Process parameter optimization system for integrated die cast
By constructing a nonlinear mapping using a Gaussian process regression model and an optimized metric tensor, and combining this with geodesic equations to optimize the process parameters of integrated die castings, the problems of low efficiency and insufficient accuracy in existing technologies are solved. This achieves efficient and stable optimization of process parameter combinations, thereby improving the quality and production efficiency of die castings.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-02
- Publication Date
- 2026-06-02
AI Technical Summary
Existing technologies rely on linear models or empirical rules to optimize the process parameters of integrated die castings, neglecting the complex interaction between process parameters and die castings, resulting in low efficiency and insufficient accuracy in optimization results.
A Gaussian process regression model is used to construct a nonlinear mapping between process parameters and quality parameters. Combined with the optimization metric tensor and geodesic equations, the global optimal solution for the combination of process parameters is searched through probabilistic modeling and geometric constraint framework.
It significantly improves the efficiency and stability of process parameter optimization, reduces reliance on manual experience, and enhances the production quality and efficiency of integrated die-cast parts.
Smart Images
Figure CN122131595A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of process optimization technology, and in particular to a process parameter optimization system for integrated die-cast parts. Background Technology
[0002] Integrated die-cast parts are metal parts that are structurally complete and multifunctional, manufactured in a single die-casting process. These parts are typically made of alloys such as aluminum, zinc, and magnesium, and are widely used in the automotive, home appliance, and aerospace industries. Integrated die-cast parts not only reduce production costs but also improve product performance and reliability.
[0003] Optimizing process parameters for integrated die-cast parts is crucial because the die-casting process involves multiple stages, such as temperature, pressure, and speed, and these parameters directly affect the quality of the casting and production efficiency. Inappropriate process parameters can lead to casting defects such as porosity, cracks, and dimensional instability, thereby reducing product performance and yield. By optimizing process parameters, casting quality can be maximized, scrap rates reduced, production efficiency improved, and product consistency and reliability ensured, meeting the high precision and high quality requirements of industrial production.
[0004] However, existing technologies typically rely on linear models or experience-based rules to fine-tune process parameters, neglecting the complex interaction between process parameters and integrated die-cast parts, resulting in inefficient and inaccurate optimization results. Summary of the Invention
[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a process parameter optimization system for integrated die castings, which can solve the technical problem that the prior art usually relies on linear models or experience-based rules to fine-tune process parameters, ignoring the complex interaction between process parameters and integrated die castings, resulting in low efficiency and insufficient accuracy of optimization results.
[0006] This invention proposes a process parameter optimization system for integrated die-casting parts, including: The acquisition module is used to acquire historical sample data of integrated die castings, including process parameter groups and corresponding integrated die casting quality parameter groups. The first construction module is used to build a Gaussian process regression model based on historical sample data to achieve a nonlinear mapping between the process parameter set and the integrated die-casting part quality parameter set; The second building module is used to combine the Gaussian process regression model to construct the optimization metric tensor of the process parameter set; The first module is used to establish a single-objective optimization function for the process parameter set, with the integrated die-casting part quality parameter set as the optimization objective. The second module is used to establish the geodesic equation of a single-objective optimization function by combining the optimization metric tensor; The solver module is used to iteratively solve the geodesic equation with the goal of minimizing a single-objective optimization function, and obtain the target set of process parameters.
[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, a Gaussian process regression model is used to accurately model the nonlinear coupling relationship between process parameters and quality parameters based on historical data. This is combined with an optimization metric tensor to construct a geometric constraint framework for the parameter space. Geodesic equations are used to guide the global search path, effectively avoiding the local optima traps caused by traditional linear models or empirical rules. High-dimensional interactions between parameters can be explored, and probabilistic modeling enables rapid searching of the parameter space, significantly improving optimization efficiency and the stability of output process parameters. Simultaneously, it reduces reliance on human experience, achieving a global optimal solution search for process parameter combinations, and improving the production quality and efficiency of integrated die-cast parts. Attached Figure Description
[0008] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.
[0009] Figure 1 This is a schematic diagram of the integrated die-casting process parameter optimization system provided in this embodiment of the invention. Detailed Implementation
[0010] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions 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, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0011] The process parameter optimization system for integrated die-casting parts provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.
[0012] Reference manual attached Figure 1 The diagram shows a schematic of the process parameter optimization system for integrated die-casting parts provided in an embodiment of the present invention.
[0013] This invention provides a process parameter optimization system for integrated die-casting parts, including: Module 1 is used to acquire historical sample data of integrated die-cast parts.
[0014] The historical sample data includes process parameter sets and corresponding integrated die-casting part quality parameter sets.
[0015] It should be noted that the process parameter set refers to the multi-dimensional input variables involved in the die casting process. The integrated die casting quality parameter set refers to the multi-dimensional performance indicators of the finished product. By comprehensively linking the dynamic mapping relationship between process parameters and quality parameters, the system can accurately capture the nonlinear coupling effect between parameters, providing data-driven decision-making basis for optimization, avoiding the limitations of traditional experience-based trial and error, and improving optimization efficiency.
[0016] In one possible implementation, the process parameter set includes injection speed, mold temperature, melt temperature, and pressurization pressure. The integrated die-casting part quality parameter set includes porosity, tensile strength, and dimensional tolerances.
[0017] Understandably, by modeling the nonlinear relationships between these parameters, global optimization of process parameters can be achieved, thereby reducing porosity, increasing tensile strength, and controlling dimensional deviations, significantly improving the quality stability and process reliability of die castings.
[0018] The first construction module 2 is used to build a Gaussian process regression model based on historical sample data to achieve a nonlinear mapping between the process parameter set and the integrated die-casting part quality parameter set.
[0019] Among them, the Gaussian process regression model is a nonparametric regression method based on a Bayesian framework. Nonlinear mapping refers to the complex nonlinear relationship between the process parameter set and the quality parameter set. This model flexibly models complex nonlinear relationships through kernel functions, while providing estimations of prediction uncertainties, offering reliable confidence guidance for subsequent optimization, and significantly improving the accuracy and robustness of parameter space modeling.
[0020] In one possible implementation, the first building module 2 is specifically used for: Establish a covariance function to describe the correlation between various process parameters in the process parameter group.
[0021] The specific formula for the covariance function is as follows: .
[0022] .
[0023] in, This represents different process parameters within a process parameter group. This represents the signal variance describing the variation of quality parameters of an integrated die-cast part with process parameters. This represents the characteristic scale diagonal matrix for each process parameter. Indicates the first i Characteristic scales of each process parameter , Represents a diagonal matrix. Represents the natural exponential function, subscript T Indicates transpose. Indicates about and The covariance function.
[0024] It should be noted that the larger the characteristic scale of the process parameters, the lower their sensitivity to the quality parameters of the integrated die-cast part. This process constructs the covariance structure among process parameters using a squared exponential kernel function, quantifies the sensitivity of each parameter to the quality parameters using the characteristic scale parameter, and characterizes the overall fluctuation amplitude using the signal variance. The nonlinear correlation between parameters is automatically captured through exponential decay, adapting to complex process interactions. The distance-based kernel function design enhances the model's responsiveness to local changes while maintaining global smoothness, significantly improving the modeling accuracy and stability of the Gaussian process under small sample conditions.
[0025] An initial Gaussian process regression model is established by combining the covariance function.
[0026] Estimate the hyperparameters of the initial Gaussian process regression model to obtain the Gaussian process regression model. The hyperparameters include the characteristic scales of each process parameter and the signal variance describing the variation of the quality parameters of the integrated die casting with the process parameters.
[0027] The formula for the Gaussian process regression model is as follows: .
[0028] .
[0029] in, Indicates process parameter group The average value of the corresponding integrated die-cast part quality parameters. Indicates the first n The first sample point m Quality parameters of an integrated die-cast part N This represents the total number of samples in the historical sample data. Indicates the first m Quality parameters of an integrated die-cast part related to the process parameter group The response function, This represents a Gaussian process.
[0030] It should be noted that this process estimates the hyperparameters (feature scale and signal variance) of the Gaussian process by maximizing the marginal likelihood function, enabling the model to adaptively adjust the weight allocation for sensitivity to process parameters and dynamically balance the overall fluctuation characteristics of quality parameters. Hyperparameter optimization makes the model more closely resemble the actual distribution of process data. Joint estimation of signal variance and feature scale enhances the model's ability to fit nonlinear relationships. The probabilistic framework-based modeling approach provides prediction confidence intervals, significantly improving the reliability of parameter space exploration and the stability of the optimization path.
[0031] Specifically, this process describes the correlation between process parameters by establishing a covariance function, and then constructs a Gaussian process regression model based on this covariance function. This enables accurate modeling of the nonlinear relationship between the quality parameters and process parameters of integrated die-cast parts. By estimating the model's hyperparameters, such as the characteristic scale and signal variance of the process parameters, the model can better capture the impact of process parameters on quality parameters. Compared to traditional linear models or empirical methods, this method can more accurately reflect the complex nonlinear relationship between process parameters and quality, thereby improving the accuracy and efficiency of optimization and ensuring that the optimal combination of process parameters can be found.
[0032] In one possible implementation, the hyperparameters of the initial Gaussian process regression model are estimated as follows: The hyperparameters of the initial Gaussian process regression model are estimated using the maximum likelihood estimation method.
[0033] Specifically, the hyperparameters of the Gaussian process regression model are optimized using the maximum likelihood estimation method, which maximizes the likelihood function of the sample data, thereby determining the specific form of the model (the likelihood function is the probability density function of the sample data; maximizing the likelihood function means the model has the highest fit to the sample data). The specific process is as follows: Select a quality parameter of the integrated die-casting part to be optimized, and organize the process parameter set and the corresponding quality parameter of the integrated die-casting part to be optimized into two sets of data. Then, calculate the covariance matrix of the two sets of process parameter sets based on the covariance function. Establish the joint likelihood function of the sample data, i.e., the organized data: ,in, This represents the average value of the quality parameters of the integrated die-cast part to be optimized. Represents the identity matrix. Represents the covariance matrix. Indicates a normal distribution. Indicates hyperparameters, express Quality parameters of the integrated die-casting part to be optimized under the given conditions Y The conditional probability density. Taking the natural logarithm of the joint likelihood function (the logarithm is a monotonically increasing function, maximizing the log-likelihood is equivalent to maximizing the original likelihood), we get: .in, Represents the natural logarithm. Represents pi (π). N This represents the total number of samples in the historical sample data. Since the log-likelihood function is non-convex (it has multiple local extrema), to maximize the log-likelihood function, numerical optimization algorithms (such as BFGS, L-BFGS, and Adam) can be used to solve for the hyperparameters corresponding to its maximum value. The update direction of the optimization process is the direction in which the partial derivatives of the log-likelihood function with respect to each hyperparameter increase (i.e., the gradient ascent direction), until the preset number of iterations is reached. Then, the hyperparameters corresponding to maximizing the log-likelihood function are selected for output, or the process terminates when the gradient norm is less than the convergence threshold.
[0034] Specifically, in practice, multiple quality parameters of integrated die-cast parts can be optimized in parallel. By assigning weights to handle the trade-offs between indicators, the optimization requirements under all quality indicator constraints can be met. The overall process is consistent with the optimization process for a single integrated die-cast part's quality parameters. This process optimizes the hyperparameters (feature scale and signal variance) of the Gaussian process regression model using the maximum likelihood estimation method, aiming to maximize the log-likelihood function of the sample data. The optimal hyperparameter combination is then iteratively solved using numerical optimization algorithms. Probability density modeling allows the model parameters to adaptively fit the data distribution, improving prediction accuracy. The gradient of the log-likelihood function guides the optimization direction, and the numerical algorithm effectively avoids local extremum traps in non-convex problems. Jointly optimizing the feature scale and signal variance, dynamically balancing parameter sensitivity and global fluctuation characteristics, can significantly enhance the model's ability to model complex nonlinear relationships and its generalization performance.
[0035] The second building module 3 is used to combine the Gaussian process regression model to construct the optimization metric tensor of the process parameter set.
[0036] Among them, the optimization metric tensor is a parameter space geometric description tool built based on the gradient information and covariance structure of the Gaussian process regression model. It is used to quantify the local sensitivity and directional rate of change of process parameters under the quality parameter optimization objective. By introducing geometric constraints on the parameter space, this tensor can guide the optimization algorithm to search along the steepest direction of quality parameter change (geodesic), effectively avoiding suboptimal solutions caused by parameter coupling in traditional optimization. At the same time, combined with the uncertainty estimation of the Gaussian process, it significantly improves the globality of parameter space exploration and the stability of the optimization path.
[0037] In one possible implementation, the second building module 3 is specifically used for: The Gaussian process regression model is used to output the quality parameters of the integrated die casting.
[0038] The specific prediction formula is as follows: .
[0039] in, Indicated in the process parameter group The predicted quality parameters of integrated die-cast parts are output by the Gaussian process regression model. Indicates the new process parameter group With process parameter group The correlation values of different parameters are the covariance function values. Indicates process parameter group The following are the quality parameters of the integrated die-cast parts. Represents the identity matrix.
[0040] It should be noted that this process quantifies the correlation between new process parameters and historical samples through the covariance function and combines it with the identity matrix to handle noise, enabling the Gaussian process regression model to dynamically output predicted quality parameters and their confidence intervals. The complex relationship between parameters and quality is flexibly modeled using a nonlinear kernel function, adapting to high-dimensional coupling scenarios. The prediction based on a probabilistic framework provides uncertainty estimation, offering a reliable basis for optimization decisions. Adaptive adjustment of the covariance function parameters ensures the model maintains high generalization ability even with small sample conditions. Through the joint calculation of the correlation matrix and its inverse, the sensitivity distribution of parameter changes to quality parameters is accurately captured, significantly improving prediction accuracy and robustness.
[0041] Establish a Jacobian matrix based on the predictive integrated quality parameters of die-cast parts.
[0042] The formula for the Jacobian matrix is as follows: .
[0043] in, , and These represent different predicted quality parameters for integrated die-cast parts: predicted porosity, predicted tensile strength, and predicted dimensional deviation. Indicates information about the new process parameter group The Jacobian matrix.
[0044] An optimized metric tensor is constructed based on the Jacobian matrix.
[0045] The formula for optimizing the metric tensor is as follows: .
[0046] .
[0047] in, Indicates information about the new process parameter group Optimization of the metric tensor This represents the weight matrix for predicted porosity, predicted tensile strength, and predicted dimensional deviation. These represent the importance weights for predicted porosity, predicted tensile strength, and predicted dimensional deviation, respectively. This represents the regularization coefficient.
[0048] Optionally, the importance weights for predicting porosity, tensile strength, and dimensional deviation can be set to 0.3, 0.5, and 0.4, respectively.
[0049] It should be noted that this optimized metric tensor can be used when the process parameters are more sensitive to quality indicators (i.e., The larger the norm, the better. The larger the eigenvalue, the shorter the local distance of the manifold, and the optimized path automatically avoids regions where parameter changes have a drastic impact on quality (regions of sharp changes). The lower the sensitivity of process parameters, The smaller the eigenvalue, the greater the local distance expansion of the manifold, allowing for a more free search of the optimization path in that direction, thus guiding the optimization towards a more stable parameter direction.
[0050] Specifically, this process predicts quality parameters under new process parameters using a Gaussian process regression model, quantifies the sensitivity of parameters to quality indicators based on the Jacobian matrix, and constructs an optimization metric tensor by combining a weight matrix and a regularization term. The Jacobian matrix captures the local gradient changes of quality parameters with respect to process parameters, providing a mathematical basis for sensitivity analysis in the parameter space. The optimization metric tensor reflects the differences in importance of different quality indicators through weight allocation, guiding the optimization path to prioritize key performance indicators. The regularization term prevents matrix ill-conditioned problems and ensures the stability of geometric constraints. Based on the joint modeling of sensitivity and geometric constraints, the optimization path automatically avoids regions with drastic parameter influence (such as steeply changing sensitive regions), searches more efficiently in low-sensitivity regions, significantly improves the global stability and convergence speed of parameter optimization, and enhances the comprehensive balancing ability for multiple objective quality indicators.
[0051] The first module 4 is used to establish a single-objective optimization function for the process parameter group with the integrated die-casting part quality parameter group as the optimization objective.
[0052] The single-objective optimization function maps the multi-dimensional quality parameters of the integrated die-cast part into a single numerical objective function, which is used to quantify the comprehensive impact of the process parameter set on the overall quality. This function avoids the complex trade-offs of multi-objective optimization by integrating the optimization requirements of multiple quality parameters. Furthermore, by combining the nonlinear modeling capabilities of Gaussian process regression, it can efficiently search for the globally optimal combination of process parameters, significantly improving optimization efficiency and the practicality of the results.
[0053] In one possible implementation, the single-objective optimization function is specifically the product of the transpose of the weight matrix and the integrated die-casting quality parameter set.
[0054] It should be noted that this single-objective optimization function uses a weight matrix to linearly weight and integrate multi-dimensional quality parameters, transforming the multi-objective optimization problem into a single-objective optimization problem. The weight allocation directly reflects the priority differences of different quality indicators (such as porosity, strength, and dimensional deviation) in actual production, ensuring that the optimization direction aligns with engineering requirements. This simplifies the complex trade-offs between multiple objectives, significantly reduces computational complexity, and improves algorithm convergence efficiency.
[0055] The second module, Module 5, is used to establish the geodesic equation of a single-objective optimization function by combining the optimization metric tensor.
[0056] The geodesic equation, constructed based on the non-Euclidean geometry of the parameter space (defined by the optimization metric tensor), is the shortest path equation used to describe the search trajectory along the direction of the fastest change in mass parameters within the parameter space. Essentially, it transforms the optimization problem into a path search problem on the parametric manifold using differential geometry, ensuring that the optimization path always extends along the "optimal direction" of quality improvement. By introducing geometric constraints into the parameter space, this method effectively avoids the deviation of the search path caused by parameter coupling in traditional Euclidean space, significantly improving the global convergence and stability of the optimization path. Furthermore, by incorporating uncertainty estimation of Gaussian processes, it further enhances its adaptability to complex nonlinear optimization problems.
[0057] In one possible implementation, the second establishing module 5 is specifically used for: Compute the Christofel notation for the optimized metric tensor.
[0058] Christofel notation is a key quantity in differential geometry describing the curvature of parameter space, used to quantify the impact of coordinate changes on the parallel translation of vectors on a manifold. In optimizing the metric tensor, it reflects the geometric nonlinearity of the parameter space by calculating the partial derivatives of the metric tensor, providing curvature correction terms for the geodesic equations. Christofel notation enables the optimization path to dynamically adapt to the nonlinear structure of the parameter space, avoids suboptimal solutions in traditional Euclidean space by introducing geometric constraints, significantly improves the global convergence and stability of the optimization algorithm, and enhances the ability to model complex parameter coupling relationships.
[0059] Geodesic equations are established based on Christofel notation.
[0060] The specific formulas for calculating the baseline and the Christofel symbol are as follows: .
[0061] .
[0062] .
[0063] in, Indicates process parameter group X Related single-objective optimization functions, This represents the transpose of the weight matrix. Indication and process parameter group X The corresponding integrated die-casting part quality parameter set, , , and They represent the first i The process parameter, the first j The process parameter, the first k The process parameters and the first l One process parameter, Indicates corresponding to the first i The process parameters and the first j The optimization metric tensor of the process parameter is the first... l Line number j Column elements, and Similarly, This represents the inverse matrix component of the optimization metric tensor. Indicates the first i The first process parameter direction j The and the first k The interaction value of each process parameter is the Christofel symbol. Indicates the adaptive step size factor. Represents a time variable. Representing time variables t Related update step size.
[0064] Among them, the adaptive step size factor is a control parameter used in the geodesic equation to dynamically adjust the parameter update step size. Its core function is to balance the convergence speed and stability of the optimization path.
[0065] Specifically, this process constructs geodesic equations in the parameter space by calculating the Christofel notation of the optimization metric tensor, transforming the optimization problem into a path search problem on a curved manifold. By quantifying the geometric curvature of the parameter space using the Christofel notation, the geodesic equations automatically adapt to the nonlinear coupling relationships between parameters, guiding the optimization path along the direction most sensitive to changes in quality parameters. Combining the gradient term of the single-objective optimization function with the inverse matrix of the metric tensor allows the optimization direction to be simultaneously driven by the quality objective and constrained by the geometric properties of the parameter space, significantly improving the globality and stability of the path search. An adaptive step size factor dynamically adjusts the search step size, avoiding the uneven convergence speed problem caused by parameter scale differences in traditional Euclidean space. Ultimately, this achieves the discovery of the globally optimal solution for the combination of process parameters, significantly improving the efficiency and robustness of die-casting quality optimization.
[0066] Solver Module 6 is used to iteratively solve the geodesic equation with the goal of minimizing the single-objective optimization function, and obtain the target process parameter set.
[0067] The target process parameter set is the optimal combination of process parameters obtained by iteratively solving the geodesic equation.
[0068] In one possible implementation, the solver module 6 is specifically used for: S601: Initialize the initial process parameter group.
[0069] S602: Calculate the initial Christofel sign and the initial gradient of the single-objective optimization function relative to the initial set of process parameters.
[0070] S603: Substitute the initial Christofel symbol and the initial gradient into the geodesic equation to obtain the initial geodesic equation.
[0071] S604: Discretize the initial geodesic equations to generate an optimized set of process parameters.
[0072] In one possible implementation, discretizing the initial geodesic equations to generate the optimized set of process parameters specifically involves: The initial geodesic equations are discretized using the fourth-order Runge-Kutta method to generate an optimized set of process parameters.
[0073] The fourth-order Runge-Kutta method is a numerical integration method that significantly improves the accuracy and stability of numerical solutions to differential equations by incorporating a weighted average of four different slopes in each calculation step. Its core idea is to balance local truncation errors and computational costs through a phased prediction-correction strategy. When discretizing geodesic equations, the fourth-order Runge-Kutta method effectively suppresses numerical oscillations by approximating continuous paths with high-order accuracy, while maintaining adaptability to the nonlinear geometric structure of the parameter space. This ensures the accuracy and stability of the optimized path calculation and significantly improves the robustness of solving complex differential equations.
[0074] Specifically, the second-order geodesic equations are first transformed into a system of first-order equations (assuming...). ,but For the initial process parameter set X and corresponding parameters v The first step is to calculate the initial time. t slope and The second step uses... , calculate Find the intermediate parameter point at time step and recalculate the intermediate parameter point. Similarly, the slope can then be obtained. and The third step is to repeat the second step, that is, use... Calculate another set of slopes and Part 4 calculate slope at time and Finally, a weighted average of the four slopes is calculated. Multiply by step size That is, to obtain the process parameter points at the next moment. In this process, the calculation of intermediate points at each step needs to be recalculated. This method captures the nonlinear effects caused by parameter variations, ensuring discretization accuracy. Through iterative processing, the process parameter points gradually approach the global optimum along the geodesic direction. All undefined parameters appearing in the above process are intermediate variables, and their relationships are given.
[0075] It should be noted that this process discretizes the geodesic equations using the fourth-order Runge-Kutta method, transforming the second-order differential equations into a system of first-order equations. A phased prediction-correction strategy is then employed to calculate the weighted average of the four slopes, approximating the nonlinear path in the parameter space with high-order accuracy. Multi-stage slope weighted averaging significantly improves the accuracy and stability of the numerical solution, effectively suppressing numerical oscillations during iteration. Dynamically capturing the geometric nonlinear effects caused by parameter changes ensures the adaptability of the optimization path to complex manifolds. Combining geodesic direction and gradient constraints enables parameter updates to possess both global search capability and local convergence stability, significantly improving the robustness and efficiency of solving high-dimensional nonlinear optimization problems.
[0076] S605: Calculate the single-objective optimization function value under the optimized process parameter group.
[0077] S606: If the single-objective optimization function value under the optimized process parameter group is less than the single-objective optimization function value under the initial process parameter group, update the initial process parameter group using the optimized process parameter group and proceed to step S608; otherwise, proceed to step S607.
[0078] S607: Reduce the update step size and return to step S602.
[0079] S608: Return to step S601 until the initial gradient norm is less than the preset initial gradient norm or the number of iterations is greater than the preset number of iterations, and output the updated target process parameter set.
[0080] Specifically, the solution process searches for the target process parameter set in the parameter space by iteratively optimizing the geodesic equation. The steps include: initializing the parameter set, calculating geometric constraints (Christofel notation) and optimization gradients, substituting these into the geodesic equation to generate a discretized optimization path, and dynamically evaluating the single objective function value to determine whether to update the parameter set or adjust the step size. The geodesic equation, based on differential geometry, guides the optimization path along the "shortest path" in the parameter space, effectively avoiding local sensitive regions (such as abrupt change sensitive regions) and improving global search capabilities. Combining gradient descent and geometric constraints in a joint optimization mechanism ensures that the search direction is consistent with the quality objective while adjusting the adaptive step size to balance convergence speed and stability. Discretization transforms the high-dimensional nonlinear optimization problem into a computable numerical iterative process, significantly improving the algorithm's practicality in complex parameter spaces.
[0081] It should be noted that those skilled in the art can set the preset initial gradient norm or the preset number of iterations according to actual needs, and this invention does not limit this.
[0082] In one possible implementation, it also includes: Update module 7 is used to update historical sample data at preset intervals.
[0083] Understandably, this process dynamically updates the historical sample database by periodically supplementing it with newly collected process and quality parameter samples, ensuring that the Gaussian process regression model and optimization framework are always modeled and iterated based on the latest data. It should be noted that those skilled in the art can set the preset duration according to actual needs, and this invention does not limit this.
[0084] In practical applications, this solution achieves global optimization of complex nonlinear parameter spaces by constructing a Gaussian process regression model. The specific process is as follows: A Gaussian process regression model is established based on historical sample data. The nonlinear coupling relationship between process parameters and quality parameters is accurately captured through the squared exponential kernel function and hyperparameter optimization. The parameter sensitivity is quantified by combining the Jacobian matrix and the optimization metric tensor, constructing geometric constraints for the parameter space. The optimal solution is searched along the shortest path of the parameter manifold using the geodesic equation and the fourth-order Runge-Kutta method. Probabilistic modeling and uncertainty estimation improve the prediction robustness under small sample conditions. Differential geometric constraints (such as Christofel notation and geodesic equations) are introduced to effectively avoid local optima traps caused by parameter coupling, ensuring the global convergence of the optimization path. A dynamic update module allows the system to adapt to the time-varying characteristics of the production environment, continuously optimizing the model and parameter combinations. A single-objective optimization function and weight allocation mechanism balance the priorities of multiple quality indicators, significantly improving optimization efficiency and the practicality of the results. This overcomes the limitations of traditional linear models and empirical trial and error, ensuring intelligent optimization of the die-casting process with high precision and high stability.
[0085] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, a Gaussian process regression model is used to accurately model the nonlinear coupling relationship between process parameters and quality parameters based on historical data. This is combined with an optimization metric tensor to construct a geometric constraint framework for the parameter space. Geodesic equations are used to guide the global search path, effectively avoiding the local optima traps caused by traditional linear models or empirical rules. High-dimensional interactions between parameters can be explored, and probabilistic modeling enables rapid searching of the parameter space, significantly improving optimization efficiency and the stability of output process parameters. Simultaneously, it reduces reliance on human experience, achieving a global optimal solution search for process parameter combinations, and improving the production quality and efficiency of integrated die-cast parts.
[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.
Claims
1. An integrated die-casting part process parameter optimization system, characterized in that, include: The acquisition module is used to acquire historical sample data of the integrated die casting, wherein the historical sample data includes process parameter groups and corresponding integrated die casting quality parameter groups; The first construction module is used to construct a Gaussian process regression model based on the historical sample data to achieve a nonlinear mapping between the process parameter set and the integrated die-casting part quality parameter set. The second construction module is used to construct the optimization metric tensor of the process parameter set by combining the Gaussian process regression model. The first module is used to establish a single-objective optimization function for the process parameter group, taking the integrated die-casting part quality parameter group as the optimization target. The second module is used to establish the geodesic equation of the single-objective optimization function by combining the optimization metric tensor; The solution module is used to iteratively solve the geodesic equation with the goal of minimizing the single-objective optimization function, so as to obtain the target process parameter set.
2. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, The process parameter set includes injection speed, mold temperature, melt temperature, and pressurization pressure; the integrated die casting quality parameter set includes porosity, tensile strength, and dimensional deviation.
3. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, The first building module is specifically used for: Establish a covariance function to describe the correlation between the various process parameters in the process parameter group; An initial Gaussian process regression model is established based on the aforementioned covariance function; Estimate the hyperparameters of the initial Gaussian process regression model to obtain the Gaussian process regression model, wherein the hyperparameters include the characteristic scales of each process parameter and the signal variance describing the variation of the quality parameters of the integrated die casting with the process parameters.
4. The process parameter optimization system for integrated die-casting parts according to claim 3, characterized in that, The estimation of the hyperparameters of the initial Gaussian process regression model specifically involves: The hyperparameters of the initial Gaussian process regression model are estimated using the maximum likelihood estimation method.
5. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, The second building module is specifically used for: The Gaussian process regression model is used to output the predicted quality parameters of the integrated die-casting part. Establish a Jacobian matrix based on the predicted integrated die-casting quality parameters; Based on the Jacobian matrix, the optimized metric tensor is constructed.
6. The process parameter optimization system for integrated die-casting parts according to claim 5, characterized in that, The single-objective optimization function is specifically the product of the transpose of the weight matrix and the integrated die-casting part quality parameter set.
7. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, The second establishment module is specifically used for: Calculate the Christofel symbol for the optimized metric tensor; The geodesic equation is established based on the Christofel notation.
8. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, The solution module is specifically used for: S601: Initialize the initial process parameter group; S602: Calculate the initial Christofel symbol under the initial set of process parameters and the initial gradient of the single-objective optimization function relative to the set of process parameters; S603: Substitute the initial Christofel symbol and the initial gradient into the geodesic equation to obtain the initial geodesic equation; S604: Discretize the initial geodesic equations to generate an optimized set of process parameters; S605: Calculate the single-objective optimization function value under the optimized process parameter set; S606: If the single-objective optimization function value under the optimized process parameter group is less than the single-objective optimization function value under the initial process parameter group, update the initial process parameter group using the optimized process parameter group and proceed to step S608; otherwise, proceed to step S607. S607: Reduce the update step size and return to step S602; S608: Return to step S601 until the initial gradient norm is less than the preset initial gradient norm or the number of iterations is greater than the preset number of iterations, and output the updated target process parameter set.
9. The process parameter optimization system for integrated die-casting parts according to claim 8, characterized in that, The discretization of the initial geodesic equations to generate an optimized set of process parameters is specifically as follows: The initial geodesic equations are discretized using the fourth-order Runge-Kutta method to generate the optimized process parameter set.
10. The process parameter optimization system for integrated die-casting parts according to claim 1, characterized in that, Also includes: The update module is used to update the historical sample data at preset intervals.