A hydrogen leakage monitoring device and method for an aerospace cryogenic environment

Abstract

The application provides a hydrogen leakage monitoring device and method for an aviation low-temperature environment, relates to the technical field of hydrogen leakage monitoring, and comprises a hydrogen monitoring sensor, a low-temperature temperature sensor, a silicone rubber resistance wire heater, a metal mesh cage, a temperature control module and a hydrogen monitoring and early warning module. The metal mesh cage is arranged in a region for monitoring hydrogen leakage. The hydrogen monitoring sensor and the low-temperature temperature sensor are both arranged in the interior of the metal mesh cage and are electrically connected with the hydrogen monitoring and early warning module and the temperature control module. The silicone rubber resistance wire heater is arranged on the metal mesh cage at intervals and is electrically connected with the temperature control module. The temperature control module is used for controlling the silicone rubber resistance wire heater to heat the gas around the metal mesh cage based on the ambient temperature. The hydrogen monitoring and early warning module is used for generating an early warning signal through the hydrogen concentration and the ambient temperature. The problem of hydrogen sensor failure caused by the fact that the aviation flight working environment temperature is lower than the hydrogen leakage monitoring sensor temperature is solved.

Description

Technical Field

[0001] This invention relates to the field of hydrogen leak monitoring technology, and specifically to a hydrogen leak monitoring device and method for use in cryogenic aviation environments. Background Technology

[0002] Hydrogen's unique physicochemical properties (small molecule, low viscosity, low density) make it prone to leakage during operation and transportation. If leaks are not detected in time, they can lead to serious safety hazards. Traditional hydrogen leak detection methods typically rely on electrochemical and catalytic combustion sensors, the performance of which is often affected by factors such as ambient temperature.

[0003] During high-altitude flight (e.g., civil aircraft typically cruise at an altitude of 11 km, with an ambient temperature of approximately -56°C), excessively low ambient temperatures can slow down or cause hydrogen sensors to malfunction (the lower limit of the permissible temperature for commonly used hydrogen concentration sensors is approximately -40°C), resulting in inaccurate monitoring results and an inability to effectively respond to hydrogen leak emergencies. Therefore, ensuring that hydrogen sensors can still operate accurately and stably under changing environmental conditions is a crucial technological challenge. Summary of the Invention

[0004] In view of this, embodiments of this specification provide a hydrogen leak monitoring device and method for use in low-temperature aviation environments, in order to solve the problem that the reaction rate of hydrogen sensors slows down or fails due to excessively low ambient temperatures during high-altitude flight, and to avoid the risk of failure of hydrogen leak monitoring caused by high-altitude low-temperature environments.

[0005] The embodiments in this specification provide the following technical solutions:

[0006] A hydrogen leak monitoring device for cryogenic aviation environments includes: Hydrogen monitoring sensor, low temperature sensor, silicone rubber resistance wire heater, metal mesh cage, temperature control module and hydrogen monitoring and early warning module; Metal mesh cages were installed in the area where hydrogen leaks were monitored; Both the hydrogen monitoring sensor and the low-temperature sensor are located inside the metal mesh cage and are electrically connected to the hydrogen monitoring and early warning module and the temperature control module. Multiple silicone rubber resistance wire heaters are spaced apart on a metal mesh cage and electrically connected to a temperature control module. The temperature control module is used to obtain the ambient temperature through a low-temperature temperature sensor, and based on the ambient temperature, to control the silicone rubber resistance wire heater to heat the gas around the metal mesh cage. The hydrogen monitoring and early warning module is used to generate early warning signals based on the hydrogen concentration obtained by the hydrogen monitoring sensor and the ambient temperature obtained by the low temperature sensor.

[0007] A method for detecting hydrogen leaks, used to detect hydrogen leaks based on a hydrogen leak detection device, includes: A hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the hydrogen leakage dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. The cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, which includes hydrogen concentration, concentration change rate, and ambient temperature. Calculate the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset, and update the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset. A multi-level runaway state classification model is constructed. The multi-level runaway state classification model is trained using historical hydrogen leakage datasets to generate a trained multi-level runaway state classification model. By training a multi-level runaway state classification model, hydrogen leakage is classified based on the hydrogen concentration obtained by the hydrogen monitoring sensor and the ambient temperature obtained by the low temperature sensor, and a classification result is generated. Based on the classification results, the hydrogen monitoring and early warning module generates a corresponding level of early warning signal.

[0008] Furthermore, a hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the hydrogen leakage dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. This cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, including: The raw data is acquired, the time interval is determined based on the sensor sampling frequency and the system response time, the concentration change rate is calculated by the hydrogen concentration and the time interval, and the concentration change rate is added to the raw data. After using linear interpolation to insert numerical values ​​into the original data with missing data, statistical methods are used to identify and remove abnormal original data. Determine the out-of-control state Y, calibrate the out-of-control state Y for each data point in the original data, and use the calibrated dataset as the historical hydrogen leak dataset.

[0009] Furthermore, the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset are calculated, including: Calculate the entropy of runaway state Y in historical hydrogen leak datasets. ,in, k represents the level of the out-of-control state. It is the frequency of occurrence of level k; Calculate the conditional entropy of each feature A. ,in, Feature A includes hydrogen concentration, rate of concentration change, and ambient temperature. These are all possible values ​​of feature A. When feature A takes the value The entropy of the runaway state Y, Feature A takes the value of The probability of; Through the entropy of the runaway state Y and the conditional entropy of each feature A The information gain IG(A) of each feature is calculated, where, ; The feature entropy H(A) of each feature A is calculated, where, ; The information gain ratio IGR(A) of each feature A is calculated using the information gain IG(A) of each feature A and the feature entropy H(A) of each feature A. ; The information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights of hydrogen concentration, concentration change rate, and ambient temperature.

[0010] Furthermore, the information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights for hydrogen concentration, concentration change rate, and ambient temperature, including: , Let i be the feature weight of the i-th feature. For the i-th feature, The characteristic weights for hydrogen concentration are: The characteristic weights are the concentration change rate. The feature weights are for ambient temperature.

[0011] Furthermore, a multi-level runaway state classification model is constructed. This model is trained using a historical hydrogen leak dataset, generating a post-trained multi-level runaway state classification model, including: Calculate the prior probabilities of the multi-level runaway state classification model, where, (Number of samples in the k-th class) / ALL For prior probability, The state is out of control, k is the level of the out-of-control state, and ALL is the total number of samples in the historical hydrogen leak dataset. Assuming that each feature follows a normal Gaussian distribution under a given state, for each level k and each feature i, calculate the mean μ{k,i} and standard deviation σ{k,i} of all samples in the historical hydrogen leak dataset for feature i at level k. The conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and the standard deviation σ{k,i}. The prior probability, conditional probability, mean μ{k,i}, standard deviation σ{k,i}, feature weights of hydrogen concentration, feature weights of concentration change rate, and feature weights of ambient temperature are used as parameters to train the multi-level runaway state classification model, generating the trained multi-level runaway state classification model.

[0012] Furthermore, the conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and standard deviation σ{k,i}, including: A conditional probability density function is constructed using the mean μ{k,i} and the standard deviation σ{k,i}, where, ), For conditional probability, For the i-th feature, Features The value of .

[0013] Furthermore, through a trained multi-level runaway state classification model, hydrogen leakage is classified based on the hydrogen concentration obtained from the hydrogen monitoring sensor and the ambient temperature obtained from the cryogenic temperature sensor, generating classification results, including: Hydrogen concentration is obtained through a hydrogen monitoring sensor; The ambient temperature is obtained using a low-temperature temperature sensor; The rate of change of concentration was calculated by measuring the hydrogen concentration. Feature vectors were constructed using hydrogen concentration, concentration change rate, and ambient temperature. , , Hydrogen concentration, The concentration change rate, Ambient temperature; The eigenvectors are analyzed using the mean μ{k,i} and standard deviation σ{k,i}. Standardize to generate standardized feature vectors. Where k is the level of the out-of-control state, and i is the feature; Through standardized feature vectors Calculate the posterior log probability of hydrogen concentration, the posterior log probability of the rate of change of concentration, and the posterior log probability of ambient temperature. By comparing the posterior log probability of hydrogen concentration, the posterior log probability of concentration change rate, and the posterior log probability of ambient temperature, the level k with the highest posterior log probability is taken as the classification result.

[0014] Furthermore, through the standardized feature vectors Calculate the posterior log probability of hydrogen concentration, the posterior log probability of the rate of change of concentration, and the posterior log probability of ambient temperature, including: level posterior log probability , For the first Level Y out of control For feature vectors, Features The weight, For prior probability, Features In state The conditional probability under the given conditions.

[0015] Furthermore, it also includes: We set overall accuracy, recall, and F1-Score, and used them as evaluation metrics. We then validated and optimized the multi-level runaway state classification model through distribution hypothesis testing, threshold optimization, and weight adjustment.

[0016] Compared with the prior art, the beneficial effects that at least one technical solution adopted in the embodiments of this specification can achieve include at least: This invention solves the problem of hydrogen sensor failure caused by the ambient temperature of aviation flight being lower than the temperature of the hydrogen leak monitoring sensor, and provides a safe, fast, and stable local temperature control device near the sensor. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an overall structural diagram of the hydrogen leakage monitoring device according to an embodiment of the present invention.

[0019] The attached figures are labeled as follows: 1. Hydrogen monitoring sensor; 2. Low temperature sensor; 3. Silicone rubber resistance wire heater; 4. Metal mesh cage; 5. Temperature control module; 6. Hydrogen monitoring and early warning module. Detailed Implementation

[0020] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] like Figure 1 As shown, a hydrogen leak monitoring device for cryogenic aviation environments includes: Hydrogen monitoring sensor 1, low temperature sensor 2, silicone rubber resistance wire heater 3, metal mesh cage 4, temperature control module 5, and hydrogen monitoring and early warning module 6.

[0023] A metal mesh cage 4 is installed in the area for monitoring hydrogen leakage. Both the hydrogen monitoring sensor 1 and the low-temperature sensor 2 are installed inside the metal mesh cage 4 and electrically connected to the hydrogen monitoring and early warning module 6 and the temperature control module 5. Multiple silicone rubber resistance wire heaters 3 are spaced apart on the metal mesh cage 4 and electrically connected to the temperature control module 5. The temperature control module 5 uses the low-temperature sensor 2 to obtain the ambient temperature and, based on the ambient temperature, controls the silicone rubber resistance wire heaters 3 to heat the gas surrounding the metal mesh cage 4. The hydrogen monitoring and early warning module 6 uses the hydrogen concentration obtained by the hydrogen monitoring sensor 1 and the ambient temperature obtained by the low-temperature sensor 2 to generate an early warning signal.

[0024] The hydrogen monitoring sensor 1 includes an electrochemical hydrogen sensor that can monitor the hydrogen concentration inside the enclosure in real time. When hydrogen leaks, the sensor detects the change in gas concentration and generates a signal output. These signals are then processed by the hydrogen monitoring and early warning module 6 to determine whether a hydrogen leak has occurred and to issue an alarm. The hydrogen monitoring sensor 1 is located inside the metal mesh cage 4 and is connected to the hydrogen monitoring and early warning module 6.

[0025] The silicone rubber resistance wire heater 3 can heat the low-temperature gas inside the metal mesh cage 4.

[0026] Temperature control module 5 heats the gas at low temperatures based on the current gas temperature state, ensuring that the gas temperature fluctuates within a suitable range monitored by the sensor. Temperature control module 5 controls the silicone rubber resistance wire heater 3 and the low-temperature sensor 2, adjusting the power of the heating elements based on real-time temperature monitoring information to maintain a stable local temperature environment.

[0027] The hydrogen monitoring and early warning module 6 analyzes the data collected by the hydrogen monitoring sensor 1, sets multiple levels of runaway states, calculates the current monitoring value of the hydrogen concentration parameter based on the multiple levels of runaway states, obtains the characteristic value that supports the occurrence of each level of runaway state, judges the runaway state of hydrogen leakage based on the characteristic value that supports the occurrence of each level of runaway state, determines the level of runaway state of the leak, and performs early warning processing. That is, based on the data monitored by hydrogen monitoring sensor 1 and temperature sensor, the system analyzes and integrates the three-dimensional features of "hydrogen concentration (C), concentration change rate (ΔC / Δt), and ambient temperature (T)" to achieve three-level classification of "minor leakage (C≤0.5% VOL, ΔC / Δt≤0.05% VOL / s), moderate leakage (0.5% VOL<C≤1% VOL, 0.05% VOL / s<ΔC / Δt≤0.1% VOL / s), and severe leakage (C>1% VOL, ΔC / Δt>0.1% VOL / s)" (i.e., the runaway state is divided into three levels, k=3), thus solving the problem of misjudgment caused by concentration fluctuations in low-temperature environments.

[0028] The multi-level runaway state classification algorithm consists of two steps: feature weight calculation and weighted Naive Bayes classification. For feature weight calculation, a historical hydrogen leak dataset (≥1000 samples) based on aviation cryogenic environments is constructed, and the information gain ratio (IGR) of each feature weight is calculated. The weighted Naive Bayes classification assumes that each feature condition is independent, constructing a posterior probability model for three levels of runaway states. The runaway state with the highest posterior probability is used as the current classification result, triggering the corresponding level of warning from the hydrogen monitoring and early warning module 6.

[0029] In an embodiment of the present invention, the hydrogen leakage monitoring method includes: A hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the hydrogen leakage dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. The cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, which includes hydrogen concentration, concentration change rate, and ambient temperature. Calculate the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset, and update the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset. A multi-level runaway state classification model is constructed. The multi-level runaway state classification model is trained using historical hydrogen leakage datasets to generate a trained multi-level runaway state classification model. The hydrogen leak was classified based on the hydrogen concentration obtained by hydrogen monitoring sensor 1 and the ambient temperature obtained by low temperature sensor 2 after training, and the classification results were generated. Based on the classification results, the hydrogen monitoring and early warning module 6 generates an early warning signal of the corresponding level. We set overall accuracy, recall, and F1-Score, and used them as evaluation metrics. We then validated and optimized the multi-level runaway state classification model through distribution hypothesis testing, threshold optimization, and weight adjustment.

[0030] Specifically, a hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the hydrogen leakage dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. This cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, including: The raw data is acquired, the time interval is determined based on the sensor sampling frequency and the system response time, the concentration change rate is calculated by the hydrogen concentration and the time interval, and the concentration change rate is added to the raw data. After using linear interpolation to insert numerical values ​​into the original data with missing data, statistical methods are used to identify and remove abnormal original data. Determine the out-of-control state Y, calibrate the out-of-control state Y for each data point in the original data, and use the calibrated dataset as the historical hydrogen leak dataset.

[0031] Specifically, the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset are calculated, including: Calculate the entropy of runaway state Y in historical hydrogen leak datasets. ,in, k represents the level of the out-of-control state. It is the frequency of occurrence of level k; Calculate the conditional entropy of each feature A. ,in, Feature A includes hydrogen concentration, rate of concentration change, and ambient temperature. These are all possible values ​​of feature A. When feature A takes the value The entropy of the runaway state Y, Feature A takes the value of The probability of; Through the entropy of the runaway state Y and the conditional entropy of each feature A The information gain IG(A) of each feature is calculated, where, ; The feature entropy H(A) of each feature A is calculated, where, ; The information gain ratio IGR(A) of each feature A is calculated using the information gain IG(A) of each feature A and the feature entropy H(A) of each feature A. ; The information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights of hydrogen concentration, concentration change rate, and ambient temperature.

[0032] Specifically, the information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights for hydrogen concentration, concentration change rate, and ambient temperature, including: , Let i be the feature weight of the i-th feature. For the i-th feature, The characteristic weights for hydrogen concentration are: The characteristic weights are the concentration change rate. The feature weights are for ambient temperature.

[0033] Specifically, a multi-level runaway state classification model is constructed. This model is trained using historical hydrogen leak datasets to generate a post-trained multi-level runaway state classification model, including: Calculate the prior probabilities of the multi-level runaway state classification model, where, (Number of samples in the k-th class) / ALL For prior probability, The state is out of control, k is the level of the out-of-control state, and ALL is the total number of samples in the historical hydrogen leak dataset. Assuming that each feature follows a normal Gaussian distribution under a given state, for each level k and each feature i, calculate the mean μ{k,i} and standard deviation σ{k,i} of all samples in the historical hydrogen leak dataset for feature i at level k. The conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and the standard deviation σ{k,i}. The prior probability, conditional probability, mean μ{k,i}, standard deviation σ{k,i}, feature weights of hydrogen concentration, feature weights of concentration change rate, and feature weights of ambient temperature are used as parameters to train the multi-level runaway state classification model, generating the trained multi-level runaway state classification model.

[0034] Specifically, the conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and standard deviation σ{k,i}, including: A conditional probability density function is constructed using the mean μ{k,i} and the standard deviation σ{k,i}, where, ), For conditional probability, For the i-th feature, Features The value of .

[0035] In one embodiment of the present invention, the weights are determined after training: hydrogen concentration C (weight 0.45), concentration change rate ΔC / Δt (weight 0.35), and ambient temperature T (weight 0.2; at low temperatures, the lower the T, the greater the interference with concentration monitoring, and the higher the weight).

[0036] Specifically, through a trained multi-level runaway state classification model, hydrogen leakage is classified based on the hydrogen concentration obtained by hydrogen monitoring sensor 1 and the ambient temperature obtained by low-temperature temperature sensor 2, generating classification results, including: The hydrogen concentration is obtained through hydrogen monitoring sensor 1; The ambient temperature is obtained through low-temperature sensor 2; The rate of change of concentration was calculated by measuring the hydrogen concentration. Feature vectors were constructed using hydrogen concentration, concentration change rate, and ambient temperature. , , Hydrogen concentration, The concentration change rate, Ambient temperature; The eigenvectors are analyzed using the mean μ{k,i} and standard deviation σ{k,i}. Standardize to generate standardized feature vectors. Where k is the level of the out-of-control state, and i is the feature; Through standardized feature vectors Calculate the posterior log probability of hydrogen concentration, the posterior log probability of the rate of change of concentration, and the posterior log probability of ambient temperature. By comparing the posterior log probability of hydrogen concentration, the posterior log probability of concentration change rate, and the posterior log probability of ambient temperature, the level k with the highest posterior log probability is taken as the classification result.

[0037] Specifically, through the standardized feature vector Calculate the posterior log probability of hydrogen concentration, the posterior log probability of the rate of change of concentration, and the posterior log probability of ambient temperature, including: level posterior log probability , For the first Level Y out of control For feature vectors, Features The weight, For prior probability, Features In state The conditional probability under the given conditions.

[0038] Beneficial effects of the embodiments of the present invention: This invention addresses the problem of hydrogen sensor failure caused by ambient temperatures in aviation operations being lower than the temperature of the hydrogen leak monitoring sensor. It provides a safe, fast, and stable method for controlling the local temperature near the sensor, effectively ensuring sensor reliability. The device employs a silicone rubber-wrapped resistance wire design, effectively preventing electrical sparks in the hydrogen environment while achieving good thermal conductivity, thus ensuring the safety of hydrogen leak monitoring. This invention can be used in aviation environments with significant temperature variations, ensuring stable operation of the low-temperature hydrogen sensor within a suitable temperature range, thereby improving the accuracy of hydrogen leak detection. The hydrogen leak monitoring device uses a non-fully enclosed structure design, improving airflow and avoiding gas accumulation and poor flow, ensuring real-time monitoring of hydrogen leaks. It integrates three-dimensional features—hydrogen concentration (C), concentration change rate (ΔC / Δt), and ambient temperature (T)—to achieve multi-level classification of runaway states.

[0039] The above description is merely a specific embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any substitution of equivalent components or equivalent changes and modifications made within the scope of protection of this patent should still fall within the scope of this patent. Furthermore, the technical features, technical features and technical solutions, and technical solutions in this invention can be freely combined and used.

Claims

1. A method for monitoring hydrogen leaks in cryogenic aviation environments, used to monitor hydrogen leaks based on a hydrogen leak monitoring device, characterized in that, The hydrogen leakage monitoring device includes: Hydrogen monitoring sensor (1), low temperature sensor (2), silicone rubber resistance wire heater (3), metal mesh cage (4), temperature control module (5) and hydrogen monitoring and early warning module (6). The metal mesh cage (4) is installed in the area where hydrogen leakage is monitored; The hydrogen monitoring sensor (1) and the low temperature sensor (2) are both installed inside the metal mesh cage (4) and are electrically connected to the hydrogen monitoring and early warning module (6) and the temperature control module (5). Multiple silicone rubber resistance wire heaters (3) are spaced apart on the metal mesh cage (4) and electrically connected to the temperature control module (5); The temperature control module (5) is used to obtain the ambient temperature through the low temperature sensor (2), and based on the ambient temperature, control the silicone rubber resistance wire heater (3) to heat the gas around the metal mesh cage (4); The hydrogen monitoring and early warning module (6) is used to generate an early warning signal based on the hydrogen concentration obtained by the hydrogen monitoring sensor (1) and the ambient temperature obtained by the low temperature sensor (2). The hydrogen leakage monitoring method includes the following steps: A hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the hydrogen leakage dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. The cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, wherein the historical hydrogen leakage dataset includes hydrogen concentration, concentration change rate, and ambient temperature. Calculate the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset, and update the feature weights of hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset. A multi-level runaway state classification model is constructed. The multi-level runaway state classification model is trained using the historical hydrogen leakage dataset to generate a trained multi-level runaway state classification model. Using the trained multi-level runaway state classification model, hydrogen leakage is classified based on the hydrogen concentration obtained by the hydrogen monitoring sensor (1) and the ambient temperature obtained by the low-temperature temperature sensor (2), generating classification results, including: The hydrogen concentration is obtained through the hydrogen monitoring sensor (1); the ambient temperature is obtained through the low-temperature temperature sensor (2); the concentration change rate is calculated based on the hydrogen concentration; and a feature vector is constructed using the hydrogen concentration, the concentration change rate, and the ambient temperature. , , Hydrogen concentration, The concentration change rate, The ambient temperature; the eigenvectors are analyzed using the mean μ{k,i} and standard deviation σ{k,i}. Standardize to generate standardized feature vectors. Where k is the level of the runaway state, and i is a feature; the standardized feature vector... Calculate the posterior log probability of hydrogen concentration, the posterior log probability of concentration change rate, and the posterior log probability of ambient temperature; compare the magnitudes of the posterior log probabilities of hydrogen concentration, concentration change rate, and ambient temperature, and take the level k with the largest posterior log probability as the classification result. Based on the classification results, the hydrogen monitoring and early warning module (6) generates an early warning signal of the corresponding level.

2. The hydrogen leakage monitoring method according to claim 1, characterized in that, A hydrogen leakage dataset under cryogenic aviation conditions is constructed, and the dataset is cleaned to generate a cleaned historical hydrogen leakage dataset. This cleaned historical hydrogen leakage dataset is then standardized to generate a historical hydrogen leakage dataset, including: The raw data is acquired, the time interval is determined based on the sensor sampling frequency and the system response time, the concentration change rate is calculated using the hydrogen concentration and the time interval, and the concentration change rate is added to the raw data. After inserting numerical values ​​into the original data with missing data using linear interpolation, statistical methods are used to identify and remove abnormal original data. Determine the out-of-control state Y, calibrate the out-of-control state Y for each data point in the original data, and use the calibrated dataset as the historical hydrogen leak dataset.

3. The hydrogen leakage monitoring method according to claim 1, characterized in that, The feature weights for hydrogen concentration, concentration change rate, and ambient temperature in the historical hydrogen leak dataset are calculated, including: Calculate the entropy of the runaway state Y in the historical hydrogen leak dataset. ,in, k represents the level of the out-of-control state. It is the frequency of occurrence of level k; Calculate the conditional entropy of each feature A. ,in, Feature A includes the hydrogen concentration, the rate of change of concentration, and the ambient temperature. These are all possible values ​​of feature A. When feature A takes the value The entropy of the runaway state Y, Feature A takes the value of The probability of; The entropy of the runaway state Y and the conditional entropy of each feature A The information gain IG(A) of each feature is calculated, where, ; The feature entropy H(A) of each feature A is calculated, where, ; The information gain ratio IGR(A) of each feature A is calculated using the information gain IG(A) of each feature and the feature entropy H(A) of each feature A, where, ; The information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights of hydrogen concentration, concentration change rate, and ambient temperature.

4. The hydrogen leakage monitoring method according to claim 3, characterized in that, The information gain ratio (IGR(A)) of all features is normalized to obtain the feature weights for hydrogen concentration, concentration change rate, and ambient temperature, including: , Let i be the feature weight of the i-th feature. For the i-th feature, The characteristic weights for hydrogen concentration are: The characteristic weights are the concentration change rate. The feature weights are for ambient temperature.

5. The hydrogen leakage monitoring method according to claim 1, characterized in that, A multi-level runaway state classification model is constructed. This model is trained using the historical hydrogen leakage dataset to generate a post-trained multi-level runaway state classification model, including: Calculate the prior probabilities of the multi-level runaway state classification model, where, (Number of samples in the k-th class) / ALL For prior probability, The state is out of control, k is the level of the out-of-control state, and ALL is the total number of samples in the historical hydrogen leak dataset; Assuming that each feature follows a normal Gaussian distribution under a given state, for each level k and each feature i, calculate the mean μ{k,i} and standard deviation σ{k,i} of all samples in the historical hydrogen leak dataset for feature i at level k. The conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and the standard deviation σ{k,i}. The prior probability, the conditional probability, the mean μ{k,i}, the standard deviation σ{k,i}, the feature weights of the hydrogen concentration, the feature weights of the concentration change rate, and the feature weights of the ambient temperature are used as parameters to train the multi-level runaway state classification model, thereby generating a trained multi-level runaway state classification model.

6. The hydrogen leakage monitoring method according to claim 5, characterized in that, The conditional probability of the multi-level runaway state classification model is calculated using the mean μ{k,i} and the standard deviation σ{k,i}, including: A conditional probability density function is constructed using the mean μ{k,i} and the standard deviation σ{k,i}, where, , For conditional probability, For the i-th feature, Features The value of .

7. The hydrogen leakage monitoring method according to any one of claims 1 to 6, characterized in that, Also includes: The overall accuracy, recall, and F1-Score are set, and the overall accuracy, recall, and F1-Score are used as evaluation indicators. The multi-level runaway state classification model is verified and optimized through distribution hypothesis testing, threshold optimization, and weight adjustment.

Images