-validation: The generalization ability of the model is evaluated through k-fold cross-validation to ensure its stable performance under different data set conditions.

[0073] e) Model performance evaluation: The model is evaluated based on multiple indicators (mean square error (MSE), root mean square error (RMSE), and coefficient of determination (R2).

[0074] f) Selecting the best model: Based on the evaluation results, the model with the best performance is finally selected for droplet size prediction. The evaluation results of the embodiment of the present invention are shown in Table 1:

[0075] Table 1 Comparison results of machine learning models

[0076]

[0077] Finally, the random forest model was selected for subsequent parameter optimization.

[0078] (4) Figure 7 As shown, the present invention uses a genetic algorithm (GA) to optimize the key parameter combination of inkjet printing to achieve precise control of droplet size. The specific steps are as follows:

[0079] a) Initialize the population: randomly generate a set of initial parameter combinations containing key parameters of inkjet printing (such as voltage amplitude, voltage bias, pulse width, ink properties, etc.) to ensure population diversity and cover a wide range of parameter space.

[0080] b) Fitness evaluation: Use the trained machine learning model to predict the droplet diameter corresponding to each parameter combination, and calculate the fitness value through the absolute value of the droplet size error. The smaller the fitness value, the closer the parameter combination is to the target size.

[0081] c) Selection: Select individuals with higher fitness as "parents" based on their fitness values ​​to pass on excellent parameter combinations to the next generation.

[0082] d) Crossover: Through the crossover operation, the key parameters of the two "parent generation" individuals are partially exchanged to generate new offspring, increase genetic diversity, and explore new parameter combinations.

[0083] e) Mutation: Use Gaussian mutation to randomly fine-tune the parameters of some offspring to avoid falling into local optimal solutions and further improve the optimization effect.

[0084] f) Termination condition: The genetic algorithm stops running according to a predetermined condition (such as reaching a set error threshold or running for a set number of generations), and the parameter combination obtained at this time is the optimal printing parameter.

[0085] Through the global search of the genetic algorithm, the optimal parameter combination that meets the target droplet size can be efficiently found, which verifies the stability and reliability of the optimization results and significantly improves the accuracy and efficiency of inkjet printing.

[0086] (5) Numerical simulation verification: The optimal parameter combination obtained by genetic algorithm optimization is input into the CFD numerical simulation model to verify whether the expected results can be obtained. According to the optimized parameters (such as voltage amplitude, voltage bias, pulse width, ink properties, etc.), the CFD simulation program is run to fully simulate the droplet generation, injection, movement and deposition process. The simulated droplet size is compared with the target size, and the error (ME) between the two is calculated to evaluate the accuracy of parameter optimization. According to the numerical simulation results, the change of droplet diameter under different Z values ​​(representing the relationship between the physical properties of the ink) is analyzed, as shown in Table 2:

[0087] Table 2: Numerical model validation results

[0088]

[0089] Multiple simulation results show that the parameter combination optimized by genetic algorithm can better meet the design requirements and provide a basis for subsequent experimental verification.

[0090] (6) Experimental verification: In this step, first select a suitable ink and measure its physical properties (such as viscosity, surface tension, density, etc.) to ensure that the experimental parameters are accurate. Subsequently, adjust the printing device parameters, input the target droplet size (such as 20μm, 40μm, 60μm, 80μm, 100μm or 120μm) based on the previously trained machine learning model, and calculate the corresponding inkjet printing parameters (such as voltage amplitude, bias, pulse width, etc.). Next, use the calculated optimal parameter combination to perform an inkjet printing experiment and generate droplets of corresponding sizes. The actual droplet size is measured by high-precision measuring equipment and compared with the target size to evaluate the printing accuracy. For different target sizes, multiple experiments are conducted to take the size average to ensure the reliability and consistency of the results. Figure 8 The droplet morphology generated by printing in some experiments is shown. Table 3 shows the comparison results between the target droplet size and the actual average droplet size in the experiment:

[0091] Table 3 Experimental verification results

[0092]

[0093] Through experimental verification, the effectiveness of genetic algorithm optimization parameters under various target sizes was demonstrated, precise control of droplet generation was achieved, and the expected printing effect was achieved.

[0094] See also Fig. 9 FIG. 1 is a system structure diagram of an embodiment of the present invention, including:

[0095] Establishing a simulation model module 901, establishing a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition;

[0096] Generate a data set module 902, using a two-dimensional axisymmetric model to generate droplet sizes under different conditions to form a data set;

[0097] Establishing a prediction model module 903, based on the generated data set, using multiple regression algorithms for comparative analysis, and selecting the optimal model as the prediction model;

[0098] Select parameter module 904 to optimize inkjet printing parameters using a genetic algorithm to find the best parameter combination for printing a target droplet size;

[0099] The parameter verification module 905 uses a prediction model to verify whether the optimal parameter combination can achieve printing of a target droplet size.

[0100] It can be seen that the present invention optimizes the parameters of inkjet printing by combining numerical simulation and machine learning, which can reduce the number of experiments and thus improve printing efficiency and printing quality.

[0101] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing inkjet printing parameters based on numerical simulation and machine learning, characterized in that: The following steps are involved: The simulation model establishment step is to establish a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition; A step of generating a data set, using a two-dimensional axisymmetric model to generate droplet sizes under different conditions to form a data set; The step of establishing a prediction model is to use a variety of regression algorithms for comparative analysis based on the generated data set and select the optimal model as the prediction model; Select parameter step, use genetic algorithm to optimize inkjet printing parameters, and find the best parameter combination to achieve target droplet size printing; In the parameter verification step, a predictive model is used to verify whether the optimal parameter combination can achieve the printing of the target droplet size.

2. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The different conditions refer to inkjet printing parameters, including: voltage amplitude, voltage bias, pulse width of the printer and physical properties of the ink; wherein the physical properties of the ink include viscosity, surface tension and density.

3. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The step of establishing a simulation model adopts a computational fluid dynamics method to simulate the droplet generation, movement and deposition in the inkjet printing process, and generates a two-dimensional axisymmetric model based on a leaky dielectric, which is used to analyze the interaction between the electric field, flow field and droplet formation, and capture the dynamic behavior of droplet generation, injection, movement and deposition.

4. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The step of generating a data set comprises the following steps: Setting key input parameters for inkjet printing, including physical properties of ink, voltage amplitude, voltage bias, and pulse width; According to the set key input parameters, the two-dimensional axisymmetric model is used to simulate the process of droplet generation, movement and deposition. After the simulation, the diameter of the deposited droplet is accurately measured, and the droplet diameter is used as the output variable, combined with the key input parameters to form structured data points. Through multiple simulations, the system changes various input parameters to generate diverse data points to form a data set.

5. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The step of establishing a prediction model comprises the following steps: Data preprocessing: dividing the dataset into training and test sets, then extracting input features for model training and normalizing or standardizing them; Model selection, training and comparing multiple regression algorithms, wherein the training uses grid search or Bayesian optimization method to adjust hyperparameters, and the generalization ability of the model is evaluated by k-fold cross validation; the comparison comprehensively evaluates the model according to multiple indicators, and selects the model with the best performance as the prediction model for predicting droplet size; the indicators include mean square error (MSE), root mean square error (RMSE) and determination coefficient (R) 2 .

6. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The step of selecting parameters comprises the following steps: In the initialization step, a set of initial parameter combinations containing key parameters of inkjet printing are randomly generated; the droplet diameter corresponding to the number combination is calculated, and the fitness value is calculated by the absolute value of the droplet size error. The smaller the fitness value, the closer the parameter combination is to the target size. Selection step, selecting two individuals with higher fitness in the population as the parent generation and the mother generation according to the fitness value; The crossover step partially exchanges the key parameters of the parent and parent generations through the crossover operation to generate new offspring; Mutation step: Use Gaussian mutation to randomly fine-tune the parameters of some new offspring, and use the fine-tuned offspring as the new population; Repeat the steps, repeat the selection steps to the variation step until the termination condition is reached, and the parameter combination obtained at this time is the optimal printing parameter; the termination condition includes reaching a set error threshold or running to a set number of generations.

7. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The step of verifying parameters comprises the following steps: Select appropriate inks and measure their physical properties; The target droplet size is input into the prediction model to calculate the corresponding key parameters of inkjet printing, and an inkjet printing experiment is performed to measure the actual droplet size; Compare the actual droplet size with the target size to evaluate printing accuracy.

8. The inkjet printing parameter optimization method based on numerical simulation and machine learning according to claim 1, characterized in that: The parameter verification step also uses a two-dimensional axisymmetric model to verify whether the printing parameter combination can achieve the printing of the target droplet size, and specifically includes the following steps: The optimal parameter combination is input into the 2D axisymmetric model; The two-dimensional axisymmetric model completely simulates the droplet generation, ejection, movement and deposition process according to the optimal parameter combination to obtain the simulated droplet size; The obtained simulated droplet size is compared with the target size, and the error between the two is calculated to determine whether the optimal parameter combination obtained by genetic algorithm optimization meets the design requirements.

9. An inkjet printing parameter optimization system based on numerical simulation and machine learning, characterized in that: include: Establish a simulation model module and build a two-dimensional axisymmetric model based on CFD to simulate the process of ink droplet generation, movement and deposition; Generate dataset module, using two-dimensional axisymmetric model to generate droplet sizes under different conditions to form dataset; Establish a prediction model module, use multiple regression algorithms to conduct comparative analysis based on the generated data set, and select the optimal model as the prediction model; Select the parameter module and use the genetic algorithm to optimize the inkjet printing parameters to find the optimal parameter combination to achieve the target droplet size printing; The verification parameter module uses a predictive model to verify whether the optimal parameter combination can achieve the printing of the target droplet size.

Images