Lightweight optimization method for spool of overflow valve with anti-stuck constraint

Abstract

The application discloses a kind of lightweight optimization methods of overflow valve spool with anti-stuck constraint.The application avoids the precision loss and local optimal dead zone problem caused by traditional discrete grid optimization by constructing continuous multi-dimensional search space and combining adaptive Kriging surrogate model with continuous domain multi-objective ant colony algorithm.The application introduces anti-stuck exponential penalty term and safety margin dynamic feedback mechanism at the bottom of the algorithm, so that the optimization process can independently perceive the dangerous boundary of radial expansion deformation of the spool and adaptively adjust the optimization strategy, balancing the contradiction among spool lightweight, dynamic response performance and super-high pressure anti-stuck structure safety from the root.

Description

Technical Field

[0001] This invention relates to the field of digital simulation and structural optimization technology for hydraulic control valves, specifically to a lightweight optimization method for the valve core of an overflow valve with anti-jamming constraints. Background Technology

[0002] Relief valves are core pressure control components in complete sets of fluid transmission equipment, widely used in high-end equipment fields such as engineering machinery, aerospace, and intelligent manufacturing. They are a key area for breakthroughs in basic components across the equipment industry chain. Currently, the equipment manufacturing industry is continuously iterating towards lightweight design, high dynamic response, and long reliable lifespan. The industry generally adopts hollow topology structures to optimize valve core mass, thereby reducing moment of inertia and improving valve opening and closing response speed.

[0003] In recent years, hydraulic equipment has been gradually upgraded towards ultra-high pressure and large flow rates, and relief valves face complex mechanical-hydraulic dynamics and fluid-structure interaction problems during opening and closing. On the one hand, to improve the dynamic response speed of the valve core, the industry generally adopts a lightweight topology with internal hollowing to reduce the valve core's moment of inertia; however, on the other hand, lightweight design reduces the valve core wall thickness, making it prone to micron-level radial expansion when facing ultra-high pressure step impacts. If the expansion exceeds the fit clearance between the valve core and the valve sleeve, irreversible serious failures such as dry friction and valve core seizure may occur. Therefore, the technical contradiction between the requirement for rapid dynamic response of the valve core and the safety constraints of the anti-seizure structure under ultra-high pressure conditions has become a core problem that urgently needs to be solved in the design of ultra-high pressure relief valve cores.

[0004] Currently, the industry generally uses intelligent heuristic algorithms to complete valve core parameter matching design, such as the multi-objective optimization scheme for overflow valves disclosed in comparative document CN119538453A, which relies on traditional discrete ant colony optimization combined with annealing algorithms to optimize the macroscopic flow and pressure characteristics of the valve. This type of scheme has several inherent defects when adapting to valve core fluid-structure interaction and multi-dimensional collaborative optimization scenarios: (1) Traditional ant colony algorithm is a discrete grid optimization architecture. For the optimization of valve core micron-level continuous size parameters, interval discrete division must be performed. To compensate for the accuracy loss caused by discrete division, secondary annealing compensation is required. The overall computation link has high redundancy. When facing a high-dimensional parameter space of strong nonlinear fluid-structure interaction, it is very easy to deviate from the global optimal interval and fall into the local optimal dead zone, resulting in inherent bias in the optimization results.

[0005] (2) The existing optimization system lacks underlying physical failure boundary constraints. The algorithm evaluation only uses macroscopic performance such as response speed and flow error as evaluation indicators, which cannot identify microscopic failure hazards such as internal stress concentration and radial expansion deformation of the valve core. The theoretical dimensions of the optimized output are prone to valve core seizure under dynamic pressure conditions, resulting in insufficient reliability of the solution for engineering implementation.

[0006] (3) Conventional simulation acceleration proxy models use global smooth covariance kernel functions, which will smooth out the peak values ​​of valve core deformation and stress change under pressure shock, resulting in distorted prediction of dangerous boundaries. If uniform encrypted sampling is used to make up for the prediction blind spot, it will greatly increase the computing power consumption of three-dimensional finite element simulation, lengthen the product iteration cycle, and make it difficult to adapt to the rapid new product development needs of equipment enterprises.

[0007] In summary, the field of digital design of fluid transmission components urgently needs a new continuous domain intelligent optimization architecture to eliminate the inherent shortcomings of traditional discrete algorithms, enabling the optimization model to autonomously identify pressure-bearing dangerous boundaries and adaptively avoid risks. This will fundamentally balance the multiple constraints and contradictions between valve core lightweighting, dynamic response performance, and structural operational safety, and support the independent research and development of hydraulic core components for high-end equipment. Summary of the Invention

[0008] To address the shortcomings of existing optimization algorithms, such as redundant discrete solutions, lack of physical safety constraints, susceptibility to local optima in high-dimensional optimization, high computational cost in simulation, and poor engineering reliability of optimization schemes, this invention aims to reconstruct an integrated mechanical-hydraulic coupling optimization framework. It embeds a deformation gap safety penalty mechanism at the algorithm's underlying layer and establishes a dual adaptive feedback closed loop of a proxy model and a continuous-domain multi-objective ant colony algorithm. This endows the optimization model with the ability to autonomously perceive dangerous boundaries in fluid-structure interaction and adaptively adjust its optimization strategy, simultaneously balancing the three mutually constraining requirements of valve core lightweighting, dynamic stability, and pressure-bearing anti-jamming safety.

[0009] To achieve the above objectives, the present invention provides a lightweight optimization method for the valve core of an overflow valve with anti-jamming constraints, comprising the following steps: S1. Construct a continuous multidimensional search space based on the multi-parameter physical constraints of the valve core; S2. Construct a machine-fluid co-simulation sample library based on the continuous multidimensional search space; S3. Based on the aforementioned mechanical-hydraulic joint simulation sample library, train an adaptive Kriging proxy model and encapsulate it as a prediction black box; S4. Based on the response predicted by the adaptive Kriging proxy model, construct a multiphysics dynamic anti-lockdown penalty fitness function; S5. Based on the anti-deadlock penalty fitness function, drive the continuous domain multi-objective ant colony algorithm to obtain the Pareto non-dominated solution set; S6. Perform TOPSIS decision based on the Pareto non-dominated solution set to obtain the final valve core shaping dimension parameters.

[0010] Preferably, in step S1, a core optimization feature vector is defined. ,in, d t For the diameter of the central damping hole, d x For the outer diameter of the valve core,δ For the clearance between the valve core and the valve bore, m For valve core quality; Based on the lower bound constraint of mass-damped orifice coupling, the upper bound constraint of annular gap leakage gap, the outer diameter boundary constraint of thick-walled cylindrical strength valve core, and the boundary constraint of dynamic response damped orifice diameter, a physically feasible solution range for continuous parameters is defined.

[0011] Preferably, in step S2, a bidirectional data interaction platform based on AMESim and ANSYS is established to extract the opening and closing response time, pressure overshoot, maximum equivalent stress, and maximum radial expansion deformation. Sample points are generated using Latin hypercube sampling, and the corresponding multi-objective physical feature outputs are obtained through joint simulation to construct a full-condition input-output mapping sample library.

[0012] Preferably, S3 includes: Based on the mapping sample library, a basic Kriging surrogate model is fitted, and the Matern 5 / 2 covariance kernel function is used for spatial correlation mapping. The model is improved by actively learning to repeatedly filter new samples and iteratively reconstruct the model; Leave-one-out cross-validation is used to evaluate the global prediction accuracy until convergence, and the converged model is encapsulated as a prediction black box.

[0013] Preferably, step S4 includes: constructing a multi-objective comprehensive fitness function that includes an anti-jamming exponential penalty term, wherein the penalty term is triggered when the radial expansion deformation of the valve core exceeds the safety threshold of the mating clearance, and parameters exceeding the safety threshold are determined to be infeasible solutions.

[0014] Preferably, S5 includes: generating a new generation of topological size parameters through a mixture of Gaussian probability density functions, inputting the generated parameters into the prediction black box to obtain physical responses and calculate fitness scores, zeroing the weights of inferior solutions according to the scores, dynamically adjusting the Gaussian kernel variance according to the anti-lockdown safety margin, and outputting a Pareto non-inferior solution set after iterative calculation.

[0015] Preferably, S6 includes: constructing a decision matrix that includes quality, response time, pressure overshoot and anti-jamming safety margin; adaptively allocating the weights of each index according to the nominal pressure level; calculating the closeness of each candidate solution to the positive and negative ideal solutions; and selecting the solution with the highest closeness as the final optimal topology size feature vector.

[0016] Preferably, in S3, the convergence condition for leave-one cross-validation is that the root mean square error of the prediction drops below 3%, or the preset maximum computing resources are reached.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention avoids the accuracy loss and local optimum dead zone problem caused by traditional discrete grid optimization by constructing a continuous multidimensional search space and combining an adaptive Kriging surrogate model with a continuous domain multi-objective ant colony algorithm. At the algorithm's underlying level, this invention introduces an anti-jamming exponential penalty term and a dynamic safety margin feedback mechanism, enabling the optimization process to autonomously perceive the dangerous boundary of radial expansion deformation of the valve core and adaptively adjust the optimization strategy. This fundamentally balances the contradictions between valve core lightweighting, dynamic response performance, and the safety of the ultra-high pressure anti-jamming structure. Simultaneously, this invention reconstructs the surrogate model based on an expectation-enhancing active learning mechanism, accurately capturing deformation peaks under limited simulation computing power, significantly reducing optimization iteration costs, and significantly improving the engineering reliability and feasibility of the high-pressure relief valve core design scheme. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a simplified structural diagram of the overflow valve according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the valve core structure according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the adaptive Kriging proxy model process according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the TOPSIS multidimensional decision output process based on engineering adaptive weights in an embodiment of the present invention.

[0020] Explanation of reference numerals in the attached figures: 1. Valve seat; 2. Spring seat; 3. Pressure regulating spring; 4. One-way spring; 5. Valve core; 6. Ball head structure. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] Example The overall structure of the overflow valve is as follows Figure 1 As shown, it mainly includes valve seat 1, spring seat 2, pressure regulating spring 3, one-way spring 4, valve core 5, and ball head structure 6. This relief valve mainly includes two working conditions: forward relief and reverse oil replenishment. In the relief state, hydraulic oil enters through port P, pushes valve core 5 through the internal oil passage of the ball head, and flows out through port T. In the oil replenishment state, hydraulic oil enters through port T, pushes the ball head and valve core 5, and flows out through port P.

[0024] Valve core 5 structure as follows Figure 2 As shown. The front section of valve core 5 has a conical structure with a central damping through-hole inside; the main design parameters include the diameter of the central damping through-hole. d t Valve core outer diameter d x The clearance between valve core 5 and valve bore δ Valve core quality m .

[0025] To address the issues of valve core 5's sluggish response, improper damping causing opening and closing oscillations, and the potential for expansion and jamming due to valve core lightweighting under high pressure, this invention proposes a valve core design method as follows: Figure 3 As shown, the main steps include: S1. Construct a continuous multidimensional search space based on the multi-parameter physical constraints of the valve core.

[0026] In this step, the core optimization feature vector is defined. And establish a continuous multidimensional search space with strong physical constraints. This space consists of core feature vectors. It is constructed over a continuous domain, and coupled boundary constraints are applied to define the value ranges of each parameter, thus delineating a physically feasible solution range for the continuous parameters. Specific constraints are as follows: ① Lower bound constraint of mass-damped orifice coupling: To meet the system's minimum natural frequency ω min ,quality m Must meet ,in K For spring stiffness; simultaneously, the diameter of the central damping orifice. d t The size of the valve core directly determines the transient throttling back pressure, limiting the ultimate volume for lightweight valve cores.

[0027] ② Upper limit constraint of annular gap leakage gap: Fit clearance δ Subject to the system's maximum permissible internal leakage Strict constraints were imposed, based on the Poisson's annular slot shear flow model. δMust meet: in, The dynamic viscosity of hydraulic oil. To match the segment length, This is the system's nominal pressure.

[0028] ③ Boundary constraints on the outer diameter of the thick-walled cylindrical strength valve core: To prevent high-pressure yielding caused by wall thinning for lightweighting, based on the theory of internal / external pressure on thick-walled cylinders, the outer diameter of the valve core is... The dynamic lower limit must be consistent with the internally hollowed-out damping orifice. This creates a safety constraint; its upper limit is limited by the inner diameter of the valve sleeve. The physical dimensions. Its initial continuous search boundary is strictly defined as: in, This refers to the allowable stress of the valve core material; This is the system's nominal pressure.

[0029] ④ Dynamic response damping aperture boundary constraints: Damping orifice The magnitude determines the system damping ratio With response time At nominal pressure If the aperture is too small, the response time will exceed the system's allowable limit. If the damping ratio is too large, it will be insufficient, easily leading to severe overshoot oscillations. Its continuous search boundary is strictly obtained by inversely calculating the response extremum and the minimum stable damping ratio: in, Indicates satisfaction Minimum diameter of the constrained damping orifice The minimum stable damping ratio that indicates the system has no severe overshoot oscillations. Indicates satisfaction Maximum diameter of the constrained damping orifice.

[0030] Thus, a physically feasible continuous multidimensional search space is obtained. .

[0031] S2. Construct a sample library for machine-liquid co-simulation based on the continuous multidimensional search space.

[0032] In this step, the search space is targeted. By combining parameters within the system, a bidirectional data interaction platform based on AMESim (one-dimensional dynamics of mechanical fluids) and ANSYS (three-dimensional explicit finite element method) was established.

[0033] In the dynamic model, two core dynamic indicators are extracted from the transient pressure-time series during valve core actuation: opening and closing response time. With pressure overshoot Pressure overshoot is defined as the peak limit pressure. Percentage deviation from rated pressure.

[0034] Recording overtuning peaks in dynamic simulation At the moment of occurrence, the transient pressure field is mapped onto a finite element mesh, and the equivalent stress distribution of the valve core according to the fourth strength theory is solved; the maximum equivalent stress in the weak region of the inner oil passage is extracted. and the maximum radial expansion deformation of the outer cylindrical surface under extreme working conditions .

[0035] Using the Latin Hypercube Sampling Algorithm (LHS) in space Intrinsic generation By using a set of sample points and co-simulating to obtain the corresponding multi-objective physical feature outputs, a full-condition input-output mapping sample library is constructed. This database fully records multidimensional input feature vectors. Multiphysics Output Response The mapping relationship is specifically defined as follows: in, For the first The structural size feature vector of the input group, i.e. ; This is the output vector of the corresponding multi-objective physical field response obtained through co-simulation. This database... This will serve as a global, full-condition experience set, providing a data foundation for subsequent training of agent models.

[0036] Thus, a full-condition input-output mapping sample library is obtained. .

[0037] S3. Based on the joint simulation sample library, train an adaptive Kriging surrogate model and encapsulate it as a prediction black box.

[0038] In this step, to address the shortcomings of excessive computational cost in 3D fluid-structure interaction simulations and the inability of conventional proxy models to capture deformation abrupt peaks, a targeted reconstruction of the conventional proxy model is performed to build a high-speed prediction engine. The construction process is as follows: Figure 4 As shown, the specific adaptive training process is as follows: (1) Initial proxy model fitting: Initial mapping sample library built based on S2 , input multidimensional feature vector With output physical response Establish mathematical relationships and fit a basic Kriging surrogate model. This is applicable to valve cores subjected to target high pressure. The non-stationary dynamic overshoot characteristics induced by a step impact are spatially correlated using a custom Matern 5 / 2 covariance kernel function. in, The correlation value is the correlation between any two sample points in space. The Euclidean distance between sample points; The scale parameter is used to control the smoothing range.

[0039] This custom kernel function, while ensuring mean-square differentiability, abandons the global absolute smoothness constraint of traditional kernel functions, endowing the surrogate model with excellent local fluctuation feature capture performance, and can accurately characterize the radial deformation abrupt behavior generated at the weak position of the valve core under step high pressure impact.

[0040] (2) EI expects to improve the active learning cycle: If only relying on the initial sample database Static fitting, within a high-dimensional unbounded continuous parameter optimization space, still results in significant errors when predicting unknown parameter intervals outside the initial sample coverage area. To accurately identify physical hazard boundaries while controlling simulation computing power, a dynamic resampling loop mechanism based on the expectation improvement criterion is designed. This mechanism allows the model to autonomously select the optimal coordinates for targeted supplementary sampling. A quantitative evaluation is performed on the predicted mean and variance of all unknown solutions within the optimization space. Critical boundary feature points with both high prediction uncertainty and high structural failure risk are automatically selected. The structural feature parameters corresponding to the maximum expected improvement index are chosen as new samples and added to the sample database. Joint simulation is used to solve for the actual physical response data corresponding to these samples, and the surrogate model is iteratively reconstructed based on the updated sample set.

[0041] (3) Model convergence accuracy verification and black-box encapsulation: After each sample point is actively added to the surrogate model, leave-one-out cross-validation is used to evaluate the global prediction accuracy. If the root mean square error of the prediction drops below 3%, or the preset maximum computing resources (the upper limit of the number of points added) are reached, the adaptive sampling loop is exited, and a high-fidelity adaptive Kriging surrogate model is finally output. This model is then encapsulated as a millisecond-level prediction black box to replace the time-consuming finite element simulation for subsequent use.

[0042] Thus, we obtain an adaptive Kriging surrogate model black box that can quickly predict multi-physics responses.

[0043] S4. Based on the surrogate model, predict the response and construct a multiphysics dynamic anti-lockdown penalty fitness function.

[0044] In this step, a multi-objective comprehensive fitness evaluation model with fluid-structure interaction boundary awareness is constructed as the core hub connecting the surrogate model prediction and the optimization algorithm evaluation link. The valve core anti-jamming safety constraint is set as a hard decision boundary for full-condition optimization. Given the deformation... With overshoot pressure Exhibiting nonlinear dynamic changes, a multi-objective comprehensive fitness solution method with built-in high-order failure penalty terms is constructed: in, Indicates the combination of design variables Predicted response time of the lower valve system Indicates the combination of design variables Equivalent stress at key locations of the lower valve core, Indicates the combination of design variables Lower valve core structure quality This indicates the response time reference value. This indicates the allowable stress reference value of the valve core. This indicates the reference mass of the valve core.

[0045] Anti-lock exponential penalty item Defined as: in, To comprehensively assess multi-objective fitness scores, a lower score indicates better performance and greater safety. For normalized weights; For taking values The maximum penalty base at this level; To prevent the scale constant from getting stuck; It is a step indicator function; This is the thermal expansion safety factor.

[0046] This fitness evaluation model has an automatic failure determination mechanism. If the expansion and deformation of the valve core due to the lightweight topology design of the candidate structural parameters exceeds the safety threshold, then the corresponding parameters will be determined. The value will increase significantly, so the set of parameters will be determined as an infeasible solution and eliminated.

[0047] Thus, a fitness function that can comprehensively evaluate dynamic performance and anti-jamming safety is obtained.

[0048] S5. Drive the continuous domain MOACO algorithm based on the fitness function to prevent deadlock, and obtain the Pareto non-dominated solution set.

[0049] In this step, the continuous domain multi-objective ant colony algorithm is used to achieve closed-loop feedback optimization. The specific process is as follows: Figure 5As shown, the prediction engine in step S3 and the evaluation referee in step S4 are embedded in the following loop: (1) Parameter generation stage: Virtual ant colony individuals do not require discrete grid partitioning; they can be directly generated using a mixture Gaussian probability density function. Generate next-generation topology size feature vectors : in, For the first in the archive The weight coefficients assigned to each high-quality solution; Gaussian kernel variance / exploration step size; For the first In the nth iteration The statistical central mean of a high-quality solution.

[0050] (2) Fitness calculation and adaptive update feedback: The weighting coefficients in the above formula With kernel variance It is not a static constant, but rather a fitness model output by S4. The driving force determines the outcome. The complete closed-loop technology flow process is as follows: (a) Rapid prediction and scoring: Each set of parameters generated by the ant colony... The input is fed into a high-precision Kriging surrogate model black box encapsulated in S3 for prediction. , The physical response is then determined; subsequently, the response result is substituted into the comprehensive fitness formula of S4. The total score is calculated in the middle.

[0051] (b) Failure penalty triggering and weighting Zeroing mechanism: The algorithm is based on The generated solutions are sorted. If the parameters generated by a certain ant trigger a maximum... If the solution is poorly scored, its overall ranking will be at the bottom. At this point, the algorithm eliminates the inferior solution, and the weight assigned to that solution is changed. When zeroed out, the corresponding Gaussian kernel distribution is directly erased.

[0052] (c) Variance based on safety margin Dynamic adjustment: For high-quality solutions that do not trigger penalties, the algorithm adjusts their Gaussian kernel variance. Anti-jamming safety margin calculated with S4 A positive correlation dynamic mapping is performed. When the predicted expansion deformation of the valve core approaches the critical safety boundary for seizing, the algorithm can identify this dangerous condition and automatically and significantly reduce it. This switches the global, large-scale exploration mode to a micrometer-level local convergence mode.

[0053] This algorithm constructs a complete closed-loop data flow: "parameter generation → surrogate prediction → fitness function evaluation → Gaussian kernel variance and weight coefficients inverse adaptive update," endowing the optimization algorithm with the ability to autonomously perceive the physical boundaries of fluid-structure interaction conditions. After multiple rounds of iterative calculations, it can stably output a Pareto nondominated solution set that balances dynamic performance and structural safety.

[0054] Thus, a multi-objective Pareto nondominated solution set oriented towards anti-lockdown constraints is obtained.

[0055] S6. Perform TOPSIS decision based on the Pareto non-dominated solution set to obtain the final valve core shaping dimension parameters.

[0056] In this step, the complete Pareto non-dominated solution set is used as the decision analysis object. The approximation ideal solution ranking method is introduced to construct an engineering decision system. By matching the high-pressure working condition requirements and setting multi-objective adaptive weights, the theoretical multi-objective optimization solution set is transformed into a valve core engineering size scheme that can be directly used for production and finalization, forming a complete design optimization closed loop.

[0057] (1) Constructing a decision indicator vector: For the TOPSIS method, extract indicators to construct a decision matrix. , Among them, topological quality Response time Pressure overshoot To minimize cost-related indicators; to prevent jamming safety margin This is a maximization benefit index. After normalizing the vectors of all candidate solutions, a standardized engineering decision matrix is ​​constructed.

[0058] For different system input pressures Assign multi-dimensional weight vectors The sum of all weights must be satisfied. Since structural seizure of the high-pressure relief valve under any target pressure is a serious malfunction, the following dynamic configuration logic is proposed: (2) Adaptive weight allocation based on operating conditions: The weights of various indicators are adjusted according to the valve's rated working pressure, with higher pressure ratings receiving the highest priority for safety margin. This dynamically balances the three key requirements of lightweight design, dynamic response, and anti-seize. The higher the level, the greater the safety margin weight. The higher the priority, and it must be the first priority to ensure that the valve core has no risk of seizure or failure under extreme operating conditions; pressure overshoot weight. With dynamic response weights Second highest priority, used to suppress high-voltage oscillations and improve response hysteresis; lightweight mass weight. Priority is placed last, solely to ensure structural safety and dynamic stability. The standardized decision matrix is ​​multiplied by the adaptive weights to obtain the weighted normalized decision matrix.

[0059] (3) Calculate the Euclidean distance from each candidate solution to the positive and negative ideal solutions: In the weighted decision matrix space, the optimal value of each column is extracted to form the positive ideal solution. Extracting the worst-case value to form the negative ideal solution Using the spatial Euclidean distance equation, the distance from each candidate solution to the positive and negative ideal solutions is calculated, thus obtaining the comprehensive relative closeness of each candidate solution. : in, For the first The comprehensive relative closeness of the solutions for each alternative feature size, with a value range of [value range missing]. The closer this value is to 1, the better the solution is at balancing lightweight design with extremely fast dynamic response and the most ample margin against jamming. For the first The weighted Euclidean distance from each alternative solution to the positive ideal solution, where the positive ideal solution represents the virtual perfect boundary where all objectives reach their optimal values ​​under the current weighted matrix; For the first The weighted Euclidean distance from each alternative solution to the negative ideal solution, where the negative ideal solution represents a virtual danger boundary where all objectives are at their worst under the current weighted matrix.

[0060] The algorithm automatically selects the overall relative similarity. The highest solution is used as the globally unique final optimal topological size feature vector. Finally, considering the fatigue limit of the material and the tolerance of the heat treatment process, the final engineering drawings for the high-pressure relief valve core, guiding actual precision manufacturing, are output.

[0061] Thus, a valve core size parameter scheme for high-pressure relief valve that can be directly used for production and finalization was obtained.

[0062] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A lightweight optimization method for the valve core of an overflow valve with anti-jamming constraints, characterized in that, Includes the following steps: S1. Construct a continuous multidimensional search space based on the multi-parameter physical constraints of the valve core; S2. Construct a machine-fluid co-simulation sample library based on the continuous multidimensional search space; S3. Based on the aforementioned mechanical-hydraulic joint simulation sample library, train an adaptive Kriging proxy model and encapsulate it as a prediction black box; S4. Based on the response predicted by the adaptive Kriging proxy model, construct a multiphysics dynamic anti-lockdown penalty fitness function; S5. Based on the anti-deadlock penalty fitness function, drive the continuous domain multi-objective ant colony algorithm to obtain the Pareto non-dominated solution set; S6. Perform TOPSIS decision based on the Pareto non-dominated solution set to obtain the final valve core shaping dimension parameters.

2. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, In S1, the core optimization feature vector is defined. ,in, d t For the diameter of the central damping hole, d x For the outer diameter of the valve core, δ For the clearance between the valve core and the valve bore, m For valve core quality; Based on the lower bound constraint of mass-damped orifice coupling, the upper bound constraint of annular gap leakage gap, the outer diameter boundary constraint of thick-walled cylindrical strength valve core, and the boundary constraint of dynamic response damped orifice diameter, a physically feasible solution range for continuous parameters is defined.

3. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, In S2, a bidirectional data interaction platform based on AMESim and ANSYS is established to extract the opening and closing response time, pressure overshoot, maximum equivalent stress, and maximum radial expansion deformation. Sample points are generated using Latin hypercube sampling, and the corresponding multi-objective physical feature outputs are obtained through joint simulation to construct a full-condition input-output mapping sample library.

4. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, S3 includes: Based on the mapping sample library, a basic Kriging surrogate model is fitted, and the Matern 5 / 2 covariance kernel function is used for spatial correlation mapping. The model is improved by actively learning to repeatedly filter new samples and iteratively reconstruct the model; Leave-one-out cross-validation is used to evaluate the global prediction accuracy until convergence, and the converged model is encapsulated as a prediction black box.

5. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, Step S4 includes: constructing a multi-objective comprehensive fitness function containing an anti-jamming exponential penalty term. The penalty term is triggered when the radial expansion deformation of the valve core exceeds the safety threshold of the mating clearance, and the parameters exceeding the safety threshold are determined to be infeasible solutions.

6. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, S5 includes: generating a new generation of topological size parameters through a mixture of Gaussian probability density functions, inputting the generated parameters into the prediction black box to obtain physical responses and calculate fitness scores, zeroing the weights of inferior solutions based on the scores, dynamically adjusting the Gaussian kernel variance based on the anti-lockdown safety margin, and outputting a Pareto non-inferior solution set after iterative calculation.

7. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 1, characterized in that, S6 includes: constructing a decision matrix that includes quality, response time, pressure overshoot and anti-jamming safety margin; adaptively allocating the weights of each index according to the nominal pressure level; calculating the closeness of each candidate solution to the positive and negative ideal solutions; and selecting the solution with the highest closeness as the final optimal topology size feature vector.

8. The lightweight optimization method for the overflow valve core with anti-jamming constraint according to claim 4, characterized in that, In S3, the convergence condition for leave-one cross-validation is that the root mean square error of the prediction drops below 3%, or the preset maximum computing resources are reached.

Images