A gas cylinder management method and system

Abstract

The application discloses a kind of gas cylinder management method and system, including, from each material gas cylinder Time series data, material characteristics and maintenance records, based on Time series data, material characteristics and maintenance records, using PC algorithm, establish gas cylinder life prediction task, according to the established gas cylinder life prediction task, screen out the related features that affect gas cylinder life;The related features are sent into graph convolutional neural network model for fusion, according to the number of features after fusion Knowledge transfer, obtain the life prediction result of each material gas cylinder;The effect of knowledge transfer is evaluated and optimized, so as to improve the accuracy of each material gas cylinder life prediction, realize the intelligentization of gas cylinder management;The application realizes the deep knowledge transfer from the data-rich gas cylinder to the rare material gas cylinder by adopting the meta-learning model and the deep residual network, and provides a basis for intelligent gas cylinder management by outputting the gas cylinder life prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas cylinder management, and in particular to a gas cylinder management method and system. BACKGROUND

[0002] As a key field of industrial safety and environmental protection, the technology of gas cylinder management has undergone significant evolution from traditional manual recording to modern intelligent management. Early gas cylinder management mainly relied on manual inspection and paper maintenance records, assessing the safety status of gas cylinders through regular checks of their appearance, pressure, and service life. With the acceleration of industrialization and the rise of Internet of Things (IoT) technology, sensor-based data collection and remote monitoring systems have gradually become the mainstream trend in gas cylinder management. For example, by installing pressure sensors and temperature sensors on gas cylinders, real-time monitoring of time series data can be achieved, and cloud computing platforms can be used for data storage and preliminary analysis, thereby improving the automation level of management. In addition, statistical models and empirical formulas are widely used in gas cylinder life prediction, attempting to infer the aging law of gas cylinders through historical data.

[0003] Although the existing technology has made great progress in the automation and dataization of gas cylinder management, its shortcomings are still significant, especially in the accuracy and adaptability of life prediction. First, traditional life prediction models are mostly based on statistical regression or fixed empirical formulas, ignoring the complex interactions between gas cylinder material properties (such as metal fatigue, corrosion resistance) and usage environment (such as temperature, humidity, maintenance frequency), resulting in limited generalization ability of the prediction results and difficulty in adapting to the decay law of gas cylinders of different materials. Second, existing methods usually only use feature concatenation or dimensionality reduction when dealing with multi-source heterogeneous data, failing to effectively capture the deep associations between data, limiting the expressive power of the prediction model. In addition, for rare material gas cylinders with small data volume, traditional transfer learning methods mostly stay at the level of shallow feature transfer, lacking a deep knowledge transfer mechanism across domains, resulting in insufficient robustness and accuracy of the model in small sample scenarios. These shortcomings jointly restrict the further development of gas cylinder management towards intelligence and high precision, therefore, there is an urgent need for a method that can break through the limitations of existing technology and promote the overall optimization of gas cylinder life prediction and management. SUMMARY

[0004] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0005] In view of the above existing problems, the present application is proposed. Therefore, the present application provides a gas cylinder management method to solve the problems mentioned in the background art.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a gas cylinder management method, comprising:

[0008] Obtain time series data, material characteristics and maintenance records from gas cylinders of each material, based on the time series data, material characteristics and maintenance records, use PC algorithm to establish gas cylinder life prediction task, and according to the established gas cylinder life prediction task, screen out related features affecting the life of gas cylinders;

[0009] Send the related features into a graph convolutional neural network model for fusion, perform knowledge transfer according to the number of fused features, and obtain the life prediction result of each material gas cylinder;

[0010] By evaluating and optimizing the knowledge transfer effect, the accuracy of the life prediction of each material gas cylinder is improved, and the intelligent management of the gas cylinder is realized.

[0011] As a preferred scheme of the gas cylinder management method of the present application, wherein: PC algorithm is used, PC algorithm is used to establish gas cylinder life prediction task, according to the established gas cylinder life prediction task, related features affecting the life of gas cylinders are screened out, including:

[0012] Define different basic collection features for each material gas cylinder to form a data set D of different basic collection features under each material gas cylinder;

[0013] Infer the causal relationship between the basic collection features under each material gas cylinder and the corresponding gas cylinder life from the data set D through PC algorithm, and generate a directed acyclic graph;

[0014] For each basic collection feature, calculate its average causal effect value and mutual information value in the corresponding gas cylinder life;

[0015] Define an importance score through the average causal effect value and the mutual information value, and select the top k basic collection features with the highest importance score;

[0016] Wherein, k is determined by the total number of basic collection features and the number of each material gas cylinder.

[0017] As a preferred scheme of the gas cylinder management method of the present application, wherein: the related features are sent into a graph convolutional neural network model for fusion, including:

[0018] Based on the graph convolutional neural network model, the first k basic collection features are regarded as graph nodes, the initial features of each graph node are generated through pre-encoding, and the correlation between the features of each graph node is calculated to define the initial edge weight of the graph node;

[0019] The graph node features and edge weights are updated, and the last updated graph node features are globally averaged and pooled to obtain the fusion features.

[0020] As a preferred scheme of the gas cylinder management method, wherein: according to the number of fused features, knowledge transfer is performed to obtain the life prediction result of each material gas cylinder, including:

[0021] The dataset D of different basic collection features of each material gas cylinder is called, if the number of fused features is less than the dataset D, knowledge transfer is adopted, and if the number of fused features is equal to or greater than the dataset D, knowledge transfer is not adopted;

[0022] Based on the meta-learning model, the deep residual network in the meta-learning model is trained through the fused features to complete the knowledge transfer;

[0023] Any two different material gas cylinders in the dataset D are taken out, and the data amount of the two material gas cylinders is in a proportional relationship, that is, B=A*k, k is a non-zero constant, A represents a gas cylinder with a small data amount, and B represents a gas cylinder with a large data amount;

[0024] B is regarded as a source domain, the source domain includes samples, a training set and a validation set; A is regarded as a target domain, the target domain includes samples, a support set and a query set;

[0025] The source domain is divided into multiple training tasks, each training task has different task difficulty;

[0026] The global parameters and local parameters of the deep residual network are initialized, the local update and global update are performed for each training task, and the global parameters and local parameters are updated through the target domain.

[0027] As a preferred scheme of the gas cylinder management method, wherein: the global parameters and local parameters are updated through the target domain, including:

[0028] The local parameter update is performed on the support set of the target domain, the local parameter update process is regarded as an inner loop, the meta-loss is calculated on the query set of the target domain to evaluate the local parameter update, the process of calculating the meta-loss is regarded as an outer loop, and the global parameters are updated;

[0029] The inner loop and outer loop processes are repeated until the meta-loss reaches a convergence state, and the life prediction result of each material gas cylinder is output.

[0030] As a preferred scheme of the gas cylinder management method, wherein: the knowledge transfer effect is evaluated and optimized, including:

[0031] In the case of knowledge transfer, the validation set is input into the meta-learning model to obtain source domain prediction values, and the source domain prediction values are compared with the validation set by calculating the error index MES to obtain the source domain accuracy; the query set is input into the meta-learning model to obtain target domain prediction values, and the target domain prediction values are compared with the query set by calculating the error index MES to obtain the target domain accuracy;

[0032] The source domain accuracy and the target domain accuracy are calculated with the meta-learning model without knowledge transfer, and if the calculation result is greater than 1, it indicates that the knowledge transfer can improve the accuracy of the meta-learning model for predicting the service life of each material gas cylinder, otherwise it indicates that the knowledge transfer effect is not ideal, and the deep residual network needs to be re-established.

[0033] In a second aspect, the present application provides a gas cylinder management system, comprising:

[0034] The gas cylinder data processing module is configured to obtain time series data, material characteristics and maintenance records from each material gas cylinder, establish a gas cylinder service life prediction task based on the time series data, material characteristics and maintenance records using a PC algorithm, and screen out related features affecting the service life of the gas cylinder according to the established gas cylinder service life prediction task.

[0035] The gas cylinder service life prediction module is configured to input the related features into a graph convolutional neural network model for fusion, perform knowledge transfer according to the number of fused features, and obtain the service life prediction result of each material gas cylinder.

[0036] The gas cylinder service life prediction optimization module is configured to evaluate and optimize the knowledge transfer effect, thereby improving the accuracy of the service life prediction of each material gas cylinder and realizing the intelligent management of the gas cylinder.

[0037] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any step of the above method when executing the computer program.

[0038] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the above method.

[0039] Compared with the prior art, the present application has the following advantages:

[0040] 1、The prior art often ignores the complex interaction between material characteristics and use environment in gas cylinder life prediction, resulting in limited generalization ability of the model, and the application establishes a gas cylinder life prediction task by adopting a PC (Peter-Clark) algorithm, screens out key features affecting the life of the gas cylinder, uses the PC algorithm to construct the dependency relationship between the features by using causal inference, and combines the time series data, material characteristics and maintenance records to accurately identify the most important variables for gas cylinder life prediction, thereby improving the adaptability of the model to different types of gas cylinders.

[0041] 2、For rare material gas cylinders with small amount of data, the traditional transfer learning method lacks a cross-domain transfer mechanism of deep knowledge, resulting in insufficient robustness and accuracy of the model in a small sample scene, the application designs a knowledge transfer scheme based on the number of fused features, and uses a meta-learning model and a deep residual network (ResNet) to realize deep knowledge transfer from data-rich gas cylinders to rare material gas cylinders, wherein the meta-learning optimizes global parameters through multi-task training and quickly adapts to small samples in the target domain, and the deep residual network captures complex nonlinear feature patterns, which helps to improve the accuracy and stability of the rare material gas cylinder life prediction, and fills the gap of the prior art in the small sample scene. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor. Among them:

[0043] Figure 1 The general flowchart of the gas cylinder management method described in an embodiment of the application;

[0044] Figure 2 The feature selection influence graph on prediction MSE described in an embodiment of the application;

[0045] Figure 3 The prediction error comparison graph under different training sample quantities described in an embodiment of the application. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0047] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0048] Secondly, the "one embodiment" or "an embodiment" referred to herein means containing a specific feature, structure or characteristic that can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate from or mutually exclusive with other embodiments.

[0049] The present application is described in detail in conjunction with the schematic diagram. In the detailed description of the embodiments of the present application, the sectional view of the device structure is partially enlarged without the general proportion for the convenience of illustration, and the schematic diagram is only an example, which should not limit the scope of protection of the present application herein. In addition, three-dimensional spatial dimensions including length, width and depth should be included in actual manufacture.

[0050] Meanwhile, in the description of the present application, it should be noted that the terms "upper, lower, inner and outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first, second or third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] In the present application, unless otherwise explicitly specified and limited, the terms "mounting, connection, connection" should be understood broadly, for example: it can be fixed connection, detachable connection or integral connection; it can also be mechanical connection, electrical connection or direct connection, it can also be indirectly connected through intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0052] Embodiment 1

[0053] Reference Figure 1 For the first embodiment of the present application, the embodiment provides a gas cylinder management method, comprising:

[0054] S1, obtaining time sequence data, material characteristics and maintenance records from gas cylinders of each material, based on the time sequence data, material characteristics and maintenance records, using PC algorithm, establishing a gas cylinder life prediction task, and screening out related features affecting the life of the gas cylinder according to the established gas cylinder life prediction task;

[0055] Specifically, the timing data is acquired by a gas cylinder sensor, and the timing data is normalized;

[0056] Specifically, the timing data is directly read using a communication protocol (such as RS-485, Modbus, CAN bus, 4-20mA analog signal, etc.) matched with the sensor. If the gas cylinder sensor has been connected to a SCADA (Supervisory Control and Data Acquisition) system or an IoT platform (such as Aliyun IoT, AWS IoT, ThingsBoard, etc.), historical data can be directly exported from the background data management of the system or platform.

[0057] It should be noted that the timing data is normalized by using Z-score to avoid the problem that the difference in the numerical range of the dimension in the timing data will cause the model to learn to be biased towards large-scale features, thereby ignoring small-scale features.

[0058] Specifically, if the gas cylinder belongs to non-public facilities, the material specifications in the gas cylinder nameplate or product manual (such as the material specifications of ISO 9809 standard gas cylinders) are queried to obtain the material characteristics of the gas cylinder. If the gas cylinder belongs to public facilities, the registration records of the industry supervision platform are queried to obtain the material characteristics of the gas cylinder.

[0059] Specifically, the maintenance labels (such as the last detection date, filling record, etc.) on the gas cylinder are checked or the electronic report of gas cylinder safety monitoring provided by the corresponding detection institution according to the “Gas Cylinder Safety Technology Supervision Regulations” is obtained, thereby obtaining the gas cylinder maintenance record.

[0060] It should be noted that the basic collection parameters include the sampling frequency (such as 1 per second), the data type (such as pressure, temperature, vibration, etc.), the material characteristics of the gas cylinder (steel, aluminum, and composite materials), and the gas cylinder maintenance record (maintenance period, maintenance time, etc.), and the data type includes timing data. For example, for a steel gas cylinder, pressure and temperature are collected by default. For a composite material gas cylinder, additional vibration data may be collected. Therefore, for gas cylinders of different materials, the basic sampling parameters are added or deleted according to the specific material.

[0061] Further, different basic collection characteristics are defined for gas cylinders of each material to form a data set D of different basic collection characteristics for each material gas cylinder.

[0062] Further, the PC algorithm is used to infer the causal relationship between the basic collection characteristics of each material gas cylinder and its corresponding gas cylinder life from the data set D, and a directed acyclic graph is generated. For example, if the directed acyclic graph is composed of “pressure → temperature → life”, then temperature is an intermediate variable.

[0063] It needs to be explained that PC (Peter-Clark) algorithm is a classical algorithm for causal discovery, aiming to infer the causal structure between variables from observational data;

[0064] Further, for each basic acquisition feature, the average causal effect value and the mutual information value corresponding to the gas cylinder life are calculated;

[0065] Specifically, the average causal effect of each basic acquisition feature on the gas cylinder life is calculated, and the formula is expressed as:

[0066] ACE(X i →Y)=E[Y∣do(X i =x1)]-E[Y∣do(X i =x0)]

[0067] Wherein, X i is a certain basic acquisition feature, x0 and x1 are expressed as the reference value of X i ; do is expressed as the intervention operation; ACE is expressed as the expected difference of a certain basic acquisition feature X i on the gas cylinder life Y;

[0068] Exemplarily, the relationship between x0 and x1 and X i is specified, if X i is the pressure of the gas cylinder, x0 can be a low pressure state, such as 50 MPa (megapascal), and x1 can be a high pressure state, such as 150 MPa (megapascal); if X i is the temperature of the gas cylinder, x0 can be room temperature, such as 20℃, and x1 can be high temperature, such as 80℃; thus it is illustrated that x0 and x1 are expressed as the reference value of X i , and are two opposite but representative reference values;

[0069] Specifically, the mutual information value of each basic acquisition feature and the gas cylinder life is calculated, and the formula is expressed as:

[0070]

[0071] Wherein, MI is expressed as the mutual information value of a certain basic acquisition feature X i and the gas cylinder life Y, the greater the value of MI(X i , Y), the stronger the dependence between the basic acquisition feature and the gas cylinder life, if MI(X i , Y) = 0, it means that the two are independent; p(x i , y) is expressed as the joint probability of X i taking x i and Y taking y, x i is expressed as X iall values of x, including x0 and x1, and similarly y represents all values of Y; for example, the probability that the pressure is 50 MPa and the cylinder lifetime is 10 years;

[0072] Further, by averaging the causal effect value and the mutual information value, an importance score is defined, and the top k basic acquisition features with the highest importance score are selected;

[0073] Specifically, the importance score S i is represented as:

[0074] S i = w1 ACE i + w2 MI i

[0075] wherein w1 and w2 respectively correspond to the weight values of ACE i and MI i ;

[0076] Specifically, k is determined by the total number of basic acquisition features and the number of gas cylinders of each material;

[0077] It should be noted that by "hooking" k (i.e., the top k basic acquisition features selected) with the total number of basic acquisition features and the number of gas cylinders of each material, the basic acquisition feature selection can be dynamically adjusted according to the data volume of different material gas cylinders, thereby realizing the "personalized" configuration of basic acquisition feature selection; because in actual application, the data volume of different material gas cylinders is often unbalanced, for example, some rare material samples are less (such as carbon fiber composite material gas cylinders), while common material samples are more (such as steel gas cylinders), so for data-rich steel gas cylinders, more features can be selected (k value is larger);

[0078] S2, the relevant features are sent into the graph convolutional neural network model for fusion, and knowledge transfer is performed according to the number of fused features to obtain the lifetime prediction result for each material gas cylinder;

[0079] Further, based on the graph convolutional neural network model, the top k basic acquisition features are regarded as graph nodes, the initial features of each graph node are generated through pre-encoding, and the correlation between the features of each graph node is calculated to define the initial edge weight of the graph node.

[0080] Specifically, the correlation between the features of each graph node is calculated using cosine similarity, and the initial edge weight of the graph node is:

[0081]

[0082] wherein, represents the initial edge weight, initial features of the nth graph node represented as pre-encoding generation, initial features of the mth graph node represented as pre-encoding generation;

[0083] Further, update the graph node features and edge weights, and perform global average pooling on the last updated graph node features to obtain the fusion features;

[0084] Specifically, the update of the graph node features is represented by the formula:

[0085]

[0086] Wherein, N(n) is the neighbor node of the graph node, d n is the degree of the nth graph node, d m is the degree of the mth graph node, W (l) represents the weight matrix, and sigma = ReLU is the activation function; l and l+1 represent the lth layer and the l+1 layer of the graph convolutional neural network, respectively;

[0087] Specifically, the update of the edge weight is represented by the formula:

[0088]

[0089] Wherein, || represents the splicing operation, W e is the training parameter;

[0090] Specifically, the fusion features H are represented by the formula:

[0091]

[0092] Wherein, L represents the total number of layers of the graph convolutional neural network, and N represents the total number of graph nodes;

[0093] Further, the data set D of different basic collection features under each material gas cylinder is called, if the number of fusion features is less than the data set D, knowledge transfer is adopted, and if the number of fusion features is equal to or greater than the data set D, knowledge transfer is not adopted.

[0094] It should be noted that assuming that the called data set D has 15 features, and the number of fusion features is only 8 features, if the 8 features are highly related to the gas cylinder life prediction (such as pressure trend, fatigue data), the knowledge transfer may still be effective; on the contrary, if the number of fusion features reaches 20, but contains many irrelevant features (such as the influence of environmental humidity on steel gas cylinders is small), then the transfer may not be effective; then on this basis, the basic collection features are screened through the importance score, which can well solve the correlation problem of the fusion features, so that the transfer becomes effective.

[0095] Further, based on the meta-learning model, the deep residual network in the meta-learning model is trained by fusing features to complete knowledge transfer.

[0096] Specifically, the deep residual network includes multiple residual blocks, each block is in the form of I=F(H)+H, F(H) represents residual mapping composed of multiple convolution layers, activation functions, etc., H is the input fused feature, and I is the output fused feature.

[0097] Further, any two different material cylinders in the data set D are taken out, and the data amount of the two material cylinders is in a proportional relationship, i.e. B=A*k, k is a non-zero constant, A represents a cylinder with a small amount of data, and B represents a cylinder with a large amount of data.

[0098] Further, B is regarded as a source domain, and the source domain includes samples, a training set and a validation set; A is regarded as a target domain, and the target domain includes samples, a support set and a query set.

[0099] It should be explained that the target domain is a specific field or task that the model can adapt to and predict in knowledge transfer, and the source domain is usually a rich data set; taking the prediction of the service life of a cylinder as an example, the service life of a carbon fiber composite cylinder is currently to be predicted, but the relevant data samples are few, and compared with the large amount of data of a steel cylinder (source domain), the knowledge in the source domain needs to be applied to the target domain through transfer learning, and the support set and the query set are equivalent to “teaching materials” and “examination questions”; assuming that the support set has 10 data samples of carbon fiber composite cylinders, each sample includes features (pressure, temperature, etc.) and corresponding service life (such as 25000 hours); then through the support set, the model can learn the characteristics of the 10 data samples of carbon fiber composite cylinders, i.e. form “teaching materials”, and then through the query set, i.e. “examination questions”, the result of the model after learning the samples is evaluated, i.e. whether the “examination questions” are guided by the “teaching materials”;

[0100] Further, the source domain is divided into multiple training tasks, each training task has different task difficulty.

[0101] Further, the global parameters θ and the local parameters θ' of the deep residual network are initialized, the local update and the global update are performed on each training task, and the global parameters and the local parameters are updated through the target domain.

[0102] It should be explained that the global parameters and the local parameters of each training task are updated through the target domain, which is a key step for the model to adapt to the target domain.

[0103] Further, the local parameter updating is performed on the support set of the target domain, the local parameter updating process is taken as an inner loop, the meta-loss is calculated on the query set of the target domain to evaluate the local parameter updating, the process of calculating the meta-loss is taken as an outer loop, and the global parameter is updated;

[0104] Specifically, before updating, not only the global parameter and the local parameter of the deep residual network need to be initialized, but also the global learning rate and the local learning rate need to be initialized, the global learning rate is set to 0.001, and the local learning rate is set to 0.01.

[0105] Specifically, the local parameter updating is performed on the support set of the target domain, which can be represented by a formula as follows:

[0106]

[0107] Wherein, α represents the local learning rate, α={α0,α1,…,α n}; represents the loss value of the nth task under the support set support, represents the gradient of the global parameter, θ′ n represents the nth local parameter updating value;

[0108] Specifically, the meta-loss L meta is calculated on the query set of the target domain, which can be represented by a formula as follows:

[0109]

[0110] Wherein, represents the loss value of the nth task under the query set query;

[0111] Specifically, the global parameter updating is represented by a formula as follows:

[0112]

[0113] Wherein, η represents the global learning rate;

[0114] Specifically, the inner loop and the outer loop process are repeated until the meta-loss reaches a convergence state, and the life prediction result of each material gas cylinder is outputted;

[0115] Further, the local learning rate is adjusted according to the task difficulty, and the variance of the training task is defined as D n , then we get:

[0116] α n =α0·exp(D n ),α0=0.01

[0117] Wherein, α0 is the initialized local learning rate;

[0118] It should be noted that the local learning rate is adjusted by the task difficulty, so that the model reduces the learning rate in the task with a large number of features to avoid overfitting, and maintains a high learning rate in the task with a constant or small number of features to accelerate the convergence of the meta-loss;

[0119] S3, by evaluating and optimizing the knowledge transfer effect, the accuracy of the life prediction of each material gas cylinder is improved, and the intelligent management of the gas cylinder is realized;

[0120] Further, in the case of knowledge transfer, the validation set is input into the meta-learning model to obtain the source domain prediction value, the source domain prediction value is compared with the validation set by calculating the error index MES, and the source domain accuracy is obtained; the query set is input into the meta-learning model to obtain the target domain prediction value, the target domain prediction value is compared with the query set by calculating the error index MES, and the target domain accuracy is obtained;

[0121] Specifically, the error index MES formula is as follows:

[0122]

[0123] Wherein, v represents the number of validation set or query set, represents the source domain prediction value or target domain prediction value, z i represents the validation set or query set; the output result of the error index MES is the source domain accuracy or the target domain accuracy;

[0124] Further, the source domain accuracy and the target domain accuracy are calculated with the meta-learning model without using knowledge transfer, if the calculation result is greater than 1, it indicates that the knowledge transfer can improve the accuracy of the meta-learning model for the life prediction of each material gas cylinder, otherwise it indicates that the knowledge transfer effect is not ideal, and the deep residual network needs to be re-established;

[0125] Specifically, the source domain accuracy and the target domain accuracy are calculated with the meta-learning model without using knowledge transfer, and the formula is as follows:

[0126]

[0127] Wherein, TLE represents the transfer coefficient, MSE source represents the source domain accuracy, MSE base represents the meta-learning model without using knowledge transfer, MSE target represents the target domain accuracy;

[0128] It should be explained that the accuracy of the life prediction of the gas cylinder of each material is improved by knowledge transfer, and the life of the gas cylinder of each material is uniformly managed according to the predicted life of the gas cylinder, which can not only reduce the workload of manual inspection, but also identify and replace the aging or damaged gas cylinder, thereby reducing the risk of accidents caused by gas cylinder failure, and providing effective support for intelligent gas cylinder management.

[0129] Further, the embodiment also provides a gas cylinder management system, comprising:

[0130] The gas cylinder data processing module is configured to obtain time series data, material characteristics and maintenance records from the gas cylinders of each material, establish a gas cylinder life prediction task based on the time series data, material characteristics and maintenance records by using a PC algorithm, and screen out related features affecting the life of the gas cylinder according to the established gas cylinder life prediction task;

[0131] The gas cylinder life prediction module is configured to input the related features into a graph convolutional neural network model for fusion, perform knowledge transfer according to the number of fused features, and obtain the life prediction result of each material gas cylinder;

[0132] The gas cylinder life prediction optimization module is configured to improve the accuracy of the life prediction of each material gas cylinder by evaluating and optimizing the knowledge transfer effect, and realize the intelligent management of the gas cylinder.

[0133] The embodiment also provides a computer device suitable for the case of the gas cylinder management method, comprising:

[0134] The memory and the processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the gas cylinder management method proposed in the above embodiment.

[0135] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. It can also be an external keyboard, touchpad or mouse, etc.

[0136] The embodiment also provides a storage medium, which stores a computer program. The computer program is executed by a processor to implement the gas cylinder management method provided by the above embodiment.

[0137] The storage medium provided by the embodiment belongs to the same inventive concept as the data storage method provided by the above embodiment. The technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0138] Embodiment 2

[0139] With reference to Figure 2 and Figure 3 The second embodiment of the present application provides a gas cylinder management method, which comprises:

[0140] In the embodiment, three gas cylinders made of different materials are selected as test objects: a steel gas cylinder (AISI 4130), an aluminum gas cylinder (AA 6061-T6), and a carbon fiber composite gas cylinder (Toray T800H / 3900-2). The test data is obtained from a gas cylinder monitoring system of an industrial gas company, covering the use records from 2022 to 2025. The data acquisition equipment includes a Honeywell TruStability pressure sensor (model: HSC series, accuracy ±0.1% FS), an Omega temperature sensor (model: THW-4, accuracy ±0.5℃), and a maintenance record database (based on MySQL 8.0). The control group uses a traditional linear regression model. Based on historical pressure and temperature data and maintenance cycles, a multiple linear regression equation is constructed to predict the service life. The SVM (Support Vector Machine) model is used. The data of the steel gas cylinder is taken as the source domain, and is transferred to the target domain (aluminum and carbon fiber gas cylinders) through feature matching.

[0141] The experimental group adopts the scheme of the application, and the steps are as follows: Z-score standardization is adopted to eliminate the dimensional difference in time series data, the gas cylinder material characteristics are obtained through a material manual, including yield strength (steel: 550 MPa, aluminum: 275 MPa, carbon fiber: 2,800 MPa) and corrosion resistance grade (ISO 9223 standard); coded as a binary feature according to the maintenance period (6 months / 12 months) and the test result (qualified / failed); the PC algorithm is used to construct a causal relationship diagram, and the importance score is calculated (weight: ACE 60%, MI 40%), and the top 5 key features are screened (steel: pressure fluctuation, temperature gradient, filling frequency, maintenance period, fatigue index; carbon fiber: vibration amplitude, temperature change rate, ultraviolet exposure time, resin thickness, filling pressure); the graph nodes are initialized, each feature node is embedded into a 128-dimensional vector, the edge weight is calculated by cosine similarity, the network structure is 2-layer graph convolution (GCN), the activation function is ReLU, and the pooling layer outputs a 256-dimensional fusion feature; the source domain (steel cylinder, data amount 10,000) and the target domain (carbon fiber cylinder, data amount 200) are divided according to the proportion B=50xA, the task difficulty is adjusted according to the local learning rate (alpha 0 =0.01) according to the task variance; the training parameters are: global learning rate η=0.001, iteration 50 times, and meta-loss convergence threshold 0.01;

[0142] Through comparison between the experimental group and the control group, part of the data is displayed in the form of an experimental graph, as shown in Figure 2 It can be seen that with the number of features decreasing from the complete feature set to the number after screening, the MSE of the three types of material cylinders (steel, aluminum and carbon fiber) all significantly decreases, the MSE of the steel cylinder decreases from a higher value to a lower value when the number of features increases from 5 to 10, and the aluminum and carbon fiber cylinders also show similar trends; it is shown that the PC algorithm removes the noise features by screening the top k features with the highest importance score, and improves the prediction accuracy of the model, the results are consistent with the effect after the PC algorithm screens the features, which proves the advantage of the application in feature selection; secondly, as shown in Figure 3 In the small sample scene (10, 20 samples) and the medium sample scene (50, 100 samples), the prediction error of the non-transfer model is large due to insufficient data, and the performance is poor; while the transfer model significantly reduces the error by using the source domain knowledge, which reflects the advantage of knowledge transfer;

[0143] In summary, through experimental comparison, it is further proved that the scheme of the application has strong applicability in the data scarce environment.

[0144] Those skilled in the art will appreciate that embodiments of the application can be devised for a variety of computer-implemented processes. Accordingly, the application can be embodied in a variety of different forms and embodiments. For example, the application can be embodied in software application form, installed on a computer readable medium, executed by a computer and / or processed by a computer using any of a wide variety of computer languages (e.g., a Java-based scripting language and / or a C++ based scripting language). Additionally, the application can be embodied in a variety of different forms and embodiments, such as a computer program product, a computer readable medium, a computer system, and / or a computer processor.

[0145] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0146] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0147] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0148] Although preferred embodiments of the application have been described, those skilled in the art will appreciate that additional modifications and changes can be made thereto without departing from the scope of the application. Accordingly, the appended claims are intended to cover all such modifications and changes as fall within the scope of the application.

[0149] Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A method for managing gas cylinders, characterized in that, include: Time-series data, material characteristics, and maintenance records are obtained from gas cylinders of each material. Based on the time-series data, material characteristics, and maintenance records, a PC algorithm is used to establish a gas cylinder life prediction task. Based on the established gas cylinder life prediction task, relevant characteristics affecting gas cylinder life are selected. The process employs a PC algorithm to establish a gas cylinder life prediction task. Based on this task, relevant features affecting gas cylinder life are selected, including: Different basic acquisition features are defined for each type of gas cylinder, forming a dataset D of different basic acquisition features for each type of gas cylinder; Using the PC algorithm, the causal relationship between the basic collection features of each type of gas cylinder and its corresponding lifespan is inferred from the dataset D, and a directed acyclic graph is generated. For each basic collected feature, calculate its average causal effect value and mutual information value over the corresponding gas cylinder lifespan; An importance score is defined using the average causal effect value and the mutual information value, and the top k basic features with the highest importance scores are selected. Where k is determined by the total number of basic acquisition features and the number of gas cylinders of each material; The relevant features are fed into a graph convolutional neural network model for fusion, and knowledge transfer is performed based on the number of fused features to obtain the life prediction results for each type of gas cylinder. By evaluating and optimizing the effect of knowledge transfer, the accuracy of life prediction for each type of gas cylinder can be improved, thereby realizing intelligent gas cylinder management. The evaluation and optimization of knowledge transfer effectiveness includes: When knowledge transfer is employed, the validation set is input into the meta-learning model to obtain the source domain prediction value. The source domain prediction value is compared with the validation set by calculating the error index MES to obtain the source domain accuracy. The query set is input into the meta-learning model to obtain the target domain prediction value. The target domain prediction value is compared with the query set by calculating the error index MES to obtain the target domain accuracy. The source domain accuracy and target domain accuracy are calculated with the meta-learning model without knowledge transfer. If the calculation result is greater than 1, it indicates that knowledge transfer can improve the accuracy of the meta-learning model in predicting the lifespan of gas cylinders of each material. Otherwise, it indicates that the knowledge transfer effect is not ideal and a new deep residual network needs to be established.

2. The gas cylinder management method as described in claim 1, characterized in that, The relevant features are fed into a graph convolutional neural network model for fusion, including: Based on the graph convolutional neural network model, the first k basic collected features are regarded as graph nodes. The initial features of each graph node are generated through precoding. At the same time, the correlation between the features of each graph node is calculated, and the initial edge weights of the graph nodes are defined. Update the graph node features and edge weights, and perform global average pooling on the last updated graph node features to obtain the fused features.

3. The gas cylinder management method as described in claim 1 or 2, characterized in that, Based on the number of fused features, knowledge transfer is performed to obtain life prediction results for gas cylinders of each material, including: If the number of fused features is less than that of the dataset D, knowledge transfer is used; if the number of fused features is equal to or greater than that of the dataset D, knowledge transfer is not used. Based on the meta-learning model, the deep residual network in the meta-learning model is trained through the fusion features to complete knowledge transfer; Take out any two different types of gas cylinders from the dataset D, and the amount of data for the two types of gas cylinders is in a positive proportional relationship, i.e., B=A×k, where k is a non-zero constant, A represents gas cylinders with sparse data, and B represents gas cylinders with abundant data. Consider B as the source domain, which includes samples, training set, and validation set; consider A as the target domain, which includes samples, support set, and query set. The source domain is divided into multiple training tasks, each with a different level of difficulty. Initialize the global and local parameters of the deep residual network, perform local and global updates for each training task, and update the global and local parameters through the target domain.

4. The gas cylinder management method as described in claim 3, characterized in that, Updating the global and local parameters through the target domain includes: Local parameter updates are performed on the support set of the target domain. The local parameter update process is used as the inner loop. The local parameter updates are evaluated by calculating the meta-loss on the query set of the target domain. The meta-loss calculation process is used as the outer loop, and the global parameters are updated. Repeat the inner and outer loop processes until the element loss reaches a convergent state, and output the life prediction results for each type of gas cylinder.

5. A gas cylinder management system, based on the gas cylinder management method according to any one of claims 1 to 4, characterized in that, include: The gas cylinder data processing module is configured to acquire time-series data, material characteristics, and maintenance records from gas cylinders of each material. Based on the time-series data, material characteristics, and maintenance records, a PC algorithm is used to establish a gas cylinder life prediction task. Based on the established gas cylinder life prediction task, relevant features affecting the gas cylinder life are selected. The gas cylinder life prediction module is configured to feed the relevant features into a graph convolutional neural network model for fusion, perform knowledge transfer based on the number of fused features, and obtain life prediction results for gas cylinders of each material. The cylinder life prediction optimization module is configured to improve the accuracy of life prediction for each type of cylinder by evaluating and optimizing the effect of knowledge transfer, thereby realizing intelligent cylinder management.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images