3D printing precision prediction method and system based on gel processing characteristics
By combining the BP-ANN model with gel performance indicators, the problem of low prediction accuracy of food-grade 3D printing performance was solved, higher prediction accuracy and wider application capabilities were achieved, and the experimental process was simplified.
Patent Information
- Application Number
- CN202510784593.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-12
- Publication Date
- 2025-09-23
AI Technical Summary
Existing food-grade 3D printing performance prediction models have low accuracy, lack comprehensive consideration of the diverse material properties of food systems, and the testing process is time-consuming and labor-intensive, which limits the research and development and upgrading of 3D printing technology.
A 3D printing accuracy prediction model based on back-propagation artificial neural network (BP-ANN) was adopted. Gel performance indicators such as water holding capacity, consistency index and flow behavior index were used as input variables. The data was enhanced by small random noise. The model was optimized with regularized loss function. A data set was established and trained to improve the prediction accuracy.
It improves the accuracy and generalization ability of 3D printing performance prediction, simplifies the R&D process, reduces the number of experiments, and provides theoretical support and technical reference for the 3D printing performance of different food systems.
Smart Images
Figure CN120690348A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to 3D printing accuracy prediction, and in particular to a 3D printing accuracy prediction method and system based on food gel processing characteristics. Background Art
[0002] Traditional food-grade 3D printing performance experimental methods are linear function methods such as response surface and partial least squares. The testing process is time-consuming and labor-intensive, which greatly limits the research and development and upgrading of food-grade 3D printing technology. With the development of machine learning technology, neural networks can effectively predict 3D printing performance. However, the existing 3D printing performance prediction models for food-grade materials using ANN or BP-ANN models have low accuracy. For example, the determination coefficient R of 3D printing accuracy prediction established by Habus et al. based on the ANN model is 2 The range is 0.725 to 0.808; Jiao et al. predicted the 3D printing performance of corn starch-based food materials based on the BP-ANN model. The prediction model they established can effectively use LF-NMR data to accurately predict the rheological properties of corn starch gels. The determination coefficient R 2 The range is 0.831 to 0.894. In addition, existing prediction models often lack comprehensive consideration of the diverse material properties of food systems, nor do they consider the economy and practicality of the measurement indicators, which to some extent limits the promotion and application of 3D printing performance prediction models. Summary of the Invention
[0003] Purpose of the invention: The purpose of the present invention is to provide a method and system for accurately predicting 3D printing accuracy based on gel performance indicators, so as to solve the problems of low prediction accuracy and lack of comprehensive consideration of the diverse material properties of food systems in the existing technology.
[0004] Technical solution: The 3D printing accuracy prediction method based on gel processing characteristics of the present invention comprises the following steps: establishing a 3D printing accuracy prediction model based on a back-propagation artificial neural network, inputting gel performance indicators into the 3D printing accuracy prediction model, and obtaining predicted 3D printing accuracy;
[0005] The gel performance indicators include water holding capacity, consistency index and flow behavior index.
[0006] Furthermore, the calculation method of the water holding capacity WHC is: Where m0 is the mass of the centrifuge tube, m1 is the total mass of the initial gel and the centrifuge tube, and m2 is the total mass of the gel and the centrifuge tube after centrifugation;
[0007] The calculation method of the consistency index K and the flow behavior index η is: the gel is subjected to a static shear test, and the curve of apparent viscosity versus shear rate is obtained by fitting: η = K·γn-1 , where γ represents the shear rate and n represents the flow behavior index.
[0008] Furthermore, during the training of the 3D printing accuracy prediction model, a small amount of random noise is added to both the input gel performance index and the output 3D printing accuracy.
[0009] Furthermore, the loss function of the 3D printing accuracy prediction model is E=(1-α)·E MSE +α·E w , where E MSE is the mean square error, E w is the sum of squares of all weights, and α is the regularization coefficient.
[0010] Furthermore, a 3D printing dataset is established for training and testing the 3D printing accuracy prediction model; the method for establishing the 3D printing dataset is as follows:
[0011] Calculate the gel performance index of the gel sample, fill the gel sample into the 3D printer, complete 3D printing, and calculate the 3D printing accuracy of the 3D printed model. The 3D printing accuracy calculation method is:
[0012]
[0013] Among them, D0 represents the design size of the target 3D printing model, D m Represents the initial size after actual printing, D R and D H They represent the 3D printing size deviations of the concentric circle diameter and height after actual printing compared with the target 3D printing model.
[0014] Furthermore, by calculating the coefficient of determination R 2 The accuracy of the 3D printing accuracy prediction model was evaluated by the root mean square error (RMSE).
[0015] The 3D printing accuracy prediction system based on gel processing characteristics of the present invention includes:
[0016] A prediction model establishment unit is used to establish a 3D printing accuracy prediction model based on a back-propagation artificial neural network. The 3D printing accuracy prediction unit is used to input gel performance indicators into the 3D printing accuracy prediction model to obtain predicted 3D printing accuracy; the gel performance indicators include water retention, consistency index and flow behavior index.
[0017] Furthermore, the calculation method of the water holding capacity WHC is: Where m0 is the mass of the centrifuge tube, m1 is the total mass of the initial gel and the centrifuge tube, and m2 is the total mass of the gel and the centrifuge tube after centrifugation;
[0018] The calculation method of the consistency index K and the flow behavior index η is: the gel is subjected to a static shear test, and the curve of apparent viscosity versus shear rate is obtained by fitting: η = K·γ n-1 , where γ represents the shear rate and n represents the flow behavior index.
[0019] In the process of training the 3D printing accuracy prediction model, a small amount of random noise is added to the input gel performance index and the output 3D printing accuracy;
[0020] The loss function of the 3D printing accuracy prediction model is E = (1-α)·E MSE +α·E w ,in EMSE is the mean square error, E w is the sum of squares of all weights, and α is the regularization coefficient.
[0021] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the computer program is loaded into the processor, the 3D printing accuracy prediction method based on gel processing characteristics is implemented.
[0022] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for predicting 3D printing accuracy based on gel processing characteristics is implemented.
[0023] Beneficial Effects: Compared with existing technologies, the present invention offers the following advantages: The "key gel properties-3D printing performance" prediction model, constructed based on the BP-ANN principle and using key material properties of composite protein-based emulsified gels as input variables, has greater generalizability and application value, while also improving the accuracy and generalization of 3D printing performance predictions. This prediction method not only simplifies the R&D process and reduces the number of experiments, but also has the potential to provide theoretical support and technical reference for predicting the 3D printing performance of other different food systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 Schematic diagram of a hexagram three-dimensional model according to an embodiment of the present invention.
[0025] Figure 2 This is a diagram of the 3D printing performance prediction model architecture of the present invention.
[0026] Figure 3 This is a diagram showing the fitting results of the 3D printing performance prediction model in an embodiment of the present invention. DETAILED DESCRIPTION
[0027] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0028] The 3D printing accuracy prediction method based on gel processing characteristics of the present invention includes the following steps.
[0029] Step 1: Create a dataset.
[0030] This embodiment first completes the 3D printing test of the material and establishes a data set. Among them, the 3D printing model used to test the 3D printing performance of the composite protein-based emulsified gel is a six-pointed star three-dimensional model, the design diameter of the six-pointed star concentric circles is 50mm, and the three-dimensional height is 6mm. First, use the Repetier-Host software to set the 3D printing parameters: the first layer height is 1.2mm, the filling rate is 80%, the nozzle diameter is 1.2mm, the nozzle movement rate is 25mm / s, the wire diameter is 10mm, and the printing temperature is 25°C. After completing the parameter setting, import the six-pointed star three-dimensional model into the software, slice it using Slice 3r, and export the G-code file. Subsequently, the composite protein-based emulsified gel sample that has been refrigerated at 4°C for standby use is taken out, placed in a room temperature environment for equilibrium for 1 hour, and the gel sample is tightly filled in the storage bin of the food-grade 3D printer to ensure that there are no bubbles mixed in between the materials.
[0031] Determine the three gel performance indicators of the gel sample: WHC, consistency index K value, and flow behavior index n value:
[0032] (1.1) Determination of water holding capacity (WHC)
[0033] Accurately weigh 2 g of gel sample into a 10 mL centrifuge tube. The mass of the centrifuge tube is recorded as m0, and the mass of the initial sample and the centrifuge tube is recorded as m1. Centrifuge at 10,000 g for 15 minutes. After centrifugation, remove the supernatant and weigh the centrifuge tube and the precipitate, recording it as m2. Repeat the experiment three times. The WHC of each gel sample is calculated using formula (1).
[0034]
[0035] (1.2) Static shear test
[0036] At a temperature of 25°C and a shear rate of 1-100s -1 The curve of the apparent viscosity of the gel sample changing with shear rate was obtained under the conditions of , and the curve was fitted using the Ostwald-de-Waele model according to formula (2).
[0037] η=K·γ n-1 (2)
[0038] Where, η represents the apparent viscosity (Pa·s), K represents the consistency index (Pa·s n ), γ represents the shear rate (s -1), n represents the flow behavior index.
[0039] After the material is loaded, start the 3D printing program and officially begin the 3D printing test. After printing is completed, take photos to record the appearance of the printed model, and use an electronic vernier caliper to measure the actual size changes of the concentric circle diameter and height of the printed hexagram model. After printing, measure the concentric circle diameters and heights of the three diagonal groups of the hexagram model and compare them with the design dimensions of the target model. Use formulas (3) and (4) to calculate the 3D printing accuracy of the gel. The printed hexagram three-dimensional model is as follows: Figure 1 As shown, Figure 1 (a) is a three-dimensional image of the six-pointed star three-dimensional model, and (b) is a top view of the six-pointed star three-dimensional model.
[0040]
[0041] Among them, D0 represents the design size of the target model, D m Represents the initial size after actual printing. R and D H They represent the 3D printing size deviations of the concentric circle diameter and height after actual printing compared with the target model.
[0042] Table 1 shows part of the original dataset.
[0043] Table 1 Examples of some original data sets for building 3D printing accuracy prediction models based on BP-ANN principle
[0044]
[0045] Step 2: Establish a 3D printing accuracy prediction model.
[0046] like Figure 2As shown in the figure, a single-output prediction model for 3D printing accuracy was constructed using the BP-ANN (back-propagation artificial neural network) principle in MATLAB R2024a software. This model predicts the 3D printing accuracy (Y) using a three-dimensional vector (X1, X2, X3) composed of three gel performance indicators (WHC, consistency index K value, and flow behavior index n value) as input. The prediction models uniformly utilize the Levenberg-Marquardt back-propagation algorithm, which is suitable for training small datasets. Each prediction model has two hidden layers: the first hidden layer contains 25 neurons, and the second hidden layer contains 26 neurons. The training, validation, and test sets are divided into 75%, 15%, and 10% portions, respectively, for model parameter optimization, hyperparameter adjustment, and model performance evaluation. As shown in Table 1, the three gel performance indicators (X1, X2, and X3) are combined into a three-dimensional vector, which serves as the neural network input vector. The prediction models uniformly utilize the Levenberg-Marquardt back-propagation algorithm, which is suitable for training small datasets, to predict the 3D printing accuracy (Y).
[0047] Since the sample size of the original data is small (n = 51 in this embodiment), a small amount of random noise is added to the input values (water holding capacity WHC, consistency index K value, flow behavior index n value) and target value (3D printing accuracy) for data enhancement. The random noise obeys the normal distribution (mean 0, standard deviation 0.01) to ensure that the enhanced data still maintains the same distribution characteristics as the original data.
[0048] In addition, during the training of the neural network, a regularization coefficient of 0.15 is added to the loss function shown in formula (5) to automatically balance the error and the sum of squares of the weights when optimizing the objective function, and to constrain the final network weights to control the complexity of the prediction model and improve the generalization ability. The coefficient of determination of the indicator for evaluating the prediction performance of the prediction model is R 2 The calculation methods of the root mean square error (RMSE) are shown in formula (6) and formula (7).
[0049] As shown in formula (5), a regularization coefficient of 0.15 is introduced into the loss function during the training process of the neural network to control the complexity of the prediction model and improve the generalization ability.
[0050] E=(1-α)·E MSE +α·E w (5)
[0051] Where E is the overall error of the neural network on the sample set, E MSE is the mean square error (the error between the prediction and the true value), E w is the sum of squares of all weights, and α is the regularization coefficient 0.15.
[0052] As shown in formula (6) and formula (7), by calculating the determination coefficient R 2 The simulation accuracy of the BP-ANN training prediction model was evaluated by the root mean square error (RMSE).
[0053]
[0054] Where n is the number of samples in the prediction dataset, y pi is the predicted value of the i-th sample in the prediction data set, y ai is the actual value of the i-th sample in the predicted data set, y mean is the average value.
[0055] Table 2 shows the performance parameters of the 3D printing accuracy prediction model of the present invention. Figure 3 The figure shows the fitting result of the 3D printing accuracy prediction model of the present invention. Figure 3 (a), (b), (c), and (d) are the fitting results of the 3D printing accuracy prediction model on the training set, validation set, test set, and the entire data set, respectively. Figure 3 As shown in Table 2, the 3D printing accuracy prediction model of the present invention performs well in the training set, validation set and test set, with R 2 The RMSE values ranged from 0.9257 to 0.9957, with RMSE values less than 1.1932. The RMSE values for all datasets showed no unusual fluctuations, indicating that the model fits the data well, maintains good predictive performance when faced with unknown data, has strong generalization capabilities, and exhibits no significant overfitting. Furthermore, the data points in the fitted graph of the model are generally distributed near the ideal diagonal line (Y=T line), indicating a high degree of consistency between the model's predicted values and the actual values.
[0056] Table 2 Performance parameters of the 3D printing accuracy prediction model of the present invention
[0057]
[0058] Step 3: Collect the gel performance indicators of the 3D printing model to be tested and input them into the trained 3D printing accuracy prediction model to obtain the predicted 3D printing accuracy.
[0059] The 3D printing accuracy prediction system based on gel processing characteristics of the present invention includes:
[0060] A prediction model establishment unit is used to establish a 3D printing accuracy prediction model based on a back-propagation artificial neural network. The 3D printing accuracy prediction unit is used to input gel performance indicators into the 3D printing accuracy prediction model to obtain predicted 3D printing accuracy; the gel performance indicators include water retention, consistency index and flow behavior index.
[0061] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the computer program is loaded into the processor, the 3D printing accuracy prediction method based on gel processing characteristics is implemented.
[0062] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for predicting 3D printing accuracy based on gel processing characteristics is implemented.
[0063] The computer-readable storage medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices, flash memory, or any other medium that can be used to store program code in the form of instructions or data structures and can be accessed by a computer. The processor is used to execute the computer program stored in the memory to implement the various steps of the method involved in the above embodiments.
Claims
1. A 3D printing accuracy prediction method based on gel processing characteristics, characterized in that: The method comprises the following steps: establishing a 3D printing accuracy prediction model based on a back-propagation artificial neural network, inputting gel performance indicators into the 3D printing accuracy prediction model, and obtaining predicted 3D printing accuracy; The gel performance indicators include water holding capacity, consistency index and flow behavior index.
2. The 3D printing accuracy prediction method based on gel processing characteristics according to claim 1, characterized in that: The calculation method of the water holding capacity WHC is: Where m0 is the mass of the centrifuge tube, m1 is the total mass of the initial gel and the centrifuge tube, and m2 is the total mass of the gel and the centrifuge tube after centrifugation; The calculation method of the consistency index K and the flow behavior index η is: the gel is subjected to a static shear test, and the curve of apparent viscosity versus shear rate is obtained by fitting: η = K·γ n-1 , where γ represents the shear rate and n represents the flow behavior index.
3. The 3D printing accuracy prediction method based on gel processing characteristics according to claim 1, characterized in that: During the training of the 3D printing accuracy prediction model, a small amount of random noise is added to both the input gel performance index and the output 3D printing accuracy.
4. The 3D printing accuracy prediction method based on gel processing characteristics according to claim 1, characterized in that: The loss function of the 3D printing accuracy prediction model is E = (1-α)·E MSE +α·E w , where E MSE is the mean square error, E w is the sum of squares of all weights, and α is the regularization coefficient.
5. The 3D printing accuracy prediction method based on gel processing characteristics according to claim 1, characterized in that: Establishing a 3D printing dataset for training and testing the 3D printing accuracy prediction model; the method for establishing the 3D printing dataset is as follows: Calculate the gel performance index of the gel sample, fill the gel sample into the 3D printer, complete 3D printing, and calculate the 3D printing accuracy of the 3D printed model. The 3D printing accuracy calculation method is: Among them, D0 represents the design size of the target 3D printing model, D m Represents the initial size after actual printing, D R and D H They represent the 3D printing size deviations of the concentric circle diameter and height after actual printing compared with the target 3D printing model.
6. The 3D printing accuracy prediction method based on gel processing characteristics according to claim 1, characterized in that: By calculating the coefficient of determination R 2 The accuracy of the 3D printing accuracy prediction model was evaluated by the root mean square error (RMSE).
7. A 3D printing accuracy prediction system based on gel processing characteristics, characterized in that: include: A prediction model establishment unit is used to establish a 3D printing accuracy prediction model based on a back-propagation artificial neural network. The 3D printing accuracy prediction unit is used to input gel performance indicators into the 3D printing accuracy prediction model to obtain predicted 3D printing accuracy; the gel performance indicators include water retention, consistency index and flow behavior index.
8. The 3D printing accuracy prediction system based on gel processing characteristics according to claim 7, characterized in that: The calculation method of the water holding capacity WHC is: Where m0 is the mass of the centrifuge tube, m1 is the total mass of the initial gel and the centrifuge tube, and m2 is the total mass of the gel and the centrifuge tube after centrifugation; The calculation method of the consistency index K and the flow behavior index η is: the gel is subjected to a static shear test, and the curve of apparent viscosity versus shear rate is obtained by fitting: η = K·γ n-1 , where γ represents the shear rate and n represents the flow behavior index. In the process of training the 3D printing accuracy prediction model, a small amount of random noise is added to the input gel performance index and the output 3D printing accuracy; The loss function of the 3D printing accuracy prediction model is E = (1-α)·E MSE +α·E w , where E MSE is the mean square error, E w is the sum of squares of all weights, and α is the regularization coefficient.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is loaded into a processor, the method for predicting 3D printing accuracy based on gel processing characteristics according to any one of claims 1 to 6 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for predicting 3D printing accuracy based on gel processing characteristics according to any one of claims 1 to 6 is implemented.