Combustion characteristic rapid prediction method suitable for sustainable aviation fuel

Abstract

The invention relates to the technical field of aviation fuel, and discloses a combustion characteristic rapid prediction method suitable for sustainable aviation fuel, comprising the following steps: step 1, acquiring fuel molecules of the sustainable aviation fuel and fuel composition data thereof; two-dimensional structure data and combustion test data of the fuel molecules are obtained, and then a fuel molecule database is formed; 2, on the basis of the two-dimensional structure data, a quantum chemistry calculation method is adopted, and basic reaction descriptors of corresponding hydrocarbon molecules are obtained; based on combustion test data, screening to obtain a physical descriptor; 3, a correlation analysis method is adopted, and target descriptors are obtained through screening; step 4, performing multi-model prediction based on the target descriptor in combination with combustion conditions; and calculating the accuracy index of each model, and screening to obtain an optimal model to predict the combustion characteristics. According to the method, the combustion characteristics of the sustainable aviation fuel can be accurately and efficiently predicted, and the prediction reliability is high.

Description

Technical Field

[0001] The invention relates to the technical field of aviation fuel, and in particular to a method for quickly predicting combustion characteristics applicable to sustainable aviation fuel. Background Art

[0002] Aviation fuel is currently mainly jet fuel made from petroleum. The rapid development of my country's civil aviation industry has led to a rapid increase in demand for aviation fuel. Therefore, the development and use of sustainable aviation fuel has become a major energy strategic issue facing my country's civil aviation and national defense.

[0003] Aviation fuel is a special oil product that provides power for aircraft engines. Its composition and performance must be certified for airworthiness and meet the model design and ground storage and transportation requirements of aircraft and engines before it can be put into use in the civil aviation market. A set of mature airworthiness certification methods for traditional petroleum-based aviation fuels have been established at home and abroad. However, for sustainable aviation fuels, which have significant differences in composition and production process from traditional petroleum-based aviation fuels, the current main reliance is on full-process airworthiness certification methods with few testing resources, complex operations, long cycles and expensive testing costs, which seriously restricts the use of sustainable aviation fuels in the civil aviation market. This has become a technical barrier affecting the innovative development of my country's civil aviation energy. Therefore, it is a major strategic need in the field of airworthiness certification for my country's civil aviation to develop efficient, economical and fast new airworthiness certification paths and methods to solve the problems of my country's sustainable aviation fuel airworthiness certification being subject to foreign testing resource monopoly and standard restrictions.

[0004] In addition to conducting basic physical and chemical property research on sustainable aviation fuels, the airworthiness certification also requires research on their combustion properties. This part of the research currently relies mainly on experimental research with complex operations, long cycles and relatively expensive test costs. In recent years, with the rapid development of combustion chemical kinetic model research, through component analysis and experimental research on different aviation fuels, detailed combustion chemical kinetic models of different compounds have been established, and kinetic simulation and prediction research on aviation fuel combustion has become a mature and mainstream method in combustion research. However, the main defect of this method is that the development and optimization of detailed combustion chemical kinetic models are highly dependent on the accuracy of the thermodynamic and chemical kinetic parameters they contain. However, the thermal and kinetic parameters currently obtained by high-precision quantum chemical calculation methods are relatively limited, and most of them rely on group contribution methods and analogy methods with low accuracy. As a result, the simulation accuracy of the constructed detailed combustion chemical kinetic model is not high and relies on the verification of experimental data. In addition, due to differences in processes and raw materials, the fuel components prepared by different manufacturers of sustainable aviation fuels vary greatly, which also brings huge challenges to the rapid construction of combustion kinetic models.

[0005] On the other hand, with the rapid development of computer technology and the continuous improvement of hydrocarbon combustion property databases in recent years, many researchers have begun to focus on machine learning methods for the prediction of sustainable aviation fuel combustion properties. In traditional research, the rapid and accurate prediction of the physical and chemical properties of molecules by establishing quantitative structure-activity relationships or models based on machine learning methods has always been the focus of research in disciplines such as chemistry and materials. With the continuous expansion of the research scope of the physical and chemical properties of various fuel molecules, many researchers have studied the combustion characteristics of the fuel molecules of interest by developing quantitative structure-activity relationships or based on machine learning methods. However, the current research objects at home and abroad are still focused on the description of a certain basic physical and chemical property of molecules with specific functions, especially for fuel molecules, the focus is on basic physical and chemical properties such as flash point, density, and freezing point, and the fitting or machine learning methods used are relatively simple. Although machine learning methods have made great progress in predicting the basic physical and chemical properties of hydrocarbons contained in aviation fuel, diesel and gasoline, research on combustion properties is still in the initial exploration stage.

[0006] It can be seen that the existing combustion characteristics analysis methods that can be used for sustainable aviation fuels all have their own defects, and there is still an urgent need for a method that can accurately and efficiently analyze the combustion characteristics of sustainable aviation fuels. Summary of the invention

[0007] The present invention aims to provide a method for quickly predicting the combustion characteristics of sustainable aviation fuel, which can accurately and efficiently predict the combustion characteristics of sustainable aviation fuel with high prediction reliability.

[0008] The basic scheme provided by the present invention is: a method for rapid prediction of combustion characteristics applicable to sustainable aviation fuel, comprising the following steps:

[0009] Step 1, obtaining fuel molecules of sustainable aviation fuel and fuel composition data thereof; the fuel molecules are hydrocarbon molecules; the fuel composition data include chemical structure characteristics and mass fraction distribution information of the fuel molecules; based on the fuel composition data, obtaining two-dimensional structure data of the fuel molecules, and obtaining combustion test data of the fuel molecules, thereby forming a fuel molecule database;

[0010] Step 2: Based on the two-dimensional structure data, a quantum chemical calculation method is used to obtain the basic reaction descriptors of the corresponding hydrocarbon molecules; based on the combustion test data, the physical descriptors are screened;

[0011] Step 3, based on the basic reaction descriptors and physical descriptors, a correlation analysis method is used to screen and obtain the target descriptors;

[0012] Step 4: Based on the target descriptor and in combination with the combustion conditions, multi-model prediction is performed; the multi-model includes at least two machine learning models; and the accuracy index of each model is calculated to screen out the optimal model for predicting the combustion characteristics.

[0013] Furthermore, the combustion test data includes high temperature combustion conditions and test results corresponding to fuel molecules of sustainable aviation fuel; the fuel molecules are hydrocarbon molecules; the high temperature combustion conditions include initial combustion temperature, initial pressure and chemical equivalence ratio.

[0014] Furthermore, the chemical equivalence ratio is the ratio of fuel molecules to oxygen entering the combustion chamber of the aircraft engine, and the value is 0.5-2.0.

[0015] Furthermore, the basic reaction descriptors include highest occupied molecular orbital energy, lowest unoccupied molecular orbital energy, core-core repulsion energy, ionization potential, solvent surface area and volume descriptors.

[0016] Further, the physical descriptors include molecular weight, initial combustion temperature, initial pressure and chemical equivalence ratio of the fuel molecules.

[0017] Furthermore, when screening the target descriptor, the following steps are included:

[0018] calculating a correlation coefficient r between descriptors, the descriptors including a basic reaction descriptor and a physical descriptor;

[0019]

[0020] Where n is the number of fuel molecules in the fuel molecule database, that is, the number of hydrocarbon molecules; and Represent the average value of descriptor X and the average value of descriptor Y, σ X and σ Y are the standard deviation of descriptor X and the standard deviation of descriptor Y, respectively.

[0021] Furthermore, when the r value between two descriptors is greater than 0.8, one of the descriptors is selected according to the preset selection criteria; and the selected descriptor is evaluated according to the preset relationship:

[0022] The preset screening criteria include: calculating the average value between a descriptor and other descriptors, deleting the descriptor with a larger average value, and then obtaining the required descriptor;

[0023] The preset relationship is that the number of descriptors N to be screened must satisfy

[0024] Furthermore, the target descriptor includes a minimum empty orbit energy and a physical descriptor.

[0025] Furthermore, the quantum chemical method is the PM6 quantum chemical method.

[0026] Furthermore, the two-dimensional structure data includes laminar flame propagation velocity of hydrocarbon molecules and corresponding topological structure information.

[0027] The working principle and advantages of the present invention are:

[0028] The invention provides a rapid prediction method for the combustion characteristics of sustainable aviation fuel, which can accurately and efficiently predict the combustion characteristics of sustainable aviation fuel with high prediction reliability; it can provide a new idea for constructing descriptors suitable for predicting the reactivity of substances such as combustion properties, and can effectively solve the problem of lack of efficient and accurate prediction model support for the prediction of the combustion properties of hydrocarbons in sustainable aviation fuel. The key points are:

[0029] First, this scheme first conducts a detailed chemical component analysis of sustainable aviation fuels and specifically constructs a hydrocarbon database that can be used for airworthiness certification. On this basis, combustion test data related to combustion tests are introduced to specifically construct a specific combustion database for sustainable aviation fuels. Secondly, based on the database, descriptors are analyzed and extracted, and the PM6 quantum chemical method with high computational efficiency, that is, the semi-empirical quantum chemical calculation method, is used to obtain basic reaction descriptors, which can improve the computational efficiency of big data analysis and more effectively describe the combustion properties related to molecular activity. Compared with existing property prediction schemes, most of the existing schemes directly select structural descriptors according to experience (such as the patent with publication number CN101339180B--a method for predicting the combustion and explosion characteristics of organic compounds based on support vector machines) or directly quote existing structural descriptor data sets (such as the patent with publication number CN116386748A--a method, device and storage medium for predicting biomass pyrolysis efficiency based on machine learning). Both of these cannot actually break away from the limitations of human subjective conjecture and are easy to affect the objectivity and reliability of the overall analysis.

[0030] Moreover, most of the existing structural descriptor data describe molecular structure information or component information. For ordinary material analysis, this descriptor data is sufficient; however, for special sustainable aviation fuels, the practical difficulty is that the fuel composition varies greatly and there are a large number of isomers, which leads to duplication and similarity in the basic information of some molecular structures or related descriptors, making it impossible to accurately describe the combustion reaction activity of these fuel molecules; in addition, the combustion environment of aviation fuel is highly specific, and it is difficult to accurately describe its reaction activity related to extreme combustion environments such as high temperature and pressure using only molecular structure descriptors, and it is difficult to consider the influence of specific combustion conditions such as temperature, pressure and chemical equivalence ratio. Therefore, if we follow the existing ideas, it is unreliable to continue to use traditional structural descriptors. On the other hand, current experimental research on the combustion characteristics of sustainable aviation fuel is constrained by the high cost and long experiment cycle. The experimental data contained in its database is relatively limited, resulting in a relatively small database. As a result, traditional machine learning methods based on molecular structure information descriptors face certain shortcomings in this regard, namely, the prediction accuracy for small databases is low. However, by introducing reaction descriptors based on quantum chemical calculations, it has been proven that it has better advantages in machine learning for small database samples related to describing reaction activity, and can obtain machine learning models with higher prediction accuracy.

[0031] In view of this, this scheme has created a new database based on molecular structure and component information, but it does not use them directly as descriptors. Instead, it uses them to extract descriptors based on them, using a quantifiable quantum chemical calculation method, and innovatively uses reaction activity as the associated target. Here, compared with the existing few analytical methods that also apply quantum chemical analysis (such as the patent with publication number CN108763855B - a method for determining the combustion performance of biodiesel), this type of scheme is still limited to the analysis of molecular structure and component information. Although the application of quantum chemical analysis realizes parameterized description, it cannot change the defects of the descriptor itself. This scheme fundamentally changes the selection principle of descriptors, intends to select reaction descriptors, and uses the reaction activity that has an objective and stable association with the combustion properties as the selection basis. It has a strong physical and chemical significance, and can maintain strong consistency under the conditions of large differences in fuel composition and strong particularity of the fuel combustion environment.

[0032] In addition, what is special is that when extracting descriptors, this scheme does not rely on any subjective speculation or experimental verification, but is based on the frontline molecular orbital theory, comprehensively considering that aviation fuel is mainly liquid phase molecules, and extracts the highest occupied orbital energy, lowest unoccupied orbital energy, nucleus-nucleus repulsion energy, ionization potential energy, solvent surface area and volume as basic reaction descriptors, which has strong objectivity and reliability.

[0033] Furthermore, this scheme also introduces physical descriptors based on combustion test data to further enhance the physical and chemical significance of the descriptors, so that they can fully describe the combustion properties of sustainable aviation fuel, a special fuel, in the specific environment of aviation from multiple dimensions. Moreover, compared with existing property prediction schemes, although some schemes have experimental descriptors (such as the patent with publication number CN117789889A - a method, system and device for screening metal organic framework materials), the description of existing schemes stops at conventional environmental factors such as temperature and pressure. In conventional analysis, this has constituted sufficient environmental conditions for combustion side reactions, so the existing schemes basically follow this setting. However, after research, this scheme found that in a specific aviation environment, taking the common application scenario of sustainable aviation fuel - aircraft engine combustion chamber as an example, inside the aircraft engine combustion chamber, due to complex environmental characteristics such as spray and air flow process, not only are there large differences in combustion temperature and pressure in different areas, but also in chemical equivalence ratio. Based on this finding, this scheme selects laminar flame propagation velocity with a wide range of chemical equivalence ratios as the target parameter, and uses temperature, pressure, and chemical equivalence ratio as physical descriptors of combustion conditions, and then considers molecular weight as an independent physical descriptor. This can fully consider the different combustion performance characterizations that may exist in a specific aviation environment, and can significantly reduce the simulation time compared with traditional simulation methods based on detailed combustion kinetics models.

[0034] Finally, it is combined with the six basic reaction descriptors as the descriptors for subsequent machine learning analysis. From the characteristics and total amount of the constructed descriptors, the descriptors introduced in this scheme have strong physical and chemical significance. At the same time, the descriptors are relatively refined, and the descriptors corresponding to the combustion conditions are highly independent from the reaction descriptors in a physical sense, which provides a simpler input for the dimensional simplification and independence research of the subsequent correlation analysis.

[0035] Second, based on the basic reaction descriptors and physical descriptors, this solution further conducted a correlation analysis, selected the target reaction descriptors, and achieved dimensional simplification of the descriptors, which can effectively avoid the high-dimensionality problem caused by excessive overlap of descriptors faced by subsequent machine learning models, and help improve the effect of subsequent machine learning. Compared with existing solutions (such as the existing solutions mentioned above - patent CN116386748, patent CN108763855B), although these solutions involve similar correlation analysis, they all ignore the determination of independence and the dimensional simplification is weak. This solution sets an efficient independence evaluation method, which can obtain the most independent descriptors, ensure that the selected descriptors and the corresponding combustion properties have a corresponding and clear mathematical relationship, which helps to improve the learning effect of subsequent machine learning models.

[0036] Third, compared with traditional experiments or combustion dynamics simulation methods, these existing methods often take months or weeks to obtain relevant combustion characteristics. However, this method can obtain high-precision prediction results on a common computing platform with a time scale of seconds, saving the complex, long-cycle and expensive combustion performance testing process faced by the airworthiness certification of sustainable aviation fuels, and can achieve rapid and efficient evaluation of the combustion properties of the main components of sustainable aviation fuels. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of a method flow chart of an embodiment of a method for rapid prediction of combustion characteristics applicable to sustainable aviation fuels according to the present invention;

[0038] Figure 2 A schematic diagram of the correlation analysis results of basic reaction descriptors and physical descriptors of an embodiment of a method for rapid prediction of combustion characteristics of sustainable aviation fuels according to the present invention;

[0039] Figure 3 The results of typical straight-chain alkanes (n-decane) of sustainable aviation fuel predicted by the optimal model screened out in an embodiment of a method for rapid prediction of combustion characteristics of sustainable aviation fuel of the present invention (the symbols represent experimental measurement results);

[0040] Figure 4 The results of typical branched alkanes (iso-octane) in sustainable aviation fuel predicted by the optimal model screened out in an embodiment of a method for rapid prediction of combustion characteristics applicable to sustainable aviation fuel of the present invention (the symbols represent experimental measurement results). DETAILED DESCRIPTION

[0041] The following is a further detailed description through specific implementation methods:

[0042] The embodiment is basically as shown in the attached Figure 1 As shown: A method for rapid prediction of combustion characteristics applicable to sustainable aviation fuels, comprising the following steps:

[0043] In this embodiment, the steps are described by taking the laminar flame propagation speed of sustainable aviation fuel as an example of the combustion characteristics.

[0044] Step 1, obtaining fuel molecules of sustainable aviation fuel and fuel composition data thereof; the fuel molecules are hydrocarbon molecules; the fuel composition data include chemical structure characteristics and mass fraction distribution information of the fuel molecules. Specifically, the main chemical composition of sustainable aviation fuel is analyzed by gas chromatography and two-dimensional gas chromatography, and the fuel composition data is obtained.

[0045] Based on the fuel composition data, the two-dimensional structure data of the fuel molecules and the combustion test data of the fuel molecules are obtained, thereby forming a fuel molecule database. The two-dimensional structure data includes the laminar flame propagation velocity of hydrocarbon molecules and the corresponding topological structure information.

[0046] The combustion test data includes high-temperature combustion conditions and test results corresponding to the fuel molecules of sustainable aviation fuel. In this embodiment, a total of 46 fuel molecules (i.e., hydrocarbon molecules) such as n-decane, methylcyclohexane, and toluene that are more common in aviation fuels are included, and 4,664 corresponding test results. The high-temperature combustion conditions include initial combustion temperature, initial pressure, and chemical equivalence ratio. The initial combustion temperature and initial pressure here refer to the initial temperature and initial pressure of the fuel molecules entering the combustion chamber; the chemical equivalence ratio is specifically the ratio of fuel molecules to oxygen entering the combustion chamber of the aircraft engine combustion chamber, and the value is 0.5-2.0.

[0047] Here, this scheme specifically introduces specific test parameters, and compared with conventional combustion characteristics analysis schemes, although a few existing schemes set experimental descriptors to introduce experimental parameters, the existing schemes only consider two parameters, temperature and pressure. However, after research, this scheme found that in a specific aviation environment, here taking the common application scenario of sustainable aviation fuel - the aircraft engine combustion chamber as an example, inside the aircraft engine combustion chamber, due to the complex environmental characteristics such as spray and air flow process, not only the combustion temperature and pressure in different areas are very different, but also the chemical equivalence ratio is very different. Based on this discovery, this scheme collects and analyzes the wide chemical equivalence ratio involved in the aircraft engine combustion chamber, which can fully consider the different combustion performance characteristics that may exist in a specific aviation environment.

[0048] Step 2: Based on the two-dimensional structure data, quantum chemical calculation methods are used to obtain the basic reaction descriptors of the corresponding hydrocarbon molecules; based on the combustion test data, physical descriptors are screened and obtained.

[0049] The quantum chemical method is the PM6 quantum chemical method, which is a semi-empirical quantum chemical method mainly used for the optimization of molecular structure, the study of reaction paths and the prediction of various physical and chemical properties. Compared with the ab initio method, PM6 simplifies the electronic structure calculation by introducing empirical parameters, thereby greatly reducing the demand for computing time and resources, achieving higher computing efficiency while maintaining relatively high accuracy.

[0050] The basic reaction descriptors include highest occupied orbital energy, lowest unoccupied orbital energy, core-core repulsion energy, ionization potential, solvent surface area, and volume descriptors.

[0051] The physical descriptors include the molecular weight of the fuel molecules, the initial combustion temperature, the initial pressure, and the chemical equivalence ratio.

[0052] Here, this scheme introduces reaction descriptors and physical descriptors that are not related to the intrinsic structure of fuel molecules, which fundamentally changes the selection principle of descriptors. The reaction descriptors selected by the former are selected based on the reaction activity that has an objective and stable correlation with the combustion properties. Under the conditions of large differences in fuel composition and strong particularity of the fuel combustion environment, they can maintain strong consistency. The physical descriptors selected by the latter can fully consider the different combustion performance representations that may exist in a specific aviation environment. The combination of the two has a strong physical and chemical significance.

[0053] Step 3: Based on the basic reaction descriptors and physical descriptors, a correlation analysis method is used to screen and obtain the target descriptors.

[0054] Here, in order to avoid the high-dimensionality problem caused by excessive overlap of descriptors faced by subsequent machine learning models, this embodiment chooses to analyze the correlation of descriptors. In addition, in view of the number and characteristics of descriptors considered in this embodiment, especially the characteristics of the strong normal distribution of the target parameters laminar flame speed and chemical equivalence ratio, this embodiment uses the Pearson correlation analysis method that is more suitable for this type of data to carry out dimensional simplification and independence analysis on the descriptors.

[0055] Specifically, when screening the target descriptor, the following steps are included:

[0056] calculating a correlation coefficient r between descriptors, the descriptors including a basic reaction descriptor and a physical descriptor;

[0057]

[0058] in, and Represent the average value of descriptor X and the average value of descriptor Y, σ X and σ Y are the standard deviation of descriptor X and the standard deviation of descriptor Y, respectively.

[0059] When the r value between two descriptors is greater than 0.8, it indicates that there is a strong correlation between them, and then one of the descriptors is screened out according to the preset screening criteria (that is, one of them is deleted); and the screened descriptors are evaluated according to the preset relationship.

[0060] The preset screening criteria include: calculating the average value between a descriptor and other descriptors, deleting the descriptor with a larger average value, and then obtaining the required descriptor.

[0061] In addition, in order to ensure that there is a corresponding and clear mathematical relationship between the screened descriptors and the corresponding combustion properties, this embodiment also uses a preset relationship to further evaluate the number of screened descriptors (N) to ensure the rationality and reliability of the descriptors, so as to facilitate the subsequent parameter optimization and model prediction of the machine learning model.

[0062] The preset relationship includes:

[0063] The number of descriptors N to be filtered must satisfy

[0064] In the above formula, n is the number of fuel molecules in the fuel molecule database, that is, the number of hydrocarbon molecules.

[0065] According to the above method, if Figure 2 As shown in the figure, the results of Pearson correlation analysis show that there is a strong independence relationship between the lowest unoccupied orbital energy calculated by quantum chemistry and other descriptors. At the same time, according to the frontier molecular orbital theory, the lowest unoccupied orbital energy is also a key parameter for describing the reactivity of molecules. Therefore, it is suitable as the input layer of the machine learning algorithm. On the other hand, combustion conditions are necessary inputs for machine learning to predict the target laminar flame speed. Therefore, combustion conditions and the lowest unoccupied orbital energy of molecules are used as input descriptors for parameter optimization of machine learning methods. That is, the target descriptor includes the lowest unoccupied orbital energy and physical descriptors.

[0066] Step 4: Based on the target descriptor and in combination with the combustion conditions, multi-model prediction is performed; the multi-model includes at least two machine learning models; and the accuracy index of each model is calculated to screen out the optimal model for predicting the combustion characteristics.

[0067] In this embodiment, the multi-model includes a degenerate tree model, a support vector machine model, a Gaussian process regression model and an integrated tree model. In practical applications, other machine learning models can also be introduced to participate in the analysis.

[0068] Specifically, the target descriptor is used as the input descriptor for parameter optimization of the machine learning method, the database is used to carry out parameter optimization, the machine learning method of each model is established, and the accuracy index of each model is calculated.

[0069] The accuracy indicators include mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R 2 ).

[0070] Specifically, the four indicators are defined as follows:

[0071]

[0072] Among them, S i,L is the experimental value, is the prediction result of the machine learning method, is the average experimental result of the test data set. In this embodiment, through the systematic analysis of the above machine learning error evaluation indicators, it is finally shown that the Gaussian process regression method is the optimal machine learning algorithm, and the corresponding four error evaluation indicators are 13.67, 3.70, 2.36 and 0.99 respectively.

[0073] In addition, this example further verifies the laminar flame speed simulation results of n-decane and isooctane, the main alternative components of sustainable aviation fuel. Figure 3 and Figure 4 The diagrams are respectively the comparison diagrams of the optimal model screened and the experimental results, indicating that the machine learning method obtained in this embodiment has good simulation accuracy for the key combustion characteristic of laminar flame speed.

[0074] The present embodiment provides a method for rapid prediction of combustion characteristics applicable to sustainable aviation fuel, which can accurately and efficiently predict the combustion characteristics of sustainable aviation fuel with high prediction reliability; it can provide new ideas for constructing descriptors applicable to prediction of the reactivity of substances such as combustion properties, and can effectively solve the problem of lack of efficient and accurate prediction model support for the prediction of combustion properties of hydrocarbons in sustainable aviation fuel, and can effectively solve the problem of high cost and long cycle of sustainable aviation fuel combustion property experiments, thereby improving the airworthiness certification capability of sustainable aviation fuel and shortening the airworthiness certification cycle.

[0075] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field are aware of all the common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement the scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for the ordinary technicians in the relevant field to implement this application. It should be pointed out that for the technicians in this field, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for rapid prediction of combustion characteristics of sustainable aviation fuel, characterized in that: The following steps are involved: Step 1, obtaining fuel molecules of sustainable aviation fuel and fuel composition data thereof; the fuel molecules are hydrocarbon molecules; the fuel composition data include chemical structure characteristics and mass fraction distribution information of the fuel molecules; based on the fuel composition data, obtaining two-dimensional structure data of the fuel molecules, and obtaining combustion test data of the fuel molecules, thereby forming a fuel molecule database; Step 2, based on the two-dimensional structure data, using quantum chemical calculation methods to obtain the basic reaction descriptors of the corresponding hydrocarbon molecules; Based on the combustion test data, physical descriptors were screened; Step 3, based on the basic reaction descriptors and physical descriptors, a correlation analysis method is used to screen and obtain the target descriptors; Step 4: Based on the target descriptor and in combination with the combustion conditions, multi-model prediction is performed; the multi-model includes at least two machine learning models; and the accuracy index of each model is calculated to screen out the optimal model for predicting the combustion characteristics.

2. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: The combustion test data includes high temperature combustion conditions and test results corresponding to the fuel molecules; the high temperature combustion conditions include initial combustion temperature, initial pressure and chemical equivalence ratio.

3. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 2, characterized in that: The chemical equivalence ratio is the ratio of fuel molecules to oxygen entering the combustion chamber of an aircraft engine, and the value is 0.5-2.

0.

4. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: The basic reaction descriptors include highest occupied orbital energy, lowest unoccupied orbital energy, core-core repulsion energy, ionization potential, solvent surface area, and volume descriptors.

5. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 2, characterized in that: The physical descriptors include the molecular weight of the fuel molecules, the initial combustion temperature, the initial pressure, and the chemical equivalence ratio.

6. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: When screening target descriptors, the following steps are included: calculating a correlation coefficient r between descriptors, the descriptors including a basic reaction descriptor and a physical descriptor; Where n is the number of fuel molecules in the fuel molecule database, that is, the number of hydrocarbon molecules; and Represent the average value of descriptor X and the average value of descriptor Y, σ X and σ Y are the standard deviation of descriptor X and the standard deviation of descriptor Y, respectively.

7. A method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 6, characterized in that: When the r value between two descriptors is greater than 0.8, one of the descriptors is screened out according to the preset screening criteria; And evaluate the selected descriptors according to the preset relationship: The preset screening criteria include: calculating the average value between a descriptor and other descriptors, deleting the descriptor with a larger average value, and then obtaining the required descriptor; The preset relationship is that the number of descriptors N to be screened must satisfy 8. The method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: The target descriptors include LEO energy and physical descriptors.

9. The method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: The quantum chemical method is the PM6 quantum chemical method.

10. The method for rapid prediction of combustion characteristics of sustainable aviation fuel according to claim 1, characterized in that: The two-dimensional structure data includes laminar flame propagation velocity of hydrocarbon molecules and corresponding topological structure information.

Images