HTPB solid propellant performance prediction and optimization system based on machine learning

Abstract

The invention discloses an HTPB solid propellant performance prediction and optimization system based on machine learning, and particularly relates to the field of solid propellant performance prediction. Comprising the following steps: respectively adding a mixture into different types of energetic compounds to obtain a plurality of HTPB solid propellants; the content of each component in each HTPB solid propellant is adjusted to obtain solid propellants with various formulas; obtaining the formation enthalpy of each energetic compound, and determining the energy characteristics of the solid propellant of each formula according to the formation enthalpy of the energetic compound and the content of each component; determining a formula descriptor of each solid propellant, wherein the formula descriptor comprises a descriptor of an energetic compound and the content of each component; and training a machine learning model by taking the formula descriptors as input and the energy characteristics of the solid propellant of each formula as output to obtain a performance prediction model. And rapid prediction of the energy performance of the HTPB solid propellant added with different types of energetic molecules can be realized.

Description

Technical Field

[0001] The present application relates to the field of solid propellant performance prediction, and in particular to a HTPB solid propellant performance prediction and optimization system based on machine learning. Background Art

[0002] In recent years, with the rapid development of new energetic compounds, the design of hydroxybutane composite solid propellants has ushered in new opportunities. These new energetic compounds have higher energy density, excellent thermal stability and better combustion characteristics, and have gradually become strong candidate materials to replace traditional ammonium salts (such as AP). In this context, the use of high-energy nitramine components (such as RDX, HMX) to replace ammonium perchlorate (AP) particles and develop hydroxybutane four-component composite propellants has become an important research direction in propellant design. During this period, most researchers used traditional thermodynamic methods to calculate energy characteristics as a reference for formulation design, but traditional thermodynamic calculation methods have many shortcomings in the calculation of energy characteristics. First, the traditional method performs high-precision formation enthalpy calculations for each solid propellant, and the calculation of formation enthalpy usually requires the use of complex quantum chemical methods (such as Gaussian, ORCA, etc.) to perform a large number of molecular trajectory optimization and thermodynamic parameter extraction. This process is not only computationally intensive and time-consuming, but also has high performance requirements for the computing platform, which is difficult to meet the needs of rapid evaluation in engineering applications. Secondly, traditional thermodynamic calculation methods do not have the potential to explore unknown energetic compounds, and the development of new energetic compounds is now the best way to break through the energy limitations of HTPB propellants. Summary of the invention

[0003] The main purpose of this application is to provide a HTPB solid propellant performance prediction and optimization system based on machine learning, aiming to solve the problem of large amount of calculation when calculating energy characteristics using traditional thermodynamic methods.

[0004] To achieve the above-mentioned purpose, the present application provides a method for obtaining a HTPB solid propellant performance prediction model based on machine learning, comprising: adding a mixture to different types of energetic compounds respectively to obtain a plurality of HTPB solid propellants, the mixture comprising AP, Al and HTPB; adjusting the content of each component in each HTPB solid propellant to obtain solid propellants of a plurality of formulations; obtaining the formation enthalpy of each energetic compound, and determining the energy characteristics of the solid propellant of each formulation according to the formation enthalpy of the energetic compound and the content of each component; determining a formulation descriptor of each solid propellant, the formulation descriptor comprising a descriptor of the energetic compound and the content of each component; taking the formulation descriptor as input and the energy characteristics of the solid propellant of each formulation as output, training a machine learning model to obtain a performance prediction model.

[0005] Optionally, the structure of the energetic compound includes carbon nitro, oxygen nitro, nitrogen nitro, gem-dinitro, nitroform, azide, furazan, oxidized furazan or tetrazole.

[0006] Optionally, the descriptors of the energetic compound include one or a combination of several of molecular composition, molecular structure, physicochemical properties, hydrogen bond characteristics, optical and physical properties, and molecular fingerprint spectrum.

[0007] Optionally, the energy characteristics include one or a combination of several of specific impulse, combustion chamber temperature, and characteristic velocity.

[0008] To achieve the above object, the present application also provides a method for predicting the performance of HTPB solid propellant based on machine learning, including: obtaining the content of each component in the solid propellant to be predicted; determining the descriptor of the energetic compound in the solid propellant to be predicted; inputting the descriptor of the energetic compound and the content of each component of the solid propellant to be predicted into the above-obtained performance prediction model to obtain the energy characteristics of the solid propellant to be predicted.

[0009] To achieve the above object, the present application also provides an optimization method for HTPB solid propellant based on machine learning, including: taking the content of all components of the solid propellant as an individual, and determining the fitness function according to the energy characteristics of the individual, and using a genetic algorithm to obtain the optimal solid propellant; wherein, the energy characteristics of the individual are predicted by the above-obtained performance prediction model.

[0010] To achieve the above object, the present application also provides a system for obtaining a performance prediction model of HTPB solid propellant based on machine learning, including: a formulation generation module for adding a mixture to different types of energetic compounds to obtain a variety of HTPB solid propellants, the mixture including AP, Al, and HTPB; adjusting the content of each component in each HTPB solid propellant to obtain solid propellants with a variety of formulations; an energy characteristic determination module for obtaining the performance data of each energetic compound and determining the energy characteristics of each formulation of the solid propellant according to the performance data of the energetic compound and the content of each component; a descriptor calculation module for determining the formulation descriptor of each solid propellant, the formulation descriptor including the descriptor of the energetic compound and the content of each component; a model generation module for training a machine learning model with the formulation descriptor as the input and the energy characteristics of each formulation of the solid propellant as the output to obtain the performance prediction model.

[0011] To achieve the above object, the present application also provides a performance prediction system for HTPB solid propellant based on machine learning, including: a content acquisition module for acquiring the content of each component in the solid propellant to be predicted; a descriptor determination module for determining the descriptors of the energetic compounds in the solid propellant to be predicted; and a prediction module for inputting the descriptors of the energetic compounds and the content of each component of the solid propellant to be predicted into the performance prediction model obtained above to obtain the energy characteristics of the solid propellant to be predicted.

[0012] To achieve the above object, the present application also provides an optimization system for HTPB solid propellant based on machine learning, including: an optimization module for taking the content of all components of the solid propellant as an individual, determining a fitness function according to the energy characteristics of the individual, and using a genetic algorithm to obtain the optimal solid propellant; wherein, the energy characteristics of the individual are predicted by the performance prediction model obtained above.

[0013] Compared with the prior art, the beneficial effects of the present application are as follows: The method for obtaining a performance prediction model of HTPB solid propellant based on machine learning of the present invention uses the descriptors of energetic compounds and the content characteristics of each component to characterize the formulation of the solid propellant, takes the multi-dimensional characteristics of energetic compounds as inputs for training, and the training of machine learning maps the relationship between the formulation and the energy characteristics at the molecular structure level, effectively avoiding the acquisition of the formation enthalpy of energetic compounds during later prediction, and reducing the computational amount during the performance prediction of solid propellants.

[0014] The performance prediction method of HTPB solid propellant based on machine learning of the present invention uses a performance prediction model to predict energy characteristics. Compared with traditional thermodynamic calculations that require coupling the formation enthalpy of accurate quantum chemistry calculations, there is no need to obtain the formation enthalpy of energetic compounds in each solid propellant to be predicted. Only by inputting the formulation descriptors of the solid propellant to be predicted, the energy performance of HTPB solid propellants with different types of energetic molecules added can be quickly predicted.

[0015] The performance optimization method of HTPB solid propellant based on machine learning of the present invention combines a genetic algorithm with a machine learning model, quickly screens out potential energetic compounds, and realizes the automatic search for the optimal performance formulation, shortening the design and research and development cycle of new solid propellants. Description of the Drawings

[0016] Figure 1 It is a schematic flow chart of a method for obtaining a performance prediction model of HTPB solid propellant based on machine learning of the present application; Figure 2 It is a schematic flow chart of a performance prediction method of HTPB solid propellant based on machine learning of the present application; Figure 3 Schematic flow chart of Embodiment 1 of a method for predicting the performance of HTPB solid propellant based on machine learning according to the present application; Figure 4 Schematic flow chart of Embodiment 2 of a method for optimizing the performance of HTPB solid propellant based on machine learning according to the present application.

[0017] The realization, functional features and advantages of the purpose of the present application will be further described with reference to the embodiments and the accompanying drawings. Specific embodiments

[0018] To make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the accompanying drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0019] The first embodiment of the present invention provides a method for obtaining a performance prediction model of HTPB solid propellant based on machine learning, as Figure 1 shown, specifically including the following steps: Step S1, adding a mixture to different types of energetic compounds (ECs) to obtain a variety of HTPB solid propellants. The mixture includes aluminum powder (Al), ammonium perchlorate (AP), and hydroxyl-terminated polybutadiene (HTPB); Specifically, a molecular data set is constructed using different types of energetic compounds and their corresponding enthalpies of formation; a mixture with an arbitrary mass ratio is added to each type of energetic compound to obtain a variety of HTPB solid propellants. Among them, the types of the mixture in each HTPB solid propellant are the same, that is, on the premise that the types of AP, Al, and HTPB remain unchanged, each energetic compound in the molecular data set is mixed with the mixture respectively to obtain HTPB solid propellants with different types of energetic compounds. In addition, the mass ratios of AP, Al, and HTPB in each HTPB solid propellant can be the same or different.

[0020] Exemplarily, the structures of the energetic compounds include carbon nitro, oxygen nitro, nitrogen nitro, gem-dinitro, nitroform, azide, furazan, oxidized furazan, or tetrazole. The performance data of the energetic compounds include the enthalpy of formation.

[0021] Step S2, adjusting the content of each component in each HTPB solid propellant to obtain solid propellants with various formulations; Step S3: Obtain the enthalpy of formation of each energetic compound, and determine the energy characteristics of the solid propellant of each formulation according to the enthalpy of formation of the energetic compound and the content of each component; wherein the energy characteristics include one or a combination of specific impulse, combustion chamber temperature, and characteristic velocity.

[0022] Specifically, input the enthalpy of formation in the energetic molecule dataset into the CEA_Wrap library, calculate the energy characteristics of the solid propellant of each formulation under preset conditions. The preset conditions can be an initial temperature of 298.15 K, a nozzle exit pressure of 0.1 MPa, and a combustion chamber pressure of 6.98 MPa. Use the calculated data as the dataset for machine learning training.

[0023] Step S4: Determine the formulation descriptor of each solid propellant. The formulation descriptor includes the descriptor of the energetic compound and the content of each component. Among them, the descriptor of the energetic compound includes one or a combination of molecular composition, molecular structure, physicochemical properties, hydrogen bond characteristics, optical and physical properties, and molecular fingerprint spectrum.

[0024] In this embodiment, since only the type of energetic compound changes in each formulation, the descriptor calculation of the energetic compound plus the content characteristics of each component can be used to characterize a formulation. Specifically, when calculating the descriptor of the energetic compound, the molecular structure of the energetic molecule can be optimized based on the molecular force field method, and then the descriptor calculation is performed on the optimized molecular structure as the descriptor of the energetic compound. Since the same substance can have multiple three-dimensional structures, a three-dimensional structure can be obtained through structure optimization, so as to make the characteristic acquisition of the energetic compound more accurate, and further ensure the accuracy of the descriptor calculation.

[0025] Exemplarily, the molecular composition includes the number of elements such as carbon, nitrogen, oxygen, and hydrogen, as well as the number of special groups such as amino groups, nitro groups, and nitramino groups. The molecular structure includes structural features such as the number of rings, heterocycles, aromatic rings, and rotatable bonds. The physicochemical properties include molecular weight, volume, oxygen balance, etc., which are used to evaluate the physical properties of the molecule. The hydrogen bond characteristics include the ability of the molecule to act as a hydrogen bond donor and acceptor. The optical and physical properties include molar refractivity, molecular planarity index, and molecular eccentricity, etc., which describe the geometric characteristics and stability of the molecule in space. The molecular fingerprint spectrum includes generating an E-state fingerprint spectrum by calculating the electronic environment of the molecule, which is commonly used in machine learning and chemical reaction prediction.

[0026] Step S5: Use the formulation descriptor as the input and the energy characteristics of the solid propellant of each formulation as the output to train a machine learning model to obtain a performance prediction model.

[0027] Exemplarily, the machine learning model may include Ridge Regression (RR), Multilayer Perceptron (MLP), AdaBoost, and K-Nearest Neighbor (KNN). In the training phase, the MLP is trained through the PyTorch framework, while the RR, AdaBoost, and KNN models are constructed using the Scikit-learn library, and grid search and cross-validation techniques are employed to fine-tune the hyperparameters of the models.

[0028] The second embodiment of the present invention provides a method for predicting the performance of HTPB solid propellant based on machine learning, as Figure 2 shown, which specifically includes the following steps: Step S10: Obtain the content of each component in the solid propellant to be predicted; Step S20: Determine the descriptors of the energetic compounds in the solid propellant to be predicted; Step S30: Input the descriptors of the energetic compounds and the content of each component of the solid propellant to be predicted into the above performance prediction model to obtain the energy characteristics of the solid propellant to be predicted.

[0029] It should be noted that the descriptor of the energetic compound in this embodiment is a multi-dimensional vector, which can be obtained by converting the SMILE code of each energetic compound. The specific conversion method is to convert the SMILE code through python to obtain the descriptor of the energetic compound.

[0030] In this embodiment, various types of energetic compounds and their corresponding enthalpies of formation are obtained in advance. The formulation descriptor of the corresponding solid propellant is determined according to the descriptor of each energetic compound and the content of each component, and used as the input; and each energetic compound is used to generate the corresponding solid propellant, and the energy characteristics of the solid propellant of each formulation are determined according to the enthalpy of formation and used as the output; a performance prediction model is trained; using this performance prediction model for energy characteristic prediction, there is no need to calculate the enthalpy of formation of each solid propellant to be predicted, and only the descriptor of the energetic compound in the solid propellant needs to be input, thereby reducing the calculation amount and calculation time, and having a lower performance requirement for the calculation platform.

[0031] Example 1: Predict the energy characteristics of a solid propellant with components Al / RDX / HTPB / AP As Figure 3 shown, determine the molecular structure of RDX represented by the SMILES code , the descriptor of the corresponding energetic compound is a 74-dimensional vector, specifically including: the number of carbon atoms, the number of nitrogen atoms, the number of oxygen atoms, the number of hydrogen atoms, the number of amino groups, the number of nitro groups, the number of carbon-nitro groups, the number of nitrogen-nitro groups, the number of oxygen-nitro groups, the number of gem-dinitro groups, the number of nitroform groups, the number of hydrogen bond donors, the number of hydrogen bond acceptors, the molecular weight, the molecular volume, the oxygen balance, the number of rings, the number of heterocycles, the number of aromatic heterocycles, the number of aromatic rings, the number of rotatable bonds, the molar refraction, the molecular planarity index, the molecular eccentricity, the molecular principal moment of inertia, and the E-state molecular fingerprint spectrum. And the formulation of the solid propellant to be predicted, the content of each component is a 4-dimensional vector, specifically: the content of HTPB is 12%, the content of Al is 18%, the content of AP is 60%, and the content of RDX is 10%; Select the performance prediction model obtained based on the MLP model, input the SMILES code and the formulation of the solid propellant to be predicted, and obtain the specific impulse of the current solid propellant as 266.1 s, the characteristic velocity as 1594. m / s, and the combustion chamber temperature as 3433 K.

[0032] The third embodiment of the present invention provides an optimization method for HTPB solid propellant based on machine learning, including: Regarding the content of all components of the solid propellant as an individual, and determining the fitness function according to the energy characteristics of the individual, and using the genetic algorithm to obtain the optimal solid propellant; wherein, the energy characteristics of the individual are predicted by the above performance prediction model. Exemplarily, the energy characteristic is the specific impulse.

[0033] Example 2 As Figure 4 shown, according to the preset constraint conditions, that is, the user defines the range of different components, HTPB > 14%, 8% < RDX < 14%, regarding the formulation as an individual, randomly generate an initial population , and ensure that all formulation variables satisfy non-negativity and the total ratio constraint; use the above performance prediction model to calculate the specific impulse of each individual , and evaluate it through the fitness function , where can represent penalty terms for factors such as formulation cost and safety. Based on the fitness value , adopt the roulette wheel or tournament selection method, and preferentially select individuals with higher fitness as parents. Then, use the single-point crossover method to recombine the genes of the parent individuals, such as exchanging variables at a random position to generate new offspring , and at the same time, through the mutation operator Adjust the values of some variables with a certain probability to enhance population diversity and prevent premature convergence. This optimization process continues iteratively, and the population evolves continuously under the actions of fitness evaluation, selection, crossover, and mutation, gradually approaching the global optimal solution. When the preset stopping conditions are reached, such as the maximum number of iterations Tmax or the fitness change is less than the convergence threshold When, the algorithm terminates and outputs the optimized solid propellant formulation with an HTPB content of 14%, an Al content of 18%, an AP content of 57.4%, and an RDX content of 10.5%, obtaining the solid propellant with the optimal energy characteristics; the specific impulse is 265.6 s, and the calculation time is less than 30 s.

[0034] It can be proved hereby that through the data-driven optimization method of this embodiment, not only can the automation degree of formulation design be improved, but also the optimization efficiency and accuracy are enhanced by combining with the performance prediction model, thereby providing a more reliable optimization scheme for propellant research and development, reducing the test cost, and improving the overall performance of the propellant.

[0035] The fourth embodiment of the present invention provides an acquisition device for a performance prediction model of HTPB solid propellant based on machine learning, including: A formulation generation module, configured to add a mixture to different types of energetic compounds respectively to obtain a variety of HTPB solid propellants, where the mixture includes AP, Al, and HTPB; adjust the content of each component in each HTPB solid propellant to obtain solid propellants with a variety of formulations; An energy characteristic determination module, configured to obtain the performance data of each energetic compound, and determine the energy characteristics of each formulated solid propellant according to the performance data of the energetic compound and the content of each component; A descriptor calculation module, configured to determine the formulation descriptor of each solid propellant, where the formulation descriptor includes the descriptor of the energetic compound and the content of each component; A model generation module, using the formulation descriptor as the input and the energy characteristics of each formulated solid propellant as the output, trains a machine learning model to obtain a performance prediction model.

[0036] The fifth embodiment of the present invention provides a performance prediction system for HTPB solid propellant based on machine learning, including: A content acquisition module, configured to acquire the content of each component in the solid propellant to be predicted; A descriptor determination module, configured to determine the descriptor of the energetic compound in the solid propellant to be predicted; A prediction module, configured to input the descriptor of the energetic compound and the content of each component of the solid propellant to be predicted into the above-obtained performance prediction model to obtain the energy characteristics of the solid propellant to be predicted.

[0037] The sixth embodiment of the present invention provides an optimization system for HTPB solid propellant based on machine learning, including: An optimization module, which takes the content of all components of the solid propellant as an individual, determines a fitness function according to the energy characteristics of the individual, and uses a genetic algorithm to obtain the optimal solid propellant; Among them, the energy characteristics of the individual are predicted by the performance prediction model obtained above.

[0038] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A method for obtaining a performance prediction model of HTPB solid propellant based on machine learning, characterized in that, Comprising: Adding a mixture to different types of energetic compounds respectively to obtain a variety of HTPB solid propellants, wherein the mixture includes AP, Al, and HTPB; Adjusting the content of each component in each HTPB solid propellant to obtain solid propellants with a variety of formulations; Obtaining the enthalpy of formation of each energetic compound, and determining the energy characteristics of each solid propellant according to the enthalpy of formation of the energetic compound in each solid propellant and the content of each component; Determining the formulation descriptor of each solid propellant, wherein the formulation descriptor includes the descriptor of the energetic compound and the content of each component; Using the formulation descriptor of each solid propellant as the input and the corresponding energy characteristics as the output to train a machine learning model to obtain a performance prediction model.

2. The method for obtaining a performance prediction model of HTPB solid propellant based on machine learning according to claim 1, wherein, The structure of the energetic compound includes carbon nitro, oxygen nitro, nitrogen nitro, gem-dinitro, nitroform, azide, furazan, oxidized furazan, or tetrazole.

3. The method for obtaining the HTPB solid propellant performance prediction model based on machine learning according to claim 1, wherein The descriptor of the energetic compound includes one or a combination of several of molecular composition, molecular structure, physicochemical properties, hydrogen bond characteristics, optical and physical properties, and molecular fingerprint spectrum.

4. The method for obtaining the HTPB solid propellant performance prediction model based on machine learning according to claim 1, wherein The energy characteristics include one or a combination of several of specific impulse, combustion chamber temperature, and characteristic velocity.

5. A method for predicting the performance of HTPB solid propellant based on machine learning, characterized in that, Comprising: Obtaining the content of each component in the solid propellant to be predicted; Determining the descriptor of the energetic compound in the solid propellant to be predicted; Inputting the descriptor of the energetic compound and the content of each component of the solid propellant to be predicted into the performance prediction model obtained in any one of claims 1-4 to obtain the energy characteristics of the solid propellant to be predicted.

6. An optimization method for HTPB solid propellant based on machine learning, characterized in that, Comprising: Taking the content of all components of the solid propellant as an individual, and determining a fitness function according to the energy characteristics of the individual, and using a genetic algorithm to obtain the optimal solid propellant; Wherein, the energy characteristics of the individual are predicted by the performance prediction model obtained in any one of claims 1-4.

7. A system for obtaining a performance prediction model of HTPB solid propellant based on machine learning, characterized in that, Comprising: A formulation generation module, configured to add a mixture to different types of energetic compounds respectively to obtain a variety of HTPB solid propellants, wherein the mixture includes AP, Al, and HTPB; adjusting the content of each component in each of the HTPB solid propellants to obtain solid propellants with a variety of formulations; An energy characteristic determination module, configured to obtain the performance data of each energetic compound, and determine the energy characteristics of each formulated solid propellant according to the performance data of the energetic compound and the content of each component; A descriptor calculation module, configured to determine the formulation descriptor of each solid propellant, wherein the formulation descriptor includes the descriptor of the energetic compound and the content of each component; A model generation module, which takes the formulation descriptor as the input and the energy characteristics of each formulated solid propellant as the output, and trains a machine learning model to obtain a performance prediction model.

8. A performance prediction system for HTPB solid propellant based on machine learning, characterized in that, Comprising: A content acquisition module, configured to obtain the content of each component in the solid propellant to be predicted; A descriptor determination module, configured to determine the descriptor of the energetic compound in the solid propellant to be predicted; A prediction module, configured to input the descriptor of the energetic compound and the content of each component of the solid propellant to be predicted into the performance prediction model obtained in claim 7 to obtain the energy characteristics of the solid propellant to be predicted.

9. An optimization system for HTPB solid propellant based on machine learning, characterized in that, Comprising: An optimization module, which takes the content of all components of the solid propellant as individuals, determines a fitness function according to the energy characteristics of the individuals, and uses a genetic algorithm to obtain the optimal solid propellant; Among them, the energy characteristics of the individuals are predicted by the performance prediction model obtained in claim 7.

Images