Body armor damage analysis method and system based on improved particle swarm optimization
By improving the particle swarm optimization algorithm to optimize sensor deployment and signal processing, the shortcomings of traditional bulletproof vest damage assessment methods have been addressed, enabling accurate damage assessment in complex battlefield environments and improving the reliability and accuracy of signal acquisition and assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG SCI-TECH UNIV
- Filing Date
- 2025-12-15
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional methods for assessing damage to body armor fail to adequately consider the dynamic stress response characteristics of body armor under different threats, resulting in insufficient signal capture capabilities of sensor networks and an inability to achieve accurate positioning and assessment in battlefield environments.
An improved particle swarm optimization algorithm was used to optimize the sensor deployment location. Combined with finite element simulation and deep neural network, signal preprocessing and impact point localization were performed, local shock wave signals were extracted, and multi-task damage assessment was conducted.
It improves the reliability of impact signal acquisition and the accuracy of damage assessment, significantly enhances the signal-to-noise ratio and the purity of damage characteristic signals, and achieves efficient damage mode classification and degree quantification.
Smart Images
Figure CN121835249A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of body armor, in particular to a body armor damage analysis method and system based on an improved particle swarm algorithm. BACKGROUND
[0002] With the development of individual protection equipment technology, it has become a key requirement to improve equipment support efficiency and combatant survivability to quickly and accurately evaluate the damage state of body armor after impact in actual combat.
[0003] The traditional body armor damage evaluation method has obvious shortcomings: Existing methods are mostly based on empirical rules and do not fully consider the dynamic stress response characteristics of body armor under different threats and different positions impacted. This non-optimized layout not only leads to insufficient ability of the sensor network to capture key signals, but also may cause waste of sensor resources or coverage blind spots, and cannot provide high-quality raw data basis for subsequent analysis.
[0004] In the battlefield environment, there are complex background noise and electromagnetic interference, and the traditional fixed parameter signal processing method cannot effectively extract the damage feature signal, resulting in the inability to accurately locate and evaluate the damage.
[0005] Therefore, the present application provides a body armor damage analysis method and system based on an improved particle swarm algorithm, which can accurately collect impact signals and accurately evaluate damage assessment when the body armor is impacted, greatly improving the reliability of impact signals and the high accuracy of damage assessment. SUMMARY
[0006] To solve the above technical problems, the purpose of the present application is to provide a body armor damage analysis method and system based on an improved particle swarm algorithm, which can accurately collect impact signals and accurately evaluate damage assessment when the body armor is impacted, greatly improving the reliability of impact signals and the high accuracy of damage assessment.
[0007] To achieve the above purpose, the present application provides the following technical solution: a body armor damage analysis method based on an improved particle swarm algorithm, comprising: Determine the layout position of the sensor on the body armor based on finite element simulation and improved genetic algorithm optimization, and collect the original impact signal generated by the impact; Perform dynamic band-pass filtering, variable step size adaptive noise reduction, and abnormal signal rejection on the original impact signal in sequence, and perform signal enhancement based on deep neural network decomposition and APFI double threshold screening to obtain standardized original impact signals; Based on the time difference of the shock wave reaching different sensors, the planar coordinates of the impact point are calculated using the time difference positioning method and the improved particle swarm optimization algorithm, and the statistical characteristics of the particle swarm convergence distribution are output. Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, signal segments within a preset dynamic range of the impact point are extracted from the standardized original impact signal and used as local shock wave signals. Based on local shock wave signals, multi-task damage assessment is performed, and damage mode classification and damage degree quantification results are output.
[0008] Preferably, the step of determining the sensor placement position on the bulletproof vest based on finite element simulation and improved genetic algorithm optimization includes: Parametric finite element modeling of bulletproof vests was performed to simulate the impact on pre-defined core and auxiliary deployment areas under various typical threats, and dynamic stress / strain spatiotemporal cloud maps during the impact process were extracted and analyzed. Based on dynamic stress / strain spatiotemporal cloud maps, the initial density distribution function of sensors in the core deployment area is constructed through gridded analysis and feature extraction. Based on the initial density distribution function of the sensors and the preset total number of sensors, the positions of the sensors in the core deployment area are iteratively optimized using an improved genetic algorithm, and the optimal set of sensor position coordinates in the core deployment area is output. Based on the optimal set of position coordinates of the sensors in the core deployment area, sparse preset and logical association are performed in the auxiliary deployment area to form the set of position coordinates of the auxiliary deployment area. Based on the optimal set of location coordinates of the sensors in the core area and the set of location coordinates of the auxiliary deployment area, sensors are deployed in the core deployment area and the auxiliary deployment area.
[0009] Preferably, the process of sequentially performing dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, and then enhancing the signal based on deep neural network decomposition and APFI dual threshold screening to obtain a standardized original impact signal includes: Based on the power spectral density of the battlefield background noise of the original impact signal, the passband frequency of the bandpass filter is dynamically calibrated, and based on the improved variable step size adaptive noise reduction algorithm, the step size update factor is adjusted by dual objective feedback of signal-to-noise ratio and mean square error to obtain the noise-reduced target signal. For the denoised target signal, a regionalized anomaly signal is initially screened based on the improved 3σ criterion of sliding window scaling; and a spike interference is accurately removed by an improved soft threshold function to obtain a signal without abnormal interference. The signal without abnormal interference is decomposed using a deep neural network decomposition method to obtain multiple intrinsic mode function components; the intrinsic mode function components are screened and reconstructed based on the APFI dual threshold of the fusion of average envelope entropy and impulse factor, and then nonlinear features are extracted to obtain the enhanced damage feature signal; The enhanced damage feature signal is subjected to standard normalization processing, and a standardized damage feature signal is obtained based on a dynamic smoothing optimization mechanism, which serves as the preprocessed original impact signal.
[0010] Preferably, the step of calculating the planar coordinates of the impact point based on the time difference of the shock wave reaching different sensors, using the time difference positioning method and an improved particle swarm optimization algorithm, and outputting the statistical characteristics of the particle swarm convergence distribution includes: Based on the timestamp data obtained after preprocessing the raw impact signals collected by sensors deployed on the bulletproof vest, combined with the sensor clock offset calibration results, a time preprocessing weighting operation is performed to obtain a high-reliability time series. Based on highly reliable time series, TDOA filtering and core weighting operations are performed to obtain a set of effective TDOAs with precision weights. Based on the effective TDOA set, the PSO iterative optimization weighted operation is performed to obtain the candidate coordinates of the impact point and the statistical characteristics of the particle swarm convergence distribution during the optimization process. Based on the candidate coordinates of the impact point and the preset verification index, a weighted inverse optimization operation is performed to obtain the planar coordinates of the impact point.
[0011] Preferably, the step of extracting signal segments within a preset dynamic range of the impact point from the standardized original impact signal as a local shock wave signal based on the plane coordinates of the impact point and the statistical characteristics of the particle swarm convergence distribution includes: Based on the plane coordinates of the impact point, the sensor with the smallest Euclidean distance from the plane coordinates of the impact point is determined from all sensor nodes and used as the reference sensor. Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, the location signal region of the impact point is calculated, and then the basic spatial radius for signal extraction is obtained. Based on the original impact signal, the peak amplitude and main frequency characteristics of the signal corresponding to the reference sensor are extracted, and the characteristic radius of effective signal propagation is calculated. Based on the basic spatial radius and feature radius, and based on the preset fusion decision rules, the final extraction radius is dynamically adjusted. Centered on the plane coordinates of the impact point, and with the final extraction radius as the domain, signals corresponding to the sensor nodes are extracted from the original impact signal to form a preliminary local signal set; Based on the average cross-correlation coefficient between signals in the preliminary local signal set, the extraction quality is verified. If the verification passes, the signal set is output as a local shock wave signal. If the verification fails, the final extraction radius is expanded based on preset rules and the extraction and verification are repeated until the verification passes or the maximum number of attempts is reached.
[0012] Preferably, the multi-task damage assessment based on local shock wave signals, and the output of damage pattern classification and damage degree quantification results, include: Based on the local shock wave signal, the corresponding time-domain features, frequency-domain features, and time-frequency features are extracted simultaneously to form time-domain feature vectors, frequency-domain feature vectors, and time-frequency feature vectors, respectively. The time-domain feature vector, frequency-domain feature vector, and time-frequency feature vector are concatenated to obtain the original high-dimensional fused feature vector. Based on the original high-dimensional fusion feature vector, feature dimensionality reduction and enhancement are performed on the original high-dimensional fusion feature vector to obtain a low-dimensional sensitive damage feature vector. Based on low-dimensional sensitive damage feature vectors and a pre-defined multi-task damage assessment model, the system outputs damage pattern classification results and damage degree quantification results.
[0013] Preferably, the original impact signal is a stress wave / strain wave signal generated inside the bulletproof vest when it is impacted.
[0014] A second aspect of the present invention also provides a bulletproof vest damage analysis system based on an improved particle swarm optimization algorithm, comprising: The data acquisition module was deployed, and the placement of the sensors on the bulletproof vest was determined based on finite element simulation and an improved genetic algorithm, and the raw impact signals generated by the impact were collected. The preprocessing module sequentially performs dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, and enhances the signal based on deep neural network decomposition and APFI dual threshold screening to obtain a standardized original impact signal. The solution module calculates the planar coordinates of the impact point based on the time difference of the shock wave reaching different sensors, using the time difference positioning method and an improved particle swarm optimization algorithm, and outputs the statistical characteristics of the particle swarm convergence distribution. The extraction module extracts signal segments within a preset dynamic range of the impact point from the standardized original impact signal, based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, as a local shock wave signal. The assessment module performs multi-task damage assessment based on local shock wave signals, and outputs damage pattern classification and damage degree quantification results.
[0015] Compared with the prior art, the beneficial effects of the present invention are: In the signal acquisition and preprocessing stage, this invention overcomes the shortcomings of traditional empirical sensor placement and fixed parameter processing methods. It obtains the dynamic mechanical response field of the bulletproof vest through finite element simulation and utilizes an improved genetic algorithm to achieve the optimal layout of the sensor network, ensuring accurate acquisition of impact signals. In signal preprocessing, a multi-level adaptive processing chain is employed, including dynamically adjusted filtering, variable step-size noise reduction, and intelligent signal enhancement, significantly improving the signal-to-noise ratio and purity of the damage characteristic signals, providing high-quality input for subsequent analysis.
[0016] In the impact location and signal extraction stages, this invention overcomes the limitation of traditional location methods that only provide coordinate estimation. While using an improved particle swarm optimization algorithm to calculate the impact point coordinates, it simultaneously extracts statistical features representing the algorithm's convergence process as a confidence index for the location results. Based on the coordinate and confidence information, and combined with the shock wave propagation law, the optimal signal extraction radius is dynamically calculated, automatically selecting a set of local signals related to the impact core. This effectively isolates far-field noise and irrelevant interference, providing highly correlated signals for damage assessment.
[0017] In the damage assessment stage, this invention solves the problem of separating damage pattern classification and severity assessment in traditional methods. For the high-quality local signal after focusing, multi-dimensional features are extracted simultaneously and input into a multi-task deep learning model, achieving integrated assessment of accurate damage pattern classification and quantitative regression of damage severity. This assessment framework not only improves analytical efficiency but also ensures consistency between qualitative judgment and quantitative analysis results in terms of physical mechanisms, significantly improving the accuracy of damage assessment. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0019] Figure 1 This is a schematic diagram of a bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm.
[0020] Figure 2 This is a schematic diagram of a bulletproof vest damage analysis system based on an improved particle swarm optimization algorithm. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0023] Example 1 like Figure 1 As shown, this embodiment discloses a bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm, including: The placement of sensors on the bulletproof vest is determined based on finite element simulation and improved genetic algorithm optimization, and the original impact signal generated by the impact is collected. In this embodiment, the original impact signal is the stress wave / strain wave signal generated inside the bulletproof vest when it is impacted.
[0024] It should be noted that the determination of the sensor placement on the bulletproof vest based on finite element simulation and improved genetic algorithm optimization includes: Parametric finite element modeling of bulletproof vests was performed to simulate the impact on pre-defined core and auxiliary deployment areas under various typical threats, and dynamic stress / strain spatiotemporal cloud maps during the impact process were extracted and analyzed. Based on dynamic stress / strain spatiotemporal cloud maps, the initial density distribution function of sensors in the core deployment area is constructed through gridded analysis and feature extraction. Based on the initial density distribution function of the sensors and the preset total number of sensors, the positions of the sensors in the core deployment area are iteratively optimized using an improved genetic algorithm, and the optimal set of sensor position coordinates in the core deployment area is output. Based on the optimal set of position coordinates of the sensors in the core deployment area, sparse preset and logical association are performed in the auxiliary deployment area to form the set of position coordinates of the auxiliary deployment area. Based on the optimal set of location coordinates of the sensors in the core area and the set of location coordinates of the auxiliary deployment area, sensors are deployed in the core deployment area and the auxiliary deployment area.
[0025] In detail, based on ballistics, traumatology, and historical data, the anterior chest region (heart and major blood vessels), the upper back region (spine), and the areas corresponding to the liver / spleen on both sides of the ribs are defined as typical high-threat areas. These areas are of paramount importance and require high-precision monitoring, thus forming the core deployment areas. Areas with relatively low probability of impact or less lethality, such as the shoulders, outer abdomen, and lumbar region, provide redundancy and range extension for the monitoring network, corresponding to auxiliary deployment areas. The typical threats include bullets of different calibers, different incident angles, and fragments of different velocities. During the extraction and analysis of the dynamic stress / strain spatiotemporal cloud map, the focus is on the propagation path, energy concentration area, and peak distribution of stress waves.
[0026] Specifically, in the cloud map analysis and feature extraction, for each grid cell, the peak value of the maximum equivalent stress and the duration of high stress in the simulation are extracted. The stress gradient field of the entire core region is calculated. Regions with drastic stress changes (large gradients) require denser sensors to capture rapid changes in the signal. A preliminary sensor density distribution function is then defined. In the formula, These are the sensor's position coordinates. For in position The maximum equivalent stress (or first principal stress) value that appears at all time steps in the finite element simulation. For in position At that point, the rate of change (magnitude) of the peak stress field in space. This indicates normalization to the [0,1] interval. and It is the weighting coefficient.
[0027] It should be noted that, in this embodiment, the goal of the improved genetic algorithm is to achieve the desired result given the total number of sensors (or budget). In this case, find from a large number of candidate locations in the core area The optimal location is determined to maximize the overall performance of the monitoring network. Specifically, this involves the two-dimensional coordinates of a sensor. A chromosome, or gene, represents a complete genetic pattern. It consists of N genes linked together sequentially. .
[0028] Randomly generate a containing Individuals (i.e.) The initial population (based on a predefined overall fitness function) is used to calculate the fitness of each individual. A roulette wheel selection or tournament selection method is employed to increase the probability of highly fit individuals being selected for the mating pool to become parents of the next generation. Two parent individuals are randomly selected from the mating pool, and crossover is performed with a certain probability. The newly generated offspring are then mutated with a lower probability. The newly generated offspring replace some of the less fit parents, forming a new generation. This process is repeated until a predefined maximum number of iterations is reached, or the optimal fitness no longer significantly improves over multiple generations. At this point, the placement scheme represented by the individual with the highest fitness in the population is the optimized sensor location.
[0029] In the auxiliary area, low-power, wake-up-enabled sensor nodes are pre-embedded at a distance much greater than the average spacing in the core area. An activation mode is set for each auxiliary node. Typically, it is associated with the spatially nearest core area sensor. All core area sensors operate continuously. When a core area sensor detects a signal exceeding a preset threshold (indicating a valid impact), it immediately wirelessly wakes up 1-2 associated auxiliary nodes, putting them into operation to jointly record the subsequent waveforms and propagation of the impact. After the impact event ends, these auxiliary nodes can enter sleep mode again to conserve energy.
[0030] The original impact signal was subjected to dynamic bandpass filtering, variable step size adaptive noise reduction, and abnormal signal removal in sequence. Then, signal enhancement was performed based on deep neural network decomposition and APFI dual threshold screening to obtain a standardized original impact signal. It should be noted that the process of sequentially performing dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, followed by signal enhancement based on deep neural network decomposition and APFI dual threshold screening, yields the standardized original impact signal, including: Based on the power spectral density of the battlefield background noise of the original impact signal, the passband frequency of the bandpass filter is dynamically calibrated, and based on the improved variable step size adaptive noise reduction algorithm, the step size update factor is adjusted by dual objective feedback of signal-to-noise ratio and mean square error to obtain the noise-reduced target signal. In this embodiment, background noise in the battlefield environment from 100Hz to 20kHz is collected in real time. The main interference frequency band is located by power spectral density analysis, and the passband of the bandpass filter is dynamically calibrated to match the characteristic frequency range of the impact damage signal. Of course, the noise range corresponding to the background noise can be adjusted.
[0031] For the denoised target signal, an improved method based on sliding window scaling is applied. The criteria are used for initial screening of regional abnormal signals; an improved soft threshold function is used to accurately remove spike interference and obtain signals without abnormal interference. In this embodiment, the improvement of the sliding window scaling The criteria for initial regional screening of anomalous signals are as follows: The signal frequency is divided into three window scales: low-frequency, mid-frequency, and high-frequency. Corresponding sliding window parameters are configured for each scale domain. The principle is that the slower the signal change, the longer the observation window; the faster the change, the shorter the window, to match the dynamic characteristics of the signal in that frequency band. For example, a 200ms sliding window is used for the low-frequency band below 50Hz in the low-frequency scale domain, a 100ms sliding window for the mid-frequency band from 50-1kHz in the mid-frequency scale domain, and a 50ms sliding window for the high-frequency band above 1kHz in the high-frequency scale domain. The signal mean and standard deviation are calculated for each window, and deviations are considered... Signal segments were marked as anomalies. Finally, the number of particles and the maximum number of iterations were set for the multi-objective PSO algorithm. The fitness function was constructed with the dual objectives of maximizing the signal-to-noise ratio and optimizing the signal feature retention. The optimal threshold of the improved soft threshold function was obtained by optimization.
[0032] For the signal without abnormal interference, a deep neural network decomposition method is used to decompose it, obtaining multiple intrinsic mode function components; based on the fusion of average envelope entropy and impulse factor... The intrinsic mode function components are screened and reconstructed using a dual threshold method, and then nonlinear features are extracted to obtain the enhanced damage feature signal. In detail, the deep neural network adopts a convolutional network structure adapted to the signal decomposition requirements. The input is a signal segment after anomaly removal, and the output is multiple sets of signals after local feature extraction by the convolutional kernel. Weight, suppressing tradition Mode aliasing defects in scenarios where impact signals and interference signals are superimposed; the aforementioned Dual threshold screening specifically involves calculating each The correlation coefficient and average envelope entropy between the component and the standard impact signal are used to screen for components that meet the dual threshold criteria. Component reconstruction; processing of the reconstructed signal The processing extracts core nonlinear features, improving the ability to distinguish between damaged and non-target signals.
[0033] The enhanced damage feature signal is subjected to standard normalization processing, and a standardized damage feature signal is obtained based on a dynamic smoothing optimization mechanism, which serves as the preprocessed original impact signal. Specifically, the adaptive window dynamic smoothing mechanism uses the local gradient of the signal as an adaptive factor. When the absolute value of the gradient is greater than or equal to the corresponding threshold (signal change is drastic), a first preset point window is used; when the absolute value of the gradient is less than the corresponding threshold (signal is stable), a second preset point window is used, where the first preset point is less than the second preset point.
[0034] Based on the time difference of the shock wave reaching different sensors, the planar coordinates of the impact point are calculated using the time difference positioning method and the improved particle swarm optimization algorithm, and the statistical characteristics of the particle swarm convergence distribution are output. It should be noted that the calculation of the impact point's planar coordinates and the output of the particle swarm convergence distribution statistical characteristics based on the time difference of the shock wave reaching different sensors, using the time difference positioning method and an improved particle swarm optimization algorithm, includes: Based on the timestamp data obtained after preprocessing the raw impact signals collected by sensors deployed on the bulletproof vest, and combined with the sensor clock offset calibration results, a time preprocessing weighting operation is performed to obtain a high-reliability time series. In this embodiment, specifically: a reference node is selected and its weight is set to 1, and the remaining nodes are based on the clock offset error relative to the reference node. The weights are calculated using the inverse variance weighting method, and the formula is as follows: Complete clock synchronization calibration. Then, based on... The criteria and median filtering are used to remove outliers. Time data with a confidence level greater than or equal to a preset threshold are retained and given high weight, while low-confidence data are removed. Finally, the calibrated and anomaly-free reception times of each node are obtained. ; Based on highly reliable time series, TDOA filtering and core weighting operations are performed to obtain a set of effective TDOAs with precision weights. In this embodiment, a full TDOA matrix is constructed. , The diagonal is 0. In the formula, The sensor nodes are deployed on the same plane. Let be the planar coordinates of the i-th sensor node. The velocity of the shock wave in the medium is known or obtained through calibration. , where is the coordinate of the impact point.
[0035] Effective filtering is performed based on filtering rules, for example Corresponding distance difference Need to meet Assign weights to each valid TDOA. It is inversely proportional to the measurement variance, forming an effective TDOA set: .
[0036] Based on the effective set of TDOAs, a PSO iterative optimization weighting operation is performed to obtain the candidate coordinates of the impact points and the statistical characteristics of the particle swarm convergence distribution during the optimization process; in this embodiment, the core weights of TDOAs are... The PSO fitness function is embedded, with the goal of minimizing the fitness value; the PSO parameter weights are dynamically adapted based on the iterative stage. After iterative updates by the particle swarm optimization, the candidate coordinates of the impact point are finally obtained.
[0037] Based on the candidate coordinates of the impact point and a preset verification metric, a weighted inverse optimization operation is performed to obtain the planar coordinates of the impact point. In this embodiment, the verification metrics include residuals, relative errors, and convergence speed. For example, if the residual is too large, the weight of the high-confidence TDOA is increased. Reduce the weight of low-confidence TDOA.
[0038] Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, signal segments within a preset dynamic range of the impact point are extracted from the preprocessed original impact signal and used as local shock wave signals. It should be noted that the extraction of signal segments within a preset dynamic range of the impact point from the standardized original impact signal, based on the plane coordinates of the impact point and the statistical characteristics of the particle swarm convergence distribution, as a local shock wave signal includes: Based on the plane coordinates of the impact point, the sensor with the smallest Euclidean distance from the plane coordinates of the impact point is determined from all sensor nodes and used as the reference sensor. Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, the location signal region of the impact point is calculated, and then the basic spatial radius for signal extraction is obtained. Based on the original impact signal, the peak amplitude and main frequency characteristics of the signal corresponding to the reference sensor are extracted, and the characteristic radius of effective signal propagation is calculated. Based on the basic spatial radius and feature radius, and based on the preset fusion decision rules, the final extraction radius is dynamically adjusted. Centered on the plane coordinates of the impact point, and with the final extraction radius as the domain, signals corresponding to the sensor nodes are extracted from the original impact signal to form a preliminary local signal set; Based on the average cross-correlation coefficient between signals in the preliminary local signal set, the extraction quality is verified. If the verification passes, the signal set is output as a local shock wave signal. If the verification fails, the final extraction radius is expanded based on preset rules and the extraction and verification are repeated until the verification passes or the maximum number of attempts is reached.
[0039] In detail, the impact point ( , ) to each sensor node ( , The Euclidean distance is: The feature extraction of the particle swarm convergence distribution statistical characteristics involves the corresponding sample mean and covariance matrix. Based on the covariance matrix, eigenvalue decomposition is performed to construct a confidence ellipse. To ensure complete coverage of the confidence ellipse, its circumcircle radius is calculated as the basic spatial radius. Using a pre-defined impact energy-spatial attenuation empirical model, combined with the signal peak amplitude and dominant frequency characteristics, the corresponding effective signal attenuation radius is calculated. and basic spatial radius The final extraction radius is obtained. The impact energy-spatial attenuation empirical model is derived through fitting analysis based on the energy spatial attenuation data of the enhanced damage characteristic signals recorded during the parameter finite element modeling and impact simulation at different impact locations. The preset rule is extended to include the final extracted radius. Updated to , If the value is greater than 1, the data is then extracted and verified again. The average cross-correlation coefficient is calculated based on the core intrinsic mode function components reconstructed after deep neural network decomposition and APFI dual threshold screening, corresponding to the standardized damage feature signal.
[0040] Based on local shock wave signals, multi-task damage assessment is performed, and damage mode classification and damage degree quantification results are output.
[0041] It should be noted that the multi-task damage assessment based on local shock wave signals, which outputs damage mode classification and damage degree quantification results, includes: Based on the local shock wave signal, the corresponding time-domain features, frequency-domain features, and time-frequency features are extracted simultaneously to form time-domain feature vectors, frequency-domain feature vectors, and time-frequency feature vectors, respectively. The time-domain feature vector, frequency-domain feature vector, and time-frequency feature vector are concatenated to obtain the original high-dimensional fused feature vector. Based on the original high-dimensional fusion feature vector, feature dimensionality reduction and enhancement are performed on the original high-dimensional fusion feature vector to obtain a low-dimensional sensitive damage feature vector. Based on low-dimensional sensitive damage feature vectors and a pre-defined multi-task damage assessment model, the system outputs damage pattern classification results and damage degree quantification results.
[0042] Specifically, the time-domain feature vector is calculated by taking the peak amplitude, rise time, pulse width, total signal energy, and root mean square of the local shock wave signal. The frequency-domain feature vector is calculated by performing a Fast Fourier Transform on the local shock wave signal to extract the dominant frequency, centroid frequency, frequency variance, and energy proportion of a specific frequency band. The time-frequency feature vector is calculated by performing a continuous wavelet transform on the local shock wave signal to generate a time-frequency energy distribution map, and extracting the singular entropy, energy concentration, and time-frequency ridge slope of the time-frequency distribution.
[0043] A pre-trained sparse autoencoder is used to perform nonlinear dimensionality reduction and feature enhancement on the original high-dimensional fused feature vector, achieving dimensionality reduction. The multi-task damage assessment model adopts an existing multi-task deep neural network assessment model. The first output branch performs damage pattern classification, outputting the probability distribution of whether the impact event belongs to back-side deformation, matrix cracking, fiber breakage, or delamination. The second output branch performs damage degree quantification, outputting estimates of the predicted back-side indentation depth, equivalent diameter of the damaged area, and residual strength reduction factor caused by the impact. Based on the probability distribution of the damage pattern classification, the estimated damage degree quantification values are weighted and corrected to obtain the final damage degree quantification result.
[0044] like Figure 2 As shown, this embodiment also discloses a bulletproof vest damage analysis system based on an improved particle swarm optimization algorithm, including: The data acquisition module was deployed, and the placement of the sensors on the bulletproof vest was determined based on finite element simulation and an improved genetic algorithm, and the raw impact signals generated by the impact were collected. The preprocessing module sequentially performs dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, and enhances the signal based on deep neural network decomposition and APFI dual threshold screening to obtain a standardized original impact signal. The solution module calculates the planar coordinates of the impact point based on the time difference of the shock wave reaching different sensors, using the time difference positioning method and an improved particle swarm optimization algorithm, and outputs the statistical characteristics of the particle swarm convergence distribution. The extraction module extracts signal segments within a preset dynamic range of the impact point from the standardized original impact signal, based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, as a local shock wave signal. The assessment module performs multi-task damage assessment based on local shock wave signals, and outputs damage pattern classification and damage degree quantification results.
[0045] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0046] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0047] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.
[0048] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0049] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.
[0050] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0051] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0052] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for analyzing bulletproof vest damage based on an improved particle swarm optimization algorithm, characterized in that, The method includes: The placement of sensors on the bulletproof vest was determined based on finite element simulation and improved genetic algorithm optimization, and the original impact signal generated by the impact was collected. The original impact signal is sequentially subjected to dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal, and then decomposed and analyzed based on a deep neural network. Signal enhancement is performed using dual threshold screening to obtain a standardized original impact signal; Based on the time difference of the shock wave reaching different sensors, the planar coordinates of the impact point are calculated using the time difference positioning method and the improved particle swarm optimization algorithm, and the statistical characteristics of the particle swarm convergence distribution are output. Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, signal segments within a preset dynamic range of the impact point are extracted from the standardized original impact signal and used as local shock wave signals. Based on local shock wave signals, multi-task damage assessment is performed, and damage mode classification and damage degree quantification results are output.
2. The bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The method of determining the sensor placement on the bulletproof vest based on finite element simulation and improved genetic algorithm optimization includes: Parametric finite element modeling of bulletproof vests was performed to simulate the impact on pre-defined core and auxiliary deployment areas under various typical threats, and dynamic stress / strain spatiotemporal cloud maps during the impact process were extracted and analyzed. Based on dynamic stress / strain spatiotemporal cloud maps, the initial density distribution function of sensors in the core deployment area is constructed through gridded analysis and feature extraction. Based on the initial density distribution function of the sensors and the preset total number of sensors, the positions of the sensors in the core deployment area are iteratively optimized using an improved genetic algorithm, and the optimal set of sensor position coordinates in the core deployment area is output. Based on the optimal set of position coordinates of the sensors in the core deployment area, sparse preset and logical association are performed in the auxiliary deployment area to form the set of position coordinates of the auxiliary deployment area. Based on the optimal set of location coordinates of the sensors in the core area and the set of location coordinates of the auxiliary deployment area, sensors are deployed in the core deployment area and the auxiliary deployment area.
3. The bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm according to claim 2, characterized in that, The process involves sequentially performing dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, followed by signal enhancement based on deep neural network decomposition and APFI dual threshold screening, to obtain a standardized original impact signal, including: Based on the power spectral density of the battlefield background noise of the original impact signal, the passband frequency of the bandpass filter is dynamically calibrated, and based on the improved variable step size adaptive noise reduction algorithm, the step size update factor is adjusted by dual objective feedback of signal-to-noise ratio and mean square error to obtain the noise-reduced target signal. For the denoised target signal, an improved method based on sliding window scaling is applied. The criteria are used for initial screening of regional abnormal signals; an improved soft threshold function is used to accurately remove spike interference and obtain signals without abnormal interference. The signal without abnormal interference is decomposed using a deep neural network decomposition method to obtain multiple intrinsic mode function components; the intrinsic mode function components are screened and reconstructed based on the APFI dual threshold of the fusion of average envelope entropy and impulse factor, and then nonlinear features are extracted to obtain the enhanced damage feature signal; The enhanced damage feature signal is subjected to standard normalization processing, and a standardized damage feature signal is obtained based on a dynamic smoothing optimization mechanism, which serves as the preprocessed original impact signal.
4. The bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm according to claim 3, characterized in that, The method, based on the time difference of the shock wave reaching different sensors, and using a time-difference positioning method and an improved particle swarm optimization algorithm, calculates the planar coordinates of the impact point and outputs the statistical characteristics of the particle swarm convergence distribution, including: Based on the timestamp data obtained after preprocessing the raw impact signals collected by sensors deployed on the bulletproof vest, combined with the sensor clock offset calibration results, a time preprocessing weighting operation is performed to obtain a high-reliability time series. Based on highly reliable time series, TDOA filtering and core weighting operations are performed to obtain a set of effective TDOAs with precision weights. Based on the effective TDOA set, the PSO iterative optimization weighted operation is performed to obtain the candidate coordinates of the impact point and the statistical characteristics of the particle swarm convergence distribution during the optimization process. Based on the candidate coordinates of the impact point and the preset verification index, a weighted inverse optimization operation is performed to obtain the planar coordinates of the impact point.
5. The bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm according to claim 4, characterized in that, The step of extracting signal segments within a preset dynamic range of the impact point from the preprocessed original impact signal, based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, as local shock wave signals includes: Based on the plane coordinates of the impact point, the sensor with the smallest Euclidean distance from the plane coordinates of the impact point is determined from all sensor nodes and used as the reference sensor. Based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, the location signal region of the impact point is calculated, and then the basic spatial radius for signal extraction is obtained. Based on the original impact signal, the peak amplitude and main frequency characteristics of the signal corresponding to the reference sensor are extracted, and the characteristic radius of effective signal propagation is calculated. Based on the basic spatial radius and feature radius, and based on the preset fusion decision rules, the final extraction radius is dynamically adjusted. Centered on the plane coordinates of the impact point, and with the final extraction radius as the domain, signals corresponding to the sensor nodes are extracted from the original impact signal to form a preliminary local signal set; Based on the average cross-correlation coefficient between signals in the preliminary local signal set, the extraction quality is verified. If the verification passes, the signal set is output as a local shock wave signal. If the verification fails, the final extraction radius is expanded based on preset rules and the extraction and verification are repeated until the verification passes or the maximum number of attempts is reached.
6. The bulletproof vest damage analysis method based on the improved particle swarm optimization algorithm according to claim 5, characterized in that, The multi-task damage assessment based on local shock wave signals, which outputs damage pattern classification and damage degree quantification results, includes: Based on the local shock wave signal, the corresponding time-domain features, frequency-domain features, and time-frequency features are extracted simultaneously to form time-domain feature vectors, frequency-domain feature vectors, and time-frequency feature vectors, respectively. The time-domain feature vector, frequency-domain feature vector, and time-frequency feature vector are concatenated to obtain the original high-dimensional fused feature vector. Based on the original high-dimensional fusion feature vector, feature dimensionality reduction and enhancement are performed on the original high-dimensional fusion feature vector to obtain a low-dimensional sensitive damage feature vector. Based on low-dimensional sensitive damage feature vectors and a pre-defined multi-task damage assessment model, the system outputs damage pattern classification results and damage degree quantification results.
7. The bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm according to claim 6, characterized in that, The original impact signal is the stress wave / strain wave signal generated inside the bulletproof vest when it is impacted.
8. A bulletproof vest damage analysis system based on an improved particle swarm optimization algorithm, implementing the bulletproof vest damage analysis method based on an improved particle swarm optimization algorithm as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module was deployed, and the placement of the sensors on the bulletproof vest was determined based on finite element simulation and an improved genetic algorithm, and the raw impact signals generated by the impact were collected. The preprocessing module sequentially performs dynamic bandpass filtering, variable step-size adaptive noise reduction, and abnormal signal removal on the original impact signal, and enhances the signal based on deep neural network decomposition and APFI dual threshold screening to obtain a standardized original impact signal. The solution module calculates the planar coordinates of the impact point based on the time difference of the shock wave reaching different sensors, using the time difference positioning method and an improved particle swarm optimization algorithm, and outputs the statistical characteristics of the particle swarm convergence distribution. The extraction module extracts signal segments within a preset dynamic range of the impact point from the standardized original impact signal, based on the plane coordinates of the impact point and the statistical characteristics of the convergence distribution of the particle swarm, as a local shock wave signal. The assessment module performs multi-task damage assessment based on local shock wave signals, and outputs damage pattern classification and damage degree quantification results.