A solid propellant combustion performance prediction method based on an AI model

Abstract

The application provides a solid propellant combustion performance prediction method based on an AI model, which is executed according to the following steps: step one, multi-modal parameter coupling design: multi-modal parameter coupling design is carried out on the solid propellant, the combustion performance parameters of the propellant to be predicted and the formula component parameters affecting the combustion performance parameters are determined, and a multi-modal parameter coupling prediction system taking the formula component parameters as input parameters and the combustion performance parameters as output parameters is established; step two, sample data is obtained; step three, a differentiated normalization standard is established; step four, a combustion performance prediction model is constructed: an AI model for predicting the combustion performance of the propellant is established, model training is carried out, and a combustion performance prediction model meeting the accuracy requirement is obtained; the AI model is a neural network model with a high-pressure compensation sub-network; and step five, prediction.

Description

Technical Field

[0001] This invention belongs to the field of solid propellant technology and relates to solid propellant performance prediction, specifically to a method for predicting the combustion performance of solid propellants based on an AI model. Background Technology

[0002] Predicting the combustion performance of solid propellants is a crucial aspect of rocket engine design. Traditional methods primarily rely on an iterative model of formulation design and experimental verification, conducting hot-fire tests or combustion bench experiments on prototype samples to directly measure parameters such as burning rate and pressure index. The cost of a single such experiment can reach hundreds of thousands of yuan, and due to safety regulations, it can only be repeated after several days. In terms of numerical simulation, although computational fluid dynamics (CFD) can simulate the flow field in the combustion chamber, it requires modeling complex processes such as multiphase flow and chemical reaction kinetics. Full 3D simulations of typical operating conditions consume millions of meshes and thousands of CPU seconds.

[0003] In recent years, data-driven methods based on artificial intelligence have demonstrated significant advantages. By systematically collecting historical experimental data and constructing a structured database, deep neural networks can efficiently uncover hidden correlations between components, processes, and performance. The trained model can output predicted values ​​for new formulations within seconds, significantly improving efficiency compared to traditional methods. However, for the prediction of the combustion performance of solid propellants, due to the complexity of propellant composition, high parameter correlation, and different influencing factors for different combustion performance parameters, current research and technologies have the following shortcomings:

[0004] First, traditional prediction models lack sufficient correlation among individual influencing parameters.

[0005] Second, the global normalization of traditional methods leads to the loss of key feature information in influencing factors.

[0006] Third, traditional network structures are not adaptable to different propellant formulation systems, which affects prediction errors. Summary of the Invention

[0007] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a solid propellant combustion performance prediction method based on an AI model, so as to solve the technical problem that the efficiency and accuracy of traditional models in the existing technology for predicting propellant combustion performance need to be further improved.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0009] A method for predicting the combustion performance of solid propellants based on an AI model, comprising the following steps:

[0010] Step 1, Multimodal parameter coupling design:

[0011] Multimodal parameter coupling design is carried out for solid propellants to determine the combustion performance parameters of the propellant to be predicted and the formulation component parameters that affect the combustion performance parameters. A multimodal parameter coupling prediction system is established with the formulation component parameters as input parameters and the combustion performance parameters as output parameters.

[0012] The combustion performance parameters include burning rate, pressure index, and temperature sensitivity coefficient.

[0013] The formulation component parameters include oxidant particle size distribution, thickness of the oxide layer on the surface of the metallic fuel, and mixing temperature process parameters.

[0014] Step 2, Obtain sample data:

[0015] The formulation components that affect the performance indicators obtained in step one are used to design a formulation. Based on the designed formulation, the propellant combustion performance parameters are experimentally measured to obtain sample data for predicting the propellant combustion performance.

[0016] Step 3: Establish differentiated normalization standards:

[0017] The sample data obtained in step two is divided into datasets. The coupling coefficient is obtained according to the three-parameter coupling formula of the formulation component parameters. The classification system is obtained by dynamic classification based on the range of the coupling coefficient. The sample data is processed by dynamic classification normalization method according to the classification system to establish differentiated normalization standards.

[0018] Step 4: Construct a combustion performance prediction model:

[0019] An AI model for predicting propellant combustion performance was established and trained to obtain a combustion performance prediction model that meets the accuracy requirements.

[0020] The AI ​​model described is a neural network model with a high-voltage compensation subnetwork added.

[0021] In the neural network model, the number of input layer nodes corresponds to the number of formulation component parameters input to the multimodal parameter coupling prediction system obtained in step one, and the number of output layer nodes corresponds to the number of combustion performance parameters output by the multimodal parameter coupling prediction system obtained in step one.

[0022] Step 5, Prediction:

[0023] Input the composition parameters of the solid propellant, use the combustion performance prediction model trained in step four to make predictions, and output the prediction results of the combustion performance parameters.

[0024] Compared with the prior art, the present invention has the following technical effects:

[0025] (I) The method of the present invention adopts multimodal parameter coupling design to construct an association network that includes material composition, microstructure, process parameters and macroscopic properties, thereby solving the problem of insufficient correlation of traditional single parameters.

[0026] (II) The present invention adopts a differentiated normalization method based on material properties, which avoids the loss of key feature information caused by traditional global normalization.

[0027] (III) The algorithm model of this invention improves the prediction accuracy by dynamically adjusting the network structure to adapt to different formulation systems.

[0028] The specific content of the present invention will be further described in detail below with reference to the embodiments. Detailed Implementation

[0029] It should be noted that, unless otherwise specified, all functions, models, and methods in this invention are functions, models, and methods known in the prior art.

[0030] Following the above technical solutions, specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent modifications made based on the technical solutions of this application fall within the protection scope of the present invention.

[0031] Example:

[0032] This embodiment presents a method for predicting the combustion performance of solid propellants based on an AI (Artificial Intelligence) model. The method is performed according to the following steps:

[0033] Step 1, Multimodal parameter coupling design:

[0034] Multimodal parameter coupling design is carried out for solid propellants to determine the combustion performance parameters of the propellant to be predicted and the formulation component parameters that affect the combustion performance parameters. A multimodal parameter coupling prediction system is established with the formulation component parameters as input parameters and the combustion performance parameters as output parameters; this solves the problem of insufficient correlation of traditional single parameters.

[0035] In this embodiment, the multimodal parameter coupling design constructs a correlation network that includes material composition, microstructure, process parameters and macroscopic properties.

[0036] In step one, the combustion performance parameters include burning rate, pressure index, and temperature sensitivity coefficient. In this embodiment, among the combustion performance parameters, burning rate is the design benchmark for engine operating time, pressure index is the core indicator of combustion stability, and temperature sensitivity coefficient is an environmental adaptability evaluation parameter.

[0037] In step one, the formulation component parameters include oxidant particle size distribution, metal fuel surface oxide layer thickness, and mixing temperature process parameters. In this embodiment, among the formulation component parameters, oxidant particle size distribution is the core parameter for combustion rate adjustment, metal fuel surface oxide layer thickness is a key microscopic parameter that directly affects propellant energy release efficiency and combustion stability, and mixing temperature process parameters are key parameters for controlling particle dispersion uniformity.

[0038] In this embodiment, a multimodal parameter coupling prediction system was finally established with oxidant particle size distribution (D10 / D50 / D90), metal fuel surface oxide layer thickness (5-150nm), and mixing temperature (-20~60℃) as input parameters, and burning rate, pressure index, and temperature sensitivity coefficient as output parameters.

[0039] Step 2, Obtain sample data:

[0040] The formulation components that affect the performance indicators obtained in step one are used to design a formulation. Based on the designed formulation, the propellant combustion performance parameters are experimentally measured to obtain sample data for predicting the propellant combustion performance.

[0041] Step two specifically includes the following steps:

[0042] Step 201: First, an experimental design is conducted for the three formulation factors: oxidant particle size distribution, metal fuel surface oxide layer thickness, and mixing temperature. Based on the different values ​​of each formulation factor, the oxidant particle size distribution is set to level 10, the metal fuel surface oxide layer thickness is set to level 10, and the mixing temperature is set to level 5. For this mixing design with different levels of each factor, a pseudo-level method is required for adjustment. At the same time, the uniformity of the pseudo-level design is calculated using the centralized L2 deviation formula. The final design table is determined based on the minimum deviation, and finally, the design scheme table for each formulation is obtained.

[0043] In this embodiment, to ensure the uniformity of the training sample data, experiments were designed for three formulation factors: oxidant particle size distribution, thickness of the oxide layer on the surface of the metal fuel, and mixing temperature.

[0044] In step 201, the formula for calculating the centering L2 bias is:

[0045] In the formula:

[0046] CD 2 Indicates the deviation value;

[0047] d represents the number of factors;

[0048] n represents the number of trials;

[0049] k represents the sequence number of the kth test point;

[0050] j represents the sequence number of the j-th dimension;

[0051] This represents the normalized coordinate of the k-th test point in the j-th dimension.

[0052] Step 202: Based on the design scheme table obtained in step 201, the particle size distribution is measured by laser diffraction, the oxide layer thickness is analyzed by XPS (X-ray photoelectron spectroscopy), and the mixing temperature is measured by an online thermal imager. Sample data is obtained with oxidant particle size distribution, metal fuel surface oxide layer thickness and mixing temperature as input parameters, and burning rate, pressure index and temperature sensitivity coefficient as output parameters.

[0053] Step 3: Establish differentiated normalization standards:

[0054] The sample data obtained in step two is divided into datasets. The coupling coefficient is obtained according to the three-parameter coupling formula of the formulation component parameters. The classification system is obtained by dynamic classification based on the range of the coupling coefficient. The sample data is processed by dynamic classification normalization method according to the classification system to establish differentiated normalization standards.

[0055] In step three, the dataset is partitioned by randomly dividing the sample data into training samples, validation samples, and test samples in a 7:2:1 ratio. In this embodiment, the training set contains 70 test data points, the validation set contains 20 test data points, and the test set contains 10 test data points.

[0056] In step three, the three-parameter coupling formula is:

[0057] In the formula:

[0058] α represents the coupling coefficient;

[0059] D 50 Indicates the median particle size of the oxidant particle size distribution;

[0060] OI oxide layer thickness;

[0061] PC indicates the mixing temperature.

[0062] In step 302, the classification system is as follows:

[0063]

[0064] In step three, the dynamic classification normalization method is as follows:

[0065] For Class I classification, oxidant particle size distribution and oxide layer thickness are jointly dominant, and the formula for the normalization standard is: In the formula, X Ⅰ This represents the normalized value of the input parameter for Category I classification level.

[0066] For Category II classification, the mixing temperature and oxide layer thickness have a dual effect, and the formula for the normalization standard is: In the formula, X Ⅱ This represents the normalized value of the input parameter for the Class II classification level.

[0067] For Category III classification levels, there is no significant synergistic effect among the parameters, and the formula for the normalization standard is: In the formula, X Ⅲ This represents the normalized value of the input parameter for the Class III classification level.

[0068] Step 4: Construct a combustion performance prediction model:

[0069] An AI model for predicting propellant combustion performance was established and trained to obtain a combustion performance prediction model that meets the accuracy requirements.

[0070] The AI ​​model is a neural network model that incorporates a high-voltage compensation subnetwork.

[0071] In the neural network model, the number of nodes in the input layer corresponds to the number of formulation component parameters input to the multimodal parameter coupling prediction system obtained in step one, and the number of nodes in the output layer corresponds to the number of combustion performance parameters output by the multimodal parameter coupling prediction system obtained in step one.

[0072] Step four specifically includes the following steps:

[0073] Step 401: Establish a neural network model that includes an input layer, at least three hidden layers, and an output layer.

[0074] Step 402: Add a high-voltage compensation sub-network. The high-voltage compensation sub-network contains three hidden layers (containing 32-32-16 neurons respectively). It is automatically activated when the oxidant content is detected to be >80%, thus realizing an adaptive network architecture based on dynamic branching structure.

[0075] In this embodiment, the high-pressure compensation subnetwork is mainly used to address the non-ideal combustion behavior of the propellant when the oxidizer content is >80%. It includes: a first hidden layer for handling the high-pressure fragmentation characteristics of oxidizer crystals, using the ELU (Exponential Linear Unit) activation function; a second hidden layer for considering the melting dynamics of the oxide layer, using the SiLU (Sigmoid Gated Linear Unit) activation function; and a third hidden layer for generating pressure-sensitive compensation terms, using a linear output function.

[0076] In this embodiment, the dynamic branching structure is implemented through a dynamic routing controller, which calculates the oxidant content threshold in real time to achieve branch signal transmission, and uses the sigmoid function to achieve a smooth transition zone of 78-82%.

[0077] In step 402, the adaptive network architecture couples the basic neural network and the high-voltage compensation sub-network, and achieves adaptive weighting of the basic prediction and the high-voltage compensation output through learnable weight coefficients α.

[0078] In step 402, the adaptive weighted implementation equation for the basic prediction and high-voltage compensation output is:

[0079]

[0080] In the formula:

[0081] This represents the output parameter results of the adaptive network;

[0082] w b Indicates the basic network weight coefficients;

[0083] w c This represents the weight coefficient of the high-voltage compensation network;

[0084] F b Indicates the output parameters of the basic network;

[0085] F c This indicates the output parameter results of the high-voltage compensation network;

[0086] α is the weighting coefficient;

[0087] AP% indicates the percentage of oxidant content;

[0088] P represents pressure.

[0089] In this embodiment, the weighting coefficient α is used to integrate the formulation and process parameters.

[0090] Step 403: Based on the adaptive network architecture obtained in step 403, network training is carried out. A two-stage training strategy is adopted. First, the basic network training compensation module is frozen and the learning rate is set to 1e-3. Then, joint fine-tuning is performed and the learning rate is set to 5e-5. The loss function adopts the dynamic weighted mean square error loss function. The training data is iteratively trained. When the error of the validation set loss function decreases by less than 1% for five consecutive rounds, the training is terminated to obtain a combustion performance prediction model that meets the accuracy requirements.

[0091] In step 403, the dynamic weighted mean squared error loss function is as follows:

[0092]

[0093] In the formula:

[0094] L represents the loss error;

[0095] n represents the number of samples;

[0096] w i Indicates dynamic weights;

[0097] y i This represents the training data values ​​for each output parameter;

[0098] This represents the predicted data values ​​for each output parameter;

[0099] clip represents the truncation function;

[0100] x i This represents the training data values ​​for each input parameter;

[0101] 'a' represents the threshold;

[0102] b represents the width of the transition interval.

[0103] Step 5, Prediction:

[0104] Input the composition parameters of the solid propellant, use the combustion performance prediction model trained in step four to make predictions, and output the prediction results of the combustion performance parameters.

[0105] Comparative Example 1:

[0106] This comparative example presents a method for predicting the combustion performance of solid propellants, which is carried out according to the following steps:

[0107] Step 1: Select only the oxidant particle size D50 as the input parameter, and output parameters include predicted burning rate, pressure index, and temperature sensitivity coefficient.

[0108] Step 2: Measure the corresponding combustion rate based on different oxidant particle size distributions to obtain sample data.

[0109] Step 3: Divide the sample data into training set, validation set and test set, and use global uniform normalization processing.

[0110] Step four: Train the data to obtain the prediction model.

[0111] Step 5: Import the data to be predicted into the prediction model for testing to obtain the prediction parameter results.

[0112] Comparative Example 2:

[0113] This comparative example presents a method for predicting the combustion performance of solid propellants. The difference between this method and the example is that, compared to step four in the example, a sigmoid activation function is used and no high-pressure compensation mechanism is set. The other steps are the same as in the example.

[0114] Using the same test data, the prediction results of propellant combustion performance using the method of the present invention, Comparative Example 1, and Comparative Example 2 were compared, and the final error results are shown in Table 1.

[0115] Table 1. Errors in combustion performance prediction results using different methods

[0116] Comparative Example 1 11.73％ 12.12％ 15.64％ Comparative Example 2 7.11％ 5.97％ 6.23％ Embodiments of the present invention 3.12％ 2.98％ 3.65％

[0117] As shown in Table 1, using the method of the present invention and the adaptive network algorithm can effectively improve the prediction efficiency and significantly improve the prediction accuracy of propellant combustion performance.

Claims

1. A method for predicting the combustion performance of solid propellants based on an AI model, characterized in that, This method is performed according to the following steps: Step 1, Multimodal parameter coupling design: Multimodal parameter coupling design is carried out for solid propellants to determine the combustion performance parameters of the propellant to be predicted and the formulation component parameters that affect the combustion performance parameters. A multimodal parameter coupling prediction system is established with the formulation component parameters as input parameters and the combustion performance parameters as output parameters. The combustion performance parameters mentioned include burning rate, pressure index, and temperature sensitivity coefficient; The formulation component parameters include oxidant particle size distribution, thickness of the oxide layer on the surface of the metallic fuel, and mixing temperature process parameters; Step 2, Obtain sample data: The formulation components that affect the performance indicators obtained in step one are used to design a formulation. Based on the designed formulation, the propellant combustion performance parameters are experimentally measured to obtain sample data for predicting the propellant combustion performance. Step 3: Establish differentiated normalization standards: The sample data obtained in step two is divided into datasets. The coupling coefficient is obtained according to the three-parameter coupling formula of the formulation component parameters. The classification system is obtained by dynamic classification based on the range of the coupling coefficient. The sample data is processed by dynamic classification normalization method according to the grading system to establish differentiated normalization standards. In step three, the three-parameter coupling formula is as follows: ; In the formula: Represents the coupling coefficient; Indicates the median particle size of the oxidant particle size distribution; Oxide layer thickness; Indicates the mixing temperature; In step three, the dynamic classification normalization method is as follows: For Category I classification levels, the formula for the normalization standard is: In the formula, express The normalized values ​​of the category classification level input parameters; For Category II classification levels, the formula for the normalization standard is: In the formula, express The normalized values ​​of the category classification level input parameters; For Category III classification levels, the formula for the normalization standard is: In the formula, Indicate The normalized values ​​of the category classification level input parameters; Step 4: Construct a combustion performance prediction model: Establish an AI model to predict propellant combustion performance and conduct model training to obtain a combustion performance prediction model that meets the accuracy requirements; The AI ​​model described is a neural network model with a high-pressure compensation sub-network. The high-pressure compensation sub-network contains three hidden layers, which are automatically activated when the oxidant content is detected to be >80%. The first hidden layer is used to handle the high-pressure fragmentation characteristics of oxidant crystals and uses the ELU activation function. The second hidden layer is used to consider the melting dynamics of the oxide layer and uses the SiLU activation function. The third hidden layer is used to generate pressure-sensitive compensation terms and uses a linear output function. In the neural network model, the number of input layer nodes corresponds to the number of formulation component parameters input to the multimodal parameter coupling prediction system obtained in step one, and the number of output layer nodes corresponds to the number of combustion performance parameters output by the multimodal parameter coupling prediction system obtained in step one. Step 5, Prediction: Input the composition parameters of the solid propellant, use the combustion performance prediction model trained in step four to make predictions, and output the prediction results of the combustion performance parameters.

2. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 1, characterized in that, Step two specifically includes the following steps: Step 201: First, an experimental design is carried out for the three formulation factors of input parameters: oxidant particle size distribution, metal fuel surface oxide layer thickness, and mixing temperature. Based on the different values ​​of each formulation factor, the oxidant particle size distribution is given as level 10, the metal fuel surface oxide layer thickness is given as level 10, and the mixing temperature is given as level 5. For such a mixing design with different levels of each factor, a pseudo-level method is needed for adjustment. At the same time, the uniformity of the pseudo-level design is calculated by using the centralized L2 deviation formula. The final design table is determined based on the minimum deviation, and finally the design scheme table of each formulation is obtained. Step 202: Based on the design scheme table obtained in step 201, the particle size distribution is measured by laser diffraction, the oxide layer thickness is analyzed by XPS depth profiling, and the mixing temperature is measured by an online thermal imager. Sample data is obtained with oxidant particle size distribution, metal fuel surface oxide layer thickness and mixing temperature as input parameters, and combustion rate, pressure index and temperature sensitivity coefficient as output parameters.

3. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 2, characterized in that, In step 201, the formula for calculating the centered L2 deviation is: ; In the formula: Indicates the deviation value; Indicates the number of factors; Indicates the number of trials; Indicates the first The sequence number of each test site; Indicates the first Dimensional sequence number; Indicates the first The test site was at the [number]th [location]. Normalized coordinates of dimension.

4. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 1, characterized in that, In step three, the classification system is as follows: 。 5. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 1, characterized in that, Step four specifically includes the following steps: Step 401: Establish a neural network model that includes an input layer, at least three hidden layers, and an output layer; Step 402: Add a high-voltage compensation sub-network. The high-voltage compensation sub-network contains three hidden layers and is automatically activated when the oxidant content is detected to be >80%, thereby realizing an adaptive network architecture based on a dynamic branch structure. Step 403: Based on the adaptive network architecture obtained in step 402, network training is carried out. A two-stage training strategy is adopted. First, the basic network training compensation module is frozen and the learning rate is set to 1e-3. Then, joint fine-tuning is performed and the learning rate is set to 5e-5. The loss function adopts the dynamic weighted mean square error loss function. The training data is iteratively trained. When the error of the validation set loss function decreases by less than 1% for five consecutive rounds, the training is terminated to obtain a combustion performance prediction model that meets the accuracy requirements.

6. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 5, characterized in that, In step 402, the adaptive network architecture couples the basic neural network and the high-voltage compensation sub-network, and achieves adaptive weighting of the basic prediction and the high-voltage compensation output through learnable weight coefficients α.

7. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 6, characterized in that, In step 402, the adaptive weighted implementation equation for the basic prediction and high-voltage compensation output is as follows: ； ； In the formula: This represents the output parameter results of the adaptive network; Indicates the basic network weight coefficients; This represents the weight coefficient of the high-voltage compensation network; Indicates the output parameters of the basic network; This indicates the output parameter results of the high-voltage compensation network; These are the weighting coefficients; Indicates the percentage of oxidant content; It indicates pressure.

8. The method for predicting the combustion performance of solid propellants based on an AI model as described in claim 5, characterized in that, In step 403, the dynamically weighted mean squared error loss function is as follows: ； ； In the formula: Indicates the loss error; Indicates the number of samples; Indicates dynamic weights; This represents the training data values ​​for each output parameter; This represents the predicted data values ​​for each output parameter; This represents the truncation function; This represents the training data values ​​for each input parameter; 'a' represents the threshold; b represents the width of the transition interval.

Images