A gas cylinder analysis method based on multi-source data

Abstract

The present application relates to the field of gas cylinder management and safety monitoring, and discloses a gas cylinder analysis method based on multi-source data, comprising: obtaining gas filling order information and gas cylinder historical gas filling information, generating a historical gas filling information sequence and determining a first loss rate of each gas cylinder; screening gas cylinders with a loss rate reaching a threshold value to form a target gas cylinder set, and obtaining customer operation information thereof; performing feature grouping based on the customer operation information, calculating the proportion of each feature group in the target gas cylinder set, and screening feature groups with a proportion reaching a threshold value as target feature groups; performing in-cylinder loss connectivity verification on the target feature group gas cylinders, counting the proportion of gas cylinders with loss connectivity, determining feature groups with a proportion reaching a threshold value as accelerated loss feature groups, and adjusting the detection period of each gas cylinder in the feature groups. Thus, the safety management of the gas cylinders is realized.

Description

Technical Field

[0001] This invention relates to the field of gas cylinder management and safety monitoring, specifically to a gas cylinder analysis method based on multi-source data. Background Technology

[0002] In existing technologies, industrial gas cylinders, as high-pressure containers, are widely used in industries such as chemical, manufacturing, medical, and specialty gases. To ensure safe use, the industry generally adopts a fixed-cycle system for inspection and disposal management, such as conducting regular inspections every year or two, and uniformly disposing of them when they reach their design life.

[0003] The following technical problems frequently exist in the existing gas cylinder management process:

[0004] First, existing gas cylinder testing methods are mainly based on fixed cycles, lacking dynamic assessment of changes in gas cylinder usage and actual wear and tear, making it difficult to achieve optimal allocation of safety management and testing resources.

[0005] Second, the existing management of gas cylinders and valves relies on periodic inspections and experience-based judgments, making it difficult to dynamically identify potentially damaged or abnormally risky gas cylinders, and also making it impossible to accurately adjust inspection cycles and replacement strategies, thus making it difficult to achieve accurate and safe maintenance and risk control.

[0006] Third, existing technologies typically rely solely on fixed time intervals or usage volume to assess equipment lifespan, lacking dynamic quantitative monitoring of actual usage status and cumulative wear and tear. This makes it difficult to identify anomalies in a timely manner, and the lifespan determination lacks traceability and structured management, leading to inaccurate assessments and increasing safety risks and maintenance uncertainties. Summary of the Invention

[0007] The summary section of this invention provides a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0008] This invention proposes a gas cylinder analysis method based on multi-source data to solve one or more of the technical problems mentioned in the background section above.

[0009] This invention provides a gas cylinder analysis method based on multi-source data, comprising: acquiring gas replenishment order information; acquiring multiple gas cylinder identifiers based on the gas replenishment order information; for each gas cylinder identifier, acquiring multiple historical gas replenishment information corresponding to it, and generating a historical gas replenishment information sequence; and determining a first loss rate corresponding to each gas cylinder identifier based on the historical gas replenishment information sequence.

[0010] From multiple cylinder identifiers, filter out cylinder identifiers whose first loss rate reaches the first threshold and summarize them to obtain a target cylinder identifier set; obtain the customer operation information corresponding to each cylinder identifier in the target cylinder identifier set;

[0011] Based on customer operation information, the cylinder identifiers in the target cylinder identifier set are grouped by features to obtain multiple feature groups; the proportion of the number of cylinder identifiers corresponding to each feature group in the target cylinder identifier set is calculated; if the proportion reaches a second threshold, at least one corresponding feature group is determined as the target feature group.

[0012] For each gas cylinder corresponding to the gas cylinder identifier in the target feature group, perform the intra-cylinder loss connectivity check to obtain the check result. The check result is characterized as whether there is loss connectivity or no loss connectivity.

[0013] The proportion of gas cylinders with loss connectivity in the target feature group is counted. If the proportion reaches the third threshold, the target feature group is determined as the accelerated loss feature group; the detection cycle of each gas cylinder in the accelerated loss feature group is adjusted.

[0014] Optionally, each historical gas replenishment information sequence includes a cylinder identifier, timestamp, internal rapid inspection information, and external rapid inspection information. The internal rapid inspection information includes the equilibrium pressure value after gas replenishment, and the external rapid inspection information includes a crack information sequence. The crack information sequence includes multiple crack information entries, and each crack information entry includes a crack identifier and a crack coordinate sequence.

[0015] Based on the historical gas replenishment information sequence, determine the first loss rate corresponding to each gas cylinder identifier, including:

[0016] For each gas cylinder identifier, based on multiple crack information sequences corresponding to multiple historical gas replenishment information, the crack information in the multiple crack information sequences is grouped according to the crack identifier to obtain multiple crack information groups. Each crack information group corresponds to a crack identifier. For each crack information group, the corresponding crack propagation degree is determined according to the timestamp order and based on the crack coordinate sequence. Based on the multiple crack propagation degrees corresponding to the multiple crack information groups, the crack propagation level corresponding to each gas cylinder identifier is determined.

[0017] Based on multiple historical gas replenishment information corresponding to multiple gas replenishment equilibrium pressure values, according to the preset theoretical gas quantity pressure range table and the preset theoretical gas quantity static pressure range table, multiple gas replenishment pressure offsets are determined, and the multiple gas replenishment pressure offsets are sorted according to the timestamp to generate a gas replenishment pressure offset sequence. The gas replenishment pressure offset sequence includes the gas replenishment pressure offset at the head of the queue and the gas replenishment pressure offset at the tail of the queue.

[0018] Based on the pressure offset after replenishment at the front and rear of the column, the pressure offset level after replenishment corresponding to the gas cylinder identifier is determined; based on the crack propagation level and the pressure offset level after replenishment, the first loss rate corresponding to each gas cylinder identifier is determined.

[0019] Optionally, for each gas cylinder corresponding to each gas cylinder identifier in the target feature group, perform an intra-cylinder loss connectivity check to obtain the check result, including:

[0020] Acquire an endoscopic image of the gas cylinder, extract multiple pitting locations from the endoscopic image, and determine the number of cracks corresponding to each pitting location; determine the pitting locations with a crack number greater than or equal to a preset threshold as target pitting locations.

[0021] For a target pitting location, determine the distance between any two target pitting locations; if the distance between at least one pair of target pitting locations is less than or equal to a preset distance threshold, and the trend of the cracks corresponding to at least one pair of target pitting locations satisfies the connectivity condition, it is determined that the gas cylinder has lossy connectivity; otherwise, it is determined that the gas cylinder does not have lossy connectivity.

[0022] Optionally, at least one pair of cracks corresponding to target pitting locations satisfy the connectivity condition, including:

[0023] For two target pitting locations, select at least one corresponding crack for each location;

[0024] Determine whether at least one crack has a mutually approaching extension direction; if it does, then determine that the trend of at least one pair of cracks corresponding to the target pitting locations satisfies the connectivity condition.

[0025] Optionally, the gas cylinder analysis method based on multi-source data of the present invention further includes:

[0026] From the gas replenishment order information, filter out the cylinder identifiers with abnormal valve replacement frequency within a preset statistical period to obtain the abnormal valve replacement cylinder sequence; identify each cylinder identifier in the abnormal valve replacement cylinder sequence as an abnormal cylinder identifier.

[0027] Obtain the valve replacement information and abnormal endoscope images corresponding to the abnormal gas cylinder identification. Based on the valve replacement information, determine the abnormal replacement level corresponding to the abnormal gas cylinder identification. Analyze the abnormal endoscope images to determine the corrosion level and loss connectivity of the cylinder mouth corresponding to the abnormal gas cylinder identification.

[0028] Based on the abnormal replacement level, the corrosion level of the cylinder neck, and the loss connectivity, the cylinder neck loss level corresponding to the abnormal gas cylinder identifier is determined according to the preset rule table. The cylinder neck loss level is low, medium, or high.

[0029] Optionally, the gas cylinder analysis method based on multi-source data of the present invention further includes:

[0030] For multiple abnormal gas cylinders with high cylinder head loss levels, identify the corresponding multiple risk characteristics. Each risk characteristic includes gas type, gas cylinder type, cylinder valve type, and customer operation information.

[0031] Based on multiple risk characteristics, the gas replenishment order information is matched to obtain multiple successfully matched gas cylinder identifiers;

[0032] For each successfully matched gas cylinder identifier among multiple successfully matched gas cylinder identifiers, obtain the corresponding pressure offset level after gas replenishment and determine it as the pressure offset level after matching and gas replenishment.

[0033] Optionally, the gas cylinder analysis method based on multi-source data of the present invention further includes:

[0034] If the pressure deviation level after matching and replenishing gas reaches the preset deviation threshold, the special endoscopic inspection data of the corresponding gas cylinder is received; the special endoscopic inspection data is analyzed to obtain the analysis results, which include the number of special pitting corrosion, the number of special cracks, and the special loss connectivity corresponding to the successfully matched gas cylinder identifier;

[0035] Based on the analysis results, the corresponding replacement information for gas cylinders and valves is generated, or the detection cycle of gas cylinders and valves is adjusted.

[0036] Optionally, the inspection cycle of the gas cylinder and its valve can be adjusted, including:

[0037] Based on the number of gas replenishment times corresponding to each successfully matched gas cylinder identifier, the pressure deviation level after matching and replenishment, the number of specific pitting corrosion, the number of specific cracks, and the connectivity of specific losses, the detection cycle adjustment amount is determined.

[0038] The updated detection cycle is determined based on the original preset detection cycle and the adjustment amount of the detection cycle.

[0039] The present invention has the following beneficial effects:

[0040] 1. Dynamic, precise, and hierarchical management of gas cylinder safety status has been achieved. Specifically, by acquiring gas replenishment order information and analyzing historical gas replenishment information sequences, the crack propagation level and pressure deviation level after gas replenishment are calculated, thereby quantifying the actual wear status of the gas cylinder. This effectively identifies high-wear cylinders and reduces the risk of missed detection. Subsequently, high-wear cylinders are grouped by feature based on customer operation information, and the proportion of each feature group is statistically analyzed to achieve hierarchical management of cylinders under different usage environments and operating conditions, ensuring that high-risk cylinders can be effectively identified and classified. Furthermore, by analyzing the pitting location and crack trend of cylinders in the target feature group, it is determined whether the wear is continuous, achieving precise identification of the internal wear structure of the cylinder and improving the ability to detect potential safety hazards. Finally, the detection cycle is dynamically adjusted according to the proportion of cylinders in the accelerated wear feature group. The detection interval is shortened for high-wear cylinders to reduce the risk of missed detection, while the detection frequency is reduced for low-wear cylinders to avoid resource waste. This optimizes the overall cylinder detection strategy and improves management efficiency and safety.

[0041] 2. Reduced safety hazards and improved detection and maintenance efficiency. Specifically, based on gas replenishment order information, cylinders with abnormal valve replacement frequency within a preset statistical period are identified, generating abnormal cylinder identifiers to achieve dynamic detection and tracking of potentially high-risk cylinders. Subsequently, the abnormal replacement level is determined based on the valve replacement records, and the corrosion level and loss connectivity of the cylinder mouth are analyzed in conjunction with abnormal endoscopic images to achieve a quantitative assessment of the cylinder interface and internal condition. Furthermore, the abnormal replacement level, cylinder mouth corrosion level, and loss connectivity are mapped to a unified cylinder mouth loss level, and multiple sets of risk features are generated for high-risk cylinders. The corresponding cylinders are matched in the gas replenishment order information to achieve dynamic tracking and identification of high-risk cylinders. Finally, the pressure deviation level after gas replenishment of the matched cylinders is obtained, and when the deviation reaches a preset threshold, specialized endoscopic data is further analyzed to obtain the number of pitting corrosion, cracks, and loss connectivity, thereby generating cylinder and valve replacement information or adjusting the detection cycle. This enables precise maintenance and preventive intervention based on actual loss and risk status, effectively reducing safety hazards and improving detection and maintenance efficiency.

[0042] 3. Improved cylinder safety and management efficiency. Specifically, by constructing a feature group template library, cylinders are classified according to gas type, cylinder type, valve type, and customer operation information, achieving standardized classification of cylinder usage conditions. Furthermore, through statistical analysis of multiple rounds of gas replenishment and loss data from sample cylinders, the loss patterns of each feature group are quantified, and a template loss rate is generated, establishing a feature group template loss benchmark table to provide a reference for dynamic cylinder life assessment. Combining the cylinder's internal control lifespan with the target template loss rate, updated cylinder lifespan information is generated, enabling dynamic lifespan management under different usage conditions and actual loss states. Finally, when the updated internal control lifespan of a cylinder reaches a preset threshold, cylinder scrapping information is automatically generated, enabling timely elimination and safe management of high-risk cylinders. This solves the problems of inaccurate lifespan assessment and difficulty in controlling potential safety risks in existing cylinder management. This method can accurately reflect the aging state of cylinders under different usage conditions, improve the ability to identify abnormal losses, and avoid premature or delayed scrapping, thereby significantly improving the safety, scientific nature, and operational efficiency of cylinder management. Attached Figure Description

[0043] The above and other features, advantages, and aspects of the various embodiments of the present invention will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0044] Figure 1 This is a flowchart of a gas cylinder analysis method based on multi-source data according to the present invention;

[0045] Figure 2 This is a flowchart of the calculation process for the first loss rate of a gas cylinder analysis method based on multi-source data according to the present invention.

[0046] Figure 3 This is a flowchart of the gas cylinder life update process of a gas cylinder analysis method based on multi-source data according to the present invention. Detailed Implementation

[0047] The invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the drawings and embodiments of the invention are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0048] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0049] It should be noted that the concepts of "first" and "second" mentioned in this invention are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0050] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0051] The names of messages or information exchanged between the various devices of this invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0052] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0053] like Figure 1 The diagram shows a flowchart of a gas cylinder analysis method based on multi-source data according to the present invention. Specifically, it includes the following steps:

[0054] Step 101: Obtain gas replenishment order information; obtain multiple gas cylinder identifiers based on the gas replenishment order information; for each gas cylinder identifier, obtain the corresponding multiple historical gas replenishment information and generate a historical gas replenishment information sequence; determine the first loss rate corresponding to each gas cylinder identifier based on the historical gas replenishment information sequence.

[0055] In some embodiments, the execution entity of the gas cylinder analysis method based on multi-source data of the present invention can be a backend server. Based on this, the backend server obtains gas replenishment order information within a target time range through a database query interface and stores the results as an internal data structure (such as a list or table). The gas replenishment order information is an electronic data unit recording gas cylinder replenishment operations, which can be obtained from the gas cylinder management system database, the front-end interactive page, or cloud storage. The gas replenishment order information includes a gas replenishment order identifier and a gas cylinder identifier. The gas replenishment order identifier is a unique number used to identify a specific gas replenishment operation. The gas cylinder identifier is a unique number used to identify the gas cylinder participating in the replenishment, and is used to associate the gas cylinder's historical replenishment data, detection records, and operation information. Next, the backend server extracts the gas cylinder identifier field from each gas replenishment order information, storing multiple gas cylinder identifiers as a set or list for easy access. Subsequently, the backend server queries the historical replenishment database through the gas cylinder identifier to obtain all historical records of the gas cylinder, i.e., multiple historical replenishment information. The retrieved historical replenishment information is then sorted by timestamp to obtain a historical replenishment information sequence. The historical gas replenishment information sequence is a collection of historical gas replenishment information for this gas cylinder, arranged chronologically. This historical replenishment information records past replenishment operations for each cylinder, including cylinder identification, timestamp, internal and external rapid inspection information, gas type, and remaining gas level before replenishment. Internal rapid inspection information records the rapid measurement of the cylinder's internal state, including the post-replenishment equilibrium pressure value. The post-replenishment equilibrium pressure value is the pressure value at which the cylinder reaches equilibrium after being replenished to the target pressure and allowed to stand. External rapid inspection information records external wear information for the cylinder, including a crack information sequence. This sequence contains multiple crack entries, each including a crack identifier and a crack coordinate sequence. The crack information sequence is a collection of cracks detected during each replenishment. Crack information refers to the data for a single crack, including the crack identifier and crack coordinate sequence (the spatial location of the crack on the cylinder surface).

[0056] Based on this, such as Figure 2 The diagram illustrates a flowchart of the calculation process for the first loss rate in a gas cylinder analysis method based on multi-source data according to the present invention. Specifically, based on the historical gas replenishment information sequence, the first loss rate corresponding to each gas cylinder identifier is determined, including:

[0057] Step 1: For each gas cylinder identifier, based on multiple crack information sequences corresponding to multiple historical gas replenishment information, group the crack information in the multiple crack information sequences according to the crack identifier to obtain multiple crack information groups. Each crack information group corresponds to a crack identifier. For each crack information group, determine the corresponding crack propagation degree according to the timestamp order and based on the crack coordinate sequence. Based on the multiple crack propagation degrees corresponding to the multiple crack information groups, determine the crack propagation level corresponding to each gas cylinder identifier.

[0058] In some embodiments, the backend server traverses the crack information sequence of each historical gas replenishment information, classifies data with the same crack identifier, and generates crack information groups. A crack information group refers to a collection of morphological feature data collected at different time points for a specific crack on the same gas cylinder. For each crack information group, the backend server sorts the crack information within the group by timestamp. It then calculates the change in crack coordinates (such as crack endpoints or pixel positions) at consecutive time points. If the crack coordinate sequence is a two-dimensional plane coordinate (x, y), the length can be calculated using Euclidean distance. The expansion amount is the current length minus the previous length. The expansion amount is then compared with a preset threshold to determine the degree of crack expansion, such as low, medium, or high expansion. Crack expansion degree describes the growth of a crack over time, including changes in length, width, or area. Next, the backend server weights or averages the expansion degrees of all crack information groups for the same gas cylinder and generates the overall crack expansion level of the gas cylinder according to a preset rule table. The crack propagation level reflects the rate of deterioration of the overall safety condition of the gas cylinder's outer wall, with possible levels including low propagation, medium propagation, and high propagation. As an example, suppose there is a crack, R1, on gas cylinder A001. The crack lengths collected in three historical gas replenishment data points are 10mm, 12mm, and 16mm, respectively, and their coordinate sequences are recorded as two-dimensional plane points. The backend server categorizes these three records into the same crack information group, R1, and sorts them by timestamp: 2025-01-01, 2025-02-01, and 2025-03-01. Next, the propagation amount is calculated; for example, the propagation amount from February to January is 12 minus 10, resulting in 2mm; similarly, the propagation amount from March to February is calculated as 4mm. Then, the degree of propagation is determined: propagation amounts greater than 3mm are classified as high propagation, and propagation amounts less than 3mm are classified as medium propagation. If there are other cracks on the gas cylinder, the backend server will weight or average the expansion degree of each crack to generate the overall crack expansion level of the gas cylinder. For example, high expansion indicates that the safety condition of the outer wall is deteriorating rapidly.

[0059] Step 2: Based on multiple historical gas replenishment information corresponding to multiple post-gas replenishment equilibrium pressure values, and according to the preset theoretical gas quantity pressure range table and the preset theoretical gas quantity static pressure range table, determine multiple post-gas replenishment pressure offsets, and sort the multiple post-gas replenishment pressure offsets according to the timestamp to generate a post-gas replenishment pressure offset sequence. The post-gas replenishment pressure offset sequence includes the first post-gas replenishment pressure offset and the last post-gas replenishment pressure offset.

[0060] In some embodiments, the backend server iterates through the historical gas replenishment information corresponding to the same gas cylinder identifier and reads and extracts multiple post-replenishment equilibrium pressure values ​​from each historical gas replenishment information. For each post-replenishment equilibrium pressure value, the backend server uses the post-replenishment equilibrium pressure value as an index to find the pressure range it falls into in a preset theoretical gas volume pressure range table, and determines the corresponding theoretical gas volume based on the pressure range. Then, it uses the theoretical gas volume as an index to find the corresponding theoretical static pressure range in a theoretical gas volume static pressure range table. The deviation between the post-replenishment equilibrium pressure value and the theoretical static pressure range is calculated. If the post-replenishment equilibrium pressure value is within the theoretical static pressure range, the offset is 0 or close to 0; if the post-replenishment equilibrium pressure value is lower or higher than the range, the difference in deviation from the range boundary is calculated. Finally, the post-replenishment pressure offset corresponding to each gas replenishment is obtained. Finally, the backend server sorts the multiple post-replenishment pressure offsets of the same gas cylinder from earliest to latest according to the timestamp of the corresponding historical gas replenishment information to obtain a post-replenishment pressure offset sequence. The backend server marks the earliest offset in the sequence as the pressure offset after replenishment at the head of the queue, and the most recent offset as the pressure offset after replenishment at the tail of the queue. The preset theoretical gas volume pressure range table is a table or database that records the pressure value range of the gas cylinder under different theoretical replenishment volumes, used to map the actual replenishment pressure value to the corresponding theoretical gas volume. The preset theoretical gas volume static pressure range table records the pressure value range of the gas cylinder after each theoretical gas volume has reached static equilibrium, used to evaluate the deviation between the actual replenishment pressure and the theoretical pressure. The replenishment pressure offset is the deviation between the actual pressure and the theoretical static pressure range for each replenishment, used to quantify the pressure deviation of the gas cylinder. The replenishment pressure offset sequence is a sequence of pressure offsets from multiple replenishments of the same gas cylinder arranged in ascending order of timestamps, used to evaluate the pressure change trend. The replenishment pressure offset at the head of the queue is the pressure offset of the earliest replenishment in the replenishment pressure offset sequence. The replenishment pressure offset at the tail of the queue is the pressure offset of the most recent replenishment in the replenishment pressure offset sequence.

[0061] Step 3: Determine the pressure offset level after gas replenishment corresponding to the gas cylinder identifier based on the pressure offset after gas replenishment at the head of the column and the pressure offset level after gas replenishment; determine the first loss rate corresponding to each gas cylinder identifier based on the crack propagation level and the pressure offset level after gas replenishment.

[0062] In some embodiments, the backend server reads the pressure offsets at the head and tail of the gas cylinder. It calculates the percentage change in offset, such as by dividing the absolute value of the difference between the tail offset and the head offset by the head offset, then multiplying by 100%. This percentage change is compared to a corresponding threshold to determine the corresponding pressure offset level after gas replenishment. Subsequently, the crack propagation level and the pressure offset level after gas replenishment are read. A table is used to map the crack propagation level and the pressure offset level to numerical values, which are then weighted and calculated to obtain the first loss rate. The pressure offset level after gas replenishment is a classification of the gas cylinder pressure change based on the changing trends of the head and tail offsets. The pressure offset level after gas replenishment can also be categorized as low offset, medium offset, or high offset. The first loss rate quantifies the early loss state of the gas cylinder; a higher first loss rate indicates more severe loss.

[0063] Step 102: From multiple gas cylinder identifiers, filter out gas cylinder identifiers whose first loss rate reaches the first threshold and summarize them to obtain a target gas cylinder identifier set; obtain the customer operation information corresponding to each gas cylinder identifier in the target gas cylinder identifier set;

[0064] In some embodiments, the first threshold is a preset first loss rate judgment standard used to filter cylinders with high loss. It can be a fixed value (e.g., 0.6) or dynamically calculated based on historical data, such as taking the top 20% of the loss rate distribution as the threshold. Based on this, the backend server iterates through each cylinder identifier and its first loss rate, comparing the first loss rate with the first threshold. Cylinder identifiers with a first loss rate reaching the first threshold are filtered out and stored as a set. The target cylinder identifier set is the set of all cylinder identifiers whose first loss rate reaches or exceeds the first threshold. Based on this, the backend server iterates through each cylinder identifier in the target set, obtaining customer operation information corresponding to each cylinder through a database or interface. This operation information is then associated with and stored with the cylinder identifier. The customer operation information records the operational behavior data and environmental data during cylinder use, including the cylinder's location, temperature and humidity data, and valve switching frequency. The cylinder's location is either a fixed or mobile gas supply room. The temperature and humidity data are the temperature and humidity of the environment where the cylinder is located. The valve switching frequency is the number or frequency of valve switching during cylinder use.

[0065] Step 103: Based on the customer's operation information, the cylinder identifiers in the target cylinder identifier set are grouped by features to obtain multiple feature groups; the proportion of the number of cylinder identifiers corresponding to each feature group in the target cylinder identifier set is calculated; if the proportion reaches the second threshold, at least one corresponding feature group is determined as the target feature group.

[0066] In some embodiments, the backend server iterates through the target cylinder identifier set. For each cylinder, it categorizes the target cylinders based on key features in the customer's operation information (usage location, temperature and humidity, valve switching frequency). Cylinders in the same group are similar in features. For example, feature group G1: fixed gas supply chamber, temperature 25°C to 30°C, valve switching once daily; feature group G2: mobile gas supply chamber, temperature 30°C to 35°C, valve switching three times daily. Each feature group contains a set of several cylinder identifiers. If a combination does not yet exist, a new feature group is created. Then, the number of cylinders in each feature group is counted and divided by the total number of target cylinder identifiers to obtain the percentage. For each feature group's percentage, if the percentage is greater than or equal to a second threshold, it is marked as a target feature group. Otherwise, it is excluded. In practice, at least one target feature group is output. If multiple feature groups have percentages reaching the threshold, all can be marked. The second threshold is determined based on the statistical results of the cylinder feature distribution in historical gas replenishment orders and is used to determine whether a feature group has a high percentage, thus determining whether it should be used as a target feature group. The target feature group is the feature group whose proportion reaches the second threshold.

[0067] Step 104: For each gas cylinder corresponding to the gas cylinder identifier in the target feature group, perform the intra-cylinder loss connectivity check to obtain the check result. The check result indicates whether there is loss connectivity or no loss connectivity.

[0068] Specifically, for each gas cylinder corresponding to each gas cylinder identifier in the target feature group, an intra-cylinder loss connectivity check is performed to obtain the check results, including:

[0069] Acquire an endoscopic image of the gas cylinder, extract multiple pitting locations from the endoscopic image, and determine the number of cracks corresponding to each pitting location; determine the pitting locations with a crack number greater than or equal to a preset threshold as target pitting locations.

[0070] In some embodiments, the backend server invokes an endoscope database or real-time acquisition interface via the gas cylinder identifier to obtain an intra-cylinder endoscopic image of the gas cylinder. This intra-cylinder endoscopic image refers to a high-resolution optical image or video frame captured by inserting an industrial endoscope into the gas cylinder cavity, recording the surface condition of the cylinder's inner wall for identifying pitting and cracks. Subsequently, the backend server uses a pre-trained deep learning model (such as Mask R-CNN or U-Net) to identify abnormal areas on the inner wall surface, identifying all pitting areas and marking their center coordinates. Using each pitting center as the origin, it detects linear pixel sets (cracks) within a preset radius area, counting the number of cracks connected to or adjacent to the pitting pit. Each pitting location includes spatial coordinate information (two-dimensional or three-dimensional) and the area size. The crack count refers to the actual number of cracks detected at that pitting location, used to assess the severity of local damage. The preset threshold is a value used to determine whether pitting has the risk of evolving into severe cracks, and can be preset based on the correspondence between the number of cracks and the degree of damage in historical endoscopic detection results. For example, a threshold of 3 means that when a pitting corrosion site is surrounded by three or more microcracks, the stress concentration at that pitting corrosion site is considered severe. Next, pitting corrosion locations with a crack number greater than or equal to a preset threshold are identified as target pitting corrosion locations. For instance, if the preset threshold is 3, only pitting corrosion locations with a crack number greater than or equal to 3 are selected as target pitting corrosion locations. Target pitting corrosion locations refer to screened damage points with a high risk of cracking.

[0071] For a target pitting location, determine the distance between any two target pitting locations; if the distance between at least one pair of target pitting locations is less than or equal to a preset distance threshold, and the trend of the cracks corresponding to at least one pair of target pitting locations satisfies the connectivity condition, it is determined that the gas cylinder has lossy connectivity; otherwise, it is determined that the gas cylinder does not have lossy connectivity.

[0072] In some embodiments, the backend server pairs the selected target pitting locations and calculates their spatial geometric distance. Specifically, Euclidean distance is calculated using pixel coordinates and converted into actual physical distance by combining the magnification and calibration parameters of the endoscope. If the distance between any two points is less than or equal to a preset distance threshold, the crack trend judgment is initiated. The preset distance threshold refers to the warning distance between two pitting locations that affect each other, and is determined based on the structural dimensions of the gas cylinder's inner wall and the spatial distribution characteristics of pitting in historical detection data. Based on this, for each of the two target pitting locations, at least one corresponding crack is selected. The coordinates of the crack ends (away from the pitting direction) are obtained, and the Euclidean distance between the ends of the two cracks is calculated. If the Euclidean distance is less than or equal to the preset endpoint distance threshold, it is determined that at least one crack has a mutually approaching extension direction, and the trend of at least one pair of cracks corresponding to the target pitting locations satisfies the connectivity condition.

[0073] Step 105: Calculate the proportion of gas cylinders with loss connectivity in the target feature group. If the proportion reaches the third threshold, the target feature group is determined as the accelerated loss feature group; adjust the detection cycle of each gas cylinder in the accelerated loss feature group.

[0074] In some embodiments, the backend server iterates through each cylinder identifier in the target feature group. The statistical verification result is the number of cylinders with loss connectivity. Dividing the statistical number by the total number of cylinders in the target feature group yields the proportion of loss connectivity. If the proportion of loss connectivity is greater than or equal to a third threshold, the target feature group is marked as an accelerated loss feature group. Otherwise, the group does not belong to the accelerated loss group. The third threshold is determined based on the proportion of loss connectivity phenomena in different cylinder groups in historical detection data; it is a preset proportion judgment standard used to determine whether the feature group belongs to the accelerated loss risk. The accelerated loss feature group is a set of cylinders with high loss risk and significant pitting and crack connectivity, requiring priority management and monitoring. Based on this, the detection cycle for cylinders corresponding to the cylinder identifiers in the accelerated loss feature group is shortened from the original preset detection cycle, such as shortening the detection cycle from once every three months to once a month. The adjusted detection cycle and the corresponding cylinder identifier are sent to the detection terminal for execution by the corresponding detection personnel.

[0075] These embodiments achieve dynamic, precise, and hierarchical management of gas cylinder safety status. Specifically, by acquiring gas replenishment order information and analyzing historical gas replenishment information sequences, the crack propagation level and pressure deviation level after gas replenishment are calculated, thereby quantifying the actual wear status of the gas cylinder. This effectively identifies high-wear cylinders and reduces the risk of missed detection. Subsequently, high-wear cylinders are grouped by feature based on customer operation information, and the proportion of each feature group is statistically analyzed to achieve hierarchical management of cylinders under different usage environments and operating conditions, ensuring that high-risk cylinders can be effectively identified and classified. Furthermore, by analyzing the pitting location and crack trend of cylinders in the target feature group, it is determined whether the wear is continuous, achieving precise identification of the internal wear structure of the cylinder and improving the ability to detect potential safety hazards. Finally, the detection cycle is dynamically adjusted according to the proportion of cylinders in the accelerated wear feature group. The detection interval is shortened for high-wear cylinders to reduce the risk of missed detection, while the detection frequency is reduced for low-wear cylinders to avoid resource waste, thereby optimizing the overall cylinder detection strategy and improving management efficiency and safety.

[0076] In some embodiments, to further address the second technical problem described in the background section, namely, "existing gas cylinder and valve management relies on periodic inspections and experience-based judgments, making it difficult to dynamically identify potentially damaged or abnormally risky gas cylinders, and also making it impossible to accurately adjust inspection cycles and replacement strategies, thus hindering accurate and safe maintenance and risk control," in some embodiments of the present invention, a gas cylinder analysis method based on multi-source data further includes:

[0077] Step 1: From the gas replenishment order information, filter out the cylinder identifiers with abnormal valve replacement frequency within a preset statistical period to obtain the abnormal valve replacement cylinder sequence; identify each cylinder identifier in the abnormal valve replacement cylinder sequence as an abnormal cylinder identifier.

[0078] In some embodiments, the backend server reads gas replenishment order information and extracts the cylinder identification field. Identical cylinder identifications appearing in different gas replenishment orders are grouped together to form a cylinder identification set. For each cylinder identification, the corresponding valve replacement record is queried, and then the timestamp in the valve replacement record is used to determine whether it falls within a preset statistical period. The valve replacement records falling within this period are counted, obtaining the number of valve replacements for that cylinder identification within the preset statistical period. A pre-set abnormal valve replacement frequency threshold is established, and the statistically obtained valve replacement frequency is compared with this threshold. If the number of valve replacements exceeds the pre-set abnormal valve replacement frequency threshold, the cylinder identification is determined to have an abnormal valve replacement frequency. The pre-set abnormal valve replacement frequency threshold is based on a preset rule of historical maintenance frequency. All cylinder identifications determined to have abnormal valve replacement frequency are added to the same data set, which is the abnormal valve replacement cylinder sequence. The preset statistical period is the time window used to analyze valve replacement behavior, and can be a fixed time period or a rolling time period. The number of cylinder valve replacements refers to the cumulative number of valve replacement operations performed on the same cylinder identifier within a preset statistical period. An abnormal number of valve replacements refers to a situation where the number of valve replacements for a particular cylinder within the preset statistical period significantly deviates from the normal maintenance frequency. The abnormal valve replacement cylinder sequence is a set of cylinder identifiers that meet the abnormal replacement criteria. An abnormal cylinder identifier refers to each cylinder identifier in the abnormal valve replacement cylinder sequence. As an example, the backend server can iterate through cylinder identifiers using SQL queries, API interfaces, or local data tables to count the number of valve replacements for each cylinder within the statistical period. For example, setting a threshold of twice per month, if cylinder A001 has 3 valve replacements in that month, then A001 is marked as an abnormal cylinder identifier. The maintenance cycle of the cylinder valve is independent of the cylinder; if the valve is replaced abnormally, it indicates that the cylinder also has a problem.

[0079] Step 2: Obtain the valve replacement information and abnormal endoscope images corresponding to the abnormal gas cylinder identification. Based on the valve replacement information, determine the abnormal replacement level corresponding to the abnormal gas cylinder identification. Analyze the abnormal endoscope images to determine the corrosion level and loss connectivity of the cylinder mouth corresponding to the abnormal gas cylinder identification.

[0080] In some embodiments, the backend server uses the abnormal cylinder identifier as a query condition to search the cylinder valve maintenance database for all cylinder valve replacement records associated with that cylinder identifier, and extracts the corresponding cylinder valve replacement information fields, including the cylinder valve identifier, total number of replacements (e.g., 3 times per year), total number of replacements for each reason (e.g., 1 time for corrosion, 2 times for wear), and the timestamp corresponding to each replacement. The replacement reason can be corrosion replacement or wear replacement. The number of cylinder valve replacements within a preset period is compared with the preset number of replacements to determine the abnormal replacement level corresponding to the abnormal cylinder identifier. For example, 5 replacements within a year is determined to be a medium replacement level. Here, cylinder valve replacement information refers to structured data that records the cylinder valve replacement behavior, used to reflect the cylinder valve maintenance or replacement situation that occurs during the use of the cylinder. Next, the backend server obtains high-definition optical images of the cylinder opening and top of the cylinder, i.e., abnormal endoscopic images, from the endoscope database or real-time acquisition interface through the cylinder identifier. Abnormal endoscopic images include cylinder opening images (for corrosion analysis) and cylinder top images (for pitting and crack analysis). After preprocessing the bottle neck image, a deep learning segmentation model (such as Mask R-CNN or U-Net) is used to identify the corrosion region, or traditional image processing methods (threshold segmentation and edge detection) are used to extract corrosion pixels. The proportion of corrosion pixels to the total number of pixels at the bottle neck is calculated, and the corrosion proportion is compared with a preset corrosion level range to obtain the bottle neck corrosion level (low corrosion / medium corrosion / high corrosion). The preset corrosion level range is a division used to classify the degree of corrosion at the bottle neck; it is not arbitrarily set but determined based on historical detection data, the specific characteristics of the gas cylinder structure, and safety management requirements, through preset rules. Subsequently, pitting corrosion detection is performed on the top image of the gas cylinder, and the coordinates of the pitting corrosion center are marked. Linear pixel sets (cracks) are detected within each pitting corrosion region, generating a crack coordinate sequence corresponding to the top image of the gas cylinder, and the number of corresponding cracks is counted. Pitting corrosion locations where the number of cracks corresponding to the top image of the gas cylinder is greater than or equal to a preset threshold are determined as the target pitting corrosion locations corresponding to the top image of the gas cylinder. For each pair of target pitting locations corresponding to the top image of the gas cylinder, the Euclidean distance between pixel coordinates is calculated and converted into actual physical distance using the endoscope magnification. If the distance between any two points is less than a preset distance threshold, the crack connectivity is assessed. The coordinates of the corresponding crack's end (away from the pitting direction) are selected, and the distance between the ends is calculated. If the distance between the ends is less than the endpoint distance threshold, the cracks are considered to be approaching each other, indicating connectivity. If at least one pair of target pitting locations meets the above conditions, the gas cylinder is considered to have lost connectivity; otherwise, it is considered not to exist. Here, abnormal endoscope images refer to image data of the top or neck area of ​​the gas cylinder acquired by an endoscope detection device for the gas cylinder corresponding to the abnormal gas cylinder identifier. The abnormal replacement level is a grade result used to characterize the degree of abnormality in the valve replacement of the abnormal gas cylinder within a preset statistical period, categorized as low replacement, medium replacement, or high replacement.The bottle neck corrosion rating is a graded result used to characterize the degree of corrosion in the bottle neck area. This rating is determined based on the corrosion characteristics of the bottle neck in abnormal endoscopic images. Loss connectivity is used to characterize whether there is a risk of spatial interconnection or tendency to penetrate the local losses in the bottle neck area or inside the bottle; loss connectivity is either yes or no.

[0081] Step 3: Based on the abnormality replacement level, the corrosion level of the cylinder neck, and the loss connectivity, determine the cylinder neck loss level corresponding to the abnormal gas cylinder identifier according to the preset rule table. The cylinder neck loss level is low, medium, or high.

[0082] In some embodiments, the preset rule table refers to a set of mapping rules pre-stored in the backend server database, used to integrate multiple risk dimensions and output a unified loss level. This table specifies the final loss level corresponding to different combinations of three variables: abnormal replacement level, cylinder neck corrosion level, and loss connectivity. For each abnormal cylinder identifier, the backend server reads its corresponding abnormal replacement level, cylinder neck corrosion level, and loss connectivity, using these three indicators as a set of comprehensive evaluation input parameters. It then matches the corresponding rule item in the preset rule table and reads the predefined cylinder neck loss level from that rule item. Each rule in the preset rule table includes at least the abnormal replacement level, cylinder neck corrosion level, loss connectivity, and the corresponding cylinder neck loss level output. The cylinder neck loss level obtained through the rule table matching is bound to the corresponding abnormal cylinder identifier and stored as the cylinder neck loss evaluation result for that abnormal cylinder. The cylinder neck loss level characterizes the comprehensive loss risk level of the cylinder neck area and is classified as low, medium, or high. By simultaneously introducing the dimensions of behavioral anomaly, structural corrosion, and loss connectivity risk, the determination of the bottle neck loss level avoids relying solely on a single detection indicator, thereby reducing the risk of misjudgment and improving the reliability of bottle neck loss assessment.

[0083] Step 4: For multiple abnormal gas cylinders with high cylinder head loss levels, determine the corresponding multiple risk characteristics. Each risk characteristic includes gas type, gas cylinder type, cylinder valve type, and customer operation information.

[0084] Step 5: Based on multiple risk characteristics, match them in the gas replenishment order information to obtain multiple successfully matched gas cylinder identifiers;

[0085] Step 6: For each successfully matched gas cylinder identifier among multiple successfully matched gas cylinder identifiers, obtain the corresponding pressure offset level after gas replenishment and determine it as the pressure offset level after matching and replenishment.

[0086] In some embodiments, the backend server first iterates through abnormal cylinder identifiers with high cylinder head wear levels. For each abnormal cylinder identifier with a high cylinder head wear level, it obtains its corresponding gas type, cylinder type, valve type, and customer operation information, and combines these four types of information to form a set of risk characteristics. This process is repeated for multiple abnormal cylinder identifiers to obtain multiple sets of risk characteristics. Specifically, the gas type characterizes the type of gas filled in the cylinder and can be obtained from refill order information or cylinder file information. The cylinder type characterizes the structural specifications or usage category of the cylinder, such as cylinder types for different purposes or with different design parameters. The valve type characterizes the structure or model type of the valve installed at the cylinder head, which is related to the sealing method and stress state of the cylinder head area. Customer operation information describes the operational behavior characteristics related to cylinder use, refilling, and maintenance, and is derived from refill order information or customer operation records. Subsequently, the backend server iterates through multiple sets of risk characteristics. For each set, it uses the gas type, cylinder type, valve type, and customer operation information as matching conditions to search the gas replenishment order information. When the gas replenishment order information corresponding to a certain abnormal cylinder identifier simultaneously meets the above matching conditions, the cylinder identifier is determined to be a successfully matched cylinder identifier. All cylinder identifiers that meet the matching conditions of any risk characteristic set are aggregated to obtain multiple successfully matched cylinder identifiers. A successfully matched cylinder identifier refers to the cylinder identifier corresponding to a cylinder whose attribute information in the gas replenishment order information meets at least one set of risk characteristic matching conditions. Finally, for each successfully matched cylinder identifier, the backend server retrieves its corresponding historical gas replenishment information and reads or calculates its post-replenishment pressure offset level. The obtained post-replenishment pressure offset level is used as the matched post-replenishment pressure offset level for the successfully matched cylinder and associated with the corresponding cylinder identifier for storage. The post-replenishment pressure offset level refers to the post-replenishment pressure offset level obtained for the successfully matched cylinder identifier and used as the basis for its subsequent risk assessment. In step four, attribute information such as gas type is read from the gas replenishment order information corresponding to the abnormal gas cylinder identifier to construct risk characteristics; in step five, based on the risk characteristics, other gas cylinders in the gas replenishment order information are matched to identify potential risk gas cylinders that are similar to the abnormal gas cylinders in terms of attribute characteristics.

[0087] Step 7: If the pressure deviation level after matching and replenishing gas reaches the preset deviation threshold, then receive the special endoscopic inspection data of the corresponding gas cylinder; analyze the special endoscopic inspection data to obtain the analysis results, which include the number of special pitting corrosion, the number of special cracks, and the special loss connectivity corresponding to the successfully matched gas cylinder identifier;

[0088] In some embodiments, a preset deviation threshold is a criterion used to determine whether the pressure deviation after gas replenishment reaches an abnormal level. It distinguishes between normal pressure deviation and abnormal pressure states with potential loss risks, and can be preset based on the statistical distribution of historical gas replenishment pressure data. The backend server compares the pressure deviation level after gas replenishment corresponding to the successfully matched cylinder identifier with the preset deviation threshold. When the pressure deviation level after gas replenishment is not lower than the preset deviation threshold, the backend server establishes a communication connection with the endoscopic inspection device or system, thereby sending a specific inspection request to the endoscopic inspection device or system and receiving specific endoscopic inspection data collected for the successfully matched cylinder. Specifically, the specific endoscopic inspection data refers to image or video data obtained by directional inspection of the inside or mouth area of ​​a specific cylinder using an endoscopic device after triggering inspection conditions. Image processing and feature recognition are performed on the received specific endoscopic inspection data to extract pitting corrosion features and crack features. Based on the analysis of specialized endoscopic inspection data, the number of pitting and cracks in the corresponding gas cylinders is counted. The spatial relationship between pitting and cracks is used to determine loss connectivity, thus forming the analysis results. By introducing specialized endoscopic inspection on top of the abnormal pressure deviation after gas replenishment, high-cost inspection of all matching gas cylinders can be avoided, thereby reducing inspection resource consumption while ensuring the accuracy of risk identification. The judgment logic for specialized loss connectivity is the same as described above; specialized loss connectivity is also either yes or no. The number of specialized pitting is the number of pitting locations identified in the specialized endoscopic inspection images. The number of specialized cracks is the number of crack features identified in the specialized endoscopic inspection images. The analysis results refer to the structural loss quantification information obtained after processing the specialized endoscopic inspection data, used to characterize the actual loss state inside the gas cylinder or at the cylinder neck area.

[0089] Step 8: Based on the analysis results, generate corresponding replacement information for gas cylinders and valves, or adjust the inspection cycle of gas cylinders and valves. Adjusting the inspection cycle of gas cylinders and valves includes: determining the inspection cycle adjustment amount based on the number of gas replenishment cycles corresponding to each successfully matched gas cylinder identifier, the pressure deviation level after matching and replenishment, the number of specific pitting corrosions, the number of specific cracks, and the connectivity of specific losses; and determining the updated inspection cycle based on the original preset inspection cycle and the inspection cycle adjustment amount.

[0090] In some embodiments, when the analysis results indicate that there is loss connectivity between specific pitting corrosion or cracks, or the number of specific cracks reaches a preset abnormal condition, or the number of specific pitting corrosions reaches a preset number of pitting corrosions, the gas cylinder or valve is determined to be unsuitable for continued use, and corresponding replacement information is generated. When the analysis results do not trigger the above replacement conditions, the gas cylinder or valve is determined to still be usable. The replacement information is management information used to indicate the replacement of gas cylinders or valves with a high risk of loss, including the specific gas cylinder identifier to be replaced, the valve identifier, the reason for replacement (e.g., high loss, crack risk, etc.), and the recommended handling method. The preset abnormal condition is a rule threshold pre-set based on historical operating data, inspection experience, or safety regulations. The preset number of pitting corrosions is pre-set based on historical inspection data.

[0091] Simultaneously, each successfully matched cylinder identifier is traversed to collect risk-related indicator data, including the number of times the cylinder has been replenished, the pressure deviation level after replenishment, the number of specific pitting corrosions, the number of specific cracks, and crack connectivity determination. These indicators are then mapped to a numerical risk score. For example, a weighted score can be preset for each indicator, and a weighted sum can be calculated to obtain the overall cylinder risk score. A preset risk table is stored on a backend server or in a database. This table maps the cylinder risk score range to the corresponding inspection cycle adjustment. The preset risk table includes risk score intervals and corresponding inspection cycle adjustment amounts. Subsequently, the risk scores are matched against the preset risk table. After finding the corresponding score interval, the corresponding inspection cycle adjustment amount is read from the table. For example, a higher total score indicates a greater cylinder risk, and the inspection cycle should be shortened; a lower total score allows for a more appropriate extension of the inspection cycle. As an example, the indicators for successfully matching cylinder identifier B002 are: 4 replenishment times, high pressure deviation level, 6 specific pitting corrosion counts, 3 specific crack counts, and "yes" specific loss connectivity. After mapping, the total score is 90. According to the preset risk table, the corresponding inspection cycle adjustment is -3 months (i.e., inspection is performed 3 months earlier than the original cycle). When determining the updated inspection cycle, the backend server adds or subtracts the calculated inspection cycle adjustment from the original preset inspection cycle to obtain the updated inspection cycle. For example, if the original preset cycle is 12 months and the adjustment is -3 months, the updated inspection cycle is 9 months. Subsequently, the updated inspection cycle is associated with the cylinder identifier and valve identifier and stored in the database, providing a basis for the next round of inspection or maintenance planning.

[0092] These embodiments reduce safety hazards and improve detection and maintenance efficiency. Specifically, based on gas replenishment order information, cylinders with abnormal valve replacement frequency within a preset statistical period are identified, generating abnormal cylinder identifiers to achieve dynamic discovery and tracking of potentially high-risk cylinders. Subsequently, the abnormal replacement level is determined based on the valve replacement record, and the corrosion level and loss connectivity of the cylinder mouth are analyzed in conjunction with abnormal endoscopic images to achieve a quantitative assessment of the cylinder interface and internal condition. Furthermore, the abnormal replacement level, cylinder mouth corrosion level, and loss connectivity are mapped to a unified cylinder mouth loss level, and multiple sets of risk features are generated for high-risk cylinders. The corresponding cylinders are matched in the gas replenishment order information to achieve dynamic tracking and identification of high-risk cylinders. Finally, the pressure deviation level after gas replenishment of the matched cylinder is obtained, and when the deviation reaches a preset threshold, specialized endoscopic data is further analyzed to obtain the number of pitting corrosion, cracks, and loss connectivity, thereby generating cylinder and valve replacement information or adjusting the detection cycle. This enables precise maintenance and preventive intervention based on actual loss and risk status, effectively reducing safety hazards and improving detection and maintenance efficiency.

[0093] In some embodiments, to further address the third technical problem described in the background section, namely, "existing technologies typically rely solely on fixed time or usage to assess equipment lifespan, lacking dynamic quantitative monitoring of actual usage and cumulative wear, making it difficult to identify anomalies in a timely manner, and resulting in inaccurate assessments and increased safety risks and maintenance uncertainties," some embodiments of the present invention construct a feature group template library and, based on the feature group template library, dynamically adjust the lifespan of gas cylinders corresponding to cylinder identifiers, including:

[0094] Step 1: For different combinations of gas type, cylinder type, valve type and customer operation information, construct multiple feature group templates and store them in the feature group template library;

[0095] In some embodiments, the backend server reads the gas type field corresponding to the cylinder identifier from the gas replenishment order information as a gas type feature. Based on the cylinder identifier, it queries the corresponding cylinder type field in the cylinder basic information table. Based on the cylinder identifier, it reads the corresponding cylinder valve type field from the most recent cylinder valve installation or replacement record. Based on the gas replenishment order information and historical gas replenishment information, it performs statistical analysis on customer operation behavior to generate corresponding customer operation information features. On this basis, the backend server standardizes the above four types of information according to a preset field structure, for example, mapping textual attributes to enumerated values ​​or category tags. Subsequently, the backend server concatenates the above four types of information according to a combination method to generate multiple different feature combinations. For example, the backend server can use "Gas Type A + Cylinder Type B + Cylinder Valve Type C + Customer Operation Type D" as one feature group template, and "Gas Type A + Cylinder Type B + Cylinder Valve Type E + Customer Operation Type D" as another feature group template. For each feature combination, a unique template identifier is generated. The gas type, cylinder type, valve type, and customer operation information within that feature combination are used as template fields, forming a feature group template data structure. The template content can include a template identifier, gas type field, cylinder type field, valve type field, and customer operation information field. Finally, the template identifier and corresponding feature fields are written to a template data table, saved as a file to the template library path, or loaded into an in-memory template library for real-time matching. During storage, the template identifier undergoes uniqueness verification to prevent duplicate templates. By constructing a feature group template library, gas cylinders under different usage environments and structural conditions are classified and modeled, providing a unified feature benchmark for subsequent statistical analysis of template loss rates based on sample gas cylinders and dynamic adjustment of cylinder lifespan. A feature group template refers to a set of features that abstractly describe a class of gas cylinders with the same or similar usage environments and structural conditions. A feature group template is not a single gas cylinder, but rather an abstraction of the characteristics of a group of gas cylinders, used to characterize the common usage and loss environment of a class of gas cylinders. The feature group template library is a data set used to store multiple feature group templates. The feature group template library can be a database table, a file system, or an in-memory data structure, and each feature group template has a unique template identifier in the feature group template library.

[0096] Step 2: For each feature group template in the feature group template library, select a preset number of initial gas cylinders as sample gas cylinders, obtain the loss data of the sample gas cylinders in multiple rounds of gas replenishment, and determine the corresponding gas cylinder loss rate based on the loss data.

[0097] In some embodiments, the backend server reads the feature combination corresponding to a certain feature group template, and filters the cylinder identifiers that meet the feature combination from the cylinder archives and gas replenishment order information; from the filtering results, cylinders whose current status is abnormal or scrapped are removed. If the number of remaining cylinders is greater than or equal to a preset number, the preset number of cylinders are selected according to time sequence, random method, or rotation strategy. If the number of remaining cylinders is less than the preset number, all cylinders that currently meet the conditions are used as sample cylinders. The preset number refers to the sample size set in advance for statistical analysis before or during operation, which can be set according to historical data volume, business scale, or empirical rules. The initial cylinders refer to cylinders that meet the basic usability conditions under the same feature group template and have not been judged as abnormal or scrapped at the beginning of the statistical stage. The backend server can filter the cylinder identifiers that are currently in use and have not been marked as abnormal or scrapped from the historical cylinder archives as initial cylinders. The sample cylinders are selected from the initial cylinders to form a representative set of cylinders used to statistically analyze the loss pattern under a certain feature group template. The initial gas cylinders form the candidate set, while the sample gas cylinders form the final statistical set. Based on this, for each sample gas cylinder, its historical gas replenishment information sequence is read according to the cylinder identifier, and the historical gas replenishment information is sorted chronologically. Then, in chronological order, the crack propagation level, pressure shift level after gas replenishment, and structural loss information related to pitting and cracks are read from the data corresponding to each round of gas replenishment. Subsequently, the above data is used to form the loss data sequence for that gas cylinder according to the number of gas replenishment rounds. A multi-round gas replenishment process refers to multiple gas replenishment operations performed on the same gas cylinder over time. Each round of gas replenishment corresponds to one historical gas replenishment record. Multi-round gas replenishment is used to reflect the loss change trend of the gas cylinder during continuous use. Loss data refers to the data set related to changes in the gas cylinder's structure or sealing performance during multi-round gas replenishment. Loss data includes at least the crack propagation level, pressure shift level after gas replenishment, and the structural loss state reflected by pitting, cracks, and loss connectivity. The loss data is obtained in the same way as the loss characteristics described earlier, such as through endoscopic image analysis, gas replenishment data statistics, or historical detection record reading. Finally, the backend server first standardizes the loss data, converting crack propagation level, post-gas replenishment pressure deviation level, and structural loss status level into corresponding equivalent loss values ​​according to preset dimension conversion rules or weighting rules for different types of loss indicators. Subsequently, the backend server weights and summarizes these equivalent loss values ​​to obtain the cumulative loss of the sample cylinder within a preset statistical period. For example, a corresponding expansion risk weight can be set for the crack propagation level, a corresponding pressure-bearing performance loss weight can be set for the post-gas replenishment pressure deviation level, and a structural weight reflecting the risk of loss connectivity can be set for the structural loss status. This converts loss information with different dimensions and physical meanings into a unified loss representation, yielding the cumulative loss of the sample cylinder within the preset statistical period.The preset statistical period is a time window set by the backend server according to a uniform time scale or usage frequency scale. For example, it could be the natural time period between the most recent detection and the current detection (e.g., 6 months, 12 months), or a usage period corresponding to several refill operations, to ensure the comparability of loss data between different sample gas cylinders. Subsequently, the backend server calculates the gas cylinder loss rate based on the cumulative loss and the length of the corresponding preset statistical period. The gas cylinder loss rate characterizes the rate or extent of cumulative loss of a gas cylinder within a unit statistical period. Specifically, the backend server can divide the cumulative loss by the length of the statistical period to obtain the loss growth value per unit time or unit usage period. Alternatively, in the case of multiple historical detection data, the loss growth slope can be calculated by comparing the changes in cumulative loss within adjacent statistical periods, and this slope can be used as the gas cylinder loss rate. For example, for a sample gas cylinder, if its crack propagation level has increased from low to medium within the past 12 months, its pressure deviation level after gas replenishment has increased from normal to abnormal, and its structural wear status has evolved from a state without connectivity risk to a state with a trend of wear connectivity, the cumulative wear amount obtained after standardization mapping and weighting is 8 equivalent wear units. The backend server can then divide these 8 equivalent wear units by 12 months to obtain the cylinder wear rate of 0.67 equivalent wear units / month, reflecting the aging trend of the cylinder under actual use conditions. Through this method, the backend server can generate a corresponding cylinder wear rate value for each sample gas cylinder. This cylinder wear rate serves as a basic quantitative indicator for subsequent wear pattern analysis, comparison with similar cylinders, risk level classification, and dynamic adjustment of the inspection cycle, ensuring that subsequent maintenance decisions have an objective, calculable, and reproducible basis.

[0098] Step 3: Statistically process the gas cylinder loss rates of multiple sample gas cylinders under the same feature group template to obtain the template loss rate corresponding to the feature group template, and bind the template loss rate with the corresponding feature group template to construct the feature group template loss benchmark table.

[0099] In some embodiments, the backend server first reads the cylinder loss rate values ​​of all sample cylinders belonging to the same feature group template from the database, forming a corresponding loss rate set. Then, the backend server performs statistical calculations on the loss rate set according to preset statistical rules to obtain a template loss rate that represents the overall loss level of the feature group template. The statistical rules may include, but are not limited to, calculating the arithmetic mean, weighted average, median, or quantile values ​​of the loss rate set. The weighting coefficients can be set based on the detection reliability of the sample cylinders, data integrity, or the length of the statistical period to avoid abnormal samples having an excessive impact on the statistical results. For example, for a certain feature group template, there are 10 sample gas cylinders belonging to it. The backend server obtains the gas cylinder loss rates corresponding to these 10 sample gas cylinders as 0.45, 0.52, 0.60, 0.48, 0.55, 0.50, 0.62, 0.47, 0.53, and 0.58 equivalent loss units / month, respectively. The backend server can average these loss rates according to preset rules to obtain the template loss rate of the feature group template, which is 0.53 equivalent loss units / month. This is used to characterize the typical loss level of this type of gas cylinder under the same usage conditions and structural configuration. After obtaining the template loss rate, the backend server binds and stores the template loss rate with the corresponding feature group template. That is, it establishes a template loss rate field in the database for each feature group template, corresponding to its unique template identifier, and records the template loss rate as the loss benchmark value of the feature group template. Therefore, the backend server constructs a feature group template loss benchmark table. Each record includes at least the feature group template identifier, template condition parameters, statistical sample size, and corresponding template loss rate, serving as a benchmark reference for subsequent individual cylinder loss comparison, abnormal loss identification, and lifespan assessment. Through this method, the backend server can establish stable and reproducible loss benchmarks for different feature group templates based on the statistical results of similar sample cylinders, providing a unified reference basis for subsequently determining whether a single cylinder experiences accelerated loss or abnormal aging. The template loss rate is a reference value used to characterize the overall loss level of this type of cylinder, obtained by statistically processing the cylinder loss rates of multiple sample cylinders under the same feature group template. The feature group template loss benchmark table associates and stores each feature group template with its corresponding template loss rate, forming a benchmark data table for subsequent lifespan adjustment and risk assessment.

[0100] Step 4: For each cylinder identifier among multiple cylinder identifiers, obtain the corresponding cylinder life information, which includes the standard life and internal control life.

[0101] In some embodiments, such as Figure 3The diagram illustrates a flowchart of the cylinder life update process in a multi-source data-based cylinder analysis method according to the present invention. Specifically, for each cylinder identifier among multiple cylinder identifiers, the backend server queries the cylinder lifecycle management database using the cylinder identifier to obtain the cylinder lifecycle information for each identifier. The cylinder lifecycle management database centrally stores basic attributes, operating status, and lifecycle management information related to the cylinder entity. The database includes a cylinder basic information table and a cylinder lifecycle management table. The cylinder basic information table stores static parameters that are determined at the factory and, in principle, do not change with use, including cylinder identifier, cylinder type, material grade, design pressure, manufacturing date, and standard life field. The standard life field records the design service life of the cylinder determined according to national standards, industry specifications, or manufacturer technical data at the time of manufacturing or initial filing. The backend server uses the cylinder identifier as the primary key or unique index to read the standard life field of the corresponding record in the cylinder basic information table. The cylinder life management table stores life-related parameters that dynamically change during use, including cylinder identification, current status (in service / pending inspection / discontinued), last inspection date, cumulative refill counts, cumulative loss value, and an internal control life field. The internal control life field is a life parameter dynamically maintained by the backend server based on actual cylinder usage, inspection results, and loss assessment results, reflecting the cylinder's continued safe service life under current operating conditions. The backend server uses the cylinder identification as the key to retrieve the corresponding internal control life field from the cylinder life management table. Finally, the backend server encapsulates the retrieved standard life and internal control life into a set of cylinder life information, binding it to the cylinder identification for subsequent life adjustment calculations. The cylinder life information is a composite parameter representing the safe service life of a cylinder, supporting subsequent life adjustment and scrapping decisions, and includes at least two components: standard life and internal control life. For example, a cylinder with the identification C015 has a standard life of 15 years, fixed from the date of manufacture. The initial internal control lifespan is set to 15 years, consistent with the standard lifespan. During subsequent use, the backend server will adjust the internal control lifespan by decreasing or advancing it based on the template wear rate of the characteristic group to which the cylinder belongs, the actual wear rate of the cylinder itself, and detection results. In this way, the backend server can simultaneously obtain both the static standard lifespan parameters and the dynamic internal control lifespan parameters for each cylinder identifier, thus providing a complete and distinguishable lifespan information foundation for subsequent lifespan updates, risk assessments, and scrapping decisions based on template wear benchmarks and individual wear status.

[0102] Step 5: Determine the previous feature group template corresponding to each gas cylinder identifier in the most recent gas replenishment process, and determine the corresponding template loss rate in the feature group template loss benchmark table, and use it as the target template loss rate.

[0103] In some embodiments, the backend server first uses the cylinder identifier as the query key to retrieve all refueling order records associated with that cylinder identifier from the refueling order database. A refueling order record is a structured data record generated during the refueling process, including the cylinder identifier, refueling timestamp, gas type, cylinder type, valve type, and customer operation information related to the refueling operation. The backend server sorts the retrieved refueling order records by their refueling timestamps, selects the refueling order record with the most recent timestamp, and identifies this refueling order record as the refueling order information corresponding to the most recent refueling process for the cylinder identifier. Subsequently, the backend server reads feature field information for feature group template matching from the refueling order information corresponding to the most recent refueling process, including gas type, cylinder type, valve type, and customer operation information. The backend server combines the read gas type, cylinder type, valve type, and customer operation information to form a complete set of feature parameters. Based on this, the backend server calls multiple pre-built feature group templates stored in the feature group template library to perform matching processing on the feature parameter set. The backend server compares the characteristic parameter set of the gas cylinder identifier with the template characteristic fields in each characteristic group template item by item. When the gas type, gas cylinder type, valve type, and customer operation information all meet the corresponding field conditions of a certain characteristic group template, it is determined that the characteristic group template matches the usage characteristics of the gas cylinder identifier in the most recent gas replenishment process, and this characteristic group template is identified as the previous characteristic group template corresponding to the gas cylinder identifier in the most recent gas replenishment process. After determining the previous characteristic group template, the backend server uses the template identifier of the previous characteristic group template as an index condition to query the characteristic group template loss benchmark table. The backend server finds the template loss rate corresponding to the previous characteristic group template identifier in the characteristic group template loss benchmark table and marks it as the target template loss rate for the gas cylinder identifier. The target template loss rate is a benchmark loss rate obtained by statistically analyzing the gas cylinder loss rates of multiple sample gas cylinders under the same characteristic conditions as the characteristic group template. It is used to reflect the typical loss level of the gas cylinder in actual use under the combined conditions of gas type, gas cylinder type, valve type, and customer operation information. The backend server establishes a storage relationship between the target template loss rate and the corresponding gas cylinder identifier, so that when updating the internal control life of the gas cylinder, analyzing the life trend, or determining its scrapping, the target template loss rate can be directly called as a unified, reusable loss reference parameter with a statistical basis.

[0104] Step 6: Based on the cylinder identification, internal control lifespan, and target template loss rate, generate the updated internal control lifespan corresponding to each cylinder identification. When the updated internal control lifespan is less than or equal to the preset lifespan threshold, generate the cylinder scrapping information corresponding to each cylinder identification.

[0105] Step 7: When the updated internal control lifespan is greater than the preset lifespan threshold, the updated internal control lifespan is combined with the corresponding standard lifespan to generate the updated cylinder lifespan information, and then associated with and stored with each cylinder identifier.

[0106] In some embodiments, when calculating the internal control lifespan update, the backend server first determines a preset lifespan update cycle. The preset lifespan update cycle refers to the time or usage window used by the backend server to uniformly update the internal control lifespan of gas cylinders. It can be a fixed time length (e.g., every month, every three months, or after each gas replenishment), or it can be a cycle triggered by usage events (e.g., a lifespan update triggered after N gas replenishment operations). Based on this, the backend server pre-stores the correspondence between target template loss rates and lifespan adjustment amounts. This correspondence can be represented by a rule table or function. For example, the backend server can preset multiple loss rate ranges and configure a corresponding lifespan adjustment coefficient or lifespan deduction amount for each loss rate range. When the backend server determines the target template loss rate corresponding to a gas cylinder identifier, it maps the target template loss rate to the corresponding loss rate range and reads the lifespan adjustment amount corresponding to that range. Within the preset lifespan update cycle, the backend server deducts the lifespan adjustment amount from the current internal control lifespan to obtain the updated internal control lifespan. The updated internal control lifespan refers to the lifespan parameter obtained by the backend server after dynamically correcting the original internal control lifespan corresponding to the cylinder identifier, combining it with the feature group template matched during the most recent gas replenishment process and its corresponding target template loss rate. This parameter reflects the remaining controllable service life of the cylinder under actual use and wear conditions. The preset lifespan threshold is a pre-configured lower limit parameter, representing the minimum lifespan limit at which the cylinder is no longer suitable for continued use in terms of safety, structural integrity, or economic maintainability. This preset lifespan threshold can be in the form of an absolute lifespan value (e.g., remaining service life less than the preset lifespan) or a relative lifespan ratio (e.g., the remaining percentage of internal control lifespan is lower than the preset percentage of the initial internal control lifespan). The backend server compares the updated internal control lifespan with the preset lifespan threshold one by one. When the updated internal control lifespan is determined to be less than or equal to the corresponding preset lifespan threshold, the cylinder is considered to have reached the scrapping trigger condition. After confirming that the cylinder has reached the scrapping trigger condition, the backend server generates corresponding cylinder scrapping information based on the cylinder identifier. The cylinder scrapping information includes the cylinder identifier, scrapping determination time, the type of lifespan parameter that triggered scrapping (updated internal control lifespan), current lifespan value, corresponding preset lifespan threshold, and a description of the reason for scrapping. The reason for scrapping can indicate that the cylinder was subjected to a high-loss characteristic group template for an extended period, causing the internal control lifespan to decay rapidly over multiple consecutive lifespan update cycles, thus reaching the lifespan threshold prematurely. The backend server can also associate the cylinder scrapping information with the characteristic group template, target template loss rate, and corresponding loss level of the most recent gas replenishment process for subsequent scrapping cause tracing, statistical analysis, or model calibration.The generated cylinder scrapping information is stored in the scrapping management database by the back-end server, and a one-to-one mapping relationship is established with the corresponding cylinder identifier. This enables the system to automatically mark the corresponding cylinder identifier as unusable based on the cylinder scrapping information during subsequent gas replenishment scheduling, testing plan generation, or inventory management, thus preventing cylinders that have reached the scrapping conditions from continuing to participate in operation or circulation.

[0107] In some embodiments, when the updated internal control lifespan is determined to be greater than the corresponding preset lifespan threshold, the backend server combines the calculated updated internal control lifespan with the standard lifespan corresponding to the cylinder identifier to generate updated cylinder lifespan information. The updated cylinder lifespan information includes at least two parameters: the standard lifespan and the updated internal control lifespan. These parameters simultaneously reflect the theoretical upper limit of the cylinder's lifespan within the regulatory limits and its dynamic remaining lifespan under actual usage conditions. By storing the standard lifespan and the updated internal control lifespan side-by-side, the backend server can simultaneously reference both types of lifespan constraints during subsequent management, thereby preventing the internal control lifespan adjustment results from exceeding safety or regulatory limits. Finally, in the cylinder lifespan management database, the backend server uses the cylinder identifier as the primary key or index field to write the updated internal control lifespan and the standard lifespan into the corresponding lifespan parameter fields, and updates or versions the existing lifespan records to ensure the traceability of lifespan information. Through this associated storage method, the backend server can directly call the latest cylinder lifespan information corresponding to the cylinder identifier during subsequent inspection cycle adjustments, scrapping determinations, risk warnings, or maintenance decisions, achieving continuous and dynamic management of the cylinder lifespan status.

[0108] These embodiments improve cylinder safety and management efficiency. Specifically, by constructing a feature group template library, cylinders are classified according to gas type, cylinder type, valve type, and customer operation information, achieving standardized classification of cylinder usage conditions. Furthermore, through statistical analysis of multiple rounds of gas replenishment and loss data from sample cylinders, the loss patterns of each feature group are quantified, and a template loss rate is generated, establishing a feature group template loss benchmark table to provide a reference for dynamic cylinder life assessment. Combining the cylinder's internal control lifespan with the target template loss rate, updated cylinder lifespan information is generated, enabling dynamic lifespan management under different usage conditions and actual loss states. Finally, when the updated internal control lifespan of a cylinder reaches a preset threshold, cylinder scrapping information is automatically generated, enabling timely elimination and safe management of high-risk cylinders. This solves the problems of inaccurate lifespan assessment and difficulty in controlling potential safety risks in existing cylinder management. This method can accurately reflect the aging state of cylinders under different usage conditions, improve the ability to identify abnormal losses, and avoid premature or delayed scrapping, thereby significantly improving the safety, scientific nature, and operational efficiency of cylinder management.

[0109] The above description is merely a selection of preferred embodiments of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to specific combinations of the above-described technical features, but also includes other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A gas cylinder analysis method based on multi-source data, characterized in that, include: Obtain gas replenishment order information, and obtain multiple gas cylinder identifiers based on the gas replenishment order information; For each gas cylinder identifier among multiple gas cylinder identifiers, obtain the corresponding multiple historical gas replenishment information and generate a historical gas replenishment information sequence; based on the historical gas replenishment information sequence, determine the first loss rate corresponding to each gas cylinder identifier, wherein each historical gas replenishment information in the historical gas replenishment information sequence includes a gas cylinder identifier, a timestamp, internal rapid inspection information and external rapid inspection information, the internal rapid inspection information includes the equilibrium pressure value after gas replenishment, the external rapid inspection information includes a crack information sequence, the crack information sequence includes multiple crack information, and each crack information includes a crack identifier and a crack coordinate sequence; From multiple cylinder identifiers, cylinder identifiers whose first loss rate reaches a first threshold are selected and aggregated to obtain a target cylinder identifier set; customer operation information corresponding to each cylinder identifier in the target cylinder identifier set is obtained. Based on the customer operation information, the cylinder identifiers in the target cylinder identifier set are grouped by features to obtain multiple feature groups; the proportion of the number of cylinder identifiers corresponding to each feature group in the target cylinder identifier set is calculated; if the proportion reaches a second threshold, at least one corresponding feature group is determined as a target feature group, wherein the features include usage location, temperature and humidity, and valve switching frequency; For each gas cylinder corresponding to the gas cylinder identifier in the target feature group, perform an in-cylinder loss connectivity check to obtain a check result, which indicates whether there is loss connectivity or no loss connectivity. The proportion of gas cylinders with loss connectivity in the target feature group is statistically analyzed. If the proportion reaches a third threshold, the target feature group is identified as an accelerated loss feature group; the detection cycle of each gas cylinder in the accelerated loss feature group is adjusted.

2. The gas cylinder analysis method based on multi-source data according to claim 1, characterized in that, The step of determining the first loss rate corresponding to each gas cylinder identifier based on the historical gas replenishment information sequence includes: For each gas cylinder identifier, based on multiple crack information sequences corresponding to multiple historical gas replenishment information, the crack information in the multiple crack information sequences is grouped according to the crack identifier to obtain multiple crack information groups. Each crack information group corresponds to a crack identifier. For each crack information group, the corresponding crack propagation degree is determined according to the timestamp order and based on the crack coordinate sequence. Based on the multiple crack propagation degrees corresponding to the multiple crack information groups, the crack propagation level corresponding to each gas cylinder identifier is determined. Based on the multiple equilibrium pressure values ​​after gas replenishment corresponding to the multiple historical gas replenishment information, multiple pressure offsets after gas replenishment are determined according to the preset theoretical gas volume pressure range table and the preset theoretical gas volume static pressure range table. The multiple pressure offsets after gas replenishment are sorted according to the timestamp to generate a pressure offset sequence after gas replenishment. The pressure offset sequence after gas replenishment includes the pressure offset after gas replenishment at the head of the queue and the pressure offset after gas replenishment at the tail of the queue. Based on the pressure offset after replenishment at the head of the column and the pressure offset after replenishment at the tail of the column, the pressure offset level after replenishment corresponding to the gas cylinder identifier is determined; based on the crack propagation level and the pressure offset level after replenishment, the first loss rate corresponding to each gas cylinder identifier is determined.

3. The gas cylinder analysis method based on multi-source data according to claim 2, characterized in that, For each gas cylinder corresponding to the gas cylinder identifier in the target feature group, an intra-cylinder loss connectivity check is performed to obtain the check result, including: Acquire an endoscopic image of the gas cylinder, extract multiple pitting locations from the endoscopic image, and determine the number of cracks corresponding to each pitting location; determine the pitting locations with a crack number greater than or equal to a preset threshold as target pitting locations. For the target pitting location, determine the distance between any two target pitting locations; when the distance between at least one pair of target pitting locations is less than or equal to a preset distance threshold, and the trend of the cracks corresponding to the at least one pair of target pitting locations satisfies the connectivity condition, determine that the gas cylinder has lossy connectivity; otherwise, determine that the gas cylinder does not have lossy connectivity.

4. The gas cylinder analysis method based on multi-source data according to claim 3, characterized in that, The trend of the cracks corresponding to at least one pair of target pitting locations satisfies the connectivity condition, including: For two target pitting locations, select at least one corresponding crack for each location; Determine whether the at least one crack has an extension direction that approaches each other; if it has an extension direction that approaches each other, then determine that the trend of at least one pair of cracks corresponding to the target pitting locations satisfies the connectivity condition.

5. The gas cylinder analysis method based on multi-source data according to claim 4, characterized in that, Also includes: From the gas replenishment order information, filter out the cylinder identifiers with abnormal cylinder valve replacement frequency within a preset statistical period to obtain an abnormal cylinder valve replacement cylinder sequence; determine each cylinder identifier in the abnormal cylinder valve replacement cylinder sequence as an abnormal cylinder identifier. Obtain the valve replacement information and abnormal endoscope image corresponding to the abnormal gas cylinder identifier, and determine the abnormal replacement level corresponding to the abnormal gas cylinder identifier based on the valve replacement information. Analyze abnormal endoscopic images to determine the corrosion level and loss connectivity of the cylinder opening corresponding to the abnormal cylinder markings; Based on the abnormal replacement level, the bottle neck corrosion level, and the loss connectivity, the bottle neck loss level corresponding to the abnormal gas cylinder identifier is determined according to a preset rule table. The bottle neck loss level is low, medium, or high.

6. The gas cylinder analysis method based on multi-source data according to claim 5, characterized in that, Also includes: For multiple abnormal gas cylinders with high cylinder head loss levels, identify the corresponding multiple risk characteristics. Each risk characteristic includes gas type, gas cylinder type, cylinder valve type, and customer operation information. Based on the multiple risk characteristics, multiple successfully matched gas cylinder identifiers are obtained by matching them in the gas replenishment order information. For each successfully matched gas cylinder identifier among multiple successfully matched gas cylinder identifiers, obtain the corresponding pressure offset level after gas replenishment and determine it as the pressure offset level after matching and gas replenishment.

7. The gas cylinder analysis method based on multi-source data according to claim 6, characterized in that, Also includes: If the pressure deviation level after the matching and replenishment reaches the preset deviation threshold, then the special endoscopic detection data of the corresponding gas cylinder is received. The specialized endoscopic detection data is analyzed to obtain analysis results, which include the number of specialized pitting corrosion, the number of specialized cracks, and the specialized loss connectivity corresponding to the successfully matched gas cylinder identifier. Based on the analysis results, corresponding replacement information for gas cylinders and valves is generated, or the detection cycle of the gas cylinders and valves is adjusted.

8. The gas cylinder analysis method based on multi-source data according to claim 7, characterized in that, The adjustment of the detection cycle of the gas cylinder and cylinder valve includes: Based on the number of gas replenishment times corresponding to each successfully matched gas cylinder identifier, the pressure deviation level after matching and replenishment, the number of specific pitting corrosion, the number of specific cracks, and the connectivity of specific losses, the detection cycle adjustment amount is determined. The updated detection cycle is determined based on the original preset detection cycle and the adjustment amount of the detection cycle.

Images