Sensor-based anti-pressure detection method and system for intangible cultural heritage wine packaging box

Abstract

The application relates to the technical field of packaging detection, and discloses a sensor-based anti-pressure detection method and system for non-heritage wine packaging boxes. The method collects surface pressure distribution data and internal stress waveforms of the packaging box through a sensor network to generate a multimodal sensor data stream. A three-dimensional digital twin model of the packaging box is constructed based on the data stream, and material elastic parameters and structure constraint conditions are initialized. An adaptive anti-pressure prediction model is generated by performing anti-pressure performance prediction training on the model, and a deformation simulation sequence is output in a virtual pressure environment. The deformation error coefficient is calculated by dynamically comparing real-time sensor data with the simulation sequence, the pressure detection device trigger threshold is adjusted according to the deformation error coefficient, and the anti-pressure detection cycle is started. In the cycle process, multi-source data are fused for spatiotemporal feature analysis, and the overall anti-pressure strength of the packaging box is evaluated. The application realizes accurate prediction and dynamic detection of the anti-pressure performance of the packaging box.

Description

Technical Field

[0001] This invention relates to the field of packaging inspection technology, specifically to a sensor-based method and system for testing the compressive strength of intangible cultural heritage wine packaging boxes. Background Technology

[0002] The packaging boxes of intangible cultural heritage wines often possess unique cultural attributes and complex structural designs, and their physical compressive strength directly affects the safety of the product during storage and transportation. Existing packaging box compressive strength testing technologies mainly rely on two types of methods. One type is destructive testing using physical pressure testing machines. This method applies continuously increasing pressure to the packaging box sample through a mechanical device until deformation or damage occurs, recording the maximum load-bearing value. While the evaluation results of this method are intuitive, it is a sampling inspection and cannot perform non-destructive evaluation on every packaging box. Furthermore, the testing process is time-consuming and cannot reflect the true response under dynamic pressure. The other type is simulation technology based on computer-aided engineering. This involves establishing a geometric model of the packaging box and assigning uniform material parameters for static mechanical analysis. This non-destructive method relies on idealized model assumptions, and its material parameters often use standard reference values, failing to fully consider the subtle differences in material properties between specific batches and the nonlinear characteristics present in the actual structure. There is a discrepancy between the simulation results and the actual behavior of the physical entity, resulting in limited prediction accuracy.

[0003] The main shortcomings of existing technical solutions lie in the static nature of the testing process and the idealization of the model. Physical testing methods cannot achieve comprehensive and non-destructive quality monitoring, while simulation technology struggles to provide high-precision predictions due to the disconnect between the model and the physical object. Conventional technologies lack a mechanism for deeply coupling sensor data from the real physical world with virtual models, and also lack the ability to dynamically optimize the testing process based on real-time monitoring data. Therefore, current testing methods are insufficient to meet the needs of efficiently, accurately, and non-destructively assessing the compressive strength of intangible cultural heritage wine packaging boxes, which possess both cultural value and practical function. The core problem this invention aims to solve is how to improve the consistency between the compressive strength testing model and the physical entity, and how to achieve adaptive adjustment of the testing threshold to adapt to individual differences and dynamic pressure environments. Summary of the Invention

[0004] The purpose of this invention is to provide a sensor-based method and system for testing the compressive strength of intangible cultural heritage wine packaging boxes, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes, the method comprising:

[0006] The surface pressure distribution data and internal stress waveform of the packaging box of intangible cultural heritage wines are collected by a high-precision sensor network to generate a multimodal sensing data stream;

[0007] Based on the multimodal sensing data stream, a three-dimensional digital twin model of the packaging box is constructed, and the material elastic parameters and structural constraints in the model are initialized.

[0008] The three-dimensional digital twin model is trained to predict its compressive performance, thereby generating an adaptive compressive performance prediction model.

[0009] The adaptive stress prediction model is run in a virtual stress environment, and the deformation simulation sequence of the packaging box is output.

[0010] Real-time acquisition of actual sensor monitoring data, dynamic comparison with the deformation simulation sequence, and calculation of deformation error coefficient;

[0011] The trigger threshold of the pressure detection device is adjusted according to the deformation error coefficient, and the pressure detection cycle is started based on the adjusted trigger threshold;

[0012] Multi-source data is integrated during the compression testing cycle, and the overall compression strength of the packaging box is evaluated through spatiotemporal feature analysis.

[0013] Preferably, the step of generating the multimodal sensing data stream includes:

[0014] Deploy a pressure sensor array at key points on the surface of the packaging box to collect continuous pressure values ​​and generate a pressure distribution map;

[0015] Strain sensors are installed at structural nodes inside the packaging box to record stress change waveforms and generate strain time series data.

[0016] Simultaneously acquire material micro-fracture signals captured by acoustic sensors to generate acoustic emission data sequences;

[0017] The pressure distribution spectrum, strain time series data, and acoustic emission data sequence are time-aligned and format-normalized to generate a multimodal sensing data stream.

[0018] Preferably, the steps for constructing the three-dimensional digital twin model include:

[0019] Analyze the geometric feature points in the multimodal sensing data stream to extract the length, width, height, and corner angles of the packaging box;

[0020] Based on the extracted dimensions and angles, a meshed 3D model of the packaging box is constructed using the finite element method;

[0021] The material elastic modulus and Poisson's ratio are derived from the multimodal sensing data stream, and the material properties of the three-dimensional model are initialized.

[0022] Set the support boundaries and load conditions of the packaging box to complete the parameter initialization of the three-dimensional digital twin model.

[0023] Preferably, the generation step of the adaptive stress resistance prediction model includes:

[0024] Prepare a historical compression test dataset, including packaging box deformation records and final compression strength labels;

[0025] Design a deep neural network architecture that includes convolutional layers for spatial feature extraction and recurrent layers for temporal dependency modeling;

[0026] The deep neural network was trained using a historical stress resistance test dataset, and the network weights and bias parameters were optimized.

[0027] Verify the prediction accuracy of the trained network on the test set and generate an adaptive stress-resistance prediction model.

[0028] Preferably, the output step of the deformation simulation sequence includes:

[0029] A virtual pressure gradient is applied to the three-dimensional digital twin model to simulate the loading process from zero to a preset maximum pressure;

[0030] Run the adaptive compressive strength prediction model to calculate the deformation and stress distribution of the packaging box under each pressure gradient;

[0031] Record the deformation curve over time to generate a deformation simulation sequence;

[0032] The output deformation simulation sequence includes the deformation peak, deformation rate, and deformation stability point.

[0033] Preferably, the calculation steps for the deformation error coefficient include:

[0034] Real-time acquisition of pressure and deformation values ​​from actual sensor monitoring data to generate actual deformation sequences;

[0035] Extract the simulated deformation values ​​at the corresponding time points from the deformation simulation sequence;

[0036] Calculate the absolute difference between the actual deformation value and the simulated deformation value, and normalize it to obtain the deformation deviation value;

[0037] The average and variance of the deformation deviation values ​​throughout the entire testing period are statistically analyzed to generate the deformation error coefficient.

[0038] Preferably, the trigger threshold adjustment step of the pressure detection device includes:

[0039] Based on the magnitude of the deformation error coefficient, the sensitivity level range is divided;

[0040] Query the preset mapping table between sensitivity levels and trigger thresholds to obtain the target trigger threshold;

[0041] Send control commands to the pressure detection device to adjust the sampling frequency and trigger threshold of its pressure sensor;

[0042] Confirm that the adjusted trigger threshold matches the current detection environment, and complete the threshold adaptation.

[0043] Preferably, the initiation step of the compression testing cycle includes:

[0044] Initialize the pressure application device, and set the pressure loading rate and maximum pressure value;

[0045] Based on the adjusted trigger threshold, the pressure application device is controlled to gradually increase the pressure;

[0046] Real-time monitoring of sensor data; when the pressure value reaches the trigger threshold, record the current deformation data.

[0047] The process of applying pressure and recording data is repeated until the preset pressure range is covered.

[0048] Preferably, the evaluation steps for the overall compressive strength include:

[0049] Multi-source data features were extracted during the testing process, including pressure distribution uniformity, deformation recovery rate, and stress concentration index.

[0050] A feature fusion algorithm is used to merge features from multiple sources into a comprehensive feature vector.

[0051] The comprehensive feature vector is input into the pre-trained stress resistance regression model to calculate the stress resistance score;

[0052] The quality grade of the packaging box is determined based on the compressive strength score, and a compressive performance evaluation report is generated.

[0053] Preferably, the present invention also includes a sensor-based intangible cultural heritage wine packaging box compression resistance detection system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-described sensor-based intangible cultural heritage wine packaging box compression resistance detection method.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] By acquiring surface pressure distribution and internal stress waveforms of packaging boxes through a high-precision sensor network, a multimodal sensor data stream is generated. The resulting 3D digital twin model more accurately reflects the actual material properties and structural characteristics of a specific packaging box. Unlike conventional techniques that use standard parameters for simulation, this method initializes the model with real data, making the elastic parameters and constraints in the model closer to the physical entity. This model construction method based on actual sensor data significantly improves the fidelity of the virtual simulation environment, allowing subsequent compression performance prediction training to be based on a more realistic physical foundation. The output results of the deformation simulation sequence therefore have higher reliability, reducing prediction errors caused by idealized assumptions in the model.

[0056] Real-time acquisition of actual sensor monitoring data and dynamic comparison with deformation simulation sequences are used to calculate the deformation error coefficient, providing a quantitative indicator for evaluating the model's prediction accuracy. Based on this coefficient, the trigger threshold of the pressure detection device is adjusted, and a pressure resistance detection loop is initiated, achieving adaptive optimization of the detection process. Conventional technologies typically use fixed, preset detection thresholds that cannot be adjusted based on the actual performance of individual packaging boxes or environmental changes. This solution, through a dynamic feedback mechanism, enables the detection system to perceive and respond to the difference between model predictions and actual behavior, automatically calibrating the detection sensitivity. This closed-loop adjustment mechanism enhances the tolerance for subtle differences among individual packaging boxes, allowing pressure resistance detection to not only rely on initial model predictions but also continuously learn and correct during the detection process, improving the intelligence and accuracy of the overall evaluation process. Attached Figure Description

[0057] Figure 1 This is a schematic diagram illustrating the working principle of the sensor-based compression resistance testing method for intangible cultural heritage wine packaging boxes described in this invention.

[0058] Figure 2 A flowchart for generating multimodal sensing data streams;

[0059] Figure 3 A flowchart for constructing a 3D digital twin model;

[0060] Figure 4 This is a histogram of the deformation error coefficient distribution.

[0061] Figure 5 This is a graph showing the relationship between loading rate and maximum pressure for different materials. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Please see Figure 1 This invention provides a sensor-based method and system for testing the compressive strength of intangible cultural heritage wine packaging boxes. The method uses a high-precision sensor network to collect surface pressure distribution data and internal stress waveforms of the packaging box, generating a multimodal sensing data stream. Based on this data stream, a three-dimensional digital twin model of the packaging box is constructed, and material elastic parameters and structural constraints are initialized. The three-dimensional digital twin model is trained to predict compressive performance to generate an adaptive compressive strength prediction model. The model is run in a virtual pressure environment to output a deformation simulation sequence. Real-time acquisition of actual sensor monitoring data and dynamic comparison with the deformation simulation sequence are performed to calculate the deformation error coefficient. The trigger threshold of the pressure detection device is adjusted according to the deformation error coefficient, and the compressive strength detection loop is started. During the loop, multi-source data is fused, and the overall compressive strength of the packaging box is evaluated through spatiotemporal feature analysis.

[0064] Example 1: See Figure 2 The generation of multimodal sensing data stream is achieved by deploying a pressure sensor array at key points on the surface of the packaging box to collect continuous pressure values ​​and generate a pressure distribution map. At the same time, strain sensors are installed at structural nodes inside the packaging box to record stress change waveforms and generate strain time series data. Simultaneously, material micro-fracture signals captured by acoustic sensors are collected to generate acoustic emission data sequences. The pressure distribution map, strain time series data, and acoustic emission data sequences are time-aligned and format-standardized to generate a multimodal sensing data stream.

[0065] In practical implementation, the generation process of multimodal sensing data stream involves the collaborative work of multiple sensor types. The initial step is to deploy a pressure sensor array at key points on the surface of the intangible cultural heritage wine packaging box. The pressure sensor array is arranged in a grid on the outer surface of the packaging box, such as the center and edge areas of the lid, bottom, and side walls. Continuous pressure values ​​are collected and converted into digital signals via a data acquisition card, generating a pressure distribution map. The pressure distribution map represents the pressure intensity at each point in a two-dimensional matrix. Strain sensors are installed at internal structural nodes of the packaging box, such as internal support frames or connections to cushioning materials. These strain sensors, using resistance strain gauges or fiber optic gratings, record stress change waveforms and generate strain time-series data. This data includes timestamps and micro-strain values, reflecting the internal deformation dynamics of the packaging box under pressure. Simultaneously, acoustic sensors capture micro-fracture signals from the material. These piezoelectric accelerometers are placed on the surface of the packaging box or inside near vulnerable areas, generating acoustic emission data sequences. These sequences include sound wave amplitude, frequency, and event counts, used to detect early material damage.

[0066] In practice, the acoustic emission signals captured by the acoustic sensors need to be processed to identify micro-fractures in the material. Amplitude thresholds and duration thresholds are set. For example, burst signals with an amplitude exceeding 60 dB and a duration of less than 100 microseconds are identified as valid micro-fracture signals in the material, while continuous signals with lower amplitude and longer duration are regarded as environmental noise and filtered out.

[0067] In some embodiments, the sampling frequency of the pressure sensor array is set to 1000 Hz to ensure the capture of rapidly changing pressure signals. The strain sensor's measurement range covers strain values ​​from zero to the yield limit of the packaging material, and the acoustic sensor's sensitivity is adjusted to detect minute acoustic emission events. Optionally, the time alignment process uses a global positioning system timestamp or internal clock synchronization mechanism to ensure that the time axes of the pressure distribution map, strain time series data, and acoustic emission data sequence are consistent. Format standardization converts different types of data into a unified JSON or CSV format, including a data header, time field, and numerical field, generating a multimodal sensing data stream. It is understood that the multimodal sensing data stream serves as the basis for subsequent processing, and its integrity depends on the accurate calibration of the sensors and the real-time nature of data acquisition.

[0068] In specific implementations, the processing of pressure distribution maps, strain time-series data, and acoustic emission data sequences also includes data preprocessing steps, such as noise and outlier removal. The pressure distribution map is smoothed for high-frequency interference using a filtering algorithm; the strain time-series data is processed using a moving average method to reduce random fluctuations; and the acoustic emission data sequences are filtered to remove background noise. In some embodiments, the generation of multimodal sensing data streams can incorporate a data fusion formula, such as using a weighted average method to integrate data from different sensors. The formula is expressed as: ,in: This represents the merged data stream. This represents pressure distribution map data. Represents strain time series data, Represents the acoustic emission data sequence. , and These are the weighting coefficients for the corresponding data, preset based on sensor accuracy and importance. Optionally, time alignment uses interpolation to ensure data point alignment, and format standardization includes data compression to reduce storage space. It is understood that the generation of multimodal sensing data streams ensures the accuracy of subsequent 3D digital twin model construction.

[0069] Example 2: See Figure 3 The construction of a 3D digital twin model involves analyzing geometric feature points in the multimodal sensing data stream, extracting the length, width, height, and corner angles of the packaging box, and constructing a meshed 3D model of the packaging box using the finite element method based on the extracted dimensions and angles. The material's elastic modulus and Poisson's ratio are derived from the multimodal sensing data stream, the material properties of the 3D model are initialized, and the support boundaries and load conditions of the packaging box are set, completing the parameter initialization of the 3D digital twin model. For the generation of the adaptive compressive strength prediction model, a historical compressive strength test dataset is prepared, including packaging box deformation records and final compressive strength labels. A deep neural network architecture is designed, including convolutional layers for spatial feature extraction and recurrent layers for temporal dependency modeling. The deep neural network is trained using the historical compressive strength test dataset, the network weights and bias parameters are optimized, and the prediction accuracy of the trained network on the test set is verified, generating the adaptive compressive strength prediction model. The output of the deformation simulation sequence applies a virtual pressure gradient to the three-dimensional digital twin model to simulate the loading process from zero to the preset maximum pressure. It runs an adaptive compressive strength prediction model to calculate the deformation and stress distribution of the packaging box under each pressure gradient, records the deformation change curve over time, and generates a deformation simulation sequence. The output deformation simulation sequence includes the deformation peak, deformation rate, and deformation stability point.

[0070] In practice, the construction of the 3D digital twin model begins with analyzing the geometric feature points in the multimodal sensing data stream, extracting the length, width, height, and corner angles of the packaging box. These geometric feature points are identified from the multimodal sensing data stream using point cloud data processing algorithms. The length, width, and height are obtained by calculating the point cloud bounding box, and the corner angles are determined by the angle between the normal vectors of adjacent point clouds. Based on the extracted dimensions and angles, a meshed 3D model of the packaging box is constructed using the finite element method. The finite element method discretizes the geometry of the packaging box into tetrahedral or hexahedral mesh elements. The 3D model is represented by node coordinates and element connection relationships. The material's elastic modulus and Poisson's ratio are derived from the multimodal sensing data stream. The elastic modulus is obtained by analyzing the slope of the stress-strain curve, and the Poisson's ratio is derived by calculating the ratio of transverse strain to longitudinal strain. The material properties of the 3D model are initialized, and the elastic modulus and Poisson's ratio are assigned to each mesh element. Support boundaries and load conditions for the packaging box are set. The support boundaries define the fixed constraints at the bottom of the packaging box, and the load conditions specify the pressure applied to the lid surface, completing the parameter initialization of the 3D digital twin model. In some embodiments, the analysis of geometric feature points uses an iterative nearest-point algorithm for point cloud registration, and the calculation of edge angles uses principal component analysis to determine the local coordinate system. Optionally, the finite element mesh size is adaptively adjusted according to the packaging box size, and the mesh refinement is set based on the expected accuracy. It can be understood that the parameter initialization of the 3D digital twin model is the foundation for subsequent simulations.

[0071] In specific implementation, the adaptive compression resistance prediction model is generated by preparing a historical compression resistance testing dataset, including packaging box deformation records and final compression strength labels. The historical compression resistance testing dataset contains multiple packaging box test cases; the deformation records are time-series displacement data, and the compression strength labels are experimentally measured failure pressure values. A deep neural network architecture is designed, including convolutional layers for spatial feature extraction and recurrent layers for temporal dependency modeling. The convolutional layers process two-dimensional data such as pressure distribution maps, and the recurrent layers use a long short-term memory network to process strain time-series data. The deep neural network is trained using the historical compression resistance testing dataset, optimizing network weights and bias parameters. The training process employs a gradient descent algorithm, and the loss function is used to measure prediction error. The prediction accuracy of the trained network on the test set is verified, generating the adaptive compression resistance prediction model. In some embodiments, the convolutional layers of the deep neural network use the ReLU activation function, and the recurrent layers are set to have 64 hidden units. Optionally, an early stopping strategy is used during training to prevent overfitting, and the accuracy is verified using root mean square error (RMSE). It is understood that the generation of the adaptive compression resistance prediction model depends on the quality of the historical data and the design of the network structure. The formula for calculating the loss function during training is expressed as:

[0072]

[0073] in: Indicates the loss value. Indicates the number of training samples. This represents the true compressive strength label of the j-th sample. This represents the predicted compressive strength value of the j-th sample.

[0074] In practical implementation, the output of the deformation simulation sequence applies a virtual pressure gradient to the 3D digital twin model, simulating the loading process from zero to a preset maximum pressure. The virtual pressure gradient is applied in a linearly increasing manner, and the preset maximum pressure is set according to the strength of the packaging box material. An adaptive compressive strength prediction model is run to calculate the deformation and stress distribution of the packaging box under each pressure gradient. The deformation is inferred from the model to output nodal displacements, and the stress distribution is calculated using Hooke's law to calculate the element stress. The deformation change curve over time is recorded to generate a deformation simulation sequence, which stores time points and corresponding deformation values ​​in array form. The output deformation simulation sequence includes the deformation peak, deformation rate, and deformation stability point. The deformation peak is the maximum value in the sequence, the deformation rate is calculated using difference to determine the instantaneous rate of change, and the deformation stability point is defined as the moment when the rate of change of deformation is below a threshold. It can be understood that the output of the deformation simulation sequence provides a reference benchmark for actual testing.

[0075] Example 3: Calculation of Deformation Error Coefficient. Real-time acquisition of pressure and deformation values ​​from actual sensor monitoring data generates an actual deformation sequence. Simulated deformation values ​​at corresponding time points are extracted from the deformation simulation sequence. The absolute difference between the actual and simulated deformation values ​​is calculated and normalized to obtain the deformation deviation value. The average and variance of the deformation deviation values ​​over the entire detection cycle are statistically analyzed to generate the deformation error coefficient. Adjustment of the Trigger Threshold of the Pressure Detection Device. Based on the magnitude of the deformation error coefficient, sensitivity level ranges are divided. A preset mapping table of sensitivity levels and trigger thresholds is consulted to obtain the target trigger threshold. A control command is sent to the pressure detection device to adjust the sampling frequency and trigger threshold of its pressure sensor. The adjusted trigger threshold is confirmed to match the current detection environment, completing threshold self-adaptation.

[0076] In practice, the calculation of the deformation error coefficient begins with the real-time acquisition of pressure and deformation values ​​from actual sensor monitoring data to generate an actual deformation sequence. The actual sensor monitoring data comes from pressure and displacement sensors installed on the packaging box. Pressure values ​​are recorded in Pascals, and deformation values ​​are recorded in millimeters. The actual deformation sequence is a data list containing timestamps, pressure values, and corresponding deformation values. Next, simulated deformation values ​​at corresponding time points are extracted from the deformation simulation sequence, which is output by an adaptive pressure prediction model. The simulated deformation value is the predicted deformation calculated by the model under a specific virtual pressure. The absolute difference between the actual and simulated deformation values ​​is calculated and normalized to obtain the deformation deviation value. The absolute difference is the absolute value obtained by subtracting the simulated deformation value from the actual deformation value. Normalization involves dividing the absolute difference by the range of simulated deformation values ​​to scale the deformation deviation value to between 0 and 1. Finally, the average and variance of the deformation deviation values ​​over the entire detection period are statistically analyzed to generate the deformation error coefficient. The deformation error coefficient is a comprehensive scalar used to quantify the overall difference between simulated and actual deformation.

[0077] In some embodiments, the sampling interval of the actual deformation sequence is set to 10 milliseconds to ensure accurate correspondence with the time points of the deformation simulation sequence, and the normalization process uses a minimum-maximum normalization method. Optionally, the deformation error coefficient can be calculated using a weighted average method, giving greater weight to the later stages of the detection cycle. Deformation error coefficient The calculation formula can be expressed as:

[0078]

[0079] in: Indicates the deformation error coefficient. This represents the total number of sampling points throughout the entire detection cycle. This represents the normalized deformation deviation value at the i-th sampling point. Indicates all The average value. It can be understood that the calculation of the deformation error coefficient provides a quantitative basis for adjusting the trigger threshold.

[0080] In practical implementation, the trigger threshold adjustment of the pressure detection device is based on the magnitude of the deformation error coefficient, dividing the sensitivity level range into high, medium, and low intervals, each corresponding to a range of deformation error coefficient values. A preset mapping table of sensitivity levels and trigger thresholds is consulted to obtain the target trigger threshold. The mapping table is stored in key-value pairs, where the key is the sensitivity level and the value is the corresponding trigger pressure threshold. Control commands are sent to the pressure detection device to adjust the sampling frequency and trigger threshold of its pressure sensor. The control commands are sent via serial port or network protocol. The sampling frequency determines the data acquisition speed, and the trigger threshold is the critical pressure value for initiating data recording. The adjusted trigger threshold is confirmed to match the current detection environment, completing threshold adaptation. The confirmation process includes applying test pressure and verifying whether data recording is triggered at the correct threshold point. In some embodiments, the sensitivity level range is set based on historical calibration data; for example, a deformation error coefficient less than 0.05 is considered a low sensitivity range. Optionally, the mapping relationship of the trigger threshold can be linear or non-linear. It can be understood that the trigger threshold adjustment of the pressure detection device achieves adaptive optimization of the detection system.

[0081] See Figure 4 This diagram illustrates the calculation process of the deformation error coefficient, clarifying that it is a quantitative indicator generated by comparing the actual deformation sequence with the simulated deformation sequence. The data in the diagram represents the statistical distribution of deformation deviation values ​​within the detection cycle, centrally reflecting the frequency characteristics of the deformation error coefficient. Its core interval corresponds to the basis for dividing the sensitivity level intervals. This distribution result is the core reference for trigger threshold adjustment: based on the concentration of the error coefficient, a preset sensitivity level and trigger threshold mapping table can be matched, thereby adjusting the sampling frequency and trigger threshold of the pressure detection device. This threshold adaptation based on error distribution allows the detection system to adapt to the individual differences of different packaging boxes, avoiding detection deviations caused by fixed thresholds. It provides quantitative support for the accurate initiation of the pressure testing cycle and is a key link in improving the intelligence and accuracy of the detection process.

[0082] Example 4: The pressure application device is initialized for the start of the pressure resistance test cycle. The pressure loading rate and maximum pressure value are set. Based on the adjusted trigger threshold, the pressure application device is controlled to gradually increase the pressure. Sensor data is monitored in real time. When the pressure value reaches the trigger threshold, the current deformation data is recorded. Pressure loading and data recording are performed in a loop until the preset pressure range is covered.

[0083] In practical implementation, the pressure application device is initialized during the start-up process of the pressure resistance testing cycle. The pressure loading rate and maximum pressure value are set. The pressure application device adopts a servo-controlled hydraulic or electric actuator. The pressure loading rate is set to increase by 50 Newtons to 500 Newtons per second. The maximum pressure value is set according to the nominal compressive strength of the intangible cultural heritage wine packaging box, for example, 1000 Newtons to 5000 Newtons. Based on the adjusted trigger threshold, the pressure application device is controlled to gradually increase the pressure. The adjusted trigger threshold is derived from the adaptive result after the deformation error coefficient is calculated. The control logic uses the trigger threshold as the step benchmark for pressure increase. Sensor data is monitored in real time. When the pressure value reaches the trigger threshold, the current deformation data is recorded. The sensor data includes pressure sensor readings and displacement sensor readings. The deformation data record includes a timestamp, pressure value, and corresponding deformation. The pressure loading and data recording are executed cyclically until the preset pressure range is covered. The pressure is gradually increased at the trigger threshold interval until the pressure reaches the preset maximum pressure value.

[0084] In some embodiments, the choice of pressure loading rate depends on the material properties of the packaging box; a higher loading rate is used for materials with higher rigidity. The current pressure value during pressure loading can be calculated using the following formula: ,in: This indicates the pressure value at the current moment. This indicates the pressure value at the previous recorded moment. This indicates the set pressure loading rate. This indicates the time interval for data recording. It's understandable that accurate pressure calculation is fundamental to cyclic control.

[0085] Optionally, the pressure loading rate and maximum pressure value can be set by referring to a pre-stored material parameter library, using different parameter combinations for packaging boxes of different materials and structures. Refer to Table 1, which shows one setting of the relationship between pressure loading rate and maximum pressure value.

[0086] Table 1: Relationship between Pressure Loading Rate and Maximum Pressure Value

[0087] Packaging box material type Pressure loading rate (Newtons per second) Maximum pressure (Newtons) Corrugated cardboard 50-100 1000-2000 rigid cardboard 100-200 2000-3500 Wood composite 200-500 3500-5000

[0088] It is understood that initiating the pressure testing cycle ensures that the testing process is conducted under controlled and gradual conditions. Real-time monitoring of sensor data is achieved through a data acquisition card, with the sampling frequency matched to the pressure loading rate. This ensures that the complete deformation response is captured at each trigger threshold point. The recorded deformation data includes not only the overall deformation but also, sometimes, the deformation of key local points for detailed analysis. In some embodiments, the cyclic execution of pressure loading and data recording is interruptible, automatically stopping to protect the sample when a sudden increase in deformation or an abnormal signal is detected. Optionally, the preset pressure range coverage can be achieved by setting multiple pressure stages, each with a different loading rate and trigger threshold.

[0089] See Figure 5 This diagram relates to the parameter configuration stage of the compression resistance testing cycle. It clarifies that the loading rate and maximum pressure of the pressure application device must be initialized before starting the compression resistance testing cycle. This diagram covers three types of materials commonly used in intangible cultural heritage wine packaging boxes: corrugated cardboard, rigid cardboard, and wood composite. The loading rate is the step benchmark for pressure increase in the testing cycle, and the maximum pressure is the critical threshold for testing. Both must be compatible with the material characteristics. The parameter matching logic in this diagram is to avoid damage to the packaging box due to non-compression resistance defects caused by improper loading rate, while ensuring that the maximum pressure covers the nominal compression strength range of the material. This ensures that the test results reflect the actual compression resistance capability. It is the core basis for parameter configuration before starting the compression resistance testing cycle, supporting the orderly execution of subsequent pressure loading and data recording, and providing parameter benchmarks for the controllability of the testing process and the validity of the results.

[0090] Example 5: Overall compressive strength assessment. Multi-source data features are extracted from the testing process, including pressure distribution uniformity, deformation recovery rate, and stress concentration index. A feature fusion algorithm is used to merge the multi-source data features into a comprehensive feature vector. The comprehensive feature vector is input into a pre-trained compressive strength regression model to calculate the compressive strength score. The quality level of the packaging box is classified according to the compressive strength score, and a compressive performance assessment report is generated.

[0091] In practice, the overall compressive strength assessment extracts multi-source data features from the testing process, including pressure distribution uniformity, deformation recovery rate, and stress concentration index. Pressure distribution uniformity is obtained by calculating the standard deviation between pressure sensor array readings. Deformation recovery rate is calculated by the ratio of deformation rebound after unloading pressure to the maximum deformation. Stress concentration index is determined by analyzing the ratio of the maximum stress value to the average stress value in the strain sensor data. A feature fusion algorithm is used to merge the multi-source data features into a comprehensive feature vector. The feature fusion algorithm uses principal component analysis or directly concatenates multiple feature values ​​into a single vector. The comprehensive feature vector is input into a pre-trained compressive strength regression model to calculate the compressive strength score. The compressive strength regression model is a machine learning model trained using historical data, such as support vector regression or random forest regression. Based on the compressive strength score, the quality grade of the packaging box is classified, and a compressive performance evaluation report is generated. The quality grades are typically divided into excellent, good, medium, and poor. The compressive performance evaluation report contains the score, grade, and key feature data in document form.

[0092] In some embodiments, the deformation recovery rate can be calculated using the following formula: ,in: Indicates the deformation recovery rate. This represents the maximum deformation during the pressure loading process. This represents the residual deformation remaining after the pressure is completely unloaded. It can be understood that the deformation recovery rate reflects the elastic properties of the packaging material.

[0093] Optionally, the calculation of pressure distribution uniformity can be based on a pressure distribution map, statistically analyzing the pressure values ​​at all effective points in the map. The stress concentration index can be determined by locating the peak region in the strain time series data and calculating its ratio to the background stress. The pre-trained compressive strength regression model receives a comprehensive feature vector as input and outputs a continuous compressive strength score, which can be set from 0 to 100. It is understood that the accuracy of the compressive strength regression model depends on the representativeness of the training data and the model's generalization ability. In some embodiments, the threshold for quality grade classification can be customized according to industry standards or customer requirements; for example, a compressive strength score above 90 is considered excellent, and 75 to 90 is considered good. Optionally, in addition to the final score and grade, the compressive performance evaluation report can also include visualization results such as pressure-deformation curves and stress distribution cloud maps for in-depth analysis. If principal component analysis is used in the feature fusion process, a set of uncorrelated new feature vectors will be generated for dimensionality reduction, thereby improving the computational efficiency and predictive stability of the compressive strength regression model.

[0094] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0095] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes, characterized in that, The method is implemented through the following process: The surface pressure distribution data and internal stress waveform of the packaging box of intangible cultural heritage wines are collected by a high-precision sensor network to generate a multimodal sensing data stream; Based on the multimodal sensing data stream, a three-dimensional digital twin model of the packaging box is constructed, and the material elastic parameters and structural constraints in the model are initialized. The three-dimensional digital twin model is trained to predict its compressive performance, thereby generating an adaptive compressive performance prediction model. The adaptive stress prediction model is run in a virtual stress environment, and the deformation simulation sequence of the packaging box is output. Real-time acquisition of actual sensor monitoring data, dynamic comparison with the deformation simulation sequence, and calculation of deformation error coefficient. The deformation error coefficient is a comprehensive scalar used to quantify the overall degree of difference between simulation and actual deformation. The trigger threshold of the pressure detection device is adjusted according to the deformation error coefficient, and the pressure detection cycle is started based on the adjusted trigger threshold; Multi-source data are integrated during the compression testing cycle, and the overall compression strength of the packaging box is evaluated through spatiotemporal feature analysis. The steps for generating the adaptive stress resistance prediction model include: Prepare a historical compression resistance test dataset, including packaging box deformation records and final compression strength labels; design a deep neural network architecture, including convolutional layers for spatial feature extraction and recurrent layers for temporal dependency modeling; train the deep neural network using the historical compression resistance test dataset, and optimize the network weights and bias parameters; verify the prediction accuracy of the trained network on the test set, and generate an adaptive compression resistance prediction model. The output steps of the deformation simulation sequence include: A virtual pressure gradient is applied to the three-dimensional digital twin model to simulate the loading process from zero to a preset maximum pressure; the adaptive compressive strength prediction model is run to calculate the deformation and stress distribution of the packaging box under each pressure gradient; the deformation change curve over time is recorded to generate a deformation simulation sequence; the output deformation simulation sequence includes the deformation peak, deformation rate, and deformation stability point, where the deformation stability point is defined as the moment when the deformation rate is below a threshold. The calculation steps for the deformation error coefficient include: The system collects pressure and deformation values ​​from actual sensor monitoring data in real time to generate an actual deformation sequence; extracts the simulated deformation values ​​at corresponding time points from the deformation simulation sequence; calculates the absolute difference between the actual deformation value and the simulated deformation value, and normalizes it to obtain the deformation deviation value; and statistically analyzes the average and variance of the deformation deviation value throughout the entire detection cycle to generate the deformation error coefficient. The trigger threshold adjustment step of the pressure detection device includes: Based on the magnitude of the deformation error coefficient, the sensitivity level range is divided; the preset mapping table between sensitivity levels and trigger thresholds is queried to obtain the target trigger threshold; a control command is sent to the pressure detection device to adjust the sampling frequency and trigger threshold of its pressure sensor; the adjusted trigger threshold is confirmed to match the current detection environment to complete threshold self-adaptation.

2. The sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes according to claim 1, characterized in that, The steps for generating the multimodal sensing data stream include: Deploy a pressure sensor array at key points on the surface of the packaging box to collect continuous pressure values ​​and generate a pressure distribution map; Strain sensors are installed at structural nodes inside the packaging box to record stress change waveforms and generate strain time series data. Simultaneously acquire material micro-fracture signals captured by acoustic sensors to generate acoustic emission data sequences; The pressure distribution spectrum, strain time series data, and acoustic emission data sequence are time-aligned and format-normalized to generate a multimodal sensing data stream.

3. The sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes according to claim 1, characterized in that, The steps for constructing the three-dimensional digital twin model include: Analyze the geometric feature points in the multimodal sensing data stream to extract the length, width, height, and corner angles of the packaging box; Based on the extracted dimensions and angles, a meshed 3D model of the packaging box is constructed using the finite element method; The material elastic modulus and Poisson's ratio are derived from the multimodal sensing data stream, and the material properties of the three-dimensional model are initialized. Set the support boundaries and load conditions of the packaging box to complete the parameter initialization of the three-dimensional digital twin model.

4. The sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes according to claim 1, characterized in that, The initiation steps of the pressure resistance test cycle include: Initialize the pressure application device, and set the pressure loading rate and maximum pressure value; Based on the adjusted trigger threshold, the pressure application device is controlled to gradually increase the pressure; Real-time monitoring of sensor data; when the pressure value reaches the trigger threshold, record the current deformation data. The process of applying pressure and recording data is repeated until the preset pressure range is covered.

5. The sensor-based method for detecting the compression resistance of intangible cultural heritage wine packaging boxes according to claim 1, characterized in that, The evaluation steps for the overall compressive strength include: Multi-source data features were extracted during the testing process, including pressure distribution uniformity, deformation recovery rate, and stress concentration index. A feature fusion algorithm is used to merge features from multiple sources into a comprehensive feature vector. The comprehensive feature vector is input into the pre-trained stress resistance regression model to calculate the stress resistance score; The quality grade of the packaging box is determined based on the compressive strength score, and a compressive performance evaluation report is generated.

6. A sensor-based compression resistance testing system for intangible cultural heritage wine packaging boxes, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the sensor-based intangible cultural heritage wine packaging box compression testing method as described in any one of claims 1 to 5.

Images