Pneumatic control valve fault diagnosis method and system of coupling data driving and ontology model

Abstract

The invention discloses a pneumatic control valve fault diagnosis method and system for coupling data driving and a body model, which can realize whole-flow fault diagnosis of a pneumatic control valve in an oil-gas pipe network. According to the method, the fault state of the pneumatic control valve is judged by collecting operation data of the pneumatic control valve and establishing a fault diagnosis PCA model of the pneumatic control valve; a pneumatic control valve fault classification SVM model is established, and fault modes are classified; and constructing a pneumatic control valve fault ontology knowledge base, performing ontology reasoning, and outputting a pneumatic control valve fault diagnosis conclusion and a maintenance method. According to the method, the advantages of a data driving method and an ontology model are combined, full-process fault diagnosis of the pneumatic control valve is achieved, the intelligent level of fault diagnosis of the pneumatic control valve is improved, and the precision and reliability of system fault diagnosis are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment fault diagnosis, and in particular to a pneumatic control valve fault diagnosis method and system coupling data drive and ontology model. Background Art

[0002] Fault diagnosis technology is a technical method to identify, locate and analyze the cause of a fault by detecting the operating status of a system or equipment. Pneumatic control valves are a typical type of control valves used in large quantities in actual production. In the process of daily monitoring and maintenance, although strict maintenance and inspection standard procedures have been formulated, with the increasing number of control valves, the development and improvement of automation technology and information technology, the station control valves and instruments are gradually developing towards intelligence, which puts forward higher requirements for equipment efficiency monitoring, daily maintenance, and fault diagnosis. The traditional fault diagnosis and maintenance methods and the instructions provided by the manufacturer lack real-time performance, resulting in the hysteresis of equipment failure awareness, excessive consumption of manpower and material resources, and it is difficult to meet the diagnosis and maintenance needs of today's equipment under complex operating conditions. It is more important and urgent to improve the intelligent level of operation management, operation monitoring, and fault diagnosis and maintenance of pneumatic control valves in the pipeline network system.

[0003] After a long period of development, the fault diagnosis of mechanical equipment is mainly based on analytical models, knowledge and data-driven. Fault diagnosis technology based on analytical models is difficult to obtain detailed system structure knowledge due to the influence of complex factors such as process and environment. For large and complex equipment, system models are generally difficult to establish, so model-based methods are no longer used alone in fault diagnosis research; although data-driven fault diagnosis methods can accurately judge the operating status and fault mode of pneumatic control valves, they cannot further analyze the faults of pneumatic control valves and cannot obtain fault knowledge such as the cause, location and maintenance method of pneumatic control valves; ontology technology has certain advantages in knowledge representation and reasoning, but due to the shortcomings of ontology technology in the analysis and learning of fault data, it is also insufficient to be used alone for pneumatic control valve fault diagnosis.

[0004] The operating state of the pneumatic control valve is affected by many factors, and the parameters are highly correlated. A single theoretical fault diagnosis method can no longer meet the systematic needs of fault diagnosis. Therefore, it is necessary to conduct research on the fusion fault diagnosis method of multiple theories to improve the applicability and accuracy of the fault diagnosis method and ensure that the pneumatic control valve can operate safely and stably. Summary of the invention

[0005] In order to solve the above problems, the present invention provides the following technical solutions:

[0006] A method for fault diagnosis of a pneumatic control valve coupled with a data-driven ontology model, characterized in that it comprises the following steps:

[0007] S1: Collect basic operating data of target pneumatic control valve;

[0008] S2: Establish a pneumatic control valve fault diagnosis PCA model based on data-driven theory to determine the fault state of the pneumatic control valve. If the pneumatic control valve has no fault, the diagnosis is terminated; if the pneumatic control valve has a fault, the fault data is collected;

[0009] S3: Establish a pneumatic control valve fault classification SVM model based on data-driven theory, input the fault data to be tested, classify the fault mode, and output the fault mode number of the pneumatic control valve;

[0010] S4: The eight-step method is used to construct the pneumatic control valve fault diagnosis ontology knowledge base, establish the semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning on the pneumatic control valve fault knowledge, and obtain the fault cause and the corresponding fault maintenance method.

[0011] It should be further explained that the basic operating data of the pneumatic control valve in step S1 includes the valve inlet pressure P in , Valve outlet pressure P out , valve flow Q, control signal CV and valve stem displacement S. The data collection scope includes normal operating condition data, historical fault data and fault data to be tested.

[0012] It should be further explained that the specific steps of S2 are:

[0013] S201: Standardize the normal operating data of the pneumatic control valve to obtain a standardized matrix, and calculate the covariance matrix and eigenvalues ​​of the normal operating data, as shown in the following formula:

[0014]

[0015] In the formula, is a standardized matrix; X is a normal operating data set; is an n-dimensional unit vector; u is the mean vector; D σ is the variance matrix; x and y are random variables in the sample data;

[0016] S202: Select the first k eigenvalues ​​in the standardized matrix whose cumulative contribution rates are greater than the threshold, and use the matrix composed of their corresponding eigenvectors as the principal element matrix, as shown in the following formula:

[0017]

[0018] Where T is the principal component score matrix; is the residual score matrix; P is the principal component loading matrix; is the residual loading matrix; the eigenvalue of the covariance matrix after λ standardization; Cl is the contribution rate;

[0019] The cumulative contribution rate threshold is set at 95%;

[0020] S203: Project the sample data to be tested to the principal component and residual subspace of the principal component model, as shown in the following formula:

[0021]

[0022] In the formula, R p is the principal element subspace; R r is the residual subspace; is the projection of the sample in the principal element subspace; is the projection of the sample in the residual subspace;

[0023] S204: Calculate the SPE and T of normal operation data at the confidence level α 2 The statistical control limits of

[0024]

[0025]

[0026] In the formula, is the SPE control limit under the confidence level α; C α is the critical value of the normal distribution; F α (k,m-1) is the critical value of F distribution;

[0027] The confidence level α was determined to be 0.95;

[0028] S205: Calculate the SPE and T of the fault data to be tested 2 The statistical formula is as follows:

[0029]

[0030] In the formula, is the i-th column of the identity matrix I; is the statistical value of each column;

[0031] S206: Comparison between SPE and T 2 The statistics and control limits are used to judge the fault. If SPE and T 2 If one of the statistical quantities exceeds the limit, it is determined that the pneumatic control valve has a fault; otherwise, it is determined that the start-up control valve has no fault and the diagnosis ends.

[0032] It should be further explained that the specific steps of S3 are:

[0033] S301: Collect historical fault data of pneumatic valves as a model training data set;

[0034] S302: A pneumatic control valve fault classification model is established based on the SVM method. The classification model adopts a "one-to-one" multi-class classification strategy, and the kernel function adopts a cubic polynomial. The optimal classification decision function is shown in the following formula:

[0035]

[0036] K(x,x i )=((x,x i )+1) 3

[0037] Where sgn is the sign function; is the Grange multiplier; K(x,x i ) is the kernel function; b * is the classification threshold;

[0038] The “one-to-one” multi-class classification strategy means that each SVM classifier is trained using only two types of pneumatic control valve fault sets in the training set. For the classification of k pneumatic control valve fault modes, it is necessary to establish SVM classifiers are used in the classification process of this method. The classification process adopts the "voting method", that is, each two-class SVM classifier will give its classification result, and the one that appears most frequently among all the classification results is the fault mode of the test sample.

[0039] S303: 70% of the model training data set is selected as a training set and 30% is selected as a test set by random allocation to train a pneumatic control valve fault classification model. When the accuracy of the training set is greater than 95% and the accuracy of the test set is greater than 90%, the model training is completed;

[0040] S304: Input the fault data to be tested into the pneumatic control valve multi-class fault mode SVM classifier for calculation, output the indication value according to the decision function, and output the fault classification result.

[0041] It should be further explained that the failure modes described in S3 can be divided into valve failure, pneumatic actuator failure, positioner failure and general failure;

[0042] The valve faults include valve body blockage / valve stem jam (fault number F1), valve plug or valve seat sediment (fault number F2), valve plug or valve seat erosion (fault number F3), valve or bushing friction increase (fault number F4), external leakage (fault number F5), internal leakage (fault number F6), medium evaporation or critical flow (fault number F7);

[0043] The pneumatic actuator failures include twisted pneumatic actuator valve stem (fault number F8), loose pneumatic actuator housing or terminals (fault number F9), diaphragm perforation of pneumatic actuator (fault number F10), and spring failure of pneumatic actuator (fault number F11);

[0044] The positioner faults include electric-pneumatic converter fault (fault number F12), rod displacement sensor fault (fault number F13), pressure sensor fault (fault number F14), and positioner feedback fault (fault number F15);

[0045] Common faults described include a drop in supply pressure to the positioner (fault number F16), unexpected pressure changes across the valve (fault number F17), a fully or partially open bypass valve (fault number F18), and a flow sensor failure (fault number F17).

[0046] It should be further explained that the specific steps of the eight-step method described in S4 are:

[0047] S401: Determine the scope of the pneumatic control valve fault diagnosis body, including fault mode number, fault type, fault cause, fault phenomenon, fault location and fault maintenance method;

[0048] S402: Observe whether there is a reusable pneumatic control valve fault diagnosis body. If not, proceed to S403. If yes, skip S403 and S404, update the original body, and proceed to S405 after the update.

[0049] S403: define a pneumatic control valve fault diagnosis ontology model, wherein the fault mode number is defined as an ontology of a pneumatic control valve fault classification number; the fault type is a specific classification of a pneumatic control valve fault; the fault cause is a cause that may cause a pneumatic control valve fault, which is classified according to multiple different parts of the pneumatic control valve; the fault phenomenon is certain special phenomena that the pneumatic control valve can exhibit under a certain fault mode of the pneumatic control valve; the fault location is the specific part of the pneumatic control valve where the fault occurs; the fault maintenance method is a maintenance method and suggestion decided by relevant management personnel for a certain pneumatic control valve fault mode;

[0050] S404: Extract key ontology terms from pneumatic regulation fault diagnosis knowledge;

[0051] S405: defining and establishing an ontology class and a class hierarchy relationship of a pneumatic control valve, wherein the class and the class hierarchy relationship include a subclass relationship and an instance relationship;

[0052] S406: defining and establishing pneumatic control valve fault diagnosis ontology attributes, wherein the ontology attributes include functionality, inverse relationship, value range, and definition domain;

[0053] S407: Create a pneumatic control valve fault diagnosis ontology instance and association;

[0054] S408: Performing an ontology consistency check on the pneumatic control valve, wherein the consistency check is implemented by the HermiT inference engine. After the check passes, the pneumatic control valve fault diagnosis ontology knowledge base is successfully established.

[0055] The present application also provides a pneumatic control valve fault diagnosis system coupled with a data-driven ontology model, the system comprising a data management module, a pneumatic control valve fault detection module, a pneumatic control valve fault mode classification module, a pneumatic control valve fault diagnosis knowledge reasoning module and a pneumatic control valve fault diagnosis knowledge management module;

[0056] The data management module is used to input the basic operation data of the target pneumatic control valve and output the pneumatic control valve fault detection data, classification data and knowledge reasoning log data;

[0057] The pneumatic control valve fault detection module is used to train and establish a data-driven pneumatic control valve fault detection PCA model, perform fault detection on a given set of pneumatic control valve operation data, determine whether the pneumatic control valve has a fault, save the corresponding fault detection record, and output the fault data;

[0058] The pneumatic control valve fault mode classification module is used to train and establish a data-driven pneumatic control valve fault mode classification SVM model, classify fault modes according to fault data, save corresponding fault mode classification records, and output fault mode classification numbers;

[0059] The pneumatic control valve fault diagnosis knowledge reasoning module is used to construct a pneumatic control valve fault diagnosis ontology knowledge base, establish a semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning according to the fault mode classification number, obtain the corresponding diagnostic conclusion, fault type, fault cause and corresponding maintenance suggestion under the pneumatic control valve fault mode, and output a pneumatic control valve fault diagnosis report;

[0060] The pneumatic control valve fault diagnosis knowledge management module is used to implement the operations of adding, editing and deleting knowledge points in the pneumatic control valve fault diagnosis ontology knowledge base.

[0061] It can be seen from the above technical solution that the present invention has the following beneficial effects:

[0062] (1) The present invention realizes the full-process fault diagnosis of pneumatic control valves by combining the advantages of data-driven methods and ontology models. This method overcomes the shortcomings of single fault diagnosis methods in analysis depth and accuracy, and can more comprehensively and accurately monitor the operating status of pneumatic control valves, classify fault modes, and reason about fault diagnosis knowledge. By complementing the advantages of the two methods, the present invention improves the intelligence level of pneumatic control valve fault diagnosis, significantly improves the accuracy and reliability of system fault diagnosis, effectively ensures the operating stability and safety of pneumatic control valves, and has certain guiding significance for the fault diagnosis of other key equipment.

[0063] (2) This application proposes to use an eight-step method to construct an ontology model. Compared with the traditional ontology construction method, it adds a pneumatic control valve fault diagnosis ontology model definition step, which can make the constructed fault diagnosis ontology structure clearer and the fault diagnosis ontology knowledge representation more standardized; it adds a fault diagnosis ontology evaluation step, which can improve the reliability of the pneumatic control valve fault diagnosis ontology. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Flow chart of the method of the present invention

[0065] Figure 2 The specific flow chart of step S2 is

[0066] Figure 3 The specific flow chart of step S3 is

[0067] Figure 4 The construction process of the eight-step pneumatic control valve fault diagnosis ontology

[0068] Figure 5 The operating data distribution of a pneumatic control valve

[0069] Figure 6 Test results of the pneumatic control valve fault detection module

[0070] Figure 7 Failure mode classification module for pneumatic control valves

[0071] Figure 8 A knowledge reasoning module for fault diagnosis of pneumatic control valves

[0072] Fig. 9 This is a schematic diagram of the pneumatic control valve fault diagnosis report. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] like Figure 1 As shown, a method and system for diagnosing a pneumatic control valve fault coupled with a data-driven ontology model, characterized in that it includes the following steps:

[0075] S1: Collect basic operating data of target pneumatic control valve;

[0076] S2: Establish a pneumatic control valve fault diagnosis PCA model based on data-driven theory to determine the fault state of the pneumatic control valve. If the pneumatic control valve has no fault, the diagnosis is terminated; if the pneumatic control valve has a fault, the fault data is collected;

[0077] S3: Establish a pneumatic control valve fault classification SVM model based on data-driven theory, input the fault data to be tested, classify the fault mode, and output the fault mode number of the pneumatic control valve;

[0078] S4: The eight-step method is used to construct the pneumatic control valve fault diagnosis ontology knowledge base, establish the semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning on the pneumatic control valve fault knowledge, and obtain the fault cause and the corresponding fault maintenance method.

[0079] It should be further explained that the basic operating data of the pneumatic control valve in step S1 includes the valve inlet pressure P in , Valve outlet pressure P out , valve flow Q, control signal CV and valve stem displacement S. The parameter collection scope includes normal operating condition data, historical fault data and fault data to be tested.

[0080] It should be further explained that the specific steps of S2 are:

[0081] S201: Standardize the normal operating data of the pneumatic control valve to obtain a standardized matrix, and calculate the covariance matrix and eigenvalues ​​of the normal operating data, as shown in the following formula:

[0082]

[0083] In the formula, is a standardized matrix; X is a normal operating data set; is an n-dimensional unit vector; u is the mean vector; Dσ is the variance matrix; x and y are random variables in the sample data;

[0084] S202: Select the first k eigenvalues ​​in the standardized matrix whose cumulative contribution rates are greater than the threshold, and use the matrix composed of their corresponding eigenvectors as the principal element matrix, as shown in the following formula:

[0085]

[0086] Where T is the principal component score matrix; is the residual score matrix; P is the principal component loading matrix; is the residual loading matrix; the eigenvalue of the covariance matrix after λ standardization; Cl is the contribution rate;

[0087] The cumulative contribution rate threshold is set at 95%;

[0088] S203: Project the sample data to be tested to the principal component and residual subspace of the principal component model, as shown in the following formula:

[0089]

[0090] In the formula, R p is the principal element subspace; R r is the residual subspace; is the projection of the sample in the principal element subspace; is the projection of the sample in the residual subspace;

[0091] S204: Calculate the SPE and T of normal operation data at the confidence level α 2 The statistical control limits of

[0092]

[0093] In the formula, is the SPE control limit under the confidence level α; C α is the critical value of the normal distribution; F α (k,m-1) is the critical value of F distribution;

[0094] The confidence level α was determined to be 0.95;

[0095] S205: Calculate the SPE and T of the fault data to be tested 2 The statistical formula is as follows:

[0096]

[0097] In the formula, is the i-th column of the identity matrix I; is the statistical value of each column;

[0098] S206: Comparison between SPE and T 2 The statistics and control limits are used to judge the fault. If SPE and T 2 If one of the statistical quantities exceeds the limit, it is determined that the pneumatic control valve has a fault; otherwise, it is determined that the start-up control valve has no fault and the diagnosis ends.

[0099] It should be further explained that the specific steps of S3 are:

[0100] S301: Collect historical fault data of pneumatic valves as a model training data set;

[0101] S302: A pneumatic control valve fault classification model is established based on the SVM method. The classification model adopts a "one-to-one" multi-class classification strategy, and the kernel function adopts a cubic polynomial. The optimal classification decision function is shown in the following formula:

[0102]

[0103] K(x,x i )=((x,x i )1) 3

[0104] Where sgn is the sign function; is the Grange multiplier; K(x,x i ) is the kernel function; b * is the classification threshold;

[0105] The “one-to-one” multi-class classification strategy means that each SVM classifier is trained using only two types of pneumatic control valve fault sets in the training set. For the classification of k pneumatic control valve fault modes, it is necessary to establish SVM classifiers are used in the classification process of this method. The classification process adopts the "voting method", that is, each two-class SVM classifier will give its classification result, and the one that appears most frequently among all the classification results is the fault mode of the test sample.

[0106] S303: 70% of the model training data set is selected as a training set and 30% is selected as a test set by random allocation to train a pneumatic control valve fault classification model. When the accuracy of the training set is greater than 95% and the accuracy of the test set is greater than 90%, the model training is completed;

[0107] S304: Input the fault data to be tested into the pneumatic control valve multi-class fault mode SVM classifier for calculation, output the indication value according to the decision function, and output the fault classification result.

[0108] It should be further explained that the failure modes described in S3 can be divided into valve failure, pneumatic actuator failure, positioner failure and general failure;

[0109] The valve faults include valve body blockage / valve stem jam (fault number F1), valve plug or valve seat sediment (fault number F2), valve plug or valve seat erosion (fault number F3), valve or bushing friction increase (fault number F4), external leakage (fault number F5), internal leakage (fault number F6), medium evaporation or critical flow (fault number F7);

[0110] The pneumatic actuator failures include twisted pneumatic actuator valve stem (fault number F8), loose pneumatic actuator housing or terminals (fault number F9), diaphragm perforation of pneumatic actuator (fault number F10), and spring failure of pneumatic actuator (fault number F11);

[0111] The positioner faults include electric-pneumatic converter fault (fault number F12), rod displacement sensor fault (fault number F13), pressure sensor fault (fault number F14), and positioner feedback fault (fault number F15);

[0112] Common faults described include a drop in supply pressure to the positioner (fault number F16), unexpected pressure changes across the valve (fault number F17), a fully or partially open bypass valve (fault number F18), and a flow sensor failure (fault number F17).

[0113] It should be further explained that the specific steps of the eight-step method described in S4 are:

[0114] S401: Determine the scope of the pneumatic control valve fault diagnosis body, including fault mode number, fault type, fault cause, fault phenomenon, fault location and fault maintenance method;

[0115] S402: Observe whether there is a reusable pneumatic control valve fault diagnosis body. If not, proceed to S403. If yes, skip S403 and S404, update the original body, and proceed to S405 after the update.

[0116] S403: define a pneumatic control valve fault diagnosis ontology model, wherein the fault mode number is defined as an ontology of a pneumatic control valve fault classification number; the fault type is a specific classification of a pneumatic control valve fault; the fault cause is a cause that may cause a pneumatic control valve fault, which is classified according to multiple different parts of the pneumatic control valve; the fault phenomenon is certain special phenomena that the pneumatic control valve can exhibit under a certain fault mode of the pneumatic control valve; the fault location is the specific part of the pneumatic control valve where the fault occurs; the fault maintenance method is a maintenance method and suggestion decided by relevant management personnel for a certain pneumatic control valve fault mode;

[0117] S404: Extract key ontology terms from pneumatic regulation fault diagnosis knowledge;

[0118] S405: defining and establishing an ontology class and a class hierarchy relationship of a pneumatic control valve, wherein the class and the class hierarchy relationship include a subclass relationship and an instance relationship;

[0119] S406: defining and establishing pneumatic control valve fault diagnosis ontology attributes, wherein the ontology attributes include functionality, inverse relationship, value range, and definition domain;

[0120] S407: Create a pneumatic control valve fault diagnosis ontology instance and association;

[0121] S408: Performing an ontology consistency check on the pneumatic control valve, wherein the consistency check is implemented by the HermiT inference engine. After the check passes, the pneumatic control valve fault diagnosis ontology knowledge base is successfully established.

[0122] The present application also provides a pneumatic control valve fault diagnosis system coupled with a data-driven ontology model, the system comprising a data management module, a pneumatic control valve fault detection module, a pneumatic control valve fault mode classification module, a pneumatic control valve fault diagnosis knowledge reasoning module and a pneumatic control valve fault diagnosis knowledge management module;

[0123] The data management module is used to input the basic operation data of the target pneumatic control valve and output the pneumatic control valve fault detection data, classification data and knowledge reasoning log data;

[0124] The pneumatic control valve fault detection module is used to train and establish a data-driven pneumatic control valve fault detection PCA model, perform fault detection on a given set of pneumatic control valve operation data, determine whether the pneumatic control valve has a fault, save the corresponding fault detection record, and output the fault data;

[0125] The pneumatic control valve fault mode classification module is used to train and establish a data-driven pneumatic control valve fault mode classification SVM model, classify fault modes according to fault data, save corresponding fault mode classification records, and output fault mode classification numbers;

[0126] The pneumatic control valve fault diagnosis knowledge reasoning module is used to construct a pneumatic control valve fault diagnosis ontology knowledge base, establish a semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning according to the fault mode classification number, obtain the corresponding diagnostic conclusion, fault type, fault cause and corresponding maintenance suggestion under the pneumatic control valve fault mode, and output a pneumatic control valve fault diagnosis report;

[0127] The pneumatic control valve fault diagnosis knowledge management module is used to implement the operations of adding, editing and deleting knowledge points in the pneumatic control valve fault diagnosis ontology knowledge base.

[0128] Taking a pneumatic control valve in a natural gas pipeline network as a research example, the method proposed in this application is used to diagnose the fault of the pneumatic control valve. First, 120 groups of historical operating data of the pneumatic control valve are collected, including 20 groups of normal operating data conditions, 520 groups of historical fault data (140 groups each for fault modes F1, F2, F6 and F16), and 180 groups of test data. After PCA model calculation, the SPE and T of the normal operating data are 0.95 at the confidence level. 2 The statistical control limits of the failure data are 0.064 and 8.0847 respectively. 2 The results of SPE statistics are shown in Table 1. As can be seen from Table 1, the accuracy of SPE statistics is between 86.43% and 99.29%, T 2 The accuracy of the statistics is 82.14% to 92.14%, and the final PCA fault detection accuracy after statistics is 90.36% to 95.36%, which meets the application requirements. Therefore, the PCA method can realize the fault detection of pneumatic control valves and lay a theoretical foundation for the subsequent pneumatic control valve fault diagnosis system.

[0129] Table 1 Pneumatic control valve fault detection results based on PCA

[0130]

[0131] In order to verify the applicability of the multi-class classification strategy and kernel function adopted by the SVM model, different multi-class classification strategies and kernel functions were selected for pneumatic control valve fault diagnosis classification, and the results are shown in Table 2. As can be seen from Table 2, in the classification results of the pneumatic control valve fault mode, the accuracy of the "one-to-one" multi-class classification strategy is higher than that of the "one-to-many" multi-class classification strategy, and the overall accuracy is between 81% and 99.5%. Among the four kernel function classifications, the "one-to-one" strategy achieved a correct classification rate of more than 92%. The overall training time of the model is between 9.09s and 19.27s. Among them, the SVM method based on the Gaussian kernel function has the shortest training time, and the SVM method based on the polynomial kernel function has the highest accuracy. There are also differences in the accuracy of the SVM methods with different polynomial kernel functions. In the classification of the four fault modes, the cubic polynomial kernel function SVM has the highest accuracy, and the "one-to-one" classification strategy has the highest accuracy of 99.5%, and the training time is also lower than other methods. The accuracy of the SVM method with the linear kernel function is similar to that of the Gaussian kernel function, but the training time is close to that of the polynomial kernel function.

[0132] In summary, the SVM method based on the cubic polynomial kernel function and the "one-to-one" multi-class classification strategy has an ideal training time and classification effect when classifying the fault modes of pneumatic control valves.

[0133] Table 2 Classification results of pneumatic control valve fault diagnosis based on SVM

[0134]

[0135] In order to verify the availability of the system established in this application, 100 sets of operating data when a pneumatic valve in a natural gas pipeline network has an F16 fault are imported into the system for full-process fault diagnosis. The distribution of operating data is as follows: Figure 5 Enter the group of test data in the fault detection interface and perform fault detection, as shown in the figure. Figure 6 As shown, the pneumatic control valve fault detection result is "fault occurred", which is in line with the expected result. At the same time, click "Perform fault mode classification" and enter the fault mode classification interface. The group of fault data is automatically entered into the continued fault data column. Click "Start classification" to perform fault mode classification, as shown in Figure 7 As shown in the figure, the fault mode result of this group of data is "F16", which is in line with the expected result. At the same time, click "Perform fault reasoning" and enter the fault diagnosis knowledge reasoning interface. The fault mode result of this group of data is automatically entered into the fault mode column. Click "Start reasoning" to perform fault knowledge reasoning, as shown in the figure. Figure 8 As shown in the figure, the fault type under this fault mode is "positioner supply pressure drops", the fault cause is "pneumatic actuator air source pressure supply failure, or air flow throttling due to occlusion effect, or excessive system air consumption or air pipe breakage and leakage", and the fault maintenance suggestion is "check the actuator and positioner, replace the pneumatic actuator air source or air pipe", which is in line with the pneumatic control valve fault diagnosis knowledge. After completing the diagnosis of fault F16, click "Form Fault Diagnosis Report", as shown in the figure. Fig. 9 As shown, this report summarizes and displays in detail the important information of the whole process diagnosis, which can provide important reference value for operation and maintenance personnel in the actual production process.

[0136] Through the fault diagnosis test, it was found that the functions of various modules of the system operate normally, the fault diagnosis results are reliable, and can meet the pneumatic control valve fault diagnosis needs in actual on-site production, verifying the stability and reliability of the pneumatic control valve fault diagnosis system based on data drive and entity developed this time.

[0137] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method and system for fault diagnosis of pneumatic control valves by coupling data drive and ontology model, characterized in that: The following steps are involved: S1: Collect basic operating data of target pneumatic control valve; S2: Establish a pneumatic control valve fault diagnosis PCA model based on data-driven theory to determine the fault state of the pneumatic control valve. If the pneumatic control valve has no fault, the diagnosis is terminated; if the pneumatic control valve has a fault, the fault data is collected; S3: Establish a pneumatic control valve fault classification SVM model based on data-driven theory, input the fault data to be tested, classify the fault mode, and output the fault mode number of the pneumatic control valve; S4: The eight-step method is used to construct the pneumatic control valve fault diagnosis ontology knowledge base, establish the semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning on the pneumatic control valve fault knowledge, and obtain the fault cause and the corresponding fault maintenance method.

2. A method for fault diagnosis of a pneumatic control valve by coupling data drive and ontology model as claimed in claim 1, characterized in that: The basic operating data of the pneumatic control valve in step S1 includes the valve inlet pressure P in , Valve outlet pressure P out , valve flow Q, control signal CV and valve stem displacement S. The data collection scope includes normal operating condition data, historical fault data and fault data to be tested.

3. A method for fault diagnosis of a pneumatic control valve coupled with data drive and ontology model as claimed in claim 1, characterized in that: The specific steps of S2 are: S201: Standardize the normal operating data of the pneumatic control valve to obtain a standardized matrix, and calculate the covariance matrix and eigenvalues ​​of the normal operating data, as shown in the following formula: In the formula, is a standardized matrix; X is a normal operating data set; is an n-dimensional unit vector; u is the mean vector; D σ is the variance matrix; x and y are random variables in the sample data; S202: Select the first k eigenvalues ​​in the standardized matrix whose cumulative contribution rates are greater than the threshold, and use the matrix composed of their corresponding eigenvectors as the principal element matrix, as shown in the following formula: Where T is the principal component score matrix; is the residual score matrix; P is the principal component loading matrix; is the residual loading matrix; the eigenvalue of the covariance matrix after λ standardization; Cl is the contribution rate; The cumulative contribution rate threshold is set at 95%; S203: Project the sample data to be tested to the principal component and residual subspace of the principal component model, as shown in the following formula: In the formula, R p is the principal element subspace; R r is the residual subspace; is the projection of the sample in the principal element subspace; is the projection of the sample in the residual subspace; S204: Calculate the SPE and T of normal operation data at the confidence level α 2 The statistical control limits of In the formula, is the SPE control limit under the confidence level α; C α is the critical value of the normal distribution; F α (k,m-1) is the critical value of F distribution; The confidence level α was determined to be 0.95; S205: Calculate the SPE and T of the fault data to be tested 2 The statistical formula is as follows: In the formula, is the i-th column of the identity matrix I; is the statistical value of each column; S206: Comparison between SPE and T 2 The statistics and control limits are used to judge the fault. If SPE and T 2 If one of the statistical quantities exceeds the limit, it is determined that the pneumatic control valve has a fault; otherwise, it is determined that the start-up control valve has no fault and the diagnosis ends.

4. A method for fault diagnosis of a pneumatic control valve by coupling data drive and ontology model as claimed in claim 1, characterized in that: The specific steps of S3 are: S301: Collect historical fault data of pneumatic valves as a model training data set; S302: A pneumatic control valve fault classification model is established based on the SVM method. The classification model adopts a "one-to-one" multi-class classification strategy, and the kernel function adopts a cubic polynomial. The optimal classification decision function is shown in the following formula: K(x,x i )=((x,x i )+1) 3 Where sgn is the sign function; is the Grange multiplier; K(x,x i ) is the kernel function; b * is the classification threshold; The "one-to-one" multi-class classification strategy means that each SVM classifier uses only two types of pneumatic control valve fault sets in the training set as training sets for training. For the classification of k pneumatic control valve fault modes, it is necessary to establish SVM classifiers are used in the classification process of this method. The classification process adopts the "voting method", that is, each two-class SVM classifier will give its classification result, and the fault mode with the highest number of occurrences among all the classification results is finally counted as the fault mode of the test sample. S303: 70% of the model training data set is selected as a training set and 30% is selected as a test set by random allocation to train a pneumatic control valve fault classification model. When the accuracy of the training set is greater than 95% and the accuracy of the test set is greater than 90%, the model training is completed; S304: Input the fault data to be tested into the pneumatic control valve multi-class fault mode SVM classifier for calculation, output the indication value according to the decision function, and output the fault classification result.

5. A method for fault diagnosis of a pneumatic control valve coupled with data drive and ontology model as claimed in claim 1, characterized in that: The failure modes described in S3 can be divided into valve failure, pneumatic actuator failure, positioner failure and general failure; The valve faults include valve body blockage / valve stem jam (fault number F1), valve plug or valve seat sediment (fault number F2), valve plug or valve seat erosion (fault number F3), valve or bushing friction increase (fault number F4), external leakage (fault number F5), internal leakage (fault number F6), medium evaporation or critical flow (fault number F7); The pneumatic actuator failures include twisted pneumatic actuator valve stem (fault number F8), loose pneumatic actuator housing or terminals (fault number F9), diaphragm perforation of pneumatic actuator (fault number F10), and spring failure of pneumatic actuator (fault number F11); The positioner faults include electric-pneumatic converter fault (fault number F12), rod displacement sensor fault (fault number F13), pressure sensor fault (fault number F14), and positioner feedback fault (fault number F15); Common faults described include a drop in supply pressure to the positioner (fault number F16), unexpected pressure changes across the valve (fault number F17), a fully or partially open bypass valve (fault number F18), and a flow sensor failure (fault number F17).

6. A method for fault diagnosis of a pneumatic control valve coupled with data drive and ontology model as claimed in claim 1, characterized in that: The specific steps of the eight-step method described in S4 are: S401: Determine the scope of the pneumatic control valve fault diagnosis body, including fault mode number, fault type, fault cause, fault phenomenon, fault location and fault maintenance method; S402: Observe whether there is a reusable pneumatic control valve fault diagnosis body. If not, proceed to S403. If yes, skip S403 and S404, update the original body, and proceed to S405 after the update. S403: define a pneumatic control valve fault diagnosis ontology model, wherein the fault mode number is defined as an ontology of a pneumatic control valve fault classification number; the fault type is a specific classification of a pneumatic control valve fault; the fault cause is a cause that may cause a pneumatic control valve fault, which is classified according to multiple different parts of the pneumatic control valve; the fault phenomenon is certain special phenomena that the pneumatic control valve can exhibit under a certain fault mode of the pneumatic control valve; the fault location is the specific part of the pneumatic control valve where the fault occurs; the fault maintenance method is a maintenance method and suggestion decided by relevant management personnel for a certain pneumatic control valve fault mode; S404: Extract key ontology terms from pneumatic regulation fault diagnosis knowledge; S405: defining and establishing an ontology class and a class hierarchy relationship of a pneumatic control valve, wherein the class and the class hierarchy relationship include a subclass relationship and an instance relationship; S406: defining and establishing pneumatic control valve fault diagnosis ontology attributes, wherein the ontology attributes include functionality, inverse relationship, value range, and definition domain; S407: Create a pneumatic control valve fault diagnosis ontology instance and association; S408: Performing an ontology consistency check on the pneumatic control valve, wherein the consistency check is implemented by the HermiT inference engine. After the check passes, the pneumatic control valve fault diagnosis ontology knowledge base is successfully established.

7. A pneumatic control valve fault diagnosis system coupled with data drive and ontology model, characterized in that: The system is used to implement a pneumatic control valve fault diagnosis method coupling data drive and ontology model as described in any one of claims 1 to 5; The system includes a data management module, a pneumatic control valve fault detection module, a pneumatic control valve fault mode classification module, a pneumatic control valve fault diagnosis knowledge reasoning module and a pneumatic control valve fault diagnosis knowledge management module; The data management module is used to input the basic operation data of the target pneumatic control valve and output the pneumatic control valve fault detection data, classification data and knowledge reasoning log data; The pneumatic control valve fault detection module is used to train and establish a data-driven pneumatic control valve fault detection PCA model, perform fault detection on a given set of pneumatic control valve operation data, determine whether the pneumatic control valve has a fault, save the corresponding fault detection record, and output the fault data; The pneumatic control valve fault mode classification module is used to train and establish a data-driven pneumatic control valve fault mode classification SVM model, classify fault modes according to fault data, save corresponding fault mode classification records, and output fault mode classification numbers; The pneumatic control valve fault diagnosis knowledge reasoning module is used to construct a pneumatic control valve fault diagnosis ontology knowledge base, establish a semantic mapping between the fault mode classification number and the ontology instance, perform ontology reasoning according to the fault mode classification number, obtain the corresponding diagnostic conclusion, fault type, fault cause and corresponding maintenance suggestion under the pneumatic control valve fault mode, and output a pneumatic control valve fault diagnosis report; The pneumatic control valve fault diagnosis knowledge management module is used to implement the operations of adding, editing and deleting knowledge points in the pneumatic control valve fault diagnosis ontology knowledge base.

Images