Gas cylinder detection method and system during inflation

Abstract

The invention provides a gas cylinder detection method and system during inflation, and relates to the technical field of gas cylinder detection, and the method comprises the steps of data acquisition, digital twin model construction, damage acquisition, damage judgment and the like. The gas cylinder can be detected in the inflation process, and the safety of the inflation process is improved.

Description

Technical Field

[0001] The present application relates to the technical field of gas cylinder detection, and particularly to a method and system for detecting a gas cylinder during inflation. Background Art

[0002] In many key fields such as industrial production, medical treatment, and scientific research, high-pressure gas cylinders, as core equipment for storing and filling compressed gases or liquefied gases, the importance of their safety is self-evident. Once a gas cylinder leaks, it will not only cause waste of resources, but may also trigger catastrophic safety accidents such as fires and explosions, posing a very serious threat to the lives and property safety of personnel.

[0003] At present, the technical means for gas cylinder safety monitoring mainly focus on the assessment of structural integrity, that is, image recognition technology is used to analyze parameters such as the flatness and depression deviation of the gas cylinder shell, so as to infer whether there are accident risks in the gas cylinder.

[0004] The above analysis method can only comprehensively detect the gas cylinder when it is in a static state. However, during the filling process, especially at the damaged points of the gas cylinder, it is more likely to be affected by the filling strategy, resulting in damage to its shell. At this time, due to the influence of the filling equipment, the image recognition technology is not sufficient to comprehensively investigate the potential risks of the gas cylinder shell. Summary of the Invention

[0005] In order to be able to detect the gas cylinder during the filling process and improve the safety of the filling process, the present application provides a method and system for detecting a gas cylinder during inflation.

[0006] In the first aspect, the present application provides a method for detecting a gas cylinder during inflation, adopting the following technical solution: A method for detecting a gas cylinder during inflation includes the following steps: Data acquisition: Obtain the point cloud data before gas cylinder filling, the strain data before filling, and the strain data during filling; Construct a digital twin model: Set the temperature boundary condition and stress boundary condition, construct a point cloud model based on the point cloud data, perform mesh division on the point cloud model to obtain mesh data; use the finite element method, input the mesh data into the temperature boundary condition to obtain the temperature distribution, and input the mesh data into the stress boundary condition to obtain the stress distribution; fuse the temperature distribution, stress distribution with the point cloud model after mesh division to obtain the digital twin model; Obtain damage: Draw a strain-time curve based on the strain data during filling, fit the strain-time curve based on the digital twin model and the strain data before filling, and obtain the real-time damage amount of the gas cylinder based on the digital twin model after fitting; Damage judgment: Determine whether the real-time damage amount is greater than the preset damage amount threshold. If so, send an alarm signal; if not, after a preset time interval, re-execute the step of obtaining the damage until the filling is completed.

[0007] This application first obtains various data before and during filling, including point cloud data before filling, strain data before filling, and strain data during filling. Subsequently, this application sets temperature boundary conditions and stress boundary conditions, and constructs a point cloud model based on the point cloud data. Subsequently, this application uses the finite element method to input the mesh data into the temperature boundary conditions and stress boundary conditions respectively, obtains the temperature distribution and stress distribution, and fuses the temperature distribution, stress distribution with the point cloud model after mesh division to construct a digital twin model, realizing the integration of multi-dimensional data and enabling the digital twin model to more comprehensively and realistically reflect the state of the gas cylinder. Subsequently, this application fits the strain-time curve based on the digital twin model and the strain data before filling, and after the fitting is completed, obtains the real-time damage amount of the gas cylinder based on the digital twin model, realizing the real-time monitoring of the damage situation of the gas cylinder. Subsequently, this application determines whether the real-time damage amount is greater than the preset damage amount threshold. If so, send an alarm signal. This application comprehensively uses various data (such as point cloud data, strain data, etc.) and advanced analysis methods (such as finite element method, digital twin technology, etc.), can accurately capture various state information of the gas cylinder during filling, realize the detailed monitoring of the damage situation of the gas cylinder, and improve the filling safety.

[0008] Optionally, in the step of constructing the digital twin model, it further includes: calculating the stress change amount based on the strain data during filling, and fusing the stress change amount to obtain the digital twin model.

[0009] This application calculates the stress change amount based on the strain data during filling and feeds it back to the digital twin model, enabling the digital twin model to reflect the actual state of the gas cylinder in real time. This application also considers the stress change amount during filling when constructing the digital twin model, reduces the deviation between the digital twin model and the actual state of the gas cylinder, and improves the accuracy and reliability of the digital twin model.

[0010] Optionally, the method further includes: Constructing a grid temperature matrix: Extract the temperature and coordinates of each grid node in the digital twin model, record the temperature of the grid node as the grid temperature, and construct a grid temperature matrix according to the grid temperature; Modeling and Classification: Map the coordinates of each grid node to the image coordinate system, map the grid temperature to the gray value of the corresponding pixel point to obtain the gray-scale image of the gas cylinder, and construct a CNN-SVM model. The CNN-SVM model includes a CNN sub-model for feature extraction and an SVM sub-model for classifying the features extracted by the CNN sub-model. Input the gray-scale image of the gas cylinder into the CNN-SVM model to obtain the classification result; analyze the classification result to obtain the analysis result. Leakage Judgment: Judge whether the analysis result is leakage. If so, send out a leakage signal; if not, re-execute the step of constructing the grid temperature matrix after a preset time interval until the filling is completed.

[0011] This application first extracts the temperature and coordinates of each grid node from the digital twin model and integrates the temperature information into the grid temperature matrix to intuitively reflect the spatial distribution of the surface temperature of the gas cylinder. Subsequently, this application maps the coordinates of each grid to the image coordinate system and maps the grid temperature to the gray value of the corresponding pixel point to obtain the gray-scale image of the gas cylinder, presenting the temperature information in an intuitive image form. Subsequently, this application constructs a CNN-SVM model. The CNN sub-model can automatically learn the feature patterns in the gray-scale image of the gas cylinder, and the SVM sub-model efficiently classifies these features. By combining the feature extraction ability of CNN and the classification ability of SVM, this application can more accurately extract the temperature anomaly features from the gray-scale image of the gas cylinder and perform classification judgment. Subsequently, this application analyzes the classification result of the CNN-SVM model. Once it is found that the classification result is leakage, a leakage signal is immediately sent out to achieve real-time monitoring and early warning of the temperature anomaly of the gas cylinder. Otherwise, after a preset time interval, the step of constructing the grid temperature matrix is re-executed to continue monitoring the gas cylinder until the filling is completed, thereby improving the safety during the entire filling process. This application introduces a temperature anomaly detection mechanism based on image processing and machine learning, and realizes the accurate identification and early warning of the temperature anomaly of the gas cylinder through a series of steps such as matrix construction, modeling and classification, and classification judgment, improving the safety of the filling process.

[0012] Optionally, after executing the step of constructing the grid temperature matrix and before executing the step of modeling and classification, it further includes: Coordinate Transformation: Take any grid node as the center point, and convert the coordinates of all grid nodes in the grid temperature matrix into relative coordinates based on the center point; In the step of modeling and classification, map the relative coordinates to the image coordinate system.

[0013] In this application, an arbitrary grid node is selected as the center point. Based on the center point, the coordinates of the remaining grid nodes in the grid temperature matrix are converted into relative coordinates. Then, the relative coordinates are mapped into the image coordinate system. By adopting the above solution, this application constructs an image corresponding to the temperature distribution on the surface of the gas cylinder. This image enables the temperature anomaly features to be presented in a more intuitive and recognizable form on the image, providing favorable conditions for subsequent feature extraction and classification judgment using the CNN-SVM model, further improving the accuracy and reliability of the gas cylinder temperature anomaly detection, and effectively ensuring the safety of the gas cylinder during the entire filling process.

[0014] Optionally, after performing the coordinate transformation step and before performing the modeling and classification step, it further includes: Construct a KD tree: Use the relative coordinates corresponding to the center point as the root node, and use the relative coordinates corresponding to the remaining grid nodes as the child nodes to construct a KD tree, and calculate the Euclidean distance between adjacent node pairs based on the KD tree; Distance judgment: Judge whether the i-th Euclidean distance is greater than the preset Euclidean distance threshold. If so, perform the step of calculating interpolation; if not, perform the step of updating the distance; Calculate interpolation: Determine the relative coordinates corresponding to the i-th Euclidean distance and the node pair in the KD tree, add new nodes between the node pairs in the KD tree, and use the linear interpolation algorithm to determine the relative coordinates corresponding to the new nodes, and perform the step of updating the distance; Update distance: Update the (i + 1)-th Euclidean distance to the i-th Euclidean distance, and perform the distance judgment step until all Euclidean distances are traversed; Update the KD tree: Update the new nodes to the KD tree to obtain a new KD tree, calculate the Euclidean distance between adjacent node pairs in the new KD tree, and perform the distance judgment step; In the modeling and classification step, map the relative coordinates of each node in the new KD tree into the image coordinate system.

[0015] In this application, the relative coordinates corresponding to the center point are used as the root node, and the relative coordinates corresponding to the remaining grid nodes are used as child nodes to construct a KD tree, obtaining an efficient spatial index structure. Subsequently, this application calculates the Euclidean distance between adjacent node pairs in the KD tree, improving the calculation efficiency. Subsequently, this application determines the spatial distance between grid nodes by judging whether the i-th Euclidean distance is greater than a preset Euclidean distance threshold. If the distance is too large, it indicates that the spatial distribution span between the nodes is large. At this time, this application will add a new node between this node pair and use the linear interpolation algorithm to determine the relative coordinates corresponding to the new node, thereby smoothing the spatial distribution of the grid. Through interpolation, this application fills the blank area between grid nodes, making the distribution of grid nodes more uniform, improving the continuity and integrity of spatial data. The new nodes after interpolation are incorporated into the KD tree, further optimizing the spatial relationship between grids. Subsequently, this application updates the (i + 1)-th Euclidean distance to the i-th Euclidean distance and repeats the step of distance judgment, realizing the cyclic iterative processing of all Euclidean distances and the comprehensive evaluation of the distances between all grid node pairs, and timely discovering and handling abnormal spatial distribution situations. This application recalculates the Euclidean distance between adjacent node pairs in the new KD tree and dynamically adjusts the KD tree structure according to the calculation results to achieve continuous optimization. Finally, this application maps the relative coordinates of each node in the new KD tree to the image coordinate system, thereby constructing a more accurate gas cylinder temperature map. Since there are more node coordinates in the new KD tree, the mapped image can more continuously reflect the temperature distribution on the surface of the gas cylinder, improving the quality and usability of the image. By constructing a KD tree, this application can efficiently organize and manage the coordinate information of grids in the grid temperature matrix, improving the calculation efficiency.

[0016] Optionally, before performing the step of data collection, the method further includes: Damage collection: Obtain the cumulative damage amount of the current gas cylinder before filling and the maximum damage amount that the same type of gas cylinder can withstand, calculate the remaining damage amount based on the cumulative damage amount and the maximum damage amount, and obtain historical data, where the historical data includes: the historical filling strategy of the same type of gas cylinder, the environmental information during filling, and the damage amount difference corresponding to each historical filling strategy; Modeling and training: Establish a neural network model and train the neural network model using the historical data to obtain a trained neural network model; Prediction: Obtain input data, where the input data includes multiple real-time filling strategies and real-time environmental information, input the input data into the trained neural network model, and obtain the predicted damage amount difference for each real-time filling strategy; Difference judgment: successively judge whether each predicted damage amount difference is greater than the remaining damage amount. If so, mark the real-time filling strategy corresponding to the predicted damage amount difference as an unavailable strategy; if not, execute the feedback step. Feedback: Mark the real-time filling strategy corresponding to the predicted damage amount difference as an available strategy, and send the available strategy to the staff.

[0017] This application obtains the cumulative damage amount of the current gas cylinder before filling and the maximum damage amount that the same type of gas cylinder can withstand, and then calculates the remaining damage amount. This application also collects historical data such as the historical filling strategies of the same type of gas cylinders, the environmental information during filling, and the damage amount difference corresponding to each historical filling strategy, providing rich data support for the subsequent neural network model training. Subsequently, this application establishes a neural network model and trains it using historical data. Through training, the neural network model can learn the complex relationship between the filling strategy, the environmental information during filling, and the damage amount difference corresponding to each filling strategy. Subsequently, this application obtains the input data (i.e., multiple real-time filling strategies and real-time environmental information) and inputs it into the trained neural network model to obtain the predicted damage amount difference corresponding to each real-time filling strategy, realizing the real-time prediction of the damage situation of different real-time filling strategies, enabling the staff to timely understand the risk degree of each real-time filling strategy. Subsequently, this application successively judges whether each predicted damage amount difference is greater than the remaining damage amount to identify the real-time filling strategies that may cause excessive damage to the gas cylinder and marks them as unavailable strategies. This step provides a clear standard and basis for the safety assessment of the filling strategy. When the gas cylinder has a certain amount of damage, this application can select the real-time filling strategy based on the predicted damage amount difference. When the predicted damage amount difference is less than the remaining damage amount, mark the real-time filling strategy corresponding to the predicted damage amount difference as an available strategy and send the available strategy to the staff, realizing the timely transmission of the filling strategy. The staff can quickly make a decision based on the feedback result and select the appropriate real-time filling strategy, thus providing intuitive and clear strategy selection suggestions for the staff and helping to optimize the filling decision.

[0018] Optionally, the feedback step further includes: setting the priority of the available strategy according to the genetic algorithm, associating the priority with the available strategy, generating a strategy priority list, and the staff selects the available strategy with the highest priority in the strategy priority list to fill the gas cylinder.

[0019] Genetic algorithms can prioritize available strategies by comprehensively considering multiple factors. By adopting the above technical solutions, staff's decisions no longer rely solely on simple empirical judgments, but are based on scientific algorithm analysis, which improves the scientificity and accuracy of decision-making. After associating priorities with available strategies, staff can quickly locate the available strategy with the highest priority, reducing the time for screening and comparison among many available strategies, thereby improving filling efficiency and enabling gas cylinders to be filled faster and safer.

[0020] Optionally, after the step of performing feedback, the method further comprises: Strategy judgment: Determine whether there is an available strategy. If so, execute the damage judgment step; if not, send a stop filling signal to the staff.

[0021] By making a strategy judgment immediately after the feedback link, this application can promptly discover whether there is an available strategy, that is, whether the gas cylinder can be filled under the calculated remaining damage amount. If not, a stop filling signal is sent in time to ensure the life safety of the staff. Through clear filling strategy judgment logic, this application can quickly determine the direction of subsequent operations. If there is an available strategy, the damage judgment will continue to make the filling process proceed in an orderly manner; if there is no available strategy, a stop filling signal will be sent directly to reduce the risk of gas cylinder damage.

[0022] Optionally, the method further comprises: Gas cylinder identification: Obtain the characteristic identification of the gas cylinder, which is a unique identification mark preset on the gas cylinder body, and determine whether the characteristic identification can be identified; If yes, based on the feature identifier, obtaining feature information carried by the feature identifier; If not, a feedback signal is sent to the staff; Filling detection: setting the feature matching criteria of the gas cylinder, using the feature matching criteria to detect the feature information, obtaining real-time detection results, and judging whether the real-time detection results meet expectations; If yes, the real-time detection result is added to the feature information, and the gas cylinder is placed in the filling area to wait for filling; If not, a feedback signal is sent to the staff.

[0023] By adopting the above scheme, the present application detects the gas cylinders to be filled, so that only the matching gas cylinders will be filled, otherwise a feedback signal will be sent to the staff. The present application can effectively prevent unidentifiable gas cylinders (not from this factory) from entering the filling process, thereby improving the reliability of the gas cylinder filling process.

[0024] In the second aspect, the present application provides a gas cylinder detection system during inflation, which adopts the following technical solution: A gas cylinder detection system during inflation, comprising: a memory and a processor, wherein the memory stores a computer-readable storage medium; When the processor processes the computer program stored on the computer-readable storage medium, the method described above is implemented.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: 1. The present application first obtains various data before and during filling, including point cloud data before filling, strain data before filling, and strain data during filling. Subsequently, the present application sets temperature boundary conditions and stress boundary conditions, and constructs a point cloud model based on the point cloud data. Subsequently, the present application uses the finite element method to input the mesh data into the temperature boundary conditions and stress boundary conditions respectively, obtains the temperature distribution and stress distribution, and fuses the temperature distribution, stress distribution with the point cloud model after mesh division to construct a digital twin model, realizing the integration of multi-dimensional data, enabling the digital twin model to more comprehensively and realistically reflect the state of the gas cylinder. Subsequently, the present application fits the strain-time curve based on the digital twin model and the strain data before filling, and after the fitting is completed, obtains the real-time damage amount of the gas cylinder based on the digital twin model, realizing the real-time monitoring of the damage condition of the gas cylinder. Subsequently, the present application determines whether the real-time damage amount is greater than a preset damage amount threshold, and if so, issues an alarm signal. The present application comprehensively uses a variety of data (point cloud data, strain data, etc.) and advanced analysis methods (finite element method, digital twin technology, etc.), can accurately capture various state information of the gas cylinder during the filling process, realize the detailed monitoring of the damage condition of the gas cylinder, and improve the filling safety.

[0026] 2. By constructing a KD tree, the present application can efficiently organize and manage the coordinate information of the grids in the grid temperature matrix, improving the calculation efficiency.

[0027] 3. By optimizing the filling strategy, the present application can select a safer and more reasonable filling strategy, thereby effectively reducing the filling risk. Description of the Drawings

[0028] Figure 1 is the flowchart of Embodiment 1 of the present application; Figure 2 is the flowchart of Embodiment 2 of the present application; Figure 3 is the flowchart of Embodiment 3 of the present application; Figure 4 is the flowchart of Embodiment 4 of the present application. Detailed Description of the Embodiments

[0029] The following is a further detailed description of the present application in conjunction with Figures 1 to 4 to the present application.

[0030] Example 1: This example discloses a method for detecting a gas cylinder during inflation. Refer to Figure 1 , the method includes: S11 data acquisition, S12 constructing a digital twin model, S13 obtaining damage, and S14 damage judgment. First, obtain the point cloud data before gas cylinder filling, the strain data before filling, and the strain data during filling; then set the temperature boundary condition and stress boundary condition, construct a point cloud model and perform mesh division on it to obtain mesh data, and then use the finite element method to input the mesh data into each boundary condition respectively to obtain the temperature distribution and stress distribution; after that, fuse the temperature distribution, stress distribution, and the point cloud model after mesh division to construct a digital twin model; subsequently, draw a strain-time curve based on the strain data during filling, fit the strain-time curve based on the digital twin model and the strain data before filling, and obtain the real-time damage amount of the gas cylinder based on the digital twin model after fitting; finally, judge whether the real-time damage amount exceeds the preset loss amount threshold. If so, give an alarm. Otherwise, after a preset time interval, collect real-time strain data again and execute S14 to obtain damage until the filling is completed. The execution process of this example is as follows: S11 data acquisition, obtain the point cloud data before gas cylinder filling, the strain data before filling, and the strain data during filling.

[0031] Use a high-precision three-dimensional laser scanner (such as a phase-type laser scanner) to perform an all-round scan of the gas cylinder, and keep the gas cylinder stationary and unobstructed during the scan. The scanner emits laser beams at a speed of millions of points per second, and calculates the spatial coordinates of each point on the gas cylinder surface by measuring the time difference between the emission and reflection of the laser beam. To ensure data integrity, it is necessary to scan the gas cylinder from multiple angles (such as the top, bottom, side wall, etc.), and finally splice the point cloud data of each angle through a registration algorithm (such as the ICP algorithm) to form a complete point cloud data set of the gas cylinder before filling. These point cloud data can intuitively present the initial shape contour and surface characteristics of the gas cylinder.

[0032] In other examples, the point cloud data of the gas cylinder before filling can also be obtained through technologies such as structured light scanning or industrial CT scanning.

[0033] Before gas cylinder filling, install multiple high-precision strain sensors on the surface of the gas cylinder. These strain sensors can real-time sense the tiny deformation of the gas cylinder under the stress state, and convert the deformation data into electrical signals for recording. Through these strain data, the stress distribution of the gas cylinder in the initial state can be understood.

[0034] During gas cylinder filling, use the installed strain sensors to continuously and real-time collect the strain change values of each monitoring point.

[0035] S12 Construct a digital twin model, and set temperature boundary conditions and stress boundary conditions according to the actual filling environment and usage requirements of the gas cylinder.

[0036] The temperature boundary conditions consider the range of ambient temperature changes that may be encountered during the filling process. Based on the filling pressure, working pressure, and safety factor of the gas cylinder, the stress boundary conditions that the gas cylinder may withstand during the filling process are calculated.

[0037] Considering factors such as the internal pressure stress, external constraint stress (i.e., the constraint of the fixed bracket), and thermal stress (stress caused by temperature changes) of the gas cylinder, a comprehensive stress boundary condition is constructed.

[0038] Using point cloud processing software, the collected point cloud data is processed and analyzed to construct a point cloud model. This point cloud model can restore the geometric shape and surface characteristics of the gas cylinder, providing an accurate geometric basis for subsequent mesh generation and finite element analysis.

[0039] Subsequently, in this embodiment, the constructed point cloud model is meshed, and the geometric model of the gas cylinder is discretized into a finite number of element meshes. By meshing the point cloud model, mesh data can be obtained, and the mesh data includes coordinate information of mesh nodes, connection relationships between mesh elements, etc.

[0040] Using the finite element method, the mesh data is respectively input into the set temperature boundary conditions and stress boundary conditions. The finite element software will perform numerical calculations based on the input mesh data to solve the temperature distribution and stress distribution of the gas cylinder under the given boundary conditions. The temperature distribution reflects the temperature change situation of each part of the gas cylinder during the filling process, and the stress distribution reveals the magnitude and direction of the stress borne by the gas cylinder at different positions.

[0041] The temperature distribution and stress distribution data obtained from the finite element analysis are fused with the point cloud model after mesh generation. Through data mapping technology, temperature and stress values are assigned to each mesh node of the point cloud model to form a digital twin model with the coupling characteristics of temperature-stress-geometry multi-physical fields.

[0042] The digital twin model is an accurate mapping of the physical entity in the virtual space, and it can reflect the physical state and behavior of the gas cylinder in real time. Through the digital twin model, the temperature and stress change situations of the gas cylinder during the filling process can be intuitively observed.

[0043] S13 Obtain damage, and draw a strain-time curve based on the strain data during filling. In the coordinate system of this curve, the vertical axis is the strain value (unit με), and the horizontal axis is the time (unit second).

[0044] Based on the digital twin model and the strain data before filling, fit the strain-time curve, that is: use the strain data before filling as the initial condition of the digital twin model, and input the strain data represented by the strain-time curve into the digital twin model in the order of acquisition time, so as to realize the simulation of the gas cylinder filling process.

[0045] After the fitting is completed, obtain the damage distribution cloud map of the gas cylinder based on the digital twin model. According to the damage distribution cloud map, the digital twin model uses a continuous damage mechanics model (such as the Lemaitre damage model or the Gurson-Tvergaard-Needleman (GTN) model) to calculate the real-time damage amount of the gas cylinder.

[0046] Taking the Lemaitre damage model as an example, the calculation model of the real-time damage amount is as follows: ; Where G is the real-time damage amount; is a material parameter, with a value of 10 3 -10 6 ; I is the elastic modulus; is the strain data at the current moment; is the stress value; s is the sensitivity of the real-time damage amount to the plastic strain, with a value range of 1-2; r is the degree of influence of the control equivalent stress on the damage evolution, with a value range of 1-2.

[0047] S14 Damage judgment: Compare the calculated real-time damage amount with the preset damage amount threshold. If the real-time damage amount is greater than the preset damage amount threshold, it means that the damage degree of the gas cylinder exceeds the expectation, and an alarm signal is immediately sent to remind relevant personnel to take necessary measures, such as adjusting the filling strategy and overhauling the gas cylinder. If the real-time damage amount is less than or equal to the preset damage amount threshold, it means that the gas cylinder is currently in a safe state. After a preset time interval, collect the real-time strain data again and execute S13 to obtain the damage, continuously monitor the damage situation of the gas cylinder until the gas cylinder is filled.

[0048] Taking a certain energy company filling a batch of high-pressure gas cylinders as an example, this embodiment is further elaborated.

[0049] For this filling, cylindrical gas cylinders are used, with a self-weight of 200 kg, a diameter of 300 mm, a height of 1000 mm, the material is steel (its elastic modulus is 210 GPa, and the Poisson's ratio is 0.3), the wall thickness is 10 mm, and the thermal expansion coefficient is 12×10 -6 / ℃.

[0050] S11 Data acquisition Using lidar scanning technology, the gas cylinder is scanned 360° in all directions. The lidar emits high-frequency laser beams and receives the reflected signals, accurately recording the three-dimensional coordinates of 1 million points on the surface of the gas cylinder (accuracy ±0.1 mm) to construct a point cloud model before filling. This point cloud model can clearly present the initial outer contour of the gas cylinder and features such as surface welds and coatings.

[0051] Twelve high-precision strain sensors (range ±1000 με, accuracy ±1 με) are evenly installed on the surface of the gas cylinder, covering key areas of the bottle body, bottleneck, and bottom.

[0052] The strain sensors record the initial stress distribution in real time. At this time, the stress at the bottleneck and the stress in the middle of the bottle body are regarded as 0.

[0053] The strain sensors will continuously collect strain data during the filling process.

[0054] S12 Construct a digital twin model Temperature boundary conditions: Set the ambient temperature to 25 °C, and the temperature gradient inside the gas cylinder during filling is -20 °C to 80 °C.

[0055] Stress boundary conditions: Internal pressure 15 MPa (uniformly acting on the inner wall) Self-weight: 200 kg. The calculation process of the bottom pressure stress of the bottle is as follows: Vertical force = self-weight of the gas cylinder × acceleration due to gravity = 200 × 9.81 = 1962 (N) Contact area ≈ 3.14 × 0.15 2 ≈ 0.707 (square meters) Self-stress is: -1962 ÷ 0.707 ≈ -27.75 kPa Circumferential stress is: - (15 × 10 6 × 0.3) ÷ (2 × 0.01) = -225 kPa Axial stress is: - (15 × 10 6 × 0.3) ÷ (4 × 0.01) = -112.5 kPa After superposition: The total stress at the bottom of the bottle is the vector sum of the self-weight stress and the internal pressure stress.

[0056] If there is a temperature gradient during the filling process, thermal stress will be further superimposed: The increase in the inner wall temperature causes thermal expansion to be blocked, generating compressive stress; the decrease in the outer wall temperature causes contraction to be blocked, generating tensile stress. The calculation process of thermal stress is as follows: ; The inner wall temperature rises by 80 - 25 = 55 degrees Celsius, and the outer wall temperature drops by -20 - 25 = -45 degrees Celsius.

[0057] Inner wall thermal stress = 210×10 9 ×12×10 -6 ×55 ≈ 138.6 MPa Outer wall thermal stress = 210×10 9 ×12×10 −6 ×(-45) ≈ -113.4 MPa When the temperature gradient is -20°C to 80°C, the inner wall thermal stress increases by 138.6 MPa and the outer wall decreases by 113.4 MPa.

[0058] Use Geomagic software to denoise and smooth the point cloud data, and construct the point cloud model of the gas cylinder. Discretize the point cloud model into 100,000 tetrahedral elements, and the mesh nodes include coordinates (such as the coordinates of a certain mesh node are (300, 0, 1000)) and element connection relationships (such as this mesh element is connected to mesh nodes A, B, C, and D).

[0059] Integrate the stress distribution, temperature distribution, and the point cloud model after mesh division to construct a digital twin model.

[0060] S13 Obtain damage Based on the strain data during filling, plot the strain-time curve. Based on the digital twin model and the strain data before filling, fit the strain-time curve. Then the real-time damage amount is the integral of the strain data.

[0061] Taking the Lemaitre damage model as an example, the calculation model of the real-time damage amount is as follows: ; For the convenience of calculation, discretize it: ; ; The values of each parameter are: is 100 MPa; the damage amount of the gas cylinder before filling is 0.08; is 0.001 s; r and s are 1.5; I is 210. The real-time damage amount is equal to the sum of the damage amount of the gas cylinder before filling and the damage change amount, that is, 0.080018.

[0062] S14 Damage judgment, compare the real-time damage amount with the preset damage amount threshold (set to 0.9). At this moment, the damage amount does not exceed the threshold. After an interval of the preset duration, re-execute S13 to obtain damage.

[0063] Through experimental verification, this embodiment shortens the maintenance response time by 50%, improves the filling efficiency by 20%, reduces the shutdown losses caused by gas cylinder damage, and saves about 500,000 yuan in annual costs.

[0064] In other embodiments, constructing the digital twin model in S2 further includes: applying the principles and related formulas of elasticity mechanics, combining the material properties and geometric dimensions of the gas cylinder, and converting the strain data into stress data. Then, calculate the difference between the real-time stress data and the stress data before filling to obtain the stress change amount. Exemplarily, the calculation model of the stress change amount is as follows: ; is the strain, and its value is equal to the difference between the real-time strain data and the strain data before filling; is the stress change amount; F is the fatigue strength coefficient of the gas cylinder; E and n are empirical constants.

[0065] Use the calculated stress change amount to update the stress distribution parameters in the digital twin model. By integrating the stress change amount into the digital twin model, the digital twin model can more accurately reflect the actual stress state of the gas cylinder during the filling process.

[0066] In this embodiment, first, the point cloud data of the gas cylinder before filling is obtained through lidar scanning technology, the strain data before and after filling is collected by strain sensors, then the temperature boundary conditions and stress boundary conditions are set based on the actual filling environment, the point cloud model is constructed and divided by point cloud processing software to obtain grid data, the finite element method is used to solve the temperature and stress distribution of the gas cylinder in combination with the boundary conditions, then the temperature distribution, stress distribution and the meshed point cloud model are fused to construct a digital twin model, the strain-time curve is fitted based on the digital twin model and the strain data during filling, the continuous damage mechanics model is used to calculate the real-time damage amount, and finally, the damage degree of the gas cylinder is judged by comparing the real-time damage amount with the preset damage amount threshold. If it exceeds the preset damage amount threshold, an alarm signal is sent, otherwise, continuous monitoring is carried out until the filling is completed. This embodiment can use the digital twin model to simulate the filling process, achieving the purpose of detecting the gas cylinder during the filling process, thereby improving the safety of the gas cylinder filling process.

[0067] Embodiment 2: Refer to Figure 2 , the difference between this embodiment and Embodiment 1 is that the method further includes: S20 constructs a grid temperature matrix, traverses each grid node in the updated digital twin model, and obtains the temperature data (i.e., grid temperature) and coordinates of each grid node.

[0068] Since the digital twin model is constructed based on the finite element method, each grid node has a unique identifier, and the corresponding temperature and coordinate information can be accurately extracted through this identifier.

[0069] Arrange the extracted grid temperatures in the order of the grids to construct a grid temperature matrix. Assume that the digital twin model is spatially divided into M×N×W grids, then the dimension of the grid temperature matrix J is M×N×W, where J(i, j, k) represents the temperature of the grid at the i-th row, j-th column, and k-th layer. Since a fixed layer is selected in this application to detect the temperature, the grid temperature matrix in this embodiment is a two-dimensional matrix.

[0070] S21 Coordinate transformation: Take any grid node as the center point, and based on the center point, convert the coordinates of all grid nodes in the grid temperature matrix into relative coordinates.

[0071] S22 Construct a KD tree: Use the relative coordinate corresponding to the center point as the root node, and use the relative coordinates corresponding to the remaining grid nodes as child nodes to construct a KD tree. Find the nearest neighbor points in the constructed KD tree, and calculate the Euclidean distance between the grid nodes corresponding to the adjacent node pairs in turn.

[0072] A KD tree is a data structure for organizing points in a multi-dimensional space. It constructs a tree structure by recursively dividing the space into two sub-spaces. During the construction process, the median is selected as the dividing point according to the dimension of the current node, and the point set is divided into two parts to construct the left and right sub-trees respectively.

[0073] S23 Distance judgment: Based on the Euclidean distance calculated in the KD tree constructed in S22, judge whether the i-th Euclidean distance is greater than the preset Euclidean distance threshold. If so, execute S24 to calculate the interpolation; if not, execute S25 to update the distance.

[0074] S24 Calculate interpolation: Determine the relative coordinates corresponding to the i-th Euclidean distance and the node pair in the KD tree, add new nodes between the node pairs in the KD tree, and use the linear interpolation algorithm to determine the relative coordinates corresponding to the new nodes, and then execute S25 to update the distance.

[0075] S25 Update distance: Update the (i + 1)-th Euclidean distance to the i-th Euclidean distance, and execute S23 distance judgment until all Euclidean distances are traversed.

[0076] S26 Update the KD tree: Update the new nodes to the KD tree to obtain a new KD tree, calculate the Euclidean distance between the adjacent node pairs in the new KD tree, and execute S23 distance judgment.

[0077] S27 Modeling and classification: Map the relative coordinates of each node in the new KD tree to the image coordinate system, map the grid temperature corresponding to each node in the new KD tree to the gray value of the corresponding pixel point to obtain the gas cylinder gray-scale image.

[0078] The calculation model for mapping the grid temperature to the gray value range of 0 - 255 is: ; where L is the gray value of the pixel corresponding to the current node; T is the current grid temperature; is the minimum value in the grid temperature; is the maximum value in the grid temperature; round(·) is the rounding operation.

[0079] Construct a CNN-SVM model, which includes a CNN sub-model for feature extraction and an SVM sub-model for classifying the features extracted by the CNN sub-model. Input the gray-scale image of the gas cylinder into the CNN-SVM model to obtain a classification result.

[0080] The CNN-SVM model first extracts features from the gray-scale image of the gas cylinder, classifies according to the extracted features, and obtains a classification result, which includes normal and leakage.

[0081] S28 Leakage judgment. Judge whether the classification result is leakage. If so, send a leakage signal; if not, execute S20 to construct a grid temperature matrix after a preset time interval until the filling is completed.

[0082] In this embodiment, first, the temperature and coordinates of each grid are extracted from the digital twin model to construct a grid temperature matrix. Then, the coordinate parameters are parameterized with an arbitrary grid node as the reference and a KD tree is constructed. Based on this tree, the Euclidean distance between adjacent nodes is calculated. By judging the size relationship between the distance and a preset threshold, linear interpolation is performed on the node pairs greater than the threshold to add new nodes. Subsequently, the KD tree is updated and the distance judgment and interpolation operations are repeated until all distance processing is completed. Then, the coordinates of the KD tree nodes are mapped to the image coordinate system and associated with the temperature to generate a gas cylinder temperature map. After grayscale processing, a gas cylinder gray-scale image is obtained. Finally, a CNN-SVM model (including a CNN sub-model and an SVM sub-model) is constructed. The gas cylinder gray-scale image is input into this model to obtain a classification result. According to the result, it is judged whether to send a leakage signal. If it does not meet the expectation, the step of constructing the grid temperature matrix is re-executed after a preset time interval until the gas cylinder filling is completed. This embodiment realizes the accurate monitoring and analysis of the temperature during the gas cylinder filling process, can timely detect potential abnormal situations such as leakage, and ensures the safety and reliability of the gas cylinder filling.

[0083] Example 3: Refer to Figure 3 , the difference between this embodiment and Embodiment 1 is that before executing S11 data acquisition, the method further includes: S31 Damage collection. Before filling the gas cylinder, obtain the cumulative damage amount of the current gas cylinder through professional detection equipment and methods. For example, use an ultrasonic detector to detect the wall thickness of the gas cylinder, and calculate the cumulative damage amount based on the change in wall thickness; or detect damage conditions such as cracks and deformations on the surface of the gas cylinder through methods such as magnetic particle flaw detection, and determine the cumulative damage amount in combination with relevant damage assessment standards.

[0084] Collect data such as the design parameters and material properties of the same type of gas cylinders, and determine the maximum damage amount that the same type of gas cylinders can withstand in combination with relevant industry standards and specifications. For example, determine the maximum damage amount through theoretical calculations or experimental tests based on factors such as the design pressure and material strength of the gas cylinder. In this embodiment, the maximum damage amount can also be determined by detecting the actual damage amount of scrapped gas cylinders of the same type and combining relevant standards.

[0085] Calculate the remaining damage amount based on the cumulative damage amount and the maximum damage amount. The calculation formula is: Remaining damage amount = Maximum damage amount - Cumulative damage amount.

[0086] Obtain the historical filling strategies of the same type of gas cylinders, including parameters such as the filling methods, filling speeds, and filling pressures of different gas cylinders under different historical conditions. These parameters will affect the damage conditions of the gas cylinders during the filling process; environmental information during filling, such as environmental factors such as temperature, humidity, and air pressure. These factors will affect the material properties of the gas cylinders and thus affect the degree of damage to the gas cylinders; and the damage amount difference corresponding to each historical filling strategy, that is, the difference between the damage amount of the gas cylinder after filling with this historical filling strategy and the damage amount before filling. Historical data can be obtained through channels such as the databases of gas cylinder filling enterprises and relevant industry reports.

[0087] S32 Modeling and training. Establish a neural network model. You can choose a neural network structure suitable for processing regression problems, such as a multi-layer perceptron (MLP), a recurrent neural network (RNN), etc. Taking the multi-layer perceptron as an example, it consists of an input layer, a hidden layer, and an output layer. The input layer receives features such as filling strategies and environmental information, the hidden layer extracts and transforms the features, and the output layer outputs the predicted damage amount difference.

[0088] Use historical data to train the neural network model. Use the historical filling strategies and environmental information in the historical data as input features, and the damage amount difference as the output label. Continuously adjust the weights and biases of the neural network through the backpropagation algorithm to minimize the error between the prediction result of the neural network model and the actual damage amount difference. Methods such as cross-validation can be used to evaluate the performance of the model during the training process.

[0089] By training the neural network model, enable it to learn the complex relationship between historical filling strategies, environmental information, and damage amount differences, so as to achieve accurate prediction of the damage amount difference of real-time filling strategies.

[0090] S33 Prediction, obtain input data, including real-time filling strategies and real-time environmental information. The filling strategy information can be obtained through historical records during previous fillings, and the real-time environmental information can be collected in real time by sensors.

[0091] Input the input data into the trained neural network model, and the trained neural network model will calculate the predicted damage amount difference corresponding to each real-time filling strategy based on the learned knowledge.

[0092] S34 Difference judgment, sequentially judge whether each predicted damage amount difference is greater than the remaining damage amount. If the predicted damage amount difference is greater than the remaining damage amount, it means that this real-time filling strategy is likely to cause damage to the gas cylinder exceeding its bearing capacity, posing a safety risk. Therefore, mark this real-time filling strategy as an unavailable strategy. Otherwise, execute S35 Feedback.

[0093] S35 Feedback, mark the real-time filling strategy corresponding to the predicted damage amount difference as an available strategy, and send the available strategy to the staff. Set the priority of the available strategy according to the genetic algorithm, and associate the priority with the available strategy to generate a strategy priority list. The staff selects the available strategy with the highest priority from this list to fill the gas cylinder.

[0094] Set the priority of the available strategy according to the genetic algorithm. The genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanisms. It can encode and evaluate various factors of the available strategy (such as filling time, cost, safety, etc.), and continuously iterate and optimize to obtain the optimal strategy priority. The process is as follows: Encode each available strategy. For example, it can be represented by the number or feature vector of the real-time filling strategy. These encodings constitute the individual population in the genetic algorithm. The fitness function is used to evaluate the quality of each individual (i.e., the real-time filling strategy). In this embodiment, the fitness function can be designed according to multiple factors, such as the length of the filling time, the magnitude of the predicted damage, etc. For example, a real-time filling strategy with a shorter filling time and a smaller predicted damage has a higher fitness. Calculate the fitness value of each individual according to the fitness function, and select individuals from the population for reproduction according to a certain selection strategy (such as roulette wheel selection, tournament selection, etc.). Individuals with higher fitness have a greater probability of being selected to ensure that excellent characteristics can be inherited to the next generation. Perform a crossover operation on the selected individuals, that is, exchange part of their encodings to generate new individuals. The crossover operation can simulate the gene recombination process in biological inheritance and increase the diversity of the population. Mutate the newly generated individuals with a certain probability, that is, randomly change part of the encoding of the individual. The mutation operation can introduce new characteristics and prevent the algorithm from falling into a local optimal solution. Repeat the above operations until the preset number of iterations is reached or the convergence condition is satisfied (such as the fitness value no longer increases significantly). The real-time filling strategy corresponding to the individual with the highest fitness in the finally obtained population is the optimal real-time filling strategy or the real-time filling strategy with a higher priority.

[0095] Associate the priority with the available strategies to obtain an association result. The staff selects the available strategy with the highest priority to fill the gas cylinder according to the association result to improve the filling efficiency and safety.

[0096] After performing the S35 feedback, the method further includes: S36 Strategy Judgment. This step is used to determine whether there are available strategies. This is because during the filling process, due to the influence of various factors (such as environmental changes, increased damage to the gas cylinder, etc.), the originally available strategies may become unavailable. The judgment basis is mainly the list of available strategies generated in the S35 feedback. If the list is empty, it means that there are no available strategies.

[0097] If there are available strategies, perform S14 Damage Judgment. If there are no available strategies, send a stop filling signal to the staff. This means that the current filling environment is very dangerous for the gas cylinder, and continuing to fill may cause the gas cylinder to be damaged or even lead to a safety accident.

[0098] This embodiment improves the safety of the filling process through S36 Strategy Judgment. If there are available strategies, perform S14 Damage Judgment to further evaluate the safety of gas cylinder filling. If there are no available strategies, send a stop filling signal to the staff to ensure the safety of the staff.

[0099] By adopting the above solution, this embodiment realizes the intelligent selection of the gas cylinder filling strategy, effectively reducing the risk during the filling process of the gas cylinder.

[0100] Embodiment 4: Refer to Figure 4 , the difference between this embodiment and Embodiment 1 is that the method further includes: S41 Gas cylinder identification: By identifying the characteristic identifier of the gas cylinder (such as bar code, QR code, RFID tag, etc.), the key information of the gas cylinder is obtained. The characteristic identifier is a unique identification mark preset on the gas cylinder body. Devices such as a barcode scanner and an RFID reader are used to read the characteristic identifier on the gas cylinder body, and it is judged whether the characteristic identifier can be recognized; If so, based on the characteristic identifier, the characteristic information carried by the characteristic identifier is obtained, and S42 filling detection is executed.

[0101] If not, a feedback signal is sent to the staff.

[0102] S42 Filling detection: Set the characteristic matching criterion for the gas cylinder, and use the characteristic matching criterion to detect the characteristic information to obtain a real-time detection result, and judge whether the real-time detection result meets the expectation; If so, the real-time detection result is added to the characteristic information, and the gas cylinder is placed in the filling area waiting for filling.

[0103] If not, a feedback signal is sent to the staff.

[0104] The characteristic matching criterion includes: Validity period detection: Whether the gas cylinder is within the inspection period.

[0105] Pressure detection: Whether the internal pressure of the gas cylinder is within the safe range.

[0106] Usage record detection: Whether the gas cylinder has any illegal usage records (such as overdue inspection, accident records).

[0107] Criterion source: National standards (such as GB / T 15382 "General Technical Conditions for Gas Cylinder Valves") and enterprise internal safety specifications, etc.

[0108] The real-time detection result meeting the expectation means that the characteristic information of the gas cylinder all meets the requirements of the characteristic matching criterion.

[0109] Through the characteristic identifier recognition and filling detection in this embodiment, the safety and compliance of the gas cylinder during the filling process can be improved.

[0110] Embodiment 5: This embodiment discloses a gas cylinder detection system during inflation. The system includes: a memory and a processor, The memory stores a computer-readable storage medium; When the processor processes the computer program stored on the computer-readable storage medium, the method described above is implemented.

[0111] The above are all preferred embodiments of this application, and the protection scope of this application is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of this application shall be covered within the protection scope of this application.

Claims

1. A method for detecting a gas cylinder during inflation, characterized in that, Including: Data acquisition: Obtain the point cloud data before gas cylinder filling, the strain data before filling, and the strain data during filling. Construct a digital twin model: Set the temperature boundary condition and stress boundary condition, construct a point cloud model based on the point cloud data, perform mesh division on the point cloud model to obtain mesh data. Adopt the finite element method, input the mesh data into the temperature boundary condition to obtain the temperature distribution, input the mesh data into the stress boundary condition to obtain the stress distribution; fuse the temperature distribution, stress distribution with the point cloud model after mesh division to obtain the digital twin model. Obtain damage: Draw a strain-time curve based on the strain data during filling, fit the strain-time curve based on the digital twin model and the strain data before filling, and obtain the real-time damage amount of the gas cylinder based on the digital twin model after fitting. Damage judgment: Judge whether the real-time damage amount is greater than the preset damage amount threshold. If so, send an alarm signal; if not, after a preset time interval, re-execute the step of obtaining damage until the filling is completed.

2. The method for detecting a gas cylinder during inflation according to claim 1, wherein In the step of constructing the digital twin model, it also includes: Calculate the stress change amount based on the strain data during filling, and fuse the stress change amount to obtain the digital twin model.

3. The gas cylinder detection method during inflation according to claim 2, wherein, The method also includes: Construct a grid temperature matrix: Extract the temperature and coordinates of each grid node in the digital twin model, record the temperature of the grid node as the grid temperature, and construct a grid temperature matrix according to the grid temperature. Modeling and classification: Map the coordinates of each grid node to the image coordinate system, map the grid temperature to the gray value of the corresponding pixel point to obtain the gas cylinder gray-scale image, construct a CNN-SVM model, the CNN-SVM model includes: a CNN sub-model for feature extraction, an SVM sub-model for classifying the features extracted by the CNN sub-model, input the gas cylinder gray-scale image into the CNN-SVM model to obtain the classification result; analyze the classification result to obtain the analysis result. Leakage judgment: Judge whether the analysis result is leakage. If so, send a leakage signal; if not, after a preset time interval, re-execute the step of constructing the grid temperature matrix until the filling is completed.

4. The method for detecting a gas cylinder during inflation according to claim 3, wherein Before executing the step of modeling and classification after executing the step of constructing the grid temperature matrix, it also includes: Coordinate transformation: Take any grid node as the center point, and convert the coordinates of all grid nodes in the grid temperature matrix to relative coordinates based on the center point. In the step of modeling and classification, map the relative coordinates to the image coordinate system.

5. The method for detecting a gas cylinder during inflation according to claim 4, wherein Before executing the step of modeling and classification after executing the step of coordinate transformation, it also includes: Construct a KD tree: Take the relative coordinates corresponding to the center point as the root node, and take the relative coordinates corresponding to the remaining grid nodes as the child nodes to construct a KD tree, and calculate the Euclidean distance between adjacent node pairs based on the KD tree. Distance judgment: Judge whether the i-th Euclidean distance is greater than the preset Euclidean distance threshold. If so, execute the step of calculating interpolation; if not, execute the step of updating the distance. Calculation interpolation: Determine the relative coordinates corresponding to the i-th Euclidean distance and the node pairs in the KD tree. Add new nodes between the node pairs in the KD tree, and use the linear interpolation algorithm to determine the relative coordinates corresponding to the new nodes, and perform the step of updating the distance; Update distance: Update the (i + 1)-th Euclidean distance to the i-th Euclidean distance, and perform the step of distance judgment until all Euclidean distances are traversed; Update KD tree: Update the new nodes to the KD tree to obtain a new KD tree, calculate the Euclidean distances between adjacent node pairs in the new KD tree, and perform the step of distance judgment; In the steps of modeling and classification, map the relative coordinates of each node in the new KD tree to the image coordinate system.

6. The gas cylinder detection method during inflation according to any one of claims 1-5, characterized in that, Before performing the step of data acquisition, the method further includes: Damage acquisition: Obtain the cumulative damage amount of the current gas cylinder before filling and the maximum damage amount that the same type of gas cylinder can withstand. Calculate the remaining damage amount based on the cumulative damage amount and the maximum damage amount, and obtain historical data, where the historical data includes: the historical filling strategies of the same type of gas cylinders, the environmental information during filling, and the damage amount difference corresponding to each historical filling strategy; Modeling and training: Establish a neural network model, and use the historical data to train the neural network model to obtain a trained neural network model; Prediction: Obtain input data, where the input data includes multiple real-time filling strategies and real-time environmental information. Input the input data into the trained neural network model to obtain the predicted damage amount difference for each real-time filling strategy; Difference judgment: Sequentially judge whether each predicted damage amount difference is greater than the remaining damage amount. If so, mark the real-time filling strategy corresponding to the predicted damage amount difference as an unavailable strategy; if not, perform the feedback step; Feedback: Mark the real-time filling strategy corresponding to the predicted damage amount difference as an available strategy, and send the available strategy to the staff.

7. The method for detecting a gas cylinder during inflation according to claim 6, characterized in that, The feedback step further includes: Set the priority of the available strategies according to the genetic algorithm, and associate the priority with the available strategies to generate a strategy priority list. The staff selects the available strategy with the highest priority from the strategy priority list to fill the gas cylinder.

8. The method for detecting a gas cylinder during inflation according to claim 7, wherein After performing the feedback step, the method further includes: Strategy judgment: Judge whether there are available strategies. If so, perform the damage judgment step; if not, send a stop filling signal to the staff.

9. The method for detecting a gas cylinder during inflation according to claim 1, characterized in that, The method further includes: Gas cylinder identification: Obtain the characteristic identifier of the gas cylinder, where the characteristic identifier is a unique identification mark preset on the gas cylinder body, and judge whether the characteristic identifier can be recognized; If so, based on the characteristic identifier, obtain the characteristic information carried by the characteristic identifier; If not, send a feedback signal to the staff; Filling detection: Set the characteristic matching criterion of the gas cylinder, and use the characteristic matching criterion to detect the characteristic information to obtain a real-time detection result, and judge whether the real-time detection result meets the expectation; If so, add the real-time detection result to the characteristic information, and place the gas cylinder in the filling area waiting for filling; If not, send a feedback signal to the staff.

10. A gas cylinder detection system during inflation, characterized in that, Includes: A memory and a processor, The memory stores a computer-readable storage medium; When the processor processes the computer program stored on the computer-readable storage medium, the method according to any one of claims 1-9 is implemented.

Images