A data knowledge fusion solid engine migration optimization method

Abstract

The application discloses a data knowledge fusion solid engine migration optimization method, comprising the following steps: mapping the performance characteristics and quality characteristics of each source domain engine to a target domain engine to obtain a plurality of low-precision sample points at a first mapping burning rate; calculating the mapping thrust-time curve of each source domain engine on the target domain engine at a second mapping burning rate, and obtaining a high-performance design domain of the target domain engine based on the error between the mapping thrust-time curve and the required thrust-time curve of the target domain engine; and iteratively optimizing the target domain engine based on the low-precision sample points and the high-performance design domain to obtain the design result of the target domain engine. The application is applied to the technical field of solid engines, can effectively improve the overall design efficiency of the solid engine, and solves the problems of a large number of design parameters, a large variable range, difficult algorithm convergence and low design efficiency during the overall design of the solid engine.

Description

Technical Field

[0001] This invention relates to the field of solid rocket motor technology, specifically a solid rocket motor migration optimization method based on data and knowledge fusion. Background Technology

[0002] Solid rocket engines are one of the most widely used propulsion systems in rockets and other space launch vehicles. Overall design is one of the most critical and challenging technologies in solid rocket engine design. Its main task is to determine the optimal design parameters and range based on the provided overall indicators, design the main design parameters of the solid rocket engine, and complete the preliminary design of the solid rocket engine.

[0003] Currently, the commonly used design methods for solid rocket motors mainly include manual methods and optimization methods. Manual methods refer to experienced engineers adjusting the initial design parameters of the motor through experimental trial and error to meet new design requirements, which usually yields good results. Optimization methods involve constructing a performance simulation model of the solid rocket motor, determining the design variables and their range, and then using intelligent optimization methods or existing surrogate model-based optimization methods to search the design domain and achieve the optimal design.

[0004] However, in practical applications, manual methods require extensive engineering experience and have a high design threshold. Furthermore, poor or unsuitable experimental results can lead to a large amount of repetitive work, wasting time and experimental costs. Optimization methods, on the other hand, involve cold starts exploring the design space each time, easily resulting in wasted computational resources. Summary of the Invention

[0005] To address the problems of numerous design parameters and a large range of variables in existing solid rocket motors, which lead to difficulties in algorithm convergence and low design efficiency during overall design, this invention provides a data-knowledge fusion-based solid rocket motor migration optimization method. This method improves the overall design efficiency of solid rocket motors and provides methodological support for efficient solid rocket motor design.

[0006] To achieve the above objectives, this invention provides a solid rocket motor migration optimization method based on data knowledge fusion, comprising the following steps:

[0007] Step 1: Under the first mapped combustion rate, the performance characteristics and quality characteristics of each source domain engine are mapped to the target domain engine to obtain several low-precision sample points, wherein the first mapped combustion rate is the same as the combustion rate of the source domain engine.

[0008] Step 2: Under the second mapped combustion rate, calculate the mapped thrust-time curve of each source domain engine on the target domain engine, and obtain the high-performance design domain of the target domain engine based on the error between the mapped thrust-time curve and the required thrust-time curve of the target domain engine. The second mapped combustion rate is obtained by back-calculation based on the working time requirement of the target domain engine and the working time of the source domain engine.

[0009] Step 3: Based on the low-precision sample points and the high-performance design domain, iteratively optimize the target domain engine to obtain the design result of the target domain engine.

[0010] Compared with the prior art, the present invention has the following beneficial technical effects:

[0011] This invention starts from existing solid rocket motor design case knowledge and uses a case history knowledge transfer mapping method to map source domain motor design cases to target domain, obtaining prior knowledge of the target domain. Through variable combustion rate mapping, a high-performance design domain is obtained, enabling efficient design of the target motor. This not only automates and lowers the threshold compared to general manual selection methods, but also improves optimization efficiency and achieves intelligent design by reusing historical knowledge, effectively meeting the intelligent design requirements of solid rocket motors. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the solid rocket motor migration optimization method based on data knowledge fusion in an embodiment of the present invention;

[0014] Figure 2 This is a schematic diagram of the thrust-time curve of a single target chamber and a single thruster in an example of an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of a front and rear wing-shaped charge in an example of an embodiment of the present invention;

[0016] Figure 4 This is a schematic diagram showing the convergence comparison results of a low-precision model guided by historical knowledge in the initial domain in an example of the embodiment of the present invention, wherein: (a) is a schematic diagram of the results of an advanced surrogate model-based optimization method, and (b) is a schematic diagram of the results of a historical knowledge-driven transfer optimization method;

[0017] Figure 5 This is a schematic diagram of a solid rocket motor migration design method in an embodiment of the present invention;

[0018] Figure 6 This is a schematic diagram of the thrust design results in an example of an embodiment of the present invention.

[0019] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0022] To address the problems of numerous and wide-ranging design variables in solid rocket motors, leading to difficulties in algorithm convergence and low design efficiency during overall design, this embodiment discloses a data-knowledge fusion-based solid rocket motor transfer optimization method (hereinafter referred to as the "transfer optimization method"). By studying the performance and mass models of solid rocket motors, and using size ratio and combustion rate ratio as benchmarks, performance characteristics and mass properties such as thrust-time curves from existing historical cases are mapped to the target domain engine propellant type. This achieves cost-free acquisition of target domain engine design knowledge. Furthermore, a high-performance design domain is obtained through variable combustion rate mapping. Finally, an inaccurate sampling method with dynamic constraint relaxation is used to achieve efficient design of the target domain engine.

[0023] refer to Figure 1 The solid rocket motor migration optimization method based on data knowledge fusion in this embodiment specifically includes the following steps 1 to 3.

[0024] Step 1: Under the first mapped combustion rate, map the performance characteristics and quality characteristics of each source domain engine to the target domain engine to obtain several low-precision sample points, wherein the first mapped combustion rate is the same as the combustion rate of the source domain engine.

[0025] In practical applications, the target domain engine's outer diameter requirement D, operating time requirement t, and time-discrete thrust requirement F are all known. Historical examples involve the source domain engine's outer diameter D. s Burning speed r s and thrust-time curve (including thrust F) s With working time t s This is also known. The design goal in this embodiment is to achieve the target domain engine's outer diameter requirement D. t Working hours t t and thrust-time curve F tIt meets the requirements of outer diameter dimension D, working time t, and time-discrete thrust F.

[0026] In the specific implementation of step 1, this embodiment maps the performance and quality characteristics of each source domain engine to the target domain under a fixed combustion rate, obtaining mapped engines that correspond one-to-one with each source domain engine. These mapped engines constitute a series of low-precision sample points. Each low-precision sample point includes one or more design variables, which are specific dimensional or performance parameters of the corresponding mapped engine, such as the canard length of the fore-and-aft wing-type propellant charge, the diameter of the canard tangent circle, the maximum pressure in the combustion chamber, and the length of the circular section of the combustion chamber. The high-performance design domain in subsequent step 2 is the range of values ​​for each design variable.

[0027] In the specific implementation process, for source domain engine i, the process of mapping its performance characteristics and quality characteristics to the target domain engine is as follows:

[0028] Step 101, obtain the outer diameter D of the source domain engine i. si Burning speed r si Working hours t si With thrust F si ;

[0029] Step 102, map source domain engine i to target domain engine i. t outer diameter D ti =D, let the mapped engine i t First mapping burn rate r ti-1 =r si ;

[0030] Step 103, calculate the mapped engine i t The size ratio λ of the source domain engine i i And the burning rate ratio β i , for λ i =D ti / D si β i =r ti-1 / r si ;

[0031] Step 104, using the size ratio λ i And the burning rate ratio β i Estimating the mapping engine i t Compared to the operating time of source domain engine i, it is t ti / t si =λ i / β i , where t ti For mapping engine i t Estimated working time;

[0032] Step 105, using the size ratio λ i Estimating the mapping engine i t The total impulse ratio of the source domain engine i is Among them, I ti For mapping engine i t The estimated total impact;

[0033] Step 106: Obtain the mapped engine i by using the total impulse ratio and the working time ratio. t The thrust ratio of the source domain engine i is F. ti / F si =I ti / t ti , of which F ti For mapping engine i t Estimated thrust;

[0034] Step 107, based on the size ratio λ i Computational mapping engine i t The mass ratio of the combustion chamber of engine i to that of source domain engine i is:

[0035]

[0036] in, They are respectively mapped engine i t The combustion chamber mass of engine i in the source domain, [σ] si , Here, [σ] represents the tensile strength and density of the combustion chamber material in the source-domain engine i, respectively. ti , They are respectively mapped engine i t The tensile strength and density of the combustion chamber material;

[0037] Step 108, based on the size ratio λ i Estimating the mapping engine i t The ratio of the propellant mass to that of the source domain engine i is: in, They are respectively mapped engine i t The propellant charge quality of the source domain engine i;

[0038] Step 109, using the size ratio λ i And the burning rate ratio β i Computational mapping engine i t The mass ratio of the nozzle to that of the source domain engine i is:

[0039]

[0040] in, They are respectively mapped engine it The nozzle mass of the source domain engine i They are respectively mapped engine i t Density of nozzle material of source domain engine i;

[0041] Step 1010, using the size ratio λ i And the burning rate ratio β i Computational mapping engine i t The mass ratio of the insulation layer of the source engine i to that of the source region engine is:

[0042]

[0043] in, They are respectively mapped engine i t The insulation quality of the source domain engine i They are respectively mapped engine i t Density of the insulation layer material of the source engine i.

[0044] The above-mentioned mapping engine i t Estimated working time t ti Estimated total impact I ti With estimated thrust F ti That is, the performance characteristics of the source domain engine i mapped onto the target domain, and the mapped engine i t Combustion chamber mass Explosive loading quality Nozzle quality Insulation layer quality This refers to the quality characteristics of the source domain engine i mapped onto the target domain.

[0045] It is worth noting that the mass mapping process in steps 107 to 1010 takes into account the density change caused by material replacement and the working time change caused by the burning rate change. Therefore, when the burning rate and the materials of each component remain unchanged before and after mapping, the mass mapping in steps 107 to 1010 can be simplified to: This reduces the computational cost of constructing low-precision samples.

[0046] Step 2: Under the second mapped combustion rate, calculate the mapped thrust-time curve of each source domain engine on the target domain engine, and obtain the high-performance design domain of the target domain engine based on the error between the mapped thrust-time curve and the required thrust-time curve of the target domain engine. The second mapped combustion rate is obtained by back-calculation based on the working time requirement of the target domain engine and the working time of the source domain engine.

[0047] In the specific implementation process, for source domain engine i, the calculation process of its mapped thrust-time curve on the target domain engine is as follows:

[0048] First, let the source domain engine i be mapped to the target domain engine i. t outer diameter D ti =D, let the source domain engine i be the mapping engine i in the target domain. t Working hours T ti =t;

[0049] Secondly, based on the mapping engine i t outer diameter D ti Working hours T ti And the outer diameter D of the source domain engine i si Burning speed r si Then, based on the size ratio formula and the combustion rate ratio formula in step 1, the second mapped combustion rate of the source domain engine i in the process of mapping to the target domain can be obtained by reverse deduction, which is:

[0050]

[0051] Where, r ti-2 The second mapped combustion rate of source domain engine i during the mapping to the target domain;

[0052] Finally, in the second mapping combustion rate r ti-2 Next, the mapping engine i is calculated using the same method as in step 1. t By combining the operating time ratio and thrust ratio of source domain engine i, the mapped thrust-time curve of source domain engine i on the target domain engine can be obtained.

[0053] In this embodiment, the specific implementation process for obtaining the high-performance design domain of the target domain engine based on the error between the mapped thrust-time curve and the demand thrust-time curve of the target domain engine is as follows:

[0054] First, the error between the mapped thrust-time curve of the mapped engine corresponding to each source domain engine and the required thrust-time curve of the target domain engine is calculated as follows:

[0055]

[0056] Among them, RMSE i The mapping of engine i in the source domain to engine i in the target domain t The error between the mapped thrust-time curve and the target domain engine's required thrust-time curve, F ti,l The mapping of engine i in the source domain to engine i in the target domain t The thrust value F at time point l in the mapped thrust-time curve. l Let G be the thrust value at time point l on the thrust-time curve of the engine in the target domain, and G be the number of time discrete points; the calculation error value RMSEi During the process, if the mapping thrust F ti If the working time is not equal to the demand thrust F, the excess part is calculated as 0kN;

[0057] Then, the errors are sorted in ascending order, and the top N1% (N1 = 5 to 10 in this embodiment) of the mapped engines are selected. From these, the mapped engine with the smallest mass that has a total impulse greater than or equal to the total impulse requirement of the target domain engine is selected as the central sample.

[0058] Next, all mapped engines are used as mapped samples, and the distances between all mapped samples and the central sample are calculated and sorted in ascending order. The design space of the envelope of the top N2% (N2 = 5-10 in this embodiment) of mapped samples is selected as the high-performance design domain of the target domain engine, that is:

[0059] Lb j =min(x 1,j ,…,x k,j ,…,x H,j ,)

[0060] Ub j =max(x 1,j ,…,x k,j ,…,x H,j ,)

[0061] Among them, Lb j Ub j Let x be the lower and upper bounds of the sample point in the j-th dimension (i.e., the j-th design variable among the sample points), respectively. k,j Let H be the value of the k-th sample point in the j-th dimension, and H be the number of mapping samples participating in the high-performance design domain envelope design.

[0062] Step 3: Iteratively optimize the target domain engine based on low-precision sample points and a high-performance design domain to obtain the design results of the target domain engine. The specific implementation process is as follows:

[0063] Step 301: A number of high-precision sample points are collected in the high-performance design domain using the recursive evolution optimized Latin hypercube experimental design method, and each high-precision sample point is stored in the elite archive. A low-precision model of the target domain engine is constructed based on each low-precision sample point.

[0064] Step 302: Construct an error model based on the prediction errors of high-precision sample points and low-precision models in the elite archives, and construct a multi-precision prediction model for the target domain engine based on the low-precision model and the error model.

[0065] Step 303: Determine the sampling criteria based on the feasibility of each high-precision sample point in the current elite archive, and use the non-precise sampling method of constraint dynamic relaxation to search for a new high-precision sample point that satisfies the sampling criteria.

[0066] Step 304: Based on the multi-precision prediction model, determine whether the new high-precision sample point is better than the worst high-precision sample point in the current elite archive:

[0067] If so, delete the worst high-precision sample point in the current elite file, update the elite file with new high-precision sample points, and then proceed to step 305.

[0068] Otherwise, proceed to step 305;

[0069] Step 305: Determine whether the best high-precision sample point in the elite archive has not been updated for M consecutive times:

[0070] If so, the best high-precision sample point in the current elite archives will be used as the design result for the solid rocket motor and output.

[0071] Otherwise, return to step 302.

[0072] In the specific implementation of step 3 of this embodiment, the concept of an elite archive is introduced to store high-precision sample points after each iteration, that is, to store the high-precision sample points with the smallest objective function value. In this embodiment, the objective function value is the overall weight of the engine while ensuring that the engine thrust performance meets the requirements; that is, high-precision sample points with smaller objective function values ​​are preferred. The number of sample points in the elite archive remains constant throughout the iteration process.

[0073] In the specific implementation of step 301, a low-precision model is constructed using the radial basis interpolation method, as follows:

[0074]

[0075] Where X represents the sample point, f L (X) represents the objective function output of the low-precision model for sample point X, N represents the number of low-precision sample points used to construct the low-precision model, and ω n These represent the weight coefficients of the radial basis functions. Let r represent the basis function with the Euclidean distance from the unknown sample to the known sample as the independent variable, and r represent the Euclidean distance between the samples.

[0076] After obtaining the low-precision model, the multi-precision prediction model of the target domain engine can be obtained by superimposing the error model on the low-precision model, as follows:

[0077] f M (X)=f L(X)+e(X)

[0078] Among them, f M (X) represents the multi-precision prediction model, and e(X) represents the error model.

[0079] In this embodiment, the error model is also constructed using the sample radial basis function method. The input is the high-precision sample points in the elite archive, and the output is the error between the actual output of the high-precision sample points and the predicted output of the multi-precision prediction model. Its structure is the same as that of the low-precision model, so it will not be described in detail in this embodiment. The actual output of the high-precision sample points refers to the overall weight of the engine obtained from actual simulation of the high-precision sample points, ensuring that the engine thrust performance meets the requirements.

[0080] In the specific implementation of step 303, this embodiment sets up a three-stage constrained sampling method based on the feasibility of each sample point in the current elite archive, specifically:

[0081] When all sample points in the elite archive are infeasible, the sampling criterion is to find sample points with smaller constraint violation values;

[0082] When some sample points in the elite archive are not feasible, the sampling criterion is to find sample points with smaller constraint violation values ​​and better objective function values;

[0083] When all sample points in the elite archive are feasible, the sampling criterion is to find sample points with better objective function values.

[0084] More specifically, a constraint conflict function is introduced during the three-stage constraint sampling process. The constraint conflict function is an important indicator that effectively characterizes the distance of the current high-precision sample point from the feasible region, and is a commonly used method in constraint optimization algorithms. Assume that a series of constraints exist in the solid rocket motor design process:

[0085] g j (X)≤0j=1,2,…,p

[0086] h k (X)=0k=p+1,p+2,…,p+m

[0087] Among them, g j (X) represents the design variable function of the j-th inequality constraint in the solid rocket motor design, h k (X) represents the design variable function of the k-th equality constraint in the solid rocket motor design, and p and m represent the inequality constraints and the number of inequality constraints, respectively.

[0088] Since equality constraints are strong constraints, they pose difficulties for the feasible region of the sample point search and localization problem. Therefore, equality constraints are usually handled as follows:

[0089] |h k (X)|-δ≤0

[0090] Here, δ represents a small parameter, which allows equality constraints to be converted into inequality constraints. In this case, a high-precision sample point X violates the constraint conflict value G of the l-th constraint in the constraint optimization model. l (X) is:

[0091]

[0092] Among them, g l (X) represents the design variable function of the l-th inequality constraint in the solid rocket motor design, h l (X) represents the design variable function of the l-th equality constraint in the solid rocket motor design;

[0093] At this point, the constraint conflict value G(X) of the high-precision sample point X that violates all constraints is:

[0094]

[0095] It is worth noting that due to the differences in scale characteristics among different variables and constraints, some constraints generate large conflict values ​​even when slightly deviating from the feasible region within the design domain. This leads to these few constraints playing a decisive role in the constraint conflict value G(X) of this sample point. Therefore, it is necessary to normalize the constraint conflict for each constraint within the population. The normalization process consists of the following two steps:

[0096] (1) Calculate the maximum conflict value of all high-precision sample points in the elite archive for a single constraint. for:

[0097]

[0098] Among them, X elite Represents elite archives;

[0099] (2) Based on the maximum conflict value After normalization, the normalized constraint conflict value G of all constraints in the constraint optimization model is obtained for the high-precision sample point X. nor (X) is:

[0100]

[0101] When G nor When (X)≤0, it indicates that the high-precision sample point X is within the feasible region, and the high-precision sample point X can be determined to be feasible; otherwise, the high-precision sample point X is determined to be infeasible.

[0102] In this embodiment, the three-stage constraint sampling is carried out based on different conditions of the elite files. The sampling criteria (the purpose of sampling) are different depending on the conditions.

[0103] When all high-precision sample points in the elite archive are infeasible, meaning at least one strong constraint exists, the goal of filling (finding the next new high-precision sample point) is to approach the boundary of the strong constraint. The sample with the smaller constraint violation compared to the high-precision sample point with the smallest constraint violation in the elite archive is selected. The sampling criterion at this point is to find high-precision sample points with smaller constraint violation values, i.e.:

[0104]

[0105] Where X0 represents the newly found high-precision sample point, G nor (X0) represents the normalized constraint conflict value of the new high-precision sample point. This represents the smallest normalization constraint conflict value in the elite files.

[0106] When some high-precision sample points in the elite archive become infeasible, the search for new high-precision sample points shifts to samples with relatively lower constraint violations and better objective values ​​compared to the best samples in the elite archive. This allows sampling of infeasible samples with high objective performance around the constraint boundaries. The sampling criterion at this point is to find high-precision sample points with smaller constraint violation values ​​and better objective function values, specifically:

[0107] Find:f(X0) <f min (X elite )

[0108]

[0109] Among them, f min (X elite () indicates the optimal target value in the elite profile. This represents the set of all infeasible high-precision sample points in the elite archive. This represents the minimum normalization constraint violation rate for infeasible high-precision sample points in the elite archives.

[0110] When all high-precision sample points in the elite archive are feasible, the new high-precision sample point searched for is selected based on the sample with a better objective value compared to the best sample in the elite archive. The search stops once a new high-precision sample point is found, thus avoiding exhaustive search of imprecise models, which may help the algorithm escape local optima. The sampling criterion at this point is to find sample points with better objective function values, specifically:

[0111] Find:f(X0) <f min (Xelite )

[0112] stG nor (X0)=0

[0113] After determining the sampling criteria based on the feasibility of each high-precision sample point in the current elite archive, sampling of new high-precision sample points can be carried out.

[0114] Step 304, the process of determining whether the new high-precision sample point is better than the worst high-precision sample point in the current elite archive based on the multi-precision prediction model, is as follows:

[0115] The high-precision sample point with the largest high-precision target simulation value in the current elite archive is defined as the worst sample point in the current elite archive.

[0116] Substitute the new high-precision sample points into the multi-precision prediction model to obtain the high-precision target simulation values ​​corresponding to the new high-precision sample points;

[0117] The high-precision target simulation value of the new high-precision sample point is compared with the high-precision target simulation value of the worst high-precision sample point in the current elite archive. If the high-precision target simulation value of the new high-precision sample point is smaller, it is determined that the new high-precision sample point is better than the worst high-precision sample point in the current elite archive.

[0118] Among them, the high-precision target simulation values ​​of high-precision sample points are obtained through simulation.

[0119] The following section provides a further explanation of the solid rocket motor migration optimization method based on data knowledge fusion in this embodiment, using specific examples.

[0120] Taking the overall design of a single-chamber, single-thrust solid rocket motor as an example, this paper presents a case study of solid rocket motor design migration. The existing design case involves 500 sample points generated during the design process of a 1400mm diameter fin-pole type propellant engine. The target engine has an outer diameter of 1000mm, a design requirement of 40s operating time, and a thrust of 300kN. The thrust requirement is as follows: Figure 2 As shown.

[0121] The optimization problem shown in the following equation is constructed to minimize the overall engine mass M(x) to improve engine performance while ensuring that the engine thrust performance meets the requirements:

[0122] min: M(x)

[0123] stt min ≤t(x)≤t max

[0124]

[0125] Where x is the design variable, M(x) is the overall mass of the engine, t(x) is the engine operating time, and t min and t max F represents the lower and upper limits of working hours. aver (x) represents the average thrust. and These represent the lower and upper limits of the average thrust.

[0126] The design variables and constraints are set in Table 1, and the configuration diagram of the front and rear wing cylindrical charge is shown in Table 1. Figure 3 . Figure 3 fw and bw represent the width of the forewing and aft wing, df and db represent the diameters of the circumcircles of the forewing and aft wing, L4 and L12 represent the lengths of the forewing and aft wing, respectively, and δ, β, α, and γ represent the forewing leading edge slope, forewing trailing edge slope, and aft wing leading edge slope, respectively, and D... p d is the outer diameter of the explosive charge. p This is the initial diameter of the inner hole for loading the explosive.

[0127] Table 1 Design Variables and Constraint Settings

[0128]

[0129]

[0130] First, for 500 sample points of the 1400mm engine in the source domain, prior knowledge of the target domain is obtained by mapping performance and mass to the target domain through constant combustion rate mapping, and a prior low-precision radial basis model is constructed. A high-performance design domain on the target domain is then selected using the variable combustion rate mapping method (i.e., step 2). Inaccurate sampling using constraint dynamic relaxation is then employed, and an error model is constructed to update and correct the low-precision model, completing the efficient design of the 1000mm engine. The specific steps are as follows:

[0131] The constant burning rate mapping maps 500 sample points from the source domain to the target domain and constructs a low-precision radial basis model for the target domain.

[0132] Variable combustion rate mapping is used to screen the high-performance design domain of target engines.

[0133] An inaccurate sampling method with constrained dynamic relaxation is used for sampling simulation. An error radial basis model is constructed to update and correct the low-precision model, thus completing the construction and continuous updating of the multi-precision model of the target domain.

[0134] If the optimization converges, the solid rocket motor design is considered complete, and the solid rocket motor design result corresponding to the current optimal solution is output.

[0135] Specifically, 500 samples of a 1400mm diameter engine from the source domain are first mapped to a 1000mm diameter engine in the target domain. The design space enveloped by these 500 source domain engine samples is used as the initial design domain. However, the initial design domain is quite large, requiring significant resources to explore during optimization, resulting in wasted computational resources. By applying a high-performance design domain acquisition method, the design domain with lower performance is reduced, saving optimization costs. The reduced high-performance design domain is shown in Table 2.

[0136] Table 2 High-performance design domain selection

[0137]

[0138]

[0139] To verify the guiding effect of the low-precision model on optimization, a solid rocket motor migration optimization method driven by historical knowledge proposed in this embodiment was used to design a solid rocket motor within a relatively large initial design domain. The initial sample points were 20, and the convergence condition was set to a maximum of 300 simulations. To verify the guiding effect of the mapped samples on optimization, an advanced surrogate model-based optimization method was used for design comparison within the same design domain. The initial sampling was set to 20, and the maximum number of simulations was set to 300. The difference was that the low-precision model was not constructed using historical case mapping. The convergence plots of the two optimizations are shown below. Figure 4 As shown. Due to the large initial design domain and the lack of prior knowledge guidance, the advanced surrogate model-based optimization method failed to find a feasible domain in 300 iterations. Figure 4 (a)), and the migration optimization method proposed in this embodiment, guided by historical knowledge, found a large number of feasible solutions and gradually converged toward the optimal solution, indicating that the historical mapping cases successfully guided the optimization of the target case.

[0140] To illustrate the effectiveness of the improved performance design domain acquisition method in this embodiment, the solid rocket motor migration optimization method was applied to optimize the high-performance design domain selected in Table 2. The initial sampling was also set to 20. If the optimal solution was not updated within 15 iterations, it was considered converged. The results are as follows: Figure 5 As shown in Table 3, the proposed method converges to an engine with an overall mass of 5493.17 kg, an operating time of 40.003 s, and an average thrust of 299.037 kN in only 60 iterations, thanks to the application of high-performance design domain and historical case knowledge. Other design parameters are shown in Table 3, and the thrust results of the final matched design are as follows. Figure 6 As shown.

[0141] Table 3 Optimal Solution Design Parameters

[0142] Forward wing length / mm 88.29 Forewing outer circle diameter / mm 655.65 Maximum pressure in combustion chamber / MPa 15.88 Combustion chamber circular section length / mm 4117.8 Wing width / mm 48.59 Rear wing length / mm 577.31 Rear wing circumscribed circle diameter / mm 454.19 Leading edge angle of the forewing / ° 71.64 Forewing trailing edge slope / ° 31.59 Aft wing leading edge slope angle / ° 12.24 Aft wing trailing edge slope angle / ° 48.21 Burning rate / mm / s 7.91 expansion ratio 11.70 Initial diameter of the charge bore / mm 285.71 Working time / s 40.003 Average thrust at room temperature / kN 299.037 Total impulse / kN·s 11962.38 Throat diameter / mm 149.652

[0143] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A solid rocket motor migration optimization method based on data knowledge fusion, characterized in that, Includes the following steps: Step 1: Under the first mapped combustion rate, the performance characteristics and quality characteristics of each source domain engine are mapped to the target domain engine to obtain several low-precision sample points, wherein the first mapped combustion rate is the same as the combustion rate of the source domain engine. For source domain engines The process of mapping its performance characteristics to the target domain engine is as follows: Source Domain Engine Mapping engine in the target domain outer diameter Source Domain Engine Mapping engine in the target domain The first mapped burn rate ,in, To meet the outer diameter requirements of the target domain engine, For source domain engine Burning rate; Computational mapping engine With source domain engine Size ratio and combustion rate ratio ,for , ,in, For source domain engine Outer diameter; By size ratio and combustion rate ratio Estimated mapping engine Heyuan Domain Engine The working hours ratio is ,in, For mapping engine Estimated working time, For source domain engine Working hours; By size ratio Estimated mapping engine Heyuan Domain Engine The total impulse ratio is ,in, For mapping engine The estimated total impact, For source domain engine Total impact; The mapped engine is obtained by using the total impulse ratio and the operating time ratio. Heyuan Domain Engine The thrust ratio is ,in, For mapping engine The estimated thrust, For source domain engine The thrust; Step 2: Under the second mapped combustion rate, calculate the mapped thrust-time curve of each source domain engine on the target domain engine, and obtain the high-performance design domain of the target domain engine based on the error between the mapped thrust-time curve and the required thrust-time curve of the target domain engine. The second mapped combustion rate is obtained by back-calculation based on the working time requirement of the target domain engine and the working time of the source domain engine. For source domain engines The calculation process of its mapped thrust-time curve on the target domain engine is as follows: Source Domain Engine Mapping engine in the target domain outer diameter Source Domain Engine Mapping engine in the target domain working hours ,in, To meet the outer diameter requirements of the target domain engine, The operating time requirements of the target domain engine; Based on mapping engine The outer diameter, operating time, and source domain engine outer diameter Burning speed Obtain the source domain engine The second mapping burn rate in the process of mapping to the target domain is: in, For source domain engine The second mapping burn rate during the mapping to the target domain process; Second mapping burn rate Below, calculate the mapping engine. With source domain engine The ratio of working time to thrust ratio is used to obtain the source domain engine. Mapped thrust-time curves on the target domain engine; Step 3: Based on the low-precision sample points and the high-performance design domain, iteratively optimize the target domain engine to obtain the design result of the target domain engine.

2. The solid rocket motor migration optimization method based on data knowledge fusion according to claim 1, characterized in that, In step 1, for the source domain engine The process of mapping its quality characteristics to the target domain engine is as follows: Based on size ratio Computational mapping engine Heyuan Domain Engine The combustion chamber mass ratio is: in, , respectively mapping engine Source Domain Engine Combustion chamber quality, , Source Domain Engine The tensile strength and density of the combustion chamber material, , respectively mapping engine The tensile strength and density of the combustion chamber material; Based on size ratio Estimated mapping engine Heyuan Domain Engine The charge mass ratio is: ,in, , respectively mapping engine Source Domain Engine The quality of the explosive charge; By size ratio and combustion rate ratio Computational mapping engine Heyuan Domain Engine The nozzle mass ratio is: in, , respectively mapping engine Source Domain Engine The quality of the nozzle, , respectively mapping engine Source Domain Engine The density of the nozzle material; By size ratio and combustion rate ratio Computational mapping engine Heyuan Domain Engine The insulation layer mass ratio is: in, , respectively mapping engine Source Domain Engine The quality of the insulation layer, , respectively mapping engine Source Domain Engine The density of the insulation layer material.

3. The solid rocket motor migration optimization method based on data knowledge fusion according to claim 1 or 2, characterized in that, In step 2, the process of obtaining the high-performance design domain of the target domain engine based on the error between the mapped thrust-time curve and the demand thrust-time curve of the target domain engine is as follows: Calculate the error between the mapped thrust-time curve of the mapped engine corresponding to each source domain engine and the required thrust-time curve of the target domain engine; Sort the errors in ascending order and select the top-ranked ones. % of the mapped engines, and select the mapped engine with the smallest mass whose total stroke is greater than or equal to the total stroke requirement of the target domain engine as the central sample; Calculate the distances between all mapped samples and the center sample, sort them in ascending order, and select the top-ranked samples. The design space of the % mapping sample envelope is used as the high-performance design domain of the target domain engine.

4. The solid rocket motor migration optimization method based on data knowledge fusion according to claim 3, characterized in that, For source domain engines The specific error between the mapped thrust-time curve of the corresponding mapped engine and the required thrust-time curve of the target domain engine is as follows: in, For source domain engine Mapping engine in the target domain The error between the mapped thrust-time curve and the target domain engine's required thrust-time curve. For source domain engine Mapping engine in the target domain The mapped thrust-time curve at time point The thrust value on top The thrust-time curve of the target domain engine at time point The thrust value on top This represents the number of discrete points in time.

5. The solid rocket motor migration optimization method based on data knowledge fusion according to claim 1 or 2, characterized in that, In step 3, the iterative optimization process for the target domain engine is as follows: Step 301: Collect several high-precision sample points in the high-performance design domain, store each high-precision sample point in the elite archive, and construct a low-precision model of the target domain engine based on each low-precision sample point. Step 302: Construct an error model based on the prediction error of the high-precision sample points in the elite archive and the low-precision model, and construct a multi-precision prediction model of the target domain engine based on the low-precision model and the error model. Step 303: Determine the sampling criteria based on the feasibility of each high-precision sample point in the current elite archive, and search for a new high-precision sample point that satisfies the sampling criteria; Step 304: Based on the multi-precision prediction model, determine whether the new high-precision sample point is better than the worst high-precision sample point in the current elite archive: If so, delete the worst high-precision sample point in the current elite file, update the elite file with new high-precision sample points, and then proceed to step 305. Otherwise, proceed to step 305; Step 305: Determine whether the optimal high-precision sample point in the elite archive has not been updated for M consecutive times: If so, the optimal high-precision sample point in the current elite archive is taken as the design result of the solid rocket motor and output; Otherwise, return to step 302.

6. The solid rocket motor migration optimization method based on data knowledge fusion according to claim 5, characterized in that, In step 303, the process of determining the sampling criteria based on the feasibility of each high-precision sample point in the current elite archive is as follows: When all high-precision sample points in the elite archive are not feasible, the sampling criterion is to find high-precision sample points with smaller constraint violation values; When the high-precision sample points in the elite archives are not feasible, the sampling criterion is to find high-precision sample points with smaller violation constraint values ​​and better objective function values; When all sample points in the elite archive are feasible, the sampling criterion is to find high-precision sample points with better objective function values.

Images