Intelligent design simulation system for through-flow component of fluid machine

Abstract

The invention relates to the technical field of fluid machinery design, and discloses a fluid machinery through-flow component intelligent design simulation system, which comprises a multi-physics field modeling module, which is used for constructing fluid mechanics, structural mechanics and heat conduction models based on preset boundary conditions and physical attributes; the intelligent optimization decision module is used for constructing and defining a through-flow performance optimization objective function; the cooperative calculation module is used for constructing a comprehensive performance evaluation model based on the candidate design parameter set; the virtual experiment feedback module is used for reconstructing a corresponding through-flow component structure model according to the multi-target evaluation result; and the scheduling control module is used for controlling algorithm path selection and computing resource allocation in the optimization process. According to the method, a method of combining multi-physics field coupling modeling and a dynamic feedback mechanism is adopted, and design parameters are continuously adjusted through simulation data fed back in real time, so that the optimization process can be dynamically updated according to actual operation conditions, and the optimization convergence speed and the overall performance are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fluid machinery design, and particularly to an intelligent design and simulation system for fluid machinery flow components. Background Art

[0002] With the progress of technology, fluid machinery flow components are increasingly widely used in various engineering systems, especially in high-precision fields such as aerospace, energy, petrochemical, etc. The design of these components is directly related to the performance and safety of the system. Traditional fluid machinery design often relies on empirical methods, and meets the design requirements through gradual iteration and manual adjustment.

[0003] Existing multi-physics field coupling simulation methods have been applied to the design of fluid machinery flow components. Traditional computational fluid dynamics (CFD) technology can provide information such as fluid flow and pressure distribution, structural mechanics simulation helps designers predict the strength and deformation of materials, and heat conduction simulation can analyze temperature distribution. These technologies usually use the finite element method (FEM) and the finite volume method (FVM) to separately process the calculations of each physical field, helping engineers optimize the design.

[0004] However, existing fluid machinery flow component simulation technologies often ignore the coupling phenomenon, making it difficult to accurately predict the performance of flow components under actual operating conditions. Most of their optimization methods rely on manual intervention, and it is easy to fall into local optimum in the multi-objective optimization process, lacking a global perspective. Even when using advanced optimization technologies such as genetic algorithms, the initial parameters and optimization paths of these methods are usually fixed and difficult to adjust in real time according to the simulation results. Therefore, the present invention provides an intelligent design and simulation system for fluid machinery flow components to solve the deficiencies existing in the prior art. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an intelligent design and simulation system for fluid machinery flow components, which solves the problems of insufficient multi-physics field coupling modeling, inflexible optimization path adjustment, and inaccurate virtual experiment feedback in the prior art.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: An intelligent design and simulation system for fluid machinery flow components, comprising: A multi-physics field modeling module, used to build fluid mechanics, structural mechanics, and heat conduction models based on preset boundary conditions and physical properties, and output coupling simulation parameters; An intelligent optimization decision-making module, used to receive the coupling simulation parameters, build an optimization objective function defining the flow performance, and generate an optimization path and a set of candidate design parameters through a behavior-driven optimization strategy and a deep reinforcement learning method; A collaborative computing module, which is used to call a multi-physics field numerical simulator based on the set of candidate design parameters to jointly calculate the response results of each physical field, construct a comprehensive performance evaluation model, and generate multi-objective evaluation results corresponding to each candidate design parameter; A virtual experiment feedback module, which is used to reconstruct the corresponding flow component structure model according to the multi-objective evaluation results corresponding to each candidate design parameter, simulate the operating conditions in a virtual simulation environment, and generate real perception feedback data as a basis for performance verification; An adjustment module, which is used to construct a design error function based on the real perception feedback data, and by comparing the preset expected performance index with the multi-objective evaluation results corresponding to the current candidate design parameters, reversely adjust the optimization strategy and dynamically update the design parameters to drive the optimization process to iterate and converge; A scheduling and control module, which is used to dynamically schedule local optimization or global search strategies according to the iterative stage and performance convergence state of the optimization process output by the adjustment module, and control the algorithm path selection and computing resource allocation in the optimization process.

[0007] Preferably, the multi-physics field modeling module includes: A fluid mechanics modeling unit, which is used to construct a fluid flow model based on preset boundary conditions and initial design parameters through the mass conservation equation and the momentum conservation equation; A structural mechanics modeling unit, which is used to construct a structural strength model of the flow component based on the fluid flow model and in combination with material property parameters, and calculate the stress and deformation fields; A heat conduction modeling unit, which is used to construct a temperature distribution model of the flow component based on the stress and deformation fields and the fluid flow model, and generate coupled simulation parameters.

[0008] Preferably, the intelligent optimization decision-making module includes: An objective function definition unit, which is used to extract flow performance indicators, structural response indicators, and thermal management indicators based on the coupled simulation parameters, and dynamically construct a multi-objective optimization objective function; A dynamic strategy optimization unit, which is used to use the multi-objective optimization objective function as an evaluation basis, initialize the optimization path, and adopt a behavior-driven optimization strategy to adjust the candidate design parameters in real time to generate a strategy; A deep reinforcement learning unit, which is used to use the candidate design parameter generation strategy as an initial state, and through the simulation feedback interaction with the collaborative computing module, gradually train the optimization strategy network, and dynamically generate a new round of high-performance candidate design parameter sets.

[0009] Preferably, the collaborative computing module includes: A multi-physics simulation unit, which is used to call the modeling process of the multi-physics modeling module to perform joint simulation on each candidate parameter according to the newly generated set of high-performance candidate design parameters, and output a data set containing fluid, structure, and thermal field results; A performance evaluation unit, which is used to construct a comprehensive performance scoring model based on the data set containing fluid, structure, and thermal field results, and dynamically map the influence relationship between design parameters and performance indicators; A data integration unit, which is used to construct a multi-objective evaluation result corresponding to the candidate design parameters based on the influence relationship between the design parameters and the performance indicators, and dynamically adjust the weight factor of the evaluation index.

[0010] Preferably, the calling of the modeling process of the multi-physics modeling module to perform joint simulation on each candidate parameter includes: Based on the newly generated set of high-performance candidate design parameters generated by the deep reinforcement learning unit, construct a multi-physics joint simulation task list; According to the design parameters in the simulation task list, dynamically configure the inlet flow velocity boundary condition of the fluid mechanics simulation, the load distribution condition of the structural mechanics simulation, and the heat source distribution condition of the heat conduction simulation; Real-time synchronize the fluid pressure field, structural stress field, and temperature field data in the joint simulation result to the performance evaluation unit, trigger the performance evaluation model to update the weight factor, and generate a multi-objective evaluation result.

[0011] Preferably, the virtual experiment feedback module includes: A model reconstruction unit, which is used to dynamically generate a three-dimensional geometric model according to the multi-objective evaluation result corresponding to the candidate design parameters; A dynamic simulation unit, which is used to call a multi-physics simulator to simulate the operating state of the flow-through component under typical working conditions based on the three-dimensional geometric model, and the boundary conditions loaded by the multi-physics simulator are dynamically configured by the joint simulation data output by the multi-physics simulation unit; A feedback generation unit, which is used to extract pressure distribution, temperature distribution, and structural strain data from the simulation data configured by the dynamic simulation unit to construct a real perception feedback data set.

[0012] Preferably, the calling of the multi-physics simulator to simulate the operating state of the flow-through component under typical working conditions includes: According to the multi-objective evaluation result, extract the candidate design parameters with the optimal flow-through efficiency and map them to the virtual simulation environment as typical working condition parameters; Based on the three-dimensional geometric model, call a multi-physics simulator to load the typical working condition parameters and perform fluid-structure-thermal coupling transient simulation; Real-time collect the inlet pressure pulsation data, trailing edge temperature gradient data, and blade strain peak data of the flow components during the fluid-structure-thermal coupling transient simulation process, and construct a structured feedback data set.

[0013] Preferably, the adjustment module includes: An error analysis unit for comparing and analyzing the real perception feedback data generated by the virtual experiment feedback module with the flow performance optimization objective function defined in the intelligent optimization decision module to construct a design error function; A strategy update unit for adjusting the optimization strategy network of the deep reinforcement learning unit based on the result of the error function by using the backpropagation algorithm and generating new strategy parameters; A parameter iteration unit for dynamically adjusting the current candidate design parameter set based on the new strategy parameters and driving the optimization process in the intelligent optimization decision module to iterate again to form a closed-loop optimization process.

[0014] Preferably, the scheduling control module includes: A state monitoring unit for real-time monitoring the design parameter update trend and performance convergence speed of the parameter iteration unit and dynamically judging the current optimization stage state; An algorithm scheduling unit for dynamically selecting the global optimization algorithm used in the global search stage or the local optimization algorithm used in the local optimization stage according to the current optimization stage state; A resource allocation unit for dynamically adjusting the calculation resource allocation strategy of each simulation task according to the local optimization algorithm, and feeding back the resource allocation strategy to the multi-physics field simulation unit to optimize the simulation execution efficiency.

[0015] There is also provided a method for intelligent design and simulation of a flow component of a fluid machine, including the following steps: Based on preset boundary conditions and physical properties, construct a multi-physics field coupling model through the mass conservation equation, momentum conservation equation, and heat conduction equation, and output the coupling simulation parameters of fluid mechanics, structural mechanics, and heat conduction; Based on the coupling simulation parameters of fluid mechanics, structural mechanics, and heat conduction, dynamically define a flow performance optimization objective function, generate an initial candidate design parameter set through a behavior-driven optimization strategy, and trigger a deep reinforcement learning mechanism to iteratively adjust the parameter generation strategy; Based on the initial candidate design parameter set, call a multi-physics field numerical simulator to perform joint calculations, generate a response data set of fluid, structure, and thermal fields, and construct a comprehensive performance evaluation model to output a multi-objective evaluation result; According to the multi-objective evaluation results, drive the parametric modeling tool to reconstruct the three-dimensional geometric model of the flow passage component, load the typical working condition parameters in the virtual simulation environment to simulate the dynamic operation behavior, and generate real perception feedback data including pressure, temperature and strain distribution; Based on the difference analysis between the real perception feedback data including pressure, temperature and strain distribution and the preset performance indicators, construct a design error function, and update the deep reinforcement learning policy network through the backpropagation algorithm to trigger a new round of optimization iteration convergence; According to the new round of optimization iteration stage and the performance convergence state, dynamically schedule the combined strategy of the global search algorithm and the local optimization algorithm, and control the priority of computing resource allocation.

[0016] The present invention provides an intelligent design and simulation system for the flow passage component of a fluid machine. It has the following beneficial effects: 1. The present invention adopts a method combining multi-physical field coupling modeling and a dynamic feedback mechanism. By continuously adjusting the design parameters through the real-time feedback simulation data, the optimization process can be dynamically updated according to the actual operating conditions. It achieves the effect of significantly improving the optimization convergence speed and global performance. Compared with the traditional static single-field optimization method in the prior art, it solves the problem that the optimization process is prone to falling into local optima and is difficult to accurately reflect the real working conditions.

[0017] 2. The present invention introduces deep reinforcement learning and behavior-driven optimization strategies, and adopts an automatically adjusted candidate design parameter generation mechanism. By automatically exploring and real-time adjusting the optimization path, it achieves a high global convergence success rate. Compared with the optimization strategies in the prior art that are easily restricted by human experience, the strategy of the present invention has stronger dynamics and adaptability, and significantly improves the design quality and efficiency.

[0018] 3. The present invention utilizes the combination of multi-physical field simulation and virtual experiment feedback to generate real perception data in real time, constructs a new feedback control mechanism, and can accurately analyze the errors in the design process and timely adjust the optimization strategy. It achieves higher simulation accuracy and real-time performance. Compared with the traditional methods in the prior art that are difficult to efficiently process complex feedback data, it solves the problem that the simulation results are affected by noise interference, resulting in inaccurate performance judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is the system architecture diagram of the present invention; Figure 2 It is the flowchart of the method steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than 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 efforts shall fall within the protection scope of the present invention.

[0021] Please refer to the attached Figure 1 , the embodiment of the present invention provides an intelligent design and simulation system for a fluid machinery flow component, including: A multi-physics field modeling module, configured to construct fluid mechanics, structural mechanics, and heat conduction models based on preset boundary conditions and physical properties, and output coupled simulation parameters; An intelligent optimization decision-making module, configured to receive the coupled simulation parameters, construct an optimization objective function defining the flow performance, and generate an optimization path and a set of candidate design parameters through a behavior-driven optimization strategy and a deep reinforcement learning method; A collaborative calculation module, configured to, based on the set of candidate design parameters, call a multi-physics field numerical simulator to jointly calculate the response results of each physical field, construct a comprehensive performance evaluation model, and generate multi-objective evaluation results corresponding to each candidate design parameter; A virtual experiment feedback module, configured to reconstruct the corresponding flow component structure model according to the multi-objective evaluation results corresponding to each candidate design parameter, simulate the operating conditions in a virtual simulation environment, and generate real perception feedback data as a basis for performance verification; An adjustment module, configured to, based on the real perception feedback data, construct a design error function, and by comparing the preset expected performance index with the multi-objective evaluation results corresponding to the current candidate design parameters, reversely adjust the optimization strategy and dynamically update the design parameters to drive the optimization process to converge iteratively; A scheduling control module, configured to, according to the iterative stage of the optimization process and the performance convergence state output by the adjustment module, dynamically schedule local optimization or global search strategies, and control the algorithm path selection and computing resource allocation in the optimization process.

[0022] For the multi-physics field modeling module, in this embodiment, it is used to establish physical field models such as fluid mechanics, structural mechanics, and heat conduction. In order to ensure the consistency of boundary conditions and the accuracy of the model in the intelligent design task, the multi-physics field modeling module needs to first complete the initial parameter setting and regional grid division, and then implement model construction and parameter output through a physical law solver.

[0023] Generally, this module dynamically generates a multi-source input parameter set based on the preset design boundary conditions and initial design parameters, comprehensively considering the multi-physics field coupling mechanism involved in the engineering objectives, and on the premise of maintaining computational stability, including but not limited to inlet flow velocity, pressure boundary, material property distribution, initial temperature field, etc.

[0024] Specifically, the multi-physics modeling module includes the following constituent units: In a possible implementation, the fluid mechanics modeling unit is used to establish a flow field model, which is based on the mass conservation equation and the momentum conservation equation. Taking incompressible flow as an example, the mass conservation equation can be expressed as , where represents the fluid velocity vector. The momentum conservation equation is further expressed as: ; where is the fluid density; is the pressure; is the fluid velocity vector; is the body force term; is the pressure gradient, representing the rate of change of pressure with spatial position; is the Laplacian operator of the flow velocity, representing the curvature or diffusion degree of the fluid velocity field; is the time derivative of the flow velocity, representing the change of fluid velocity with time; is the non-linear term, representing the convection effect of the fluid and the change of fluid velocity. This system of equations is discretely solved by a numerical solver to obtain the spatial distributions of the velocity field and the pressure field.

[0025] As an option, the structural mechanics modeling unit is activated synchronously after the fluid modeling is completed, and it is used to analyze the stress and deformation response behavior of the components. The stress-strain model follows the linear elastic theory or the non-linear constitutive relationship.

[0026] The heat conduction modeling unit is then used to supplement the analysis of the coupled influence of the thermal field on the fluid and structural responses, and to solve the temperature field distribution based on the steady-state heat conduction or transient heat conduction equation. Under steady-state conditions, the control equation is: ; where is the thermal conductivity; is the temperature field variable; is the volume heat source term; represents the temperature gradient; is the gradient operator.

[0027] In this embodiment, the variable transfer is coordinated through the multi-field unified physical interface to realize the coupling relationship modeling between multiple physical quantities. For example, the temperature field affects the material properties of the structural mechanics module (such as the change of elastic modulus with temperature), and the structural deformation will also affect the boundary conditions of the flow field (such as the change of the flow channel).

[0028] In some embodiments, the model is solved by combining the finite element method (FEM) and the finite volume method (FVM). FEM is applicable to stress analysis and heat conduction sub-models, while FVM is used for the discretization of flow field variables.

[0029] Specifically, an automatic mesh encryption strategy can be introduced in the building block to enhance the simulation accuracy in complex regions such as vortex regions and wall boundaries. The specific encryption criterion can be dynamically triggered according to the change rates of field quantities such as velocity gradient and pressure gradient.

[0030] For three-dimensional simulation tasks, the module supports an adaptive time step adjustment mechanism to ensure computational convergence and physical time consistency. This mechanism is iteratively adjusted based on local error estimation, effectively improving the stability and accuracy of multi-field simulations.

[0031] For the intelligent optimization decision-making module, in this embodiment, it is used to receive the coupled simulation parameters output by the multi-physics field modeling module. By constructing a multi-objective optimization function and a dynamic policy generation mechanism, it realizes the intelligent screening and iterative optimization of candidate design parameters. This module maps the physical field simulation results to the design space through a data-driven approach, combines a behavior-driven optimization strategy and a deep reinforcement learning algorithm to form an adaptive decision-making framework for complex engineering problems.

[0032] The intelligent optimization decision-making module includes an objective function definition unit, a dynamic policy optimization unit, and a deep reinforcement learning unit. The objective function definition unit extracts key performance indicators based on the coupled simulation parameters and constructs a multi-objective optimization function. The dynamic policy optimization unit generates an initial optimization path according to the evaluation results of the objective function, and the deep reinforcement learning unit trains the policy network through interaction with the simulation feedback and dynamically adjusts the parameters to generate a policy.

[0033] Specifically, the objective function definition unit defines the comprehensive performance index through the following formula: ; In the formula, is the design parameter vector (such as blade geometric parameters, material properties, etc.); is the flow-through efficiency; is the maximum equivalent stress; is the highest temperature field value; is the weight coefficient, satisfying ; is the objective function. This formula normalizes the aerodynamic efficiency, structural strength, and thermal management indicators to achieve multi-objective collaborative optimization.

[0034] In some embodiments, the weight coefficient can be dynamically adjusted according to the design stage. For example, set in the stage where aerodynamic performance is prioritized; adjust to This dynamic adjustment mechanism is implemented through an external configuration file or an interactive interface.

[0035] The dynamic policy optimization unit uses a behavior-driven optimization algorithm to generate an initial candidate parameter set. Its policy function can be expressed as: ; In the formula, represents the initial policy selection function; represents the performance vector corresponding to the design parameters; is the target performance vector; represents the parameter vector that minimizes the objective function among all possible ; is the design parameter for the th iteration. This unit solves the above optimization problem through the gradient descent method or the genetic algorithm and outputs a candidate parameter set that meets the constraint conditions .

[0036] The deep reinforcement learning unit constructs a policy network based on the Actor-Critic framework. The Actor network generates parameter adjustment actions, and the Critic network evaluates the state value function , where the state includes the current design parameters, the deviation of the objective function, and the historical optimization trajectory. The network update follows the following policy gradient rule: ; In the formula, represents the parameters of the policy network (Actor); represents the policy objective function; represents the probability of taking action in state obtained based on the current policy parameters ; represents the logarithmic probability of the policy in the current action selection; represents the gradient (derivative) of the policy function with respect to the parameter under the condition of ; represents the action value function; represents the expectation under the policy (the expectation value calculation is based on the trajectory samples generated by the current policy). This unit obtains simulation feedback data through interaction with the collaborative computing module and stores the transition samples in an experience replay pool , where the reward is calculated from the improvement rate of the objective function.

[0037] As an option, the deep reinforcement learning unit gradually converges to the optimal policy at the initial stage of training. The network structure can be selected as a multi-layer perceptron (MLP) or a graph neural network (GNN), and the latter is suitable for the design parameter space with topological relevance.

[0038] Specifically, when the design parameters include the parametric control point coordinates of the blade geometry, the graph neural network encodes the spatial adjacency relationship between the control points into a graph structure, and captures the influence of local features on the global performance through a message passing mechanism. This implementation method can improve the generalization ability of the policy network to complex geometric changes.

[0039] In a possible implementation, the module introduces an early stopping mechanism to prevent overfitting. When the average reward on the validation set has not improved for 5 consecutive epochs, the training is terminated and the current optimal policy is saved. This mechanism is implemented through a monitoring window, and the window size can be adjusted according to the dataset size.

[0040] For the collaborative computing module, in this embodiment, based on the set of candidate design parameters generated by the intelligent optimization decision module, it calls a multi-physics numerical simulator to jointly calculate the response results of each physical field, constructs a comprehensive performance evaluation model, and outputs multi-objective evaluation results. This module realizes the dynamic mapping between design parameters and performance indicators through multi-physics joint simulation and data integration, and provides a verification basis for the virtual experiment feedback module.

[0041] The collaborative computing module includes a multi-physics simulation unit, a performance evaluation unit, and a data integration unit. The multi-physics simulation unit generates a joint simulation task according to the set of candidate design parameters, the performance evaluation unit constructs a scoring model based on the simulation results, and the data integration unit dynamically adjusts the output multi-objective evaluation results through weight factors.

[0042] Specifically, the multi-physics simulation unit calls numerical simulators for fluid mechanics, structural mechanics, and heat conduction for joint calculation. The fluid mechanics simulation is based on the Reynolds-averaged Navier-Stokes equations, the structural mechanics simulation solves the stress field by the finite element method, and the heat conduction simulation calculates the temperature field distribution based on Fourier's law.

[0043] The performance evaluation unit constructs a comprehensive performance scoring model through the following formula: ; In the formula, is the comprehensive score of the th group of design parameters; is the value of the th performance indicator; and are the index normalization boundaries; is the dynamic weight coefficient. The weight coefficient is dynamically adjusted according to the requirements of the design stage; represents the total number of performance indicators.

[0044] The data integration unit maps the response results of each physical field to a unified design space. In one implementation, principal component analysis (PCA) is used for dimensionality reduction to extract key performance features and reduce data redundancy. Its mathematical process can be expressed as: ; In the formula, is the original high-dimensional data set; is the eigenvector matrix; is the coordinate of the design space after dimensionality reduction.

[0045] When a certain simulation diverges due to grid distortion, the abnormal parameter combination is automatically marked and local grid encryption is triggered to ensure the simulation stability by dynamically adjusting the grid size. This mechanism is based on the residual convergence threshold and the accuracy order of the discretization format to ensure reliable calculation of complex parameter combinations.

[0046] For the virtual experiment feedback module, in this embodiment, according to the multi-objective evaluation results output by the collaborative calculation module, the flow passage component structure model is reconstructed, and the operating conditions are simulated in the virtual simulation environment to generate real perception feedback data as the basis for performance verification. This module realizes the virtual verification of candidate design parameters through dynamic modeling and multi-physical field coupling simulation, providing a basis for error analysis for the adjustment module.

[0047] The virtual experiment feedback module includes a model reconstruction unit, a dynamic simulation unit, and a feedback generation unit. The model reconstruction unit drives the parametric modeling tool to generate a three-dimensional geometric model based on the key geometric feature parameters in the multi-objective evaluation results. The dynamic simulation unit calls the multi-physical field simulator to load the boundary conditions in the joint simulation data and perform dynamic response simulations under typical operating conditions. The feedback generation unit extracts the pressure, temperature, and strain distribution data from the simulation results to construct a structured feedback data set.

[0048] Generally, the model reconstruction unit realizes the generation of the geometric model through the following steps: Analyze the blade profile control point coordinates, flow passage cross-section size, and material distribution parameters in the candidate design parameters; Construct the three-dimensional blade surface based on non-uniform rational B-spline (NURBS) curves; Discretize the geometric model into finite element meshes and associate material properties with boundary identifiers.

[0049] In one possible implementation, the dynamic simulation unit uses the following thermo-fluid-solid coupling control equations to describe the operating conditions: ; In the formula, is the fluid density; is the velocity field; is the pressure; is the dynamic viscosity; is the coefficient of thermal expansion; is the temperature field; is the reference temperature; is the gravitational acceleration vector; is the Laplacian operator of the flow velocity; is the time derivative of the flow velocity. This equation couples the influence of the temperature field on the flow and reflects the thermal convection effect.

[0050] As an option, the stress-strain behavior in the structural response analysis is described by the following constitutive relation: ; In the formula, is the stress tensor; is the elastic stiffness tensor; is the strain tensor; is the coefficient of thermal expansion; is the temperature change; is the Kronecker symbol. This equation quantifies the influence of the temperature gradient on the structural deformation.

[0051] Specifically, the dynamic simulation unit performs the following key operations: Screen the candidate parameters with the optimal flow-through efficiency according to the multi-objective evaluation results and map them to the inlet flow velocity, temperature boundary and structural constraint conditions of the virtual environment; Call the transient solver to perform fluid-structure-thermal coupling simulation, and the time step is adaptively adjusted based on the Courant-Friedrichs-Lewy (CFL) condition; Real-time monitor the vortex shedding frequency and the risk of structural resonance, and trigger the local grid refinement strategy to ensure the calculation stability.

[0052] The feedback generation unit constructs the verification basis through the following data processing process: Extract the spectral characteristics of the pressure pulsation in the fluid domain and calculate the proportion of the main frequency energy; Quantify the maximum value of the temperature gradient and the spatial distribution uniformity in the trailing edge region; Identify the strain concentration region on the blade surface and count the frequency of the peak strain.

[0053] For the adjustment module, in this embodiment, based on the real perception feedback data generated by the virtual experiment feedback module, a design error function is constructed. By comparing the preset performance indicators with the multi-objective evaluation results of the current candidate design parameters, the optimization strategy is adjusted backward and the design parameters are dynamically updated to drive the iterative convergence of the optimization process. This module ensures the dynamic optimization and performance convergence of the system in the multi-physical field coupling design space through a closed-loop mechanism of error analysis, strategy update, and parameter iteration.

[0054] The adjustment module includes an error analysis unit, a strategy update unit, and a parameter iteration unit. The error analysis unit receives the pressure distribution, temperature gradient, and structural strain data output by the virtual experiment feedback module, and conducts a comparative analysis with the flow performance optimization objective function defined in the intelligent optimization decision module to construct a design error function. The strategy update unit adjusts the parameters of the deep reinforcement learning strategy network using the backpropagation algorithm based on the error function results to generate new optimization strategy instructions. The parameter iteration unit dynamically adjusts the candidate design parameter set according to the updated strategy parameters and triggers the intelligent optimization decision module to start a new round of optimization iteration.

[0055] Specifically, the error analysis unit quantifies the design error through the following formula: ; In the formula, is the error index of the design parameter ; is the simulation value of the th performance indicator (such as flow efficiency, maximum stress value); is the preset target value; is the error weight coefficient; is the total number of error indicators. This formula eliminates the dimension difference through normalization and weights and synthesizes the multi-objective performance deviation.

[0056] The strategy update unit updates the parameters of the deep reinforcement learning strategy network using the backpropagation algorithm. The network parameter update rule is: ; In the formula, is the strategy network parameter of the th iteration; is the learning rate; is the regularization coefficient. The second regularization term is used to prevent the mutation of strategy parameters and improve the optimization stability.

[0057] In a possible implementation, the parameter iteration unit introduces an adaptive step-size mechanism. The parameter update step-size is dynamically adjusted according to the magnitude of the error gradient: a smaller step-size is adopted when the gradient is large to avoid oscillation, and the step-size is increased when the gradient is small to accelerate convergence. This mechanism is controlled by a gradient threshold and a step-size coefficient to ensure a stable and efficient optimization process. The deep reinforcement learning unit For the scheduling control module, in this embodiment, according to the optimization iteration stage and the performance convergence state, the global search or local optimization strategy is dynamically scheduled to control the computing resource allocation and the algorithm path selection. This module coordinates the data interaction between the multi-physics field simulation unit and the intelligent optimization decision module by real-time sensing the design parameter update trend and the simulation task load characteristics, ensuring that the system achieves efficient convergence in the complex parameter space.

[0058] The scheduling control module consists of a state monitoring unit, an algorithm scheduling unit, and a resource allocation unit. The state monitoring unit tracks the design parameter update sequence output by the parameter iteration unit in real time, and dynamically determines the current optimization stage state by quantifying the convergence rate and the performance fluctuation amplitude. The algorithm scheduling unit calls the global search or local optimization algorithm according to the stage state, and the resource allocation unit dynamically allocates computing resources based on the task complexity and the historical execution efficiency, forming a closed-loop control logic.

[0059] In some embodiments, the state monitoring unit calculates the design parameter convergence rate through the following formula: ; In the formula, is the design parameter vector at the th iteration; is the design parameter vector at the th iteration; is a very small positive number (usually taken as ) to prevent the denominator from being zero. When ( is a preset threshold, such as 0.01), it is determined that the system enters the local optimization stage and the algorithm switching mechanism is triggered.

[0060] Specifically, the resource allocation unit dynamically configures computing resources through the following function: ; In the formula, is the resource allocation ratio of the th simulation subtask, is the task priority weight (obtained by calculating the deviation rate of the objective function), is the historical average simulation time, is a non-linear adjustment factor (usually taken as 0.5 - 1.2), is the total number of current parallel tasks. This formula passes through the exponential term Strengthen the resource tilt for high-time-consuming tasks to alleviate the simulation bottleneck.

[0061] In a possible implementation, the algorithm scheduling unit adopts a two-layer decision-making mechanism: when the performance change rate (the calculation formula is the weighted average of the change rates of each objective function value) exceeds the threshold, forcefully enable the global search algorithm to reset the optimization path; when iterating continuously for 3 times, switch to a local optimization algorithm such as the quasi-Newton method. This mechanism is implemented through a status flag register to avoid oscillations caused by frequent switching.

[0062] As an option, the resource allocation unit supports an elastic scaling strategy. When it is monitored that the time consumption of a certain simulation subtask exceeds 150% of the estimated value, an instruction to expand the computing nodes is automatically triggered. This strategy is implemented through the Kubernetes container orchestration framework to ensure load balancing in a distributed simulation environment.

[0063] An intelligent design simulation method for the flow components of a fluid machine described below can be correspondingly referred to with an intelligent design simulation system for the flow components of a fluid machine described above.

[0064] Please refer to the appendix Figure 2 , the present invention also provides an intelligent design simulation method for the flow components of a fluid machine, including the following steps: Based on preset boundary conditions and physical properties, construct a multi-physical field coupling model through the mass conservation equation, momentum conservation equation, and heat conduction equation, and output the coupling simulation parameters of fluid mechanics, structural mechanics, and heat conduction; Based on the coupling simulation parameters of fluid mechanics, structural mechanics, and heat conduction, dynamically define the optimization objective function of the flow performance, generate an initial candidate design parameter set through a behavior-driven optimization strategy, and trigger a deep reinforcement learning mechanism to iteratively adjust the parameter generation strategy; Based on the initial candidate design parameter set, call a multi-physical field numerical simulator to perform joint calculations, generate a response data set of the fluid, structure, and thermal fields, and construct a comprehensive performance evaluation model to output a multi-objective evaluation result; According to the multi-objective evaluation result, drive a parametric modeling tool to reconstruct the three-dimensional geometric model of the flow component, load typical working condition parameters in a virtual simulation environment to simulate dynamic operation behaviors, and generate real perception feedback data including pressure, temperature, and strain distributions; Based on the difference analysis between the real perception feedback data including pressure, temperature, and strain distributions and the preset performance indicators, construct a design error function, update the deep reinforcement learning policy network through the backpropagation algorithm, and trigger a new round of optimization iteration convergence; According to the new round of optimization iteration stage and performance convergence state, dynamically schedule the combined strategy of the global search algorithm and the local optimization algorithm to control the priority of computing resource allocation.

[0065] The method of this embodiment can be used to implement the above system embodiment, and its principle and technical effects are similar, which will not be elaborated here.

[0066] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent design and simulation system for the flow components of a fluid machinery, characterized in that, Including: A multi-physics field modeling module, which is used to build fluid mechanics, structural mechanics, and heat conduction models based on preset boundary conditions and physical properties, and output coupled simulation parameters; An intelligent optimization decision-making module, which is used to receive the coupled simulation parameters, build an objective function for optimizing the flow performance, and generate an optimization path and a set of candidate design parameters through a behavior-driven optimization strategy and a deep reinforcement learning method; A collaborative calculation module, which is used to call a multi-physics field numerical simulator to jointly calculate the response results of each physical field based on the set of candidate design parameters, build a comprehensive performance evaluation model, and generate multi-objective evaluation results corresponding to each candidate design parameter; A virtual experiment feedback module, which is used to reconstruct the corresponding flow component structure model according to the multi-objective evaluation results corresponding to each candidate design parameter, simulate the operating conditions in a virtual simulation environment, and generate real perception feedback data as a basis for performance verification; An adjustment module, which is used to build a design error function based on the real perception feedback data, compare the preset expected performance index with the multi-objective evaluation results corresponding to the current candidate design parameters, reverse-adjust the optimization strategy, and dynamically update the design parameters to drive the optimization process to converge iteratively; A scheduling control module, which is used to dynamically schedule local optimization or global search strategies according to the iterative stage of the optimization process and the performance convergence state output by the adjustment module, and control the algorithm path selection and computing resource allocation in the optimization process.

2. The intelligent design and simulation system for a fluid machinery flow component according to claim 1, wherein The multi-physics field modeling module includes: A fluid mechanics modeling unit, which is used to build a fluid flow model based on preset boundary conditions and initial design parameters through the mass conservation equation and the momentum conservation equation; A structural mechanics modeling unit, which is used to build a structural strength model of the flow component based on the fluid flow model and combine material property parameters, and calculate the stress and deformation fields; A heat conduction modeling unit, which is used to build a temperature distribution model of the flow component based on the stress and deformation fields and the fluid flow model, and generate coupled simulation parameters.

3. The intelligent design and simulation system for the flow components of a fluid machine according to claim 1, characterized in that The intelligent optimization decision-making module includes: An objective function definition unit, which is used to extract flow performance indicators, structural response indicators, and thermal management indicators based on the coupled simulation parameters, and dynamically build a multi-objective optimization objective function; A dynamic strategy optimization unit, which is used to use the multi-objective optimization objective function as an evaluation basis, initialize the optimization path, and adopt a behavior-driven optimization strategy to adjust the candidate design parameters in real time to generate a strategy; A deep reinforcement learning unit, which is used to use the candidate design parameter generation strategy as an initial state, gradually train the optimization strategy network through simulation feedback interaction with the collaborative calculation module, and dynamically generate a new round of high-performance candidate design parameter sets.

4. An intelligent design and simulation system for fluid machinery flow components according to claim 3, characterized in that The collaborative calculation module includes: A multi-physics field simulation unit, which is used to call the modeling process of the multi-physics field modeling module to jointly simulate each candidate parameter according to the generated new round of high-performance candidate design parameter sets, and output a data set containing fluid, structural, and thermal field results; A performance evaluation unit, which is used to build a comprehensive performance scoring model based on the data set containing fluid, structural, and thermal field results, and dynamically map the influence relationship between design parameters and performance indicators; A data integration unit, which is used to construct multi-objective evaluation results corresponding to candidate design parameters based on the influence relationship between the design parameters and the performance indicators, and dynamically adjust the weight factors of the evaluation indicators.

5. An intelligent design and simulation system for fluid machinery flow components according to claim 4, characterized in that, The process of jointly simulating each candidate parameter by invoking the modeling process of the multi-physics field modeling module includes: Based on the new round of high-performance candidate design parameter set generated by the deep reinforcement learning unit, constructing a multi-physics field joint simulation task list; According to the design parameters in the simulation task list, dynamically configuring the inlet flow velocity boundary condition of the fluid mechanics simulation, the load distribution condition of the structural mechanics simulation, and the heat source distribution condition of the heat conduction simulation; Real-time synchronizing the fluid pressure field, structural stress field, and temperature field data in the joint simulation results to the performance evaluation unit, triggering the comprehensive performance scoring model to update the weight factors, and generating multi-objective evaluation results.

6. The intelligent design and simulation system for a fluid machinery flow component according to claim 4, wherein The virtual experiment feedback module includes: A model reconstruction unit, which is used to dynamically generate a three-dimensional geometric model according to the multi-objective evaluation results corresponding to the candidate design parameters; A dynamic simulation unit, which is used to simulate the operating state of the flow-through component under typical working conditions by invoking a multi-physics field simulator based on the three-dimensional geometric model, and the boundary conditions loaded by the multi-physics field simulator are dynamically configured by the joint simulation data output by the multi-physics field simulation unit; A feedback generation unit, which is used to extract pressure distribution, temperature distribution, and structural strain data from the simulation data configured by the dynamic simulation unit to construct a real perception feedback data set.

7. The intelligent design and simulation system for the flow components of a fluid machinery according to claim 6, wherein The process of simulating the operating state of the flow-through component under typical working conditions by invoking a multi-physics field simulator includes: According to the multi-objective evaluation results, extracting the candidate design parameters with the optimal flow-through efficiency and mapping them to the virtual simulation environment as typical working condition parameters; Based on the three-dimensional geometric model, invoking a multi-physics field simulator to load the typical working condition parameters and perform fluid-structure-thermal coupling transient simulation; Real-time collecting the inlet pressure pulsation data, trailing edge temperature gradient data, and blade strain peak data of the flow-through component during the fluid-structure-thermal coupling transient simulation process to construct a structured feedback data set.

8. The intelligent design and simulation system for a fluid machinery flow component according to claim 3, characterized in that, The adjustment module includes: An error analysis unit, which is used to compare and analyze the real perception feedback data generated by the virtual experiment feedback module with the flow-through performance optimization objective function defined in the intelligent optimization decision module to construct a design error function; A strategy update unit, which is used to adjust the optimization strategy network of the deep reinforcement learning unit by using the backpropagation algorithm based on the error function result and generate new strategy parameters; A parameter iteration unit, which is used to dynamically adjust the current candidate design parameter set based on the new strategy parameters and drive the optimization process in the intelligent optimization decision module to iterate again to form a closed-loop optimization process.

9. An intelligent design and simulation system for fluid machinery flow components according to claim 8, characterized in that, The scheduling control module includes: A state monitoring unit, which is used to monitor the design parameter update trend and performance convergence speed of the parameter iteration unit in real time and dynamically judge the current optimization stage state; An algorithm scheduling unit, which is used to dynamically select the global optimization algorithm used in the global search stage or the local optimization algorithm used in the local optimization stage according to the current optimization stage state; A resource allocation unit, which is used to dynamically adjust the computing resource allocation strategy for each simulation task according to the local optimization algorithm, and the resource allocation strategy is fed back to the multi-physics field simulation unit to optimize the simulation execution efficiency.

10. A method for intelligent design and simulation of a fluid machinery flow component, applied to a system for intelligent design and simulation of a fluid machinery flow component according to any one of claims 1-9, characterized in that, It includes the following steps: Based on preset boundary conditions and physical properties, construct a multi-physics field coupling model through the mass conservation equation, momentum conservation equation and heat conduction equation, and output the coupling simulation parameters of fluid mechanics, structural mechanics and heat conduction; Based on the coupling simulation parameters of fluid mechanics, structural mechanics and heat conduction, dynamically define the objective function for optimizing the flow performance, generate an initial candidate design parameter set through the behavior-driven optimization strategy, and trigger the deep reinforcement learning mechanism to iteratively adjust the parameter generation strategy; Based on the initial candidate design parameter set, call the multi-physics field numerical simulator to perform joint calculations, generate the response data sets of the fluid, structure and thermal fields, and construct a comprehensive performance evaluation model to output the multi-objective evaluation results; According to the multi-objective evaluation results, drive the parametric modeling tool to reconstruct the three-dimensional geometric model of the flow-through component, load the typical working condition parameters in the virtual simulation environment to simulate the dynamic operation behavior, and generate the real perception feedback data including pressure, temperature and strain distributions; Based on the difference analysis between the real perception feedback data including pressure, temperature and strain distributions and the preset performance indicators, construct a design error function, update the deep reinforcement learning policy network through the backpropagation algorithm, and trigger a new round of optimization iteration convergence; According to the new round of optimization iteration stage and performance convergence state, dynamically schedule the combined strategy of the global search algorithm and the local optimization algorithm to control the computing resource allocation priority.

Images