Compressor cylinder fault monitoring method and hydrogen compressor

Abstract

The present application relates to the technical field of equipment health state monitoring, and provides a compressor cylinder fault monitoring method and a hydrogen compressor, wherein the method comprises: acquiring a compression data index of a target cylinder corresponding to at least one compressor working cycle; determining an input feature vector corresponding to the compression data index; inputting the input feature vector into each leakage prediction model in a plurality of preset leakage prediction models, to predict a cylinder state of the target cylinder; and determining whether the target cylinder has a cylinder fault according to the cylinder state of the target cylinder determined by each leakage prediction model. Thus, the technology of fusing a plurality of machine learning models is adopted to identify various cylinder state results of the target cylinder, thereby ensuring comprehensive coverage and high accuracy of the finally monitored cylinder fault.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment health state monitoring, and particularly relates to a compressor cylinder fault monitoring method and a hydrogen compressor. BACKGROUND

[0002] As an important clean energy to achieve the double carbon goal, with the popularization of hydrogen energy, the construction of hydrogen refueling stations is also gradually accelerating.

[0003] As the core equipment in the hydrogen refueling station, the operation stability of the hydrogen compressor directly affects the operation of the whole station, so the hydrogen refueling station has an urgent need for a health state monitoring system of the hydrogen compressor, especially an effective implementation scheme for the state monitoring or fault diagnosis of the cylinder of the hydrogen compressor.

[0004] At present, the industry has not yet proposed a better technical solution to the above problems. SUMMARY

[0005] The present application provides a compressor cylinder fault monitoring method and a hydrogen compressor to at least solve the defect that the existing technology cannot comprehensively and accurately monitor the state of the cylinder of the hydrogen compressor, which makes it difficult for the hydrogen compressor to ensure stable operation.

[0006] The present application provides a compressor cylinder fault monitoring method, which comprises: obtaining compression data indicators of a target cylinder corresponding to at least one compressor working cycle; the compression data indicators include: intake pressure parameters, compression ratio parameters, compression polytropic index, expansion polytropic index, expansion process volume difference and compression process volume difference; determining an input feature vector corresponding to the compression data indicators; inputting the input feature vector into each leakage prediction model in a plurality of preset leakage prediction models to predict the cylinder state of the target cylinder for each leakage prediction model; each leakage prediction model is used to predict the input feature vector based on a corresponding cylinder state set, and the cylinder state sets corresponding to each leakage prediction model are different; the cylinder state set contains at least one of the following preset cylinder states: normal cylinder state, suction valve leakage state, exhaust valve leakage state and piston ring leakage state; determining whether the target cylinder has a cylinder fault according to the cylinder state of the target cylinder determined by each leakage prediction model, wherein the cylinder fault includes a sealing element failure fault and a valve damage fault.

[0007] According to the present application, a compressor cylinder fault monitoring method is provided, each of the leakage prediction models adopts a binary classification support vector machine, a first leakage prediction model corresponds to a cylinder state set containing a normal cylinder state and an air inlet valve leakage state, a second leakage prediction model corresponds to a cylinder state set containing a normal cylinder state and an air outlet valve leakage state, a third leakage prediction model corresponds to a cylinder state set containing a normal cylinder state and a piston ring leakage state, a fourth leakage prediction model corresponds to a cylinder state set containing an air inlet valve leakage state and an air outlet valve leakage state, a fifth leakage prediction model corresponds to a cylinder state set containing an air outlet valve leakage state and a piston ring leakage state, and a sixth leakage prediction model corresponds to a cylinder state set containing an air inlet valve leakage state and a piston ring leakage state.

[0008] According to the present application, a compressor cylinder fault monitoring method is provided, the hyperparameters in each of the leakage prediction models are iteratively optimized by a particle swarm optimization algorithm, the hyperparameters include a penalty factor and a radial kernel function parameter.

[0009] According to the present application, a compressor cylinder fault monitoring method is provided, the determination of the input feature vector corresponding to the compression data index includes: normalizing each type of index parameter in the compression data index to determine the corresponding normalized feature vector; and performing principal component analysis dimension reduction processing on the normalized feature vector to determine the corresponding input feature vector.

[0010] According to the present application, a compressor cylinder fault monitoring method is provided, the determination of whether the target cylinder has a cylinder fault according to the cylinder state of the target cylinder determined by each of the leakage prediction models includes: determining the confidence degree corresponding to the cylinder state of each of the target cylinders; determining the statistical confidence degree corresponding to each of the preset cylinder states according to the confidence degree corresponding to the cylinder state of each of the target cylinders; determining the target preset cylinder state corresponding to the target cylinder according to the statistical confidence degree corresponding to each of the preset cylinder states; determining that the target cylinder does not have a cylinder fault in the case of determining that the target preset cylinder state is a normal cylinder state; determining that the target cylinder has a valve damage fault in the case of determining that the target preset cylinder state is an air inlet valve leakage state and / or an air outlet valve leakage state; and determining that the target cylinder has a sealing element failure fault in the case of determining that the target preset cylinder state is a piston ring leakage state.

[0011] According to the present application, a compressor cylinder fault monitoring method is provided, the compressor comprises a plurality of cylinders, each of the cylinders is connected to a leakage flow meter through a corresponding electromagnetic valve, wherein, after determining whether the target cylinder has a cylinder fault according to the cylinder state of the target cylinder determined by each of the leakage prediction models, the method further comprises: in the case of determining that the target cylinder does not have a cylinder fault, controlling to close the target electromagnetic valve to turn on the leakage flow meter to the piston ring of the target cylinder to collect the seal running state of the piston ring in the compressor working period; inputting the seal running state into a preset life prediction model to predict the remaining life of the corresponding seal; the remaining life of the seal represents the remaining time during which the seal of the target cylinder can work normally; generating a replacement reminder notification based on the remaining life of the seal, the replacement reminder notification defines a message notification reminding a user to replace the seal.

[0012] According to the present application, a compressor cylinder fault monitoring method is provided, each of the electromagnetic valves is sequentially turned on according to a preset period, and the preset period is greater than or equal to the compressor working period.

[0013] According to the present application, a compressor cylinder fault monitoring method is provided, the target cylinder corresponding at least one compressor working period is obtained by acquiring a compression data index, comprising: for each compressor working period, collecting the compression working parameters of the target cylinder in the compressor working period, the compression working parameters include cylinder pressure parameters and cylinder volume parameters; determining the pressure-volume relationship of the target cylinder corresponding to the compressor working period according to the compression working parameters of the target cylinder; the pressure-volume relationship represents the relationship between the cylinder pressure parameters and the cylinder volume parameters; determining the compression data index according to the pressure-volume relationship.

[0014] According to the present application, a compressor cylinder fault monitoring method is provided, the compressor is a liquid-driven piston compressor, wherein, the compression working parameters of the target cylinder in the compressor working period are collected by: based on a preset pressure sensor group, collecting the gas pressure parameters of the target cylinder corresponding to the compressor working period; the pressure sensor group comprises a first pressure sensor and a second pressure sensor symmetrically arranged on both sides of the screw of the target cylinder; and based on a preset hydraulic flow meter, detecting the volume of hydraulic oil acting on the target cylinder, and determining the cylinder volume parameters of the target cylinder corresponding to the compressor working period according to the volume of the hydraulic oil.

[0015] The application further provides a hydrogen compressor comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the compressor cylinder fault monitoring method according to any one of the above when executing the program.

[0016] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps of the compressor cylinder fault monitoring method according to any one of the above when executed by a processor.

[0017] The compressor cylinder fault monitoring method and the hydrogen compressor provided by the application can obtain the compression data index of at least one compressor working cycle corresponding to the target cylinder, determine the input feature vector corresponding to the compression data index, predict the cylinder state corresponding to the input feature vector by using multiple leakage prediction models, and comprehensively analyze multiple preset cylinder states of the target cylinder, so as to identify the cylinder fault of the target cylinder. Thus, the technology of fusing multiple machine learning models is used to identify the cylinder state results of the target cylinder, and the comprehensive coverage and high accuracy of the finally monitored cylinder fault are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0019] Figure 1 A flow chart of an example of the compressor cylinder fault monitoring method according to the application is shown;

[0020] Figure 2 A flow chart of another example of the compressor cylinder fault monitoring method according to an embodiment of the application is shown;

[0021] Figure 3 A structural schematic diagram of an example of the hydrogen compressor suitable for applying the compressor cylinder fault monitoring method according to an embodiment of the application is shown;

[0022] Figure 4 A real-time PV diagram of the target cylinder corresponding to the compressor working cycle according to an embodiment of the application is shown;

[0023] Figure 5 An operation flow chart of an example of processing the input feature vector by using multiple leakage prediction models according to an embodiment of the application is shown;

[0024] Figure 6 A flowchart illustrating another example of a compressor cylinder fault monitoring method according to an embodiment of the present invention is shown;

[0025] Figure 7 This is a schematic diagram of the structure of the hydrogen compressor provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] Figure 1 A flowchart illustrating an example of a compressor cylinder fault monitoring method provided according to the present invention is shown.

[0028] Regarding the execution subject of the method in the embodiments of the present invention, it can be any controller or processor with computing or processing capabilities to achieve the goal of monitoring the cylinders of the compressor. In some examples, it can be integrated into the compressor through software, hardware, or a combination of software and hardware.

[0029] like Figure 1 As shown, in step S110, compression data indicators for at least one compressor working cycle corresponding to the target cylinder are obtained. The compression data indicators include: intake pressure parameters, compression ratio parameters, compression polytropic index, expansion polytropic index, expansion process volume difference, and compression process volume difference.

[0030] It should be understood that compressor duty cycle is a technical term used in this field to refer to the various stages experienced during the complete duty cycle of a compressor, namely the intake stage, compression stage, exhaust stage, and expansion stage. Furthermore, the compressor can be a single-cylinder compressor or a multi-cylinder compressor, where the target cylinder is not specifically designated and can refer to any cylinder in a multi-cylinder compressor that is subject to fault monitoring.

[0031] In one example of this invention, a potential or known index analysis module is installed in the compressor, which can intelligently obtain compression data indicators for each stage of the corresponding working cycle. Furthermore, the compression data indicators are acquired through communication and interaction with the index analysis module. In another example of this invention, various sensors can be installed in the target cylinder to detect working parameters at each stage of the compressor's working cycle, and the corresponding compression data indicators are obtained through calculation.

[0032] In step S120, the input feature vector corresponding to the compressed data index is determined.

[0033] Here, the compressed data index can be converted by various feature processing methods to obtain the corresponding input feature vector. In some embodiments, the feature dimension corresponding to the input feature vector is matched with the parameter type of the index parameter in the compressed data index, that is, a single input feature vector is composed of 6 features.

[0034] It should be noted that in a compressor working cycle, there is mutual influence between different index parameters in the compressed data index, for example, the intake pressure parameter and the compression ratio parameter. By using various decoupling feature processing algorithms, the compressed data index is decoupled and reduced in dimension to obtain the input feature vector, which can better guarantee the classification effect of the model. More details will be described below in combination with other examples.

[0035] In step S130, for each of the plurality of leakage prediction models, the input feature vector is input into the leakage prediction model to predict the cylinder state of the target cylinder. Here, each leakage prediction model can be a machine learning model, and the training sample set of each leakage prediction model can include compressed data index samples and corresponding cylinder state labels.

[0036] Specifically, each leakage prediction model is used to predict the corresponding input feature vector based on the corresponding cylinder state set, and the cylinder state sets corresponding to each leakage prediction model are different. The cylinder state set includes at least one of the following preset cylinder states: normal cylinder state, suction valve leakage state, exhaust valve leakage state, and piston ring leakage state. It should be understood that the cylinder state set can also include other types of cylinder states not described herein, and all belong to the implementation scope of the embodiments of the present application.

[0037] In some embodiments, each leakage prediction model respectively predicts different preset cylinder states based on the input feature vector. In one example, the leakage prediction model r is used to detect whether there is a normal cylinder state, the leakage prediction model p is used to detect whether there is a suction valve leakage state, the leakage prediction model m is used to detect whether there is an exhaust valve leakage state, and the leakage prediction model n is used to detect whether there is a piston ring leakage state. In another example, the leakage prediction model can classify and identify multiple preset cylinder states, such as normal cylinder state, suction valve leakage state, and piston ring leakage state.

[0038] In step S140, according to the cylinder state of the target cylinder determined by each leakage prediction model, it is determined whether the target cylinder has a cylinder fault, and the cylinder fault includes a sealing element failure fault and a valve damage fault.

[0039] It should be understood that the cylinder failure can also include other types of failures, such as a high-pressure pressure-bearing component failure or a hydraulic oil leakage failure, etc.

[0040] In some embodiments, when the cylinder state of each target cylinder indicates that the target cylinder does not have a leakage risk, it is determined that the target cylinder does not have a cylinder failure. In addition, when the cylinder state of the target cylinder determined by the leakage prediction model u is the suction valve leakage state or the exhaust valve leakage state, it is determined that the target cylinder has a valve damage failure, and further, when the cylinder state of the target cylinder determined by the leakage prediction model v is the piston ring leakage state, it can be determined that the target cylinder also has a sealing component failure.

[0041] Through the embodiments of the present application, for each type of preset cylinder state related to the cylinder failure, a plurality of leakage prediction models are used for identification respectively, and the prediction results of each state are comprehensively analyzed for the compressor cylinder failure monitoring, so as to realize the all-around monitoring of the cylinder failure.

[0042] Figure 2 A flow chart of another example of the compressor cylinder failure monitoring method according to the embodiments of the present application is shown.

[0043] As shown in Figure 2 , in step S210, the compression data indicators of the target cylinder corresponding to at least one compressor working cycle are obtained.

[0044] In some embodiments, the working parameters of the target cylinder in the compressor working cycle are collected, and the corresponding compression data indicators are obtained through calculation and analysis. Figure 3 A structural schematic diagram of an example of a hydrogen compressor suitable for applying the compressor cylinder failure monitoring method according to the embodiments of the present application is shown. As shown in Figure 3 , the hydrogen compressor includes a primary compressor 351 and a secondary compressor 353, and each stage of compressor includes two cylinders for reciprocating compression movement of a screw rod.

[0045] Specifically, based on a preset pressure sensor group, the gas pressure parameters of the target cylinder corresponding to each compressor working cycle are collected, the pressure sensor group includes a first pressure sensor and a second pressure sensor symmetrically arranged on both sides of the screw rod of the target cylinder, for example Figure 3 , the pressure variable chip 360 in the figure, so that the pressure in the cylinder can be detected without opening a pressure tapping hole. Further, based on a preset hydraulic flow meter, the volume of the hydraulic oil acting on the target cylinder is detected, for example Figure 3The hydraulic flow meter 310 or 330 in the hydraulic oil circuit 300 is installed to determine the cylinder volume parameter of the target cylinder corresponding to the compressor working cycle according to the volume of the hydraulic oil. Thus, during the operation of the compressor system, the pressure in the cylinder can be transmitted to the screw rod as the cylinder barrel and the end cover of the cylinder are pressed on the oil cylinder by the screw rod, and the dynamic pressure in the cylinder can be accurately calculated by monitoring the stress change of the screw rod through the strain gauges symmetrically arranged on the screw rod, for example, by monitoring the average value of the strain gauge outputs on the two screw rods on the same side of the cylinder to calculate the dynamic pressure of the cylinder. In addition, since the dynamic stroke of the cylinder and the oil cylinder is always consistent, the hydraulic flow meter is installed on the oil side circuit to monitor the flow and pressure of the hydraulic oil under each stroke in real time, calculate the dynamic volume V of the cylinder, obtain the real-time PV diagram of the compressor operation, and complete the overall health status diagnosis of the cylinder of the compressor.

[0046] In step S220, each type of index parameter in the compression data index is normalized and feature processed to determine the corresponding normalized feature vector.

[0047] Figure 4 The real-time PV diagram of the target cylinder corresponding to the compressor working cycle according to an embodiment of the application is shown. As shown, it shows each stage in the compressor working cycle, including the exhaust process, the compression process, the suction process and the expansion process. Figure 4

[0048] The calculation details of each compression data index are as follows:

[0049] Since the change of the load pressure will affect the shape of the P-V diagram, the intake pressure and the compression ratio are selected as the features.

[0050] The calculation method of the intake pressure feature x1 is as follows:

[0051] x1=P s Formula (1)

[0052] Wherein, P s represents the intake pressure of the target cylinder.

[0053] The calculation method of the compression ratio feature x2 is as follows:

[0054]

[0055] Wherein, P d represents the exhaust pressure of the target cylinder.

[0056] ​In addition, only when all the valves are logically closed, can the presence of a leak in the valve be detected. For example, if a leak occurs in the suction valve during the expansion stage, the gas in the gas source bottle will leak through the suction valve into the cylinder with lower pressure, which will cause the shape of the expansion curve to change. Therefore, the compression stage and the expansion stage of the P-V diagram are selected as the process variable index. Specifically, the average polytropic index of the thermodynamic process can be calculated by the area ratio of the curve to the coordinate axis.

[0057] The average compression polytropic index x3 is calculated as follows:

[0058]

[0059] The average expansion polytropic index x4 is calculated as follows:

[0060]

[0061] The expansion process volume difference x5 is calculated as follows:

[0062] x5 = V b -V a Equation (5)

[0063] The compression process volume difference x6 is calculated as follows:

[0064] x6 = V c -V d Equation (6)

[0065] wherein a, b, c and d are different time points in the compressor working cycle. Specifically, a is the exhaust valve closing point, b is the intake valve opening point, c is the intake valve closing point, and d is the exhaust valve opening point.

[0066] Further, for each compressor working cycle, the corresponding normalized feature vector x = (x1, x2, x3, x4, x5, x6) is obtained based on the above six features. Thus, the compression data index features are extracted based on the PV diagram, so that better compression data indexes can be extracted under the complex working conditions of the hydrogen compressor in the hydrogen filling station with large fluctuations in the inlet and outlet pressure range.

[0067] In step S230, the normalized feature vector is subjected to principal component analysis dimension reduction processing to determine the corresponding input feature vector.

[0068] It should be noted that the 6 features in the normalized feature vector can be mutually influenced. In order to make the features relatively independent and make the SVM classification effect better, the PCA (principal component analysis) algorithm is used to convert the 6-dimensional feature into a k-dimensional (k < 6) vector. Specifically, first, for the input m normalized feature vectors, each normalized feature vector has n-dimensional features (i.e., 6-dimensional features), an n*m matrix is formed, and each row vector is normalized to obtain a data matrix X, wherein the rows of the matrix represent a certain feature (such as the first row representing the pressure feature of all data), and the columns represent a data (the first column represents the 6 features of the first sample). The n row vectors are obtained by dividing the matrix X by rows, and the covariance matrix Σ is calculated by combining two by two. The eigenvectors and eigenvalues of the covariance matrix Σ are obtained by using SVD matrix decomposition, and the eigenvalues are arranged in descending order, and the corresponding eigenvectors are arranged in descending order as a whole. Further, the eigenvectors of the first k largest eigenvalues are selected to form a projection matrix W, and the projection matrix W is multiplied by the original matrix X to obtain a reduced matrix Y = WX, the original data matrix X is converted into a k*m matrix Y, and the feature dimension reduction decoupling goal is achieved.

[0069] In step S240, for each of the plurality of leakage prediction models, the input feature vector is input into the leakage prediction model to predict the cylinder state of the target cylinder.

[0070] Specifically, as Figure 5 An operation flowchart showing an example of processing an input feature vector using a plurality of leakage prediction models according to an embodiment of the present application is shown, each of which uses a binary classification support vector machine, the first leakage prediction model 521 corresponds to a cylinder state set containing a normal cylinder state a and an intake valve leakage state b, the second leakage prediction model 522 corresponds to a cylinder state set containing a normal cylinder state a and an exhaust valve leakage state c, the third leakage prediction model 523 corresponds to a cylinder state set containing a normal cylinder state a and a piston ring leakage state d, the fourth leakage prediction model 524 corresponds to a cylinder state set containing an intake valve leakage state b and an exhaust valve leakage state c, the fifth leakage prediction model 525 corresponds to a cylinder state set containing an exhaust valve leakage state c and a piston ring leakage state d, and the sixth leakage prediction model 526 corresponds to a cylinder state set containing an intake valve leakage state b and a piston ring leakage state d.

[0071] In one aspect, in some embodiments, the cylinder states predicted by the individual binary classification support vector machines are counted by voting to obtain a final state identification result for the target cylinder. For example, the above-mentioned six binary classification support vector machines (i.e., 521-526) give one classification result (such as: state categories b, a, d, b, c, b, respectively), and the six results are counted, and the final state identification result for the target cylinder is the suction valve leakage state b, as the number of votes for b is the largest.

[0072] On the other hand, in step S240, the confidence corresponding to the cylinder state of each target cylinder is determined.

[0073] For example, each binary classification support vector machine can output a classification result, and for each classification result, the confidence determined in the model prediction process is output, for example, the binary classification support vector machine 521 obtains the confidence for the cylinder states a and b as 0.2 and 0.8, respectively.

[0074] In step S250, for each preset cylinder state, the statistical confidence corresponding to the preset cylinder state is determined according to the confidence corresponding to the cylinder state of each target cylinder.

[0075] Specifically, according to the four preset cylinder states, the cylinder states of the target cylinder determined by the six binary classification support vector machines (i.e., 521-526) are classified to obtain the statistical confidence under the corresponding preset cylinder state. For example, the confidence for the output cylinder state b obtained by the binary classification support vector machines 521, 524, and 526 is statistically analyzed and calculated to obtain the statistical confidence for the suction valve leakage state.

[0076] In step S260, the target preset cylinder state corresponding to the target cylinder is determined according to the statistical confidence corresponding to each preset cylinder state.

[0077] In one example of the embodiment of the present application, the statistical confidence corresponding to the cylinder normal state, the suction valve leakage state, the exhaust valve leakage state, and the piston ring leakage state is compared, and the preset cylinder state corresponding to the maximum statistical confidence is determined as the target preset cylinder state, thereby outputting a unique preset cylinder state to facilitate rapid positioning of the corresponding fault type. In another example of the embodiment of the present application, the statistical confidence corresponding to each preset cylinder state is compared with the corresponding confidence threshold, respectively, and the preset cylinder state exceeding the confidence threshold is determined as the target preset cylinder state, at this time, multiple different preset cylinder states can be output to facilitate comprehensive evaluation of the cylinder fault.

[0078] Through the embodiments of the present application, the cylinder states predicted by each binary classification support vector machine and the corresponding confidence are comprehensively considered, compared with the state result voting statistics, and the high accuracy and comprehensive coverage of the finally determined target preset cylinder state can be ensured.

[0079] In step S270, whether the target cylinder has one or more cylinder faults is determined according to the target preset cylinder state.

[0080] Specifically, in the case where the target preset cylinder state is determined to be the cylinder normal state, it is determined that the target cylinder does not have a cylinder fault, in the case where the target preset cylinder state is determined to be the suction valve leakage state and / or the exhaust valve leakage state, it is determined that the target cylinder has a valve damage fault, and in the case where the target preset cylinder state is determined to be the piston ring leakage state, it is determined that the target cylinder has a sealing element failure fault.

[0081] It should be noted that the training sample set of each leakage prediction model includes the compression data index sample and the cylinder state label of the corresponding cylinder state set, for example, the cylinder state label of the first leakage prediction model 521 is the cylinder normal state a or the suction valve leakage state b, and the cylinder state label of the second leakage prediction model 522 is the cylinder normal state a or the exhaust valve leakage state c. Through the above training sample set, the setting of the model parameters (i.e., the parameters of the model itself) is completed.

[0082] In addition, the setting of the hyperparameters of each leakage prediction model can use an algorithm to find the optimal. The fault diagnosis problem of the compressor system is a linearly inseparable multi-classification problem, and the penalty factor C and the radial kernel function parameter γ of the SVM have a great influence on the model effect. In some embodiments, the grid search method is used to find the optimal hyperparameters, but the grid search range is difficult to determine and the search efficiency is slow.

[0083] Further, in some examples of the embodiments of the present application, the hyperparameters in each leakage prediction model are iteratively optimized by a particle swarm optimization algorithm. It should be understood that the particle swarm optimization algorithm is only used in the training stage of the model, and after the optimal hyperparameters are determined and the hyperparameter setting of each leakage prediction model is completed, the particle swarm optimization algorithm does not need to be used again in the prediction stage of the model.

[0084] It should be noted that the particle swarm algorithm is a kind of random global optimization technology, which finds the optimal region in a complex search space through the interaction between particles. Using the particle swarm algorithm to optimize the structural parameters of the support vector machine can quickly converge to find the optimal solution and improve the learning ability of the SVM.

[0085] Specifically, the operation of iterative optimization by the particle swarm optimization algorithm is as follows: first, set the value range of the penalty factor C and the radial kernel function parameter g to be optimized, and the initialization parameters such as the local search ability of the parameter, the maximum number of population, the maximum number of iterations, etc. Then, a group of particles are randomly initialized in the value range of the penalty factor C and the radial kernel function parameter g. The position information and speed information of each particle are initialized, and then the historical optimal position Pbesti of each particle is set as the current position, and the current position of the best particle in the population is set as Gbest. Further, the fitness value of each particle is calculated. Then, the historical optimal value Pbesti and the global optimal fitness value Gbest of each particle are updated, and the speed and position of each particle are updated. Further, it is judged whether the maximum number of iterations is reached, and if yes, the operation is stopped, otherwise it is returned to the calculation of the fitness value of each particle to continue iteration.

[0086] Figure 6 A flow chart of another example of the compressor cylinder fault monitoring method according to an embodiment of the application is shown.

[0087] As Figure 6 shown, in step S610, according to the cylinder state of the target cylinder determined by each leakage prediction model, it is determined whether the target cylinder has a cylinder fault.

[0088] For more details of step S610, please refer to the description of the operation in the flowchart shown in the above Figure 1 or Figure 2 herein.

[0089] It should be noted that during the operation of the hydrogen compressor, the sealing element (or sealing ring) of the cylinder piston ring of the compressor is more likely to be damaged than other components (such as valves, screws, etc.), so the health status of the sealing element needs to be paid special attention to avoid piston ring leakage failure of the compressor.

[0090] In step S620, in the case where it is determined that the target cylinder does not have a cylinder fault, from a plurality of preset electromagnetic valves, the target electromagnetic valve is controlled to close the piston ring of the target cylinder to the leakage flowmeter to collect the running state of the sealing element of the piston ring during the operation cycle of the compressor.

[0091] Here, the seal running state can represent the overall deformation state of the seal in one or more of the compressor working cycles. It should be noted that when the target cylinder does not have a piston ring leakage fault, the working state of the seal in the compressor working cycle should be: the inner and outer lips are tightly attached to the inner and outer walls and move backward slightly. The root of the seal is pressed backward, and as the pressure increases, the hydraulic oil also increases the force on the seal, and the overall deformation of the seal also increases. When the working cycle is completed and the oil cylinder has no pressure, the seal returns to its original state. Therefore, the deformation amount of the seal at each stage of the compressor working cycle can be counted to obtain the corresponding seal running state.

[0092] In some embodiments, the compressor includes a plurality of cylinders, and each cylinder is connected to a leakage flow meter through a corresponding electromagnetic valve. Specifically, the leakage flow meter 320 is connected to the piston ring of each cylinder through a corresponding electromagnetic valve, i.e., a leakage detection circuit is installed on the gas side isolation chamber of the compressor, and the leakage flow meter 320 is connected to the piston ring of each cylinder through each circuit via electromagnetic valves V01, V02, V03, and V04.

[0093] In some examples of the embodiments of the present application, each electromagnetic valve is sequentially turned on according to a preset period, and the preset period is greater than or equal to the working cycle of the compressor. For example, the electromagnetic valves V01, V02, V03, and V04 are sequentially turned on at a time interval of two minutes, so that only a single leakage flow meter 320 is needed, and real-time and comprehensive monitoring of the seal state of the piston ring of multiple cylinders is achieved through switching of the electromagnetic valves, without the need to deploy a corresponding leakage flow meter for each cylinder, thereby saving the equipment deployment cost of the cylinder fault monitoring system.

[0094] In step S630, the seal running state is input into a preset life prediction model to predict the remaining life of the corresponding seal, and the remaining life of the seal represents the remaining time during which the seal ring of the target cylinder can work normally.

[0095] Here, the life prediction model can use various non-limiting machine learning models, such as deep neural networks, etc. In some embodiments, each training sample in the training sample set of the life prediction model includes a seal working state sample and a corresponding seal remaining life label, for example, data from various life tests of the seal ring is sorted to obtain training samples for the life prediction model, so as to achieve accurate prediction of the life of the seal.

[0096] In step S640, a replacement reminder notification is generated based on the remaining life of the seal, and the replacement reminder notification defines a message notification reminding the user to replace the seal.

[0097] In some embodiments, when it is detected that the remaining life of the seal is approaching the end of its life, for example, one week or one month in advance, a replacement reminder notification is sent to the terminal, for example, a replacement reminder notification is sent to the management terminal of the hydrogenation station, so as to realize early notification of the maintenance time for the seal ring by predicting the remaining life of the seal, and to reduce the probability of seal failure of the compressor to the maximum extent.

[0098] In some examples of the embodiments of the present application, some other types of fault detection modules are also deployed for the compressor to realize secondary confirmation or verification of the faults of the cylinder of the compressor. Referring to the examples in Figure 3 The acoustic emission sensor 340 is installed for the valve position, and by comparing the real-time expansion (compression) process with the average amplitude of the acoustic emission signal curve in the expansion (compression) process under the fault-free state, it can be quickly identified whether the cylinder has a valve damage fault.

[0099] In addition, as shown in Figure 3 The hydraulic flow meters 310 and 330 are installed on the hydraulic oil inlet pipeline, which can measure the instantaneous flow on one side of the cylinder during the operation of the compressor with high precision, and then calculate the hydraulic oil volume under single stroke by integrating calculation combined with the specific structure parameters of the compressor. Further, it is compared with the normal oil volume, and when it is higher than the normal oil volume, it is determined that the compressor has a hydraulic oil leakage fault.

[0100] Through the embodiments of the present application, the operating state of the compressor is monitored in real time, comprehensive and overall state monitoring and fault diagnosis for the cylinder of the compressor are realized, the safety of the compressor system is improved, and predictive maintenance can be realized, the unexpected failure of the hydrogenation station is reduced, the operation and maintenance cost of the whole station is effectively reduced, and the safety and reliability of the system are improved, and the hydrogenation station can be orderly and controllably operated.

[0101] Figure 7 An example of a schematic diagram of the physical structure of a hydrogen compressor is shown in Figure 7As shown, the hydrogen compressor can include a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communications bus 740. The processor 710 can invoke the logic instructions in the memory 730 to execute the compressor cylinder fault monitoring method, which includes: obtaining a compression data index of a target cylinder corresponding to at least one compressor working cycle; the compression data index includes: an intake pressure parameter, a compression ratio parameter, a compression polytropic index, an expansion polytropic index, an expansion process volume difference, and a compression process volume difference; determining an input feature vector corresponding to the compression data index; inputting the input feature vector into each of a plurality of preset leakage prediction models to predict the cylinder state of the target cylinder for each of the plurality of preset leakage prediction models; each of the leakage prediction models is used to predict the input feature vector based on a corresponding cylinder state set, and the cylinder state sets corresponding to each of the leakage prediction models are different; the cylinder state set contains at least one of the following preset cylinder states: a normal cylinder state, an air suction valve leakage state, an exhaust valve leakage state, and a piston ring leakage state; determining whether the target cylinder has a cylinder fault according to the cylinder state of the target cylinder determined by each of the leakage prediction models, wherein the cylinder fault includes a sealing element failure fault and a valve damage fault.

[0102] In addition, the logic instructions in the memory 730 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0103] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, enable the computer to perform the compressor cylinder fault monitoring method provided by any of the above methods, the method comprising: obtaining a compression data indicator of a target cylinder corresponding to at least one compressor working cycle; the compression data indicator comprising: an intake pressure parameter, a compression ratio parameter, a compression polytropic index, an expansion polytropic index, an expansion process volume difference, and a compression process volume difference; determining an input feature vector corresponding to the compression data indicator; inputting the input feature vector into each of a plurality of preset leakage prediction models to predict a cylinder state of the target cylinder; each of the leakage prediction models is configured to predict the input feature vector based on a corresponding set of cylinder states, and the set of cylinder states corresponding to each of the leakage prediction models is different; the set of cylinder states comprises at least one of the following preset cylinder states: a normal cylinder state, an air suction valve leakage state, an exhaust valve leakage state, and a piston ring leakage state; and determining whether the target cylinder has a cylinder fault based on the cylinder state of the target cylinder determined by each of the leakage prediction models, wherein the cylinder fault comprises a seal failure fault and a valve damage fault.

[0104] In yet another aspect, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a compressor cylinder fault monitoring method provided by any of the above methods, the method comprising: obtaining a compression data indicator of a target cylinder corresponding to at least one compressor working cycle; the compression data indicator comprising: an intake pressure parameter, a compression ratio parameter, a compression polytropic index, an expansion polytropic index, an expansion process volume difference, and a compression process volume difference; determining an input feature vector corresponding to the compression data indicator; inputting the input feature vector into each of a plurality of preset leakage prediction models to predict a cylinder state of the target cylinder; each of the leakage prediction models is configured to predict the input feature vector based on a corresponding set of cylinder states, and the set of cylinder states corresponding to each of the leakage prediction models is different; the set of cylinder states comprises at least one of the following preset cylinder states: a normal cylinder state, an air suction valve leakage state, an exhaust valve leakage state, and a piston ring leakage state; and determining whether the target cylinder has a cylinder fault based on the cylinder state of the target cylinder determined by each of the leakage prediction models, wherein the cylinder fault comprises a seal failure fault and a valve damage fault.

[0105] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0106] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0107] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring compressor cylinder faults, characterized in that, The method includes: Obtain compression data indicators for at least one compressor working cycle corresponding to the target cylinder; the compression data indicators include: intake pressure parameters, compression ratio parameters, compression polytropic index, expansion polytropic index, expansion process volume difference, and compression process volume difference; Determine the input feature vector corresponding to the compressed data index; For each of the preset multiple leakage prediction models, the input feature vector is input to the leakage prediction model to predict the cylinder state of the target cylinder; each leakage prediction model is used to predict the corresponding input feature vector based on the corresponding cylinder state set, and the cylinder state sets corresponding to each leakage prediction model are different; the cylinder state set includes at least one of the following preset cylinder states: normal cylinder state, intake valve leakage state, exhaust valve leakage state, and piston ring leakage state; Based on the cylinder state of the target cylinder determined by each of the leakage prediction models, it is determined whether the target cylinder has a cylinder fault, which includes seal failure fault and valve damage fault. Each of the aforementioned leakage prediction models employs a binary classification support vector machine. The cylinder state set corresponding to the first leakage prediction model includes the normal cylinder state and the intake valve leakage state; the cylinder state set corresponding to the second leakage prediction model includes the normal cylinder state and the exhaust valve leakage state; the cylinder state set corresponding to the third leakage prediction model includes the normal cylinder state and the piston ring leakage state; the cylinder state set corresponding to the fourth leakage prediction model includes the intake valve leakage state and the exhaust valve leakage state; the cylinder state set corresponding to the fifth leakage prediction model includes the exhaust valve leakage state and the piston ring leakage state; and the cylinder state set corresponding to the sixth leakage prediction model includes the intake valve leakage state and the piston ring leakage state.

2. The compressor cylinder fault monitoring method according to claim 1, characterized in that, The hyperparameters in each leakage prediction model are iteratively optimized using a particle swarm optimization algorithm, and the hyperparameters include a penalty factor and a radial kernel function parameter.

3. The compressor cylinder fault monitoring method according to claim 1, characterized in that, Determining the input feature vector corresponding to the compressed data index includes: The various index parameters in the compressed data index are normalized to determine the corresponding normalized feature vector. The normalized feature vectors are subjected to principal component analysis for dimensionality reduction to determine the corresponding input feature vectors.

4. The compressor cylinder fault monitoring method according to any one of claims 2-3, characterized in that, The step of determining whether the target cylinder has a cylinder fault based on the cylinder state of the target cylinder determined by each of the leakage prediction models includes: Determine the confidence level corresponding to the cylinder state of each of the target cylinders; For each of the preset cylinder states, the statistical confidence level corresponding to the preset cylinder state is determined based on the confidence level corresponding to the cylinder state of each target cylinder. Based on the statistical confidence levels corresponding to the various preset cylinder states, the target preset cylinder state corresponding to the target cylinder is determined; If the target preset cylinder state is determined to be a normal cylinder state, it is determined that the target cylinder does not have a cylinder fault; If the target preset cylinder is determined to be in an intake valve leakage state and / or an exhaust valve leakage state, it is determined that the target cylinder has a valve damage fault; and If the target preset cylinder is determined to be in a piston ring leakage state, it is determined that the target cylinder has a seal failure fault.

5. The compressor cylinder fault monitoring method according to claim 1, characterized in that, The compressor comprises multiple cylinders, each of which is connected to a leakage flow meter via a corresponding solenoid valve. The method further includes, after determining whether the target cylinder has a cylinder fault based on the cylinder state determined by each of the leakage prediction models, the method further includes: If it is determined that there is no cylinder fault in the target cylinder, the target solenoid valve is closed from among a plurality of preset solenoid valves to conduct the leakage flow meter to the piston ring of the target cylinder, so as to collect the sealing status of the piston ring during the working cycle of the compressor. The operating status of the seal is input into a preset life prediction model to predict the remaining life of the corresponding seal; the remaining life of the seal represents the remaining time that the seal of the target cylinder can work normally. A replacement reminder notification is generated based on the remaining lifespan of the seal. The replacement reminder notification defines a message notification to remind the user to replace the seal.

6. The compressor cylinder fault monitoring method according to claim 5, characterized in that, Each of the solenoid valves is sequentially activated according to a preset cycle, which is greater than or equal to the compressor's operating cycle.

7. The compressor cylinder fault monitoring method according to claim 1, characterized in that, The acquisition of compression data indicators for at least one compressor working cycle corresponding to the target cylinder includes: For each compressor working cycle, the compression working parameters of the target cylinder are collected within the compressor working cycle. The compression working parameters include cylinder pressure parameters and cylinder volume parameters. Based on the compression operating parameters of the target cylinder, the pressure-volume relationship of the target cylinder corresponding to the working cycle of the compressor is determined; the pressure-volume relationship represents the relationship between the cylinder pressure parameters and the cylinder volume parameters. The compression data indicators are determined based on the pressure-volume relationship.

8. The compressor cylinder fault monitoring method according to claim 7, characterized in that, The compressor is a hydraulically driven piston compressor, wherein the collection of compression operating parameters of the target cylinder during the compressor's working cycle includes: Based on a preset pressure sensor group, gas pressure parameters corresponding to the compressor's working cycle are collected for the target cylinder; the pressure sensor group includes a first pressure sensor and a second pressure sensor symmetrically arranged on both sides of the screw of the target cylinder; and The volume of hydraulic fluid acting on the target cylinder is detected by a preset hydraulic flow meter, and the cylinder volume parameter of the target cylinder corresponding to the working cycle of the compressor is determined based on the volume of the hydraulic fluid.

9. A hydrogen compressor, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the compressor cylinder fault monitoring method as described in any one of claims 1-8.

Images