Quantum-inspired adaptive weighted method and system for solving multi-scale combustion operators of an aero-engine and storage medium

Abstract

The application discloses a quantum-inspired aero-engine combustion multi-scale operator adaptive weighting and numerical solving method and system and a storage medium, and belongs to the technical field of computational fluid dynamics and combustion numerical simulation. The method comprises the following steps: in the combustion numerical simulation process, control equations of an aero-engine combustion process are decomposed into multiple physical operators; in the numerical iteration process, contribution characteristic quantities of each physical operator are calculated online; a quantum-inspired weight state vector is constructed based on the contribution characteristic quantities; the weight state is adaptively updated according to the contribution characteristic quantity change; the solving strategy of each physical operator is adaptively scheduled based on the updated weight state; and time advancing is completed and results are output. The application can dynamically respond to the change of a dominant physical mechanism in the combustion process, significantly improves the calculation efficiency and numerical stability under the premise of ensuring the numerical precision and conservation.

Description

Technical Field

[0001] This invention belongs to the field of computational fluid dynamics and combustion numerical simulation technology, specifically involving a quantum-inspired adaptive weighted and numerical solution method, system and storage medium for multi-scale operators in aero-engine combustion. Background Technology

[0002] The flow and combustion processes within aero-engine combustion chambers exhibit typical characteristics of strong coupling across multiple time and scales, involving the simultaneous action of various physical processes such as convective transport, molecular diffusion, turbulent mixing, chemical reactions, and heat release. To address such problems, engineering and research typically employ computational fluid dynamics (CFD) numerical solution methods based on the discretization of governing equations, combined with appropriate time-progression strategies to complete transient evolution calculations.

[0003] In existing combustion CFD solutions, common approaches include using explicit or implicit time-progression methods, operator splitting to reduce coupling complexity, and adaptive time-step adjustment strategies based on error or stability constraints. However, in reactive flow calculations, because chemical reaction source terms typically exhibit significant rigidity, solutions often require smaller time steps or more stringent iterative convergence control. Furthermore, in regions of strong shear or high Reynolds number turbulence, convection terms and turbulent mixing-related operators are equally sensitive to numerical stability and accuracy. Therefore, within the same computational domain, at the same time, in different regions, or at different operating stages (such as ignition, two-stage ignition, flame stabilization, and lean / rich fuel transition), the dominance of different physical processes on the solution can vary significantly.

[0004] Commonly used "fixed strategies" or adaptive strategies focused solely on local errors / stability in existing technologies often fail to simultaneously address: solution efficiency under different operating conditions and dominant physical mechanisms, numerical stability and convergence in strongly coupled and rigid stages, and dynamic response capability to differences in contributions from different physical operators. Furthermore, while some existing methods can reduce the computational burden of chemistry to some extent by dynamically selecting subsets of chemical mechanisms or adjusting chemical integration strategies, these methods typically focus on the pruning and integral control of the chemical subsystem itself, making it difficult to cover the dominant changes in operators such as convection, diffusion, and turbulent mixing at different stages.

[0005] Therefore, there is an urgent need for a numerical solution method for multi-scale and multi-physical processes of combustion, which can evaluate the relative contributions of each physical operator online and realize adaptive weighting and solution strategy scheduling across operator levels to improve computational efficiency and enhance stability and robustness under complex conditions. Summary of the Invention

[0006] The purpose of this invention is to address the problem of balancing computational efficiency and numerical stability caused by the strong coupling of multi-scale and multi-physics processes in numerical simulation of aero-engine combustion. It provides a quantum-inspired adaptive weighting and numerical solution method, system, and storage medium for multi-scale operators in aero-engine combustion, so as to achieve online response to changes in the contribution of multi-physics operators and dynamic optimization allocation of computational resources.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a quantum-inspired adaptive weighted summation and numerical solution method for multi-scale operators in aero-engine combustion, comprising the following steps: Establishment of operator decomposition and solution framework for multi-physics governing equations: In the combustion numerical simulation process, the governing equations describing the combustion process of aero-engines are decomposed into operators according to different physical processes, forming a solution framework composed of multiple physical operators. The physical operators include at least convection transport operators, molecular diffusion operators, chemical reaction source term operators, and turbulence or subgrid-scale model operators.

[0009] Online calculation of operator contribution characteristics: During numerical iteration, the contribution characteristics of each physical operator are calculated online to characterize the degree of influence of the operator on the evolution of the numerical solution, such as the relative dominance of the operator on the evolution of the numerical solution under the current operating conditions, at the current time step, or within the current spatial region. The contribution characteristics may consist of one or more of the following: the magnitude of the change in solution variables before and after the operator's action, the operator residual or conservation error index, the spectral energy or modal intensity information related to the operator, and characteristics related to the time scale, stiffness, or local nonlinear intensity.

[0010] Construction of Quantum-Inspired Weighted States: Based on the aforementioned operator contribution characteristics, a weighted state vector representing each physical operator is constructed, where each physical operator corresponds to a weight component. The construction of the weighted states draws inspiration from the idea in quantum mechanics that state amplitudes characterize the contribution of system components, treating different physical operators as quantum-like state components. The weighted states satisfy preset constraints through normalization and other methods to ensure the stability and consistency of the numerical solution process.

[0011] Adaptive Evolution and Update of Weighted States: During numerical time progression, the weighted states are adaptively updated based on changes in operator contribution characteristics, allowing them to dynamically adjust with the evolution of the combustion process and changes in the dominant physical mechanisms. Weight updates can be performed at each time step or within a predetermined time interval, or when specific triggering conditions are met. These triggering conditions include sudden increases in residuals, increased stiffness, or switching of the dominant operator. Monotonicity, consistency, or smoothness constraints can be introduced into the weighted evolution process to avoid numerical oscillations or instability.

[0012] Operator Adaptive Weighting and Solution Strategy Scheduling Based on Weight State: After obtaining the updated weight state, the numerical solution strategy for each physical operator is adaptively scheduled based on this weight state. The adaptive scheduling methods include, but are not limited to: assigning different time sub-steps to different physical operators, selecting between explicit and implicit solution methods, dynamically setting the iteration count or convergence tolerance for each operator, and adjusting the solution order of multi-operator splitting. Within a time-progression loop, the solution strategy scheduling includes at least two of the following scheduling methods simultaneously: allocation of time sub-steps and sub-step sizes, switching between explicit and implicit solution methods, allocation of iteration counts or convergence tolerance, and adjustment of the operator splitting execution order.

[0013] The process of advancing the time step and outputting numerical results is as follows: After solving the operator based on weighted state scheduling, the flow field and combustion field variables are updated, and the process proceeds to the next time step. The above calculation of operator contribution characteristics, weighted state updates, and solution strategy scheduling constitute a closed-loop control flow. This process is repeated until the numerical simulation of the entire aero-engine combustion process is completed and the corresponding calculation results are output.

[0014] Secondly, the present invention provides a quantum-inspired adaptive weighted and numerical solution system for multi-scale operators in aero-engine combustion, comprising: The operator decomposition module is used to decompose the control equations of the combustion process in aero-engines into multiple physical operators; The contribution evaluation module is used to calculate the contribution characteristics of each physical operator online. The weight construction module is used to construct a weighted state vector representing each physical operator based on the contribution feature. The weight update module is used to adaptively update the weight state based on changes in the contribution feature. The scheduling and execution module is used to adaptively schedule the numerical solution strategy of each physical operator based on the updated weight state. The adaptive scheduling method includes at least two of the following scheduling methods: time substep number and substep size allocation, explicit and implicit solution method switching, iteration number or convergence tolerance allocation, and operator split execution order adjustment. The iterative control module is used to control the time propagation cycle until the numerical simulation of the entire aero-engine combustion process is completed and the calculation results are output.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention evaluates the contribution characteristics of different physical operators in combustion numerical simulation online and adaptively schedules the operator solution strategy based on a quantum-inspired weighted state evolution mechanism. This enables dynamic response to the dominant changes in multi-scale and multi-physical processes, and rationally allocates computational resources under different operating conditions, different time stages, and different spatial regions, thereby improving the overall solution efficiency and numerical stability under complex combustion conditions.

[0017] 2. The operator-level adaptive weighting and scheduling method adopted in this invention starts from the overall solution process and comprehensively considers the relative influence of various physical processes such as convection, diffusion, chemical reaction and turbulence on the evolution of numerical solutions. Compared with the existing technology that only performs dynamic pruning or integral control on the chemical subsystem, it can coordinate the numerical advancement relationship between different physical operators at a higher level.

[0018] 3. This invention introduces a quantum-inspired weighted state representation method, which maps the relative importance of different physical operators to the same weighted state space. Operator scheduling decisions are realized through the adaptive evolution of weighted states. Compared with traditional adaptive strategies based on empirical thresholds or single error indices, it has better continuity and consistency, and improves the robustness of numerical simulation of complex combustion processes.

[0019] 4. The method proposed in this invention does not rely on real quantum computing hardware. It introduces the idea of ​​quantum-inspired weight evolution under the classical computing framework, which has good engineering feasibility, is compatible with the operator splitting framework of existing combustion CFD solvers, and is easy to integrate into existing numerical computing processes. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the quantum-inspired adaptive weighted numerical solution method for multi-scale operators in aero-engine combustion, as described in an embodiment of the present invention. Detailed Implementation

[0022] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0023] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Example 1 This embodiment provides a quantum-inspired adaptive weighted numerical solution method for multi-scale operators in aero-engine combustion, applied to transient combustion numerical simulation within the aero-engine combustion chamber. The numerical solution object is a discrete system comprising mass conservation equations, momentum conservation equations, energy conservation equations, and multi-component transport equations; the solution process includes at least numerical operators corresponding to convective transport, molecular diffusion, chemical reaction source terms, and turbulence (or sub-grid scale) models. The computing platform can be a general-purpose CPU / GPU cluster, and the solver can employ a finite volume, finite difference, or finite element discrete framework; this embodiment of the invention does not impose any limitations on these aspects.

[0025] like Figure 1 As shown, the method described in this embodiment includes the following steps: Step 1: Establishing the operator decomposition and solution framework for multiphysics governing equations In the numerical simulation of combustion, the control equations describing the combustion process of aero-engines are first decomposed into operators according to different physical processes, forming a solution framework composed of multiple physical operators.

[0026] Specifically, at the start of the calculation, each term in the governing equations is decomposed into an operator according to the physical mechanism it describes, resulting in a set of physical operators. Where N is the total number of operators, and each operator These correspond to processes such as convective transport, diffusion, chemical reaction source terms, and turbulence models. In this embodiment, the physical operators include: a convective transport operator (describing the convective transport of mass, momentum, energy, and components), a molecular diffusion operator (describing molecular viscous diffusion and heat conduction), a chemical reaction source term operator (describing component formation and energy release caused by chemical reactions), and a turbulence model operator or a subgrid-scale model operator (describing subgrid-scale or Reynolds stress effects). Each physical operator corresponds to a different discretization method and time progression process during numerical solution, collectively forming the basic computational unit for combustion numerical simulation.

[0027] To ensure stable solution startup, initial weights are set for each physical operator. The initial weights can take the same value (e.g., all are constants). Alternatively, different initial values ​​can be given based on empirical rules; at the same time, the weight update cycle (e.g., update once per time step or once every several time steps) and the constraints on weight changes (e.g., upper and lower limits of weights, smoothing coefficients, etc.) can be set to control the frequency and magnitude of weight adjustments.

[0028] This step decomposes the complex combustion control equations into multiple operators with clear physical meanings, laying the foundation for subsequent independent evaluation and differentiated scheduling of each physical process, enabling computational resources to be optimally allocated based on the actual contribution of different physical processes.

[0029] Step 2: Online calculation of operator contribution features During the numerical iteration process, the contribution feature quantity corresponding to each physical operator is calculated online to characterize the degree of influence of the operator on the evolution of the numerical solution, that is, the relative contribution of the operator to the evolution of the numerical solution in the current working condition, the current time step, or the current spatial region. The contribution feature quantity is the basis for the subsequent construction of weighted states, and its value directly reflects the dominance of the corresponding operator in the current stage.

[0030] Specifically, during each time step, for each operator... Before and after the operator update, the changes in solution variables (such as velocity, pressure, temperature, composition, etc.) are recorded, forming the input of the operator's contribution feature quantity. The output of the contribution feature quantity is a non-negative scalar or vector index. , used to characterize the operator The relative influence of the current time step (or a local spatial region) on the solution evolution. The contributing features include one or more of the following: the magnitude of the change in solution variables before and after the operator's action, the operator residual or conservation error index, the spectral energy or modal intensity information related to the operator, and features related to the time scale or rigidity.

[0031] In this embodiment, the contributing feature quantity The following combination of indicators is used to comprehensively evaluate the operator contribution: a) The norm of the state increment caused by the operator action:

[0032] in, Indicates due to operator The increment of the solution state vector (such as certain conserved variables or original variable components) caused by the action; This represents the selected norm (e.g., the Euclidean norm) used to measure the magnitude of the increment. This metric directly reflects the magnitude of the operator's correction to the solution; the larger the correction magnitude, the greater the contribution of the operator to the evolution at the current time step.

[0033] b) Norm or conservation error index of the residuals of the discrete equations corresponding to the operator:

[0034] in, Operator The residual vector (or conserved error) of the discrete equation at the current time step, where the norm can be chosen as the 2-norm, ∞-norm, etc., to characterize the overall magnitude of the residual or error. This index reflects the degree of equilibrium of the physical process described by the operator; the larger the residual or error, the more significant the change in the process, and the greater its contribution to the evolution of the solution.

[0035] c) Indicators reflecting rigidity or time scale:

[0036] in, For the current time step, For operators The timescale of local features related to the described process. When When the value is close to or greater than 1, it indicates that the process is relatively "rigid" (i.e., the local feature time scale is smaller than or close to the current time step), and a more robust strategy is needed to solve it.

[0037] d) Indicators related to spectral energy or modal intensity: By performing Fourier spectral analysis or modal decomposition on the solution field, the energy of a specific frequency band is considered as an indicator. Then:

[0038] in, Indicates the first At each time step, the operator The energy measure of the corresponding process within the local spectral range. Similarly, the change in a specific mode coefficient can also be used as an indicator. This indicator reflects the energy of the operator. The effect on the spectral composition or modal structure of the solution.

[0039] The final contribution feature can be obtained by weighting the above four indicators. By comprehensively evaluating the contribution features in multiple dimensions, the actual contribution of each physical operator in the current numerical state can be captured more comprehensively and accurately, avoiding the evaluation bias that may be caused by a single indicator, and providing a reliable basis for the accurate construction of subsequent weighted states.

[0040] Step 3: Construction of Quantum-Inspired Weighted States Based on the contribution feature quantity obtained in step 2 Where N is the total number of operators, a weighted state vector representing the relative contribution of each physical operator is constructed. Each physical operator corresponds to a weighted component. The construction of the weighted states draws on the idea in quantum mechanics that state amplitudes characterize the contribution of the system's composition. Different physical operators are regarded as quantum-like state components, and their weights reflect the relative contribution of the corresponding operators in the current numerical state.

[0041] The specific construction process is as follows: First, to avoid the impact of differences in the units of measurement of different indicators on the weight update, Dimensionless or normalized processing is performed. In this embodiment, a characteristic dimension is selected for each index. ,Will Dimensionless transformation (if If a quantity already has a unified dimension or is dimensionless, it can be used directly.

[0042] Secondly, the weights of each operator are calculated based on the processed contribution features. To standardize the weight scale and facilitate comparison, the weights can be normalized and necessary constraints can be imposed. Specifically, the temporary weights of each operator are made to have a monotonically increasing relationship with their contribution features: a monotonically increasing function is defined. (For example, in the form of an exponential or power function), the contributing features are mapped to temporary weights: The monotonically increasing function ensures that the larger the contributing feature, the larger the corresponding temporary weight, which conforms to the consistency rule that weight increases with contribution.

[0043] Finally, the final weights are obtained through normalization, and normalization constraints are applied to the weight states:

[0044] After the above normalization process, it can be guaranteed that the sum of all weight components satisfies (That is, normalized to the preset constant 1).

[0045] It should be noted that adjustments can be made as needed. Other constraints are imposed to satisfy preset conditions, such as restricting each weight component to a given upper and lower limit range. Alternatively, the rate of change of weights between adjacent time steps can be constrained to maintain smoothness or monotonicity, thereby avoiding numerical oscillations caused by frequent and drastic switching of the solution strategy. Or, the maximum change in weights at each time step can be specified, or a gradual update strategy can be adopted for the weights to ensure relatively stable weight evolution.

[0046] To embody quantum-inspired ideas, this embodiment further employs a "quasi-amplitude" construction method:

[0047] in, For class magnitude variables, weights In form, it corresponds to the squared magnitude of the quantum-like state amplitude, thus reflecting the relative importance of each operator to the evolution of the current numerical state in the form of "similar probability amplitude (and its normalization constraint)".

[0048] It should be noted that the above magnitude construction is only an illustrative implementation of quantum-inspired ideas. Those skilled in the art can also use other equivalent weighted state construction methods or magnitude-weight mapping methods (such as a type of normalization constraint that maintains global constraints and comparability, modulus square mapping or its equivalent transformation) to achieve the same weighted state representation and comparison effect. The embodiments of the present invention do not limit this.

[0049] This step utilizes a quantum-inspired weighted state construction method to uniformly map the relative contributions of different physical operators to the same weighted state space, and applies normalization constraints to ensure the consistency of the weighted states. Compared to traditional adaptive strategies based on empirical thresholds or single error indices, this method offers better continuity and comparability, providing a unified decision-making basis for subsequent operator scheduling.

[0050] Step 4: Adaptive Evolution and Update of Weighted States During numerical time progression, the weight state is adaptively updated based on changes in the operator contribution characteristics, allowing the weight state to dynamically adjust as the combustion process evolves and the dominant physical mechanisms change. This adaptive update of the weight state can be performed periodically at each time step or within a predetermined time interval, or it can occur when specific triggering conditions are met.

[0051] This embodiment employs a periodic update method, meaning the weights are recalculated at each time step. Additionally, an event-triggered update mechanism is included as a supplementary measure: an extra update is triggered when preset trigger conditions are met. These trigger conditions include, but are not limited to: Sudden increase in residuals: A sudden increase in residuals at a certain step, such as when the residuals at a certain step exceed a preset multiple (e.g., 3 times) of the residuals at the previous time step. Increased rigidity: Local rigidity is significantly increased, such as the characteristic timescale of chemical reaction operators being significantly reduced; The dominant operator changes: the relative magnitudes of the components in the weighted state vector change significantly (e.g., the largest weighted component changes).

[0052] The aforementioned event triggering mechanism can adjust the weights in a timely manner when needed, thereby enhancing the robustness and efficiency of the solution.

[0053] Because the weight evolution process follows the consistency rule of "increased contribution, increased weight": that is, if the relative contribution of an operator to the solution evolution increases in the current stage, the corresponding weight should increase accordingly, and vice versa. In practical implementation, to prevent drastic changes in weights between adjacent time steps, a smoothing constraint is introduced during the adaptive update process of the weight state to adjust the update, so that the weight state evolves smoothly between adjacent time steps, thereby achieving gradual evolution. In this embodiment, the smoothing constraint is implemented using a gradual update formula:

[0054] in, The weights of the previous time step. The new, unsmoothed weights are calculated based on the contribution features at the current time step. For smoothing coefficients, .

[0055] In this embodiment, the following is taken =0.5, so that the new weights integrate historical and current information. By selecting a smaller value... This allows the weights to gradually approach the new value over multiple steps, avoiding sudden, drastic changes. When When the value is 1, it means that smoothing is not performed, and the new weights calculated in the current time step are used directly. The smoothing constraint can effectively avoid frequent switching of the solution strategy due to drastic changes in weights, thereby avoiding numerical oscillations.

[0056] This step combines periodic updates with event triggering, ensuring timely responses of the weight state to changes in the dominant physical mechanism while avoiding unnecessary frequent updates. The introduction of smoothing constraints further guarantees the stability of weight evolution, making scheduling decisions continuous and thus improving the robustness of numerical simulations of complex combustion processes.

[0057] Step 5: Operator Adaptive Weighting and Solution Strategy Scheduling Based on Weight State Obtain the updated weight state Subsequently, the numerical solution strategies for each physical operator are adaptively scheduled based on this weight state, and the advancement calculations of each operator are executed accordingly. The adaptive scheduling methods include, but are not limited to: assigning different time sub-steps to different physical operators, selecting between explicit and implicit solution methods, dynamically setting the iteration count or convergence tolerance for each operator, and adjusting the solution order for multi-operator splitting. Within a time-advancement loop, the solution strategy scheduling method must simultaneously include at least two of the following scheduling methods: allocation of time sub-steps and sub-step sizes, switching between explicit and implicit solution methods, allocation of iteration counts or convergence tolerances, and adjustment of the operator splitting execution order.

[0058] The specific scheduling methods are as follows: a) Allocation of time substeps and substep lengths according to Different time substeps are assigned to different operators, so that operators with larger weights are assigned to finer time advances, while operators with smaller weights use relatively larger substeps.

[0059] In this embodiment, the allocation of time substeps and substep sizes is performed as follows:

[0060] in, The preset total number of sub-steps (K=10 in this embodiment). For the first The weights of each operator To be assigned to the Number of substeps for each operator For the floor operator, For the global time step, For the first The local time substep size of each operator.

[0061] Since the weights have been normalized to quantities that sum to 1, This is the expected number of sub-steps allocated according to the weighted proportions. Rounding up ensures... The value is an integer, ensuring that at least the preset total number of sub-steps is calculated.

[0062] This allocation method allows operators that contribute significantly (such as rigid chemical reaction source terms) to receive finer time advances (smaller substeps), thus ensuring numerical stability; while operators that contribute less use larger substeps to reduce computational overhead, thereby improving overall computational efficiency while ensuring numerical accuracy and stability.

[0063] b) Allocation of iteration count or convergence tolerance The global iterative computation budget or convergence tolerance is allocated among different operators according to their weights, so that the dominant operators receive more iterative resources or stricter convergence requirements (i.e., smaller error tolerances), while the minor operators are appropriately relaxed to optimize the utilization of computational resources.

[0064] In this embodiment, the number of iterations is allocated as follows:

[0065] in, This represents the maximum number of iterations allowed globally (in implicit or nonlinear iteration processes). For the first The weights of each operator For the first The weights of each operator The total number of operators, To be assigned to the The upper limit of the number of iteration steps for each operator.

[0066] When the weights are normalized to a sum of 1, the above formula simplifies to: .

[0067] Similarly, for the convergence criterion of implicit iteration, different convergence tolerances can be assigned to different physical processes based on the weight magnitude. For example, let the global convergence tolerance be... Then the convergence tolerance of the i-th operator can be set as: This makes operators with larger weights subject to stricter convergence requirements (smaller convergence limits). This ensures that the main processes are solved more accurately.

[0068] This allocation method prioritizes the allocation of limited computing resources to the currently dominant physical processes, avoiding the waste of too many iterations on secondary processes, thereby significantly improving computational efficiency while ensuring overall solution accuracy.

[0069] c) Switching between explicit and implicit solution methods When an operator has a large weight, or its corresponding characteristic index indicates an increased risk to stability, an implicit time-progression method (or increased implicit iteration intensity) is used to enhance stability. Conversely, when an operator has a small weight and numerical stability allows, explicit time-progression is used to reduce computational overhead. For example, for the chemical reaction operator with the highest weight, the solution can be switched from explicit Euler method to implicit BDF scheme to avoid numerical instability due to excessive contribution.

[0070] The explicit and implicit adaptive switching enables the solution method to be dynamically adjusted according to the current characteristics of the physical process. The implicit method is used to ensure stability for stability-sensitive operators (such as rigid chemical reactions), while the explicit method is used to improve efficiency for non-rigid operators, thus achieving a balance between stability and efficiency.

[0071] d) Adjustment of operator splitting execution order The execution order of operators in the splitting solution is dynamically adjusted based on the current weight state. Generally, operators that are more sensitive to numerical stability or error propagation can be executed first or later to suppress the accumulation and amplification of errors.

[0072] In this embodiment, the following rule is used to adjust the order: if the weight of a certain operator is significantly higher than that of other operators in the current stage (e.g., ... A weight >0.5 indicates that the physical process plays a dominant role in the solution evolution. Therefore, this operator should be prioritized in the time step (e.g., solve this term first in the Strang split) to reduce the impact of the process error on the overall simulation. Conversely, for operators with very small weights, they can be scheduled to be executed in the position where their impact on the solution is least significant.

[0073] By adaptively adjusting the sequence, the error propagation of the dominant process can be further suppressed, thereby improving the stability and accuracy of numerical simulation.

[0074] This embodiment employs two scheduling strategies simultaneously: substep allocation and iteration budget allocation. This allows the operator contribution evaluation results to simultaneously affect both time scale control and numerical stability control. Compared to methods relying solely on a single adaptive parameter adjustment, multi-strategy joint scheduling avoids the failure of the trade-off between stability and efficiency, achieving collaborative optimization of computational resources across multiple dimensions. For example, when the weight of a chemical reaction operator increases significantly, the system not only allocates a smaller substep (time dimension) but also allocates a larger iteration budget (iteration dimension), and may switch it to implicit solution (algorithm dimension), forming a multi-dimensional collaborative guarantee to ensure numerical stability during the rigid phase.

[0075] Step 6: Complete the time progression and output the numerical results. After completing the operator solution based on weighted state scheduling, the flow field and combustion field variables are updated, and the process proceeds to the next time step. The above-mentioned operator contribution feature calculation, weighted state update, and solution strategy scheduling constitute a closed-loop control flow, which is executed at each global time step (or when the triggering condition is met) to drive the online update of the propulsion configuration of each operator.

[0076] Repeat the above process (i.e., repeat steps 2 to 5) until the numerical simulation of the entire aero-engine combustion process is completed and the corresponding calculation results are output.

[0077] The method described in this embodiment uses closed-loop feedback control to ensure that the solution strategy always matches the current dominant physical mechanism, thereby achieving dynamic optimization of computational resources. While ensuring numerical accuracy and conservation, it significantly improves computational efficiency and numerical stability under complex combustion conditions.

[0078] Example 2 This embodiment provides a quantum-inspired adaptive weighted numerical solution system for multi-scale operators in aero-engine combustion, used to implement the method described in Embodiment 1. The system includes the following modules: Operator decomposition module: This module decomposes the control equations of the aero-engine combustion process into operator decompositions based on different physical processes, forming a solution framework composed of multiple physical operators. These physical operators include at least convection transport operators, molecular diffusion operators, chemical reaction source term operators, and turbulence model operators or subgrid-scale model operators.

[0079] Contribution evaluation module: used to calculate the contribution characteristics of each physical operator online during numerical iteration. The contribution characteristics include one or more of the following: the magnitude of the change in solution variables before and after the operator's action, the operator residual or conservation error index, the spectral energy or modal intensity information related to the operator, and characteristics related to the time scale or stiffness.

[0080] The weight construction module is used to construct a weighted state vector representing the relative contribution of each physical operator based on contribution feature quantities, where each physical operator corresponds to a weight component. The weight construction module performs dimensionless or normalized processing on the contribution feature quantities and applies normalization constraints to the weighted states so that the sum of all weight components is a preset constant.

[0081] Weight update module: Used to adaptively update the weight state based on changes in contribution features. The weight update module supports both periodic update and event-triggered update modes, and introduces smoothing constraints to ensure smooth evolution of the weight state between adjacent time steps.

[0082] The scheduling and execution module is used to adaptively schedule the numerical solution strategies of each physical operator based on the updated weight state. The adaptive scheduling method includes at least two of the following scheduling methods: time substep allocation or substep size allocation, explicit and implicit solution method switching, iteration number or convergence tolerance allocation, and operator splitting execution order adjustment.

[0083] Iterative control module: Used to control the time advance loop, repeatedly calling the above modules to perform contribution evaluation, weight update and strategy scheduling, numerical simulation of the entire aero-engine combustion process and output calculation results.

[0084] The implementation methods of the above modules are described in detail in Example 1, and will not be repeated here.

[0085] Example 3 This embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in Embodiment 1. The computer-readable storage medium can be any medium capable of storing program code, including but not limited to: read-only memory (ROM), random access memory (RAM), magnetic disk, optical disk, flash memory, etc.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A quantum-inspired adaptive weighted numerical solution method for multi-scale operators in aero-engine combustion, characterized in that, Includes the following steps: Step 1: Decompose the control equations describing the combustion process of an aero-engine into operators according to different physical processes to form a solution framework composed of multiple physical operators; Step 2: Calculate the corresponding contribution feature for each physical operator online. The contribution feature is used to characterize the degree of influence of the operator on the evolution of the numerical solution. Step 3: Based on the contribution feature, construct a weight state vector representing each physical operator, where each physical operator corresponds to a weight component; Step 4: Based on the changes in the contribution feature quantity, the weight state is adaptively updated so that the weight state is dynamically adjusted as the combustion process evolves and the dominant physical mechanism changes. Step 5: Based on the updated weight state, adaptively schedule the numerical solution strategy of each physical operator. The adaptive scheduling method includes at least two of the following scheduling methods: time substep number and substep size allocation, explicit and implicit solution method switching, iteration number or convergence tolerance allocation, and operator split execution order adjustment. Step 6: After solving the operator based on weighted state scheduling, update the flow field and combustion field variables, and proceed to the next time step. Repeat steps 2 to 5 until the numerical simulation of the entire aero-engine combustion process is completed and the calculation results are output.

2. The method according to claim 1, characterized in that, The physical operators mentioned in step 1 include at least convection transport operators, molecular diffusion operators, chemical reaction source term operators, and turbulence model operators or subgrid-scale model operators.

3. The method according to claim 1, characterized in that, The contributing features mentioned in step 2 include one or more of the following: the magnitude of the change in the solution variables before and after the operator is applied, the operator residual or conservation error index, the spectral energy or modal intensity information related to the operator, and features related to the time scale or stiffness.

4. The method according to claim 1, characterized in that, In step 3, when constructing the weighted state vector, the contribution feature quantity is dimensionless or normalized, and a normalization constraint is applied to the weighted state so that the sum of all weight components is a preset constant.

5. The method according to claim 1, characterized in that, The adaptive update in step 4 adopts either periodic update or event-triggered update; the conditions for event triggering include residual burst, stiffness enhancement, or switching of the dominant operator.

6. The method according to claim 1, characterized in that, The adaptive update described in step 4 introduces a smoothing constraint, which enables the weight states to evolve smoothly between adjacent time steps. This smoothing constraint is achieved through an asymptotic update formula: in, The weights of the previous time step. The new, unsmoothed weights are calculated based on the contribution features at the current time step. For smoothing coefficients, .

7. The method according to claim 1, characterized in that, In step 5, the number of time substeps and the substep size are allocated as follows: in, The preset total number of sub-steps, For the first The weights of each operator To be assigned to the Number of substeps for each operator For the floor operator, For the global time step, For the first The local time substep size of each operator.

8. The method according to claim 1, characterized in that, The allocation of iteration counts in step 5 is performed as follows: in, This represents the maximum number of iterations allowed globally. For the first The weights of each operator For the first The weights of each operator The total number of operators, To be assigned to the The upper limit of the number of iteration steps for each operator.

9. A quantum-inspired adaptive weighted numerical solution system for multi-scale operators in aero-engine combustion, characterized in that, include: The operator decomposition module is used to decompose the control equations of the combustion process in aero-engines into multiple physical operators; The contribution evaluation module is used to calculate the contribution characteristics of each physical operator online. The weight construction module is used to construct a weighted state vector representing each physical operator based on the contribution feature. The weight update module is used to adaptively update the weight state based on changes in the contribution feature. The scheduling and execution module is used to adaptively schedule the numerical solution strategy of each physical operator based on the updated weight state. The adaptive scheduling method includes at least two of the following scheduling methods: time substep number and substep size allocation, explicit and implicit solution method switching, iteration number or convergence tolerance allocation, and operator split execution order adjustment. The iterative control module is used to control the time propagation cycle until the numerical simulation of the entire aero-engine combustion process is completed and the calculation results are output.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-8.

Images