Intelligent selection method and system for occupant protection and restraint system solutions based on AI model

Abstract

The present application discloses an AI-model-based intelligent selection method and system for an occupant protection restraint system, which relates to the field of artificial intelligence technology. The method comprises: obtaining occupant protection restraint system data of multiple modalities; integrating test data, simulation model data, and expert experience data into a multimodal input data set; performing feature extraction on the multimodal input data set using feature extraction models corresponding to multiple different modalities to obtain multimodal feature data of the multimodal input data set; training prediction models for multiple different selection targets using the multimodal feature data to output a reference selection scheme for a vehicle occupant protection restraint system for vehicle design; obtaining an optimal restraint system selection scheme that meets design requirements from multiple reference selection schemes for a vehicle occupant protection restraint system based on the expert experience data; and updating the selection model based on simulation test results corresponding to the optimal restraint system selection scheme and the expert experience data.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an AI model-based intelligent selection method and system for an occupant protection and restraint system solution. Background Art

[0002] Finite element simulation technology, a key automotive safety design tool, can reduce the cost and resource investment associated with physical crash testing during vehicle development. Currently, most traditional simulation methods rely on numerical simulation techniques, using CAE (Computer Aided Engineering) finite element software. This method uses detailed numerical simulation of the vehicle's occupant protection and restraint systems to simulate the dynamic response and injury of occupants during a collision, thereby evaluating the protective performance of the occupant protection and restraint systems. Numerical simulation models for occupant protection and restraint systems are typically complex, requiring detailed modeling of all system components within the passenger compartment, including the vehicle body, seatbelts, airbags, seats, steering column, instrument panel, accelerator pedal, and floor, positioning of the crash dummy, and setting of contact parameters between components. The accuracy and applicability of these models depend on numerous factors, including modeling methods, material parameters, boundary conditions, and contact parameters. Improper handling of these factors can lead to inaccurate simulation results, ultimately deviating from actual results. Simultaneously, when faced with multiple simulation results, professional technicians must analyze and select a feasible and effective solution, often relying heavily on the experience of experienced engineers.

[0003] With the rapid development of the automobile industry, domestic and foreign automobile collision safety evaluation regulations have been upgraded and revised. The requirements for automobile collision performance have become increasingly stringent, and the number of collision conditions has increased. The difficulty of simulation technology has also increased. Faced with the large number of development conditions and the high difficulty, the use of traditional simulation methods has problems such as long development cycle and inaccurate effect evaluation. Summary of the Invention

[0004] The purpose of this application is to provide an AI-based intelligent selection method and system for occupant protection and restraint system solutions.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In the first aspect, the present application provides an AI-based intelligent selection method for an occupant protection and restraint system solution, including:

[0007] Acquiring occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data includes at least: test data, simulation model data, and expert experience data;

[0008] Integrating the test data, simulation model data, and expert experience data into a multimodal input data set;

[0009] Performing feature extraction on the multimodal input data set using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input data set;

[0010] In the process of training the prediction models for multiple different selection targets using the multimodal feature data, a reference selection scheme for a vehicle occupant protection and restraint system is output for vehicle design, and an optimal restraint system selection scheme that meets the design requirements is obtained from the multiple reference selection schemes for the vehicle occupant protection and restraint system based on the expert experience data;

[0011] The prediction model is updated based on simulation test results corresponding to the optimal restraint system selection scheme and expert experience data.

[0012] Optionally, the step of obtaining occupant protection restraint system data of multiple modes includes:

[0013] Acquire occupant protection restraint system data of multiple different modalities from multiple data sources; wherein both the test data and the simulation model data include: numerical data, which are restraint system parameters; collision results, which include at least: head displacement, head injury, neck force, chest pressure displacement and score; curve data, which include at least: collision pulse curve; text data, which include at least: restraint system solution decision-making method and selection conclusion;

[0014] The occupant protection restraint system data is cleaned by the following methods:

[0015] For the numerical data, identify and eliminate abnormal data, and perform denoising and smoothing correction processing;

[0016] Perform continuity and integrity checks on the curve data, traverse and eliminate incomplete data, and the incomplete data includes at least: data with obvious discontinuities and data that does not show the entire collision process;

[0017] The text data is subjected to semantic screening and noise filtering to remove irrelevant text information and noise that interferes with semantics.

[0018] Optionally, the step of performing feature extraction on the multimodal input dataset using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input dataset includes:

[0019] For the curve data after data cleaning, the following three methods are used for feature extraction:

[0020] Extracting time domain features from the original curve, wherein the time domain features include at least: pulse duration, pulse energy, and pulse effective value;

[0021] Performing Fourier transform on the original curve, converting the original curve from the time domain to the frequency domain to extract frequency domain features, wherein the frequency domain features include at least: primary frequency amplitude, secondary frequency amplitude, frequency energy, frequency center, and spectrum entropy;

[0022] According to the waveform characteristics of the original curve, the original curve is fitted into a data function, and the waveform characteristics of the fitting function are extracted. The fitting types of the original curve fitting include: double trapezoidal wave, peak rectangular wave, and half sine wave;

[0023] For text data after data cleaning, knowledge extraction, knowledge representation and data encoding are performed respectively to convert the text data features into vector features that can be recognized by the training model.

[0024] Optionally, the step of integrating the test data, simulation model data and expert experience data into a multimodal input data set includes:

[0025] performing normalization processing on each numerical data in the occupant protection restraint system data;

[0026] The standardized features of various types of preprocessed data are fused to form a multimodal feature dataset within a unified scale.

[0027] Optionally, in the process of training the prediction models for multiple different selection targets using the multimodal feature data, the steps of outputting a reference selection scheme for a vehicle occupant protection and restraint system for vehicle design, and obtaining an optimal restraint system selection scheme that meets design requirements from the multiple reference selection schemes for the vehicle occupant protection and restraint system based on the expert experience data include:

[0028] Determine the selection target based on the passenger type, where the passenger type includes at least: driver, co-driver, rear passenger, and child;

[0029] Under the selected selection target, independent data sets are collated based on the working conditions. The input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities. The regression target quantities include at least: head injury, neck force, chest pressure displacement, and score.

[0030] Based on independent data sets representing various operating conditions, the multimodal feature data is trained using regression and classification prediction algorithms to perform multiple target quantity occupant protection injury prediction tasks. Each input feature represents a restraint system selection scheme, and the classification target quantity and regression target quantity serve as the basis for screening restraint system selection schemes for reference selection of vehicle occupant protection restraint system schemes.

[0031] A natural language processing model is used to train the reference selection scheme of the vehicle occupant protection restraint system in combination with the characteristic parameters of expert experience data to perform decision analysis on the reference selection scheme of the restraint system and output the optimal restraint system selection scheme that meets the restraint system score and design requirements.

[0032] Optionally, the step of updating the prediction model based on simulation test results and expert experience data corresponding to the optimal constraint system selection solution includes:

[0033] After finite element simulation or actual vehicle test verification is performed using the screened optimal restraint system selection scheme that meets the restraint system score and design requirements, a new data set is obtained based on the collision result data and expert experience data generated during the simulation or test process, and the prediction model is updated and corrected in combination with the new data set.

[0034] Secondly, this application provides an AI-based intelligent selection system for occupant protection and restraint systems, including:

[0035] an acquisition module, configured to acquire occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data includes at least: test data, simulation model data, and expert experience data;

[0036] An extraction module, configured to integrate the test data, simulation model data, and expert experience data into a multimodal input data set;

[0037] Performing feature extraction on the multimodal input data set using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input data set;

[0038] a training module for outputting a reference selection scheme for a vehicle occupant protection restraint system for vehicle design during a process of training a plurality of prediction models for different selection targets using the multimodal feature data, and obtaining an optimal restraint system selection scheme that meets design requirements from the plurality of reference selection schemes for the vehicle occupant protection restraint system based on the expert experience data;

[0039] An updating module is used to update the prediction model based on simulation test results corresponding to the optimal constraint system selection scheme and expert experience data.

[0040] Optionally, the acquisition module is further configured to:

[0041] Acquire occupant protection restraint system data of multiple different modalities from multiple data sources; wherein both the test data and the simulation model data include: numerical data, which are restraint system parameters; collision results, which include at least: head displacement, head injury, neck force, chest pressure displacement and score; curve data, which include at least: collision pulse curve; text data, which include at least: restraint system solution decision-making method and selection conclusion;

[0042] The occupant protection restraint system data is cleaned by the following methods:

[0043] For the numerical data, identify and eliminate abnormal data, and perform denoising and smoothing correction processing;

[0044] Perform continuity and integrity checks on the curve data, traverse and eliminate incomplete data, and the incomplete data includes at least: data with obvious discontinuities and data that does not show the entire collision process;

[0045] The text data is subjected to semantic screening and noise filtering to remove irrelevant text information and noise that interferes with semantics.

[0046] Optionally, the extraction module is further configured to:

[0047] For the curve data after data cleaning, the following three methods are used for feature extraction:

[0048] Extracting time domain features from the original curve, wherein the time domain features include at least: pulse duration, pulse energy, and pulse effective value;

[0049] Performing Fourier transform on the original curve, converting the original curve from the time domain to the frequency domain to extract frequency domain features, wherein the frequency domain features include at least: primary frequency amplitude, secondary frequency amplitude, frequency energy, frequency center, and spectrum entropy;

[0050] According to the waveform characteristics of the original curve, the original curve is fitted into a data function, and the waveform characteristics of the fitting function are extracted. The fitting types of the original curve fitting include: double trapezoidal wave, peak rectangular wave, and half sine wave;

[0051] For text data after data cleaning, knowledge extraction, knowledge representation and data encoding are performed respectively to convert the text data features into vector features that can be recognized by the training model.

[0052] Optionally, the extraction module is further configured to:

[0053] performing normalization processing on each numerical data in the occupant protection restraint system data;

[0054] The standardized features of various types of preprocessed data are fused to form a multimodal feature dataset within a unified scale.

[0055] Optionally, the training module is further used to:

[0056] Determine the selection target based on the passenger type, where the passenger type includes at least: driver, co-driver, rear passenger, and child;

[0057] Under the selected selection target, independent data sets are collated based on the working conditions. The input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities. The regression target quantities include at least: head injury, neck force, chest pressure displacement, and score.

[0058] Based on independent data sets representing various operating conditions, the multimodal feature data is trained using regression and classification prediction algorithms to perform multiple target quantity occupant protection injury prediction tasks. Each input feature represents a restraint system selection scheme, and the classification target quantity and regression target quantity serve as the basis for screening restraint system selection schemes for reference selection of vehicle occupant protection restraint system schemes.

[0059] A natural language processing model is used to train the reference selection scheme of the vehicle occupant protection restraint system in combination with the characteristic parameters of expert experience data to perform decision analysis on the reference selection scheme of the restraint system and output the optimal restraint system selection scheme that meets the restraint system score and design requirements.

[0060] Optionally, the update module is further configured to:

[0061] After finite element simulation or actual vehicle test verification is performed using the screened optimal restraint system selection scheme that meets the restraint system score and design requirements, a new data set is obtained based on the collision result data and expert experience data generated during the simulation or test process, and the prediction model is updated and corrected in combination with the new data set.

[0062] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-mentioned method for intelligent selection of an occupant protection restraint system solution based on an AI model.

[0063] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-mentioned AI model-based intelligent selection methods for an occupant protection restraint system solution.

[0064] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned AI model-based intelligent selection methods for occupant protection restraint system solutions.

[0065] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0066] The present application provides an AI-based model-based intelligent selection method and system for occupant protection restraint system solutions. By combining occupant protection restraint system data of different modalities with expert experience data to generate multimodal sample feature data, multiple models can learn richer features and more complex correlations. This training method not only improves the accuracy of the model, but also enables the model to adapt to multiple task objectives. The prediction model can quickly respond to the relevant parameter inputs of the vehicle design solution and output the optimal occupant protection restraint system selection solution based on these parameters, realizing full process automation from data collection to solution output, greatly improving the efficiency and accuracy of vehicle occupant protection restraint system selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0068] Figure 1 A flowchart of an AI-based intelligent selection method for an occupant protection and restraint system solution provided in one embodiment of the present application;

[0069] Figure 2 A schematic diagram illustrating the effects of an AI-based intelligent selection method for an occupant protection and restraint system solution provided in one embodiment of the present application;

[0070] Figure 3 A flowchart of an AI-based intelligent selection system for an occupant protection and restraint system according to an embodiment of the present application is provided;

[0071] Figure 4 A schematic diagram of the functional modules of an AI-based intelligent selection system for an occupant protection and restraint system according to one embodiment of the present application;

[0072] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0074] like Figure 1 As shown, some embodiments of the present application provide an AI model-based intelligent selection method for an occupant protection and restraint system solution, which includes the following steps 101 to 105. Among them:

[0075] Step 101 : Acquire occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data at least includes: test data, simulation model data, and expert experience data.

[0076] In the examples of this application, modal data includes, but is not limited to, R&D test data (such as slide tests and real-car crash test data), R&D simulation data (simulation model data built using CAE software), and expert experience data (expert knowledge and historical technical reports). The acquisition of this data provides a solid foundation for subsequent model construction.

[0077] Step 102: Integrate the test data, simulation model data, and expert experience data into a multimodal input data set.

[0078] Step 103: Perform feature extraction on the multimodal input dataset using feature extraction models corresponding to multiple different modalities to obtain multimodal feature data of the multimodal input dataset.

[0079] In the embodiments of this application, after acquiring diverse data, feature extraction is performed on each type of data. This step uses a feature extraction model that matches the modality to extract key features. The purpose of feature extraction is to convert the raw data into a form that the model can understand and learn from, allowing for subsequent analysis and prediction.

[0080] In the embodiments of the present application, since occupant protection and restraint system data involves multiple types (such as numerical values, curves, and text), these modalities have different feature representations. Therefore, it is necessary to fuse these different types of feature data at the feature layer or decision layer to form unified multimodal feature data. Multimodal feature fusion can fully utilize the information from various modalities and improve the model's predictive performance. Specifically, at the feature layer, different types of features are fused by concatenating or weighted summing the extracted numerical, text, and curve features. This method preserves the rich information of the original features, but requires addressing the dimensionality and size differences between the different features. At the decision layer, sub-models are trained separately for each type of data, and then the prediction results of each sub-model are fused at the decision layer. Fusion can be performed using methods such as voting and weighted averaging, with weights assigned based on the prediction accuracy and reliability of the sub-models.

[0081] Furthermore, by combining the advantages of feature-level fusion and decision-level fusion, we first perform feature-level fusion on some features, and then perform decision-level fusion on the fused features. This approach can more flexibly handle different types of data features and improve the model's predictive performance.

[0082] Step 104: In the process of training the prediction models of multiple different selection targets using the multimodal feature data, a reference selection scheme for the vehicle occupant protection restraint system is output for vehicle design, and an optimal restraint system selection scheme that meets the design requirements is obtained from the multiple reference selection schemes for the vehicle occupant protection restraint system based on the expert experience data.

[0083] In the embodiments of this application, after obtaining multimodal feature data, a prediction model is trained using this data. The choice of prediction model depends on the mission objectives and can include classification models, regression models, and others. During the training process, the model parameters are adjusted by updating the algorithm to enable the model to accurately predict the performance or effectiveness of the occupant protection restraint system.

[0084] In this embodiment of the present application, the predictive model is used to guide the selection of a vehicle's occupant protection and restraint system. By inputting relevant vehicle design parameters, the model outputs the optimal occupant protection and restraint system selection plan. This fully automated process, from data collection to solution output, significantly improves the efficiency and accuracy of occupant protection and restraint system development.

[0085] For example, refer to Figure 2 First, the collision law is determined to determine the passive safety collision conditions, and then the various requirements and configuration parameters of the front collision, as well as the various requirements and configuration parameters of the side collision, are determined. By inputting them into the prediction model, the injury results and injury values ​​of different parts of the occupants, as well as the selection schemes of the front collision restraint system and the side collision restraint system can be obtained.

[0086] Step 105 : updating the prediction model based on simulation test results corresponding to the optimal constraint system selection solution and expert experience data.

[0087] In an embodiment of the present application, a test set or a validation set can be used to evaluate the performance of the model, including indicators such as accuracy, recall, F1 score, and R2. Feedback and updates are performed on the model based on the evaluation results. The model's hyperparameters, update algorithms, or fusion strategies are adjusted based on the evaluation results to improve the performance of the model. At the same time, new features or data can be introduced to further improve the accuracy of the model. When generating a solution for the occupant protection restraint system, excellent design cases and common problems in manual development experience are fully considered. After the candidate solutions are generated by the data mining model, they are further updated and adjusted in combination with expert experience data. Expert experience data is used to screen and sort the solutions to select the solution that best meets actual needs and expected goals.

[0088] In the embodiments of the present application, a test set or a validation set is used to evaluate the performance of the model, including indicators such as accuracy, recall, F1 score, and R2. Feedback and updates to the model are performed based on the evaluation results.

[0089] Based on the evaluation results, model hyperparameters, algorithm updates, or fusion strategies can be adjusted to improve model performance. New features or data can also be introduced to further enhance model accuracy. When generating occupant protection restraint system solutions, consider both excellent design examples and common problems from manual development experience. After candidate solutions are generated through the data mining model, they are further updated and adjusted based on expert feedback. Expert experience is leveraged to screen and rank the solutions, selecting the one that best meets actual needs and desired objectives.

[0090] Validate the selected solution in a simulation environment, simulating the occupant protection effects under different crash scenarios. Further adjust and update the solution based on simulation results and expert experience. Field testing or simulation experiments are conducted, where conditions permit, to verify the solution's effectiveness and collect additional feedback for subsequent improvements.

[0091] The embodiment of the present application combines occupant protection restraint system data of different modalities with expert experience data to obtain multimodal sample feature data, so that multiple models can learn richer features and more complex associations. This training method not only improves the accuracy of the model, but also enables the model to adapt to multiple task objectives. The prediction model can quickly respond to the relevant parameter inputs of the vehicle design plan and output the optimal occupant protection restraint system selection plan based on these parameters, realizing the automation of the entire process from data collection to plan output, greatly improving the efficiency and accuracy of the vehicle occupant protection restraint system selection.

[0092] Optionally, step 101 includes:

[0093] Step 1011: Acquire occupant protection restraint system data of multiple different modalities from multiple data sources; wherein both the test data and the simulation model data include: numerical data, which are restraint system parameters; collision results, which include at least: head displacement, head injury, neck force, chest pressure displacement and score; curve data, which include at least: collision pulse curve; and text data, which include at least: restraint system solution decision-making method and selection conclusion.

[0094] In the embodiments of the present application, the data sources include but are not limited to R&D test data, R&D simulation data, and R&D experience data.

[0095] R&D test data is derived from historical data on R&D vehicle models and actual data obtained through slide tests and real-vehicle crash tests. This data includes occupant injury values ​​for various body parts, restraint system solutions, crash impulses, restraint system parameters, crash results, and scores. This data is authentic and reliable, and represents validation data from various stages of the vehicle development process, directly reflecting the actual effectiveness of the restraint system.

[0096] R&D simulation data is derived from historical simulation model data for the R&D vehicle model. Using computer simulation technology, a simulation model of the occupant restraint system is constructed. This model, based on experimental correlation analysis, offers a high degree of accuracy and can accurately simulate occupant injuries and dynamic responses under various collision scenarios. The R&D simulation data includes individual occupant injury values, restraint system solutions, collision impulses, restraint system parameters, collision results, and scores. Detailed 3D data features can be extracted from this R&D simulation data.

[0097] Expert experience data is derived from historical technical reports on R&D vehicle models and expert knowledge. This R&D experience data includes restraint system solution decision-making methods, technical roadmaps for restraint system upgrades, problem-solving solutions, and selection conclusions. This expert experience data contains extensive experience in occupant protection restraint system design and provides proven and effective options.

[0098] Step 1012: Clean the occupant protection and restraint system data by:

[0099] For the numerical data, identify and eliminate abnormal data, and perform denoising and smoothing correction processing;

[0100] In the embodiment of the present application, numerical data includes restraint system parameters, such as seat belt parameters, airbag parameters, and seat parameters, as well as collision results, such as head displacement, head injury, neck force, chest pressure displacement, and scores.

[0101] Abnormal data typically includes NaN values ​​(e.g., null or missing values) and outliers. NaN values ​​directly indicate missing data, while outliers can arise from measurement errors or equipment failures. By setting appropriate thresholds or using statistical methods (such as boxplots), outliers can be identified and removed to ensure data accuracy.

[0102] Numerical data may contain noise, which is random fluctuations or small errors. This noise can be caused by the precision limitations of measurement equipment or environmental factors. Denoising methods, such as moving averages and median filters, can smooth data curves and reduce the impact of noise on data analysis.

[0103] Smoothing correction processing further ensures that the data can still reflect the real physical phenomenon or trend after smoothing.

[0104] The curve data is subjected to continuity and integrity checks, and incomplete data is traversed and eliminated. The incomplete data at least includes: data with obvious discontinuities and data that does not show the entire collision process.

[0105] In this embodiment, the curve data is traversed to check for significant discontinuities or data that does not represent the entire collision process. Discontinuities may be caused by interruptions in data recording or sensor failure, while data that does not represent the entire collision process may be caused by interruptions in the simulation process, improper data capture, or measurement range limitations.

[0106] Curve data with obvious discontinuities or that do not show the entire collision process should be eliminated to ensure the integrity and reliability of the data.

[0107] After removing incomplete data, the curve needs to be identified and smoothed. Key point identification aims to extract key characteristic points in the curve, such as peaks, valleys, and inflection points, for subsequent feature extraction. Curve smoothing uses an algorithm to smooth the curve to reduce the impact of noise or abnormal fluctuations on the analysis results.

[0108] The text data is subjected to semantic screening and noise filtering to remove irrelevant text information and noise that interferes with semantics.

[0109] In an embodiment of the present application, there may be irrelevant text information and noise that interferes with semantics in the text data. These noises may be caused by unclear text expression, grammatical errors, or spelling errors. Through semantic screening and noise filtering, these irrelevant information and noise can be removed to ensure the accuracy and consistency of the text data. Semantic screening can be performed based on natural language processing technology, such as word segmentation, part-of-speech tagging, named entity recognition, etc., to extract key information from the text. Noise filtering can be achieved through text cleaning algorithms, such as removing stop words, punctuation marks, and special characters.

[0110] Optionally, step 103 includes: extracting features from the cleaned curve data using the following three methods:

[0111] Extracting time domain features from the original curve, wherein the time domain features include at least: pulse duration, pulse energy, and pulse effective value;

[0112] In the embodiment of the present application, time domain feature extraction is a feature directly obtained from the original curve data, which reflects the characteristics of the curve in the time dimension. For curve data in the occupant protection restraint system data (such as collision pulse curve, dummy damage curve, etc.). Pulse duration refers to the time from the start to the end of the collision process, that is, the length of time from the starting point to the end point of the curve. This feature reflects the duration of the collision process and is of great significance for evaluating occupant injuries and restraint system performance. Pulse energy is the integral value of the pulse curve within the pulse duration, which represents the total energy transferred during the collision process. The larger the pulse energy, the higher the risk of injury to the occupants. The effective value of the pulse represents the average level of the pulse value during the entire duration, and is another important indicator for evaluating pulse intensity and injury potential.

[0113] Performing Fourier transform on the original curve, converting the original curve from the time domain to the frequency domain to extract frequency domain features, wherein the frequency domain features include at least: primary frequency amplitude, secondary frequency amplitude, frequency energy, frequency center, and spectrum entropy;

[0114] Frequency domain feature extraction is performed by converting the original curve from the time domain to the frequency domain, thereby obtaining the curve's characteristics in the frequency dimension. This conversion is typically achieved through a Fourier transform. Frequency domain feature extraction includes the following aspects: The primary frequency amplitude is the amplitude corresponding to the frequency with the highest amplitude in the spectrum, reflecting the main vibration component in the curve. The secondary frequency amplitude is the amplitude corresponding to the frequency with the second highest amplitude in the spectrum, excluding the primary frequency, providing more information about the curve's vibration components. Frequency energy is the sum of the energy of all frequency components in the spectrum, representing the total energy distribution of the curve in the frequency domain. The frequency center is the weighted average of all frequency components in the spectrum (weighted by the energy of each frequency component) and reflects the main frequency location of the curve's vibration. Spectral entropy is the randomness or uncertainty of the energy distribution of each frequency component in the spectrum, providing information about the complexity of the curve's vibration.

[0115] According to the waveform characteristics of the original curve, the original curve is fitted into a data function, and the waveform characteristics of the fitting function are extracted. The fitting types of the original curve fitting include: double trapezoidal wave, peak rectangular wave, and half sine wave;

[0116] In the embodiment of the present application, the fitting function feature extraction is to fit the original curve into a mathematical function and extract features from the fitting function. For curve data in the occupant protection restraint system data, common fitting types include double trapezoidal waves, peak rectangular waves, half-sine waves, etc. The fitting function feature extraction includes the following aspects: Double trapezoidal waves simplify the entire collision process into two trapezoidal waves, representing two collision stages. The algorithm automatically identifies the discontinuity position of the collision stage, and fits the pulse energy of each stage while keeping it consistent. The extracted features include the position of each corner point of the trapezoidal wave and the pulse value. The peak rectangular wave fits the complex peak fluctuation into a rectangular wave according to the energy consistency based on the peak distribution. The extracted features are each peak and the corresponding time. The half-sine wave simplifies the pulse into two segments of half-sine functions, identifies the dividing points of different stages through the algorithm, and extracts the amplitude, period and phase features of the two half-sine waves by fitting the function.

[0117] For text data after data cleaning, knowledge extraction, knowledge representation and data encoding are performed respectively to convert the text data features into vector features that can be recognized by the training model.

[0118] In the embodiments of the present application, cleaned text data (such as expert experience data on restraint system solution decision-making methods and selection conclusions) requires knowledge extraction, knowledge representation, and data encoding to convert it into vector features that can be recognized by the training model. Knowledge extraction involves extracting expert experience related to occupant restraint system solution selection from text data. This includes key information such as decision-making methods and selection conclusions. Natural language processing techniques such as word segmentation, part-of-speech tagging, and named entity recognition can be used to extract key information from the text. Knowledge representation involves normalizing the extracted expert experience so that the model can recognize and utilize it. Common knowledge representation methods include structured data formats (such as JSON and XML) or knowledge graphs. These methods can transform expert experience into a form that can be recognized by the model. Data encoding is the process of converting text data into vector features. This is typically achieved through text vectorization techniques such as bag-of-words models, TF-IDF, and word embeddings. These techniques convert text data into high-dimensional vectors, where each dimension represents the frequency or importance of a word or phrase. Using these vector features, the model can recognize and utilize the information in the text data.

[0119] Optionally, step 102 includes:

[0120] Step 1021 , normalizing each numerical data in the occupant protection and restraint system data;

[0121] In step 1022 , the pre-processed standardized features of each type of data are fused to form a multimodal feature dataset within a unified scale.

[0122] In the present embodiment, continuous numerical data such as airbag volume and pore size are standardized using a normalization method. Each parameter is normalized by subtracting its mean and dividing by its standard deviation. Refer to formula (1), where X is the raw data, μ is the mean, and σ is the standard deviation. The processed data has a mean of 0 and a variance of 1, which is close to a normal distribution and is suitable for scenarios requiring distance or similarity calculations.

[0123] Z=(X-μ) / σ (1)

[0124] The airbag volume range, pore size range, seatbelt force limiter range, and steering column crush travel range all have clear upper and lower bounds. A normalization method (Min-Max transformation) is used to linearly transform the raw data into the range [0, 1]. Refer to formula (2), where Xmin and Xmax are the minimum and maximum values ​​of the data, respectively, and Xnorm represents the normalized data. This method is simple and intuitive, and can maintain the distribution trend of the original data.

[0125] Xnorm=(X-Xmin) / (Xmax- Xmin) (2)

[0126] Finally, the standardized features of each data type are fused. This can be achieved through methods such as concatenation and weighted summation. The fused multimodal feature dataset will contain rich information from different data types, providing strong support for subsequent model training. The benefit of data fusion is that it can integrate complementary information from different data types, improving the model's predictive performance and stability. Furthermore, data fusion helps the model better understand and learn data features, resulting in more accurate predictions.

[0127] Optionally, step 104 includes:

[0128] Step 1041 , determining a selection target based on passenger types, where the passenger types include at least: driver, front passenger, rear passengers, and children.

[0129] In the embodiment of the present application, in the design of the occupant protection restraint system, it is first necessary to clarify the occupant type, because different types of occupants have different risks of injury and require different protection measures in a collision accident. Occupant types include at least: main driver, front passenger, rear passengers, and children. These types of occupants differ in terms of position, body size, physiological characteristics, etc. in the vehicle, so they need to be considered separately. According to the occupant type, the corresponding selection goal is determined. The selection goal refers to selecting the most suitable restraint system type and parameters for a specific occupant type to minimize injuries in a collision accident.

[0130] Step 1042, under the selected selection target, organize the independent data sets based on the working conditions; wherein the input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities, and the regression target quantities include at least: head injury, neck force, chest pressure displacement, and score.

[0131] In the examples of this application, data is organized based on operating conditions, within the selected model selection objective. Operating conditions refer to the specific state of the vehicle during a collision, such as collision speed, collision angle, and collision type. Different operating conditions can have different impacts on occupant injuries, so data needs to be organized based on these operating conditions.

[0132] Input features include crash pulse features and restraint system parameter features. The crash pulse features describe the mechanical properties during a collision, such as pulse duration and pulse energy. The restraint system parameter features describe the configuration and performance of the restraint system, such as seatbelt type, number of airbags, and seat position.

[0133] The output features include classification target quantity and regression target quantity.

[0134] Categorical targets may be used to determine the severity or type of occupant injury in a collision. Regression targets include at least head injury, neck force, chest pressure displacement, and scores. These values ​​are important indicators for assessing occupant injury severity and restraint system performance.

[0135] Based on the working condition criteria, the input and output features are combined into independent data sets. Each data set corresponds to a working condition and contains all relevant data and features under that working condition.

[0136] Step 1043: Based on independent data sets representing each operating condition, the multimodal feature data is trained using regression and classification prediction algorithms to perform an occupant protection injury prediction task for multiple target quantities. Each input feature represents a restraint system selection solution, and the classification target quantity and regression target quantity serve as a basis for screening restraint system selection solutions for reference in selecting vehicle occupant protection restraint system solutions.

[0137] In this embodiment, multimodal feature data is trained using regression and classification prediction algorithms based on independent datasets representing various operating conditions. The regression algorithm is used to predict continuous output feature values, such as head injury and neck force. The classification algorithm is used to determine the type and severity of occupant injury.

[0138] Each input feature represents a restraint system selection scheme. This means that for each restraint system configuration and parameter combination, a feature vector can be generated for all operating conditions. The classification target quantity and the regression target quantity serve as the basis for screening restraint system selection schemes. By comparing the target quantity values ​​of different selection schemes under different operating conditions, the optimal selection scheme can be selected.

[0139] Step 1044 , using a natural language processing model, combined with the vehicle occupant protection restraint system reference selection scheme and expert experience data characteristic parameters for training, to conduct decision analysis on the restraint system reference selection scheme, and output an optimal restraint system selection scheme that meets the restraint system score and design requirements.

[0140] In the embodiments of this application, a natural language processing model is used to process and analyze expert experience data. This expert experience data may include textual information such as the decision-making method and selection conclusions for the constraint system selection. The natural language processing model can convert this textual information into structured data, facilitating subsequent analysis and processing.

[0141] The training is conducted by combining the reference selection scheme of the vehicle occupant protection and restraint system with the characteristic parameters of expert experience data. Through the training model, the decision logic and empirical rules used by experts in the selection process can be learned, thus automating the selection decision.

[0142] Perform a decision analysis on the reference restraint system selection options and output the optimal restraint system selection option that meets both the restraint system score and design requirements. The decision analysis process may involve comparing and evaluating multiple selection options to select the optimal one. The output selection option should meet the restraint system score requirements (e.g., safety performance score) and design requirements (e.g., cost, feasibility, etc.).

[0143] Optionally, step 105 includes: performing finite element simulation or actual vehicle test verification using the screened optimal restraint system selection scheme that meets the restraint system score and design requirements, obtaining a new data set based on the collision result data and expert experience data generated during the simulation or test process, and updating and correcting the prediction model in combination with the new data set.

[0144] In the embodiments of this application, both finite element simulation and actual vehicle testing generate a large amount of collision result data. This collision result data includes occupant injury indicators (such as head injury, neck force, chest pressure displacement, etc.), restraint system response data (such as seat belt tension, airbag deployment, etc.), and vehicle collision dynamics data (such as collision speed, collision angle, etc.). Expert experience data is the conclusions and recommendations drawn from expert observations and analysis of collision tests. This data may include expert evaluations of selected options, interpretations of collision results, and predictions of future selection directions.

[0145] The collision results data and expert experience data are integrated to form a new data set. This data set includes the performance of the selected solutions in actual collisions as well as the evaluation and suggestions of experts, which is an important basis for subsequent model updates and revisions.

[0146] The initial prediction model was built based on theoretical assumptions and a limited dataset, potentially subject to bias and limitations. New datasets, obtained through finite element simulation and real-vehicle testing, can more accurately reflect the performance of the selected solution in actual crashes. Therefore, the prediction model needs to be updated and revised based on the new dataset to improve its accuracy and reliability.

[0147] Use new datasets to train and optimize prediction models. Machine learning algorithms and statistical analysis methods can be used to mine and analyze new datasets, extracting more valuable features and information. These features and information are then incorporated into the prediction model to update and revise it.

[0148] By updating and revising the model, we can improve the accuracy and reliability of the prediction model, providing stronger support for subsequent model selection decisions. At the same time, it can also promote the continuous advancement and development of occupant protection and restraint system technology, providing better protection for occupant safety.

[0149] Reference Figure 3 , which shows a flow chart of an intelligent selection system for an occupant protection and restraint system solution based on an AI model provided in an embodiment of the present application:

[0150] Among them, the main process includes active data collection, and the collected data includes but is not limited to R&D vehicle test data and simulation model data. These data can be obtained from the data update process, and the collected data can also include expert knowledge and historical cases obtained from the expert experience library.

[0151] After the collected data is standardized, it can be pre-processed through data cleaning, data conversion, data standardization, etc., and the cleaned data can be stored. The data cleaning process can also refer to the rules and strategies in the expert experience library.

[0152] During the training of the prediction model, the data update process can update the training dataset in real time. The training process adapts to the characteristics of various data types, enabling multimodal feature fusion for different modalities and selecting appropriate models and training methods. The training process can also reference rules and strategies from the expert experience library.

[0153] Once the prediction model meets the required accuracy, it can be used to predict the selection of vehicle occupant protection and restraint system solutions. During this process, expert feedback from the expert experience database can be used to constrain the model output boundary conditions, set goals, and evaluate the model output results. Furthermore, the data update process can iteratively train the prediction model using the updated data.

[0154] Traditional occupant protection and restraint system development faces numerous challenges, including insufficient data quality, difficulty processing multi-source heterogeneous data, and difficulty applying complex algorithms. The present application embodiments address these issues by introducing artificial intelligence (AI) models. First, by utilizing AI models such as data standardization, feature extraction, and multimodal feature fusion, the present application embodiments effectively process large amounts of inconsistent and missing data and fuse multi-source heterogeneous data, thereby improving data quality and consistency. This innovation significantly enhances the model's adaptability to data, making the development of occupant protection and restraint systems more reliable.

[0155] Secondly, the embodiments of the present application solve the problem of the model's lack of detachment from actual application scenarios by integrating expert knowledge into the AI ​​model. The effective integration of expert knowledge improves the system's decision-making accuracy and reliability, avoiding the risk of being out of touch with reality. In addition, the intelligent selection method adopted in the embodiments of the present application not only simplifies the complexity of algorithm processing, but also improves the generalization ability of the model, making the selection of occupant protection and restraint systems more efficient and accurate, thereby greatly improving the efficiency and accuracy of the research and development of automobile occupant protection and restraint systems, and bringing significant progress to the relevant technical fields.

[0156] Based on the same inventive concept, embodiments of the present application also provide an AI-based intelligent selection system for occupant protection and restraint systems, which is used to implement the aforementioned AI-based intelligent selection method for occupant protection and restraint systems. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more AI-based intelligent selection system embodiments provided below can be found in the limitations of the AI-based intelligent selection method for occupant protection and restraint systems described above and will not be further elaborated here.

[0157] In an exemplary embodiment, Figure 4 As shown, an AI model-based intelligent selection system 20 for an occupant protection and restraint system solution is provided, comprising:

[0158] An acquisition module 201 is configured to acquire occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data includes at least: test data, simulation model data, and expert experience data;

[0159] An extraction module 202 is configured to integrate the test data, simulation model data, and expert experience data into a multimodal input data set;

[0160] Performing feature extraction on the multimodal input data set using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input data set;

[0161] a training module 203 configured to output a reference selection scheme for a vehicle occupant protection and restraint system for vehicle design during a process of training prediction models for a plurality of different selection targets using the multimodal feature data, and to obtain an optimal restraint system selection scheme that meets design requirements from the plurality of reference selection schemes for the vehicle occupant protection and restraint system based on the expert experience data;

[0162] The updating module 204 is configured to update the prediction model based on simulation test results and expert experience data corresponding to the optimal constraint system selection solution.

[0163] Optionally, the acquisition module 201 is further configured to:

[0164] Acquire occupant protection restraint system data of multiple different modalities from multiple data sources; wherein both the test data and the simulation model data include: numerical data, which are restraint system parameters; collision results, which include at least: head displacement, head injury, neck force, chest pressure displacement and score; curve data, which include at least: collision pulse curve; text data, which include at least: restraint system solution decision-making method and selection conclusion;

[0165] The occupant protection restraint system data is cleaned by the following methods:

[0166] For the numerical data, identify and eliminate abnormal data, and perform denoising and smoothing correction processing;

[0167] Perform continuity and integrity checks on the curve data, traverse and eliminate incomplete data, and the incomplete data includes at least: data with obvious discontinuities and data that does not show the entire collision process;

[0168] The text data is subjected to semantic screening and noise filtering to remove irrelevant text information and noise that interferes with semantics.

[0169] Optionally, the extraction module 202 is further configured to:

[0170] For the curve data after data cleaning, the following three methods are used for feature extraction:

[0171] Extracting time domain features from the original curve, wherein the time domain features include at least: pulse duration, pulse energy, and pulse effective value;

[0172] Performing Fourier transform on the original curve, converting the original curve from the time domain to the frequency domain to extract frequency domain features, wherein the frequency domain features include at least: primary frequency amplitude, secondary frequency amplitude, frequency energy, frequency center, and spectrum entropy;

[0173] According to the waveform characteristics of the original curve, the original curve is fitted into a data function, and the waveform characteristics of the fitting function are extracted. The fitting types of the original curve fitting include: double trapezoidal wave, peak rectangular wave, and half sine wave;

[0174] For text data after data cleaning, knowledge extraction, knowledge representation and data encoding are performed respectively to convert the text data features into vector features that can be recognized by the training model.

[0175] Optionally, the extraction module 202 is further configured to:

[0176] performing normalization processing on each numerical data in the occupant protection restraint system data;

[0177] The standardized features of various types of preprocessed data are fused to form a multimodal feature dataset within a unified scale.

[0178] Optionally, the training module 203 is further configured to:

[0179] Determine the selection target based on the passenger type, where the passenger type includes at least: driver, co-driver, rear passenger, and child;

[0180] Under the selected selection target, independent data sets are collated based on the working conditions. The input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities. The regression target quantities include at least: head injury, neck force, chest pressure displacement, and score.

[0181] Based on independent data sets representing various operating conditions, the multimodal feature data is trained using regression and classification prediction algorithms to perform multiple target quantity occupant protection injury prediction tasks. Each input feature represents a restraint system selection scheme, and the classification target quantity and regression target quantity serve as the basis for screening restraint system selection schemes for reference selection of vehicle occupant protection restraint system schemes.

[0182] A natural language processing model is used to train the reference selection scheme of the vehicle occupant protection restraint system in combination with the characteristic parameters of expert experience data to perform decision analysis on the reference selection scheme of the restraint system and output the optimal restraint system selection scheme that meets the restraint system score and design requirements.

[0183] Optionally, the updating module 204 is further configured to:

[0184] After finite element simulation or actual vehicle test verification is performed using the screened optimal restraint system selection scheme that meets the restraint system score and design requirements, a new data set is obtained based on the collision result data and expert experience data generated during the simulation or test process, and the prediction model is updated and corrected in combination with the new data set.

[0185] The embodiment of the present application obtains multimodal sample feature data by combining occupant protection restraint system data of different modalities with expert experience data, so that multiple models can learn richer features and more complex correlation relationships. This training method not only improves the accuracy of the model, but also enables the model to adapt to multiple task objectives. The prediction model can quickly respond to the relevant parameter input of the vehicle design plan and output the optimal occupant protection restraint system selection plan based on these parameters, realizing the full process automation from data collection to plan output, greatly improving the efficiency and accuracy of vehicle occupant protection restraint system selection.

[0186] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store intelligent selection data of occupant protection restraint system solutions based on AI models. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements an intelligent selection method for occupant protection restraint system solutions based on AI models.

[0187] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0188] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0189] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0190] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0192] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0193] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0194] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0195] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An AI-based intelligent selection method for occupant protection and restraint system solutions, characterized by: The AI ​​model-based intelligent selection method for occupant protection and restraint system solutions includes: Acquiring occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data includes at least: test data, simulation model data, and expert experience data; Integrating the test data, simulation model data, and expert experience data into a multimodal input data set; Performing feature extraction on the multimodal input data set using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input data set; In the process of training the prediction models for multiple different selection targets using the multimodal feature data, a reference selection scheme for a vehicle occupant protection and restraint system is output for vehicle design, and an optimal restraint system selection scheme that meets the design requirements is obtained from the multiple reference selection schemes for the vehicle occupant protection and restraint system based on the expert experience data; Updating the prediction model based on simulation test results and expert experience data corresponding to the optimal restraint system selection scheme; The steps of outputting a reference selection scheme for a vehicle occupant protection and restraint system for vehicle design during the process of training the prediction models for multiple different selection targets using the multimodal feature data, and obtaining an optimal restraint system selection scheme that meets the design requirements from the multiple reference selection schemes for the vehicle occupant protection and restraint system based on the expert experience data include: Determine the selection target based on the passenger type, where the passenger type includes at least: driver, co-driver, rear passenger, and child; Under the selected selection target, independent data sets are collated based on the working conditions. The input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities. The regression target quantities include at least: head injury, neck force, chest pressure displacement, and score. Based on independent data sets representing various operating conditions, the multimodal feature data is trained for an occupant protection injury prediction task using multiple target quantities using regression and classification prediction algorithms; wherein each input feature represents a restraint system selection scheme, and the classification target quantity and the regression target quantity serve as a basis for screening restraint system selection schemes for reference selection of vehicle occupant protection restraint system schemes; A natural language processing model is used to train the reference selection scheme of the vehicle occupant protection restraint system in combination with the characteristic parameters of expert experience data to perform decision analysis on the reference selection scheme of the restraint system and output the optimal restraint system selection scheme that meets the restraint system score and design requirements.

2. The AI ​​model-based intelligent selection method for occupant protection and restraint system solutions according to claim 1 is characterized in that: The step of obtaining occupant protection restraint system data of multiple modes includes: Acquire occupant protection restraint system data of multiple different modalities from multiple data sources; wherein both the test data and the simulation model data include: numerical data, which are restraint system parameters, including collision results, which include at least: head displacement, head injury, neck force, chest pressure displacement and score; curve data, which include at least: collision pulse curve; text data, which include at least: restraint system solution decision-making method and selection conclusion; The occupant protection restraint system data is cleaned by the following methods: For the numerical data, identify and eliminate abnormal data, and perform denoising and smoothing correction processing; Perform continuity and integrity checks on the curve data, traverse and eliminate incomplete data, and the incomplete data includes at least: data with obvious discontinuities and data that does not show the entire collision process; The text data is subjected to semantic screening and noise filtering to remove irrelevant text information and noise that interferes with semantics.

3. The AI ​​model-based intelligent selection method for occupant protection and restraint system solutions according to claim 2 is characterized in that: The step of extracting features from the multimodal input dataset using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input dataset comprises: For the curve data after data cleaning, the following three methods are used for feature extraction: Extracting time domain features from the original curve, wherein the time domain features include at least: pulse duration, pulse energy, and pulse effective value; Performing Fourier transform on the original curve, converting the original curve from the time domain to the frequency domain to extract frequency domain features, wherein the frequency domain features include at least: primary frequency amplitude, secondary frequency amplitude, frequency energy, frequency center, and spectrum entropy; According to the waveform characteristics of the original curve, the original curve is fitted into a data function, and the waveform characteristics of the fitting function are extracted. The fitting types of the original curve fitting include: double trapezoidal wave, peak rectangular wave, and half sine wave; For text data after data cleaning, knowledge extraction, knowledge representation and data encoding are performed respectively to convert the text data features into vector features that can be recognized by the training model.

4. The AI ​​model-based intelligent selection method for occupant protection and restraint system solutions according to claim 2 is characterized in that: The step of integrating the test data, simulation model data and expert experience data into a multimodal input data set includes: performing normalization processing on each numerical data in the occupant protection restraint system data; The standardized features of various types of preprocessed data are fused to form a multimodal feature dataset within a unified scale.

5. The AI ​​model-based intelligent selection method for occupant protection and restraint system solutions according to claim 1 is characterized in that: The step of updating the prediction model based on the simulation test results corresponding to the optimal constraint system selection scheme and expert experience data includes: After finite element simulation or actual vehicle test verification is performed using the screened optimal restraint system selection scheme that meets the restraint system score and design requirements, a new data set is obtained based on the collision result data and expert experience data generated during the simulation or test process, and the prediction model is updated and corrected in combination with the new data set.

6. An AI-based intelligent selection system for occupant protection and restraint systems, characterized by: The AI ​​model-based intelligent selection system for occupant protection and restraint systems includes: an acquisition module, configured to acquire occupant protection and restraint system data of multiple modes, wherein the occupant protection and restraint system data includes at least: test data, simulation model data, and expert experience data; An extraction module, configured to integrate the test data, simulation model data, and expert experience data into a multimodal input data set; Performing feature extraction on the multimodal input data set using feature extraction models corresponding to a plurality of different modalities to obtain multimodal feature data of the multimodal input data set; a training module for outputting a reference selection scheme for a vehicle occupant protection restraint system for vehicle design during a process of training a plurality of prediction models for different selection targets using the multimodal feature data, and obtaining an optimal restraint system selection scheme that meets design requirements from the plurality of reference selection schemes for the vehicle occupant protection restraint system based on the expert experience data; An updating module, configured to update the prediction model based on simulation test results corresponding to the optimal constraint system selection scheme and expert experience data; The training module is further used to determine the selection target according to the passenger type, and the passenger type includes at least: driver, co-driver, rear passenger, and child; Under the selected selection target, independent data sets are collated based on the working conditions. The input features include collision pulse features and restraint system parameter features, and the output features include classification target quantities and regression target quantities. The regression target quantities include at least: head injury, neck force, chest pressure displacement, and score. Based on independent data sets representing various operating conditions, the multimodal feature data is trained for an occupant protection injury prediction task using multiple target quantities using regression and classification prediction algorithms; wherein each input feature represents a restraint system selection scheme, and the classification target quantity and the regression target quantity serve as a basis for screening restraint system selection schemes for reference selection of vehicle occupant protection restraint system schemes; A natural language processing model is used to train the reference selection scheme of the vehicle occupant protection restraint system in combination with the characteristic parameters of expert experience data to perform decision analysis on the reference selection scheme of the restraint system and output the optimal restraint system selection scheme that meets the restraint system score and design requirements.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the intelligent selection method for an occupant protection restraint system solution based on an AI model according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent selection method for an occupant protection restraint system solution based on an AI model according to any one of claims 1 to 5 are implemented.

9. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the steps of the intelligent selection method of an occupant protection restraint system solution based on an AI model according to any one of claims 1 to 5.

Images