Rearview mirror support parameter optimization system based on cloud edge

Abstract

The invention belongs to the field of computer-aided engineering and design optimization, and particularly relates to a rearview mirror bracket parameter optimization system based on a cloud edge, which comprises a multi-source excitation scene modeling unit used for calling multi-source vibration excitation data in a cloud database and performing mathematical fusion processing on the multi-source vibration excitation data to construct a total excitation spectrum; the dynamic response agent model unit is used for obtaining vibration response power spectrum density; the dynamic definition evaluation unit is used for obtaining a dynamic visual definition index and carrying out discrimination processing on the dynamic visual definition index to obtain an unqualified signal, a qualified signal or an excellent signal; and the collaborative optimization control unit is used for responding to the qualified signal or the excellent signal, constructing a comprehensive objective function based on the dynamic visual definition index, the bracket quality and the material cost, and performing cloud parameter optimization processing to obtain a hierarchical optimization scheme. According to the system, the reality fidelity of simulation analysis is improved from the source.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided engineering and design optimization, specifically a rearview mirror bracket parameter optimization system based on cloud edge computing. Background Technology

[0002] As the automotive industry continues to demand higher standards for driving experience and safety performance, vibration control of rearview mirrors during driving, especially ensuring dynamic visual clarity, has become a key technical indicator in vehicle development. As a key structural component connecting the mirror body to the vehicle body, the design of the rearview mirror bracket directly affects the vibration transmission characteristics and the final image quality.

[0003] Currently, the traditional rearview mirror bracket design process largely relies on the designer's personal engineering experience and costly computer-aided engineering simulation. This process typically involves the designer proposing an initial design based on experience, followed by performance verification through finite element analysis, which takes several hours or even longer. If the vibration performance does not meet the standards, the design parameters need to be manually adjusted and the lengthy simulation verification process needs to be repeated. This iterative approach, which relies on manual trial and error, has many drawbacks: its development cycle is long and inefficient, and it cannot meet the design needs of rapid iteration; the high computational cost limits the number of design solutions that can be explored; the quality of the design results heavily depends on the designer's experience level, and there is a lack of direct quantitative correlation between complex physical vibration data and the driver's intuitive visual perception, resulting in unclear optimization objectives.

[0004] Therefore, how to transform the traditional rearview mirror bracket design, which relies on high computing costs and designer experience, into an intelligent and automated design optimization process driven by data and empowered by models, in order to systematically discover innovative design solutions that balance performance, cost, and lightweighting, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention discloses a rearview mirror bracket parameter optimization system based on cloud edge computing. Specifically, the technical solution of this invention is as follows: A cloud edge-based rearview mirror bracket parameter optimization system includes: The multi-source excitation scenario modeling unit is used to retrieve multi-source vibration excitation data from the cloud database and perform mathematical fusion processing on the multi-source vibration excitation data to construct the total excitation spectrum; The dynamic response proxy model unit is used to call a preset AI proxy model to perform real-time predictive analysis based on the total excitation spectrum and the current design parameters input by the designer, so as to obtain the vibration response power spectral density. The dynamic sharpness evaluation unit is used to combine the vibration response power spectral density and the preset visual sensitivity function to calculate and analyze the dynamic sharpness index, obtain the dynamic visual sharpness index, and perform discrimination processing on the dynamic visual sharpness index to obtain unqualified signal, qualified signal or excellent signal. The collaborative optimization control unit is used to respond to qualified or excellent signals, construct a comprehensive objective function based on dynamic visual clarity index, bracket quality and material cost, and perform cloud-based parameter optimization to obtain a hierarchical optimization scheme.

[0006] Preferably, the multi-source vibration excitation data includes road surface excitation data, powertrain excitation data, and aerodynamic excitation data; the mathematical fusion processing is as follows: Obtain the weighting coefficients of road surface excitation data, powertrain excitation data, and aerodynamic excitation data under specific virtual test conditions; The aforementioned incentive data and their corresponding weighting coefficients are then weighted and superimposed to construct the total incentive spectrum.

[0007] Preferably, the AI ​​agent model is constructed in the following way: Using the total excitation spectrum and design parameters as load inputs, transient dynamic finite element analysis is performed to obtain a high-precision simulation dataset; A supervised learning paradigm is used to train a high-precision simulation dataset to obtain an AI agent model that can accurately fit the nonlinear relationship between input and output.

[0008] Preferably, the dynamic sharpness index calculation and analysis process is as follows: The total ambiguity is determined by integrating the vibration response power spectral density with a preset visual sensitivity function. The reference ambiguity is determined by integrating the preset reference vibration reference spectrum and the preset visual sensitivity function. The total ambiguity is then compared with the baseline ambiguity to generate a dynamic visual sharpness index.

[0009] Preferably, the reference vibration reference spectrum is a preset vibration reference that represents unacceptable vibration based on the vehicle manufacturer's design specifications and user experience standards.

[0010] Preferably, the discrimination process for the dynamic visual acuity index is as follows: When the dynamic visual clarity index is less than or equal to the preset performance baseline threshold, an unqualified signal is generated; A qualified signal is generated when the dynamic visual clarity index is greater than the minimum performance threshold and less than or equal to the preset excellent performance threshold. When the dynamic visual clarity index is greater than the excellent performance threshold, an excellent signal is generated.

[0011] Preferably, the process of constructing the comprehensive objective function is as follows: Obtain the performance weighting coefficient, quality weighting coefficient, and cost weighting coefficient set by the designer; Determine target reference values ​​for the normalized dynamic visual acuity index, support mass, and material cost, respectively. The dynamic visual acuity index, bracket quality, and material cost are normalized with their respective target reference values ​​to obtain normalized performance, normalized quality, and normalized cost items. The normalized performance term, normalized quality term, and normalized cost term are linearly combined using corresponding weighting coefficients to construct a comprehensive objective function.

[0012] Preferably, the tiered optimization scheme includes the highest resolution scheme, the lightest scheme, and the best overall scheme; the tiered optimization scheme is pushed to the edge device where the designer is located for decision-making.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This system improves the realism of simulation analysis from the source. Traditional designs often use simplified load inputs, which are difficult to realistically reflect the complex vibration environment that rearview mirrors experience in actual service. The multi-source excitation scenario modeling unit in this system retrieves multi-source vibration excitation data covering road surface, powertrain, and aerodynamics from a cloud database, and performs mathematical fusion processing based on the weighting coefficients of specific virtual test conditions to construct a total excitation spectrum that can comprehensively reflect the complex coupled vibration characteristics. This ensures that the load input for subsequent dynamic analysis has high fidelity and improves the realism, effectiveness, and reliability of the optimization results. 2. This system achieves real-time and interactive design feedback, transforming the R&D model. Traditional design processes are constrained by the lengthy computational cycle of finite element analysis, requiring designers to wait hours or even longer for performance feedback after each parameter modification, hindering efficient iterative design. The dynamic response proxy model unit in this system replaces the high-computing finite element analysis with an AI proxy model trained in the cloud. Deployed at the edge where the designer is located, this model can perform millisecond-level real-time predictions based on the total excitation spectrum and the current design parameters input by the designer, instantly outputting the vibration response power spectral density. This mode transforms the static and lagging verification process into a dynamic and interactive design exploration, significantly compressing the R&D cycle. 3. This system establishes a quantitative evaluation system directly related to user subjective perception. Traditional technologies for evaluating vibration performance often rely on interpreting complex physical spectra, resulting in a lack of direct, quantitative correlation between the results and the driver's actual visual experience, leading to subjectivity and ambiguity in the evaluation process. The dynamic clarity evaluation unit in this system innovatively combines the vibration response power spectral density with a preset visual sensitivity function and compares it with a reference vibration baseline spectrum representing unacceptable vibration levels, ultimately generating a standardized dynamic visual clarity index. This index transforms complex physical quantities into a single, intuitive performance indicator directly linked to end-user perception, and further classifies them as unqualified, qualified, or excellent signals, providing a clear, objective, and engineering-guiding decision-making basis for performance evaluation and automated control logic. 4. This system upgrades manual trial-and-error parameter tuning to a systematic multi-objective intelligent optimization process. Traditional optimization processes rely heavily on designers' experience for manual trial and error, which is not only inefficient but also makes it difficult to find the globally optimal solution among multiple conflicting objectives such as performance, quality, and cost. The collaborative optimization control unit in this system can automatically construct a comprehensive objective function based on the dynamic visual clarity index, bracket quality, and material cost after receiving a qualified or excellent signal. This unit uses cloud computing power to run multi-objective optimization algorithms, performing tens of thousands of high-speed virtual iterations to systematically explore the entire design space. Finally, the system extracts hierarchical optimization solutions such as the highest clarity solution, the lightest solution, and the best overall solution and pushes them to the edge. This data- and algorithm-driven intelligent recommendation mode replaces the tuning process that relies on manual experience, and can systematically discover innovative design solutions that take into account multiple objectives, significantly improving the market competitiveness of the final product. Attached Figure Description

[0014] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. Example 1:

[0016] Please see Figure 1 A cloud edge-based rearview mirror bracket parameter optimization system includes: The multi-source excitation scenario modeling unit is used to retrieve multi-source vibration excitation data from the cloud database and perform mathematical fusion processing on the multi-source vibration excitation data to construct the total excitation spectrum; The dynamic response proxy model unit is used to call a preset AI proxy model to perform real-time predictive analysis based on the total excitation spectrum and the current design parameters input by the designer, so as to obtain the vibration response power spectral density. The dynamic sharpness evaluation unit is used to combine the vibration response power spectral density and the preset visual sensitivity function to calculate and analyze the dynamic sharpness index, obtain the dynamic visual sharpness index, and perform discrimination processing on the dynamic visual sharpness index to obtain unqualified signal, qualified signal or excellent signal. The collaborative optimization control unit is used to respond to qualified or excellent signals, construct a comprehensive objective function based on dynamic visual clarity index, bracket quality and material cost, and perform cloud-based parameter optimization to obtain a hierarchical optimization scheme.

[0017] This invention provides a cloud-edge based rearview mirror bracket parameter optimization system. The system aims to transform the traditional rearview mirror bracket design process, which relies on high computational costs and designer experience, into an intelligent and automated design optimization process driven by data, empowered by artificial intelligence models, and featuring cloud-edge collaboration. In this embodiment, the system is deployed on a cloud-edge collaborative architecture, where the cloud is responsible for large-scale data storage, high-performance model training, and multi-objective optimization, while the edge, i.e., the designer's workstation, is responsible for real-time access to lightweight models and interactive design. Logically, the system consists of four collaborative core units, forming a complete technical closed loop from problem definition to solution recommendation. The multi-source excitation scenario modeling unit aims to provide high-fidelity quantified load input for subsequent dynamic response analysis. This unit retrieves multi-source vibration excitation data collected and standardized under real driving scenarios from a cloud database. Through a specific mathematical fusion processing method, this unit combines multiple independent excitation data into a total excitation spectrum that can comprehensively reflect the complexity of a specific driving scenario. This processing ensures the accuracy of the source data for simulation analysis and improves the practical effectiveness of the optimization results. The Dynamic Response Proxy Model Unit aims to predict the vibration response of rearview mirror brackets under given excitation in real time with extremely low computational cost, thereby solving the technical bottleneck of traditional finite element analysis being too time-consuming and unable to support interactive design. At the core of this unit is a pre-trained AI proxy model in the cloud. When a designer inputs a set of current design parameters at the edge, such as rib thickness and material, the unit immediately invokes this AI proxy model and, combined with the total excitation spectrum generated in previous steps, predicts the vibration response power spectral density of the rearview mirror lens mounting point within milliseconds. This allows for immediate performance feedback for every parameter modification by the designer. Furthermore, the unit includes an input parameter verification module. When the design parameters input by the designer exceed the feasible domain covered by the AI ​​model during training, the system will issue a warning, indicating that the accuracy of the prediction result may not be guaranteed, thus ensuring the validity of the prediction.

[0018] The purpose of the dynamic visual acuity assessment unit is to transform the complex physical quantities of vibration response into a single, quantifiable dynamic visual acuity index directly related to the driver's subjective perception. This unit combines the vibration response power spectral density predicted by an artificial intelligence model with a preset visual sensitivity function characterizing the human eye's sensitivity to vibrations at different frequencies. Through integration, it obtains the quantified total ambiguity. This total ambiguity is then compared with a benchmark representing an unacceptable vibration level to generate a standardized dynamic visual acuity index. This index intuitively reflects the performance level of the current design. Furthermore, the unit performs discrimination based on this index, outputting clear signals of non-compliance, compliance, or excellence, providing a basis for subsequent optimization control decisions. The Collaborative Optimization Control Unit aims to automatically explore a broad design space and intelligently recommend a set of optimal design solutions based on multiple objectives set by the designer, such as performance, quality, and cost. This unit is activated when the Dynamic Visual Clarity Evaluation Unit determines that the current solution is qualified or excellent. It constructs a comprehensive objective function that linearly combines three core indicators: Dynamic Visual Clarity Index, bracket quality, and material cost. This unit utilizes cloud computing power, guided by this comprehensive objective function, to run a multi-objective optimization algorithm for tens of thousands of high-speed virtual iterations, finds a set of Pareto optimal solutions, and selects several representative hierarchical optimization solutions from them to push to the designer. This invention constructs a closed-loop, automated rearview mirror bracket parameter optimization system through the collaborative work of the four units mentioned above. It transforms the simulation analysis process, which takes several hours, into real-time prediction, converts complex physical vibration data into intuitive visual clarity indicators, and upgrades the tuning process, which relies on manual trial and error, into an algorithm-driven intelligent recommendation mode. This greatly improves R&D efficiency, reduces reliance on the personal experience of designers, and can systematically discover innovative design solutions that balance performance, cost, and lightweight design, significantly enhancing the market competitiveness of the final product.

[0019] Example 2: Multi-source vibration excitation data includes road surface excitation data, powertrain excitation data, and aerodynamic excitation data; the mathematical fusion processing is as follows: Obtain the weighting coefficients of road surface excitation data, powertrain excitation data, and aerodynamic excitation data under specific virtual test conditions; The aforementioned incentive data and their corresponding weighting coefficients are then weighted and superimposed to construct the total incentive spectrum.

[0020] This embodiment is a specific implementation of the multi-source excitation scenario modeling unit described in Embodiment 1. In this embodiment, the multi-source vibration excitation data refers to three types of core datasets retrieved by the system from the cloud-based real road excitation database. Their function is to provide basic data for the reproduction of real driving scenarios. These three types of data include: Road excitation data: represented by the power spectral density of vibration acceleration of different road surfaces at different vehicle speeds; Powertrain excitation data: This refers to the frequency and amplitude of vibrations generated by the engine or electric motor in various speed ranges and transmitted through the vehicle structure to the rearview mirror mounting point. Aerodynamic excitation data: represented by the pressure fluctuation spectrum caused by wind noise or vehicle vibration at different vehicle speeds and wind speeds; To reproduce the real coupled vibration environment, the mathematical fusion process in this embodiment is based on the excitation superposition principle in linear system theory. The above multi-source excitation data are fused by weighted superposition to construct a unified total excitation spectrum. To accurately construct this total excitation spectrum, this embodiment introduces the total excitation power spectral density function. The calculation method is as follows:

[0021] in: Total excitation power spectral density function, in units of Its source is the calculation obtained by this formula, which serves as the sole input load for subsequent dynamic response analysis; Angular frequency, measured in rad / s; No. The power spectral density function of a single-source excitation, in units of Its source is road surface, powertrain, or aerodynamic excitation data from the aforementioned cloud database; No. The weight coefficients of the various excitation sources are dimensionless. They are derived from empirical values ​​calibrated by test engineers based on a large amount of real vehicle test data for specific virtual test conditions, such as high-speed cruising or bumpy urban roads, and are stored in conjunction with the test conditions. The total number of stimulus sources, an integer, whose source is determined by the currently selected virtual test case; This formula transforms multiple isolated excitation data from different physical sources into a single mathematical model that can comprehensively and accurately reflect the complex coupled vibration characteristics under specific driving scenarios and can be directly used for engineering calculations. During the calculation, the system first obtains the corresponding weighting coefficients based on the virtual test conditions selected by the designer. Set up the dataset and then analyze the various incentive data. Its corresponding Weighted superposition is performed; the total excitation spectrum constructed in this way Compared to using a single excitation source for analysis, it can more realistically simulate the complex vibration environment that rearview mirrors experience in actual service, thereby significantly improving the source fidelity of the simulation analysis and the realistic reliability of the final optimization results. This embodiment ensures that the construction process of the total excitation spectrum has clear physical meaning and engineering operability by clearly defining the sources of excitation data, namely road surface, powertrain, and aerodynamics, and by adopting a weighted superposition method based on operating condition weight coefficients. This greatly improves the simulation fidelity of the virtual test scenario to the real driving environment.

[0022] Example 3: AI agent models are constructed in the following ways: Using the total excitation spectrum and design parameters as load inputs, transient dynamic finite element analysis is performed to obtain a high-precision simulation dataset; A supervised learning paradigm is used to train a high-precision simulation dataset to obtain an AI agent model that can accurately fit the nonlinear relationship between input and output.

[0023] This embodiment is a specific implementation of the artificial intelligence proxy model construction method in the dynamic response proxy model unit described in Embodiment 1; its core technical motivation is to use an artificial intelligence model with extremely fast computing speed to replace or proxy the computationally expensive finite element analysis process, thereby realizing real-time performance feedback of the design; the construction of this model follows a two-stage process; The first stage involves generating a high-precision simulation dataset. This process is executed on a cloud computing cluster, with the aim of providing a high-quality training dataset for training artificial intelligence models. Specifically, it utilizes design experimentation methods such as Latin hypercube sampling to generate hundreds to thousands of well-covered design parameter vectors within a designer-preset design space, such as the thickness range of key ribs and a list of optional material elastic moduli. The total excitation spectrum generated in Example 2 As a standard load input, for each set of design parameter vectors A complete high-precision transient dynamic finite element analysis was performed to calculate the vibration response power spectral density at the rearview mirror lens mounting point. The inputs and outputs are combined into data pairs to form the basic dataset required for training artificial intelligence models. The second stage is the training of the AI ​​agent model. This embodiment adopts a supervised learning paradigm, specifically using models such as deep neural networks and gradient boosting trees to accurately fit the complex nonlinear relationship between input and output. After training, the resulting AI agent model... This can be expressed as the following functional relationship:

[0024] in: Vibration response power spectral density predicted by the artificial intelligence surrogate model, in units of Its origin is the artificial intelligence agent model. Calculated in real time; : A trained artificial intelligence agent model, a non-linear function mapping relationship, which internally embeds network weights and biases obtained by learning from a high-precision simulation dataset; Total excitation spectrum, in units of Its source is provided by the multi-source excitation scenario modeling unit; The vector of design parameters currently input by the designer, which includes multiple design variables such as the thickness of key ribs, angles, and the elastic modulus of materials; The model After training and optimization are completed in the cloud, the data is synchronized to the edge device where the designer is located. This embodiment details the construction path of the artificial intelligence proxy model, which is to generate high-precision training data through high-precision finite element simulation, and then obtain a high-speed prediction model through supervised learning. This method ensures that the proxy model is not only fast, but also has guaranteed prediction accuracy. The finite element simulation process, which originally required several hours or even longer, is encapsulated into an ultra-high-speed predictor that can be called in real time at the edge, providing designers with instant and interactive performance feedback for parametric design. This is a key technological breakthrough in realizing the intelligent design process.

[0025] Example 4: The dynamic resolution index calculation and analysis process is as follows: The total ambiguity is determined by integrating the vibration response power spectral density with a preset visual sensitivity function. The reference ambiguity is determined by integrating the preset reference vibration reference spectrum and the preset visual sensitivity function. The total ambiguity is then compared with the baseline ambiguity to generate a dynamic visual sharpness index.

[0026] The reference vibration spectrum is a preset vibration standard that represents an unacceptable vibration level, based on the vehicle manufacturer's design specifications and user experience standards.

[0027] This embodiment is a specific implementation of the dynamic sharpness index calculation and analysis process in the dynamic sharpness evaluation unit described in Embodiment 1. This process aims to create a single, quantifiable index that can directly measure the core technical indicator, namely the driver's visual sharpness. Its calculation logic is: calculate the total blur caused by vibration and weighted by human eye perception, compare it with a benchmark blur that represents unacceptable blur, and obtain a normalized sharpness index. The specific calculation process consists of three steps: The first step is to determine the total ambiguity by integrating the vibration response power spectral density with a preset visual sensitivity function; the visual sensitivity function... This refers to a standardized human visual perception model pre-built in the cloud, derived from publicly available psychophysical research. It is a dimensionless normalization function that quantifies the human eye's sensitivity to image movement of different frequencies and amplitudes; total ambiguity. The calculation method is as follows:

[0028] in: Total ambiguity, in units of This represents the mean square acceleration integral value after weighting for visual sensitivity under the current design. The visual sensitivity function is dimensionless and originates from a standardized model pre-built in the cloud. Predicted vibration response power spectral density, in units of Its source is obtained through real-time prediction by the dynamic response surrogate model unit; The cutoff frequency that the human eye is sensitive to, measured in rad / s, is derived from the upper limit of integration preset based on psychophysical experimental data. The second step, to ensure the effectiveness of the dynamic visual acuity index calculation, sets the reference vibration baseline spectrum to a non-zero spectrum, thus guaranteeing that the calculated baseline ambiguity is always positive. The baseline ambiguity is determined by integrating the preset reference vibration baseline spectrum with a preset visual sensitivity function. The reference vibration baseline spectrum is a preset, unacceptable vibration baseline based on the vehicle manufacturer's design specifications and user experience standards. This is a fixed spectrum, its purpose being to provide a clear and quantifiable performance baseline for acuity assessment; its unit is consistent with the predicted vibration response power spectral density. The calculation method for the reference ambiguity is as follows:

[0029] The third step involves comparing the total ambiguity with the baseline ambiguity to generate a dynamic visual sharpness index; the dynamic visual sharpness index... It is a dimensionless final evaluation index, and its calculation method is as follows:

[0030] The physical meaning of this index is clearly defined: when When, it corresponds to the ideal state of no vibration; when When this occurs, it indicates that the visual clarity of the current design is just at an unacceptable critical level; when If the value is less than a certain threshold, it indicates that the design performance is inferior to that critical level. This embodiment introduces the human eye visual sensitivity function. and unacceptable vibration reference spectrum And construct a normalized dynamic visual sharpness index The complex, multi-dimensional physical data of vibration response has been successfully transformed into a single quantitative evaluation index that is directly linked to the subjective perception of end users and has a clear physical meaning. This approach enables the performance evaluation results to directly guide engineering design, greatly improving the relevance and effectiveness of optimization objectives.

[0031] Example 5: The process for determining the dynamic visual sharpness index is as follows: When the dynamic visual clarity index is less than or equal to the preset performance baseline threshold, an unqualified signal is generated; A qualified signal is generated when the dynamic visual clarity index is greater than the minimum performance threshold and less than or equal to the preset excellent performance threshold. When the dynamic visual clarity index is greater than the excellent performance threshold, an excellent signal is generated.

[0032] This embodiment is a specific implementation of the discrimination process of the dynamic visual acuity index in the dynamic sharpness evaluation unit described in Embodiment 1; the purpose of this process is to, based on the dynamic visual acuity index calculated in the preceding steps, The performance of the current design scheme is clearly and automatically quantified and graded, thereby generating clear signals to guide subsequent design decisions or optimization processes; In this embodiment, the discrimination process relies on two preset thresholds: Performance baseline threshold: Its function is to define whether the design meets the most basic engineering requirements; in this embodiment, this threshold is set to 0, which is related to the dynamic visual acuity index. The time represents the physical meaning of being at an unacceptable critical level; Excellent performance threshold This is a preset value greater than 0, for example Its function is to differentiate between acceptable and excellent performance levels; this threshold is set based on the OEM's requirements for a high-end user experience, for example... Represents its perceptual ambiguity Compared to the critical value It has already decreased by 50%; The discrimination and processing logic is as follows: When the dynamic visual clarity index is less than or equal to the preset performance baseline threshold, i.e. The system generates a non-compliance signal; this signal means that the visual stability of the current design fails to meet the most basic engineering requirements and must be significantly modified or redesigned. When the dynamic visual clarity index is greater than the minimum performance threshold but less than or equal to the preset excellent performance threshold, i.e. The system generates a pass signal; this signal means that the design meets engineering specifications and is acceptable, but there is still room for optimization; this signal will trigger the collaborative optimization control unit to perform further optimization based on the current solution; When the dynamic visual sharpness index is greater than the excellent performance threshold, that is The system generates an excellent signal; this signal means that the design exhibits excellent visual stability and can meet the most demanding driving scenario requirements; this signal will also trigger the co-optimization control unit to explore whether better performance in terms of quality or cost can be achieved while maintaining excellent performance. This embodiment transforms continuous sharpness indices into discrete signals of unqualified, qualified, and excellent that can be directly used for program control by setting explicit and physically meaningful thresholds. This design achieves automated and standardized rating of the performance of the design scheme, avoids the subjectivity and ambiguity of human judgment, provides clear and reliable triggering conditions for the subsequent initiation of multi-objective optimization control logic, and ensures the logical rigor of the entire automated optimization process.

[0033] Example 6: The process of constructing the comprehensive objective function is as follows: Obtain the performance weighting coefficient, quality weighting coefficient, and cost weighting coefficient set by the designer; Determine target reference values ​​for the normalized dynamic visual acuity index, support mass, and material cost, respectively. The dynamic visual acuity index, bracket quality, and material cost are normalized with their respective target reference values ​​to obtain normalized performance, normalized quality, and normalized cost items. The normalized performance term, normalized quality term, and normalized cost term are linearly combined using corresponding weighting coefficients to construct a comprehensive objective function.

[0034] This embodiment is a concrete implementation of the construction process of the comprehensive objective function in the collaborative optimization control unit described in Embodiment 1. The purpose of this construction process is to unify multiple design objectives with different dimensions and potential conflicts, namely performance, quality, and cost, into a single, dimensionless comprehensive utility function, thereby providing a clear direction for multi-objective optimization algorithms. The underlying logic of this construction process lies in first obtaining the weights and the target, then normalizing, and finally linearly combining them: The first step is to obtain the performance weighting coefficient, quality weighting coefficient, and cost weighting coefficient set by the designer; weighting coefficients It is a dimensionless parameter that reflects the degree of emphasis placed on different objectives in the current design project; for example, in a performance-priority project, the designer can set... These weighting coefficients are input by the designer through the interactive interface at the edge, and satisfy the constraint that their sum is 1; The second step is to determine target reference values ​​for the normalized dynamic visual acuity index, support mass, and material cost, respectively. Its function is to serve as a normalization benchmark to eliminate the influence of different physical dimensions; in this embodiment, its sources are as follows: Performance target reference value This refers to the excellent performance threshold; the quality target reference value. and cost target reference value Designers can set these parameters based on project design goals, such as target weight and target cost, and these values ​​must be positive real numbers to ensure the validity of the normalized calculation; before constructing the comprehensive objective function, the system will... and These three target reference values ​​are verified, if or If the input value is less than or equal to zero, the designer will be prompted to enter a valid positive real number to avoid calculation errors and ensure the stability of the optimization algorithm. The third step is to normalize the dynamic visual acuity index, bracket quality, and material cost with their respective target reference values. This step converts each sub-target into a dimensionless relative value, resulting in normalized performance, normalized quality, and normalized cost terms. The fourth step involves linearly combining the normalized performance, normalized quality, and normalized cost terms using their corresponding weighting coefficients to construct the final comprehensive objective function. In this embodiment, the optimization objective is to maximize sharpness while minimizing quality and cost; therefore, the overall objective function is defined as:

[0035] in: The objective function is dimensionless, and the optimization algorithm aims to find the design parameter vector that maximizes the value of this function.

[0036] Design parameter vector Constrained within the feasible region Within, feasible region The feasible range of various geometric parameters determined by the manufacturing process is defined; Respectively with design The resolution index, quality, and cost of the link are derived from calculations in previous steps or directly from design parameters; The dimensionless weight coefficients for each item are derived from the designer's input; The target reference values ​​used for normalizing various objectives are derived from the designer's or system presets. This embodiment transforms the designer's multi-dimensional design goals into a precise, single-dimensional, computable optimization goal through a clear and mathematically expressible construction process. This method not only enables efficient solutions to multi-objective optimization problems, but also provides the system with great flexibility through the setting of weight coefficients. It can adapt to the differentiated design requirements of different project stages and product positioning, thereby finding a Pareto optimal solution that is more in line with actual engineering requirements.

[0037] Example 7: The tiered optimization schemes include the highest resolution scheme, the lightest scheme, and the best overall scheme; the tiered optimization schemes are pushed to the edge device where the designer is located for decision-making.

[0038] This embodiment is a concrete implementation of the hierarchical optimization scheme of the output result of the collaborative optimization control unit described in Embodiment 1 after completing the cloud parameter optimization process; the purpose of this process is to extract and organize a large number of Pareto optimal solutions searched by the cloud optimization engine into several design schemes with clear technical orientation and representativeness, so that designers can intuitively understand and make quick decisions. Optimize the engine in the cloud to synthesize the objective function. To guide this process, after tens of thousands of high-speed virtual iterations using an AI agent model and a clarity evaluation model, a Pareto front solution set is obtained. To facilitate designer decision-making, the collaborative optimization control unit does not directly push this massive solution set, but rather performs post-processing and filtering to ultimately generate a hierarchical optimization scheme. In this embodiment, the scheme includes at least the following three categories: Highest resolution scheme: This scheme corresponds to the dynamic visual sharpness index at the Pareto front. The design point that achieves the maximum value; choosing this option means maximizing visual stability within acceptable cost and quality limits; Lightest option: This option corresponds to the stent mass at the Pareto front. The design point that reaches the minimum value; choosing this scheme means maximizing the weight reduction target while meeting the most basic performance requirements. The optimal solution: This solution corresponds to the one that, on the Pareto front, makes the overall objective function... The design point that achieves the maximum value; it represents the solution that achieves the best balance between performance, quality, and cost based on the weights set by the designer. This set of hierarchical optimization schemes is not just a list of parameters, but is pushed to the edge device where the designer is located in a format containing rich engineering information; the pushed content usually includes: computer-aided design parameters for each scheme, and the corresponding dynamic visual clarity index. Rating, stent quality and material costs The accurate estimation and visualized performance simulation results provide comprehensive data support for designers' final decisions. The hierarchical optimization scheme proposed in this embodiment transforms the output of the optimization algorithm from a single computational result into a set of strategic options with clear engineering orientation. This approach greatly enhances the user-friendliness and practicality of human-computer interaction, parsing complex Pareto front information into concrete solutions with clear engineering inclinations that are easy for designers to understand. This enables designers to make high-level trade-offs and decisions based on comprehensive data, significantly improving the quality of the final design decision and the intelligence level of the entire R&D process. It should be noted that the model constructed in this embodiment mainly focuses on the three core design objectives of performance, quality, and cost. In future work, this framework can be further expanded by introducing additional evaluation modules and objective functions such as fatigue life and manufacturing process constraints to build a more comprehensive multidisciplinary design optimization platform.

[0039] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0040] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A rearview mirror bracket parameter optimization system based on cloud edge computing, characterized in that, include: The multi-source excitation scenario modeling unit is used to retrieve multi-source vibration excitation data from the cloud database and perform mathematical fusion processing on the multi-source vibration excitation data to construct the total excitation spectrum; The dynamic response proxy model unit is used to call a preset AI proxy model to perform real-time predictive analysis based on the total excitation spectrum and the current design parameters input by the designer, so as to obtain the vibration response power spectral density. The dynamic sharpness evaluation unit is used to combine the vibration response power spectral density and the preset visual sensitivity function to calculate and analyze the dynamic sharpness index, obtain the dynamic visual sharpness index, and perform discrimination processing on the dynamic visual sharpness index to obtain unqualified signal, qualified signal or excellent signal. The collaborative optimization control unit is used to respond to qualified or excellent signals, construct a comprehensive objective function based on dynamic visual clarity index, bracket quality and material cost, and perform cloud-based parameter optimization to obtain a hierarchical optimization scheme.

2. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, Multi-source vibration excitation data includes road surface excitation data, powertrain excitation data, and aerodynamic excitation data; the mathematical fusion processing is as follows: Obtain the weighting coefficients of road surface excitation data, powertrain excitation data, and aerodynamic excitation data under specific virtual test conditions; The aforementioned incentive data and their corresponding weighting coefficients are then weighted and superimposed to construct the total incentive spectrum.

3. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, AI agent models are constructed in the following ways: Using the total excitation spectrum and design parameters as load inputs, transient dynamic finite element analysis is performed to obtain a high-precision simulation dataset; A supervised learning paradigm is used to train a high-precision simulation dataset to obtain an AI agent model that can accurately fit the nonlinear relationship between input and output.

4. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, The dynamic resolution index calculation and analysis process is as follows: The total ambiguity is determined by integrating the vibration response power spectral density with a preset visual sensitivity function. The reference ambiguity is determined by integrating the preset reference vibration reference spectrum and the preset visual sensitivity function. The total ambiguity is then compared with the baseline ambiguity to generate a dynamic visual sharpness index.

5. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 4, characterized in that, The reference vibration spectrum is a preset vibration standard that represents an unacceptable vibration, based on the vehicle manufacturer's design specifications and user experience standards.

6. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, The process for determining the dynamic visual acuity index is as follows: When the dynamic visual clarity index is less than or equal to the preset performance baseline threshold, an unqualified signal is generated; A qualified signal is generated when the dynamic visual clarity index is greater than the minimum performance threshold and less than or equal to the preset excellent performance threshold. When the dynamic visual clarity index is greater than the excellent performance threshold, an excellent signal is generated.

7. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, The process of constructing the comprehensive objective function is as follows: Obtain the performance weighting coefficient, quality weighting coefficient, and cost weighting coefficient set by the designer; Determine target reference values ​​for the normalized dynamic visual acuity index, support mass, and material cost, respectively. The dynamic visual acuity index, bracket quality, and material cost are normalized with their respective target reference values ​​to obtain normalized performance, normalized quality, and normalized cost items. The normalized performance term, normalized quality term, and normalized cost term are linearly combined using corresponding weighting coefficients to construct a comprehensive objective function.

8. The rearview mirror bracket parameter optimization system based on cloud edge as described in claim 1, characterized in that, The tiered optimization schemes include the highest resolution scheme, the lightest scheme, and the best overall scheme; the tiered optimization schemes are pushed to the edge device where the designer is located for decision-making.

Images