Liquid hydrogen container heat leakage prediction and pressure increase characteristic detection method and system

Abstract

This invention discloses a method and system for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container. The method involves establishing a three-dimensional thermal flow field model of the liquid hydrogen container; calculating the heat flux density at the gas-liquid interface based on the three-dimensional thermal flow field model to adjust the liquid hydrogen evaporation rate; obtaining the total heat leakage through finite element simulation and experimental data co-calibration; calculating the static evaporation rate based on the total heat leakage; establishing a nonlinear mapping model between the static evaporation rate and the pressure rise rate using response surface regression; monitoring the wall temperature distribution of the liquid hydrogen container in real time; calibrating the thermophysical parameters using a dynamic calibration algorithm via an LSTM neural network; and correcting the output results of the three-dimensional thermal flow field model. By establishing a three-dimensional thermal flow field model of the liquid hydrogen container, the nonlinear response relationship between the evaporation rate and the pressure rise rate of the liquid hydrogen container is clarified, and a cloud-edge collaborative detection system is constructed to accurately predict the heat leakage performance of the liquid hydrogen container and achieve efficient pressure rise rate detection.

Description

Technical Field

[0001] This invention relates to the field of cryogenic storage and transportation equipment testing technology, and in particular to a method and system for predicting heat leakage and detecting pressure rise characteristics of liquid hydrogen containers. Background Technology

[0002] With the development of hydrogen energy technology, liquid hydrogen storage tanks are widely used as critical infrastructure in aerospace, energy, and transportation. Liquid hydrogen containers generally adopt a vacuum multi-layer insulation structure, and their insulation performance directly determines the evaporation loss and operational safety of the liquid hydrogen container. However, current methods for assessing heat leakage in liquid hydrogen containers are mostly empirical estimates, ignoring the nonlinear thermophysical properties of liquid hydrogen in ultra-low temperature environments of -253℃ and the temperature-density stratification phenomenon, resulting in insufficient prediction accuracy and large prediction errors. At the same time, the lack of a quantitative correlation model between BOR (Boil-off Rate, static evaporation rate) and PRR (Pressure Rise Rate, pressure rise rate) makes it difficult to quantitatively detect the pressure rise rate. Summary of the Invention

[0003] 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 predicting heat leakage and detecting pressure rise characteristics of liquid hydrogen containers. By establishing a three-dimensional thermal flow field model of the liquid hydrogen container, the nonlinear response relationship between the evaporation rate and the pressure rise rate of the liquid hydrogen container is clarified, thereby improving the prediction accuracy, accurately predicting the heat leakage performance of the liquid hydrogen container, and achieving efficient detection of the pressure rise rate.

[0004] On one hand, embodiments of the present invention provide a method for predicting heat leakage and detecting pressure boosting characteristics of a liquid hydrogen container, including:

[0005] A three-dimensional thermal flow field model of a liquid hydrogen container is established. The three-dimensional thermal flow field model includes the temperature-density layered structure of the liquid hydrogen storage area, the time-varying characteristics of the thermal conductivity of the multilayer insulation material with temperature, the influence factors of the vibration mode of the outer wall structure on the thermal boundary layer, and the heat transfer mechanism of the transition flow-free molecular flow coupling of gas molecules in the vacuum layer.

[0006] The heat flux density at the gas-liquid interface is calculated based on the three-dimensional heat flow field model to adjust the amount of liquid hydrogen evaporation.

[0007] The total heat loss was obtained by calibrating the data through finite element simulation and experimental data.

[0008] Calculate the static evaporation rate based on the total heat loss.

[0009] A nonlinear mapping model between static evaporation rate and pressure rise rate was established using response surface regression.

[0010] The temperature distribution of the liquid hydrogen container wall is monitored in real time, and the thermal property parameters are calibrated using a dynamic calibration algorithm through an LSTM neural network to correct the output results of the three-dimensional thermal flow field model.

[0011] According to some embodiments of the present invention, establishing a three-dimensional thermal flux model of the liquid hydrogen container includes:

[0012] A three-dimensional model including the liquid hydrogen storage area, gas-liquid interface, inner wall, multi-layer insulation and outer wall structure was built using the COMSOL finite element platform to intuitively characterize each physical parameter;

[0013] Establish a vacuum layer thermodynamic model and construct a vacuum environment to suppress heat conduction, heat convection and heat radiation heat transfer.

[0014] The thermophysical properties of liquid hydrogen were fitted as a function of temperature.

[0015] A gas-liquid stratification model was established to simulate the gas-liquid situation inside a liquid hydrogen container, and the Boussinesq model was used to simulate the natural convection heat transfer behavior.

[0016] According to some embodiments of the present invention, the thermal properties of liquid hydrogen are modeled using the specific heat formula, the thermal conductivity formula, and the density formula, including:

[0017] The thermal properties of liquid hydrogen are modeled using a specific heat formula, which is:

[0018] C p (T) = a T + b T · T + c T · T 2 + d T · T 3

[0019] In the formula, C p Specific heat is expressed in Kelvin (K); T is absolute temperature, a T b T c T d T These are the temperature polynomial fitting coefficients;

[0020] The thermal properties of liquid hydrogen are modeled using a thermal conductivity formula, which is:

[0021] k(T) = e T · exp(f T / T)

[0022] In the formula, k is the thermal conductivity, T is the absolute temperature, and e T、 f T These are material constants;

[0023] The thermal properties of liquid hydrogen are modeled using a density formula, which is:

[0024] ρ(T) = g T / (1 + h T In the formula T), ρ is the density, and g T h T is the fitting coefficient, and T is the absolute temperature.

[0025] According to some embodiments of the present invention, the static evaporation rate is calculated based on the total heat loss using the following formula:

[0026] BOR = Q(t) / L v

[0027] In the formula, BOR is the static evaporation rate, in % / h; Q(t) is the total heat loss, in W; L v The latent heat of vaporization is expressed in J / kg.

[0028] According to some embodiments of the present invention, the nonlinear mapping model is as follows:

[0029] PRR = a + b·BOR + c·Fill + d·t hold

[0030] In the formula, PRR is the pressurization rate, in MPa / h; BOR is the static evaporation rate, in % / h; Fill is the initial fill rate; t hold , where is the settling time in hours; a, b, c, and d are the fitting coefficients.

[0031] According to some embodiments of the present invention, the dynamic calibration algorithm includes:

[0032] Deploy lightweight LSTM models on edge computing nodes;

[0033] Real-time correction of thermophysical property fitting coefficients, calibration frequency ≥10Hz, temperature fluctuation suppression ratio ≥95%.

[0034] According to some embodiments of the present invention, the detection of a high-pressure liquid hydrogen container includes:

[0035] Establish a supercritical liquid hydrogen density model:

[0036]

[0037] In the formula, ρ sc The density of liquid hydrogen, ρ 0 is the base density, p is the internal pressure of the container, a is the pressure coefficient, and b is the exponential factor.

[0038] Added circumferential stress analysis module:

[0039]

[0040] In the formula, σ θ Let p be the circumferential stress, p be the internal pressure of the container, and r be the inner radius of the container. i Let t be the radius of the i-th layer of the container's inner structure, and t be the wall thickness.

[0041] On the other hand, embodiments of the present invention provide a liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection system, comprising:

[0042] A multimodal sensing unit, comprising a fiber optic temperature sensor, a vacuum sensor, and a piezoelectric thin film sensor, wherein the fiber optic temperature sensor is used to monitor the temperature distribution of the liquid hydrogen container wall in real time.

[0043] An edge computing unit, equipped with an ARM processor, is used to support the nonlinear mapping between real-time static evaporation rate and pressure boost rate;

[0044] The backend server is configured with a GPU cluster for running COMSOL finite element simulation and response surface regression models.

[0045] According to some embodiments of the present invention, the fiber optic grating temperature sensor detects the temperature gradient by Bragg wavelength shift, the vacuum sensor detects the change in vacuum layer thickness by diffraction time difference method, and the piezoelectric thin film sensor is embedded in multiple insulation layers to monitor material creep.

[0046] The embodiments of the present invention have at least the following beneficial effects:

[0047] By constructing a three-dimensional thermal flow field model of a liquid hydrogen container and coupling heat conduction, vaporization, and natural convection, a two-way prediction and verification of heat leakage and pressurization rate of the liquid hydrogen container is achieved, significantly improving the accuracy and efficiency of heat leakage prediction and pressurization rate detection. First, a heat conduction equation coupled with a vacuum layer gas molecular dynamics model is proposed to achieve regional heat transfer simulation, reducing heat leakage prediction errors and overcoming the technical bottleneck of dynamic prediction of cryogenic insulation failure. Second, a dynamic calibration algorithm based on an LSTM neural network is developed, which corrects thermophysical parameters in real time through edge computing nodes, suppressing temperature fluctuation interference, shortening the model response time to within 5 seconds, and improving model robustness. Third, it represents a leap from simple heat leakage detection to lifetime prediction, providing early warning of local defects such as insulation layer damage. The detection system includes a multimodal sensing unit, an edge computing unit, and a backend server, and integrates fiber optic temperature sensors, vacuum sensors, and piezoelectric thin film sensors to construct a cloud-edge collaborative detection system. Through the deep integration of mechanism modeling and data-driven approaches, the difficulty of quantitative characterization of vacuum insulation failure is solved, prediction accuracy is improved, the detection cycle is shortened, and the detection cost is reduced.

[0048] 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

[0049] 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:

[0050] Figure 1 This is a flowchart of the liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection method according to an embodiment of the present invention;

[0051] Figure 2 A flowchart illustrating the method for predicting heat leakage and detecting pressure boosting characteristics of a liquid hydrogen container in this invention;

[0052] Figure 3 The flowchart of the liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection method of this invention is a process for modeling the liquid hydrogen thermal property parameters using specific heat formula, thermal conductivity formula and density formula.

[0053] Figure 4 This is a block diagram of the liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection system according to an embodiment of the present invention.

[0054] Figure label:

[0055] Multimodal sensing unit 100, edge computing unit 200, backend server 300. Detailed Implementation

[0056] 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.

[0057] 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.

[0058] 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.

[0059] In the description of this invention, unless otherwise explicitly defined, the terms "setting", "installing", "connecting" and "linking" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0060] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Please see Figure 1 This embodiment provides a method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container, mainly including steps S101~S106:

[0062] S101. Establish a three-dimensional thermal flow field model for the liquid hydrogen container. The three-dimensional thermal flow field model includes the temperature-density layered structure of the liquid hydrogen storage area, the time-varying characteristics of the thermal conductivity of the multilayer insulation material with temperature, the influence factors of the vibration mode of the outer wall structure on the thermal boundary layer, and the heat transfer mechanism of the transition flow-free molecular flow coupling of gas molecules in the vacuum layer.

[0063] S102. Calculate the heat flux density at the gas-liquid interface based on the three-dimensional heat flow field model to adjust the amount of liquid hydrogen evaporation.

[0064] S103. The total heat loss is obtained through finite element simulation and experimental data calibration.

[0065] S104. Calculate the static evaporation rate based on the total heat loss.

[0066] S105. A nonlinear mapping model between static evaporation rate and pressurization rate is established using response surface regression.

[0067] S106. Real-time monitoring of the wall temperature distribution of liquid hydrogen container, and calibration of thermal property parameters using a dynamic calibration algorithm via LSTM neural network to correct the output results of the three-dimensional thermal flow field model.

[0068] Specifically, in the three-dimensional thermal flow field model, the temperature-density stratification structure of the liquid hydrogen reservoir refers to the gradient distribution of temperature along the height direction caused by heat conduction, convection, or external heat leakage within the liquid hydrogen container, forming a stable stratified structure with a high upper temperature and a low lower temperature. The density gradient of liquid hydrogen due to temperature differences results in a vertical distribution where cold liquid sinks and hot liquid rises, thus coupling with the temperature stratification to form a composite structure. The time-varying thermal conductivity of multilayer insulation materials refers to the dynamic change in thermal conductivity of the material during service due to environmental conditions (such as temperature, humidity, and pressure) or the passage of time. This characteristic directly affects the thermal resistance stability and long-term insulation performance of the multilayer insulation system. The influence of external wall vibration modes on the thermal boundary layer refers to how external structural vibration alters the fluid flow state, heat transfer path, and boundary layer stability, thereby affecting the thermal boundary layer thickness, temperature distribution, and heat transfer efficiency. Vibration mode parameters include frequency, amplitude, and vibration direction. The transitional flow-free molecular flow coupled heat transfer mechanism refers to the process in which gas molecules, under moderate pressure (between viscous flow and molecular flow), are simultaneously affected by intermolecular collisions (viscous flow characteristics) and molecular-to-wall collisions (free molecular flow characteristics). Influencing factors include molecular weight, specific heat ratio, viscosity, characteristic size, and surface roughness.

[0069] By monitoring data in real time, such as measuring wall temperature using fiber Bragg grating temperature sensors and vacuum decay rate using vacuum sensors, the thermal properties (e.g., thermal conductivity) and boundary conditions (e.g., heat leakage) in the model are dynamically corrected. For example, if the fiber Bragg grating temperature sensor detects an abnormal increase in wall temperature, the LSTM neural network adjusts the thermal conductivity parameter and recalculates the heat leakage and static evaporation rate (BOR). This reduces heat leakage prediction errors and lowers the correlation model error between the static evaporation rate (BOR) and the pressure rise rate (PRR). The calculation of heat leakage and static evaporation rate (BOR) is based on multiphysics coupled simulation and closed-loop calibration of experimental data, combined with real physical mechanisms such as temperature stratification and material nonlinearity, ensuring the engineering applicability and high accuracy of the results. A single measurement can simultaneously obtain heat leakage and static evaporation rate (BOR), eliminating the need for step-by-step measurements; by monitoring the static evaporation rate (BOR) in real time, early warnings of abnormal liquid hydrogen evaporation, such as insulation layer damage, can be provided; the error is controlled within ±3%, reducing the number of repeated tests and lowering maintenance costs.

[0070] The evaporation rate is related to the heat flux density. The heat flux density at the gas-liquid interface is calculated using a three-dimensional heat flow field model. Controlling the heat flux density can regulate the evaporation rate and prevent excessive pressure. Integrating Fourier's law heat conduction equation and the Navier-Stokes fluid dynamics equation, the control equations are discretized using the finite volume method to simultaneously solve for the temperature, velocity, and pressure field distributions at the gas-liquid interface. Correlation analysis between heat flux density and parameters such as temperature gradient and material thermal conductivity reveals the microscopic mechanism of heat conduction at the gas-liquid interface. The heat flux density at the gas-liquid interface within the storage tank is calculated to predict the pressure change trend over time, such as the nonlinear relationship between pressurization time and heat flux density. This provides a basis for pressure control strategies, optimizes heat transfer efficiency, and ensures the safety of liquid hydrogen storage.

[0071] Please see Figure 2 The establishment of the three-dimensional thermal flow field model of the liquid hydrogen container in step S101 above includes:

[0072] S201. Use the COMSOL finite element platform to establish a three-dimensional model including the liquid hydrogen storage area, gas-liquid interface, inner wall, multi-layer insulation and outer wall structure to intuitively characterize each physical parameter.

[0073] S202. Establish a thermodynamic model of the vacuum layer and construct a vacuum environment to suppress heat conduction, heat convection and heat radiation heat transfer.

[0074] S203. Model the thermal properties of liquid hydrogen using specific heat formula, thermal conductivity formula, and density formula;

[0075] S204. Establish a gas-liquid stratification model to simulate the gas-liquid situation inside the liquid hydrogen container, and use the Boussinesq model to simulate natural convection heat transfer behavior.

[0076] The total heat loss Q(t) was calibrated by combining COMSOL finite element simulation with measured data from fiber optic grating sensors.

[0077] In some embodiments of the present invention, the vacuum layer thermodynamic model includes the Knudsen number criterion and the vacuum degree decay model:

[0078] Knudsen number criterion: Kn = 2λ gas / D pore

[0079] In the formula, λ gas D is the mean free path of gas molecules, measured in meters. pore The aperture of the vacuum layer is in meters (m).

[0080] Vacuum degree decay model:

[0081]

[0082] In the formula, P vac V represents the vacuum pressure, in Pa; Q(t) represents the total heat loss, in W; V vac The volume of the vacuum layer is expressed in meters (m). 3 ;T vac T is the temperature of the vacuum layer. env For ambient temperature; S leak External leakage rate, in Pa·m 3 / s; α is the surface desorption coefficient, in Pa. -1 ·s -1 .

[0083] Please see Figure 3 The above-mentioned step S203 involves modeling the thermal properties of liquid hydrogen using the pyrometric formula, thermal conductivity formula, and density formula, including:

[0084] S301. Model the thermal properties of liquid hydrogen using the specific heat formula, which is:

[0085] C p (T) = a T + b T · T + c T · T 2 + d T · T 3

[0086] In the formula, C p Specific heat is expressed in Kelvin (K); T is absolute temperature, a T b T c T d T These are the temperature polynomial fitting coefficients;

[0087] S302. The thermal properties of liquid hydrogen are modeled using the thermal conductivity formula, which is:

[0088] k(T) = e T · exp(f T / T)

[0089] In the formula, k is the thermal conductivity, T is the absolute temperature, and e T、 f T These are material constants;

[0090] S303. The thermal properties of liquid hydrogen are modeled using the density formula, which is:

[0091] ρ(T) = g T / (1 + h T In the formula T), ρ is the density, and g T hT is the fitting coefficient, and T is the absolute temperature.

[0092] The static evaporation rate in step S104 above, calculated based on the total heat loss, is calculated using the following formula:

[0093] BOR = Q(t) / L v

[0094] In the formula, BOR is the static evaporation rate, in % / h; Q(t) is the total heat loss, in W; L v The latent heat of vaporization is expressed in J / kg.

[0095] The nonlinear mapping model in step S105 above is:

[0096] PRR = a + b·BOR + c·Fill + d·t hold

[0097] In the formula, PRR is the pressurization rate, in MPa / h; BOR is the static evaporation rate, in % / h; Fill is the initial fill rate; t hold , where is the settling time in hours; a, b, c, and d are the fitting coefficients.

[0098] In some embodiments of the present invention, the LSTM neural network model is as follows:

[0099]

[0100] In the formula, For the predicted output, i.e. the corrected thermal conductivity, x t The input feature vector includes temperature, pressure, and time; h t This is the hidden layer state vector.

[0101] In some embodiments of the present invention, the dynamic calibration algorithm includes:

[0102] Deploy lightweight LSTM models on edge computing nodes to correct thermal property fitting coefficients in real time, with a calibration frequency ≥10Hz and a temperature fluctuation suppression ratio ≥95%;

[0103] Federated learning enables privacy protection of multi-container data and joint model training, meeting GDPR compliance requirements.

[0104] In some embodiments of the present invention, the dynamic calibration algorithm further includes:

[0105] The multi-objective optimization module performs dual optimization by minimizing prediction error and suppressing model complexity;

[0106] The online learning mechanism uses the AdaBoost algorithm to dynamically adjust the weights of the base classifiers to adapt to the container aging process.

[0107] The objective function for minimizing the prediction error is:

[0108]

[0109] In the formula, To predict heat loss for the model, This represents the actual measured value of heat loss.

[0110] The objective function for reducing model complexity is:

[0111]

[0112] In the formula, λ is the regularization coefficient and w is the model weight.

[0113] It should be noted that heat leakage is affected by multiple factors such as temperature, pressure, and material aging, exhibiting strong nonlinear characteristics. By minimizing the prediction error, the model can adaptively learn the heat leakage characteristics under complex operating conditions, such as transient thermal disturbances caused by vibration, ensuring that the model has a high-precision response to dynamic changes in heat leakage. The regularization term is equivalent to applying a smoothing constraint to the changes in thermal property parameters (such as thermal conductivity and specific heat), avoiding drastic fluctuations in parameters due to sensor noise or local anomalies; L2 regularization applies a penalty to the model weights w, suppressing overfitting and preventing the model from deviating from the true physical laws due to noisy data or outliers.

[0114] In some embodiments of the present invention, the detection of high-pressure liquid hydrogen containers includes:

[0115] Establish a supercritical liquid hydrogen density model:

[0116]

[0117] In the formula, ρ sc The density of liquid hydrogen, ρ 0 is the base density, p is the internal pressure of the container, a is the pressure coefficient, and b is the exponential factor.

[0118] Added circumferential stress analysis module:

[0119]

[0120] In the formula, σ θ Let p be the circumferential stress, p be the internal pressure of the container, and r be the inner radius of the container. i Let t be the radius of the i-th layer of the container's inner structure, and t be the wall thickness.

[0121] Please see Figure 4This embodiment also discloses a liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection system, used to implement the above-mentioned liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection method, including:

[0122] The multimodal sensing unit 100 includes a fiber optic temperature sensor, a vacuum sensor, and a piezoelectric thin film sensor. The fiber optic temperature sensor is used to monitor the temperature distribution of the liquid hydrogen container wall in real time.

[0123] Edge computing unit 200, equipped with an ARM processor, is used to support the nonlinear mapping of real-time static evaporation rate and pressure boost rate;

[0124] Backend server 300 is configured with a GPU cluster to run COMSOL finite element simulations and response surface regression models. It should be noted that backend server 300 can be a cloud platform. It can also compare real-time monitoring data with a historical cloud database to generate a comprehensive evaluation report on the thermal insulation performance of the liquid hydrogen container. This report includes: prediction of remaining service life based on Weibull distribution; 3D localization of local defects; and early warning of vacuum life 6-12 months in advance.

[0125] In some embodiments of the present invention, the fiber optic grating temperature sensor detects the temperature gradient by Bragg wavelength shift, with a Bragg wavelength of 1550 nm and a spatial resolution of 5 cm; the vacuum sensor uses time-of-flight diffraction to detect changes in vacuum layer thickness at a frequency of 40 kHz, and is used to detect the vacuum attenuation rate with an accuracy of ±0.1 Pa / h; the piezoelectric thin film sensor is embedded in multiple layers of insulation to monitor material creep with a sensitivity of 0.1 με.

[0126] Example 1: Static Evaporation Rate Prediction

[0127] A certain type of vehicle-mounted liquid hydrogen storage tank (500L volume, 30mm wall thickness, insulation layer consisting of 10 layers of Al / glass fiber reflector + polyurethane foam) underwent heat leakage testing at an ambient temperature of 25℃ and an initial filling rate of 80%.

[0128] (1) Model building

[0129] A three-dimensional thermal flow field model was built using COMSOL Multiphysics, with a mesh size of 0.5 mm (locally refined to 0.1 mm at the gas-liquid interface).

[0130] Boundary conditions are set as follows: the inner wall is subjected to heat absorption by liquid hydrogen evaporation (Neumann condition), and the outer wall is subjected to heat transfer by natural convection (convection coefficient 5 W / (m²·K)).

[0131] (2) Calibration of thermophysical parameters

[0132] The thermal conductivity of liquid hydrogen was fitted using a temperature polynomial:

[0133] k(T) = 0.138 × exp(830 ÷ T)

[0134] Where, k(T) is the thermal conductivity of liquid hydrogen, in W / (m·K); T is the absolute temperature, in K, with a value ranging from 20 K to 300 K (corresponding to -253℃ to 27℃) for typical operating temperatures of liquid hydrogen; 0.138 is the empirical fitting coefficient, reflecting the baseline value of thermal conductivity of liquid hydrogen at low temperatures; and 830 is a constant related to the micro-vibration characteristics of the material.

[0135] The thermal conductivity of the multilayer insulation material was measured by the transient hot wire method, and the error between the experimental value and the model input was ≤2%.

[0136] (3) Data collection

[0137] Wall temperature distribution was monitored using a fiber optic temperature sensor (16 channels, wavelength range 1529-1561nm) at a sampling rate of 10Hz.

[0138] The vacuum decay rate is measured in real time using a vacuum sensor (40kHz), with an accuracy of ±0.1Pa / h.

[0139] (4) Comparison between simulation and experiment

[0140] The simulation calculation shows that the total heat loss in 24 hours is Q(t) = 43.6W, corresponding to a static evaporation rate BOR = 0.351% / h;

[0141] The experiment measured the vacuum level from 1 Pa to 50 Pa over 24 hours, and the static evaporation rate BOR was calculated to be 0.357% / h using the ideal gas equation.

[0142] Error analysis: The relative error Δ = (0.357 - 0.351) / 0.351 × 100% = 1.71%, which meets the ±3% accuracy requirement.

[0143] (5) Verification of technical effects

[0144] Model correction: By dynamically calibrating the thermal conductivity parameter using an LSTM neural network, the temperature fluctuation suppression ratio was improved from 75% to 95%.

[0145] Vacuum lifetime prediction: combining vacuum decay curve (dP) vac ( / dt = -0.03Pa / h), the predicted vacuum layer failure time is 2.1 years.

[0146] Example 2: Dynamic Calibration Application

[0147] A dynamic calibration experiment was conducted on a liquid hydrogen storage tank under the condition of liquid filling and discharging circulation (filling rate 50% → 90% → 50%, circulation cycle 6 hours).

[0148] (1) Sensor data acquisition

[0149] Wall temperature fluctuation amplitude: The temperature difference reaches 15℃ during the filling stage and returns to 5℃ during the discharging stage;

[0150] Vacuum fluctuation: During liquid filling, the pressure drops from 10 due to the cold trap effect. -3 Pa rises to 10 -2 Pa, slowly recovers after drainage.

[0151] (2) LSTM neural network calibration

[0152] Model structure: Input layer (temperature, pressure, time) → Hidden layer (64 neurons) → Output layer (corrected thermal conductivity);

[0153] Training data: 30 sets of historical data on fluid filling and discharging conditions were selected, and the loss function was mean square error (MSE).

[0154] Calibration frequency: 10Hz, real-time correction of thermal conductivity parameters.

[0155] (3) Performance verification

[0156] Temperature fluctuation suppression: The standard deviation of wall temperature decreased from 0.5℃ to 0.025℃ after calibration (suppression ratio 95%).

[0157] Model response time: reduced from 30 seconds in the traditional method to 5 seconds, adapting to transient conditions during tank start-up and shutdown.

[0158] (4) Verification of technical effectiveness:

[0159] Noise immunity: Even under the interference of vacuum sensor noise (signal-to-noise ratio 30dB), the BOR calculation error remains below 1.5%;

[0160] Federated learning results: By jointly training with data from 5 storage tanks, the model's generalization error was reduced by 40%.

[0161] Example 3: High-Pressure Liquid Hydrogen Container Detection

[0162] A high-pressure test was conducted on a 35MPa vehicle-mounted liquid hydrogen storage tank (1.2m in diameter, 40mm in wall thickness, with a composite structure of nano-aerogel and reflective screen for insulation).

[0163] (1) Validation of the supercritical model:

[0164] State equation parameter: a = 0.034 MPa -1 b=6.2;

[0165] Density calculation: Calculate the density ρ under conditions of 35 MPa and -253℃. sc =708 kg / m 3 The error between the experimental and experimental values ​​is ≤1.2%.

[0166] (2) Circumferential stress analysis:

[0167] Calculation conditions: internal pressure 35MPa, inner radius 0.6m, wall thickness 0.04m;

[0168] Formula application:

[0169]

[0170] Safety factor verification: The material yield strength is 180MPa, and the safety factor is 1.52, which meets the industry standard.

[0171] (3) Accelerated life test:

[0172] Test conditions: accelerated aging at 80℃ constant temperature, initial pressure of vacuum layer 10. -3 Pa;

[0173] Failure criteria: Vacuum level rises to 1 Pa or heat leakage rate increases by 50%;

[0174] Lifespan prediction: Based on the Weibull model (β=1.5, η=2000h), the lifespan under normal operating conditions is predicted to be ≥15 years.

[0175] (4) Verification of technical effects

[0176] The model was successfully applied to a 50MPa stationary liquid hydrogen storage tank, with a heat leakage prediction error of ±2.8%, demonstrating high-pressure adaptability.

[0177] The following details the specific steps for calculating the total heat loss and static evaporation rate:

[0178] 1. Calculate the total heat loss Q(t)

[0179] The results were obtained through joint calibration using finite element simulation and experimental data. The specific steps are as follows:

[0180] (1) Model building

[0181] A three-dimensional thermal flow field model was created using COMSOL Multiphysics, including:

[0182] Liquid hydrogen storage area (VOF method for tracking gas-liquid interface);

[0183] Multi-layer thermal insulation structure (10 layers of Al / glass fiber reflector + polyurethane foam).

[0184] Vibration modes of the outer wall (10-50Hz frequency);

[0185] Vacuum layer heat transfer mechanism (transition flow-free molecular flow coupling).

[0186] Its boundary conditions are:

[0187] Inner wall: liquid hydrogen evaporation is endothermic, Neumann conditions;

[0188] External wall: Ambient temperature convective heat transfer, Robin conditions.

[0189] (2) Simulation solution

[0190] Discrete control equations: ,

[0191]

[0192] Input parameters:

[0193] Temperature distribution: monitored in real time using a fiber Bragg grating temperature sensor;

[0194] Thermophysical parameters: based on temperature polynomial fitting.

[0195] (3) Experimental calibration

[0196] Under conditions of 80% filling rate and ambient temperature of 25℃, the container wall temperature distribution and heat leakage were measured by heat balance experiment.

[0197] The simulation results were compared with the experimental data, and the thermal conductivity parameters were dynamically calibrated using an LSTM neural network (calibration frequency ≥ 10 Hz). The total heat loss was finally obtained as Q(t) = 43.6 W, and the simulation error was ≤ 2.1%.

[0198] 2. Calculate the static evaporation rate BOR

[0199] BOR = Q(t) / L v = 43.6 / (447 × 10 3 ) × 3600 = 0.351% / h;

[0200] The total heat loss Q(t) is 43.6 W, and the latent heat of vaporization L v It is 447 kJ / kg.

[0201] 3. Data validation and error control

[0202] (1) Experimental verification

[0203] The accuracy of BOR is indirectly verified by monitoring pressure changes inside the container using pressure sensors.

[0204] The measured BOR value was 0.357% / h, which deviated from the simulation result by 1.71%.

[0205] (2) Sources of error

[0206] Model simplification: Ignore local turbulence effects, such as microscale flow at the gas-liquid interface;

[0207] Sensor accuracy: The fiber optic grating temperature sensor has a temperature measurement error of ±0.5℃;

[0208] Material aging: The thermal conductivity of insulation materials drifts over time, which can be corrected using the Arrhenius equation.

[0209] By constructing a multi-physics coupled model and an intelligent detection system, bidirectional prediction and verification of heat leakage and pressurization rate of liquid hydrogen containers are achieved, significantly improving the accuracy and efficiency of heat leakage prediction and pressurization rate detection. First, a heat conduction equation coupled with a vacuum layer gas molecular dynamics model is proposed to achieve regional heat transfer simulation, reducing the heat leakage prediction error from ±15% of traditional methods to ±3%, breaking through the technical bottleneck of dynamic prediction of cryogenic insulation failure. Second, a dynamic calibration algorithm based on LSTM neural networks is developed, which corrects thermophysical parameters (such as the cubic coefficient of temperature) in real time through edge computing nodes, suppressing temperature fluctuation interference, shortening the model response time to within 5 seconds, and improving model robustness. Third, it represents a leap from simple heat leakage detection to lifetime prediction, providing early warning of local defects such as insulation layer damage. The detection system includes a multi-modal sensing unit, an edge computing unit, and a back-end server, and integrates fiber optic temperature sensors, vacuum sensors, and piezoelectric thin-film sensors to construct a cloud-edge collaborative detection system; it also supports a unified detection standard for high-pressure liquid hydrogen containers above 35MPa. By deeply integrating mechanistic modeling and data-driven approaches, this method addresses industry pain points such as the difficulty in quantitatively characterizing vacuum insulation failures and the lack of testing standards for high-pressure scenarios. This method boasts high accuracy and strong adaptability, making it widely applicable for factory testing, periodic inspections, and fault early warning of liquid hydrogen containers. It improves prediction accuracy, shortens testing cycles, and reduces testing costs, providing a systematic solution for the safe operation and energy efficiency optimization of liquid hydrogen storage and transportation equipment.

[0210] 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 predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container, characterized in that, include: A three-dimensional thermal flow field model of a liquid hydrogen container is established. The three-dimensional thermal flow field model includes the temperature-density layered structure of the liquid hydrogen storage area, the time-varying characteristics of the thermal conductivity of the multilayer insulation material with temperature, the influence factors of the vibration mode of the outer wall structure on the thermal boundary layer, and the heat transfer mechanism of the transition flow-free molecular flow coupling of gas molecules in the vacuum layer. The heat flux density at the gas-liquid interface is calculated based on the three-dimensional heat flow field model to adjust the amount of liquid hydrogen evaporation. The total heat loss was obtained by calibrating the data through finite element simulation and experimental data. Calculate the static evaporation rate based on the total heat loss. A mapping model between static evaporation rate and pressure rise rate was established using response surface regression. The temperature distribution of the liquid hydrogen container wall is monitored in real time, and the thermal property parameters are calibrated using a dynamic calibration algorithm through an LSTM neural network to correct the output results of the three-dimensional thermal flow field model.

2. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 1, characterized in that, The establishment of the three-dimensional thermal flow field model of the liquid hydrogen container includes: A three-dimensional model including the liquid hydrogen storage area, gas-liquid interface, inner wall, multi-layer insulation and outer wall structure was built using the COMSOL finite element platform to intuitively characterize each physical parameter; Establish a vacuum layer thermodynamic model and construct a vacuum environment to suppress heat conduction, heat convection and heat radiation heat transfer. The thermal properties of liquid hydrogen are modeled using specific heat formula, thermal conductivity formula, and density formula. A gas-liquid stratification model was established to simulate the gas-liquid situation inside a liquid hydrogen container, and the Boussinesq model was used to simulate the natural convection heat transfer behavior.

3. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 2, characterized in that, The process of modeling the thermal properties of liquid hydrogen using specific heat formulas, thermal conductivity formulas, and density formulas includes: The thermal properties of liquid hydrogen are modeled using a specific heat formula, which is: C p (T) = a T + b T · T + c T · T 2 + d T · T 3 In the formula, C p Specific heat is expressed in Kelvin (K); T is absolute temperature, a T b T c T d T These are the temperature polynomial fitting coefficients; The thermal properties of liquid hydrogen are modeled using a thermal conductivity formula, which is: k(T) = e T · exp(f T / T) In the formula, k is the thermal conductivity, T is the absolute temperature, and e T、 f T These are material constants; The thermal properties of liquid hydrogen are modeled using a density formula, which is: ρ(T) = g T / (1 + h T In the formula T), ρ is the density, and g T h T is the fitting coefficient, and T is the absolute temperature.

4. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 1, characterized in that, The static evaporation rate is calculated based on the total heat loss using the following formula: BOR = Q(t) / L v In the formula, BOR is the static evaporation rate, in % / h; Q(t) is the total heat loss, in W; L v The latent heat of vaporization is expressed in J / kg.

5. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 1, characterized in that, The mapping model is as follows: PRR = a + b· BOR + c· Fill + d· t hold In the formula, PRR is the pressurization rate, in MPa / h; BOR is the static evaporation rate, in % / h; Fill is the initial fill rate; t hold , where is the settling time in hours; a, b, c, and d are the fitting coefficients.

6. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 1, characterized in that, The LSTM neural network model is as follows: In the formula, For the predicted output, i.e. the corrected thermal conductivity, x t The input feature vector includes temperature, pressure, and time; h t This is the hidden layer state vector.

7. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 3, characterized in that, The dynamic calibration algorithm includes: Deploy lightweight LSTM models on edge computing nodes; Real-time correction of fitting coefficients, calibration frequency ≥10Hz, temperature fluctuation suppression ratio ≥95%.

8. The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container according to claim 1, characterized in that, Inspection of high-pressure liquid hydrogen containers includes: Establish a supercritical liquid hydrogen density model: In the formula, ρ sc The density of liquid hydrogen, ρ 0 is the base density, p is the internal pressure of the container, a is the pressure coefficient, and b is the exponential factor. Added circumferential stress analysis module: In the formula, σ θ Let p be the circumferential stress, p be the internal pressure of the container, and r be the inner radius of the container. i Let t be the radius of the i-th layer of the container's inner structure, and t be the wall thickness.

9. A system for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container, characterized in that, The method for predicting heat leakage and detecting pressure rise characteristics of a liquid hydrogen container as described in any one of claims 1 to 8 includes: A multimodal sensing unit, comprising a fiber optic temperature sensor, a vacuum sensor, and a piezoelectric thin film sensor, wherein the fiber optic temperature sensor is used to monitor the temperature distribution of the liquid hydrogen container wall in real time. An edge computing unit, equipped with an ARM processor, is used to support the nonlinear mapping between real-time static evaporation rate and pressure boost rate; The backend server is configured with a GPU cluster for running COMSOL finite element simulation and response surface regression models.

10. The liquid hydrogen container heat leakage prediction and pressure boosting characteristic detection system according to claim 9, characterized in that, The fiber optic temperature sensor detects the temperature gradient by Bragg wavelength shift, the vacuum sensor detects the change in vacuum layer thickness by diffraction time difference method, and the piezoelectric thin film sensor is embedded in multiple insulation layers to monitor material creep.

Images