A robust design method for electromagnetic relay based on preference set

Abstract

The application relates to the field of optimal design, and discloses a robustness optimization design method for an electromagnetic relay based on a preference set, which is used for achieving a better balance between performance and cost of the electromagnetic relay. The method comprises the following steps: generating a performance index loss value through a performance calculation model according to an input design parameter combination and fluctuation parameters; generating a parameter perturbation sample set through a fluctuation parameter statistical distribution sampling; obtaining a preference weight set set by a designer; and then constructing a weighted robustness evaluation index. The application takes the weighted robustness evaluation index as a target function, iteratively optimizes the design parameter combination by using a gradient-based optimization algorithm, outputs an optimization result when a convergence condition is met, and improves the robustness of the electromagnetic relay.

Description

Technical Field

[0001] This invention relates to the field of optimization design, and more particularly to a robust optimization design method for electromagnetic relays based on preference sets. Background Technology

[0002] In electronic equipment and electrical control systems, electromagnetic relays play a crucial role as a key switching element. They connect and disconnect circuits through electromagnetic force and are widely used in aerospace, automotive electronics, industrial automation, communication equipment, and many other fields. The stability and reliability of their performance directly affect the operational efficiency and safety of the entire system. With continuous technological advancements and increasingly complex application scenarios, the performance requirements for electromagnetic relays are becoming increasingly stringent. They must not only meet basic electrical functional requirements but also maintain robust performance under the influence of various uncertainties.

[0003] Traditional electromagnetic relay design methods often focus on optimizing performance under ideal conditions, while neglecting the various fluctuations and uncertainties present in the actual operating environment. Although this design approach can improve the nominal performance of the relay to some extent, its performance may significantly degrade or even fail when faced with complex changes in actual operating conditions, thus affecting the stable operation of the entire system. Existing technologies often focus only on one or a few key performance indicators of electromagnetic relays, while ignoring the mutual influence and trade-offs between different performance indicators, as well as the differences in designers' preferences for different performance indicators, making it difficult to achieve optimal overall performance. When faced with uncertainties, traditional design methods lack effective risk assessment mechanisms and cannot accurately quantify the potential losses and risks of performance indicators under fluctuating conditions, making it difficult to formulate targeted robust optimization strategies.

[0004] Therefore, we propose a robust optimization design method for electromagnetic relays based on preference sets to address the above problems. Summary of the Invention

[0005] This invention provides a robust optimization design method for electromagnetic relays based on preference sets, which enables electromagnetic relays to achieve a better balance between performance and cost.

[0006] The first aspect of this invention provides a robust optimization design method for electromagnetic relays based on preference sets. This method includes: generating performance index loss values ​​using a performance calculation model based on an input combination of electromagnetic relay design parameters and fluctuation parameters related to manufacturing tolerances, material properties, and operating conditions; generating a parameter perturbation sample set by sampling according to the statistical distribution of the fluctuation parameters; obtaining a set of preference weights set corresponding to each performance index, set by the designer; constructing a weighted robustness evaluation index using the performance index loss values, the parameter perturbation sample set, and the preference weight set; and iteratively optimizing the electromagnetic relay design parameter combination using the weighted robustness evaluation index as the objective function, outputting the optimized design parameter combination when a convergence condition is met.

[0007] Optionally, in a first implementation of the first aspect of the present invention, the distribution characteristics of each fluctuation parameter are determined to form a parameter distribution characteristic description; based on the parameter distribution characteristic description, original parameter samples are generated to form an original parameter sample set; according to the physical coupling relationship between each fluctuation parameter in the electromagnetic relay, the original parameter sample set is subjected to correlation processing and correction to generate a parameter disturbance sample set.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the following steps are taken: receiving input from the designer regarding pairwise importance comparisons of the plurality of performance indicators, generating an importance comparison matrix; calculating an unnormalized preliminary weight vector based on the importance comparison matrix; performing a consistency check on the importance comparison matrix; and when the check passes, normalizing the preliminary weight vector to generate a preference weight set.

[0009] Optionally, in a third implementation of the first aspect of the present invention, based on the parameter perturbation sample set and a predefined confidence level parameter, the performance index loss value is calculated to generate a conditional risk value set for each performance index; based on the conditional risk value set and the performance index loss value, a risk contribution matrix describing the distribution of the loss of each performance index above its risk value is constructed; the contribution values ​​in the risk contribution matrix are integrated with the corresponding weights in the preference weight set to obtain a weighted conditional risk value; and based on the weighted conditional risk value, a weighted robustness evaluation index is generated.

[0010] Optionally, in a fourth implementation of the first aspect of the present invention, generating a set of conditional risk values ​​for each performance index includes: analyzing the distribution of the performance index loss values ​​on the parameter perturbation sample set based on a preset tail risk probability threshold to generate an initial risk threshold; calculating the portion of each performance index loss value that exceeds its corresponding initial risk threshold according to the initial risk threshold to generate an over-threshold loss sample; and generating a set of conditional risk values ​​by aggregating and calculating the over-threshold loss samples.

[0011] Optionally, in the fifth implementation of the first aspect of the present invention, a set of initial values ​​for design parameters to be optimized are set to form an initial design parameter combination; based on the initial design parameter combination, the processes of generating performance index loss values, generating parameter perturbation sample sets, obtaining preference weight sets, and constructing weighted robustness evaluation indices are invoked to calculate the current objective function value; a gradient-based optimization algorithm is used to generate a new design parameter combination based on the current objective function value and its gradient information with respect to the design parameters; the new design parameter combination is used as input, and the calculation process of the weighted robustness evaluation index is repeatedly executed to obtain an updated objective function value; it is determined whether the difference between the updated objective function value and the current objective function value satisfies a preset convergence threshold; if it does not satisfy the threshold, the new design parameter combination is used as the current electromagnetic relay design parameter combination, and the parameter update and evaluation process is iteratively executed; if the threshold is satisfied, the new design parameter combination is output as the optimized design parameter combination.

[0012] Optionally, in the sixth implementation of the first aspect of the present invention, a gradient-based optimization algorithm is used to dynamically adjust the step size according to the risk volatility at the current design point: ; Among them, P k This is the current parameter vector; The basic learning rate matrix; The gradient vector; As an acceleration factor; is the coefficient of variation of the conditional risk value.

[0013] Optionally, in the seventh implementation of the first aspect of the present invention, the method further includes: generating an independent verification sample set based on a sampling rule different from the parameter perturbation sample set; inputting the optimized design parameter combination and the verification sample set into the performance calculation model to calculate the verification performance index loss value; calculating the verification robustness evaluation value of the optimized design parameter combination based on the verification performance index loss value and the preference weight set; and confirming the effectiveness of the optimized design parameter combination when the verification robustness evaluation value is better than a preset verification threshold.

[0014] Optionally, in an eighth implementation of the first aspect of the present invention, when the verification robustness evaluation value is not better than the verification threshold, the method further includes: identifying one or more dominant performance indicators that lead to insufficient robustness based on the distribution of the verification performance indicator loss values; Based on the identification results of the dominant performance indicators, the allowable variation range of the design parameter combination of the electromagnetic relay is adjusted to generate an updated design constraint boundary; based on the updated design constraint boundary, an initial design parameter combination is regenerated, and the construction and optimization process of the weighted robustness evaluation index is re-executed.

[0015] The mechanism of this invention is as follows: the set of preference weights determined by the designer through the pairwise comparison method is deeply integrated with the risk contribution generated based on CVaR to construct a weighted robustness evaluation index that reflects personalized risk tolerance; Beneficial effects: By introducing fluctuation parameters of manufacturing tolerances, material properties and operating conditions, and generating a parameter disturbance sample set based on the statistical distribution of these parameters, the uncertainties in the actual operating environment are fully considered, providing a rich data foundation for robust optimization design. By obtaining the set of preference weights set by the designer, and combining the performance index loss value and the parameter perturbation sample set, a weighted robustness evaluation index is constructed, realizing a comprehensive evaluation of multiple performance indicators of electromagnetic relays, and fully considering the mutual influence between different performance indicators and the differences in the designer's preferences. Based on the parameter perturbation sample set and predefined confidence level parameters, the conditional risk value set of each performance indicator is calculated, and a risk contribution matrix is ​​constructed. This enables a quantitative assessment of the potential loss and risk of performance indicators under volatile conditions, providing a basis for formulating targeted robust optimization strategies. A gradient-based optimization algorithm, combined with a weighted robustness evaluation index as the objective function, is used to iteratively optimize the combination of electromagnetic relay design parameters. This effectively improves the computational efficiency and convergence speed of the optimization design, avoids getting trapped in local optima, and achieves a robust optimization design with high precision and high efficiency. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of an embodiment of the robust optimization design method for electromagnetic relays based on preference sets in this invention. Figure 2 This is a schematic diagram of another embodiment of the electromagnetic relay robustness optimization design method based on preference sets in this invention. Figure 3 This is a schematic diagram of the design parameters and fluctuation parameters of the electromagnetic relay in an embodiment of the present invention. Detailed Implementation

[0017] This invention provides a robust optimization design method for electromagnetic relays based on preference sets, enabling a better balance between performance and cost. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the robust optimization design method for electromagnetic relays based on preference sets in this invention includes: 101. Based on the combination of input electromagnetic relay design parameters, as well as the fluctuation parameters of manufacturing tolerances, material properties and operating conditions, a set of performance index loss values ​​are generated through a performance calculation model. It is understood that the executing entity of this invention can be a robust optimization design device for electromagnetic relays based on preference sets, or it can be a terminal or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example of the executing entity.

[0019] It should be noted that the design goal of a certain type of electromagnetic relay is to ensure that its pull-in voltage remains stable near the rated value, while minimizing contact bounce. In this step, the server needs to quantify the performance of the initial design parameter combination under the influence of fluctuation factors.

[0020] Load a set of initial electromagnetic relay design parameters: Key design parameters: armature thickness (1.5mm), coil turns (500 turns), spring stiffness (0.8N / mm). Fluctuation parameters (considering manufacturing tolerances, material variations, and operating condition fluctuations): Manufacturing tolerances: armature thickness deviation ±0.05mm, coil turns deviation ±10 turns. Material property fluctuations: spring stiffness coefficient variation ±5%. Operating condition fluctuations: coil drive voltage fluctuates within ±5% of rated value, ambient temperature varies between -10℃ and 50℃.

[0021] The following models are integrated to simulate the dynamic behavior of the relay: Finite element analysis model: used to calculate electromagnetic attraction characteristics, inputting armature dimensions and coil parameters; the model outputs electromagnetic attraction curves under different air gaps. Multibody dynamics model: simulates the relay's operation process (armature movement, contact collision), used to calculate closing time and contact bounce amplitude. This model incorporates parameters such as spring stiffness to analyze their impact on motion characteristics. Experimentally calibrated surrogate model: to reduce computational costs, the server uses historical experimental data to train a simplified model (a fast attraction calculation model based on nonlinear regression) for quickly predicting key performance indicators.

[0022] The following process is used to generate performance index loss values: Baseline performance calculation: Input the nominal design parameters (without fluctuations) into the performance calculation model to obtain the baseline performance values. The nominal pull-in voltage is calculated to be 12.0V, and the nominal contact bounce amplitude is 0.15mm. Disturbance sample generation and simulation: Based on the statistical distribution of the fluctuation parameters (assuming the armature thickness follows a normal distribution with a tolerance of ±0.05mm), 500 sets of disturbance parameter samples are generated using Monte Carlo sampling. Each set of samples represents a possible manufacturing or operating condition fluctuation scenario. Substitute each set of disturbance parameters into the performance calculation model to calculate the corresponding pull-in voltage and contact bounce amplitude. A simulation result may show that when the armature thickness is 1.52mm, the coil turns are 495, and the ambient temperature is 45℃, the pull-in voltage is 12.3V, and the bounce amplitude is 0.18mm.

[0023] Calculate the performance index loss values: Pull-in voltage loss value: Calculate the root mean square error (RMSE) between the pull-in voltage of 500 simulations and the nominal value (12.0V). If the standard deviation of the pull-in voltage in the simulation results is 0.2V, the loss value can be quantified as 0.2 (the smaller the value, the more robust the pull-in voltage). Contact bounce loss value: Similarly, calculate the RMSE between the bounce amplitude and the nominal value (0.15mm). If the fluctuation range of the bounce amplitude is 0.02mm, the loss value can be quantified as 0.02.

[0024] Output a set of performance index loss values: [Pull-in voltage loss: 0.2, Contact bounce loss: 0.02]. This set of values ​​reflects the possible deviation of the core performance indicators under the expected fluctuations of the current design parameter combination.

[0025] 102. Based on the statistical distribution of the fluctuation parameters, generate a sample set of parameter perturbations through sampling; It should be noted that the key fluctuation parameters affecting the performance of the electromagnetic relay, identified in step 101, are generally divided into three categories, and reasonable statistical distribution models are set for them to describe their fluctuation characteristics: Manufacturing tolerance: Armature thickness, with a nominal design value of 1.5 mm, considering processing variations, its fluctuation is set to follow a normal distribution with a standard deviation of 0.02 mm (i.e., a tolerance zone of ±0.06 mm corresponds to approximately 3 standard deviations). Material property fluctuation: Spring stiffness coefficient, with a nominal value of 0.8 N / mm, affected by material batches, its fluctuation follows a normal distribution with a coefficient of variation (the ratio of standard deviation to mean) of 2.5%. Operating condition fluctuation: Coil drive voltage, rated at 12 volts, but may fluctuate within ±0.5 volts in actual use, is set to follow a uniform distribution; Ambient temperature, expected to vary between -10℃ and 50℃, is also set to a uniform distribution.

[0026] To efficiently cover the possible combinations of these fluctuating parameters, a Monte Carlo sampling method is employed. This method is particularly suitable for generating a large number of random samples based on probability distributions to simulate the actual fluctuations of parameters. Sample size determination: To obtain statistically reliable results, the server is set to generate 1000 sets of parameter perturbation samples. This number is sufficient to reflect the statistical characteristics of parameter fluctuations without causing excessive computation. Sampling execution: For each fluctuating parameter, independent random sampling is performed according to its defined statistical distribution. For parameters following a normal distribution (armature thickness, spring stiffness), the sampling process generates normally distributed random numbers that conform to their mean (nominal value) and standard deviation; a sample value for armature thickness might be 1.52 mm. For parameters following a uniform distribution (drive voltage, ambient temperature), sampling generates random numbers with equal probability within a given range (11.5 V to 12.5 V); a sample value for drive voltage might be 12.2 V. Combined Samples: The server combines each sampling result for each parameter to form a complete set of parameter perturbation samples. These 1000 independent sampling results are organized into a parameter perturbation sample set. Each row of this set represents a specific combination of fluctuating parameters, describing a specific fluctuation scenario that the electromagnetic relay may encounter during manufacturing or operation.

[0027] The generated parameter perturbation sample set may include the following examples of the first few data sets: Table 1 below: Table 1 This set of parameter disturbance samples lays the foundation for subsequent steps (104 Constructing a weighted robustness evaluation index). Each set of disturbance samples will be substituted into the performance calculation model to evaluate the performance of the electromagnetic relay under that specific fluctuation condition, thereby comprehensively quantifying the robustness of the design scheme.

[0028] 103. Obtain the set of preference weights corresponding to each performance index set by the designer; It should be noted that the performance indicators to be evaluated are obtained from step 101. In this embodiment, three core indicators are focused on: Pull-in voltage stability: This is crucial for reflecting the reliability of the relay operation; the smaller the loss value, the more stable the pull-in voltage under different fluctuations. Contact bounce amplitude: This directly affects contact life and contact reliability; the smaller the loss value, the better the contact bounce control. Closing time: This relates to the operating speed of the relay; the smaller the loss value, the faster the closing process.

[0029] Designers assign preference weights to the above performance indicators based on the specific application scenarios and design goals of the product. Weights are typically between 0 and 1, and the sum of all indicator weights is 1. The following shows two weight allocation schemes under different design goals: For relays used in industrial automation control systems, reliability and long lifespan are the primary goals in weighting for high-reliability industrial control. Therefore, designers prioritize ensuring the stability of the pull-in voltage and minimizing contact bounce. Pull-in voltage stability weight: 0.5; Contact bounce amplitude weight: 0.4; Closing time weight: 0.1; Design logic: In this scenario, preventing malfunctions and reducing contact wear are far more important than millisecond-level speed improvements.

[0030] For power protection devices designed for rapid response, the operating speed of relays used in power system protection, which require rapid fault disconnection, is crucial. The weighting is as follows: Closing time weight: 0.6; Pull-in voltage stability weight: 0.3; Contact bounce amplitude weight: 0.1. The design logic is that rapid disconnection to protect downstream equipment is the core task, therefore closing time has the highest weight. Stability and bounce still need to be guaranteed, but the requirements can be appropriately relaxed.

[0031] After setting the weights through the graphical interface (slider, numerical input box) provided by the server, the system will prompt whether the total weight is 1 and confirm. The server obtains and stores this set of preferred weights. The weight set of Example A is {pull-in voltage stability: 0.5, contact bounce amplitude: 0.4, closing time: 0.1}.

[0032] 104. Construct a weighted robustness evaluation index using performance index loss values, parameter perturbation sample sets, and preference weight sets; It should be noted that the following key data have been obtained from the previous steps: Performance index loss values ​​(from step 101): For the current combination of design parameters to be evaluated, the loss values ​​of three key performance indicators under parameter perturbations were obtained through performance calculation models (finite element analysis, nonlinear dynamic model) and Monte Carlo simulation. These loss values ​​are usually expressed as root mean square error (RMSE) or standard deviation, quantifying the uncertainty of performance caused by fluctuations: pull-in voltage stability loss: 0.20V; contact bounce amplitude loss: 0.018mm; closing time loss: 0.15ms; Parameter disturbance sample set (from step 102): This set (1000 samples) has been used to calculate the above loss value, and its statistical characteristics are implicit in the loss value. Preference weight set (from step 103): Based on the product's positioning for high-reliability industrial control, the designer set the following preference weights: pull-in voltage stability weight: 0.5; contact bounce amplitude weight: 0.4; closing time weight: 0.1; The evaluation metrics are constructed according to the following process: Metric normalization: To avoid the influence of different dimensions and orders of magnitude of performance metrics, the loss values ​​are normalized. The server scales the loss values ​​using the maximum and minimum values ​​of all candidate design schemes. In the current iteration, the pull-in voltage stability loss for all design schemes is a maximum of 0.35V and a minimum of 0.10V.

[0033] The maximum loss of contact bounce amplitude is 0.030mm, and the minimum is 0.010mm.

[0034] The maximum closing time loss is 0.25ms, and the minimum is 0.08ms.

[0035] The normalized loss value for the current design (loss values ​​of 0.20V, 0.018mm, and 0.15ms) is then: Normalized pull-in voltage stability loss = (0.20 - 0.10) / (0.35 - 0.10) = 0.4; Normalized loss of contact bounce amplitude = (0.018 - 0.010) / (0.030 - 0.010) = 0.4; Normalized loss for closure time = (0.15 - 0.08) / (0.25 - 0.08) ≈ 0.411; Weighted summation is performed by applying weights: the normalized loss value is multiplied by the corresponding preference weight and summed to obtain the comprehensive weighted robustness evaluation index (WRI) of the design scheme.

[0036] WRI=(0.4×0.5)+(0.4×0.4)+(0.411×0.1)=0.2+0.16+0.0411=0.4011; The calculated weighted robustness index (WRI) value (0.4011 in this example) is the overall performance score of the current design parameter combination under expected fluctuations. The smaller the value, the less sensitive the overall performance of the design scheme is to external fluctuations after considering the designer's preferences, i.e., the better the robustness. This WRI value is used as the objective function for iterative optimization in step 105, guiding the search process towards a more robust design scheme.

[0037] 105. Using the weighted robustness evaluation index as the objective function, iteratively optimize the combination of electromagnetic relay design parameters. When the convergence condition is met, output the optimized design parameter combination.

[0038] It should be noted that the weighted robustness evaluation index calculated in step 104 is used as the objective function for optimization. The smaller the index value, the better the overall robustness of the design under fluctuating conditions. The initial design parameter combination is set as follows: armature thickness 1.5 mm, coil turns 500, and spring stiffness 0.8 N / mm, with a weighted robustness evaluation index of 0.4011. The optimization objective is to find a set of design parameters that minimizes this index.

[0039] An improved differential evolution algorithm is used for iterative optimization, which can effectively handle multi-parameter optimization and avoid getting trapped in local optima. The optimization process must meet the basic engineering constraints of electromagnetic relays, such as minimum pull-in voltage and maximum allowable contact bounce. Initialization and first iteration: The algorithm randomly generates an initial population containing 50 sets of design parameters, each set of parameters being within a preset reasonable range (armature thickness 1.3-1.7mm). The server executes steps 101 to 104 for each set of parameters in the population to calculate its respective weighted robustness evaluation index. After the first iteration, the optimal parameter combination is set as follows: armature thickness 1.55mm, coil turns 490, spring stiffness 0.82N / mm, corresponding to an index of 0.35.

[0040] Population Update and Multiple Iterations: The algorithm performs mutation, crossover, and selection operations on the current population to generate new parameter combinations for testing. In the second iteration, a new set of parameters (armature thickness 1.52 mm, coil turns 510, spring stiffness 0.78 N / mm) may be found to further reduce the index to 0.28. This process is repeated, and the optimal index trend for each generation of the population is as follows: 0.4011 (initial) → 0.35 → 0.28 → 0.25 → 0.23 → 0.221 → 0.2205 → 0.2201.

[0041] Two convergence conditions are set to be met simultaneously: Improvement threshold: The improvement in the optimal weighted robustness evaluation index is less than 0.05% over 10 consecutive iterations. Maximum number of iterations: The total number of iterations does not exceed 200 to prevent infinite loops. During implementation, the algorithm found the following parameter combination at the 125th iteration: armature thickness 1.48mm, coil turns 505, spring stiffness 0.83N / mm, with a weighted robustness evaluation index of 0.2201. Subsequently, until the 135th iteration, the optimal index did not change significantly (improvement less than 0.0005), therefore, the optimization was deemed converged.

[0042] Output this optimized design parameter combination. Verification shows that, considering manufacturing tolerances and operating condition fluctuations, this solution outperforms the initial design in terms of pull-in voltage stability and contact bounce control, thus meeting robustness requirements.

[0043] Please see Figure 2 and Figure 3 Another embodiment of the robust optimization design method for electromagnetic relays based on preference sets in this invention includes: 201. Based on the combination of input electromagnetic relay design parameters, as well as the fluctuation parameters of manufacturing tolerances, material properties and operating conditions, a set of performance index loss values ​​are generated through a performance calculation model. Specifically, design target values ​​are set for each performance index; the combination of electromagnetic relay design parameters and the combination of sample parameters selected from the parameter disturbance sample set are input into the performance calculation model to calculate a set of actual performance index values; the actual performance index values ​​are compared with their corresponding design target values ​​to obtain a set of original performance deviations; the original performance deviations are processed based on a predetermined normalization benchmark and converted into a set of dimensionless performance index loss values.

[0044] It should be noted that, taking a general-purpose electromagnetic relay of model JZC-22F as an example, the equivalent magnetic circuit method and dynamic simulation model are used as the "performance calculation model". We select two key performance indicators: electromagnetic attraction force (the greater the better) and attraction time (the shorter the better).

[0045] Specific implementation steps: Setting design target values: Designers clarify the ideal targets for each indicator. The design target value for electromagnetic attraction force is set at 60 millinewtons (mN). The design target value for attraction time is set at 10 milliseconds (ms).

[0046] Parameter Input and Model Calculation: Input Design Parameter Combination: Select a set of current design center values, with 5000 coil turns, a core diameter of 4.0 mm, and a working air gap of 0.8 mm. Introduce Parameter Disturbance Sample: Select a specific disturbance sample from the sample set generated in step 202. Set the sample display: Due to manufacturing tolerances, the actual core diameter is 3.95 mm (too small); due to environmental fluctuations, the coil temperature increases, leading to a 5% increase in resistance. Perform Calculation: Input the above-mentioned actual parameters with deviations into the simulation model. After model calculation, output the actual performance index value under this condition: The actual electromagnetic attraction force is calculated to be 54 millinewtons (due to the thinner core and increased resistance, the magnetic flux decreases).

[0047] The actual suction time was calculated to be 12 milliseconds (due to reduced suction, resulting in slower movement). Calculation of original performance deviation: Compare the actual value with the target value. Suction deviation: Subtract the actual 54 mN from the target 60 mN, resulting in a deviation of 6 millinewtons (indicating insufficient performance). Time deviation: Subtract the target 10 ms from the actual 12 ms, resulting in a deviation of 2 milliseconds (indicating delay). Normalization to generate loss values: To allow for subsequent weighted calculations of metrics in different units (millinewtons and milliseconds), they need to be converted to dimensionless loss values.

[0048] Set normalization baseline: Set a preset "maximum acceptable deviation range" as the baseline. The maximum allowable deviation for suction power is 15 millinewtons, and the maximum allowable deviation for time is 5 milliseconds.

[0049] Calculate the loss values: Suction loss value = 6 millinewtons (deviation) divided by 15 millinewtons (reference) = 0.4. Time loss value = 2 milliseconds (deviation) divided by 5 milliseconds (reference) = 0.4.

[0050] For this set of specific parameter inputs, step 201 outputs a set of performance index loss values: {Suction loss: 0.4, Time loss: 0.4}.

[0051] 202. Based on the statistical distribution of the fluctuation parameters, generate a sample set of parameter perturbations through sampling; Specifically, based on historical data or engineering experience, the distribution characteristics of each fluctuation parameter are determined to form a parameter distribution characteristic description; based on the parameter distribution characteristic description, a set of original parameter samples is generated using the Monte Carlo sampling method to form the original parameter sample set; according to the physical coupling relationship between each fluctuation parameter in the electromagnetic relay, the original parameter sample set is subjected to correlation processing and correction to generate a parameter disturbance sample set.

[0052] It should be noted that, continuing with the aforementioned JZC-22F electromagnetic relay scenario, this step aims to generate a set of "virtual test samples" that closely resemble the real physical environment, for subsequent simulation of the relay's performance under various extreme or normal conditions.

[0053] Determine parameter distribution characteristics (defining fluctuation sources): Based on historical production data (QC data) and product specifications, engineers identified three main fluctuation parameters and their statistical characteristics: Parameter A (Manufacturing Tolerance): Initial working air gap between the core and armature. The design nominal value is 0.5 mm. According to factory CPK data analysis, this parameter follows a normal distribution with a mean of 0.5 mm and a standard deviation of 0.01 mm. Parameter B (Material Properties): Elastic modulus of the contact spring. The nominal value is 110 GPa. Due to supplier batch variations, this parameter follows a truncated normal distribution, with fluctuations limited to ±5% of the nominal value. Parameter C (Operating Conditions): Ambient temperature. Based on automotive application scenarios, this parameter does not follow a normal distribution but rather a uniform distribution, with a value range covering -40 degrees Celsius to 85 degrees Celsius, representing various possible environments for vehicles from extremely cold to extremely hot temperatures.

[0054] Monte Carlo sampling generates the original sample (generating random numbers): the number of sampling iterations is set to 2000. Data is independently extracted using a computer random number generator based on the above distribution characteristics.

[0055] Taking sample number 101 as an example, the result of random computer sampling might be: Initial working air gap: 0.512 mm (slightly larger than the nominal value). Elastic modulus: 109.5 GPa (slightly lower than the nominal value). Ambient temperature: 75 degrees Celsius (high temperature environment). At this point, these three data points are statistically independent, and physical influences have not yet been considered, constituting the "original parameter sample set".

[0056] Correlation Handling and Correction (Introducing Physical Coupling): In real physics, some parameters are not entirely independent. In electromagnetic relays, the coil resistance (another key parameter, nominally 100 ohms) has a strong physical coupling relationship with ambient temperature (the resistivity of copper increases with temperature). Identifying the Coupling: In the original sample set, although we did not directly extract the manufacturing tolerances of the coil resistance, changes in ambient temperature directly cause a significant drift in the coil resistance. Performing Correction: For the 75°C high temperature in sample number 101, the system corrects the coil resistance according to the temperature coefficient of resistance rule for copper wire. Although the nominal resistance is 100 ohms at room temperature (20°C), at 75°C (a temperature rise of 55°C), the resistance value needs to increase by approximately 22%.

[0057] Sample generation: Correct sample number 101 to a set of perturbation samples containing coupling effects: [Air gap: 0.512 mm, elastic modulus: 109.5 GPa, ambient temperature: 75 degrees Celsius, actual coil resistance: 122 ohms].

[0058] After 2000 such processing steps, a "parameter perturbation sample set" containing 2000 rows of data is output. Each row of data represents a "virtual relay entity" that is physically reasonable and includes manufacturing errors and harsh environmental conditions.

[0059] 203. Obtain the set of preference weights corresponding to each performance index, as set by the designer; Specifically, the system receives input from the designer comparing the importance of multiple performance metrics pairwise and generates an importance comparison matrix. Based on the importance comparison matrix, a set of unnormalized preliminary weight vectors is calculated. The importance comparison matrix is ​​then subjected to a consistency check. If the check passes, the preliminary weight vectors are normalized to generate a set of preference weights.

[0060] It should be noted that the JZC-22F electromagnetic relay will continue to be used as the subject. To balance the conflicts between different performance indicators, the design team needs to clarify the relative importance of each indicator. Three key performance indicators were selected: Indicator A (electromagnetic attraction force), Indicator B (pull-in time), and Indicator C (contact bounce time). Among these, the contact bounce time directly affects contact life and signal transmission stability, and is considered the most critical; electromagnetic attraction force ensures operational reliability, which is secondary; and fine-tuning the pull-in time at the millisecond level has the least impact on this application.

[0061] Constructing an importance comparison matrix (pairwise comparisons): The designers used a 1-9 scale to compare and score the three indicators pairwise. The scores were input into the system as follows: Suction vs. Time: Suction was considered "slightly more important" than time, assigned a value of 3. Rebound vs. Suction: Rebound was considered "slightly more important" than suction, assigned a value of 2. Rebound vs. Time: Rebound was considered "significantly more important" than time, assigned a value of 5. The corresponding reverse comparison (Time vs. Suction) was automatically calculated using the reciprocal (1 / 3).

[0062] The resulting importance comparison matrix data is as follows (row order: suction, time, bounce): First row (suction): [1,3,0.5]; Second row (time): [0.33,1,0.2]; Third row (bounce): [2,5,1]; Calculating the initial weight vector: The system calculates the eigenvectors of the above matrix. The matrix data is processed using the geometric mean or square root method to calculate a set of unnormalized eigenvector values. Intermediate calculation results show that the original weight scores for the three indicators are approximately: suction score 1.82, time score 0.65, and bounce score 3.38.

[0063] To prevent logical contradictions in the designer's judgments (e.g., A is more important than B, B is more important than C, yet C is judged to be more important than A), the system automatically calculates the consistency ratio (CR). In this example, after calculating the largest eigenvalue and the consistency index, the CR value is 0.009. Since 0.009 is much less than the standard threshold of 0.1, the system determines that the importance matrix passes the consistency test, indicating that the designer's preference logic is clear and self-consistent.

[0064] Generate a set of preference weights (normalization): Normalize the initial weight scores to make the sum of all weights equal to 1.

[0065] Suction weight = 1.82 / (1.82 + 0.65 + 3.38) ≈ 0.311; Time weight = 0.65 / (1.82+0.65+3.38)≈0.111; The bounce weight = 3.38 / (1.82 + 0.65 + 3.38) ≈ 0.578; Step 203 outputs a defined set of preference weights: {Electromagnetic attraction force: 0.311, attraction time: 0.111, contact bounce time: 0.578}.

[0066] 204. Construct a weighted robustness evaluation index using performance index loss values, parameter perturbation sample sets, and preference weight sets; Specifically, based on the parameter perturbation sample set and predefined confidence level parameters, the performance index loss value is calculated to generate a set of conditional risk values ​​for each performance index; based on the set of conditional risk values ​​and the performance index loss values, a risk contribution matrix describing the distribution of each performance index loss above its risk value is constructed; the contribution values ​​in the risk contribution matrix are integrated with the corresponding weights in the preference weight set to obtain a set of weighted conditional risk values; based on the weighted conditional risk values, a weighted robustness evaluation index is generated.

[0067] Furthermore, a set of conditional risk values ​​for each performance index is generated, including: analyzing the distribution of performance index loss values ​​on the parameter perturbation sample set based on a preset tail risk probability threshold, and generating a set of initial risk thresholds; calculating the portion of each performance index loss value that exceeds its corresponding initial risk threshold based on the initial risk thresholds, and generating a set of over-threshold loss samples; and generating a set of validated conditional risk values ​​that reflect extremely adverse situations by aggregating and calculating the over-threshold loss samples.

[0068] It should be noted that the system sets the risk probability threshold and generates the initial risk threshold: the "tail risk probability threshold" is set to 5% (i.e., a confidence level of 95%), aiming to focus on the worst 5% of extreme cases. The system performs statistical analysis on the loss value distribution of 2000 samples. Taking indicator C (touch bounce time) as an example, the 2000 loss values ​​are sorted from smallest to largest, and the value located at the 95th percentile is found. The value of this percentile is set to 0.82, which means that in 95% of cases, the bounce time loss value is less than 0.82, while those exceeding 0.82 are considered "extreme risks". Similarly, the threshold for indicator A (suction force) is 0.75, and the threshold for indicator B (time) is 0.50.

[0069] Generate over-threshold loss samples and conditional risk values ​​(CVaR): The system filters out all "over-threshold loss samples" (i.e., the 100 worst samples) that exceed the above threshold.

[0070] Indicator C (Rebound): The average loss value of these 100 extreme samples is calculated to be 1.10. This means that under the worst-case extreme conditions, the average loss of contact rebound performance is as high as 1.10.

[0071] Indicator A (Suction): The mean of the last 100 samples is calculated, resulting in a conditional risk value of 0.80.

[0072] Indicator B (Time): The mean of the last 100 samples is calculated, yielding a conditional risk value of 0.60. This generates a set of conditional risk values ​​reflecting extremely unfavorable scenarios: {0.80, 0.60, 1.10}.

[0073] Construct and weight the risk contribution matrix: Integrate the above risk values ​​with the preference weights determined in step 203 to calculate: Attraction risk contribution: The conditional risk value of 0.80 is multiplied by the weight of 0.311, and the result is approximately 0.249.

[0074] Time risk contribution: The conditional risk value of 0.60 multiplied by the weight of 0.111 results in approximately 0.067.

[0075] Contribution of bounce risk: The conditional risk value of 1.10 multiplied by the weight of 0.578 yields approximately 0.636. It can be seen that, due to the highest weight and greatest extreme risk associated with the bounce time, it dominates the contribution to the total risk.

[0076] Generate a weighted robustness evaluation index: Add the above three weighted contribution values: 0.249 plus 0.067 plus 0.636, to get a total of 0.952.

[0077] The output value of 0.952 is used as the current "weighted robustness evaluation index". This value represents the level of potential system risk under the current design parameters, taking into account manufacturing tolerances, environmental fluctuations, and designer preferences.

[0078] 205. Using the weighted robustness evaluation index as the objective function, iteratively optimize the combination of electromagnetic relay design parameters. When the convergence condition is met, output the optimized design parameter combination.

[0079] Specifically, a set of initial values ​​for the design parameters to be optimized is set to form an initial design parameter combination. Based on the initial design parameter combination, the processes of generating performance index loss values, generating parameter perturbation sample sets, obtaining preference weight sets, and constructing weighted robustness evaluation indices are invoked to calculate the current objective function value. A gradient-based optimization algorithm is used to generate a new set of design parameter combinations based on the current objective function value and its gradient information with respect to the design parameters. The new design parameter combination is used as input, and the calculation process of the weighted robustness evaluation indices is repeated to obtain an updated objective function value. It is determined whether the difference between the updated objective function value and the current objective function value meets a preset convergence threshold. If it does not meet the threshold, the new design parameter combination is used as the current electromagnetic relay design parameter combination, and the parameter update and evaluation process is iteratively executed. If the threshold is met, the new design parameter combination is output as the optimized design parameter combination.

[0080] It should be noted that the initial design parameter combination for the JZC-22F relay is P0 = {coil turns: 5000 turns, core diameter: 4.0mm}, and the calculated weighted robustness evaluation index value (i.e., objective function value) is J0 = 0.952. Our goal is to find the parameter combination that minimizes J through iteration, and we set the convergence threshold Ɛ = 0.001.

[0081] Specific implementation steps: Gradient information acquisition (sensitivity analysis): The system calculates the gradient by performing a small perturbation near the initial point. Increasing the number of coil turns by 10 turns (5010 turns) yields J'=0.948, and the change is... Gradient with respect to the number of turns .

[0082] Increasing the core diameter by 0.05 mm (to 4.05 mm) yields J'' = 0.930, the change is... Gradient with respect to diameter This indicates that increasing the number of turns and the diameter can reduce the risk, with the diameter being more sensitive.

[0083] To accelerate convergence and avoid getting trapped in local optima, this embodiment employs a risk-adaptive gradient update formula, dynamically adjusting the step size based on the risk volatility at the current design point: Among them, P k This is the current parameter vector; Set the base learning rate matrix (500 for the number of turns and 0.2 for the diameter to balance the dimensions). The gradient vector; The acceleration factor is set to 0.5. This is the coefficient of variation of the conditional risk value calculated in the previous steps (set to 0.4 here, indicating that the risk distribution fluctuates greatly and the search step size needs to be increased).

[0084] Turns update: (Rounded to 5000 turns. Because the change is too small, the system may automatically and dynamically adjust the learning rate later. This is only for demonstration purposes. The actual algorithm will accumulate momentum.)

[0085] Diameter update: Millimeters.

[0086] Iteration and convergence judgment: combining new parameters Substitute the values ​​of {5000 turns, 4.11 mm} into the model and repeat steps 201-204. The new objective function value J1 = 0.885 is calculated. Convergence is then determined by calculating the difference. Since 0.067 > 0.001 (the preset threshold), the convergence condition was not met.

[0087] The system uses P1 as the new current parameter and repeats the gradient calculation and update process described above.

[0088] After 12 iterations, the parameter update amount becomes small, and the objective function value stabilizes at J. final =0.420, and the difference between two consecutive iterations is less than 0.001. The system determines convergence and outputs the optimized design parameter combination: {number of coil turns: 5200 turns, core diameter: 4.25mm}. At this point, the robustness of the relay is significantly improved, and the expected failure risk is greatly reduced.

[0089] 206. Based on a sampling rule different from that of the parameter perturbation sample set, generate an independent verification sample set; input the optimized design parameter combination and the verification sample set into the performance calculation model to calculate a set of verification performance index loss values; based on the verification performance index loss values ​​and the set of preference weights, calculate the verification robustness evaluation value of the optimized design parameter combination; when the verification robustness evaluation value is better than the preset verification threshold, confirm that the optimized design parameter combination is effective.

[0090] Specifically, when the robustness evaluation value is not better than the verification threshold, the method further includes: identifying one or more dominant performance indicators that lead to insufficient robustness based on the distribution of the loss values ​​of the verification performance indicators; adjusting the allowable variation range of the design parameter combination of the electromagnetic relay according to the identification results of the dominant performance indicators, and generating a set of updated design constraint boundaries; and regenerating an initial design parameter combination based on the updated design constraint boundaries, and re-executing the construction and optimization process of the weighted robustness evaluation index.

[0091] It should be noted that after optimization in step 205, the following design parameter combination was obtained for verification: {number of coil turns: 5200 turns, core diameter: 4.25mm}. At this point, a rigorous "stress test" is required to confirm the effectiveness of the design.

[0092] Generating an independent set of validation samples: To avoid "data leakage" or overfitting, this step no longer uses the simple random sampling in step 202, but instead employs the Latin hypercube sampling (LHS) rule. This rule forces the samples to cover the marginal distribution of the fluctuation parameters (a rare combination of extremely high temperatures and extremely low manufacturing precision). The system generates 500 independent validation samples, which focus more on boundary conditions.

[0093] Calculating the robustness evaluation value: The optimized design parameters and these 500 validation samples are input into the model. The validation performance index loss value is calculated. Combined with the preference weights, the validation robustness evaluation value is calculated to be 0.48. Judgment: The preset validation threshold is 0.45 (the smaller the value, the more robust). Since 0.48 > 0.45, the validation is judged to have failed. Identifying the dominant performance index and adjusting constraints: The system analyzes the distribution of validation results and finds that the main reason for the evaluation value exceeding the standard is "contact bounce time". Data shows that at a high number of turns of 5200, although the suction force is extremely stable, the excessive suction force causes violent armature impact, and the risk of bounce increases sharply.

[0094] Adjustment Decision: Since excessive suction caused the bounce problem, the design space for generating excessive suction needs to be limited. Update Boundary: The upper limit of the design constraint for "coil turns" is lowered from 6000 turns to 5100 turns, forcing the optimization algorithm to find a balance in a lower excitation range.

[0095] Restart Optimization: Based on the updated constraint boundary (number of turns ≤ 5100), the system automatically jumps back to step 205 to regenerate the initial point and start a new round of iteration. Data verification comparison table 2: Table 2 This step successfully intercepted a design scheme that performed well under normal distribution but might fail due to "excessive backflip" under extreme boundary conditions, thus ensuring product reliability.

[0096] The present invention also provides a robust optimization design device for electromagnetic relays based on preference sets. The device includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the robust optimization design method for electromagnetic relays based on preference sets in the above embodiments.

[0097] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the preference set-based electromagnetic relay robust optimization design method.

[0098] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0099] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0100] The above-described 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 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. Such 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.

Claims

1. A robust optimization design method for electromagnetic relays based on preference sets, characterized in that, include: Based on the combination of input electromagnetic relay design parameters, as well as the fluctuation parameters of manufacturing tolerances, material properties and operating conditions, a performance index loss value is generated through a performance calculation model. Based on the statistical distribution of the fluctuation parameters, a parameter perturbation sample set is generated by sampling. Obtain the set of preference weights corresponding to each performance metric, as set by the designer; A weighted robustness evaluation index is constructed using the performance index loss value, the parameter perturbation sample set, and the preference weight set. Using the weighted robustness evaluation index as the objective function, the combination of electromagnetic relay design parameters is iteratively optimized. When the convergence condition is met, the optimized design parameter combination is output.

2. The robust optimization design method for electromagnetic relays based on preference sets according to claim 1, characterized in that, Determine the distribution characteristics of each fluctuation parameter to form a parameter distribution characteristic description; Based on the parameter distribution characteristics described above, original parameter samples are generated to form an original parameter sample set. Based on the physical coupling relationship between the fluctuation parameters in the electromagnetic relay, the original parameter sample set is subjected to correlation processing and correction to generate a parameter disturbance sample set.

3. The robust optimization design method for electromagnetic relays based on preference sets according to claim 2, characterized in that, Receive input from the designer regarding pairwise importance comparisons of the multiple performance metrics, and generate an importance comparison matrix; Based on the importance comparison matrix, the unnormalized preliminary weight vector is calculated; A consistency check is performed on the importance comparison matrix. If the check passes, the initial weight vector is normalized to generate a set of preference weights.

4. The robust optimization design method for electromagnetic relays based on preference sets according to claim 3, characterized in that, Based on the parameter perturbation sample set and the predefined confidence level parameter, the loss value of the performance index is calculated to generate a set of conditional risk values ​​for each performance index. Based on the set of conditional risk values ​​and the performance index loss values, a risk contribution matrix is ​​constructed to describe the distribution of each performance index loss above its risk value. The contribution values ​​in the risk contribution matrix are integrated with the corresponding weights in the preference weight set to obtain the weighted conditional risk value; Based on the weighted conditional risk value, a weighted robustness evaluation index is generated.

5. The robust optimization design method for electromagnetic relays based on preference sets according to claim 4, characterized in that, The set of conditional risk values ​​for generating each performance indicator includes: Based on a preset tail risk probability threshold, the distribution of the performance index loss value on the parameter perturbation sample set is analyzed to generate an initial risk threshold. Based on the initial risk threshold, calculate the portion of each performance indicator loss value that exceeds its corresponding initial risk threshold, and generate an over-threshold loss sample. By aggregating and calculating the loss samples exceeding the threshold, a set of conditional risk values ​​is generated.

6. The robust optimization design method for electromagnetic relays based on preference sets according to claim 4, characterized in that, Set initial values ​​for a set of design parameters to be optimized, thus forming the initial design parameter combination; Based on the initial design parameter combination, the process of generating performance index loss value, generating parameter perturbation sample set, obtaining preference weight set, and constructing weighted robustness evaluation index is invoked to calculate the current objective function value; A gradient-based optimization algorithm is used to generate a new combination of design parameters based on the current objective function value and its gradient information with respect to the design parameters. By taking the new combination of design parameters as input, the calculation process of the weighted robustness evaluation index is repeated to obtain the updated objective function value; Determine whether the difference between the updated objective function value and the current objective function value satisfies a preset convergence threshold; If the conditions are not met, the new combination of design parameters will be used as the current combination of electromagnetic relay design parameters, and the parameter update and evaluation process will be executed iteratively. If satisfied, the new combination of design parameters will be output as the optimized combination of design parameters.

7. The robust optimization design method for electromagnetic relays based on preference sets according to claim 6, characterized in that, A gradient-based optimization algorithm is used to dynamically adjust the step size based on the risk volatility at the current design point: ； Among them, P k This is the current parameter vector; The basic learning rate matrix; The gradient vector; As an acceleration factor; is the coefficient of variation of the conditional risk value.

8. The robust optimization design method for electromagnetic relays based on preference sets according to claim 1, characterized in that, Also includes: An independent set of verification samples is generated based on a sampling rule that differs from the parameter perturbation sample set. The optimized design parameter combination and the verification sample set are input into the performance calculation model to calculate the verification performance index loss value. Based on the verification performance index loss value and the preference weight set, the verification robustness evaluation value of the optimized design parameter combination is calculated. When the verification robustness evaluation value is better than the preset verification threshold, the optimized design parameter combination is confirmed to be effective.

9. The robust optimization design method for electromagnetic relays based on preference sets according to claim 8, characterized in that, When the validation robustness evaluation value is not better than the validation threshold, the method further includes: Based on the distribution of the loss values ​​of the verification performance metrics, identify one or more dominant performance metrics that lead to insufficient robustness. Based on the identification results of the dominant performance indicators, the allowable variation range of the design parameter combination of the electromagnetic relay is adjusted to generate an updated design constraint boundary. Based on the updated design constraint boundaries, a new combination of initial design parameters is generated, and the construction and optimization process of the weighted robustness evaluation index is re-executed.

Images