A hydrogen fuel cell engine fault diagnosis system

Abstract

The present invention discloses a hydrogen fuel cell engine fault diagnosis system, which relates to the technical field of fault diagnosis and includes an analysis module, a construction module, and a compensation module. The system extracts a first feature, performs cluster analysis to obtain a data class, calculates a first probability, obtains a target data class, establishes a first mapping relationship and a second mapping relationship, constructs an interference degree model, and compensates and updates the second mapping relationship. The present invention reduces misjudgments through cluster analysis and probability calculation, takes external interference factors into account, and improves diagnostic accuracy. The system can dynamically update the mapping relationship and interference degree model based on historical data to adapt to changes in the operating state of the equipment. By establishing a mapping relationship between the first feature and the fault type, new fault modes can be quickly identified. By analyzing the characteristics of vibration intensity, equipment downtime can be reduced. Combined with cluster analysis and probability calculation, the system can handle complex fault modes and improve overall robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a hydrogen fuel cell engine fault diagnosis system. Background Art

[0002] In recent years, nondestructive testing (NDT) technology has made significant progress in hydrogen fuel cell fault diagnosis. A research team at the University of Science and Technology of China (USTC) has developed a NDT based on electromagnetic field changes. By attaching a magnetic sensor matrix to the exterior of the PEMFC (Peripheral Electron Molecular Flux) (PEMFC), NDT collects magnetic field distribution data for fault diagnosis. This technology not only avoids the interference with battery performance associated with traditional invasive testing methods, but also enables real-time monitoring of dynamic changes in battery status, resulting in high diagnostic accuracy and practicality.

[0003] Currently, a Chinese invention patent with the announcement number CN113594510B discloses a fault diagnosis method for a fuel cell system based on a cloud platform. The method integrates and analyzes vehicle operation data and hydrogenation data through the cloud platform, and realizes rapid fault location and diagnosis for fuel cell engine performance failure caused by hydrogenation environmental pollution, etc., and solves the problem of rapid location and diagnosis of fuel cell engine performance failure or fault caused by hydrogen fuel purity problems. However, the related technology does not verify the fault according to different signals, which is not conducive to the accuracy of diagnosis. It does not compensate for the measurement error caused by interference factors such as the environment for the operation data, lacks the accuracy of judging the data source, and has certain limitations. Summary of the Invention

[0004] The technical problem solved by the present invention is that the related technology does not verify the fault according to different signals, which is not conducive to the accuracy of diagnosis, does not compensate for the measurement error caused by the operating data according to interference factors such as the environment, lacks the accuracy of judging the source of the data, and has certain limitations.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a hydrogen fuel cell engine fault diagnosis system, comprising an analysis module, a construction module and a compensation module;

[0006] The analysis module extracts the first feature of the historical vibration intensity and performs cluster analysis on the corresponding historical operation data to obtain data classes;

[0007] The construction module obtains a first probability of each data class corresponding to the first feature according to the data class, obtains a target data class corresponding to the first feature according to the first probability, establishes a first mapping relationship between the first feature and the corresponding historical fault type, and establishes a second mapping relationship between the target data class and the first feature;

[0008] The compensation module constructs an interference degree model according to historical natural data, and compensates and updates the second mapping relationship according to the interference degree model.

[0009] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system of the present invention, wherein: the analysis module obtains historical fault types and their corresponding historical vibration intensities;

[0010] The historical fault types include deionized glycol inlet low pressure fault, current set point deviation fault, deionized glycol inlet high temperature fault, hydrogen inlet and outlet low pressure difference fault, hydrogen supply pressure low fault, reverse stack current fault, hydrogen exhaust voltage low fault, air pressure low fault, hydrogen leakage fault, stack voltage low fault and deionized glycol outlet temperature voltage high fault;

[0011] The historical operating data include inlet hydrogen pressure, inlet coolant temperature, total stack current, stack voltage, coolant flow, stack outlet hydrogen pressure, stack inlet hydrogen pressure, coolant inlet pressure, air inlet pressure, stack coolant outlet temperature, stack coolant inlet temperature, air compressor inlet temperature, air compressor outlet temperature, air compressor voltage and air compressor current.

[0012] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system of the present invention, the extraction logic of the first feature includes:

[0013] Obtain any historical fault type and its corresponding historical vibration intensity, remove outliers and normalize the historical vibration intensity so that the values ​​of each historical vibration intensity are distributed between 0 and 1;

[0014] Each historical vibration intensity is converted into frequency domain data through Fourier transform, the power spectrum density of the historical vibration intensity is calculated through autocorrelation analysis, and the power spectrum density is set as the first feature of the historical fault type.

[0015] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system described in the present invention, the clustering logic of the data class includes:

[0016] Obtaining any type of historical operating data corresponding to the historical fault type, calculating an average value of the historical operating data, setting a first value as a change gradient, calculating a first sum of the average value and the first value, calculating a first difference between the average value and the first value, setting the average value, the first sum, and the first difference as first cluster centers, respectively, assigning each piece of historical operating data to the first cluster center that is closest to it, where the closest distance is represented by the smallest absolute value of the difference between the historical operating data and the cluster center, to obtain three first classifications;

[0017] Calculate the average value of each first classification, set the average value of each first classification as the second cluster center, and classify each historical running data into the second cluster center with the closest distance, to obtain three second classifications;

[0018] Compare the data in the first classification with the data in the second classification. When the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification. Otherwise, repeat the above steps until the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification.

[0019] The final classification is set as the data class.

[0020] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system of the present invention, the calculation logic of the first probability includes:

[0021] Obtaining a first feature corresponding to the historical fault type, obtaining any type of historical operation data corresponding to the first feature, and classifying each piece of historical operation data into a corresponding data class;

[0022] Count the first times the same data type appears, and the total number of times the data type appears;

[0023] Calculate the ratio of the first number to the total number of times and set it as the first probability of the corresponding data class for the first feature.

[0024] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system described in the present invention, the acquisition logic of the target data class includes:

[0025] The first probabilities are sorted in descending order, a data class corresponding to the first probability with the largest value is selected, and the data class is set as the target data class corresponding to the first feature.

[0026] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system described in the present invention, wherein: by inputting the current first feature into the first mapping relationship, the corresponding fault type is obtained, and by inputting the current first feature into the second mapping relationship, the corresponding target data class is obtained.

[0027] As a preferred embodiment of the hydrogen fuel cell engine fault diagnosis system described in the present invention, the system comprises: obtaining current operating data, classifying the current operating data to obtain a current data class, comparing the current data class with a target data class, counting a first number of the current data class and the target data class being identical, setting the number of types of the current operating data to a second number, calculating a second ratio of the first number to the second number, setting the second value to a second ratio threshold, and comparing the second ratio with the second value;

[0028] When the second ratio is greater than or equal to the second value, the corresponding fault type is set as the current fault type;

[0029] Otherwise, a no-fault signal is output.

[0030] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system of the present invention, wherein: the historical natural data includes ambient temperature and ambient humidity;

[0031] The logic for compensating the second mapping relationship includes:

[0032] Set the first temperature and the first humidity as the standard temperature and standard humidity, obtain the first characteristic quantity of the historical fault type corresponding to the standard temperature and standard humidity and the average value of the historical operation data of any type, calculate the third difference and the fourth difference between any temperature and any humidity and the standard temperature and standard humidity, calculate the sum of the third difference and the fourth difference, set the sum of the third difference and the fourth difference as the independent variable, obtain the corresponding historical operation data at any temperature and any humidity, calculate the average value of the corresponding historical operation data at any temperature and any humidity, calculate the difference between the average value of the historical operation data corresponding to the standard temperature and standard humidity and the average value of the corresponding historical operation data at any temperature and any humidity, record it as the target difference, set the sum of the third difference and the fourth difference as the independent variable, use the target difference as the dependent variable, construct an interference degree model, use the coefficient of the interference degree model as the compensation coefficient, calculate the difference between 1 and the compensation coefficient, record it as the proportional parameter, calculate the product of the corresponding historical operation data and the proportional parameter, record it as the first product, obtain the target data class corresponding to the first product, and update the data class in the second mapping relationship to the target data class corresponding to the first product.

[0033] As a preferred solution of the hydrogen fuel cell engine fault diagnosis system of the present invention, the calculation expression of the interference degree model is:

[0034] ;

[0035] Wherein, Y is the target difference, X is the sum of the third difference and the fourth difference, C is the compensation coefficient, the third difference is the difference between the temperature and the standard temperature, and the fourth difference is the difference between the humidity and the standard humidity.

[0036] The beneficial effects of the present invention are as follows: through cluster analysis and probability calculation, the system can analyze fault characteristics from multiple dimensions, reduce misjudgments, take into account external interference factors (such as environmental noise, equipment aging, etc.), and can dynamically adjust the diagnosis model to further improve the accuracy of diagnosis. It can dynamically update the mapping relationship and interference model based on historical data to adapt to changes in the operating status of the equipment and avoid diagnostic failures caused by equipment aging or environmental changes. By establishing a mapping relationship between the first feature and the fault type, the system can quickly identify new fault modes and adjust the diagnosis strategy in a timely manner. By analyzing the characteristics of vibration intensity, the system can detect potential faults in advance and reduce equipment downtime. Combined with cluster analysis and probability calculation, the system can handle complex fault modes and improve overall robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of the basic flow of a hydrogen fuel cell engine fault diagnosis system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0039] Example, see Figure 1 , as one embodiment of the present invention, provides a hydrogen fuel cell engine fault diagnosis system, including an analysis module, a construction module and a compensation module;

[0040] The analysis module extracts the first feature of the historical vibration intensity and performs cluster analysis on the corresponding historical operation data to obtain data classes;

[0041] The construction module obtains a first probability of each data class corresponding to the first feature according to the data class, obtains a target data class corresponding to the first feature according to the first probability, establishes a first mapping relationship between the first feature and the corresponding historical fault type, and establishes a second mapping relationship between the target data class and the first feature;

[0042] The compensation module constructs an interference degree model according to historical natural data, and compensates and updates the second mapping relationship according to the interference degree model.

[0043] Through cluster analysis and probability calculation, the present invention enables the system to analyze fault characteristics from multiple dimensions, reduce misjudgments, take into account external interference factors (such as environmental noise, equipment aging, etc.), and dynamically adjust the diagnosis model to further improve the accuracy of diagnosis. It can dynamically update the mapping relationship and interference degree model based on historical data to adapt to changes in the equipment's operating status and avoid diagnostic failures caused by equipment aging or environmental changes. By establishing a mapping relationship between the first feature and the fault type, the system can quickly identify new fault modes and adjust the diagnosis strategy in a timely manner. By analyzing the characteristics of vibration intensity, the system can detect potential faults in advance and reduce equipment downtime. Combined with cluster analysis and probability calculation, the system can handle complex fault modes and improve overall robustness.

[0044] The analysis module obtains historical fault types and their corresponding historical vibration intensities;

[0045] Historical fault types include deionized glycol inlet low pressure fault, current set point deviation fault, deionized glycol inlet high temperature fault, low hydrogen inlet and outlet pressure differential fault, low hydrogen supply pressure fault, reverse stack current fault, low hydrogen exhaust voltage fault, low air pressure fault, hydrogen leakage fault, low stack voltage fault, and deionized glycol outlet temperature and voltage high fault;

[0046] Historical operating data includes inlet hydrogen pressure, inlet coolant temperature, total stack current, stack voltage, coolant flow rate, stack outlet hydrogen pressure, stack inlet hydrogen pressure, coolant inlet pressure, air inlet pressure, stack coolant outlet temperature, stack coolant inlet temperature, air compressor inlet temperature, air compressor outlet temperature, air compressor voltage, and air compressor current.

[0047] The extraction logic of the first feature includes:

[0048] Obtain any historical fault type and its corresponding historical vibration intensity, remove outliers and normalize the historical vibration intensity so that the values ​​of each historical vibration intensity are distributed between 0 and 1;

[0049] Each historical vibration intensity is converted into frequency domain data through Fourier transform, and the power spectrum density of the historical vibration intensity is calculated through autocorrelation analysis. The power spectrum density is set as the first feature of the historical fault type.

[0050] In specific implementations, outliers in vibration intensity data are identified and removed through statistical methods (such as the 3σ principle) or visualization methods (such as boxplots). This step reduces the impact of noise on subsequent analysis and ensures the reliability of feature extraction. Normalization eliminates the influence of different dimensions and ranges, giving the data a uniform numerical range for subsequent analysis. Frequency domain analysis captures periodic and frequency characteristics that are difficult to detect in time domain signals, helping to distinguish different fault modes. Power spectral density clearly displays the frequency components and energy distribution of the signal and is an important indicator for characterizing vibration signals. As the first feature, power spectral density has a clear physical meaning (the energy distribution of the frequency components), making it easier for engineers to understand and interpret diagnostic results, further improving diagnostic efficiency. Combined with subsequent compensation modules, this feature extraction method can adapt to complex operating conditions such as equipment aging and environmental changes, further improving system adaptability.

[0051] The clustering logic of the data class includes:

[0052] Obtain any type of historical operating data corresponding to a historical fault type, calculate the average value of the historical operating data, set the first value as the change gradient, calculate the first sum of the average value and the first value, calculate the first difference between the average value and the first value, set the average value, the first sum, and the first difference as the first cluster center, respectively, assign each historical operating data to the first cluster center that is closest to it, where the closest distance is represented by the smallest absolute value of the difference between the historical operating data and the cluster center, and obtain three first classifications;

[0053] Calculate the average value of each first classification, set the average value of each first classification as the second cluster center, and classify each historical running data into the second cluster center with the closest distance, to obtain three second classifications;

[0054] Compare the data in the first classification with the data in the second classification. When the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification. Otherwise, repeat the above steps until the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification.

[0055] Set the final classification to the data class.

[0056] In specific implementation, by calculating the average value of historical operating data and combining it with gradient changes, the initial cluster centers are scientifically set, reducing the impact of randomness on the clustering results. Through multiple iterative optimizations of clustering, the data distribution is more accurately identified, improving clustering accuracy. By comparing the first and second classifications, the stability of the clustering results is ensured. Iterations are only stopped when the two clustering results are the same, ensuring the stability and reliability of the clustering results. By changing the gradient setting, it can adapt to different data distributions and working conditions, improving the adaptability of the clustering method.

[0057] The calculation logic of the first probability includes:

[0058] Obtain a first feature corresponding to a historical fault type, obtain any type of historical operation data corresponding to the first feature, and classify each historical operation data into a corresponding data class;

[0059] Count the first times the same data type appears, and the total number of times the data type appears;

[0060] Calculate the ratio of the first number to the total number of times and set it as the first probability of the corresponding data class for the first feature.

[0061] The acquisition logic of the target data class includes:

[0062] The first probabilities are sorted in descending order, the data class corresponding to the first probability with the largest value is selected, and the data class is set as the target data class corresponding to the first feature.

[0063] By inputting the current first feature into the first mapping relationship, the corresponding fault type is obtained, and by inputting the current first feature into the second mapping relationship, the corresponding target data class is obtained.

[0064] In specific implementations, the probability of each data class occurring is calculated by counting the number and total number of occurrences of the same data class. This probability-based diagnostic method more accurately reflects the likelihood of a fault occurring and improves the accuracy of fault diagnosis. The probabilistic model is robust to noise and random interference because it focuses on the statistical patterns of the data rather than specific individual differences. This probability-based diagnostic method can effectively respond to external interference and reduce false positives and false negatives. Probability calculation is an efficient mathematical operation that quickly produces results. This efficient calculation method significantly shortens fault diagnosis time and improves the real-time performance of the system. It is applicable to a variety of fault modes and operating conditions and can effectively identify different types of faults. This probability-based diagnostic method has broad applicability and adapts to complex operating conditions such as equipment aging and environmental changes.

[0065] Obtaining current running data, classifying the current running data to obtain a current data class, comparing the current data class with a target data class, counting a first number of the current data class and the target data class being identical, setting the number of types of the current running data to a second number, calculating a second ratio of the first number to the second number, setting the second value as a second ratio threshold, and comparing the second ratio with the second value;

[0066] When the second ratio is greater than or equal to the second value, the corresponding fault type is set as the current fault type;

[0067] Otherwise, a no-fault signal is output.

[0068] Among them, the first quantity represents the same number of the current data class and the target data class. Since the data class and the target data class are obtained through the clustering algorithm, and the clustering algorithm updates the distance between the data and the cluster center until the iteration does not change the cluster center, the cluster center and the data class are both represented as numerical values. For example, in the K-means clustering algorithm, the basic process is as follows: initialize the cluster center, randomly select K data points as the initial cluster centers (centroids), assign data points to the nearest cluster center, for each data point, calculate its distance to each cluster center (usually using Euclidean distance), assign the data point to the nearest cluster center to form K clusters, update the cluster center, and for each cluster, calculate the average of all data points in the cluster. The mean is used as the new cluster center, and the steps are repeated and iterated continuously until the cluster center no longer changes (or changes very little) or the maximum number of iterations is reached. The present application obtains a data class through this algorithm. The first mapping relationship is expressed as the correspondence between the characteristics of the vibration data (first characteristics) and the types of historical faults. Specifically, any first characteristic may correspond to multiple types of historical faults. Therefore, the existing technology cannot obtain the fault type through vibration characteristics, but determines the fault characteristics by studying and calculating multidimensional data. The technical innovation of the present invention lies in mapping the fault characteristics through vibration signals, and obtaining the final and unique (target data class) by jointly mapping the fault characteristics with the numerical data class of the fault type classified by the operating data (second mapping), thereby determining the fault type;

[0069] Input each current operation data (current operation data of time series) into the first mapping relationship. Each set of current operation data can obtain multiple or one fault types, so each current operation data obtains multiple fault types. By inputting any current operation data into the second mapping relationship, a unique fault type is obtained (there may be multiple fault types occurring at the same time, so the unique representation is a unique group of fault types). The current data class is equivalent to a series of repeated values ​​(vector series), and the target data class is equivalent to a unique value (vector). The first value is expressed as: The second quantity is the same as the number of the only numerical value (vector), and is expressed as the total quantity contained in the sequence (vector sequence). Therefore, the ratio of the first quantity to the second quantity is calculated. In fact, in order to calculate the accuracy of the fault judgment model, the fault type is output and displayed only when the accuracy is greater than or equal to the second value. When the accuracy is less than the second value, the second mapping relationship is compensated. This simplifies the complex steps of calculating the power spectrum density, improves the speed of calculation, improves the accuracy of fault judgment, avoids the computational complexity of indifferent compensation, improves the rapid responsiveness of the system, and has significant beneficial effects.

[0070] The first number refers to the number of elements in the current data class obtained by classifying the current running data that are identical to the target data class. Specifically, the current data class obtained by classifying the current running data is compared with the target data class obtained by the second mapping relationship, and the number of elements in the current data class that are identical to the target data class is counted.

[0071] In one implementation, the system acquired and classified the current operating data, resulting in the current data class: [2.1, 2.1, 3.0, 2.1, 3.0, 4.2]. The target data class is 2.1. The current data class is: [2.1, 2.1, 3.0, 2.1, 3.0, 4.2]. The target data class is: 2.1. The first number is the number of elements in the current data class that are equal to 2.1, that is, 3. This first number will be used in subsequent ratio calculations to determine whether to trigger a fault signal.

[0072] In the specific implementation, by classifying the current operating data into the corresponding data class, the fault mode is identified more accurately, and by calculating the second ratio and comparing it with the second value, it is scientifically determined whether a fault has occurred, thereby improving the accuracy of fault diagnosis. The setting of the second ratio threshold is adjusted according to the actual working conditions and diagnostic requirements, thereby enhancing the robustness of the system. When the second ratio is less than the second value, a no-fault signal is output, thereby avoiding false alarms and improving the stability of the system.

[0073] Historical natural data include ambient temperature and ambient humidity;

[0074] The logic for compensating the second mapping relationship includes:

[0075] Set the first temperature and the first humidity as the standard temperature and standard humidity, obtain the first characteristic quantity of the corresponding historical fault type under the standard temperature and standard humidity and the average value of the historical operation data of any type, calculate the third difference and the fourth difference between any temperature and any humidity and the standard temperature and standard humidity, calculate the sum of the third difference and the fourth difference, set the sum of the third difference and the fourth difference as the independent variable, obtain the corresponding historical operation data under any temperature and any humidity, calculate the average value of the corresponding historical operation data under any temperature and any humidity, calculate the difference between the average value of the historical operation data under the standard temperature and standard humidity and the average value of the corresponding historical operation data under any temperature and any humidity, record it as the target difference, set the sum of the third difference and the fourth difference as the independent variable, take the target difference as the dependent variable, construct an interference degree model, use the coefficient of the interference degree model as the compensation coefficient, calculate the difference between 1 and the compensation coefficient, record it as the proportional parameter, calculate the product of the corresponding historical operation data and the proportional parameter, record it as the first product, obtain the target data class corresponding to the first product, and update the data class in the second mapping relationship to the target data class corresponding to the first product.

[0076] The calculation expression of the interference model is:

[0077] ;

[0078] Wherein, Y is the target difference, X is the sum of the third difference and the fourth difference, C is the compensation coefficient, the third difference is the difference between the temperature and the standard temperature, and the fourth difference is the difference between the humidity and the standard humidity.

[0079] In specific implementation, by considering the impact of ambient temperature and humidity on fault characteristics, the interference of external environmental changes on the diagnosis results is compensated, and the fault diagnosis results are dynamically adjusted based on the interference degree model, which can more accurately reflect the actual operating status of the equipment. The impact of temperature and humidity changes on equipment operation is taken into account, so that the system can adapt to more complex operating environments. The compensation coefficient and proportional parameters are quickly calculated through the interference degree model, which improves the diagnostic efficiency, avoids unnecessary maintenance due to misdiagnosis caused by environmental interference, and reduces maintenance costs.

[0080] Through cluster analysis and probability calculation, the present invention enables the system to analyze fault characteristics from multiple dimensions, reduce misjudgments, take into account external interference factors (such as environmental noise, equipment aging, etc.), and dynamically adjust the diagnosis model to further improve the accuracy of diagnosis. It can dynamically update the mapping relationship and interference degree model based on historical data to adapt to changes in the equipment's operating status and avoid diagnostic failures caused by equipment aging or environmental changes. By establishing a mapping relationship between the first feature and the fault type, the system can quickly identify new fault modes and adjust the diagnosis strategy in a timely manner. By analyzing the characteristics of vibration intensity, the system can detect potential faults in advance and reduce equipment downtime. Combined with cluster analysis and probability calculation, the system can handle complex fault modes and improve overall robustness.

[0081] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that modifications or equivalent replacements of the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A hydrogen fuel cell engine fault diagnosis system, characterized in that: Includes analysis module, construction module and compensation module; The analysis module extracts the first feature of the historical vibration intensity and performs cluster analysis on the corresponding historical operation data to obtain data classes; The construction module obtains a first probability of each data class corresponding to the first feature according to the data class, obtains a target data class corresponding to the first feature according to the first probability, establishes a first mapping relationship between the first feature and the corresponding historical fault type, and establishes a second mapping relationship between the target data class and the first feature; The compensation module constructs an interference degree model according to historical natural data, and compensates and updates the second mapping relationship according to the interference degree model; The calculation logic of the first probability includes: Obtaining a first feature corresponding to the historical fault type, obtaining any type of historical operation data corresponding to the first feature, and classifying each piece of historical operation data into a corresponding data class; Count the first times the same data type appears, and the total number of times the data type appears; Calculate the ratio of the first number to the total number of times and set it as the first probability of the corresponding data class for the first feature; The acquisition logic of the target data class includes: Sort the first probabilities in descending order, select the data class corresponding to the first probability with the largest value, and set the data class as the target data class corresponding to the first feature; By inputting the current first feature into the first mapping relationship, a corresponding fault type is obtained; by inputting the current first feature into the second mapping relationship, a corresponding target data class is obtained; Acquiring current operating data, classifying the current operating data to obtain a current data class, comparing the current data class with a target data class, counting a first number of instances in which the current data class is identical to the target data class, setting the number of types of the current operating data to a second number, calculating a second ratio of the first number to the second number, setting the second value as a second ratio threshold, and comparing the second ratio with the second value; When the second ratio is greater than or equal to the second value, the corresponding fault type is set as the current fault type; Otherwise, output no fault signal; The historical natural data include ambient temperature and ambient humidity; The logic for compensating the second mapping relationship includes: Set the first temperature and the first humidity as the standard temperature and standard humidity, obtain the first characteristic value of the historical fault type corresponding to the standard temperature and standard humidity and the average value of the historical operation data of any type, calculate the third difference and the fourth difference between any temperature and any humidity and the standard temperature and standard humidity, calculate the sum of the third difference and the fourth difference, set the sum of the third difference and the fourth difference as the independent variable, obtain the historical operation data corresponding to the any temperature and any humidity, calculate the average value of the corresponding historical operation data at any temperature and any humidity, calculate the difference between the average value of the historical operation data corresponding to the standard temperature and standard humidity and the average value of the corresponding historical operation data at any temperature and any humidity, record it as the target difference, set the sum of the third difference and the fourth difference as the independent variable, use the target difference as the dependent variable, construct an interference degree model, use the coefficient of the interference degree model as the compensation coefficient, calculate the difference between 1 and the compensation coefficient, record it as the proportional parameter, calculate the product of the corresponding historical operation data and the proportional parameter, record it as the first product, obtain the target data class corresponding to the first product, and update the data class in the second mapping relationship to the target data class corresponding to the first product; The calculation expression of the interference degree model is: ； Wherein, Y is the target difference, X is the sum of the third difference and the fourth difference, C is the compensation coefficient, the third difference is the difference between the temperature and the standard temperature, and the fourth difference is the difference between the humidity and the standard humidity.

2. A hydrogen fuel cell engine fault diagnosis system according to claim 1, characterized in that: The analysis module obtains historical fault types and their corresponding historical vibration intensities; The historical fault types include deionized glycol inlet low pressure fault, current set point deviation fault, deionized glycol inlet high temperature fault, hydrogen inlet and outlet low pressure difference fault, hydrogen supply pressure low fault, reverse stack current fault, hydrogen exhaust voltage low fault, air pressure low fault, hydrogen leakage fault, stack voltage low fault and deionized glycol outlet temperature voltage high fault; The historical operating data include inlet hydrogen pressure, inlet coolant temperature, total stack current, stack voltage, coolant flow, stack outlet hydrogen pressure, stack inlet hydrogen pressure, coolant inlet pressure, air inlet pressure, stack coolant outlet temperature, stack coolant inlet temperature, air compressor inlet temperature, air compressor outlet temperature, air compressor voltage and air compressor current.

3. A hydrogen fuel cell engine fault diagnosis system according to claim 1, characterized in that: The extraction logic of the first feature includes: Obtain any historical fault type and its corresponding historical vibration intensity, remove outliers and normalize the historical vibration intensity so that the values ​​of each historical vibration intensity are distributed between 0 and 1; Each historical vibration intensity is converted into frequency domain data through Fourier transform, the power spectrum density of the historical vibration intensity is calculated through autocorrelation analysis, and the power spectrum density is set as the first feature of the historical fault type.

4. A hydrogen fuel cell engine fault diagnosis system according to claim 1, characterized in that: The clustering logic of the data class includes: Obtaining any type of historical operating data corresponding to the historical fault type, calculating an average value of the historical operating data, setting a first value as a change gradient, calculating a first sum of the average value and the first value, calculating a first difference between the average value and the first value, setting the average value, the first sum, and the first difference as first cluster centers, respectively, assigning each piece of historical operating data to the first cluster center that is closest to it, where the closest distance is represented by the smallest absolute value of the difference between the historical operating data and the cluster center, to obtain three first classifications; Calculate the average value of each first classification, set the average value of each first classification as the second cluster center, and classify each historical running data into the second cluster center with the closest distance, to obtain three second classifications; Compare the data in the first classification with the data in the second classification. When the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification. Otherwise, repeat the above steps until the data in the first classification is the same as the data in the second classification, stop clustering and obtain each final classification. The final classification is set as the data class.

Images