Monitoring alarm method and system for air compressor system

Abstract

The present invention relates to a monitoring and alarm method and system for an air compressor system. The method comprises the following steps: obtaining historical data of monitoring variables of the air compressor system, and dividing the monitoring variables into oil system related variables, electrical system related variables and gas system related variables according to the subsystems to which the monitoring variables belong; performing process analysis on each subsystem to obtain an initial alarm propagation path of the monitoring variables of each subsystem; performing correlation analysis on the historical data of the monitoring variables based on a transfer entropy analysis algorithm, and correcting the initial alarm path according to the correlation analysis result to obtain a final alarm propagation path; constructing a principal component analysis model according to the monitoring variables when the air compressor is normal; judging whether the current system has a fault based on the monitoring data at the current moment according to the principal component analysis model, and if so, judging the fault monitoring variable, and giving an alarm prompt according to the final alarm propagation path.

Description

Technical Field

[0001] The present invention relates to the technical field of air compressor monitoring and alarm technology, and in particular to a monitoring and alarm method and system for an air compressor system. Background Art

[0002] Process monitoring ensures the normal operation of a process by detecting, identifying, and eliminating abnormal process characteristics or behaviors. Therefore, process monitoring has become a key measure for effectively ensuring the safe and stable operation of process industries. For typical continuous process industries, data from distributed control systems (DCSs) is currently used to monitor variable limits caused by actuator or sensor failures, control system component failures, or abnormal equipment status. While these systems provide anomaly alarms, they are based solely on whether input and output data exceed their limits. Momentary overshoots are eliminated by the control system. Furthermore, the closed-loop feedback loop between decision-making and control, coupled with the coupling between system units, creates causal relationships between input and output, leading to false alarms and difficulty in tracing anomalies. For discrete industrial processes, monitoring of the local operating status of key processing equipment primarily relies on single-source characteristic parameters. These monitoring systems ignore data correlation and causal relationships, resulting in unreliable results and an inability to comprehensively monitor the entire process and the quality of intermediate products. Data-driven intelligent process monitoring solutions, based on the collection and analysis of process operation data, avoid complex process reaction mechanisms, offer high flexibility and versatility, and have been widely adopted in process industry monitoring.

[0003] Alarm management has always been a key concern for process plant operators. Many petrochemical accidents worldwide are linked to improper alarm management. Modern petrochemical production plants and auxiliary facilities feature numerous alarm settings, and frequent alarm messages not only put pressure on operators but also lead to oversight of important alarms, resulting in unnecessary production losses and reduced safety. Therefore, developing an effective alarm strategy and implementing an alarm management system are key measures to standardize and improve the alarm system. Current DCS control systems place greater emphasis on alarm management, with specialized alarm management systems designed. These systems collect, record, analyze, manage, and optimize process and control system alarm information, helping operators identify and address issues arising in the production process and control systems promptly. Through functions such as alarm grouping, prioritization, alarm adjustment, alarm filtering, alarm suppression, and alarm information diagnostics and analysis, these systems maximize the effectiveness of the alarm system and enhance production safety. In a complex and ever-changing work environment, timely resolution of unexpected faults is paramount to improving work efficiency. In the actual factory production process, when multiple production lines are working at the same time, a fault in a certain section often affects the production process of the entire production line or even the entire factory, resulting in the inability to accurately locate the fault location, delaying the production cycle, affecting the progress of the entire production line or the entire production, and greatly reducing production efficiency. Therefore, in order to accurately find the location of the fault and reduce time waste, it becomes an important issue to solve the actual production problem for working technicians. Summary of the Invention

[0004] In view of the above analysis, the embodiments of the present invention aim to provide a monitoring and alarm method and system for an air compressor system, so as to solve the existing problems of being unable to accurately locate the fault location and having low alarm efficiency.

[0005] In one aspect, an embodiment of the present invention provides a monitoring and alarm method for an air compressor system, comprising the following steps:

[0006] Obtain historical data of air compressor system monitoring variables, and classify the monitoring variables into oil system-related variables, electrical system-related variables, and gas system-related variables according to the subsystem to which the monitoring variables belong;

[0007] Conduct process analysis on each subsystem to obtain the initial alarm propagation path of each subsystem's monitoring variables;

[0008] Based on the transfer entropy analysis algorithm, the historical data of the monitoring variables are analyzed for correlation, and the initial alarm path is modified according to the correlation analysis results to obtain the final alarm propagation path;

[0009] A principal component analysis model is constructed based on the monitoring variables when the air compressor is normal; for the monitoring data at the current moment, the principal component analysis model is used to determine whether the current system has a fault. If so, the fault monitoring variables are determined, and an alarm prompt is issued according to the final alarm propagation path.

[0010] Based on the further improvement of the above technical solution, the correlation analysis of the monitoring variables based on the transfer entropy analysis algorithm includes:

[0011] Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem;

[0012] Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

[0013] Furthermore, the transfer relationship of the monitored variables within the subsystem is calculated in the following way:

[0014] S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed;

[0015] S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable;

[0016] S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

[0017] Furthermore, the following formula is used to calculate the transfer relationship t between the monitoring variables X→Y :

[0018] t X→Y =t(x|y)-t(y|x)

[0019]

[0020]

[0021] Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, x i+1 Indicates the value of variable x at time i+1, y i+1 represents the value of variable y at time i+1, P(x i+1 ,xi ,y i ) represents x i+1 ,x i ,y i The joint probability, P(y i+1 ,y i ,x i ) represents y i+1 ,y i ,x i The joint probability, P(x i+1 |x i ) means that under the condition x i Next x i+1 The probability, P(y i+1 |y i ) means that under condition y i Next i+1 The probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

[0022] Furthermore, the principal component analysis model is constructed based on the monitoring variables when the air compressor is normal in the following way:

[0023] Standardize the sample data of monitoring variables and calculate the covariance matrix;

[0024] Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0025] According to the cumulative variance percentage, k eigenvalues ​​and corresponding eigenvectors are taken to construct the principal component analysis model.

[0026] Furthermore, for the monitoring data at the current moment, judging whether the current system has a fault according to the principal component analysis model includes:

[0027] Standardize the monitoring data at the current moment;

[0028] according to Calculate the squared prediction error SPE, where X ij represents the measured value of the jth monitored variable at the i-th moment, represents the principal component model prediction value of the jth monitoring variable at the i-th moment;

[0029] If the square prediction error SPE is greater than the first threshold, it is determined that there is an error in the current system.

[0030] On the other hand, an embodiment of the present invention provides a monitoring and alarm system for an air compressor system, comprising the following modules:

[0031] The monitoring variable acquisition module is used to obtain the historical data of the air compressor system monitoring variables, and classify the monitoring variables into oil system related variables, electrical system related variables and gas system related variables according to the subsystem to which the monitoring variables belong;

[0032] The initial path determination module is used to perform process analysis on each subsystem to obtain the initial alarm propagation path of the monitoring variables of each subsystem;

[0033] The final path determination module is used to perform correlation analysis on the historical data of the monitoring variables based on the transfer entropy analysis algorithm, and to modify the initial alarm path according to the correlation analysis results to obtain the final alarm propagation path;

[0034] The alarm module is used to build a principal component analysis model based on the monitoring variables when the air compressor is normal; for the monitoring data at the current moment, it is used to determine whether there is a fault in the current system based on the principal component analysis model. If so, the fault monitoring variables are determined, and an alarm prompt is issued according to the final alarm propagation path.

[0035] Based on the further improvement of the above technical solution, the final path determination module performs correlation analysis on the monitoring variables based on the transfer entropy analysis algorithm, including:

[0036] Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem;

[0037] Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

[0038] Furthermore, the final path determination module calculates the transfer relationship of the monitoring variables within the subsystem in the following way:

[0039] S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed;

[0040] S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable;

[0041] S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

[0042] Furthermore, the final path determination module uses the following formula to calculate the transfer relationship t between the monitoring variables X→Y :

[0043] t X→Y =t(x|y)-t(y|x)

[0044]

[0045]

[0046] Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, x i+1 Indicates the value of variable x at time i+1, y i+1 represents the value of variable y at time i+1, P(x i+1 ,x i ,y i ) represents x i+1 ,x i ,y i The joint probability, P(y i+1 ,y i ,x i ) represents y i+1 ,y i ,x i The joint probability, P(x i+1 |x i ) means that under the condition x i Next x i+1 The probability, P(y i+1 |y i ) means that under condition y i Next i+1 The probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

[0047] Compared with the existing technology, the present invention constructs an initial alarm propagation path by analyzing the air compressor system, and corrects the alarm propagation path according to the correlation analysis results of the monitoring variables, so as to obtain an alarm propagation path that is more complex during actual operation. Through the principal component analysis model, when a fault occurs in the system, an alarm can be issued in time, and a fault tracing prompt can be given according to the alarm propagation path, which is convenient for operators to promptly and accurately discover and locate the source of the fault, and repair the fault in time, thereby improving production efficiency.

[0048] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like parts throughout the drawings.

[0050] Figure 1 This is a flow chart of a monitoring and alarm method for an air compressor system according to an embodiment of the present invention;

[0051] Figure 2 This is a block diagram of a monitoring and alarm system for an air compressor system according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the air compressor mechanism according to an embodiment of the present invention;

[0053] Figure 4 Schematic diagram of the alarm path of the 1# air compressor according to an embodiment of the present invention;

[0054] Figure 5 Schematic diagram of an alarm path of an air compressor system according to an embodiment of the present invention;

[0055] Figure 6 Schematic diagram of the transmission relationship of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0057] A specific embodiment of the present invention discloses a monitoring and alarm method for an air compressor system, such as Figure 1 As shown, the following steps are included:

[0058] S1. Obtain historical data of monitoring variables of the air compressor system, and classify the monitoring variables into oil system-related variables, electrical system-related variables, and gas system-related variables according to the subsystem to which the monitoring variables belong;

[0059] S2. Perform process analysis on each subsystem to obtain the initial alarm propagation path of each subsystem monitoring variable;

[0060] S3. Perform correlation analysis on the historical data of the monitoring variables based on the transfer entropy analysis algorithm, and modify the initial alarm path according to the correlation analysis results to obtain the final alarm propagation path;

[0061] S4. Construct a principal component analysis model based on the monitoring variables when the air compressor is normal; for the monitoring data at the current moment, determine whether there is a fault in the current system based on the principal component analysis model. If so, determine the fault monitoring variables and issue an alarm prompt based on the final alarm propagation path.

[0062] The present invention constructs an initial alarm propagation path by analyzing the air compressor system, and corrects the alarm propagation path according to the correlation analysis results of the monitoring variables, so as to obtain an alarm propagation path that is more complex during actual operation. Through the principal component analysis model, when a fault occurs in the system, an alarm can be issued in time, and a fault tracing prompt can be given according to the alarm propagation path, so that the operator can find and locate the source of the fault in a timely and accurate manner, repair the fault in time, and improve production efficiency.

[0063] During implementation, historical and real-time data of air compressor system monitoring variables can be obtained through the DCS system.

[0064] The implementation is illustrated using the Chaoyu HD-DCS system at a refinery's three-compressor station. The station includes two air compressors (#1 and #3), a nitrogen compressor, and four drying systems. This system primarily monitors and analyzes the air compressors' alarms.

[0065] The three-stage air compressor system is composed of multiple devices connected together. The devices affect each other, and the internal variables of each device are also related. Figure 3 The air compressor structure mainly consists of three parts, namely the oil system, gas system and motor system of the air compressor, which correspond to Figure 3 The air compressor's oil machine equipment stores and uses lubricating oil, the compressor's three-section air compression piping, and the three-stage impeller bearings connecting the sections. These serve as the basis for classification and modeling of the compressor's internal variable subsystems. To facilitate accurate alarm path analysis, the air compressor subsystem is first divided into three subsystems: the oil system, the electrical system, and the gas system. The monitored variables are then divided into variables related to the oil system, the electrical system, and the gas system. The main monitored variables for the air compressor are shown in Table 1. Variables related to the oil system include 3, 4, 5, 7, 8, and 21; variables related to the electrical system include 11, 12, 13, 14, 15, 16, 17, 18, 19, and 20; and variables related to the gas system include 1, 2, 6, 9, 10, 11, 12, 13, and 22.

[0066] Table 1 Some monitoring variables of air compressor

[0067] 1 PT-101 / 1 1# air compressor system pressure (inlet) 2 PT-102 / 1 1# air compressor exhaust pressure (outlet) 3 PT-105 / 1 1# air compressor oil pressure (is it currently in normal condition) 4 PT-106 / 1 1# air compressor bearing oil pressure (is it currently in normal condition) 5 PT-108 / 1 1# air compressor filter front oil pressure (is it currently normal) 6 PT-113 / 1 1# air compressor inlet air filter differential pressure (is it currently in normal state) 7 PDI-108 / 1 Calculation of differential pressure of oil filter of No. 1 air compressor 8 TE-101 / 1 1# air compressor oil temperature (is it currently in normal state) 9 TE-103 / 1 1# air compressor second stage inlet temperature (is it currently in normal state) 10 TE-104 / 1 1# air compressor stage 3 inlet temperature (is it currently in normal state) 11 VT-110 / 1 1# air compressor first stage vibration (2 out of 3 to determine vibration) 12 VT-111 / 1 1# air compressor second stage vibration (2 out of 3 to determine vibration) 13 VT-112 / 1 1# air compressor three-stage vibration (2 out of 3 to determine vibration) 14 YS-108 / 1 1# air compressor motor operation signal (start, stop) 0-1 15 IE-114 / 1 1# air compressor drive motor current 16 IY-114 / 1 1# air compressor surge control point (algorithm output point) 17 QY-114 / 1 1# air compressor surge count 18 UY-114 / 1 1# air compressor surge monitoring motor (current) 19 HS-112 / 1 1# air compressor emergency stop (manual button) 0-1 20 EI-100 / 1 1# air compressor voltage module 380v 21 LSL-102 / 1 1# air compressor oil level switch 0-1 22 HV3011 Filter outlet butterfly valve switch

[0068] The variables related to the occurrence of air compressor surge failure include continuous variables 1, 2, 6, 11, 12, 13 and discrete variables 16 and 17; the variables related to the temperature change of the air compressor include continuous variables 3, 4, 5, 8, 9, 10 and discrete variables 14 and 21;

[0069] The variables related to the pressure change inside the air compressor are continuous variables 1, 2, 6, 7 and discrete variable 19. The initial alarm propagation paths of the monitoring variables of the three subsystems inside the air compressor obtained based on the process experience knowledge analysis are as follows:

[0070] Air compressor oil system: 21→3→8→5→7

[0071] Air compressor electromechanical system: 11→12→13→18→16→17; 19→20→14→15

[0072] Air compressor system: 6→11→12→13; 6→1→2

[0073] After the initial alarm propagation path is established, the correlation analysis of the monitoring variables is performed through transfer entropy to correct the initial alarm path.

[0074] Specifically, the correlation analysis of monitoring variables based on the transfer entropy analysis algorithm includes:

[0075] Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem;

[0076] Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

[0077] That is, correlation analysis is performed within the subsystem and between subsystems.

[0078] Specifically, the monitoring variables within the subsystem are calculated using the following method to calculate the transfer relationship of the variables:

[0079] S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed;

[0080] S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable;

[0081] S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

[0082] For the monitoring variables between each subsystem, the transfer relationship is calculated pairwise.

[0083] Transfer entropy is a method for deriving causality between time series based on probability distributions and Shannon entropy statistics. It indicates how much the uncertainty of the current state of Y is reduced by knowing the history of the source variable X and the target variable Y, compared to the reduction in uncertainty of Y simply from the history of Y. This information is asymmetric between the flow of information from Y to X and from X to Y, and this asymmetry establishes a driver-response relationship.

[0084] Specifically, the following formula is used to calculate the transfer entropy between monitoring variables:

[0085]

[0086]

[0087] According to t X→Y =t(x|y)-t(y|x) calculates the transfer relationship between monitoring variables;

[0088] Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, P(x i+1 ,x i ,y i ) represents x i+1 ,x i ,y i The joint probability, P(x i+1 |x i ) represents the conditional probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

[0089] According to the above calculation formula, the transfer relationship between the continuous variables in each monitoring variable within the subsystem and between subsystems is calculated respectively. If |t X→Y If | is less than or equal to the second threshold, it is considered that there is no transmission relationship between variables x and y. If there is a direct connection between x and y in the initial alarm propagation path, the connection is deleted. X→Y | is greater than the second threshold, if t X→Y >0, indicating that information is transferred from variable x to y. If t X→Y<0, indicating that the information is transferred from variable y to x. If there is no connection between variables x and y in the initial alarm propagation path, then add the association relationship. If there is a connection with the wrong direction, then correct the connection direction.

[0090] When calculating the transfer entropy relationship between variables, these transfer causal relationships can be expressed using chain propagation graphics. For example, for multivariate time variables x, y, z, there may be loops. Based on the size of the transfer relationship value, the path with the smallest transfer relationship is removed. For example, Figure 6 (a) shows the loop, if t X→Y and t Y→Z Both are greater than t X→Z , then the transitive relationship is modified to Figure 6 (b).

[0091] In addition to calculating the transfer entropy for continuous variables, it's also necessary to consider the impact of discrete variables on fault path propagation. In actual industrial processes, discrete variables often serve as the starting or end points of process systems. Discrete variables include alarm data such as switch variables, alarm points, and alarm count statistics obtained from the DCS system. Specifically, in the air compressor process, multiple switch variables are included. Based on process knowledge, the impact of these discrete variables on continuous variables is evaluated and added to the alarm path diagram. For example, discrete variable 22 (filter outlet butterfly valve switch) is evaluated based on process knowledge to have an impact on variable 6 (compressor inlet air filter differential pressure), so it is added to the alarm path diagram.

[0092] Construct the final alarm path propagation diagram of the entire 1# air compressor subsystem as follows Figure 4As shown in Figure 4, the alarm group for air compressor #1 consists of variables and the alarm logic propagation relationships between them. The oil level switch variable 21 controls signal propagation throughout the compressor oil system, directly affecting the compressor's oil pressure (3). This influences the compressor oil temperature (8), the oil pressure before the compressor filter (5), the oil filter differential pressure calculation (7), and the bearing oil pressure (4). Variable #4 affects the compressor's second-stage inlet temperature (9) in the air system. The filter outlet butterfly valve from the air filter system regulates the compressor's inlet filter differential pressure (6), which in turn affects the compressor's system pressure (1) and the compressor's outlet exhaust pressure (2). The compressor's inlet differential pressure (6) also affects the compressor's second-stage inlet temperature (9) and third-stage inlet temperature (10). Together with the compressor's outlet pressure, it controls the compressor's cooler inlet and outlet temperatures in the compressed air cooling system. When vibration is factored in, the inlet differential pressure (6) also directly affects the compressor's first-stage vibration (11), second-stage vibration (12), and third-stage vibration (13). Vibration can affect the system pressure (1). Finally, the compressor's attached surge detection motor unit, the surge monitoring motor 18, and the surge control point 16 influence the surge count 17. The root variable for surge count monitoring is the compressor's outlet pressure 2, whose changes directly reflect whether the compressor is experiencing a surge fault.

[0093] According to this method, the causal relationship propagation path diagram of other equipment can also be obtained. The propagation path subgraphs of each equipment are linked together according to process knowledge, which reflects the relationship between the equipment. The fault propagation path diagram of the total system is as follows: Figure 5 shown.

[0094] When implemented, the fault propagation path diagram can be displayed on the DCS system.

[0095] After obtaining the fault propagation path diagram, the principal component analysis model is used to determine whether the current system has a fault. When a fault occurs, an estimated traceback reminder is performed.

[0096] Principal component analysis (PCA), also known as principal component analysis (PCA), is a commonly used method in multivariate statistical analysis. Its basic approach is to prioritize the direction of change based on the variance of the data. Principal components are then derived in this order, and these principal components are independent of each other. This tool can be used to refine information about changes and reduce the complexity of data analysis. It can project multivariate-dependent, high-dimensional data into an independent, low-dimensional data space, significantly reducing the difficulty of directly analyzing multidimensional, complex process variables. It can be used for data simplification, data denoising, data compression, modeling, singular value detection, and variable selection.

[0097] Specifically, the principal component analysis model is constructed based on the monitoring variables when the air compressor is normal in the following way:

[0098] S41, standardize the monitoring variable sample data and calculate the covariance matrix;

[0099] During implementation, the data of various monitoring variables stored in the DCS of the 1# air compressor in the previous month can be obtained and standardized preprocessing can be performed on the monitoring variable data. During implementation, invalid data can be eliminated. For example, if there is no data for a variable at a certain moment, it is considered invalid data and the data at that moment is eliminated. For example, there are a total of m monitoring variables, and data of n sampling points are extracted. The sample data is X n×m .

[0100] It should be noted that before calculating the covariance matrix, the data of each monitoring variable is first standardized, that is, the value of each monitoring variable is subtracted from the mean of the monitoring variable. The data after such processing has a mean of 0, and the data after standardization is

[0101] Calculate the normalized data The covariance matrix of

[0102] S42, calculating the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0103] Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, and sort the eigenvalues ​​and corresponding eigenvectors according to the size of the eigenvalues.

[0104] S43. Take k eigenvalues ​​and corresponding eigenvectors according to the cumulative variance percentage to construct a principal component analysis model.

[0105] The number of principal components is determined based on the cumulative variance percentage. Specifically, according to the formula Calculate the cumulative variance percentage. If the cumulative variance percentage of the first k eigenvalues ​​is greater than a third threshold, such as 90%, then take the first k eigenvalues ​​and their corresponding eigenvectors to construct a principal component analysis model. Among them, p l represents the i-th eigenvector, X i Represents the monitoring data at the i-th moment.

[0106] After the principal component analysis model is established, whether the current system has a fault is determined based on the monitoring variable values ​​at the current moment.

[0107] Specifically, for the monitoring data at the current moment, judging whether the current system has a fault according to the principal component analysis model includes:

[0108] The monitoring data at the current moment is normalized; that is, the mean value of each monitoring variable is subtracted from the value of the monitoring variable at the current moment.

[0109] according to Calculate the squared prediction error SPE, where X ijrepresents the measured value of the jth monitored variable at the i-th moment, represents the principal component model prediction value of the jth monitoring variable at the i-th moment;

[0110] in, According to the formula Calculated.

[0111] If the square prediction error SPE is greater than the first threshold, it is determined that there is an error in the current system.

[0112] During implementation, the first threshold may be determined according to system monitoring accuracy.

[0113] According to the squared prediction error SPE, it can be determined whether the system currently has a fault. If a fault exists, it is necessary to further determine which or which monitoring variables have a fault.

[0114] When implementing, according to The contribution of each monitored variable to the fault is calculated, and variables with a contribution greater than a fourth threshold are considered to be fault variables.

[0115] After obtaining the fault variables, the fault source can be traced and an alarm reminder can be issued according to the final alarm path diagram.

[0116] A specific embodiment of the present invention discloses a monitoring and alarm system for an air compressor system, such as Figure 2 As shown, it includes the following modules:

[0117] The monitoring variable acquisition module is used to obtain the historical data of the air compressor system monitoring variables, and classify the monitoring variables into oil system related variables, electrical system related variables and gas system related variables according to the subsystem to which the monitoring variables belong;

[0118] The initial path determination module is used to perform process analysis on each subsystem to obtain the initial alarm propagation path of the monitoring variables of each subsystem;

[0119] The final path determination module is used to perform correlation analysis on the historical data of the monitoring variables based on the transfer entropy analysis algorithm, and to modify the initial alarm path according to the correlation analysis results to obtain the final alarm propagation path;

[0120] The alarm module is used to build a principal component analysis model based on the monitoring variables when the air compressor is normal; for the monitoring data at the current moment, it is used to determine whether there is a fault in the current system based on the principal component analysis model. If so, the fault monitoring variables are determined, and an alarm prompt is issued according to the final alarm propagation path.

[0121] Preferably, the final path determination module performs correlation analysis on the monitoring variables based on the transfer entropy analysis algorithm, including:

[0122] Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem;

[0123] Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

[0124] Preferably, the final path determination module calculates the transfer relationship of the monitoring variables within the subsystem in the following manner:

[0125] S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed;

[0126] S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable;

[0127] S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

[0128] Preferably, the final path determination module uses the following formula to calculate the transfer relationship t between the monitoring variables X→Y :

[0129] t X→Y =t(x|y)-t(y|x)

[0130]

[0131]

[0132] Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, x i+1 Indicates the value of variable x at time i+1, y i+1 represents the value of variable y at time i+1, P(x i+1 ,x i ,y i ) represents x i+1 ,x i ,y i The joint probability, P(y i+1 ,y i ,x i ) represents y i+1 ,y i ,x iThe joint probability, P(x i+1 |x i ) means that under the condition x i Next x i+1 The probability, P(y i+1 |y i ) means that under condition y i Next i+1 The probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

[0133] The above method embodiment and system embodiment are based on the same principle, and their related parts can be used for reference, and can achieve the same technical effect. The specific implementation process can be found in the above embodiment and will not be repeated here.

[0134] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0135] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A monitoring and alarm method for an air compressor system, characterized in that: The following steps are involved: Obtain historical data of air compressor system monitoring variables, and classify the monitoring variables into oil system-related variables, electrical system-related variables, and gas system-related variables according to the subsystem to which the monitoring variables belong; Conduct process analysis on each subsystem to obtain the initial alarm propagation path of each subsystem's monitoring variables; Based on the transfer entropy analysis algorithm, the historical data of the monitoring variables are subjected to correlation analysis. According to the correlation analysis results, the initial alarm propagation path is modified to obtain the final alarm propagation path. A principal component analysis model is constructed based on the monitoring variables when the air compressor is normal; based on the monitoring data at the current moment, the principal component analysis model is used to determine whether the current system has a fault. If so, the fault monitoring variables are determined, and an alarm prompt is issued based on the final alarm propagation path; Correlation analysis of monitoring variables based on transfer entropy analysis algorithm includes: Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem; Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

2. The monitoring and alarm method for an air compressor system according to claim 1, characterized in that: The transfer relationship of the monitoring variables within the subsystem is calculated in the following way: S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed; S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable; S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

3. The monitoring and alarm method for an air compressor system according to claim 1, characterized in that: The following formula is used to calculate the transfer relationship t between monitoring variables X→Y : t X→Y =t(x|y)-t(y|x) Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, x i+1 Indicates the value of variable x at time i+1, y i+1 represents the value of variable y at time i+1, P(x i+1 ,x i ,y i ) represents x i+1 ,x i ,y i The joint probability, P(y i+1 ,y i ,x i ) represents y i+1 ,y i ,x i The joint probability, P(x i+1 |x i ) means that under the condition x i Next x i+1 The probability, P(y i+1 |y i ) means that under condition y i Next i+1 The probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

4. The monitoring and alarm method for an air compressor system according to claim 1, characterized in that: The principal component analysis model is constructed based on the monitoring variables when the air compressor is normal in the following way: Standardize the sample data of monitoring variables and calculate the covariance matrix; Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; According to the cumulative variance percentage, k eigenvalues ​​and corresponding eigenvectors are taken to construct the principal component analysis model.

5. The monitoring and alarm method for an air compressor system according to claim 1, characterized in that: For the monitoring data at the current moment, judging whether the current system has a fault according to the principal component analysis model includes: Standardize the monitoring data at the current moment; according to Calculate the squared prediction error SPE, where X ij represents the measured value of the jth monitored variable at the i-th moment, represents the principal component model prediction value of the jth monitoring variable at the i-th moment; If the square prediction error SPE is greater than the first threshold, it is determined that there is an error in the current system.

6. A monitoring and alarm system for an air compressor system, characterized in that: Includes the following modules: The monitoring variable acquisition module is used to obtain the historical data of the air compressor system monitoring variables, and classify the monitoring variables into oil system related variables, electrical system related variables and gas system related variables according to the subsystem to which the monitoring variables belong; The initial path determination module is used to perform process analysis on each subsystem to obtain the initial alarm propagation path of the monitoring variables of each subsystem; The final path determination module is used to perform correlation analysis on the historical data of the monitoring variables based on the transfer entropy analysis algorithm, and to modify the initial alarm propagation path according to the correlation analysis results to obtain the final alarm propagation path; The alarm module is used to build a principal component analysis model based on the monitoring variables when the air compressor is normal; based on the monitoring data at the current moment, it determines whether the current system has a fault based on the principal component analysis model. If so, it determines the fault monitoring variables and issues an alarm prompt based on the final alarm propagation path; The final path determination module performs correlation analysis on the monitoring variables based on the transfer entropy analysis algorithm, including: Based on the transfer entropy analysis algorithm, the transfer relationship of the continuous variables in the monitoring variables within each subsystem is calculated to obtain the correlation relationship of the monitoring variables within each subsystem; Based on the transfer entropy analysis algorithm, the transfer relationship of continuous variables in the monitoring variables between subsystems is calculated to obtain the correlation relationship between the monitoring variables between subsystems.

7. The monitoring and alarm system for an air compressor system according to claim 6, characterized in that: The final path determination module calculates the transfer relationship of the monitoring variables within the subsystem in the following way: S301, sorting the monitoring variables in order from the inlet to the outlet of the air compressor according to the process position to obtain a monitoring variable sequence; taking the first monitoring variable in the monitoring variable sequence as the variable to be analyzed; S302, sequentially taking the next monitoring variable to the last monitoring variable of the variable to be analyzed in the monitoring variable sequence as the current variable, and calculating the transfer relationship between the variable to be analyzed and the current variable; S303: If the variable to be analyzed is the last variable in the monitoring variable sequence, the analysis ends; otherwise, the next monitoring variable after the variable to be analyzed in the monitoring variable sequence is used as the variable to be analyzed, and the process returns to step S302.

8. The monitoring and alarm system for an air compressor system according to claim 6, characterized in that: The final path determination module uses the following formula to calculate the transfer relationship t between monitoring variables X→Y : t X→Y =t(x|y)-t(y|x) Among them, x i Indicates the value of variable x at time i, y i represents the value of variable y at time i, x i+1 Indicates the value of variable x at time i+1, y i+1 represents the value of variable y at time i+1, P(x i+1 ,x i ,y i ) represents x i+1 ,x i ,y i The joint probability, P(y i+1 ,y i ,x i ) represents y i+1 ,y i ,x i The joint probability, P(x i+1 |x i ) means that under the condition x i Next x i+1 The probability, P(y i+1 |y i ) means that under condition y i Next i+1 The probability, P(x i ,y i ) represents x i ,y i The joint probability of n is the number of sampling points.

Images