Low temperature adiabatic cylinder static evaporation rate test method and system

Abstract

The application discloses a kind of low-temperature adiabatic gas cylinder static evaporation rate test method and system, it is related to low-temperature adiabatic gas cylinder detection technical field.Continuously collect gas bottle internal pressure and temperature data, judge whether gas bottle reaches thermal equilibrium steady state, output test start instruction when steady state condition satisfies;Load preset adaptive control strategy parameters;With preset high sampling frequency continuously collect mass flow, pressure, gas temperature, ambient temperature and ambient pressure;According to ambient temperature and ambient pressure, the measured mass flow is compensated on line, and digital twin simulation is carried out;After the measured time reaches the minimum measurement time, continuously evaluate adaptive termination determination condition;When adaptive termination determination condition is all satisfied, calculate total evaporation mass and average static evaporation rate.Through the front and rear linkage model of multimodal fusion, digital twin online simulation, reinforcement learning adaptive parameter optimization and uncertainty evaluation, shorten test cycle, improve the reliability of test result.

Description

Technical Field

[0001] This invention relates to the field of cryogenic insulated gas cylinder testing technology, and in particular to a method and system for testing the static evaporation rate of cryogenic insulated gas cylinders. Background Technology

[0002] Cryogenic insulated gas cylinders are refillable, mobile pressure vessels primarily used for storing and transporting cryogenic media. Cryogenic media refer to substances that provide a sub-room temperature environment through heat transfer; these are liquefied gases or fluids used in refrigeration or cryogenic processes to obtain and maintain a cryogenic state, such as liquid nitrogen, liquid oxygen, liquid argon, liquid hydrogen, and liquefied natural gas (LNG). The insulation performance of cryogenic insulated gas cylinders directly affects the storage time and transportation safety of cryogenic liquids. Static evaporation rate is the most important technical indicator for measuring the insulation performance of cryogenic insulated gas cylinders. It refers to the percentage of gas mass evaporated due to external heat leakage within 24 hours after the cylinder reaches thermal equilibrium at its rated fill rate.

[0003] Currently, existing methods for testing the static evaporation rate of cryogenic insulated gas cylinders have the following problems: 1. Long test preparation time, relying on fixed settling time or simple thresholds. Traditional testing methods require a settling time of over 48 hours after filling to ensure the gas cylinder reaches thermal equilibrium. This lack of adaptability to gas cylinders of different specifications, insulation structures, and media leads to significant time waste. While some rapid testing methods propose using pressure change rate as a criterion, they only employ discrete thresholds comparing pressure change rate deviations of no more than 10% every 10 minutes. This results in weak disturbance resistance and a lack of physical modeling of the pressure evolution process, making them prone to misjudgments due to accidental fluctuations or sensor drift.

[0004] 2. The testing process is open-loop controlled and lacks adaptive capabilities. After the test starts, existing technologies generally use a fixed duration (usually 24 hours) to measure the evaporation rate. Even if the gas cylinder has already entered a stable evaporation state, it is still necessary to wait for the full duration, resulting in low testing efficiency. At the same time, when encountering abnormalities such as ambient temperature fluctuations, sensor noise, or flow meter zero drift during the measurement period, it cannot be automatically identified and responded to, requiring manual intervention or abandoning the entire test.

[0005] 3. The modules are isolated before and after the test, and the parameters are not linked. The thermodynamic characteristics of the gas cylinders obtained in the steady-state identification stage (such as relaxation time and exponential decay coefficient) are not used to guide the parameter setting in the measurement stage, resulting in a rigid control strategy that cannot adapt to the dynamic differences of different gas cylinders.

[0006] 4. Lack of quantitative evaluation of test result quality. Existing technology only outputs an evaporation rate value without providing any information about data integrity, measurement stability, or uncertainty. This makes it difficult for inspectors to judge the reliability of the results, affecting the authority of the test conclusions.

[0007] 5. Low level of intelligence and inability to self-evolve. Existing testing methods and rules are fixed, making it impossible to learn from historical tests to optimize control strategies, simulate and predict future operating conditions, and proactively respond to unknown anomalies. Summary of the Invention

[0008] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a method and system for testing the static evaporation rate of cryogenic insulated gas cylinders. Through multimodal fusion of front-end and back-end linkage models, online digital twin simulation, reinforcement learning adaptive parameter optimization, and automatic uncertainty assessment, it achieves intelligent and efficient static evaporation rate testing throughout the entire process, shortens the testing cycle, and improves the reliability and traceability of test results.

[0009] On one hand, embodiments of the present invention provide a method for testing the static evaporation rate of a cryogenic insulated gas cylinder, comprising: After the cryogenic insulated gas cylinder is filled, the internal pressure and temperature data of the gas cylinder are continuously collected, the pressure growth rate sequence is calculated, and the gas cylinder is judged to have reached thermal equilibrium steady state based on the pressure growth rate sequence. When the steady state condition is met, a test start command is output. The measurement process is initialized by loading preset adaptive control strategy parameters, which include minimum measurement duration, maximum measurement duration, stability judgment threshold, and drift-free slope threshold. During the measurement, mass flow rate, pressure, gas phase temperature, ambient temperature and ambient pressure are continuously collected at a preset high sampling frequency, and the collected measured data are validated in real time. Based on the measured data that has passed real-time validity verification, the measured mass flow rate is compensated online according to the ambient temperature and ambient pressure, converted to the equivalent mass flow rate under standard environmental conditions, and the measured smooth flow rate sequence is obtained by using the moving average method. Digital twin simulation is performed, using the current measured data as boundary conditions, to simulate the thermodynamic evolution trend of the gas cylinder in a future preset time period in real time, generate a predicted flow rate sequence, and compare the predicted flow rate sequence with the measured smooth flow rate sequence in real time. When the deviation continuously exceeds a preset threshold, an abnormal alarm is triggered. After the measured duration reaches the minimum measurement duration, the adaptive termination criteria are continuously evaluated. The adaptive termination criteria include: within the most recent sliding time window, the relative standard deviation of the smoothed flow rate is less than the stability judgment threshold; the absolute value of the linear fitting slope of the smoothed flow rate within the same window is less than the drift-free slope threshold; and no abnormal alarms are triggered or the ambient temperature fluctuation is within the allowable range during the measurement. When all adaptive termination criteria are met, the total evaporation mass is calculated, and the average static evaporation rate is calculated by combining the initial liquid mass of the gas cylinder and the actual measurement time.

[0010] According to some embodiments of the present invention, determining whether the gas cylinder has reached a thermal equilibrium steady state based on the pressure increase rate sequence includes: The evolution of the pressure growth rate is divided into four characteristic stages: the period of violent relaxation, the period of exponential decay, the period of quasi-steady-state oscillation, and the period of complete steady state. An exponential decay model is used to perform nonlinear least squares fitting on the data during the exponential decay period to obtain the decay rate coefficient, steady-state rate, and goodness of fit. The exponential decay model is: r(t) = r0e -λt + r ss， In the formula, r(t) is the pressure increase rate at time t, λ is the decay rate coefficient, and r0 is the initial rate. ss For steady-state rate, R 2 For goodness of fit; When the system detects that it has entered a fully steady state and continues for multiple consecutive time windows, the goodness of fit is greater than a preset threshold, and the relative deviation between the measured steady-state rate and the theoretical steady-state rate is less than a preset deviation threshold, the steady-state condition is determined to be satisfied.

[0011] According to some embodiments of the present invention, the real-time validity verification of the collected measured data includes: Single-point physical range check to determine whether the values ​​collected by each sensor are within the preset reasonable range. If MM consecutive sampling points exceed the range, it is determined to be an L3 level system fault. Check the rationality of the rate of change, calculate the rate of change of mass flow and the rate of change of pressure. If the rate of change exceeds the theoretical maximum rate of change threshold, mark the data point as a slight disturbance of level L1 and do not terminate the measurement. Multi-sensor consistency verification is based on the real gas equation of state. The equivalent outflow rate is calculated according to the rate of change of pressure and the rate of change of temperature, and compared with the measured flow rate. When the relative deviation exceeds the consistency threshold, it is judged as a severe disturbance of level L2.

[0012] According to some embodiments of the present invention, the online compensation of the measured mass flow rate based on ambient temperature and ambient pressure includes: Temperature compensation is performed; the temperature compensation formula is: ; In the formula, The instantaneous mass flow rate after temperature compensation. To measure instantaneous mass flow rate, The ambient temperature at the current moment. The boiling point of the low-temperature medium. Standard ambient temperature; If a volumetric flow meter is used, pressure compensation is added. The pressure compensation formula is: ; In the formula, The final compensated instantaneous mass flow rate, The instantaneous mass flow rate after temperature compensation. The current atmospheric pressure. Standard atmospheric pressure The absolute temperature under standard conditions. The current vapor phase temperature; The average length of the sliding window adaptively adjusts between 5 and 10 minutes, and the window length increases as the fluctuation of the environment increases.

[0013] According to some embodiments of the present invention, performing digital twin simulation includes: A thermodynamic digital twin model of a gas cylinder is established, which includes: a one-dimensional unsteady-state heat conduction equation for the insulation layer, a phase change kinetic equation for the medium, and a pressure-temperature-density coupling relationship based on the real gas equation of state. Boundary conditions of pressure, temperature, and ambient temperature are synchronously measured every 5 minutes, and the expected mass flow rate sequence for the next 1-2 hours is solved using the finite difference method. The relative deviation between the measured smoothed flow rate and the simulated predicted flow rate is calculated in real time. If the deviation exceeds 15% continuously and lasts for more than 10 minutes, it is determined that there is an unmodeled anomaly, triggering a Level 2 severe disturbance. When the gas cylinder thermodynamic digital twin model predicts that the flow rate fluctuation will be less than 0.5% within the next 30 minutes, the measurement will be terminated early as an auxiliary criterion, even if the statistical conditions within the current sliding window have not been fully met.

[0014] According to some embodiments of the present invention, the calculation of the average static evaporation rate includes: Calculate the cumulative evaporation mass: ; In the formula, The total mass of liquid evaporated during the measurement period. Test startup time, The test ends at this time. The final compensated instantaneous mass flow rate, It is a time differential element; The average static evaporation rate is calculated based on the cumulative evaporated liquid mass and the initial liquid mass during the measurement period. The calculation formula is as follows: ; In the formula, The average static evaporation rate, The total mass of liquid evaporated during the measurement period. The initial liquid mass in the gas cylinder. This refers to the actual measurement duration.

[0015] According to some embodiments of the present invention, after the step of calculating the average static evaporation rate, the method includes: The uncertainty of the evaporation rate measurement results is evaluated based on the GUM method, and the combined standard uncertainty and expanded uncertainty are calculated. To assess Type A uncertainty, the relative standard uncertainty is calculated based on the repeatability of the smoothed flow series during the measurement period. The formula is as follows: ; In the formula, For type A relative standard uncertainty, To smooth the experimental standard deviation of the flow series, This is the average value for smoothing the flow rate; Type B uncertainty assessment is performed, including uncertainty introduced by pressure sensor accuracy, uncertainty introduced by temperature sensor accuracy, uncertainty introduced by mass flow meter accuracy, uncertainty introduced by filling rate measurement error, and uncertainty introduced by medium physical property parameters (density, boiling point, latent heat) error. The sensitivity coefficients of each uncertainty component are calculated according to the measurement model transfer formula, and the combined relative standard uncertainty is calculated using the following formula: ; In the formula, To synthesize the relative standard uncertainty, For type A relative standard uncertainty, Let i be the i-th type B relative standard uncertainty component. The expanded uncertainty is calculated using the following formula: ; In the formula, To expand the uncertainty, k is the coverage factor. For the combined relative standard uncertainty; The output result is: In the formula, The average static evaporation rate, This is the absolute value of the expanded uncertainty.

[0016] According to some embodiments of the present invention, after the step of calculating the average static evaporation rate, the method further includes: Confidence scores are calculated based on data integrity, steady-state maintenance, number of outliers, environmental stability, and model consistency. If the weighted total score is below 60 points, the test is deemed invalid. If the weighted total score is between 60 and 85, the expert collaboration process will be triggered. If the weighted total score is greater than 85 points, it will be automatically recognized.

[0017] According to some embodiments of the present invention, after the step of calculating the average static evaporation rate, the method further includes: The entire test process data is stored as experience samples in the experience pool, and reinforcement learning training is triggered asynchronously to update the adaptive control strategy parameters. The entire test process data includes gas cylinder characteristics, steady-state parameters, measurement process actions, environmental conditions, and test results. The reinforcement learning training employs a deep Q-network or a proximal policy optimization algorithm, whose state space, action space, and reward function are defined as follows: State space: includes cylinder volume, medium type, current test stage, current decay rate coefficient, current ambient temperature fluctuation range, and the ratio of measured duration to predicted stabilization duration; Action space: including incremental adjustment of stability judgment threshold, incremental adjustment of minimum measurement duration, and selection of environmental compensation model (linear model or radiation-convection combined model). Reward function: ; In the formula, R is the reward function value, α is the efficiency term weighting coefficient, and T std For standard test duration, D meas S represents the actual measurement duration, S represents the confidence score, and N represents the actual measurement duration. abort The number of interruptions due to abnormalities is represented by α, β, and γ, which are weighting coefficients. Trigger model training, and distribute the updated policy network parameters after offline verification.

[0018] In another aspect, embodiments of the present invention provide a static evaporation rate testing system for cryogenic insulated gas cylinders, used to implement the above-described method for testing the static evaporation rate of cryogenic insulated gas cylinders. The system includes: The data acquisition module includes a pressure sensor, a temperature sensor array, a mass flow meter, and an ambient temperature and humidity sensor, used to acquire raw measurement data; The steady-state identification module is used to calculate the pressure growth rate, extract features, divide the evolution stages, perform exponential decay fitting, and output the steady-state parameter set. The test process control module includes sub-modules for measurement initialization, data validity verification, environmental compensation calculation, adaptive termination determination, and evaporation rate calculation. The digital twin simulation module is used to build and solve the thermodynamic model of the gas cylinder and predict future flow trends in real time. The reinforcement learning optimization module is used to train deep reinforcement learning models and optimize control policy parameters. The uncertainty and confidence assessment module is used to calculate Type A and Type B uncertainties based on the GUM method, output the expanded uncertainty, and calculate the confidence-weighted total score.

[0019] The method and system for testing the static evaporation rate of cryogenic insulated gas cylinders according to embodiments of the present invention have at least the following beneficial effects: First, it abandons the traditional 48-hour static setting followed by 24-hour fixed measurement model. Instead, it uses a four-stage evolutionary identification process to dynamically determine steady-state conditions. Combined with digital twin online simulation prediction and reinforcement learning parameter self-optimization, it achieves adaptive adjustment of measurement duration, shortening the testing cycle. Second, it automates the entire process from filling completion, steady-state identification, test initiation, measurement process control, adaptive termination, to uncertainty assessment, eliminating the need for manual intervention. This achieves intelligent closed-loop control throughout the entire process, eliminating the inconsistencies caused by traditional methods relying on manual experience to determine steady-state conditions and manually recording data, thus improving the repeatability and traceability of test results. Third, it integrates multi-source data such as pressure, temperature, and flow rate, performing physical self-consistency cross-validation based on the real gas law. This fundamentally avoids making incorrect judgments based on erroneous data. Multi-parameter fusion and thermodynamic consistency verification enhance the reliability and safety of test results. Fourth, it establishes a reinforcement learning closed loop, continuously optimizing control strategy parameters through historical test data. The average testing time for similar gas cylinders decreases with the increase in the number of tests, achieving continuous optimization and improving testing efficiency. Fifth, the combined standard uncertainty and expanded uncertainty of the evaporation rate are automatically assessed based on the GUM method. At the same time, a multi-dimensional confidence scoring system is constructed to provide inspectors with intuitive quantitative indicators of the reliability of the results and enhance the authority of the results.

[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of the static evaporation rate test method for a cryogenic insulated gas cylinder according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the pressure growth rate evolution process of the static evaporation rate test method for cryogenic insulated gas cylinders according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the temperature compensation and moving average effect of the static evaporation rate test method for low-temperature insulated gas cylinders according to an embodiment of the present invention. Figure 4This is a schematic diagram of the ambient temperature change curve for the static evaporation rate test method of a low-temperature insulated gas cylinder according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the evolution trajectory of the multidimensional feature vector in the feature space of the static evaporation rate test method for cryogenic insulated gas cylinders according to an embodiment of the present invention. Figure 6 This is a radar chart showing the confidence score of the static evaporation rate test method for low-temperature insulated gas cylinders according to an embodiment of the present invention. Figure 7 This is a functional block diagram of the static evaporation rate testing system for cryogenic insulated gas cylinders according to an embodiment of the present invention. Detailed Implementation

[0022] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0023] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0024] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," "exceeding," etc. are understood to exclude the stated number, and "above," "below," "within," etc. are understood to include the stated number. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the order of the indicated technical features.

[0025] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] This embodiment provides a method for testing the static evaporation rate of a low-temperature insulated gas cylinder. Please refer to [link to relevant documentation]. Figure 1 The method for testing the static evaporation rate of cryogenic insulated gas cylinders mainly includes steps S101 to S107: S101. After the cryogenic insulated gas cylinder is filled, continuously collect the internal pressure and temperature data of the gas cylinder, calculate the pressure growth rate sequence, determine whether the gas cylinder has reached thermal equilibrium steady state based on the pressure growth rate sequence, and output the test start command when the steady state conditions are met.

[0027] S102. Initialize the measurement process and load the preset adaptive control strategy parameters, which include minimum measurement duration, maximum measurement duration, stability judgment threshold, and drift-free slope threshold.

[0028] S103. During the measurement, mass flow rate, pressure, gas phase temperature, ambient temperature and ambient pressure are continuously collected at a preset high sampling frequency, and the collected measured data are verified in real time.

[0029] S104. Based on the measured data that has passed the real-time validity verification, the measured mass flow rate is compensated online according to the ambient temperature and ambient pressure, converted to the equivalent mass flow rate under standard environmental conditions, and the measured smooth flow rate sequence is obtained by using the moving average method.

[0030] S105. Perform digital twin simulation, using the current measured data as boundary conditions, to simulate the thermodynamic evolution trend of the gas cylinder in the future preset time period in real time, generate a predicted flow rate sequence, and compare the predicted flow rate sequence with the measured smooth flow rate sequence in real time. When the deviation continuously exceeds the preset threshold, trigger an abnormal alarm.

[0031] S106. After the measured duration reaches the minimum measurement duration, continuously evaluate the adaptive termination judgment conditions. The adaptive termination judgment conditions include: within the most recent sliding time window, the relative standard deviation of the smoothed flow is less than the stability judgment threshold; within the same window, the absolute value of the linear fitting slope of the smoothed flow is less than the drift-free slope threshold; and no abnormal alarm is triggered or the ambient temperature fluctuation is within the allowable range during the measurement.

[0032] S107. When all adaptive termination conditions are met, calculate the total evaporation mass and, in conjunction with the initial liquid mass of the gas cylinder and the actual measurement time, calculate the average static evaporation rate.

[0033] In step S101 above, determining whether the gas cylinder has reached thermal equilibrium steady state based on the pressure increase rate sequence includes: The evolution of the pressure growth rate is divided into four characteristic stages: the period of violent relaxation, the period of exponential decay, the period of quasi-steady-state oscillation, and the period of complete steady state. The exponential decay model is used to perform nonlinear least squares fitting on the data during the exponential decay period to obtain the decay rate coefficient, steady-state rate, and goodness of fit. The exponential decay model is: r(t) = r0e -λt + r ss， In the formula, r(t) is the pressure increase rate at time t, λ is the decay rate coefficient, and r0 is the initial rate. ss For steady-state rate, R 2 For goodness of fit; When the system detects that it has entered a fully steady state and continues for multiple consecutive time windows, the goodness of fit is greater than a preset threshold, and the relative deviation between the measured steady-state rate and the theoretical steady-state rate is less than a preset deviation threshold, the steady-state condition is determined to be satisfied.

[0034] In step S103 above, real-time validity verification is performed on the collected measured data, including: Single-point physical range check to determine whether the values ​​collected by each sensor are within the preset reasonable range. If MM consecutive sampling points exceed the range, it is determined to be an L3 level system fault. Check the rationality of the rate of change, calculate the rate of change of mass flow and the rate of change of pressure. If the rate of change exceeds the theoretical maximum rate of change threshold, mark the data point as a slight disturbance of level L1 and do not terminate the measurement. Multi-sensor consistency verification is based on the real gas equation of state. The equivalent outflow rate is calculated according to the rate of change of pressure and the rate of change of temperature, and compared with the measured flow rate. When the relative deviation exceeds the consistency threshold, it is judged as a severe disturbance of level L2.

[0035] In step S104 above, online compensation is performed on the measured mass flow rate based on ambient temperature and ambient pressure, including: Temperature compensation is performed; the temperature compensation formula is: ; In the formula, The instantaneous mass flow rate after temperature compensation. To measure instantaneous mass flow rate, The ambient temperature at the current moment. The boiling point of the low-temperature medium. Standard ambient temperature; If a volumetric flow meter is used, pressure compensation is added. The pressure compensation formula is: ; In the formula, The final compensated instantaneous mass flow rate, The instantaneous mass flow rate after temperature compensation. The current atmospheric pressure. Standard atmospheric pressure The absolute temperature under standard conditions. The current vapor phase temperature; The average length of the sliding window adaptively adjusts between 5 and 10 minutes, and the window length increases as the fluctuation of the environment increases.

[0036] In step S105 above, performing digital twin simulation includes: A thermodynamic digital twin model of the gas cylinder is established, which includes: a one-dimensional unsteady-state heat conduction equation for the insulation layer, a phase change kinetic equation for the medium, and a pressure-temperature-density coupling relationship based on the real gas equation of state. Boundary conditions of pressure, temperature, and ambient temperature are synchronously measured every 5 minutes, and the expected mass flow rate sequence for the next 1-2 hours is solved using the finite difference method. The relative deviation between the measured smoothed flow rate and the simulated predicted flow rate is calculated in real time. If the deviation exceeds 15% continuously and lasts for more than 10 minutes, it is determined that there is an unmodeled anomaly, triggering a Level 2 severe disturbance. When the gas cylinder thermodynamic digital twin model predicts that the flow rate fluctuation will be less than 0.5% within the next 30 minutes, the measurement will be terminated early as an auxiliary criterion, even if the statistical conditions within the current sliding window have not been fully met.

[0037] In step S107 above, the calculation of the average static evaporation rate includes: Calculate the cumulative evaporation mass: ; In the formula, The total mass of liquid evaporated during the measurement period. Test startup time, The test ends at this time. The final compensated instantaneous mass flow rate, It is a time differential element; The average static evaporation rate is calculated based on the cumulative evaporated liquid mass and the initial liquid mass during the measurement period. The calculation formula is as follows: ; In the formula, The average static evaporation rate, The total mass of liquid evaporated during the measurement period. The initial liquid mass in the gas cylinder. This refers to the actual measurement duration.

[0038] Following the step of calculating the average static evaporation rate in step S107 above, the following is included: The uncertainty of the evaporation rate measurement results is evaluated based on the GUM method, and the combined standard uncertainty and expanded uncertainty are calculated. To assess Type A uncertainty, the relative standard uncertainty is calculated based on the repeatability of the smoothed flow series during the measurement period. The formula is as follows: ; In the formula, For type A relative standard uncertainty, To smooth the experimental standard deviation of the flow series, This is the average value for smoothing the flow rate; Type B uncertainty assessment is performed, including uncertainty introduced by pressure sensor accuracy, uncertainty introduced by temperature sensor accuracy, uncertainty introduced by mass flow meter accuracy, uncertainty introduced by filling rate measurement error, and uncertainty introduced by medium physical property parameters (density, boiling point, latent heat) error. The sensitivity coefficients of each uncertainty component are calculated according to the measurement model transfer formula, and the combined relative standard uncertainty is calculated using the following formula: ; In the formula, To synthesize the relative standard uncertainty, For type A relative standard uncertainty, Let i be the i-th type B relative standard uncertainty component. The expanded uncertainty is calculated using the following formula: ; In the formula, To expand the uncertainty, k is the coverage factor. For the combined relative standard uncertainty; The output result is: The formula includes the contribution ratio of each component; where, The average static evaporation rate, This is the absolute value of the expanded uncertainty.

[0039] Following the step of calculating the average static evaporation rate in step S107 above, the method further includes: Confidence scores are calculated based on data integrity, steady-state maintenance, number of outliers, environmental stability, and model consistency. If the weighted total score is below 60 points, the test is deemed invalid. If the weighted total score is between 60 and 85, the expert collaboration process will be triggered. If the weighted total score is greater than 85 points, it will be automatically recognized.

[0040] Following the step of calculating the average static evaporation rate in step S107 above, the method further includes: The entire test process data is stored as experience samples in the experience pool, and reinforcement learning training is triggered asynchronously to update the adaptive control strategy parameters. The entire test process data includes gas cylinder characteristics, steady-state parameters, measurement process actions, environmental conditions, and test results. Reinforcement learning training employs Deep Q-Network (DQN) or Proximal Policy Optimization (PPO) algorithms, with their state space, action space, and reward function defined as follows: State space: includes cylinder volume, medium type, current test stage, current decay rate coefficient, current ambient temperature fluctuation range, and the ratio of measured duration to predicted stabilization duration; Action space: including incremental adjustment of stability judgment threshold, incremental adjustment of minimum measurement duration, and selection of environmental compensation model (linear model or radiation-convection combined model). Reward function: ; In the formula, R is the reward function value, α is the efficiency term weighting coefficient, and T std For standard test duration, D meas S represents the actual measurement duration, S represents the confidence score, and N represents the actual measurement duration. abort The number of interruptions due to abnormalities is represented by α, β, and γ, which are weighting coefficients. Trigger model training, and distribute the updated policy network parameters after offline verification.

[0041] Please see Figure 7 This embodiment also provides a static evaporation rate testing system for cryogenic insulated gas cylinders, used to implement the above-described method for testing the static evaporation rate of cryogenic insulated gas cylinders. The system includes: The data acquisition module 100 includes a pressure sensor, a temperature sensor array, a mass flow meter, and an ambient temperature and humidity sensor, used to acquire raw measurement data; The steady-state identification module 200 is used to calculate the pressure growth rate, extract features, divide the evolution stages, perform exponential decay fitting, and output the steady-state parameter set. The test process control module 300 includes sub-modules for measurement initialization, data validity verification, environmental compensation calculation, adaptive termination determination, and evaporation rate calculation. The Digital Twin Simulation Module 400 is used to establish and solve the thermodynamic model of the gas cylinder and predict future flow trends in real time. The reinforcement learning optimization module 500 is used to train deep reinforcement learning models and optimize control policy parameters. The uncertainty and confidence assessment module 600 is used to calculate Type A and Type B uncertainties based on the GUM method, output the expanded uncertainty, and calculate the confidence weighted total score.

[0042] The following is a detailed description of the static evaporation rate test method for cryogenic insulated gas cylinders provided in the embodiments of the present invention: 1. Test preparation and steady-state identification After the cryogenic insulated gas cylinder is filled, internal pressure and temperature data are continuously collected, and the sequence of pressure increase rate over time is calculated. Based on the evolution characteristics of the pressure increase rate sequence, it is automatically determined whether the gas cylinder has reached thermal equilibrium steady state.

[0043] Please see Figure 2 The evolution of the pressure growth rate is divided into four characteristic stages: the period of violent relaxation (stage I), the period of exponential decay (stage II), the period of quasi-steady-state oscillation (stage III), and the period of complete steady state (stage IV).

[0044] The exponential decay model (r(t)=r0e) is adopted. -λt + r ss Nonlinear least squares fitting was performed on the data from stage II to obtain the decay rate coefficient λ, the initial rate r0, and the steady-state rate r. ss and goodness of fit R 2 When the system is detected to have entered a fully steady state (stage IV) and this state persists for at least N consecutive time windows (N≥3), and (R... 2 >0.95), measured steady-state rate vs. theoretical steady-state rate r ss,th When the relative deviation is less than 5%, the steady-state condition is deemed met, a test start command is output, and the steady-state parameter set is transmitted to the test process control module 300. The steady-state parameter set includes the decay rate coefficient λ and the goodness-of-fit R. 2 and theoretical steady-state rate r ss,th .

[0045] 2. Test Initialization In response to the test start command, the system performs initialization operations: record the start time t. start Initial pressure and temperature; loading adaptive control strategy parameters pre-optimized by the reinforcement learning agent, including minimum measurement duration D. min (e.g., 4-8 hours), maximum measurement duration D max (e.g., 24 hours), stability threshold σ th (e.g., 1%~2%), drift-free slope threshold K th (e.g., 0.00042 g / s / h); Initiate high-frequency data acquisition (1~5 Hz).

[0046] 3. Data acquisition and real-time validity verification during measurement. Continuous collection of mass flow rate Pressure P(t), vapor phase temperature Ambient temperature and environmental pressure Perform three-level validation on each set of data: Single-point physical range check: Check if each parameter is within the preset reasonable range. If 10 consecutive sampling points exceed the range, it is judged as an L3 level system fault, and the test is stopped.

[0047] Reasonableness check of the rate of change: Calculate the rate of change of flow. and pressure change rate If it exceeds the theoretical maximum rate of change threshold (e.g.) If the data point is marked as a minor disturbance of level L1, the measurement is recorded but not stopped.

[0048] Multi-sensor consistency verification: Based on the real gas law (such as the SRK equation), the equivalent outflow rate is calculated from the rate of change of pressure and the rate of change of temperature. , and the measured flow Comparison. A relative deviation exceeding 15% is classified as a Level 2 severe disturbance.

[0049] 4. Online Environmental Compensation Please see Figure 3 and Figure 4 Environmental compensation was performed on the measured data that passed the validity verification, converting them to standard environmental conditions (standard ambient temperature 20℃, standard atmospheric pressure 101.325kPa). The compensation formula is as follows: The temperature compensation formula is: ; In the formula, The instantaneous mass flow rate after temperature compensation. To measure instantaneous mass flow rate, The ambient temperature at the current moment. The boiling point of the low-temperature medium. Standard ambient temperature; If a volumetric flow meter is used, pressure compensation is added. The pressure compensation formula is: ; In the formula, The final compensated instantaneous mass flow rate, The instantaneous mass flow rate after temperature compensation. The current atmospheric pressure. Standard atmospheric pressure The absolute temperature under standard conditions. The current vapor phase temperature; Among them, standard atmosphere =101.325 kPa, absolute temperature under standard conditions The temperature is 273.15K, at the standard ambient temperature. The temperature is 20°C.

[0050] A smoothed flow series is obtained by performing a moving average on the compensated flow series (with a window length of 5-10 minutes, adaptively adjusted according to environmental fluctuations). .

[0051] 5. Online simulation and prediction of digital twins The digital twin model includes: a one-dimensional unsteady-state heat conduction equation for the insulation layer (solved by the finite difference method), a phase change dynamics model of the medium, and a pressure-temperature-density coupling relationship based on the real gas equation of state.

[0052] Every 5 minutes, the digital twin model uses the current measured boundary conditions (such as pressure, temperature, and ambient temperature) as input to simulate the expected mass flow rate sequence for the next 1-2 hours. The system calculates the relative deviation between the measured smoothed flow rate and the simulated predicted flow rate in real time. If the deviation exceeds 15% continuously for more than 10 minutes, an unmodeled anomaly is identified (such as insulation layer damage, impure medium, or severe sensor drift), triggering a Level 2 severe disturbance.

[0053] When the digital twin model predicts that the flow fluctuation will be less than 0.5% within the next 30 minutes, this will be used as an auxiliary criterion to terminate the measurement in advance, even if the statistical conditions of the current sliding window have not been fully met.

[0054] 6. Adaptive Termination Decision The measured duration has reached the minimum measured duration D. min Subsequently, the following termination conditions will be continuously evaluated: Stability condition: Within the most recent sliding time window W (120~240 minutes), the relative standard deviation of the smoothed flow is less than the stability judgment threshold.

[0055] No drift condition: The absolute value of the slope of the linear fit of the smooth flow within the same window is less than the no drift slope threshold.

[0056] Abnormal event conditions: No L2 or higher abnormal events occurred during the measurement period, and the ambient temperature fluctuation range was less than ±5℃.

[0057] Digital twin auxiliary condition (optional): The digital twin predicts that traffic fluctuations within the next 30 minutes will be less than 0.5%.

[0058] The measurement will automatically terminate when the stability condition, drift-free condition, and abnormal event condition are all met simultaneously. If the measurement duration reaches the maximum measurement duration... If the termination conditions are still not met, the process will be forcibly terminated and a warning message will be generated, such as "The measurement is not sufficiently stable".

[0059] 7. Evaporation rate calculation After the measurement is terminated, calculate the cumulative evaporation mass: ; In the formula, The total mass of liquid evaporated during the measurement period. Test startup time, The test ends at this time. The final compensated instantaneous mass flow rate, It is a time differential element.

[0060] To facilitate calculation, a simplified formula can be used to calculate the cumulative evaporation mass: m evap = ×D meas ; In the formula, The total mass of liquid evaporated during the measurement period. To smooth the average flow rate, This refers to the actual measurement duration.

[0061] The initial liquid mass is calculated based on the filling rate, cylinder geometric volume, and medium density. The average static evaporation rate is then calculated. The calculation formula is: ; In the formula, The average static evaporation rate, The total mass of liquid evaporated during the measurement period. The initial liquid mass in the gas cylinder. This refers to the actual measurement duration; .

[0062] 8. Uncertainty assessment Based on the GUM method, the combined standard uncertainty of evaporation rate measurement results is automatically evaluated.

[0063] (1) Type A uncertainty assessment: Calculate the standard deviation based on the repeatability of the smoothed flow series during the measurement period. Relative standard uncertainty .

[0064] (2) Type B uncertainty assessment: Considering the accuracy of the pressure sensor (tolerance ±0.25%FS), the accuracy of the temperature sensor (±0.2℃), the accuracy of the mass flow meter (±0.5% of the reading), the filling rate measurement error (±2%), and the error of the medium's physical properties (such as density, boiling point, latent heat) (±1%). Based on the evaporation rate calculation formula, the cumulative evaporation mass m is respectively... evap Initial liquid mass m liquid and measurement duration D meas By taking the partial derivatives, we obtain the relative sensitivity coefficients of each input quantity. The sensitivity coefficients reflect the degree to which the error of the input quantity is amplified or reduced, and thus affects the final result.

[0065] Calculations show that the absolute value of the relative sensitivity coefficients of these three input quantities to the evaporation rate is ±1, meaning that the relative error of each input quantity will be proportionally transferred to the relative uncertainty component of the evaporation rate. Therefore, the relative standard uncertainty of each component is as follows: 0.5% introduced by the mass flow meter accuracy, 2.0% introduced by the filling rate error, and 1.0% introduced by the medium density error, etc.

[0066] (3) Calculate the combined relative standard uncertainty: ; In the formula, To synthesize the relative standard uncertainty, For type A relative standard uncertainty, Let i be the i-th type B relative standard uncertainty component; (4) Calculate the expanded uncertainty using the following formula: ; In the formula, To expand the uncertainty, k is the coverage factor. For the combined relative standard uncertainty; (5) The output result is: In the formula, The average static evaporation rate, This is the absolute value of the expanded uncertainty.

[0067] 9. Confidence Score The confidence score is calculated based on the following dimensions: Data integrity (weight 15%): 100 points are awarded for valid data points ≥ 95%, and 2 points are deducted for every 5% decrease.

[0068] Steady-state maintenance (weight 30%): The relative standard deviation of smoothed flow within the termination window is <0.5% for 100 points, <1% for 80 points, and <2% for 60 points.

[0069] Number of abnormal events (weight 25%): 100 points for no L2 or above events, 20 points deducted for each L2 event, and L3 / L4 events are directly deemed invalid.

[0070] Environmental stability (weight 15%): During the measurement period, the standard deviation of ambient temperature <1℃ earns 100 points, <2℃ earns 80 points, and <3℃ earns 60 points.

[0071] Model consistency (weight 15%): 100 points for a deviation of <5% between the calculated flow rate and pressure drop rate, 80 points for <10%, and 60 points for <15%. Model consistency is defined as the relative deviation between the measured flow rate and the equivalent flow rate calculated from the pressure change rate.

[0072] A weighted total score below 60 points indicates the test is invalid; scores between 60 and 85 points trigger an expert collaboration process; scores above 85 points are automatically approved. Upon triggering remote collaboration, a data snapshot is automatically generated, including pressure-time curves, flow-time curves, temperature-time curves, abnormal event logs, and digital twin simulation mismatch records, and pushed to experts. Experts review the data online and make decisions: approving the results, suggesting retesting, adjusting control parameters, or dispatching maintenance personnel. The system records expert decisions and subsequent results, transforming expert experience into labeled samples and incorporating them into the reinforcement learning training dataset.

[0073] 10. Enhance self-optimization and knowledge updating of learning. After each test is completed, the entire process data of this test is stored in the experience pool as an experience sample. The entire process data includes gas cylinder characteristics, steady-state parameters, measurement process actions, environmental conditions and test results.

[0074] Optimization algorithms using deep Q-networks or near-end strategies: State space: cylinder volume, medium type, current test stage, current decay rate coefficient, current ambient temperature fluctuation range, and the ratio of measured duration to predicted stabilization duration.

[0075] Action space: Incremental adjustment of stability judgment threshold (±0.2%), incremental adjustment of minimum measurement duration (±0.5 hours), and selection of environmental compensation model (linear model or radiation-convection combined model).

[0076] The reward function is: ; In the formula, R is the reward function value. T represents the efficiency term weighting coefficient. std For standard test duration, D meas S represents the actual measurement duration, S represents the confidence score, and N represents the actual measurement duration. abort The number of interruptions due to abnormalities is represented by α, β, and γ, which are weighting coefficients; for example, α=0.4, β=0.4, and γ=0.2.

[0077] After accumulating multiple test samples, such as every 50 test samples, a background training is triggered to update the policy network. The new policy is then deployed after offline validation (i.e., simulation on historical datasets), enabling the continuous evolution of the control policy.

[0078] The following explanation uses the static evaporation rate test of a 175L insulated gas cylinder as an example.

[0079] 1. System Configuration Test object: High-vacuum multilayer insulated gas cylinder, designed for storing liquid nitrogen, nominal working pressure 1.6MPa, volume 175L; and equipped with the following sensors: Pressure sensor: range 0~2.5MPa, accuracy 0.25 grade, installed in the gas phase space at the top of the gas cylinder.

[0080] Vapor phase temperature sensor: PT100 platinum resistance thermometer, measuring range -200℃~+50℃, accuracy class A.

[0081] Mass flow meter: thermal mass flow meter, range 0~2 g / s, accuracy ±0.5%FS.

[0082] Ambient temperature sensor: accuracy ±0.2℃.

[0083] Environmental pressure sensor: accuracy ±0.1 kPa.

[0084] Edge node control utilizes an industrial computer and is equipped with control valves for mass flow meters, solenoid valves for gas path switching, and alarm devices. Edge nodes perform real-time tasks such as data acquisition, filtering, environmental compensation, digital twin simulation, and termination determination; the cloud server handles reinforcement learning training, multi-station collaborative scheduling, and remote collaborative services. The digital twin model is solved using the finite difference method with a step size of 0.1 seconds, simulating the next 120 minutes every 5 minutes.

[0085] 2. Steady-state identification and parameter transfer between preceding and following links After the gas cylinders were filled (85% filling rate), pressure data collection began. The pressure increase rate evolution curve is shown below. Figure 2 As shown. Approximately 0.5 hours after filling, the system detected the entry into the exponential decay period (Stage II). Using a nonlinear least squares fitting exponential decay model, λ = 0.008 min was obtained. -1 R 2 =0.974, r ss,meas =0.031 kPa / min.

[0086] Theoretical steady-state rate r ss,th Based on the nominal static evaporation rate of the gas cylinder and the current ambient temperature, the result is 0.031 kPa / min, which is less than 5% relative to the measured value.

[0087] After reaching a fully steady state (Phase IV) and maintaining this state for three windows (30 minutes), a test start command is output, and the parameter set is passed to the test process control module 300. The test process control module 300 sets the stability threshold σ based on the decay rate coefficient λ = 0.008 (which is moderately slow). th =1.5% (tightened from the default 1.8%), minimum measurement duration D min =5 hours (longer than the default 4 hours) to accommodate a slower balancing process.

[0088] 3. Data verification and environmental compensation during measurement After the test starts, data is collected at a frequency of 2Hz.

[0089] In the second hour, the ambient temperature slowly rose from 24.8℃ to 25.3℃, and at the same time, the pressure sensor showed a single-point jump (0.2MPa→0.21MPa→0.2MPa).

[0090] The data validity verification module identifies points where the rate of change exceeds the threshold, marking them as slight L1 perturbations. Median filtering is then used to remove these points without affecting the overall data quality.

[0091] Real-time calculation of corrected flow rate: The original instantaneous flow rate is 0.021 g / s, the ambient temperature is 25.1℃, the boiling point of liquid nitrogen is -196℃, the standard temperature is 20℃, the temperature compensation coefficient is (25.1+196) / (20+196)=1.024, and the corrected instantaneous flow rate is 0.0215 g / s.

[0092] After a 10-minute moving average, the smoothed flow rate average was approximately 0.021 g / s, with a significant reduction in flow rate fluctuations and a decrease in the standard deviation from 0.0015 g / s to 0.0006 g / s. This smoothed flow rate average was used as the benchmark for average flow rate during subsequent measurements and for subsequent calculations.

[0093] 4. Digital Twin Online Simulation and Early Warning Please see Figure 5 Two hours after the test started, the digital twin model found that the measured flow rate of 0.021 g / s and the simulated predicted flow rate of 0.019 g / s deviated by more than 15% for 10 consecutive minutes.

[0094] The disturbance was identified as a severe L2 level disturbance, and the cause was analyzed: the simulation model was based on nominal insulation parameters, but the measured flow rate was higher than expected, indicating possible localized aging of the insulation layer. The system issued a warning and automatically extended the minimum measurement duration D. min Up to 5 hours later, in subsequent measurements, the actual flow rate stabilized at 0.021 g / s, which is the average value of the smoothed flow rate. The value is 0.021 g / s. The digital twin early warning system prevents premature system termination under erroneous assumptions, ensuring test reliability.

[0095] 5. Adaptive Termination of Measurement After 5 hours of measurement, the minimum measurement duration has been reached, and the assessment of termination conditions begins.

[0096] Take the smoothed flow rate data for the most recent W=120 minutes and calculate the relative standard deviation σ of the smoothed flow rate. rel =0.8%, σ rel <σ th The slope of the linear fit is k= 0.00021 g / s / h, k <k thNo L2 events occurred during the measurement period, and the ambient temperature fluctuated by ±0.5℃. The ambient temperature slowly increased from 24.8℃ to 25.3℃ during the measurement period, and fluctuated around the average value of 24.9℃ over the 5-hour measurement period. The standard deviation of the temperature data was calculated to be approximately 0.3℃.

[0097] All conditions were met, and the system automatically terminated the measurement. The actual measurement time was 5 hours, saving 19 hours.

[0098] 6. Evaporation rate calculation (1) Given conditions Geometric volume of gas cylinder: V total =175L=0.175m 3 ; Fill rate: η = 85% = 0.85; Liquid nitrogen density (at normal pressure and boiling point): ρ LN2 =808 kg / m3 = 0.808 kg / L; Actual measurement duration: D meas =5 × 3600 = 18000 seconds; Standard day length: T day =86400 seconds / day.

[0099] Calculate the cumulative evaporation mass during the measurement period: m evap = ×D meas =0.021g / s×18000s=378g=0.378kg.

[0100] (2) Calculate the initial liquid mass Calculate the initial liquid mass in the gas cylinder: M liquid = V total ×η×ρ LN2 = 175×0.85×0.808 = 175×0.6868=120.19kg.

[0101] (3) Calculate the daily evaporation mass ratio Calculate the proportion of the evaporated mass to the initial liquid mass per unit time: .

[0102] In comparison, the 5-hour adaptive measurement results of the present invention are basically consistent with the results of the traditional 24-hour fixed measurement method, with smaller errors and a time saving of 79.2%.

[0103] 7. Uncertainty assessment Type A uncertainty was calculated: the standard deviation of the smoothed flow rate series was 0.0006 g / s, and the average flow rate was 0.021 g / s; .

[0104] The following Type B uncertainty components were determined: pressure sensor ±0.25%FS (corresponding to a relative error of approximately 0.3% at a pressure of 0.2MPa), temperature sensor ±0.2℃ (affecting gas density calculation by approximately 0.2%), mass flow meter ±0.5%, filling rate error ±2%, and medium density error ±1%.

[0105] Combined relative standard uncertainty: .

[0106] Calculate the expanded uncertainty: ; Calculate the absolute value of the expanded uncertainty: .

[0107] Output: 1.51%±0.11% / d, k=2; the component contributions are as follows: Type A uncertainty accounts for 60%; the filling rate error accounts for 30% of the Type B uncertainty components, and other components account for 10%.

[0108] This uncertainty level is consistent with the engineering practice of static evaporation rate testing of cryogenic insulated gas cylinders, indicating that the measurement results have engineering reliability. It should be noted that the Type A uncertainty is obtained through statistical analysis of the measurement series, and its source is the fluctuation introduced by the repeatability of the mass flow meter measurement; in the combined standard uncertainty, u A The largest component, contributing approximately 60%, is Type B uncertainty, assessed using non-statistical methods and primarily based on prior information such as sensor accuracy and operational errors. Fill rate, determined by weighing or level gauges, introduces errors during cryogenic liquid filling operations due to factors such as weighing equipment accuracy, level reading errors, and deviations between the actual and nominal cylinder volumes. Fill rate measurement errors are typically controlled within ±2%, contributing approximately 30% of the uncertainty. The sum of the squares of the other components contributes approximately 10%.

[0109] 8. Confidence score Please see Figure 6 During the test, one L1 event and one L2 event occurred, affecting approximately 2% of the data. The confidence score was calculated as follows: Data integrity 98% → 100 points, Steady-state maintenance 0.8% → 80 points, Outlier events (one L1, one L2) → L2 deduction 20 points → 80 points, Environmental stability standard deviation 0.3℃ → 100 points, Model consistency bias 12% → 60 points. The weighted total score was calculated as follows: 0.15×100+0.3×80+0.25×80+0.15×100+0.15×60 = 83 points.

[0110] If the weighted total score is between 60 and 85, expert collaboration will be automatically triggered.

[0111] Data snapshots (including curves, digital twin mismatch records, and anomaly logs) are pushed to experts. After reviewing the data online, the experts conclude that the L2 disturbance is caused by a slow rise in ambient temperature, the correction is sufficient, and the result is reliable. They approve the result and add comments such as, "Note the trend of insulation degradation; it is recommended to shorten the next inspection cycle." The expert's decision is recorded and stored in the knowledge base.

[0112] 9. Reinforcement Learning Self-Optimization Comparison Initially, the default strategy is used, with a stability threshold σ. th =1.5%, minimum measurement duration D min =4h.

[0113] After 200 tests on different types of gas cylinders, reinforcement learning revealed that: For high-performance gas cylinders with a decay rate coefficient λ > 0.04, the stability judgment threshold σ is set as follows: th When the confidence level is relaxed to 2.0%, the confidence level can still be maintained at >90 points, and the average measurement time is reduced from 4.5h to 3.5h.

[0114] For aging gas cylinders with a decay rate coefficient λ < 0.005, tighten σ th Setting the value to 1.0% can prevent accidental termination.

[0115] After the new strategy was implemented, the average testing time for similar gas cylinders was reduced by 22%, and the false termination rate decreased by 50%.

[0116] The embodiments of the present invention have the following beneficial effects: 1. Full-process intelligent closed-loop control: The entire process, from filling completion, steady-state identification, test initiation, measurement process control, adaptive termination to uncertainty assessment, is fully automated without manual intervention, improving testing efficiency and result consistency. Compared to traditional 48-hour static placement and 24-hour fixed measurement, the total testing cycle can be shortened to 6-10 hours, improving efficiency by more than 75%.

[0117] 2. Enhanced adaptability through front-to-back linkage: The exponential decay parameters and goodness of fit obtained in the steady-state stage are automatically used to set the termination threshold and environmental compensation coefficient in the measurement stage, so that the control strategy matches the actual thermodynamic characteristics of the gas cylinder and avoids rigid parameter settings.

[0118] 3. Digital twins enhance predictive capabilities: By enabling online simulations to detect anomalies (such as insulation layer damage) in advance, and assisting in the early termination of measurements, testing time is further shortened. In gas cylinder tests simulating insulation layer aging, the digital twin issues a deviation warning 30 minutes in advance, avoiding invalid measurements.

[0119] 4. Reinforcement learning enables self-evolution: By continuously optimizing control strategies through historical data, the average testing time for similar gas cylinders decreases as the number of tests increases.

[0120] 5. Uncertainty Assessment and Confidence Score: The output includes expanded uncertainty, enhancing the reliability of the results, facilitating comparison and mutual recognition between different laboratories, and improving the authority of the test. By quantifying the confidence score, testing personnel can intuitively judge the test quality, aiding in decisions regarding retesting or release, and reducing the risk of misjudgment. Data is automatically pushed when confidence is low, allowing experts to make remote decisions, reducing anomaly handling time from an average of 2 hours to 20 minutes.

[0121] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for testing the static evaporation rate of a low-temperature insulated gas cylinder, characterized in that, include: After the cryogenic insulated gas cylinder is filled, the internal pressure and temperature data of the gas cylinder are continuously collected, the pressure growth rate sequence is calculated, and the gas cylinder is judged to have reached thermal equilibrium steady state based on the pressure growth rate sequence. When the steady state condition is met, a test start command is output. The measurement process is initialized by loading preset adaptive control strategy parameters, which include minimum measurement duration, maximum measurement duration, stability judgment threshold, and drift-free slope threshold. During the measurement, mass flow rate, pressure, gas phase temperature, ambient temperature and ambient pressure are continuously collected at a preset high sampling frequency, and the collected measured data are validated in real time. Based on the measured data that has passed real-time validity verification, the measured mass flow rate is compensated online according to the ambient temperature and ambient pressure, converted to the equivalent mass flow rate under standard environmental conditions, and the measured smooth flow rate sequence is obtained by using the moving average method. Digital twin simulation is performed, using the current measured data as boundary conditions, to simulate the thermodynamic evolution trend of the gas cylinder in a future preset time period in real time, generate a predicted flow rate sequence, and compare the predicted flow rate sequence with the measured smooth flow rate sequence in real time. When the deviation continuously exceeds a preset threshold, an abnormal alarm is triggered. After the measured duration reaches the minimum measurement duration, the adaptive termination criteria are continuously evaluated. The adaptive termination criteria include: within the most recent sliding time window, the relative standard deviation of the smoothed flow rate is less than the stability judgment threshold; the absolute value of the linear fitting slope of the smoothed flow rate within the same window is less than the drift-free slope threshold; and no abnormal alarms are triggered or the ambient temperature fluctuation is within the allowable range during the measurement. When all adaptive termination criteria are met, the total evaporation mass is calculated, and the average static evaporation rate is calculated by combining the initial liquid mass of the gas cylinder and the actual measurement time.

2. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, The step of determining whether the gas cylinder has reached thermal equilibrium steady state based on the pressure increase rate sequence includes: The evolution of the pressure growth rate is divided into four characteristic stages: the period of violent relaxation, the period of exponential decay, the period of quasi-steady-state oscillation, and the period of complete steady state. An exponential decay model is used to perform nonlinear least squares fitting on the data during the exponential decay period to obtain the decay rate coefficient, steady-state rate, and goodness of fit. The exponential decay model is: r(t) = r0e -λt + r ss， In the formula, r(t) is the pressure growth rate at time t, λ is the decay rate coefficient, and r0 is the initial rate. ss For steady-state rate, R 2 For goodness of fit; When the system detects that it has entered a fully steady state and continues for multiple consecutive time windows, the goodness of fit is greater than a preset threshold, and the relative deviation between the measured steady-state rate and the theoretical steady-state rate is less than a preset deviation threshold, the steady-state condition is determined to be satisfied.

3. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, The real-time validity verification of the collected measured data includes: Single-point physical range check to determine whether the values ​​collected by each sensor are within the preset reasonable range. If MM consecutive sampling points exceed the range, it is determined to be an L3 level system fault. Check the rationality of the rate of change, calculate the rate of change of mass flow and the rate of change of pressure. If the rate of change exceeds the theoretical maximum rate of change threshold, mark the data point as a slight disturbance of level L1 and do not terminate the measurement. Multi-sensor consistency verification is based on the real gas equation of state. The equivalent outflow rate is calculated according to the rate of change of pressure and the rate of change of temperature, and compared with the measured flow rate. When the relative deviation exceeds the consistency threshold, it is judged as a severe disturbance of level L2.

4. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, The online compensation for the measured mass flow rate based on ambient temperature and ambient pressure includes: Temperature compensation is performed; the temperature compensation formula is: ； In the formula, The instantaneous mass flow rate after temperature compensation. To measure instantaneous mass flow rate, The ambient temperature at the current moment. The boiling point of the low-temperature medium. Standard ambient temperature; If a volumetric flow meter is used, pressure compensation is added. The pressure compensation formula is: ； In the formula, The final compensated instantaneous mass flow rate, The instantaneous mass flow rate after temperature compensation. The current atmospheric pressure. Standard atmospheric pressure The absolute temperature under standard conditions. The current vapor phase temperature; The average length of the sliding window adaptively adjusts between 5 and 10 minutes, and the window length increases as the fluctuation of the environment increases.

5. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, The digital twin simulation includes: A thermodynamic digital twin model of a gas cylinder is established, which includes: a one-dimensional unsteady-state heat conduction equation for the insulation layer, a phase change kinetic equation for the medium, and a pressure-temperature-density coupling relationship based on the real gas equation of state. Boundary conditions of pressure, temperature, and ambient temperature are synchronously measured every 5 minutes, and the expected mass flow rate sequence for the next 1-2 hours is solved using the finite difference method. The relative deviation between the measured smoothed flow rate and the simulated predicted flow rate is calculated in real time. If the deviation exceeds 15% continuously and lasts for more than 10 minutes, it is determined that there is an unmodeled anomaly, triggering a Level 2 severe disturbance. When the gas cylinder thermodynamic digital twin model predicts that the flow rate fluctuation will be less than 0.5% within the next 30 minutes, the measurement will be terminated early as an auxiliary criterion, even if the statistical conditions within the current sliding window have not been fully met.

6. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, The calculation of the average static evaporation rate includes: Calculate the cumulative evaporation mass: ； In the formula, The total mass of liquid evaporated during the measurement period. Test startup time, The test ends at this time. The final compensated instantaneous mass flow rate, It is a time differential element; The average static evaporation rate is calculated based on the cumulative evaporated liquid mass and the initial liquid mass during the measurement period. The calculation formula is as follows: ； In the formula, The average static evaporation rate, The total mass of liquid evaporated during the measurement period. The initial liquid mass in the gas cylinder. This refers to the actual measurement duration.

7. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 6, characterized in that, Following the step of calculating the average static evaporation rate, the following steps are included: The uncertainty of the evaporation rate measurement results is evaluated based on the GUM method, and the combined standard uncertainty and expanded uncertainty are calculated. To assess Type A uncertainty, the relative standard uncertainty is calculated based on the repeatability of the smoothed flow series during the measurement period. The formula is as follows: ； In the formula, For type A relative standard uncertainty, To smooth the experimental standard deviation of the flow series, This is the average value for smoothing the flow rate; Type B uncertainty assessment is performed, including uncertainty introduced by pressure sensor accuracy, uncertainty introduced by temperature sensor accuracy, uncertainty introduced by mass flow meter accuracy, uncertainty introduced by filling rate measurement error, and uncertainty introduced by medium physical property parameters (density, boiling point, latent heat) error. The sensitivity coefficients of each uncertainty component are calculated according to the measurement model transfer formula, and the combined relative standard uncertainty is calculated using the following formula: ； In the formula, To synthesize the relative standard uncertainty, For type A relative standard uncertainty, Let i be the i-th type B relative standard uncertainty component; The expanded uncertainty is calculated using the following formula: ； In the formula, To expand the uncertainty, k is the coverage factor. For the synthesis of relative standard uncertainty; The output result is: In the formula, Let U be the average static evaporation rate, and U be the absolute value of the expanded uncertainty.

8. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 6, characterized in that, Following the step of calculating the average static evaporation rate, the method further includes: Confidence scores are calculated based on data integrity, steady-state maintenance, number of outliers, environmental stability, and model consistency metrics. If the weighted total score is below 60 points, the test is deemed invalid. If the weighted total score is between 60 and 85, the expert collaboration process will be triggered. If the weighted total score is greater than 85 points, it will be automatically recognized.

9. The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to claim 1, characterized in that, Following the step of calculating the average static evaporation rate, the method further includes: The entire test process data is stored as experience samples in the experience pool, and reinforcement learning training is triggered asynchronously to update the adaptive control strategy parameters. The entire test process data includes gas cylinder characteristics, steady-state parameters, measurement process actions, environmental conditions, and test results. The reinforcement learning training employs a deep Q-network or a proximal policy optimization algorithm, whose state space, action space, and reward function are defined as follows: State space: includes cylinder volume, medium type, current test stage, current decay rate coefficient, current ambient temperature fluctuation range, and the ratio of measured duration to predicted stabilization duration; Action space: including incremental adjustment of stability judgment threshold, incremental adjustment of minimum measurement duration, and selection of environmental compensation model (linear model or radiation-convection combined model). Reward function: ； In the formula, R is the reward function value, α is the efficiency term weighting coefficient, and T std For standard test duration, D meas S represents the actual measurement duration, S represents the confidence score, and N represents the actual measurement duration. abort The number of interruptions due to abnormalities is represented by α, β, and γ, which are weighting coefficients. Trigger model training, and distribute the updated policy network parameters after offline verification.

10. A static evaporation rate testing system for a cryogenic insulated gas cylinder, characterized in that, The method for testing the static evaporation rate of a cryogenic insulated gas cylinder according to any one of claims 1 to 9 includes: The data acquisition module includes a pressure sensor, a temperature sensor array, a mass flow meter, and an ambient temperature and humidity sensor, used to acquire raw measurement data; The steady-state identification module is used to calculate the pressure growth rate, extract features, divide the evolution stages, perform exponential decay fitting, and output the steady-state parameter set. The test process control module includes sub-modules for measurement initialization, data validity verification, environmental compensation calculation, adaptive termination determination, and evaporation rate calculation. The digital twin simulation module is used to build and solve the thermodynamic model of the gas cylinder and predict future flow trends in real time. The reinforcement learning optimization module is used to train deep reinforcement learning models and optimize control policy parameters. The uncertainty and confidence assessment module is used to calculate Type A and Type B uncertainties based on the GUM method, output the expanded uncertainty, and calculate the confidence-weighted total score.

Images