Method for predicting crushing performance of multi-cavity profile for automotive and related apparatus

Abstract

This invention discloses a method and related equipment for predicting the crush performance of multi-cavity profiles for automobiles, belonging to the field of aluminum profile production technology. This invention constructs a sample library by acquiring sample feature data and corresponding crush performance index data from multiple historical profile samples, and obtains candidate feature data for candidate profiles. It determines the matching weight of each crush mechanism feature based on the correlation between each crush mechanism feature and the crush performance index data in the sample library. Based on the candidate feature data, sample feature data, and matching weights, it determines the similarity between the candidate profile and each historical profile sample, and determines the maximum similarity and matching samples. Based on the maximum similarity and a preset similarity threshold, it determines the crush performance prediction index of the candidate profile, and based on the maximum similarity and the number of matching samples, it determines the confidence level of the prediction result, achieving the beneficial effect of reducing the number of new simulations or experiments.

Description

Technical Field

[0001] This invention relates to the field of aluminum profile manufacturing technology, and in particular to a method and related equipment for predicting the crush performance of multi-cavity profiles for automobiles. Background Technology

[0002] In vehicle development, automakers implement quota management for aluminum usage per vehicle to control costs and achieve lightweighting goals. Vehicle quality targets are broken down into systems and components, and the usage of multi-cavity profiles is limited in the early design stages. Designers must ensure crush performance meets these limits. The conceptual design phase generates numerous candidate cross-section designs, each requiring rapid evaluation of its crush performance. Currently, obtaining crush performance data primarily relies on crush tests, finite element simulations, and data modeling. However, testing and simulation are costly and time-consuming, making it impossible to cover all candidate designs. While data modeling enables rapid prediction, historically validated profile data lacks a unified structured description method, and the similarity between new and historical cross-sections cannot be quantified, hindering direct reuse. Furthermore, existing modeling methods only output numerical values, not the reliability of the prediction results, leaving engineers unable to determine when to accept the data directly and when further verification is necessary.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method and related equipment for predicting the crush performance of multi-cavity automotive profiles, aiming to reduce the number of additional simulations or experiments. To achieve the above objective, this invention provides a method for predicting the crush performance of multi-cavity automotive profiles, characterized in that the method includes the following steps: The sample feature data and corresponding crushing performance index data of multiple historical profile samples are obtained to form a sample library, and the candidate feature data of candidate profiles are obtained. The sample feature data includes historical cross-sectional morphology feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphology feature data and candidate crushing mechanism feature data. The matching weight of each crushing mechanism feature is determined based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; The similarity between the candidate profile and each historical profile sample is determined based on the candidate feature data, the sample feature data, and the matching weight, and the maximum similarity and matching sample are determined. The crush performance prediction index of the candidate profile is determined based on the maximum similarity and the preset similarity threshold, and the confidence level of the prediction result is determined based on the maximum similarity and the number of matching samples.

[0005] Optionally, the step of determining the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data, and the matching weight includes: The morphological similarity is determined based on the degree of difference between the candidate cross-sectional morphological feature data and the historical cross-sectional morphological feature data of each historical profile sample, wherein the smaller the degree of difference, the greater the morphological similarity. The sub-similarity between the candidate crushing mechanism feature data and the historical crushing mechanism feature data of each historical profile sample is determined according to the feature group, and the sub-similarity of each feature group is weighted according to the matching weight to determine the mechanism similarity. The similarity between the morphological similarity and the mechanistic similarity is weighted and summed to obtain the similarity between the candidate profile and each of the historical profile samples.

[0006] Optionally, the weight corresponding to the mechanism similarity is greater than the weight corresponding to the morphological similarity.

[0007] Optionally, the step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold includes: When the maximum similarity is not less than the preset similarity threshold, the weight of each matching sample is determined, wherein the weight of each matching sample is positively correlated with the corresponding similarity, and the sum of the weights of all matching samples is normalized. Based on the weight of each matched sample, the crushing performance index data corresponding to the crushing performance evaluation index of each matched sample are weighted and summed to obtain the predicted value of the crushing performance evaluation index of the candidate profile, wherein the crushing performance index data is the data corresponding to at least one crushing performance evaluation index. The crush performance prediction index of the candidate profile is determined based on the predicted value of each of the crush performance evaluation indices.

[0008] Optionally, the step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold includes: Based on the sample feature data and corresponding crush performance index data of each historical profile sample in the sample library, a training set and a validation set are constructed. The preset prediction model is trained according to the training set, and the prediction error of the trained preset prediction model is determined according to the verification set. When the prediction error meets the preset accuracy requirement, the trained preset prediction model is determined as the crushing performance prediction model. When the maximum similarity is less than the preset similarity threshold, the candidate feature data is input into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile.

[0009] Optionally, after the step of inputting the candidate feature data into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile when the maximum similarity is less than the preset similarity threshold, the method further includes: When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the crushing performance prediction model is retrained based on the updated sample library.

[0010] Optionally, after the steps of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold, and determining the confidence level of the prediction result based on the maximum similarity and the number of matched samples, the method further includes: When the confidence level meets the preset confidence condition, the crushing performance prediction index and the confidence level are output. When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the matching weight of each crushing mechanism feature is re-determined according to the updated sample library.

[0011] Furthermore, to achieve the above objectives, the present invention also provides a prediction system for the crushing performance of multi-cavity profiles for automobiles, the prediction system comprising: The acquisition module is used to acquire sample feature data and corresponding crushing performance index data of multiple historical profile samples to form a sample library, and to acquire candidate feature data of candidate profiles. The sample feature data includes historical cross-sectional morphology feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphology feature data and candidate crushing mechanism feature data. The identification module is used to determine the matching weight of each crushing mechanism feature based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; The matching module is used to determine the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data and the matching weight, and to determine the maximum similarity and the matching sample; The prediction module is used to determine the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold, and to determine the confidence level of the prediction result based on the maximum similarity and the number of matching samples.

[0012] Furthermore, to achieve the above objectives, the present invention also provides a device for predicting the crush performance of multi-cavity automotive profiles. The device for predicting the crush performance of multi-cavity automotive profiles includes: a memory, a processor, and a program for predicting the crush performance of multi-cavity automotive profiles stored in the memory and executable on the processor. The program for predicting the crush performance of multi-cavity automotive profiles is configured to implement the steps of the method for predicting the crush performance of multi-cavity automotive profiles as described above.

[0013] In addition, to achieve the above objectives, the present invention also provides a storage medium storing a prediction program for the crush performance of a multi-cavity automotive profile, wherein when the prediction program for the crush performance of a multi-cavity automotive profile is executed by a processor, the program implements the steps of the prediction method for the crush performance of a multi-cavity automotive profile as described above.

[0014] This invention proposes a method for predicting the crush performance of multi-cavity profiles for vehicles. This method constructs a sample library by acquiring sample feature data and corresponding crush performance index data from multiple historical profile samples, and obtains candidate feature data for candidate profiles. It determines the matching weight based on the correlation between each crush mechanism feature and the crush performance index data, thereby determining the similarity, maximum similarity, and matching samples between the candidate profile and each historical profile sample. Finally, it determines the crush performance prediction index and confidence level based on the maximum similarity and a preset similarity threshold. This method can fully reuse historically validated profile data to quickly predict the crush performance of candidate sections and provide the reliability of the prediction results. It enables engineers to determine when to directly accept the prediction results and when further verification is necessary, minimizing the number of additional simulations or experiments while ensuring prediction reliability. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of a device for predicting the crush performance of automotive multi-cavity profiles in the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the method for predicting the crushing performance of multi-cavity profiles for vehicles according to the present invention. Figure 3 This is a flowchart illustrating the second embodiment of the method for predicting the crushing performance of multi-cavity profiles for vehicles according to the present invention.

[0016] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a device for predicting the crushing performance of automotive multi-cavity profiles in the hardware operating environment involved in the embodiments of the present invention.

[0019] like Figure 1 As shown, the device for predicting the crush performance of multi-cavity automotive profiles may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, an interactive device 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The interactive device 1003 may include a display screen or an input unit such as a keyboard. Optionally, the interactive device 1003 may also be connected to the communication bus via standard wired or wireless interfaces. The network interface 1004 may optionally include standard wired or wireless interfaces (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0020] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the predictive device for the crush performance of multi-cavity profiles for vehicles, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0021] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and a program for predicting the crush performance of multi-cavity automotive profiles.

[0022] exist Figure 1 In the vehicle multi-cavity profile crush performance prediction device shown, the network interface 1004 is mainly used for data communication with other devices; the interactive device 1003 is mainly used for data interaction with the user; the processor 1001 and memory 1005 in the vehicle multi-cavity profile crush performance prediction device of the present invention can be set in the vehicle multi-cavity profile crush performance prediction device, the vehicle multi-cavity profile crush performance prediction device calls the vehicle multi-cavity profile crush performance prediction program stored in the memory 1005 through the processor 1001, and executes the vehicle multi-cavity profile crush performance prediction method provided in the embodiment of the present invention.

[0023] This invention provides a method for predicting the crush performance of multi-cavity automotive profiles, referring to... Figure 2 , Figure 2 This is a flowchart illustrating a first embodiment of a method for predicting the crush performance of multi-cavity automotive profiles according to the present invention. In this embodiment, the method for predicting the crush performance of multi-cavity automotive profiles includes: Step S1: Obtain sample feature data and corresponding crush performance index data of multiple historical profile samples to form a sample library, and obtain candidate feature data of candidate profiles. The sample feature data includes historical cross-sectional morphological feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphological feature data and candidate crushing mechanism feature data. In this embodiment, the historical profile samples are multi-cavity profiles that have undergone crush tests or crush simulation verification in historical vehicle projects, such as extruded profiles with multi-cavity cross-sections used in parts like the front longitudinal beam, bumper beam energy absorption box, sill beam, and battery pack edge beam. The sample feature data for each historical profile sample includes historical cross-sectional morphology feature data and historical crush mechanism feature data. The corresponding crush performance index data is data from at least one crush performance evaluation index obtained through crush tests or crush simulation verification. The crush performance evaluation index includes one or more of the following: initial peak crushing force, average crushing force, total energy absorption, specific energy absorption, and crushing force efficiency. The candidate profiles are newly generated cross-sectional schemes during the conceptual design phase. The candidate feature data includes candidate cross-sectional morphology feature data and candidate crush mechanism feature data, and their feature types are the same as those of the sample feature data. It should be noted that the cross-sectional morphology feature data is obtained by directly extracting statistical data from the normalized cross-sectional profile, representing only the overall morphology of the cross-section, decoupled from the absolute dimensions of the profile, and not dependent on prior mechanical knowledge. The crushing mechanism feature data is obtained by deconstructing the normalized cross-sectional profile into multiple basic crushing energy-absorbing units according to crushing association rules, determining the unit characteristics of each basic crushing energy-absorbing unit, and performing statistical processing, representing the deformation and energy absorption behavior of the cross-section during the crushing process. The same feature type is repeatedly extracted in different projects with the same caliber, thereby ensuring that the sample feature data in the sample library is of the same data type as the candidate feature data of the candidate profiles, and thus can be directly compared.

[0024] Step S2: Determine the matching weight of each crushing mechanism feature based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; In this embodiment, the degree of correlation characterizes the influence of each crushing mechanism feature on crushing performance indicators, and can be determined using data analysis methods such as correlation analysis or feature importance ranking. The matching weight is positively correlated with the degree of correlation, that is, the stronger the correlation between the crushing mechanism feature and crushing performance, the greater its weight in the matching. Since the matching weight is determined by actual data in the sample library rather than by human experience, it can objectively reflect the true influence of each mechanism feature on crushing performance. Furthermore, the matching weight can be re-determined as the sample library is updated; the more complete the sample library, the more accurate the weight characterization.

[0025] Step S3: Determine the similarity between the candidate profile and each historical profile sample based on the candidate feature data, the sample feature data, and the matching weight, and determine the maximum similarity and the matching sample; In this embodiment, based on the candidate feature data, the sample feature data, and the matching weight, the similarity between the candidate profile and each historical profile sample in the sample library is calculated one by one. All similarities are then sorted in descending order, and the maximum value is taken as the maximum similarity. A predetermined number of historical profile samples with the highest similarity ranking are selected as the matching samples. The predetermined number can be a fixed value or set according to the size of the sample library. This provides a quantitative basis for subsequent branch scheduling and weighted prediction.

[0026] Step S4: Determine the crush performance prediction index of the candidate profile based on the maximum similarity and the preset similarity threshold, and determine the confidence level of the prediction result based on the maximum similarity and the number of matching samples.

[0027] Specifically, the preset similarity threshold is used to determine whether there are historical schemes in the sample library that are sufficiently similar to the candidate profile. The preset similarity threshold can be a fixed value or adaptively set according to the size and distribution of the sample library. The confidence level represents the credibility of the current prediction result. The confidence level is positively correlated with the maximum similarity and is also related to the number of matching samples. The more matching samples, the higher the confidence level. By uniformly scheduling the prediction path through the similarity threshold, it is possible to directly retrieve and reuse similar schemes when they exist in the library, and to make a fallback prediction by the model when there are no similar schemes in the library. This maximizes the reuse rate of historical data while ensuring prediction accuracy.

[0028] In this embodiment, a sample library is constructed by acquiring sample feature data and corresponding crush performance index data of multiple historical profile samples, and candidate feature data of candidate profiles are obtained. The matching weight is determined according to the degree of correlation between each crush mechanism feature and crush performance index data, and then the similarity, maximum similarity and matching sample between the candidate profile and each historical profile sample are determined. Finally, the crush performance prediction index is determined according to the maximum similarity and the preset similarity threshold, and the confidence level is determined. This allows for full reuse of historically verified profile data, rapid prediction of the crush performance of candidate sections and provision of the credibility of the prediction results. This enables engineers to determine when the prediction results can be directly accepted and when supplementary verification is necessary, minimizing the number of additional simulations or experiments while ensuring the credibility of the prediction.

[0029] Furthermore, based on the first embodiment, a second embodiment of the method for predicting the crushability of multi-cavity profiles for automobiles according to the present invention is proposed. In this embodiment, reference is made to... Figure 3 The step of determining the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data, and the matching weight includes: Step S31: Determine the morphological similarity based on the degree of difference between the candidate cross-sectional morphological feature data and the historical cross-sectional morphological feature data of each historical profile sample, wherein the smaller the degree of difference, the greater the morphological similarity. In this embodiment, the morphological similarity is determined based on the degree of difference between the candidate cross-sectional morphological feature data and the historical cross-sectional morphological feature data of each historical profile sample. The smaller the degree of difference, the greater the morphological similarity. Specifically, the types of cross-sectional morphological feature data include: chamber statistical features, rib statistical features, wall thickness statistical features, and contour statistical features. Chamber statistical features may include at least one of the following: number of chambers, the proportion of the area of ​​each chamber to the total cross-sectional area, and the distribution statistics of chambers along the height direction of the cross-section. Rib statistical features include at least one of the following: number of ribs, and the distribution statistics of the angle between the ribs and the cross-sectional contour. Wall thickness statistical features may include at least one of the following: the mean and dispersion of the wall thickness distribution along the cross-sectional contour. Contour statistical features may include: contour symmetry and the aspect ratio of the circumscribed contour. Since the cross-sectional morphological feature data is decoupled from the absolute dimensions and orientation of the cross-section after normalization, only shape information is retained, thereby providing a unified comparable benchmark for historical profile data of different vehicle models.

[0030] Step S32: Determine the sub-similarity between the candidate crushing mechanism feature data and the historical crushing mechanism feature data of each historical profile sample according to the feature group, and perform weighted processing on the sub-similarity of each feature group according to the matching weight to determine the mechanism similarity; In this embodiment, the crushing mechanism feature data is organized by feature groups. These feature groups include: corner unit feature groups, panel feature groups, chamber circumferential unit feature groups, chamfer unit feature groups, intersection node feature groups, and width difference unit feature groups. Specifically, the corner unit feature group may include the number and distribution of right-angled corner units, T-shaped intersection nodes, cross intersection nodes, and free edge ends; the number and type of corner units directly determine at least one of the following: the average crushing force level. The panel feature group may include: width statistics for each panel; panel width determines at least one of the half-wavelengths of the crushing fold. The chamber circumferential unit feature group may include: perimeter statistics for each chamber; at least one of the following: the correlation between perimeter and energy absorption potential. The chamfer unit feature group may include: chamfer shape statistics; these significantly affect the initial peak crushing force and the folding initiation position. The intersection node feature group may include: node type and number distribution; these significantly affect at least one of the crushing deformation modes. The width difference unit feature group may include: the degree of difference between the maximum and minimum panel widths; these significantly relate to at least one of the overall instability deformation modes. Sub-similarity between candidate profiles and each historical profile sample is determined according to feature groups, and the sub-similarity of each feature group is weighted according to the matching weight to obtain the mechanism similarity. Since each feature group carries proprietary information directly related to crushing energy absorption, the mechanism similarity can truly reflect the degree of similarity between the two profiles in crushing deformation and energy absorption behavior.

[0031] Step S33: Perform a weighted summation of the morphological similarity and the mechanistic similarity to obtain the similarity between the candidate profile and each of the historical profile samples.

[0032] In this embodiment, the morphological similarity and the mechanistic similarity are weighted and summed to obtain the similarity between the candidate profile and each of the historical profile samples. Compared with the method of directly comparing the similarity of all features with equal weight, the weighted synthesis allows the similarity to simultaneously characterize the overall morphological similarity of the cross-section and the similarity of the crushing mechanism, thereby improving the matching accuracy.

[0033] In this embodiment, morphological similarity is determined based on the degree of difference between candidate cross-sectional morphological feature data and historical cross-sectional morphological feature data. Sub-similarity of crushing mechanism features is determined by feature group and weighted according to matching weights to obtain mechanism similarity. Then, the weighted sum of morphological similarity and mechanism similarity yields the overall similarity. This allows similarity matching to be characterized by both the overall cross-sectional morphology and the crushing mechanism, with the matching result dominated by mechanism features carrying crushing mechanical meaning, thus improving the correlation between the matching result and actual crushing performance. Specifically, the similarity calculation process is as follows: the degree of difference between the candidate cross-sectional morphological feature data and the historical cross-sectional morphological feature data of each historical profile sample is calculated; the smaller the degree of difference, the greater the morphological similarity. Optionally, the similarity between corresponding features is calculated by normalizing the above data. Specifically, the similarity can be calculated using the Euclidean distance algorithm: The difference between the candidate profile and the historical profile sample for each feature after normalization is calculated one by one. The square root of the sum of the squares of each difference is taken to obtain the difference distance between the two. Then, 1 is added to the difference distance and the reciprocal is taken to obtain the similarity of the corresponding feature. The smaller the difference distance, the greater the similarity. Alternatively, Manhattan distance or cosine distance can also be used. For each feature group, the sub-similarity of the candidate crushing mechanism feature data and the historical crushing mechanism feature data of the historical profile sample under that feature group is calculated. The sub-similarity of each feature group is multiplied by the corresponding matching weight and summed to obtain the mechanism similarity. The feature group with a stronger correlation to crushing performance has a larger matching weight and contributes more to the mechanism similarity. The morphological similarity and the mechanism similarity are then multiplied by their corresponding weights and summed to obtain the similarity between the candidate profile and the historical profile sample. The weight corresponding to the mechanism similarity is greater than the weight corresponding to the morphological similarity. By using weighted calculations differentiated according to the degree of correlation, the similarity can be dominated by the mechanistic features that carry the meaning of crushing mechanics, thereby improving the actual correlation between similarity and crushing performance.

[0034] Furthermore, the weight corresponding to the mechanism similarity is greater than the weight corresponding to the morphological similarity.

[0035] In this embodiment, the weight corresponding to the mechanism similarity is greater than the weight corresponding to the morphological similarity, meaning that mechanism similarity dominates the synthesized similarity. It should be noted that cross-sectional morphological feature data only characterizes the overall shape of the cross-section; profiles with similar shapes may not have similar crushing performance. In contrast, crushing mechanism feature data carries crushing mechanics meaning and is truly related to crushing performance. By making mechanism similarity dominant, the matching results are more correlated with actual crushing performance, avoiding mismatches caused by shape similarity alone.

[0036] Furthermore, based on the first or second embodiment, a third embodiment of the method for predicting the crush performance of multi-cavity profiles for automobiles according to the present invention is proposed. In this embodiment, the step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold includes: When the maximum similarity is not less than the preset similarity threshold, the weight of each matching sample is determined, wherein the weight of each matching sample is positively correlated with the corresponding similarity, and the sum of the weights of all matching samples is normalized. In this embodiment, when the maximum similarity is not less than the preset similarity threshold, it indicates that there are historical schemes in the sample library that are sufficiently similar to the candidate profile. The retrieval branch is then executed: the weight of each matching sample is determined. The weight of each matching sample is positively correlated with its corresponding similarity; that is, the higher the similarity of the matching sample, the greater its weight in the prediction. Furthermore, the sum of the weights of all matching samples is normalized. This ensures that the prediction result is primarily contributed by the most similar matching sample.

[0037] Based on the weight of each matched sample, the crushing performance index data corresponding to the crushing performance evaluation index of each matched sample are weighted and summed to obtain the predicted value of the crushing performance evaluation index of the candidate profile, wherein the crushing performance index data is the data corresponding to at least one crushing performance evaluation index. In this embodiment, each crushing performance evaluation index is processed separately: the crushing performance index data corresponding to the crushing performance evaluation index of each matched sample are weighted and summed according to their corresponding weights to obtain the predicted value of the crushing performance evaluation index of the candidate profile; the crushing performance index data is the data corresponding to at least one crushing performance evaluation index, such as initial peak crushing force, average crushing force, total energy absorption, specific energy absorption, and crushing force efficiency. Since the crushing performance index data of the matched samples have all been verified by crushing tests or crushing simulations, and the selection of matched samples is dominated by crushing mechanism characteristics, the predicted value output by the retrieval branch is essentially a weighted reuse of verified data from samples with similar mechanisms, and its credibility is directly supported by similarity.

[0038] The crush performance prediction index of the candidate profile is determined based on the predicted value of each of the crush performance evaluation indices.

[0039] In this embodiment, when there are multiple crushing performance evaluation indicators, the above-described weighted summation process is repeatedly performed on each crushing performance evaluation indicator. The predicted values ​​of each crushing performance evaluation indicator are then aggregated and determined as the crushing performance prediction indicators of the candidate profile. This allows for the output of multiple crushing performance prediction indicators for the candidate profile at once, meeting the multi-indicator evaluation requirements in the scheme design stage.

[0040] In this embodiment, the weight of each matching sample is determined when the maximum similarity is not less than a preset similarity threshold, and the crushing performance index data of each matching sample are weighted and summed according to the weight of each matching sample to obtain the predicted value of each crushing performance evaluation index, thereby determining the crushing performance prediction index of the candidate profile. This enables the weighted reuse of historically verified data with similar mechanisms, and allows for the rapid output of reliable prediction results without the need for additional simulations or experiments.

[0041] Furthermore, based on any of the above embodiments, a fourth embodiment of the method for predicting the crush performance of multi-cavity profiles for automobiles according to the present invention is proposed. In this embodiment, the step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold includes: Based on the sample feature data and corresponding crush performance index data of each historical profile sample in the sample library, a training set and a validation set are constructed. In this embodiment, based on the sample feature data and corresponding crush performance index data of each historical profile sample in the sample library, a training set and a validation set are constructed according to a preset ratio. The training set is used for training model parameters, and the validation set is used for evaluating model prediction errors. This provides a data foundation for model training and ensures the objectivity of error evaluation.

[0042] The preset prediction model is trained according to the training set, and the prediction error of the trained preset prediction model is determined according to the verification set. When the prediction error meets the preset accuracy requirement, the trained preset prediction model is determined as the crushing performance prediction model. In this embodiment, the preset prediction model is a supervised learning model, such as a neural network model, a gradient boosting tree model, or a support vector regression model. The input dimension of the preset prediction model is consistent with the feature dimension of the candidate feature data, and the output is a crushing performance index. Specifically, training is performed using sample feature data from the training set as input and the corresponding crushing performance index data as the desired output. During training, cross-validation is used to evaluate the prediction error, and the prediction error of the trained preset prediction model is determined based on the validation set. When the prediction error meets the preset accuracy requirement, training is completed, and the trained preset prediction model is determined as the crushing performance prediction model. This ensures that the prediction accuracy of the crushing performance prediction model meets engineering requirements.

[0043] When the maximum similarity is less than the preset similarity threshold, the candidate feature data is input into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile.

[0044] In this embodiment, when the maximum similarity is less than the preset similarity threshold, it indicates that there are no historical schemes in the sample library that are sufficiently similar to the candidate profile, meaning the candidate profile is outside the coverage of the sample library. At this point, the model branch is executed: the candidate feature data is input into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile. It should be noted that the crush performance prediction model is only invoked under the above conditions; that is, the model branch only processes candidate profiles outside the coverage of the sample library. This allows the retrieval branch and the model branch to perform their respective functions, maximizing the reuse rate of historical data while ensuring prediction accuracy.

[0045] Furthermore, after the step of inputting the candidate feature data into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile when the maximum similarity is less than the preset similarity threshold, the method further includes: When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the crushing performance prediction model is retrained based on the updated sample library.

[0046] In this embodiment, when the confidence level does not meet the preset confidence condition, it indicates that the reliability of the current prediction result is insufficient and cannot be directly accepted. In this case, a crush simulation or crush test is performed on the candidate profile for supplementary verification. The verification feature data and verification index data obtained from the verification are extracted and written into the sample library. The crush performance prediction model is then retrained based on the updated sample library. Retraining can be performed incrementally or fully. It should be noted that the supplementary verification is triggered only once when the confidence level is insufficient, rather than being performed indiscriminately on every candidate scheme. This ensures that each supplementary verification targets the weak points covered by the sample library, minimizing the number of additional simulations or tests while maintaining prediction reliability.

[0047] In this embodiment, a training set and a validation set are constructed based on a sample library. A preset prediction model is trained and determined to be a crush performance prediction model when the prediction error meets the preset accuracy requirements. When the maximum similarity is less than the preset similarity threshold, candidate feature data is input into the crush performance prediction model to obtain crush performance prediction indicators. When the confidence level does not meet the preset confidence conditions, a crush simulation or crush test is performed on the candidate profile, the validation data is written into the sample library, and the model is retrained. Thus, the model branch can perform fallback predictions for candidate profiles outside the coverage of the sample library, and the model is continuously optimized as the sample library is updated.

[0048] Furthermore, based on any of the above embodiments, a fifth embodiment of the method for predicting the crush performance of multi-cavity profiles for automobiles according to the present invention is proposed. In this embodiment, after the steps of determining the crush performance prediction index of the candidate profile based on the maximum similarity and the preset similarity threshold, and determining the confidence level of the prediction result based on the maximum similarity and the number of matching samples, the method further includes: When the confidence level meets the preset confidence condition, the crushing performance prediction index and the confidence level are output. In this embodiment, the preset confidence condition is that the confidence level is not less than a preset confidence threshold. The preset confidence threshold can be a fixed value or adaptively set according to the sample library size and distribution. When the confidence level meets the preset confidence condition, it indicates that the reliability of the prediction result meets the acceptance standard. The crushing performance prediction index and the confidence level are output, and engineers can directly accept the prediction results for scheme selection. This ensures that each direct acceptance of the prediction result has a clear confidence level basis.

[0049] When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the matching weight of each crushing mechanism feature is re-determined according to the updated sample library.

[0050] In this embodiment, when the confidence level does not meet the preset confidence condition, a crush simulation or crush test is performed on the candidate profile. The verified feature data and verification index data obtained are written into the sample library, and the matching weight of each crush mechanism feature is re-determined based on the updated sample library. It should be noted that as the coverage of the sample library increases, the proportion of subsequent candidate profiles hitting the retrieval branch increases, and the frequency of triggering supplementary verification gradually decreases; the matching weight becomes more accurate with the accumulation of sample library data, and the matching accuracy improves synchronously, further reducing the long-term cost of the solution.

[0051] In this embodiment, when the confidence level meets the preset confidence conditions, the crush performance prediction index and confidence level are output. When the confidence level does not meet the preset confidence conditions, a crush simulation or crush test is performed on the candidate profile. The verification feature data and verification index data obtained are written into the sample library and the matching weight of each crush mechanism feature is re-determined. This ensures that every direct adoption prediction result has a confidence level basis, and each supplementary verification targets the weak links covered by the sample library. As the coverage of the sample library increases, the frequency of triggering supplementary verification gradually decreases, minimizing the number of new verifications and further reducing long-term usage costs.

[0052] Furthermore, this invention also proposes a prediction system for the crushing performance of multi-cavity automotive profiles, characterized in that the prediction system for the crushing performance of multi-cavity automotive profiles includes: The acquisition module is used to acquire sample feature data and corresponding crushing performance index data of multiple historical profile samples to form a sample library, and to acquire candidate feature data of candidate profiles. The sample feature data includes historical cross-sectional morphology feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphology feature data and candidate crushing mechanism feature data. The identification module is used to determine the matching weight of each crushing mechanism feature based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; The matching module is used to determine the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data and the matching weight, and to determine the maximum similarity and the matching sample; The prediction module is used to determine the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold, and to determine the confidence level of the prediction result based on the maximum similarity and the number of matching samples.

[0053] Furthermore, this invention also proposes a device for predicting the crush performance of multi-cavity automotive profiles, characterized in that the device includes: a memory, a processor, and a program for predicting the crush performance of multi-cavity automotive profiles stored in the memory and executable on the processor, wherein the program for predicting the crush performance of multi-cavity automotive profiles is configured to implement the steps of the method for predicting the crush performance of multi-cavity automotive profiles as described above.

[0054] Furthermore, this embodiment of the invention also proposes a storage medium storing a prediction program for the crush performance of automotive multi-cavity profiles. When the prediction program for the crush performance of automotive multi-cavity profiles is executed by a processor, it implements the steps of the prediction method for the crush performance of automotive multi-cavity profiles described above.

[0055] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0056] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0057] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0058] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for predicting the crush performance of multi-cavity profiles for automobiles, characterized in that, The method for predicting the crush performance of multi-cavity profiles for automobiles includes the following steps: The sample feature data and corresponding crushing performance index data of multiple historical profile samples are obtained to form a sample library, and the candidate feature data of candidate profiles are obtained. The sample feature data includes historical cross-sectional morphology feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphology feature data and candidate crushing mechanism feature data. The matching weight of each crushing mechanism feature is determined based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; The similarity between the candidate profile and each historical profile sample is determined based on the candidate feature data, the sample feature data, and the matching weight, and the maximum similarity and matching sample are determined. The crush performance prediction index of the candidate profile is determined based on the maximum similarity and the preset similarity threshold, and the confidence level of the prediction result is determined based on the maximum similarity and the number of matching samples.

2. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 1, characterized in that, The step of determining the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data, and the matching weight includes: The morphological similarity is determined based on the degree of difference between the candidate cross-sectional morphological feature data and the historical cross-sectional morphological feature data of each historical profile sample, wherein the smaller the degree of difference, the greater the morphological similarity. The sub-similarity between the candidate crushing mechanism feature data and the historical crushing mechanism feature data of each historical profile sample is determined according to the feature group, and the sub-similarity of each feature group is weighted according to the matching weight to determine the mechanism similarity. The similarity between the morphological similarity and the mechanistic similarity is weighted and summed to obtain the similarity between the candidate profile and each of the historical profile samples.

3. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 2, characterized in that, The weight corresponding to the mechanism similarity is greater than the weight corresponding to the morphological similarity.

4. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 1, characterized in that, The step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and the preset similarity threshold includes: When the maximum similarity is not less than the preset similarity threshold, the weight of each matching sample is determined, wherein the weight of each matching sample is positively correlated with the corresponding similarity, and the sum of the weights of all matching samples is normalized. Based on the weight of each matched sample, the crushing performance index data corresponding to the crushing performance evaluation index of each matched sample are weighted and summed to obtain the predicted value of the crushing performance evaluation index of the candidate profile, wherein the crushing performance index data is the data corresponding to at least one crushing performance evaluation index. The crush performance prediction index of the candidate profile is determined based on the predicted value of each of the crush performance evaluation indices.

5. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 1, characterized in that, The step of determining the crush performance prediction index of the candidate profile based on the maximum similarity and the preset similarity threshold includes: Based on the sample feature data and corresponding crush performance index data of each historical profile sample in the sample library, a training set and a validation set are constructed. The preset prediction model is trained according to the training set, and the prediction error of the trained preset prediction model is determined according to the verification set. When the prediction error meets the preset accuracy requirement, the trained preset prediction model is determined as the crushing performance prediction model. When the maximum similarity is less than the preset similarity threshold, the candidate feature data is input into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile.

6. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 5, characterized in that, After the step of inputting the candidate feature data into the crush performance prediction model to obtain the crush performance prediction index of the candidate profile when the maximum similarity is less than the preset similarity threshold, the method further includes: When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the crushing performance prediction model is retrained based on the updated sample library.

7. The method for predicting the crush performance of multi-cavity profiles for automobiles as described in claim 6, characterized in that, After the steps of determining the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold, and determining the confidence level of the prediction result based on the maximum similarity and the number of matched samples, the method further includes: When the confidence level meets the preset confidence condition, the crushing performance prediction index and the confidence level are output. When the confidence level does not meet the preset confidence conditions, a crushing simulation or crushing test is performed on the candidate profile, the verification feature data and verification index data obtained are written into the sample library, and the matching weight of each crushing mechanism feature is re-determined according to the updated sample library.

8. A predictive system for the crush performance of multi-cavity profiles for automobiles, characterized in that, The system for predicting the crushing performance of multi-cavity profiles for automobiles includes: The acquisition module is used to acquire sample feature data and corresponding crushing performance index data of multiple historical profile samples to form a sample library, and to acquire candidate feature data of candidate profiles. The sample feature data includes historical cross-sectional morphology feature data and historical crushing mechanism feature data, and the candidate feature data includes candidate cross-sectional morphology feature data and candidate crushing mechanism feature data. The identification module is used to determine the matching weight of each crushing mechanism feature based on the degree of correlation between each crushing mechanism feature in the sample library and the crushing performance index data; The matching module is used to determine the similarity between the candidate profile and each of the historical profile samples based on the candidate feature data, the sample feature data and the matching weight, and to determine the maximum similarity and the matching sample; The prediction module is used to determine the crush performance prediction index of the candidate profile based on the maximum similarity and a preset similarity threshold, and to determine the confidence level of the prediction result based on the maximum similarity and the number of matching samples.

9. A device for predicting the crush performance of multi-cavity automotive profiles, characterized in that, The device for predicting the crush performance of multi-cavity automotive profiles includes: a memory, a processor, and a program for predicting the crush performance of multi-cavity automotive profiles stored in the memory and executable on the processor, wherein the program for predicting the crush performance of multi-cavity automotive profiles is configured to implement the steps of the method for predicting the crush performance of multi-cavity automotive profiles as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a prediction program for the crush performance of automotive multi-cavity profiles. When the prediction program for the crush performance of automotive multi-cavity profiles is executed by a processor, it implements the steps of the prediction method for the crush performance of automotive multi-cavity profiles as described in any one of claims 1 to 7.

Images