Method for determining physical property prediction model of aerospace kerosene, physical property prediction method and equipment

Abstract

The invention provides a method for determining a physical property prediction model of aerospace kerosene, a physical property prediction method and equipment. The determination method of the aerospace kerosene physical property prediction model comprises the steps that a structured data set with temperature and pressure as feature vectors and physical properties as target values is formed based on experimental data of aerospace kerosene physical properties, and physical property parameters comprise density, dynamic viscosity, heat conductivity and specific heat capacity at constant pressure; dividing the structured data set into a training set and a test set, and performing feature matrix standardization processing on feature vectors in the training set and the test set; training the selected at least two machine learning models based on the training set until convergence is achieved, and obtaining at least two prediction models; and evaluating the prediction capabilities of the at least two prediction models for different physical property parameters based on the test set, and determining the prediction models for different physical property parameters. According to the method, the prediction speed and prediction precision of the physical property parameters are effectively improved, and meanwhile, the engineering application requirements under wide working conditions are met.

Description

Technical Field

[0001] This invention relates to the field of aerospace propellant technology, and in particular to a method for determining a predictive model of aerospace kerosene properties, a property prediction method, and equipment. Background Technology

[0002] As a core component of rocket propulsion systems, liquid oxygen-kerosene engines directly determine the rocket's carrying capacity and flight reliability through key performance indicators such as thrust and specific impulse. As the core fuel for this type of engine, the accurate characterization and dynamic adaptation of the physical properties of aerospace kerosene are essential prerequisites for ensuring the engine's efficient and stable operation.

[0003] The core physical properties of aerospace kerosene mainly include density, dynamic viscosity, thermal conductivity, and specific heat capacity. Each parameter has a clear engineering correlation with engine performance: density directly determines the design size of the fuel tank, the amount of fuel carried, and the heat sink characteristics of the fuel itself, thus affecting the engine's endurance and thermal protection design; dynamic viscosity is a key factor affecting the selection of fuel delivery pipelines, pump matching, and combustion atomization effect. Unreasonable viscosity characteristics will lead to abnormal fuel delivery resistance and uneven atomization particle size, ultimately affecting combustion efficiency; thermal conductivity and specific heat capacity are directly related to the heat transfer efficiency and temperature field distribution in the engine combustion chamber, playing a decisive role in the thermal management design and operational stability of the combustion system.

[0004] Compared to ordinary kerosene such as aviation kerosene, aerospace kerosene has significant advantages in engineering applications, including a wider operating range and higher precision requirements for physical properties. Specifically, during rocket engine operation, aerospace kerosene must withstand wide temperature fluctuations from -30℃ to 400℃ and extreme pressure changes from 0.1 to 70 MPa. Furthermore, the precision requirements for testing these physical properties during engine design, simulation, and testing are far higher than those for ordinary civilian kerosene.

[0005] Currently, the acquisition of aerospace kerosene physical properties mainly relies on experimental measurement methods. These methods require specialized precision testing equipment such as high-precision vibrating tube densitometers and laser dynamic viscometers, involving repeated testing through multiple parallel experiments to obtain property data. However, this experimental method has inherent technical limitations: firstly, the testing process is time-consuming, with long cycles for single full-parameter tests, failing to meet the rapid acquisition requirements for physical properties in scenarios such as real-time engine simulation and dynamic optimization; secondly, the testing process consumes a large amount of high-purity aerospace kerosene samples, significantly increasing testing costs, and sample loss is difficult to effectively reduce through process optimization, hindering large-scale engineering applications.

[0006] To address the aforementioned shortcomings of experimental measurement methods, researchers attempted to obtain the physical properties of aerospace kerosene using traditional formula fitting methods. Specifically, this included employing classic models such as the DIPPR 801 density formula, the Sutherland dynamic viscosity formula, and the polynomial specific heat capacity formula, performing fitting calculations based on limited experimental data to achieve rapid estimation of physical properties. However, practice has shown that these traditional formula fitting methods have significant limitations in their applicability: they can only guarantee fitting accuracy within a narrow pressure range. Under extreme high-pressure conditions of 20 MPa and above, the fitting error generally exceeds 10%, failing to meet the precise requirements for physical properties during the high-pressure operation phase of engines. This severely restricts the application of this method in the design and simulation of engines under all operating conditions.

[0007] To meet the requirements for fitting physical parameters across a wide pressure range, existing technologies typically employ a segmented approach, dividing the entire pressure range into multiple sub-ranges and performing fitting calculations and parameter optimizations separately for each sub-range. However, this approach further increases the overall complexity of the fitting method. It not only requires significant additional effort for range division and boundary condition calibration but also suffers from discontinuous fitting results across ranges and large boundary transition errors, causing considerable inconvenience for practical engineering applications and failing to fundamentally address the need for high-precision, convenient acquisition of aerospace kerosene physical parameters across all operating conditions.

[0008] In summary, current methods for obtaining the physical properties of aerospace kerosene (experimental measurement and traditional formula fitting) all have insurmountable technical defects and cannot meet the actual needs of liquid oxygen kerosene engines for full-condition operation, real-time simulation optimization, and engineering applications. Therefore, developing a technology for obtaining the physical properties of aerospace kerosene that is adaptable to wide temperature and pressure conditions, has high fitting accuracy, and is easy to use has become an urgent technical problem to be solved in this field.

[0009] In view of this, the present invention is hereby proposed. Summary of the Invention

[0010] The technical problem to be solved by this invention is to overcome at least some of the shortcomings of the prior art and provide a method for determining a prediction model for the physical properties of aerospace kerosene. Based on experimentally measured physical property data of aerospace kerosene, at least two machine learning models are trained to obtain a corresponding number of candidate prediction models. Subsequently, the prediction capabilities of each candidate prediction model are systematically evaluated for different physical property parameters such as density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity. Finally, a dedicated prediction model adapted to different physical property parameters is selected and determined for the prediction of physical property parameters of aerospace coal. This method can effectively improve the prediction speed and accuracy of physical property parameters, while adapting to the engineering application requirements under a wide range of working conditions.

[0011] To address the aforementioned technical problems, a first aspect of the present invention provides a method for determining a predictive model for the physical properties of aerospace kerosene, comprising: Experimental data based on the physical properties of aerospace kerosene were used to form a structured dataset with [temperature, pressure] as feature vectors and physical properties as target values. The physical property parameters include density, dynamic viscosity, thermal conductivity, and specific heat capacity at constant pressure. The structured dataset is divided into a training set and a test set, and the feature vectors in the training set and the test set are subjected to feature matrix standardization. Based on the training set, at least two selected machine learning models are trained until convergence is achieved, resulting in at least two prediction models; and The predictive capabilities of the at least two prediction models for different physical property parameters are evaluated based on the test set, and prediction models for different physical property parameters are determined.

[0012] In some implementations, the step of performing feature matrix normalization on the feature vectors in the training and test sets includes: The original data of feature vectors in the training and test sets respectively X The distribution is transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1, resulting in standardized data. , represented as: Where υ is the original data based on the feature vectors in the training set. X The obtained mean, σ, is the original data based on the feature vectors in the training set. X The obtained standard deviation.

[0013] In some implementations, at least two machine learning models include at least two of the following: ridge regression model, random forest model, and extreme gradient boosting model.

[0014] In some implementations, the experimental data for the physical properties of aerospace kerosene ranges from -30 to 400°C and from 0.1 to 70 MPa.

[0015] In some implementations, the evaluation metrics for the prediction model include the coefficient of determination R. 2 The mean absolute percentage error (MAPE) is defined as follows: in, y i Indicates the first i The true value of each sample y i pred Indicates the first i The predicted value for each sample, y i avg This represents the average of all real samples.

[0016] In some implementations, the structured dataset is divided into training and test sets using a random (random_state=42) partitioning method, with 80% of the data in the structured dataset used as the training set and 20% used as the test set.

[0017] In some implementations, the types of aerospace kerosene in the experimental data on aerospace kerosene properties include synthetic components and components produced by the cracking of industrial feedstocks.

[0018] A second aspect of this invention is to provide a method for predicting the physical properties of aerospace kerosene. An input dataset with [temperature, pressure] as the feature vector is formed based on test data of aerospace kerosene. Perform feature matrix standardization on the input dataset; The standardized input dataset is then input into the prediction model for different physical property parameters determined according to the method described above for predicting the physical properties of aerospace kerosene. These physical property parameters include density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity. Output predicted values ​​for different physical property parameters.

[0019] A third aspect of the present invention is to provide a storage medium storing one or more programs that, when executed by a processor, implement the determination method of the aerospace kerosene property prediction model described above or the property prediction method of aerospace kerosene.

[0020] A fourth aspect of the present invention is to provide an electronic device comprising: Processor; and The memory is communicatively connected to the processor. The memory stores a program that can be executed by a processor. When the program is executed by the processor, the processor can execute the determination method of the aerospace kerosene property prediction model or the aerospace kerosene property prediction method described above.

[0021] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art.

[0022] (1) The method for determining the aerospace kerosene property prediction model provided by the present invention is based on the experimentally measured aerospace kerosene property data, training at least two machine learning models to obtain a corresponding number of candidate prediction models; then, for different physical property parameters such as density, dynamic viscosity, isobaric specific heat capacity and thermal conductivity, the prediction ability of each candidate prediction model is systematically evaluated; finally, a special prediction model adapted to different physical property parameters is selected and determined for the prediction of aerospace coal chemical physical property parameters. This method can effectively improve the prediction speed and accuracy of physical property parameters, and at the same time adapt to the engineering application requirements under a wide range of working conditions.

[0023] (2) The method for determining the prediction model of aerospace kerosene provided by the present invention selects at least two machine learning models from the ridge regression model, random forest model and extreme gradient boosting model for training and screening. This method can give full play to the performance advantages of each model. Through multi-model comparison and evaluation, the optimal model that is suitable for different physical property parameters is selected in a targeted manner, thereby greatly reducing the prediction error of each physical property parameter in the full working condition range and solving the core pain point of insufficient accuracy of traditional fitting methods under extreme working conditions.

[0024] (3) The method for determining the aerospace kerosene property prediction model provided by the present invention uses experimental data of artificially synthesized aerospace kerosene components and aerospace kerosene components produced by industrial raw material pyrolysis to train the machine learning model. By incorporating the property data of two different component systems, the prediction model obtained by training can simultaneously adapt to the property parameter prediction requirements of multi-component mixed aerospace kerosene produced by industrial pyrolysis process and single-component aerospace kerosene prepared by artificial synthesis or purification, comprehensively covering the typical component types of aerospace kerosene and greatly expanding the engineering application scope of the model. Attached Figure Description

[0025] The accompanying drawings, as part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation of the invention. Obviously, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without creative effort. In the drawings: Figure 1 This is a flowchart illustrating a method for determining a prediction model of aerospace kerosene properties according to an exemplary embodiment of the present invention. Figure 2 This is a flowchart illustrating a method for predicting the physical properties of aerospace kerosene according to an exemplary embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device provided according to an exemplary embodiment of the present invention.

[0026] In the picture: 600. Electronic device; 601. Processor; 602. Memory; 603. Bus; 604. Communication interface.

[0027] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0029] In the description of this invention, it should be noted that the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0030] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] As described in the background section, existing related technologies have the problem that the methods for obtaining the physical property parameters of aerospace kerosene cannot meet the actual needs of liquid oxygen kerosene engines for full-condition operation, real-time simulation optimization, and engineering applications.

[0032] Based on this, the present invention provides a method for determining a prediction model for the physical properties of aerospace kerosene, comprising: forming a structured dataset based on experimental data of aerospace kerosene physical properties, with [temperature, pressure] as feature vectors and physical properties as target values, wherein the physical property parameters include density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity; dividing the structured dataset into a training set and a test set, and performing feature matrix standardization processing; training at least two selected machine learning models based on the training set until convergence is achieved, thereby obtaining at least two prediction models; and evaluating the prediction capabilities of the at least two prediction models for different physical property parameters based on the test set, thereby determining the prediction model for different physical property parameters.

[0033] This scheme is based on experimentally measured aerospace kerosene physical property data. It trains at least two machine learning models to obtain a corresponding number of candidate prediction models. Then, it systematically evaluates the prediction capabilities of each candidate prediction model for different physical property parameters such as density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity. Finally, it selects and determines the exclusive prediction model that is suitable for different physical property parameters for the prediction of aerospace coal chemical physical property parameters. This method can effectively improve the prediction speed and accuracy of physical property parameters, while adapting to the engineering application requirements under a wide range of working conditions.

[0034] Figure 1 The flowchart illustrates a method for determining a prediction model of aerospace kerosene according to an exemplary embodiment of the present invention.

[0035] like Figure 1 As shown, the method for determining the predictive model of aerospace kerosene includes the following steps: S110. Experimental data based on the physical properties of aerospace kerosene are used to form a structured dataset with [temperature and pressure] as feature vectors and physical properties as target values. S120. Divide the structured dataset into a training set and a test set, and perform feature matrix standardization. S130. Train at least two selected machine learning models based on the training set until convergence is achieved, thereby obtaining at least two prediction models; and S140. Evaluate the predictive ability of the at least two prediction models for different physical property parameters based on the test set, and determine the prediction model for different physical property parameters.

[0036] It should be understood that the steps shown in the method for determining the predictive model of aerospace kerosene properties are not exclusive. The method for determining the predictive model of aerospace kerosene properties may also include additional steps not shown and / or the steps shown may be omitted. The scope of the present invention is not limited in this respect. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. In addition, unless explicitly defined or contradicted by the context, the specific steps included in the method described in the present invention are not limited to the order in which they are described, but can be performed in any order or in parallel. Refer below to Figure 1 Describe steps S110 to S140 in detail.

[0037] S110 In step S110, the experimental data is processed.

[0038] Specifically, thermophysical data acquisition experiments were conducted on three types of aerospace kerosene: kerosene A, kerosene B, and kerosene C, to obtain raw experimental data. Kerosene A was artificially synthesized, while kerosene B and kerosene C were derived from fossil fuel pyrolysis. The raw experimental data included temperature, pressure, and physical property parameters, including density. ρ Dynamic viscosity μ Thermal conductivity k Specific heat capacity at constant pressure C p .

[0039] The original experimental data were processed to obtain a structured dataset. The feature vectors of this structured dataset are temperature and pressure, i.e. X =[ T , p The target value is density. ρ Dynamic viscosity μ Thermal conductivity k Specific heat capacity at constant pressure C p One of them, namely Y = ρ or Y = μ or Y = k or Y = C p .

[0040] It should be noted that if the obtained structured dataset includes data with missing values ​​for temperature, pressure, or physical property parameters, such data needs to be removed. However, it is necessary to ensure that the structured dataset includes at least 200 complete sets of data, covering a temperature range of -30 to 400°C and a pressure range of 0.1 to 70 MPa.

[0041] S120 In step S120, the structured dataset is divided and the feature matrix is ​​standardized.

[0042] Specifically, the structured dataset is divided into training and test sets using a random (random_state=42) partitioning method, with 80% of the data in the structured dataset used as the training set and 20% used as the test set.

[0043] Feature vectors in the training and test sets X =[ T , p The steps for standardizing the feature matrix include: standardizing the feature vectors in the training and test sets respectively. X =[T , p The original data X The distribution is transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1, resulting in standardized data. This makes the distribution pattern of features uniform.

[0044] Specifically, it can be expressed as equation ①: ① Where υ is the original data based on the feature vectors in the training set. X The obtained mean, σ, is the original data based on the feature vectors in the training set. X The obtained standard deviation.

[0045] In the above scheme, the original data of the feature vectors in the training set... X Calculate the mean and standard deviation, then transform them into a standard normal distribution, while using the original data of the feature vectors of the test set. X The process of transforming to a standard normal distribution does not require reprocessing the original data of the feature vectors in the test set. X To calculate the mean and standard deviation, only the original data of the feature vectors in the training set is needed. X The mean and standard deviation calculated can be transformed into a standard normal distribution. This method is more robust to outliers because outliers have a much smaller impact on the mean and standard deviation than on extreme values, and will not over-compress the distribution of normal data.

[0046] S130 In step S130, at least two machine learning models are trained based on the data in the training set to obtain a corresponding number of candidate prediction models.

[0047] The aforementioned at least two machine learning models include at least two of the following: Ridge Regression (RR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). This can bring the following beneficial effects: First, it significantly improves the accuracy of physical property prediction, adapting to the needs of wide-range operating conditions. The Ridge Regression (RR) model has excellent anti-multicollinearity capabilities, effectively handling the linear correlation between different physical property parameters (density, dynamic viscosity, etc.) of aerospace kerosene and input features such as temperature and pressure, avoiding prediction bias caused by feature redundancy; the Random Forest (RF) model, based on the concept of ensemble learning, can accurately capture the complex nonlinear mapping relationship between input features and physical property parameters, adapting to the nonlinear variation law of physical property parameters of aerospace kerosene under wide temperature (-30℃~400℃) and wide pressure (0.1~70MPa) operating conditions; the Extreme Gradient Boosting (XGBoost) model, through gradient iterative optimization and regularization constraints, has both high-precision fitting and strong generalization capabilities, effectively reducing prediction errors under extreme operating conditions (such as high pressure of 20MPa and above). This invention selects at least two of the above models for training and screening, which can give full play to the performance advantages of each model. Through multi-model comparison and evaluation, the optimal model suitable for different physical property parameters can be selected in a targeted manner. For example, for physical property parameters with significant linear correlation, the ridge regression model can be selected, and for parameters with prominent nonlinear characteristics, the random forest or extreme gradient boosting model can be selected. This significantly reduces the prediction error of each physical property parameter in the entire working condition range and solves the core pain point of insufficient accuracy of traditional fitting methods under extreme working conditions.

[0048] Secondly, it ensures rapid prediction capabilities to meet the needs of real-time engineering simulation. Compared to the lengthy testing of traditional experimental methods and the iterative calculations of traditional formula fitting methods, the selected Ridge Regression, Random Forest, and Extreme Gradient Boosting models all possess efficient computational characteristics: the Ridge Regression model has a simple structure, short training and prediction time, and can quickly output physical property parameters; although the Random Forest and Extreme Gradient Boosting models are ensemble models, they can balance computational efficiency and fitting accuracy through reasonable parameter optimization, and can output prediction results instantly after training without complex piecewise calculations and iterative optimizations. The prediction model determined by this invention can significantly improve the prediction speed of physical property parameters while ensuring high accuracy, completely solving the problems of long time consumption and inability to meet the dynamic needs of real-time engine simulation of traditional methods, and providing strong support for rapid decision-making in engineering scenarios.

[0049] Third, it lowers the threshold for engineering applications and enhances the practicality and adaptability of the method. On the one hand, the three selected models are all mature and easy-to-implement machine learning models in the engineering field, which do not require complex hardware equipment and professional algorithm development capabilities. The training and deployment process is simple, which greatly reduces the difficulty of use for engineering technicians compared to the cumbersome operations such as segmented processing and repeated parameter optimization of traditional fitting methods. On the other hand, by selecting multiple models to adapt to different physical property parameters, it is not necessary to build a single model for all physical property parameters. The training model can be flexibly selected according to the actual needs of the engineering (such as focusing on the prediction of a certain type of physical property parameter or adapting to specific working conditions). This not only improves the flexibility of the method, but also ensures that the prediction results of each physical property parameter are adapted to the accuracy and speed requirements of engineering applications, avoiding the performance trade-off problem when a single model is adapted to multiple parameters.

[0050] Fourth, it improves the reliability and stability of prediction results and reduces risks in engineering applications. By selecting at least two models for training and evaluation, rather than relying on a single model, the limitations of a single model in predicting specific operating conditions and physical parameters can be effectively avoided—for example, the underfitting of the ridge regression model in strongly nonlinear scenarios, and the overfitting risk of the random forest model with small sample data. Through performance comparison and screening of multiple models, a more robust prediction model can be selected, ensuring that the prediction results of physical parameters remain stable and reliable under different operating conditions and different data samples. This reduces the risks in engineering aspects such as engine design and simulation optimization caused by prediction deviations, and provides a guarantee for the safe and efficient operation of rocket liquid oxygen kerosene engines.

[0051] Fifth, it reduces the cost of obtaining physical property parameters, achieving both economic efficiency and environmental friendliness. This invention trains a model based on existing experimental data. After determining the optimal prediction model, subsequent predictions of physical property parameters no longer require extensive testing with precision instruments and consumption of aerospace kerosene samples. Compared to traditional experimental methods, this significantly reduces sample loss and equipment investment costs. Simultaneously, the rapid prediction process reduces the time cost associated with testing cycles, achieving a balance between economic efficiency and environmental friendliness, further enhancing the feasibility of the method in large-scale engineering applications.

[0052] S140 In step S140, the predictive capabilities of each candidate prediction model are systematically evaluated using data from the test set for different physical property parameters such as density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity; finally, a specific prediction model suitable for different physical property parameters is selected.

[0053] Specifically, the evaluation metrics for prediction models include the coefficient of determination R. 2 The mean absolute percentage error (MAPE) is defined as shown in equations ② and ③, respectively: ② ③ in, y i Indicates the first i The true value of each sample y i pred Indicates the first i The predicted value for each sample, y i avg This represents the average of all real samples.

[0054] In the above scheme, the model evaluation index, the coefficient of determination R... 2 It is used to measure the ability of a regression model to explain the variation in the dependent variable. In Equation ②, the numerator is the residual sum of squares (SSE), which represents the unexplained variation in the model, and the denominator is the total sum of squares (SST), which represents the total variation in the target value. The range of the coefficient of determination R² is (-∞, 1]. The larger the value of the coefficient of determination R², the better the model fit.

[0055] Mean Absolute Percentage Error (MAPE) measures the relative accuracy of a model's predictions and directly reflects the proportion of prediction error. The range of MAPE is (0, ∞). A smaller MAPE value indicates a better model fit.

[0056] Table 1 shows the prediction results for aerospace kerosene density from different prediction models. ρ Calculation results As shown in Table 1, the density of aerospace kerosene... ρ It shows a linear correlation with temperature and pressure, and the coefficient of determination R 2 The values ​​for both the Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models are close to 1. For the Mean Absolute Percentage Error (MAPE), the values ​​for both models are close to 0, indicating that both models fit the trend well and have high accuracy, with the XGBoost model performing better. The Ridge Regression (RR) model, however, performs poorly with larger errors, especially regarding the density of kerosene A. ρ Fitting results.

[0057] In other words, when it is necessary to determine the density of aerospace kerosene... ρ When making predictions, the pre-trained Extreme Gradient Boosting (XGBoost) model should be used first, and the Random Forest (RF) model can be used as a fallback.

[0058] Table 2 shows the prediction results for the dynamic viscosity of aerospace kerosene using different prediction models. μ Calculation results The results in Table 2 show that the dynamic viscosity of aerospace kerosene is... μ The linear correlation with temperature and pressure is weak, and the coefficient of determination R0 is low. 2 The numerical values ​​for both the Random Forest (RF) model and the Extreme Gradient Boosting (XGBoost) model are close to 1. For the Mean Absolute Percentage Error (MAPE), the numerical values ​​for both models are close to 0, indicating that both can fit the dynamic viscosity of aerospace kerosene well. μ The relationship with temperature / pressure is highly accurate, with the Extreme Gradient Boosting (XGBoost) model performing better. Ridge Regression (RR) models, however, perform poorly with large errors. But the Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models are not limited to linear models.

[0059] In other words, when it is necessary to control the dynamic viscosity of aerospace kerosene... μ When making predictions, the pre-trained Extreme Gradient Boosting (XGBoost) model should be used first, and the Random Forest (RF) model can be used as a fallback.

[0060] Table 3 shows the prediction results for the thermal conductivity of aerospace kerosene using different prediction models. k Calculation results As shown in Table 3, the thermal conductivity of aerospace kerosene is... k It has a certain linear correlation with temperature and pressure, and the coefficient of determination R 2 The numerical values ​​for both the Random Forest (RF) model and the Extreme Gradient Boosting (XGBoost) model are close to 1. For the Mean Absolute Percentage Error (MAPE), the numerical values ​​for both models are close to 0, indicating that both can fit the thermal conductivity of aerospace kerosene well. k The relationship with temperature / pressure is highly accurate, with the Random Forest (RF) model performing better. The Ridge Regression (RR) model performs slightly worse with larger errors. However, the Random Forest (RF) model and the Extreme Gradient Boosting (XGBoost) model have strong generalization capabilities.

[0061] In other words, when the thermal conductivity of aerospace kerosene needs to be... k When making predictions, a pre-trained Random Forest (RF) model should be used first, and an Extreme Gradient Boosting (XGBoost) model can be used as a fallback.

[0062] Table 4 shows the predictions for the isobaric specific heat capacity of aerospace kerosene using different prediction models. C p Calculation results As shown in Table 4, the specific heat capacity of aerospace kerosene at constant pressure is...C p There is almost no linear correlation with temperature and pressure, and the coefficient of determination R0 shows no significant correlation. 2 The numerical values ​​for both the Random Forest (RF) model and the Extreme Gradient Boosting (XGBoost) model are close to 1. For the Mean Absolute Percentage Error (MAPE), the numerical values ​​for both models are close to 0, indicating that both can fit the thermal conductivity of aerospace kerosene well. C p The relationship with temperature / pressure is highly accurate, with the Extreme Gradient Boosting (XGBoost) model performing better. Ridge Regression (RR) models cannot fit the isobaric specific heat capacity of aerospace kerosene. C p The relationship with temperature / pressure. However, the Random Forest (RF) model and the Extreme Gradient Boosting (XGBoost) model have strong generalization capabilities.

[0063] In other words, when it is necessary to determine the specific heat capacity of aerospace kerosene at constant pressure... C p When making predictions, the pre-trained Extreme Gradient Boosting (XGBoost) model should be used first, and the Random Forest (RF) model can be used as a fallback.

[0064] Furthermore, this invention provides four comparative examples, which respectively compare the prediction models and fitting formulas determined by the above-described determination method for density. ρ Dynamic viscosity μ Thermal conductivity k Specific heat capacity at constant pressure C p Its predictive ability. Specifically: Comparative Example 1 Aerospace kerosene density was obtained by quadratic curve fitting. ρ The relationship between temperature and temperature is shown in equation ④ below: ④ Among them, the fitting parameters a =1062.5、 b 1 = -0.7957 b 2 = 1.189 × 10 -4 .

[0065] Table 5 shows the density of aerospace kerosene. ρ When making predictions, the mean absolute percentage error (MAPE) of the quadratic curve fitting method and the extreme gradient boosting (XGBoost) model is compared. As shown in Table 5, for the mean absolute percentage error (MAPE), the values ​​obtained by fitting the aerospace kerosene density at different temperatures using the extreme gradient boost (XGBoost) model are all smaller than those obtained by fitting using the quadratic curve method. Therefore, the fitting effect of the extreme gradient boost (XGBoost) model is generally better than that of the quadratic curve fitting method, indicating that the density prediction method of the extreme gradient boost (XGBoost) model is more accurate.

[0066] Comparative Example 2 Logarithmic form of dynamic viscosity of aerospace kerosene μ The relationship between temperature and pressure is shown in equation ⑤ below: ⑤ Among them, when the pressure is in the range of 0.3~40MPa (inclusive), the fitting parameters are... Z 0 = 1.19071 × 10 2 , Z 1 = -5.54439 × 10 3 , Z 2 = 2.12645 × 10 1 , Z 3 = 1.56939 × 10 -2 , E 0 = -1.00333 × 10 -1 , E 1 = 1.81220 × 10 1 , E 2 = 3.68335 × 10 -2 , A =-4.84325、 B =-2.69211×10 -10 .

[0067] Table 6 shows the effect of pressure range of 0.3~40MPa (inclusive) on the dynamic viscosity of aerospace kerosene. μ When making predictions, the mean absolute percentage error (MAPE) of the logarithmic fitting method and the extreme gradient boosting (XGBoost) model are compared. When the pressure is in the range of 40 (excluding) to 70 MPa, the fitting parameters are... Z 0 = 19.83459 × 10 1 , Z 1 = -1.47900 × 10 4 , Z 2 = 2.72896 × 10 1 , Z 3 = 3.06729 × 10 -1 , E0 = 6.05223 × 10 1 , E 1 = 1.81220 × 10 1 , E 2 = 3.68335 × 10 -2 , A =-4.84325、 B =-2.69211×10 -10 .

[0068] Table 7 shows the effect of pressure range of 40 (excluding) to 70 MPa on the dynamic viscosity of aerospace kerosene. μ When making predictions, the mean absolute percentage error (MAPE) of the logarithmic fitting method and the extreme gradient boosting (XGBoost) model are compared. As shown in Tables 6 and 7, for the mean absolute percentage error (MAPE), the values ​​obtained by fitting the dynamic viscosity of aerospace kerosene under different temperatures and pressures using the extreme gradient boost (XGBoost) model are all smaller than those obtained by fitting using the logarithmic curve. Therefore, the fitting effect of the extreme gradient boost (XGBoost) model on the dynamic viscosity of aerospace kerosene under different temperatures and pressures is significantly better than that of the piecewise logarithmic formula fitting method. This indicates that the dynamic viscosity prediction method of the extreme gradient boost (XGBoost) model is more accurate and convenient to apply, without the need for piecewise pressure processing.

[0069] Comparative Example 3 thermal conductivity k The fitting formula for the changes in temperature and pressure is shown in equation ⑥: ⑥ Where T is temperature in K, p is pressure in MPa, and the other symbols are fitting parameters, see Table 8.

[0070] Table 8 shows the fitting parameters for the thermal conductivity of aerospace kerosene. Table 9 shows the thermal conductivity of aerospace kerosene. k When making predictions, the mean absolute percentage error (MAPE) of the fitting formula method and the random forest (RF) model is compared. As shown in Table 9, for the mean absolute percentage error (MAPE), the random forest (RF) model was used to assess the thermal conductivity of aerospace kerosene under different temperatures and pressures. kThe values ​​obtained by fitting are all smaller than those obtained by fitting using the fitting formula. Therefore, the fitting effect of the Random Forest (RF) model is generally better than that of the formula fitting method, indicating that the Random Forest (RF) model has higher accuracy in predicting thermal conductivity.

[0071] Comparative Example 4 Isobaric specific heat capacity C p The relationship between temperature and temperature is shown in equation ⑦ below: ⑦ in, C p Specific heat capacity at constant pressure of the fluid, in kJ / (kg·K); T The temperature is given under the corresponding operating conditions, and the unit is K. a 1. a 2. a 3. a 4 represents the corresponding coefficients in the fitting formula. a 1 = 1.34552 a 2 = -0.00161 a 3 = 1.28833 × 10 -5 , a 4 = -7.95719 × 10 -9 .

[0072] Table 10 shows the specific heat capacity of aerospace kerosene at constant pressure. C p When making predictions, the mean absolute percentage error (MAPE) of the formula fitting method and the extreme gradient boosting (XGBoost) model are compared. As shown in Table 10, for the mean absolute percentage error (MAPE), the extreme gradient boosting (XGBoost) model was used to determine the isobaric specific heat capacity of aerospace kerosene at different temperatures. C p The values ​​obtained by fitting are all smaller than those obtained by fitting using the fitting formula. Therefore, the fitting effect of the extreme gradient boost (XGBoost) model is generally better than that of the formula fitting method, indicating that the thermal conductivity prediction method of the extreme gradient boost (XGBoost) model is more accurate.

[0073] Figure 2 The flowchart of a method for predicting the physical properties of aerospace kerosene according to an exemplary embodiment of the present invention is shown.

[0074] like Figure 2 As shown, the method for predicting the physical properties of aerospace kerosene includes the following steps: S210. Based on test data of aerospace kerosene, an input dataset is formed with [temperature, pressure] as the feature vector; S220. Perform feature matrix standardization on the input dataset; S230. Input the standardized input dataset into the prediction model for different physical property parameters; and S240: Output predicted values ​​for different physical property parameters.

[0075] Specifically, when it is necessary to obtain the physical properties (density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity) of aerospace kerosene under different operating conditions (temperature, pressure), an input dataset with [temperature, pressure] as the feature vector can be formed based on the temperature and pressure under different operating conditions in step S210. Then, in step S220, the input dataset is subjected to feature matrix standardization processing. It should be noted here that the process of performing feature matrix standardization processing on the input dataset can refer to the method used in the aforementioned step S120, and will not be repeated here.

[0076] In step S230, temperature and / or pressure are input into prediction models for different physical property parameters. For example, when the density, dynamic viscosity, or isobaric specific heat capacity of aerospace kerosene is required, the data from the input dataset is input into an Extreme Gradient Boosting (XGBoost) model, thereby outputting the density, dynamic viscosity, or isobaric specific heat capacity of aerospace kerosene under that operating condition in step S240. When the thermal conductivity of aerospace kerosene is required, the data from the input dataset is input into a Random Forest (RF) model, thereby outputting the thermal conductivity of aerospace kerosene under that operating condition in step S240.

[0077] Figure 3 The structure of an electronic device provided according to an exemplary embodiment of the present invention is shown.

[0078] like Figure 3 As shown, the electronic device 600 includes a processor 601 and a memory 602. The memory 602 is communicatively connected to the processor 601. The memory 602 stores programs executable by the processor. When the program is executed by the processor, the processor 601 can execute the aforementioned method for determining the physical property prediction model of aerospace kerosene or the method for predicting the physical properties of aerospace kerosene.

[0079] Figure 3 The electronic device shown also includes a bus 603 and a communication interface 604. The processor 601, the communication interface 604, and the memory 602 are connected via the bus 603.

[0080] The memory 602 may include high-speed random access memory (RAM), or it may also include non-volatile memory 602, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 604 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 603 may be an ISA bus, PCI bus, or EISA bus, etc. The bus 603 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus 603 or one type of bus 603.

[0081] Processor 601 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 601 or by software instructions. Processor 601 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor, or processor 601 can be any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 602. The processor 601 reads the information in memory 602 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0082] An exemplary embodiment of the present invention also provides a computer-readable storage medium storing a computer program. When the computer program is called and executed by the processor 601, the computer-executable instructions cause the processor 601 to implement the above-mentioned method for determining the physical property prediction model of aerospace kerosene or the method for predicting the physical properties of aerospace kerosene. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0083] The computer program product of the method for determining the physical property prediction model of aerospace kerosene or the method for predicting the physical property of aerospace kerosene provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0084] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and / or electronic equipment can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for determining a prediction model for the physical properties of aerospace kerosene, characterized in that, include: Experimental data based on the physical properties of aerospace kerosene were used to form a structured dataset with [temperature, pressure] as feature vectors and physical properties as target values. The physical property parameters include density, dynamic viscosity, thermal conductivity, and specific heat capacity at constant pressure. The structured dataset is divided into a training set and a test set, and the feature vectors in the training set and the test set are subjected to feature matrix standardization. Based on the training set, at least two selected machine learning models are trained until convergence is achieved, resulting in at least two prediction models; and The predictive capabilities of the at least two prediction models for different physical property parameters are evaluated based on the test set, and prediction models for different physical property parameters are determined.

2. The method for determining the prediction model of aerospace kerosene according to claim 1, characterized in that, The steps for standardizing the feature matrix of the feature vectors in the training and test sets include: The original data of feature vectors in the training and test sets respectively X The distribution is transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1, resulting in standardized data. , represented as: Where υ is the original data based on the feature vectors in the training set. X The obtained mean, σ, is the original data based on the feature vectors in the training set. X The obtained standard deviation.

3. The method for determining the prediction model of aerospace kerosene according to claim 1, characterized in that, At least two machine learning models are required, including at least two of the following: Ridge Regression, Random Forest, and Extreme Gradient Boosting.

4. The method for determining the prediction model of aerospace kerosene according to claim 1, characterized in that, The experimental data on the physical properties of aerospace kerosene ranges from -30 to 400°C and from 0.1 to 70 MPa.

5. The method for determining the prediction model of aerospace kerosene according to any one of claims 1 to 4, characterized in that, Evaluation metrics for predictive models include the coefficient of determination (R²). 2 The mean absolute percentage error (MAPE) is defined as follows: in, y i Indicates the first i The true value of each sample y i pred Indicates the first i The predicted value for each sample, y i avg This represents the average of all real samples.

6. The method for determining the prediction model of aerospace kerosene according to any one of claims 1 to 4, characterized in that, The structured dataset was divided into training and test sets using a random (random_state=42) partitioning method, with 80% of the data in the structured dataset used as the training set and 20% used as the test set.

7. The method for determining the prediction model of aerospace kerosene according to any one of claims 1 to 4, characterized in that, The experimental data on the physical properties of aerospace kerosene include both synthetically produced components and components generated from the cracking of industrial raw materials.

8. A method for predicting the physical properties of aerospace kerosene, characterized in that, An input dataset with [temperature, pressure] as the feature vector is formed based on test data of aerospace kerosene. Perform feature matrix standardization on the input dataset; The standardized input dataset is input into the prediction model for different physical property parameters determined by the method for determining the prediction model of aerospace kerosene according to any one of claims 1 to 7, wherein the physical property parameters include density, dynamic viscosity, specific heat capacity at constant pressure, and thermal conductivity; and Output predicted values ​​for different physical property parameters.

9. A storage medium, characterized in that, The storage medium stores one or more programs, which, when executed by a processor, implement the method for determining the aerospace kerosene property prediction model according to any one of claims 1-7 or the aerospace kerosene property prediction method according to claim 8.

10. An electronic device, characterized in that, include: processor; as well as The memory is communicatively connected to the processor. The memory stores a program that can be executed by a processor. When the program is executed by the processor, the processor is able to execute the method for determining the physical property prediction model of aerospace kerosene according to any one of claims 1 to 7 or the method for predicting the physical properties of aerospace kerosene according to claim 8.

Images