Fault Diagnosis Method, System, Device and Storage Medium Based on Qualitative Trend Analysis and Five-State Bayesian Network

Abstract

The present invention discloses a fault diagnosis method, system, device and storage medium based on qualitative trend analysis and five-state Bayesian network, aiming to solve the defective technical problem in the prior art that the fault diagnosis method only has two conditions of normal and abnormal, and the two states cannot describe in detail the variable change trend and the influence relationship of positive and negative effects between variables. The present invention conducts fault reasoning and diagnosis in a multi-variable time series system. First, multi-variable time window data and two linear segments are defined and the optimal segmentation point is obtained. According to the optimal segmentation point, the two linear segments are solved. Then, the latter linear segment is described by five states of normal, rising, falling, positive step and negative step. A causal relationship network is constructed according to the causal relationship of variables. The qualitative trend analysis method is used to judge the state of each variable in real time, and the five-state Bayesian network is used for reasoning and diagnosis to judge the root cause of the current system failure and the fault propagation path.

Description

Technical Field

[0001] The present invention belongs to the field of fault diagnosis, and relates to a fault diagnosis method, system, device and storage medium based on qualitative trend analysis and five-state Bayesian network. Background Art

[0002] When a multi-variable time series system fails, multiple variables often deviate abnormally. It is very difficult for staff to quickly judge the root cause of the failure based on experience. The purpose of fault reasoning and diagnosis is to infer and diagnose the root cause of the system failure according to the change trends of abnormal variables, and give the fault propagation path. The results of fault reasoning and diagnosis have important guiding significance for staff to understand the system state and fault types, and can also assist staff in manually adjusting the operating variables according to the identification and diagnosis results to make the system adjust to its normal state as soon as possible and avoid more serious accidents.

[0003] Existing fault reasoning and diagnosis methods usually only judge variables as normal and abnormal variables, and the number of states of each variable is 2, which cannot reflect the change trends of each variable. The change trends of variables also play an important role for staff to understand the system state.

[0004] Other Bayesian network fault diagnosis methods usually only have 2 states (normal, abnormal). In actual situations, only two states cannot describe in detail the change trends of variables and the influence relationship of positive and negative effects between variables. For example, for a positive influence relationship, when the cause variable rises, the result variable rises accordingly, and when the cause variable falls, the result variable falls accordingly; for a negative influence relationship, when the cause variable rises, the result variable falls accordingly, and when the cause variable falls, the result variable rises accordingly. Summary of the Invention

[0005] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide a fault diagnosis method, system, device and storage medium based on qualitative trend analysis and five-state Bayesian network, aiming to solve the defective technical problems in the prior art that the fault diagnosis method only has two conditions of normal and abnormal, and only two states cannot describe in detail the change trends of variables and the influence relationship of positive and negative effects between variables.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] The present invention proposes a fault diagnosis method based on qualitative trend analysis and five-state Bayesian network, including:

[0008] Sorting multi-variable time series data into multi-variable time window data;

[0009] Finding the optimal segmentation point through multivariate time-window data to obtain two linear segments, and describing the linear segments with five trend states;

[0010] Constructing a variable causal relationship network, and based on the variable causal relationship network, constructing a Bayesian network with five states, and calculating the conditional probability information between node states in the Bayesian network with five states;

[0011] Using the conditional probability information to diagnose the actual trend state of the variable, and obtaining the cause of the fault and the fault propagation path.

[0012] Preferably, the multivariate time-series data has two dimensions: one variable dimension, representing each variable; one time dimension, representing each sampling point;

[0013] Organizing the multivariate time-series data into multivariate time-window data with a fixed time length, where L is from 10 min to 10 h.

[0014] Preferably, the expressions of the two linear segments are shown in Formulas (1) and (2):

[0015]

[0016]

[0017] where i represents the sampling moment within the multivariate time-window data, and respectively represent the start moments of the two linear segments; and respectively represent the measured values of the variable in the two linear segments at and moments; p1 and p2 are the slopes of the two linear segments;

[0018] Using the least squares method for fitting to obtain the slopes p1 and p2 of the two linear segments, and the calculation methods of the slopes p1 and p2 of the two linear segments are shown in Formulas (3) and (4):

[0019]

[0020]

[0021] where t k represents the segmentation point of the two linear segments, t k = 5, 6, …, L – 5; 1 to t k is the previous segment, and t k + 1 to L is the latter segment; and respectively represent the average values of the sampling moments within two linear segments, and respectively represent the average values of the measured values within two linear segments.

[0022] Preferably, the traversal method is adopted to find the segmentation point The segmentation point The calculation formula is as shown in formula (5):

[0023]

[0024] Take the sum of the absolute values of the differences between the predicted value and and the measured values y1 and y2 as the fitting error, and find the minimum fitting error as the segmentation point between the two linear segments

[0025] Preferably, the latter segment is described by normal, rising, falling, positive step, and negative step;

[0026] When |p| ≤ p th and |I d | > h tc and i d > 0, the latter segment is a positive step;

[0027] When |p| ≤ p th and |I d | > h tc and I d < 0, the latter segment is a negative step;

[0028] When |p| ≤ p th and |I d | ≤ h tc , the latter segment remains unchanged;

[0029] When |p| > p th and p > 0, the latter segment is rising;

[0030] When |p| > p th and p < 0, the latter segment is falling;

[0031] Among them, I d represents the difference at the connection point of the two adjacent linear segments; p th is used to compare with the change slope p of the variable linear segment to judge whether the trend of the linear segment remains unchanged; h tc is used to compare with the difference I d at the connection point of the two adjacent linear segments of the variable to judge whether there is a step change when the trend of the latter linear segment remains unchanged.

[0032] Preferably, in the variable causality network, each node represents a variable, and the causal relationship between variables is represented by a one-way arrow. The cause node points to the result node, indicating that an abnormality in the cause variable may lead to an abnormality in the result variable;

[0033] According to the variable causality network, a five-state Bayesian network is constructed, and the flow relationship of variable information in the system is represented by conditional probability, and its definition is shown in formula (6):

[0034] G=(V,E,Ψ) (6)

[0035] Calculate the conditional probability information between node states in the five-state Bayesian network: According to the change trend states of each variable, the maximum likelihood estimation method is used to calculate the conditional probability information between node states in the five-state Bayesian network.

[0036] Preferably, use the conditional probability information to diagnose the actual trend state of the variable:

[0037] Update the trend state of each variable into the constructed five-state Bayesian network;

[0038] According to the actual trend state of each variable, perform inference diagnosis through the Bayesian formula, as shown in formula (7):

[0039]

[0040] Bayesian inference starts from the abnormal variable with the largest real-time deviation degree, sets this variable to the corresponding abnormal state as 100%, and calculates the state probabilities of its cause variables;

[0041] Find the variable that best matches the actual abnormal state and has the largest abnormal probability as its abnormal cause;

[0042] Successively search for the abnormal nodes with faults step by step from downstream to upstream along the causal relationship until it is found that the current node has no parent node, or all cause nodes upstream of the current node are normal;

[0043] The abnormal variable where the search finally stops is the root cause variable that causes the system failure, and the propagation path of the fault from upstream to downstream is given.

[0044] A system for a fault diagnosis method based on qualitative trend analysis and five-state Bayesian network proposed by the present invention includes:

[0045] A multi-variable time window data acquisition module, which is used to organize multi-variable time series data into multi-variable time window data;

[0046] A linear segment acquisition module, which is used to find the optimal segmentation point through multivariate time window data to obtain two linear segments, and describe the linear segments with five trend states;

[0047] A Bayesian network construction module, which is used to construct a variable causal relationship network, and construct a Bayesian network with five states according to the variable causal relationship network, and calculate the conditional probability information between the node states in the Bayesian network with five states;

[0048] A fault cause and fault propagation path acquisition module, which is used to diagnose the actual trend state of the variable by using the conditional probability information to obtain the fault cause and fault propagation path.

[0049] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network are implemented.

[0050] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network are implemented.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention provides a fault diagnosis method based on qualitative trend analysis and five-state Bayesian network. By defining two linear segments, the optimal segmentation point of the two linear segments is calculated in combination with multivariate time window data. Through the optimal segmentation point, the two linear segments can be solved, and then the latter linear segment obtained by the solution is described with 5 change trends. The 5 change trends are sent into the constructed Bayesian network, and the conditional probability information between the node states in the Bayesian network is used to judge the cause of the fault and the propagation path. The present invention proposes a five-state Bayesian network method, combines the trend state information of each variable, and performs real-time inference diagnosis to finally find the root cause of the current system failure and the fault propagation path.

[0053] Further, in the multivariate time window data, the change trend in the latter part of the time is more important for the fault diagnosis of the current system. Therefore, we divide the data into two linear segments for fitting, so as to extract the more accurate change trend state in the latter part of the time.

[0054] The system of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network proposed by the present invention adopts a modular idea to realize the causes of faults and the fault propagation paths, making each module independent of each other, which is convenient for the unified management of each module. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is the flowchart of the fault diagnosis method of the present invention;

[0056] Figure 2 It is the flowchart of fault diagnosis modeling and online diagnosis of the present invention;

[0057] Figure 3 It is five trend charts for describing linear segments of the present invention;

[0058] Figure 4 It is the change trend analysis chart of the linear segment of the present invention;

[0059] Figure 5 It is the network diagram for constructing variable causal relationships of the present invention;

[0060] Figure 6 It is the fault diagnosis result chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0062] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0063] The following further describes the present invention in detail with reference to the accompanying drawings:

[0064] Our method consists of three parts: (1) A two-stage qualitative trend analysis method is proposed to judge the state of each variable in real time, which is described by five states (normal, rising, falling, positive step, negative step); (2) A causal relationship network is constructed according to the causal relationship between variables; (3) A five-state Bayesian network is proposed for reasoning and diagnosis to judge the root cause of the current system failure and the fault propagation path.

[0065] A fault reasoning and diagnosis method based on qualitative trend analysis and five-state Bayesian network proposed by the present invention, as Figure 1 shown, includes:

[0066] Organize multi-variable time series data into multi-variable time window data;

[0067] Find the optimal segmentation point through multi-variable time window data to obtain two linear segments, and describe the linear segments with five trend states;

[0068] Construct a variable causal relationship network, and according to the variable causal relationship network, construct a five-state Bayesian network, and calculate the conditional probability information between node states in the five-state Bayesian network;

[0069] Use the conditional probability information to diagnose the actual trend state of the variable, and obtain the cause of the fault and the fault propagation path.

[0070] A fault diagnosis method based on qualitative trend analysis and five-state Bayesian network, as Figure 2 shown, the specific implementation steps are as follows:

[0071] Step S01: Offline modeling stage. For a multi-variable time series system, collect multi-variable time series data for a period of time from the historical database as the historical data used in the offline modeling stage. The historical data has two dimensions, one is the variable dimension, representing each variable; the other is the time dimension, representing each sampling point.

[0072] Step S02: Organize the historical data into multi-variable time window data with a fixed time length (L = 10 minutes - 10 hours).

[0073] Use the two-stage qualitative trend analysis method to determine the trend state of each variable, as shown in steps S03 - S05.

[0074] Step S03: The two-stage qualitative trend analysis method needs to use two linear segments to fit the time window data. Define the linear segments with the following formulas. The expressions of the two linear segments are shown in formulas (1) and (2):

[0075]

[0076]

[0077] Among them, \(i\) represents the sampling moment within the multi-variable time window data. and respectively represent the start moments of two linear segments; and respectively represent the measured values of the variable in two linear segments at and moments; \(p1\) and \(p2\) are the slopes of two linear segments.

[0078] The least squares method is used for fitting to obtain the slopes \(p1\) and \(p2\) of two linear segments. The calculation methods of the slopes \(p1\) and \(p2\) of two linear segments are shown in formulas (3) and (4):

[0079]

[0080]

[0081] Among them, \(t\) k represents the segmentation point of two linear segments, \(t\) k = 5, 6, …, L–5; 1 to \(t\) k is the previous segment, \(t\) k +1 to L is the latter segment; and respectively represent the average values of the sampling moments within two linear segments, and respectively represent the average values of the measured values within two linear segments.

[0082] Preferably, in the multi-variable time window data, the change trend of the latter part of the time is more important for the fault diagnosis of the current system. Therefore, we divide the data into two linear segments for fitting, so as to extract the change trend state of the more accurate latter part of the time.

[0083] Step S04: Trend extraction uses the traversal method to find the segmentation point Two linear segments are respectively fitted with two unary linear equations \(y1(i)\) and \(y2(i)\) by the least squares method. The calculation formula of the segmentation point is shown in formula (5):

[0084]

[0085] The sum of the absolute values of the differences between the predicted values and and the measured values \(y1\) and \(y2\) is used as the fitting error, and the minimum fitting error is found as the segmentation point of two linear segments

[0086] Step S05: After obtaining two linear segments, the latter segment needs to be described using normal, rising, falling, positive step, and negative step, as Figure 3 shown.

[0087] When |p| ≤ p th and |I d | > h tc and I d > 0, the latter segment is a positive step;

[0088] When |p| ≤ p th and |I d | > h tc and I d < 0, the latter segment is a negative step;

[0089] When |p| ≤ p th and |I d | ≤ h tc , the latter segment remains unchanged;

[0090] When |p| > p th and p > 0, the latter segment is rising;

[0091] When |p| > p th and p < 0, the latter segment is falling;

[0092] Among them, I d represents the difference at the connection point of the front and back two linear segments; p th is used to compare with the change slope p of the variable linear segment to judge whether the trend of the linear segment remains unchanged; h tc is used to compare with the difference I d at the connection point of the front and back two variable linear segments to judge whether there is a step change when the trend of the latter linear segment remains unchanged. The trend of the variable is classified using the Figure 4 shown rules.

[0093] Step S06: Construct a variable causal relationship network based on expert experience knowledge. The complex process-level flow is decomposed into multiple simple equipment-level units in a modular manner, and then for each variable node, analyze which other variables may be abnormal when it is abnormal. A variable causal relationship network is as Figure 5 shown, where each node represents a variable, and a unidirectional arrow represents the causal relationship between variables. The cause node points to the result node, indicating that the abnormality of the cause variable may lead to the abnormality of the result variable.

[0094] Step S07: Construct a five-state Bayesian network G according to the variable causal relationship network in Step S06. The trend states of each variable node in this network include five types: normal, rising, falling, positive step, and negative step. The Bayesian network model G takes the variable V as nodes and the causal relationship between variables E as edges. Each variable has five states Ψ ∈ {normal, rising, falling, positive step, negative step}, and represents the information flow relationship of variables in a complex system through conditional probability, as defined in formula (6):

[0095] G = (V, E, Ψ) (6)

[0096] Step S08: Use the change trend states of each variable in historical data and adopt the maximum likelihood estimation method to calculate the conditional probability information between node states in the five-state Bayesian network. Since the node values are discrete values, the maximum likelihood estimation can be divided into two steps:

[0097] (1) Count the frequency of each case appearing in the conditional probability table. For example, for a case (y = "rising" | x = "rising") in a variable relationship x → y, it is necessary to count the number of times x = "rising" in the data and the number of times y = "rising" when x = "rising";

[0098] (2) Convert the counted frequency into probability. For example, for a case (y = "rising" | x = "rising") in a variable relationship x → y, its probability parameter can be calculated according to the formula.

[0099]

[0100] In the real-time application stage, first organize the real-time data into multi-variable time window data with a fixed time length (L = 10 min - 10 h), and determine the trend state of each variable by using the two-stage qualitative trend analysis method in Steps S02 - S05 in the same way.

[0101] Update the trend state of each variable to the constructed five-state Bayesian network. According to the actual trend state of each variable, use the conditional probability information obtained in Step S08 and perform inference diagnosis through the Bayesian formula:

[0102] (1) Update the trend state of each variable to the constructed five-state Bayesian network;

[0103] (2) According to the actual trend state of each variable, perform inference diagnosis through the Bayesian formula. For example, for a variable relationship x → y, to calculate the probability that x = "rising" when y = "rising", it can be calculated according to the following Bayesian formula, as shown in formula (8):

[0104]

[0105] Bayesian inference starts from the abnormal variable with the largest deviation from the real-time situation, sets this variable to 100% corresponding to the abnormal state, and calculates the state probabilities of its causal variables.

[0106] (3) Find the variable that best matches the actual abnormal state and has the largest abnormal probability as its abnormal cause.

[0107] (4) Gradually search for the abnormal nodes with faults from downstream to upstream along the causal relationship in turn until there is no parent node for the current node, or all the cause nodes upstream of the current node are normal.

[0108] (5) The abnormal variable where the search finally stops is the root cause variable leading to the system fault, and the propagation path of the fault from upstream to downstream is given.

[0109] As Figure 6 shown, variable 1 represents the fault cause (yellow represents the cause), and can give the fault propagation path in the form of a graph. Each variable can use different colors to represent its abnormal state (red represents rising or positive step, blue represents falling or negative step).

[0110] The system of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network proposed by the present invention includes:

[0111] A multi-variable time window data acquisition module, which is used to organize multi-variable time series data into multi-variable time window data;

[0112] A linear segment acquisition module, which is used to find the best segmentation point through the multi-variable time window data to obtain two linear segments, and describe the linear segments with five trend states;

[0113] A Bayesian network construction module, which is used to construct a variable causal relationship network, and construct a five-state Bayesian network according to the variable causal relationship network, and calculate the conditional probability information between the node states in the five-state Bayesian network;

[0114] A fault cause and fault propagation path acquisition module, which is used to diagnose the actual trend state of the variable by using the conditional probability information, and obtain the fault cause and fault propagation path.

[0115] Specifically, the linear segment acquisition module is used to find the best segmentation point through the multi-variable time window data to obtain two linear segments, and describe the latter linear segment with five trend states.

[0116] The terminal device provided by an embodiment of the present invention includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0117] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention.

[0118] The terminal device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0119] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0120] The memory may be used to store the computer program and / or modules, and the processor realizes various functions of the terminal device by running or executing the computer program and / or modules stored in the memory and by invoking the data stored in the memory.

[0121] If the modules / units integrated in the terminal device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0122] The above content is only for explaining the technical idea of the present invention, and the protection scope of the present invention cannot be limited thereby. Any modification made on the basis of the technical solution according to the technical idea proposed by the present invention falls within the protection scope of the claims of the present invention.

Claims

1. A fault diagnosis method based on qualitative trend analysis and a five-state Bayesian network, characterized in that, Including: Organize multi-variable time series data into multi-variable time window data; Find the optimal segmentation point through the multi-variable time window data to obtain two linear segments, and describe the linear segments using five trend states; Construct a variable causal relationship network, and based on the variable causal relationship network, construct a Bayesian network with five states, and calculate the conditional probability information between node states in the Bayesian network with five states; Use the conditional probability information to diagnose the actual trend state of the variable, and obtain the cause of the fault and the fault propagation path; The expressions of the two linear segments are shown in formulas (1) and (2): where i represents the sampling moment within the multi-variable time window data, and respectively represent the start moments of two linear segments; and respectively represent the measured values of the variable in the two linear segments at and moments; p1 and p2 are the slopes of the two linear segments; Use the least squares method for fitting to obtain the slopes p1 and p2 of the two linear segments. The calculation methods of the slopes p1 and p2 of the two linear segments are shown in formulas (3) and (4): where t k represents the segmentation point of two linear segments, and t k = 5, 6, …, L – 5; 1 to t k is the previous segment, and t k + 1 to L is the subsequent segment; and respectively represent the average values of the sampling times within the two linear segments, and respectively represent the average values of the measured values within the two linear segments; Use the traversal method to find the segmentation point Segmentation point The calculation formula is shown in Formula (5) as follows: The predicted value and The sum of the absolute values of the differences between the measured values y1 and y2 is used as the fitting error, and the splitting point of the two linear segments is found by seeking the minimum fitting error 2. The fault diagnosis method based on qualitative trend analysis and five-state Bayesian network according to claim 1, wherein The multi-variable time series data has two dimensions: one variable dimension, representing each variable; one time dimension, representing each sampling point; Organize the multi-variable time series data into multi-variable time window data with a fixed time length, where L is 10 min to 10 h.

3. The fault diagnosis method based on qualitative trend analysis and five-state Bayesian network according to claim 1, characterized in that Describe the latter segment using normal, rising, falling, positive step, and negative step; When |p| ≤ p th and |I d | > h tc and I d > 0, the latter segment is a goose step; When |p| ≤ p th and |I d | > h tc and I d < 0, the latter segment is a negative step; When |p| ≤ p th and |I d | ≤ h tc , the latter segment remains unchanged; When |p| > p th and p > 0, the subsequent segment is ascending; When |p| > p th and p < 0, the latter segment is decreasing; Among them, I d represents the difference between the connection points of the two linear segments before and after; p th is used to compare with the change slope p of the variable linear segment to determine whether the trend of the linear segment remains unchanged; h tc is used to compare with the difference I between the connection points of the two linear segments before and after of the variable d to determine whether there is a step change when the trend of the latter linear segment remains unchanged.

4. The fault diagnosis method based on qualitative trend analysis and five-state Bayesian network according to claim 1, wherein In the variable causal relationship network, each node represents a variable, and a one-way arrow is used to represent the causal relationship between variables. The cause node points to the result node, indicating that the abnormality of the cause variable may lead to the abnormality of the result variable; According to the variable causal relationship network, construct a five-state Bayesian network, and represent the variable information flow relationship of the system through conditional probability. Its definition is shown in formula (6): G = (V, E, Ψ) (6) Calculate the conditional probability information between node states in the five-state Bayesian network: According to the change trend states of each variable, use the maximum likelihood estimation method to calculate the conditional probability information between node states in the five-state Bayesian network.

5. The fault diagnosis method based on qualitative trend analysis and five-state Bayesian network according to claim 4, characterized in that Use the conditional probability information to diagnose the actual trend state of the variable: Update the trend state of each variable to the constructed five-state Bayesian network; According to the actual trend state of each variable, perform inference diagnosis through the Bayesian formula, as shown in formula (7): The Bayesian inference starts from the non-normal variable with the largest real-time deviation degree, sets the variable to the corresponding abnormal state of 100%, and calculates the state probabilities of its cause variables; Find the variable that best matches the actual abnormal state and has the largest abnormal probability as its abnormal cause; Sequentially search for abnormal nodes where the fault occurs from downstream to upstream along the causal relationship until the current node has no parent node, or all cause nodes upstream of the current node are normal; The finally stopped abnormal variable is the root cause variable that causes the system fault, and the fault propagation path from upstream to downstream is given.

6. A system adopting the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network according to any one of claims 1 to 5, characterized in that Including: A multi-variable time window data acquisition module, which is used to organize multi-variable time series data into multi-variable time window data; A linear segment acquisition module, which is used to find the optimal segmentation point through the multi-variable time window data to obtain two linear segments, and describe the linear segments using five trend states; A Bayesian network construction module, which is used to construct a variable causal relationship network, and construct a Bayesian network with five states according to the variable causal relationship network, and calculate the conditional probability information between node states in the Bayesian network with five states; A fault cause and fault propagation path acquisition module, which is used to diagnose the actual trend state of variables by using the conditional probability information, and obtain the fault cause and fault propagation path.

7. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the fault diagnosis method based on qualitative trend analysis and five-state Bayesian network described in any one of claims 1 to 5.

Images