Three-phase separator fault cause positioning method and system, processing equipment and storage medium

Abstract

The invention relates to a three-phase separator fault cause positioning method and system, a processing device and a storage medium. The method comprises the following steps: determining key loop information and parameter information of a three-phase separator; constructing an exception reasoning network structure, calculating possibility levels among node variables in the exception reasoning network structure, and converting the exception reasoning network structure into a directed weighted graph in a query matrix storage form; determining the lowest activation value of a fault cause node in the abnormal reasoning network structure based on the possibility level between node variables in the abnormal reasoning network structure by adopting a particle swarm optimization algorithm; fault phenomenon nodes in the operation process of the three-phase separator are obtained through an industrial field alarm system; all the fault cause nodes of the industrial field alarm system are determined based on the directed weighted graph in the query matrix storage form, the lowest activation value of the fault cause nodes in the anomaly reasoning network structure and the obtained fault phenomenon nodes, and the method can be widely applied to the field of petroleum and petrochemical engineering.

Description

Technical Field

[0001] The invention relates to the field of petroleum and petrochemicals, and in particular to a method, system, processing equipment and storage medium for locating the cause of a three-phase separator failure. Background Art

[0002] Oil is one of the most important energy sources in today's human society and the most important driving force of the country's industrial pillar industry. It is called the "blood of industry". The world's marine oil resources are very rich. The production of marine oil and gas has continued to increase in recent years, alleviating the difficulties caused by the shortage of land oil resources. As a typical carrier of oil and gas processing equipment, offshore oil production platforms are far away from the land, with concentrated equipment layout and a large number of flammable and explosive substances. Once a major accident occurs, it will cause extremely serious consequences to personnel, equipment and marine environment, and it is difficult to rescue and escape personnel. The oil and gas processing system is an important part of offshore oil and gas field development activities. Its main process includes oil, gas and water separation, crude oil stabilization, light hydrocarbon recovery, etc., which can play a role in stabilizing oil and gas field production, maintaining a balance between crude oil extraction and sales, and ensuring the quality of crude oil and natural gas. The produced fluid from offshore platform oil wells cannot be used directly. It is a typical multi-component system.

[0003] The three-phase separation of produced fluid from offshore oil wells is to use the different physical and chemical properties of oil, gas and water to separate oil, gas and water in a certain way (such as gravity sedimentation, packing separation and centrifugation). As a key separation equipment for processing produced fluid from oil wells, how to ensure the long-term stability and efficiency of the separation effect of the three-phase separator under the load of marine environment such as wind, waves and currents has become an urgent problem to be solved. In the actual production process, the operating status of the three-phase separator affects the production efficiency and personnel safety, but the on-site diagnosis method is still mainly based on the participation of operators and inspectors, which is difficult to meet the needs of digitalization, integration and intelligence of offshore oil production platforms. Once an abnormality occurs in the three-phase separator, it will affect the crude oil processing efficiency and oil content of produced water of the offshore platform, and more seriously affect the life safety of on-site personnel. Therefore, it is an important task for maintenance personnel to quickly and accurately locate the cause of the failure. The development of a method for locating the cause of the three-phase separator failure has important theoretical and engineering significance, which will provide reference information for on-site operators to accurately locate the root cause, clearly identify the propagation path, and quickly make remedial measures. At present, as an important process equipment in offshore oilfield engineering, the function of the three-phase separator is to separate oil, gas and water by utilizing the different properties among the three phases. Its separation efficiency or operating performance of oil, gas and water will directly affect the quality of the exported crude oil and gas. Therefore, engineering technicians and scholars continuously improve and optimize the three-phase separator according to the actual exploitation environment of the oilfield and the requirements of the produced fluid properties, so as to make it have better economy and adaptability and meet more separation requirements.

[0004] There are many methods for cause location in the existing technology, including analysis methods based on process data and analysis methods based on qualitative knowledge. Among them, the method based on process data is to construct a causal topology diagram by mining the correlation information between industrial process variables through historical data to achieve cause location. Since this type of method involves less mechanism and process knowledge, has a low degree of limitation and is easy to implement, it is widely favored by scholars at home and abroad. It mainly includes cross-correlation analysis, Granger causality analysis, and transfer entropy. However, the analysis method based on process data is heavily dependent on data samples, but in actual industrial processes, it is relatively scarce to obtain fault data samples, and it is very common to have no target data samples. Therefore, the research direction is turned to the analysis method based on qualitative knowledge. The method based on qualitative knowledge uses qualitative ideas to analyze and characterize the correlation and causal relationship between various units, subsystems, links, etc. in the industrial process, and uses this to judge the generation and propagation of faults. It mainly includes methods such as symbolic directed graphs, adjacency matrices, fault trees, Petri nets, and Bayesian networks. Bayesian networks are favored by researchers for their unique uncertainty knowledge expression form, strong probability expression ability, and incremental learning characteristics of comprehensive prior knowledge. Although massive amounts of data are stored in the distributed control systems of industrial processes, most of the data are normal data of the smooth operation of the device. Even if there are problems with the data, they are usually unlabeled fault data, resulting in limited available fault samples. In the case of fewer fault samples, the application of Bayesian networks is extremely lacking. Therefore, a cause location method that does not rely on fault data is urgently needed to ensure the production efficiency of the device and the safety of personnel. Summary of the invention

[0005] In view of the above problems, the purpose of the present invention is to provide a method, system, processing equipment and storage medium for locating the cause of a three-phase separator fault that effectively combines multi-source dynamic information that can reflect the essence of a production process.

[0006] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a method for locating the cause of a three-phase separator fault is provided, comprising:

[0007] According to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system, determine the key circuit information and parameter information of the three-phase separator;

[0008] Based on the key loop information and parameter information of the three-phase separator, an abnormal reasoning network structure is constructed, the possibility level between node variables in the abnormal reasoning network structure is calculated, and the abnormal reasoning network structure is converted into a directed weighted graph in the form of query matrix storage;

[0009] The particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure;

[0010] Obtain the fault phenomenon nodes during the operation of the three-phase separator through the industrial field alarm system;

[0011] Based on the directed weighted graph stored in the form of a query matrix, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon nodes, all the fault cause nodes of the industrial field alarm system are determined.

[0012] Furthermore, based on the key loop information and parameter information of the three-phase separator, an abnormal reasoning network structure is constructed, the possibility level between node variables in the abnormal reasoning network structure is calculated, and the abnormal reasoning network structure is converted into a directed weighted graph in the form of a query matrix storage, including:

[0013] Based on the process data related to the three-phase separator and the key circuit information and parameter information of the three-phase separator, the corresponding relationship between the fault phenomenon node and the fault cause node is constructed to form an abnormal reasoning network structure;

[0014] Using fuzzy theory knowledge, the possibility level between node variables in the abnormal reasoning network structure is calculated;

[0015] The anomaly reasoning network structure is converted into a directed weighted graph in the form of query matrix storage.

[0016] Furthermore, the use of fuzzy theoretical knowledge to calculate the likelihood levels between node variables in the abnormal reasoning network structure includes:

[0017] Classify the language of expert evaluation;

[0018] Using trapezoidal and triangular fuzzy numbers, the fuzzy membership is determined;

[0019] Based on the determined fuzzy membership, a comprehensive fuzzy number corresponding to the expert evaluation language is determined;

[0020] The determined comprehensive fuzzy number is defuzzified to obtain the possibility level between node variables in the abnormal reasoning network structure.

[0021] Furthermore, the particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure, including:

[0022] Determine the value range of the minimum activation value of the fault cause node in the abnormal reasoning network structure;

[0023] Based on the probability level between node variables in the abnormal reasoning network structure, a test case set is constructed. The construction process of the test case set is to find out the most likely fault phenomenon node set corresponding to each fault cause node;

[0024] Based on the constructed test case set and the determined range of the minimum activation value, the particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure.

[0025] Furthermore, the possibility level between node variables in the abnormal reasoning network structure is used to construct a test case set, including:

[0026] Determine the set of fault cause nodes in the abnormal reasoning network structure;

[0027] Determine a set of fault phenomenon nodes corresponding to each fault cause node in the abnormal reasoning network structure;

[0028] According to the process data, a very likely fault phenomenon node set is determined in the fault phenomenon node set, and the condition for the fault cause node to be activated is that all fault phenomenon nodes in the fault phenomenon node set are activated;

[0029] An initial test case set is constructed based on a highly likely fault phenomenon node set, wherein each test case includes a fault cause node and a highly likely phenomenon node set.

[0030] Furthermore, the particle swarm optimization algorithm is used based on the constructed test case set and the determined value range of the minimum activation value to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure, including:

[0031] Initialize the minimum activation value for each fault cause node and use the calculated minimum activation value as input;

[0032] The fault phenomenon nodes in the test case are input into the abnormal reasoning network structure as the input conditions of reasoning. The abnormal reasoning network structure activates the corresponding related fault cause nodes according to the input fault phenomenon nodes.

[0033] Using the matching procedure, the cumulative fitness of all test cases in the current iteration is calculated;

[0034] In this round of iteration, the current local optimal value and global optimal value are recorded;

[0035] Update the particle's speed, position and weight coefficient according to the current local optimal value and global optimal value;

[0036] Determine whether the current state meets the preset iteration termination condition. If the maximum number of iterations has been reached, the termination condition is met and the final result is output; if the maximum number of iterations has not been reached, continue to the next round of iteration.

[0037] Furthermore, the directed weighted graph stored in the form of a query matrix, the lowest activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node determine all the fault cause nodes of the industrial field alarm system, including:

[0038] Find the directed edge corresponding to the fault phenomenon node in the fault phenomenon column of the query matrix, and record the corresponding possibility level as the activation value of the fault cause node;

[0039] Compare the activation value of the fault cause node with the minimum activation value. If the activation value is greater than the minimum activation value, the fault cause node is activated; otherwise, the fault cause node is not activated.

[0040] The activated fault cause node is used as a fault phenomenon node, and is searched again in the fault phenomenon column of the query matrix until a certain node is inferred, and the activated fault cause node vector is empty or the activated fault cause node is not found in the fault phenomenon column, at which time the inference process ends;

[0041] Output all fault cause nodes and propagation paths and corresponding normalized probability values.

[0042] In a second aspect, a three-phase separator fault cause locating system is provided, comprising:

[0043] A sensor network building module is used to determine the key circuit information and parameter information of the three-phase separator according to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system;

[0044] An abnormal reasoning network construction module is used to construct an abnormal reasoning network structure based on key loop information and parameter information of the three-phase separator, calculate the possibility level between node variables in the abnormal reasoning network structure, and convert the abnormal reasoning network structure into a directed weighted graph in the form of query matrix storage;

[0045] A minimum activation value determination module is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure by using a particle swarm optimization algorithm;

[0046] A fault phenomenon node acquisition module is used to acquire the fault phenomenon nodes during the operation of the three-phase separator through the industrial field alarm system;

[0047] The fault cause node determination module is used to determine all the fault cause nodes of the industrial field alarm system based on the directed weighted graph stored in the query matrix form, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node.

[0048] In a third aspect, a processing device is provided, comprising computer program instructions, wherein the computer program instructions, when executed by the processing device, are used to implement the steps corresponding to the above-mentioned method for locating the cause of a three-phase separator fault.

[0049] In a fourth aspect, a computer-readable storage medium is provided, on which computer program instructions are stored, wherein the computer program instructions, when executed by a processor, are used to implement the steps corresponding to the above-mentioned method for locating the cause of a three-phase separator fault.

[0050] The present invention adopts the above technical solution, which has the following advantages:

[0051] 1. The present invention realizes rapid and accurate positioning of the cause of three-phase separator failure by comprehensively processing multiple alarm information of the industrial field alarm system, combining abnormal reasoning network structure and fuzzy processing, which can not only improve the accuracy and efficiency of diagnosis, reduce manual intervention and misdiagnosis rate, but also significantly enhance the intelligence and automation level of the industrial field alarm system.

[0052] 2. The present invention uses triples to deconstruct the abnormal reasoning network structure, making the abnormal reasoning network structure clearer, easier to expand and maintain, and further reducing maintenance costs.

[0053] 3. The present invention can timely discover and deal with potential safety hazards, ensure production safety, and protect the lives of on-site personnel.

[0054] 4. The present invention has broad promotion value and can be applied to the field of abnormal diagnosis of other industrial equipment, providing strong support for industrial intelligence and safe production.

[0055] In summary, the present invention can be widely used in the field of petroleum and petrochemicals. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings:

[0057] Figure 1 It is a schematic diagram of constructing an abnormal reasoning network structure provided by an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of a process flow of a three-phase separator provided by one embodiment of the present invention;

[0059] Figure 3 is a schematic diagram of a fuzzy membership function provided by an embodiment of the present invention;

[0060] Figure 4 is a schematic diagram of a comprehensive fuzzy number provided by an embodiment of the present invention;

[0061] Figure 5 It is a schematic diagram of a directed weighted graph and two representation methods provided by an embodiment of the present invention;

[0062] Figure 6 is a schematic diagram of an optimization process of a minimum activation value of a cause node provided by an embodiment of the present invention;

[0063] Figure 7 It is a schematic diagram of an online activation reasoning process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0065] It should be understood that the terms used in the text are only for the purpose of describing specific example embodiments, and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used in the text may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0066] Although the terms first, second, third, etc. can be used in the text to describe multiple elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms can only be used to distinguish an element, component, region, layer or section from another region, layer or section. Unless the context clearly indicates, terms such as "first", "second" and other numerical terms do not imply order or sequence when used in the text. Therefore, the first element, component, region, layer or section discussed below can be referred to as the second element, component, region, layer or section without departing from the teaching of the example embodiments.

[0067] In an offshore floating production storage and offloading unit, the horizontal three-phase separator is the core separation equipment. Once a failure occurs, it may cause significant economic losses and catastrophic consequences. Therefore, it is crucial to accurately and quickly identify the cause of the failure of the three-phase separator. However, in actual three-phase separator systems, the available fault samples are usually limited, resulting in the inability of traditional diagnostic methods to accurately identify the cause of the failure. In addition, although the fault diagnosis theory based on Bayesian networks has made great progress in recent years, the application of Bayesian networks is extremely lacking when there are fewer fault samples. Therefore, there is an urgent need for a cause location method that does not rely on fault data to ensure the production efficiency of the device and the safety of personnel. The present invention proposes a new method for locating the cause of a three-phase separator, namely, an abnormal reasoning network. An embodiment of the present invention provides a method for locating the cause of a three-phase separator fault, comprising: determining key loop information and parameter information of the three-phase separator according to the structural characteristics and monitoring points of the three-phase separator in an industrial field alarm system; constructing an abnormal reasoning network structure based on the key loop information and parameter information of the three-phase separator, calculating the possibility level between node variables in the abnormal reasoning network structure, and converting the abnormal reasoning network structure into a directed weighted graph in the form of a query matrix storage; using a particle swarm optimization algorithm, based on the possibility level between node variables in the abnormal reasoning network structure, determining the minimum activation value of the fault cause node in the abnormal reasoning network structure; obtaining the fault phenomenon node during the operation of the three-phase separator through the industrial field alarm system; determining all the fault cause nodes of the industrial field alarm system based on the directed weighted graph in the form of a query matrix storage, the minimum activation value of the fault cause node in the abnormal reasoning network structure, and the obtained fault phenomenon node. The present invention realizes rapid and accurate positioning of the cause of three-phase separator failure by comprehensively processing multiple alarm information of the industrial field alarm system, combining abnormal reasoning network structure and fuzzy processing. It can not only improve the accuracy and efficiency of diagnosis, reduce manual intervention and misdiagnosis rate, but also significantly enhance the intelligence and automation level of the industrial field alarm system.

[0068] Example 1

[0069] like Figure 1 As shown, this embodiment provides a method for locating the cause of a three-phase separator fault, comprising the following steps:

[0070] 1) According to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system, determine the key circuit information and parameter information of the three-phase separator.

[0071] Specifically, based on the alarm information of the industrial field alarm system, the alarm information of the industrial field alarm system in the same time period is comprehensively considered. Due to the uncertainty of the raw materials of the three-phase separator, and the different equipment status, external environment, process technology, etc., the production conditions are complex and changeable. The industrial field alarm system may generate multiple alarm information, and there may be redundancy or correlation between these alarm information. It may not be timely to manually process a single alarm information. Therefore, the present invention changes the research idea from the existing manual processing of a single alarm information to the comprehensive processing of several alarm information, so as to improve the accuracy and rapidity of the method. Therefore, the specific process of this step is:

[0072] 1.1) Obtain relevant information about the three-phase separator in the industrial field alarm system, including PID flow chart, process configuration screen and alarm information. The process flow of the three-phase separator is as follows: Figure 2 shown.

[0073] 1.2) Based on the relevant information of the three-phase separator, according to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system, the key circuit information and parameter information of the three-phase separator are obtained.

[0074] Among them, the key loop information includes the water phase level control loop LIC2001, the gas pressure control loop PIC2007 and the oil phase level control loop LIC2003, etc., and the parameter information includes the water phase level LIC2001_PV, the gas pressure PIC2007_PV, the oil phase level LIC2003_PV and the tank temperature TI2005, etc.

[0075] 2) Based on the process data related to the three-phase separator and the key circuit information and parameter information of the three-phase separator, an abnormal reasoning network structure is constructed, specifically:

[0076] 2.1) Obtain process data related to the three-phase separator.

[0077] Specifically, the process data related to the three-phase separator include training manuals, operation manuals and inspection records.

[0078] 2.2) Identify the phenomena and causes of faults in the training manual and inspection records, and build an abnormal reasoning network structure based on the key circuit information and parameter information of the three-phase separator, build the correspondence between the fault phenomenon nodes and the fault cause nodes, and form an abnormal reasoning network structure.

[0079] Specifically, in actual industrial processes, a fault phenomenon may have multiple fault causes, and a fault cause may be represented by multiple fault phenomena. Define fault phenomenon nodes and fault cause nodes, build a connection correspondence between fault phenomenon nodes and fault cause nodes, and form an abnormal reasoning network structure. Figure 1In the module for constructing the abnormal reasoning network structure, the node collection is used to identify the fault phenomenon and fault cause when the fault occurs in the training manual and inspection records as nodes, and the node connection is the connection relationship from the fault phenomenon to the fault cause.

[0080] Specifically, in the abnormal reasoning network structure, all the fault phenomenon nodes involved are related to the key circuit information and parameter information. When the key circuit and parameter alarm occur, the fault phenomenon nodes in the abnormal reasoning network structure are activated.

[0081] 3) Using fuzzy theory knowledge, calculate the possibility level between node variables in the abnormal reasoning network structure.

[0082] Specifically, the probability level between nodes in the abnormal reasoning network structure exists between the parent node and the child node. The specific meaning is that when the parent node occurs, the probability of the child node occurring is in the range of 0 to 10, and the larger the number, the greater the probability of occurrence. The probability level between node variables is expressed as LR(V2|V1), where V1 is the parent node and V2 is the child node. Due to the lack of statistical data, the present invention introduces fuzzy theory knowledge to calculate the probability level between node variables in the abnormal reasoning network structure. The specific process is:

[0083] 3.1) Classify the expert evaluation language.

[0084] Specifically, the typical estimate of human working memory capacity is 7±2, which means that the appropriate number of comparisons for humans to judge at one time is between 5 and 9. Therefore, the expert evaluation language is divided into 7 levels: very high (VH), high (H), relatively high (RH), medium (M), relatively low (RL), low (L) and very low (VL).

[0085] 3.2) Using trapezoidal and triangular fuzzy numbers, determine the fuzzy membership.

[0086] Specifically, in the uncertainty analysis of expert evaluation, fuzzy membership is a commonly used representation method, and its value is generally between 0 and 1. The prior art shows that trapezoidal and triangular membership functions are the most common types of fuzzy membership. These two fuzzy numbers are favored by researchers because of their advantages over other fuzzy numbers in measuring the sensitivity of language information. Therefore, the present invention uses trapezoidal and triangular fuzzy numbers to define the mapping relationship between fuzzy language and numerical values. Fuzzy membership function is as follows: Figure 3 shown.

[0087] 3.3) Based on the determined fuzzy membership, determine the comprehensive fuzzy number corresponding to the expert evaluation language.

[0088] Specifically, the present invention synthesizes fuzzy numbers by linear weighted summation, and the synthesis process is:

[0089]

[0090] Among them, M i (x) is the comprehensive fuzzy number of the possibility level on the i-th directed edge; w j is the weight value of the jth expert, satisfying that the sum of the weights is 1, that is, f ji is the fuzzy number of the semantic judgment of the jth expert on the possibility level of the i-th directed edge, i = 1, 2, ..., m, j = 1, 2, ... n, there are m directed edges and n experts in total. In the above formula (1), the summation operation rule is: Assuming there are two trapezoidal fuzzy numbers f1 = (a1, b1, c1, d1), f2 = (a2, b2, c2, d2), the summation operation of the two trapezoidal fuzzy numbers is a1, b1, c1, d1 are the position parameter information of the trapezoidal fuzzy number f1 respectively, and a2, b2, c2, d2 are the position parameter information of the trapezoidal fuzzy number f2 respectively.

[0091] Specifically, factors such as the knowledge and experience of experts will affect the evaluation results. To ensure the accuracy of the results, each expert is given a corresponding weight. Suppose the expert group consists of n experts, and the weight of each expert is w j , the expert weight calculation process is:

[0092] ① Calculate the similarity S(R) of the fuzzy numbers between each pair of experts i , R j ).

[0093] More specifically, let expert E i and E j The evaluation languages ​​correspond to the fuzzy numbers R i =(a1, b1, c1, d1, w1) and R j =(a2, b2, c2, d2, w2), combined with the center point distance, span, center width, perimeter and area of ​​the trapezoidal fuzzy number, the similarity S of the fuzzy numbers between each pair of experts is = (R i , R j ) is calculated as:

[0094]

[0095] Where k is the distance between the center points; z is the span; h is the center width; P(R i ), P(R j ) are fuzzy numbers R i = the circumference of (a1, b1, c1, d1, w1) and R j = the perimeter of (a2, b2, c2, d2, w2); S(R i ), S(Rj ) are fuzzy numbers R i = the area of ​​(a1, b1, c1, d1, w1) and R j =the area of ​​(a2, b2, c2, d2, w2).

[0096] ② Based on the calculated similarity S(R i , R j ), construct the similarity matrix SM of n experts:

[0097]

[0098] Among them, S ij =S(R i , R j ), if the fuzzy number R i =(a1, b1, c1, d1, w1) and R j =(a2, b2, c2, d2, w2) are the same, then S ij =S(R i , R j )=0.

[0099] ③ Based on the constructed similarity matrix SM, calculate the weights of n experts.

[0100] More specifically, the maximum eigenvalue λ of the similarity matrix SM is obtained by the following formulas (4) and (5): max and the corresponding eigenvector W = [w1, w2, ..., w n ] T :

[0101] |SM-λE|=0 (4)

[0102] SM*W=λ max (5)

[0103] Among them, the required feature vector W is normalized and becomes the weight of each expert.

[0104] ④ The expert weight and the fuzzy number corresponding to the expert evaluation language are synthesized by the above formula (1) to obtain the comprehensive fuzzy number corresponding to the expert evaluation language.

[0105] 3.4) Defuzzify the determined comprehensive fuzzy number to obtain the possibility level between the node variables in the abnormal reasoning network structure.

[0106] Specifically, to obtain the possibility level between node variables, it is necessary to convert the comprehensive fuzzy number obtained in the previous step into an accurate value. This embodiment adopts the left-right fuzzy sorting method to realize the defuzzification of fuzzy language. The conversion process is:

[0107] 3.4.1) Define the maximum fuzzy set and the minimum fuzzy set:

[0108]

[0109] Among them, f max (x) is the maximum fuzzy set, i.e., the left-valued function; f min (x) is the minimum fuzzy set, i.e., the right-valued function. The calculation method of the comprehensive fuzzy membership function is as follows: Figure 4 shown.

[0110] 3.4.2) Determine the maximum fuzzy set f defined max (x) and the minimum fuzzy set f min (x) and the comprehensive fuzzy number M i (x), and get the leftmost value F of the intersection L (x) and the rightmost value F R (x), and calculate the likelihood level LR(x) between node variables according to the following formula (10):

[0111] F R (x) = sup[M i (x)∧f max (x) (8)

[0112] F L (x) = sup[M i (x)∧f min (x) (9)

[0113]

[0114] 4) Convert the anomaly reasoning network structure into a directed weighted graph in the form of query matrix storage.

[0115] Specifically, in the abnormal reasoning network structure, a fault cause node may be activated by multiple fault phenomenon nodes, and a fault phenomenon node may participate in activating multiple fault cause nodes, which makes the connections between nodes in the abnormal reasoning network structure complicated. Therefore, for the convenience of storage and calculation, the abnormal reasoning network structure is converted into a directed weighted graph in graph theory to realize the deconstruction of the abnormal reasoning network.

[0116] More specifically, in traditional graph algorithms, the two most common basic structures for representing graphs are adjacency matrices and triples. A simple directed weighted graph can be represented as Figure 5 (a). Figure 5 As shown in (b), the adjacency matrix A = (a ij ) N*NThe number of elements with a value of zero is greater than the number of non-zero elements, and the distribution of non-zero elements is irregular. If dense matrices are still used for storage, it will be extremely inefficient. This embodiment introduces the concept of triples to store non-zero elements in the adjacency matrix. The triple storage form of the directed weighted graph [A_row, A_col, A_val] stores the index values ​​of the rows and columns of non-zero elements and the non-zero elements corresponding to the index values ​​respectively. The triple representation is equivalent to the "query matrix" in this embodiment, and the formation of the query matrix is ​​the deconstruction of the abnormal reasoning network. In the abnormal reasoning network structure, the query matrix [A_row, A_col, A_val] stores the phenomenon node sequence number, the cause node sequence number and the possibility level between the nodes respectively. In the query matrix, each row corresponds to a directed edge in the abnormal reasoning network structure, and converting the graph structure into a query matrix facilitates computer storage and calculation.

[0117] 5) Using the particle swarm optimization algorithm, based on the probability level between node variables in the abnormal reasoning network structure, the minimum activation value of the fault cause node in the abnormal reasoning network structure is determined, specifically:

[0118] 5.1) Determine the range of the minimum activation value of the fault cause node in the abnormal reasoning network structure.

[0119] Specifically, there are two types of nodes in the abnormal reasoning network structure, namely, fault cause nodes and fault phenomenon nodes. The minimum activation value of the fault phenomenon node depends on whether the node event occurs. When the fault phenomenon node occurs, the node is activated; otherwise, the node is not activated. The minimum activation value of the fault cause node depends on its corresponding fault phenomenon node and the possibility level on the directed edge.

[0120] More specifically, the minimum activation value MAV (V i )for:

[0121]

[0122] Among them, T i is the minimum activation value of the fault cause node, assuming that the fault cause node v i There are k fault phenomenon nodes W=(v i-k , v i-k+1 , …, v i-1 ), the probability level between k fault phenomenon nodes and the fault cause node is (LR(v i |v i-k ), LR(v i |v i-k+1 ),…,LR(v i |v i-1 )), then the minimum activation value T i The value range of is:

[0123]

[0124] 5.2) Based on the probability level between the node variables in the abnormal reasoning network structure, a test case set TC is constructed. The construction process of the test case set TC is to find out the most likely fault phenomenon node set corresponding to each fault cause node, specifically:

[0125] 5.2.1) Determine the fault cause node set V = {v1, v2, ..., v n}.

[0126] 5.2.2) Determine the fault phenomenon node set P corresponding to each fault cause node in the abnormal reasoning network structure i = {v i1 , v i2 , ..., v ik}.

[0127] 5.2.3) According to the probability level between the fault phenomenon node and the fault cause node, in the fault phenomenon node set P i = {v i1 , v i2 , ..., v ik} to determine the most likely fault phenomenon node set Q i = {v i1 , v i2 , ..., v im}, find out the most likely fault phenomenon node set based on the process data as follows: i The conditions for activation are: fault phenomenon node set Q i All fault phenomenon nodes in the are activated.

[0128] Example: The fault cause node v1 corresponds to the fault phenomenon node set {v 11 , v 12 , ..., v 1k}; the probability levels of the fault phenomenon node and the fault cause node are (LR(v1|v 11 ), LR(v1|v 12 ),…,LR(v1|v 1k )), select the combinations that account for more than 80% of the sum of the possibility levels in the possibility levels. Assuming that the possibility levels of the fault phenomenon node and the fault cause node are [9, 8, 3] respectively, the fault phenomenon nodes with possibility levels 9 and 8 are regarded as the very likely fault phenomenon node set of the fault cause node.

[0129] 5.2.4) Based on the most likely fault phenomenon node set Qi = {v i1 , v i2 , ..., v im}, construct the initial test case set TC = {(v1, Q1), (v2, Q2), ..., (v n , Q n )}, where each test case includes the fault cause node v i and the most likely phenomenon node set Q i .

[0130] 5.3) Based on the constructed test case set TC and the determined range of the minimum activation value, the particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure.

[0131] Specifically, to reasonably set the minimum activation value T i In this embodiment, the particle swarm optimization algorithm is introduced to optimize the overall parameters. i Set to the spatial position of the particle and use the test case set for parameter optimization.

[0132] More specifically, the particle swarm optimization algorithm is an evolutionary computing method based on swarm intelligence. It searches for the optimal parameter value within a value range by particles. Considering the global optimization problem and local optimization ability of the particle swarm, this embodiment provides an improved particle swarm optimization algorithm based on linear decay inertia weight. The particle motion speed and motion position update formulas are shown in equations (13) and (14):

[0133]

[0134] w=w max -(w max -w min )*iter / maxiter (15)

[0135] in, They are the speed at the current moment, the speed at the next moment, the position at the current moment, and the position at the next moment. are the individual extreme value and the group extreme value respectively; c1, c2 are learning factors respectively; r1, r2 are random numbers in the range of [0, 1] respectively; w is the inertia weight coefficient; w max and w min are the maximum inertia weight coefficient and the minimum inertia weight coefficient, iter and maxiter are the current iteration number and the maximum iteration number, respectively. The above formula (15) is the formula for the linear decay of the inertia weight with the iteration number.

[0136] More specifically, to evaluate the minimum activation value T of the fault cause node in each iterationi In this embodiment, the most likely phenomenon node set Q i and the minimum activation value T i Input into the abnormal reasoning network structure for online activation reasoning, and the abnormal reasoning network structure outputs the fault cause node set V i opt is the input of the fitness function, and the output of the fitness function is V i opt With V i exp The degree of consistency, the minimum activation value T of the fault cause node i The optimization process is as follows Figure 6 As shown, the specific process is:

[0137] 5.3.1) According to the above formula (12), the minimum activation value is initialized for each fault cause node, and the calculated minimum activation value is used as input to provide a basis for the subsequent operation of the abnormal reasoning network structure.

[0138] 5.3.2) Input the fault phenomenon node in the test case into the abnormal reasoning network structure as the input condition of reasoning. The abnormal reasoning network structure activates the corresponding related fault cause node according to the input fault phenomenon node, providing a basis for the next matching procedure.

[0139] 5.3.3) Using the matching procedure, calculate the cumulative fitness of all test cases in the current iteration.

[0140] Specifically, the matching program generates a possible set of fault cause nodes based on the inference results of the fault cause nodes, calculates the fitness value of each fault cause node and the test case, accumulates the fitness of all test cases, and outputs the cumulative fitness of all test cases in this iteration to provide a basis for optimization.

[0141] 5.3.4) In this round of iteration, the current local optimal value and global optimal value are recorded to facilitate algorithm update and optimization.

[0142] 5.3.5) According to the current local optimal value and global optimal value, update the particle's speed, position and weight coefficient to prepare for the next round of iteration.

[0143] 5.3.6) Determine whether the current state meets the preset iteration termination condition. If the maximum number of iterations has been reached, the termination condition is met, the process ends, and the final result is output; if the maximum number of iterations has not been reached, go to step 5.3.2) and continue the next round of iteration.

[0144] 6) Obtain the fault signs during the operation of the three-phase separator through the industrial field alarm system (i.e., the fault phenomenon node in the abnormal reasoning network structure), and the fault phenomenon node is activated at this time.

[0145] 7) If Figure 7 As shown in the figure, based on the directed weighted graph stored in the query matrix, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node, all the fault cause nodes of the industrial field alarm system are determined, specifically:

[0146] 7.1) Find the directed edge corresponding to the fault phenomenon node in the fault phenomenon column of the query matrix, and record the corresponding possibility level as the activation value of the fault cause node.

[0147] Specifically, if the fault cause nodes of multiple directed edges are the same, the possibility levels of different directed edges are accumulated to the fault cause node and recorded as the activation value of the fault cause node.

[0148] 7.2) Compare the activation value of the fault cause node with the minimum activation value. If the activation value is greater than the minimum activation value, the fault cause node is activated; otherwise, the fault cause node is not activated.

[0149] 7.3) Take the activated fault cause node as the fault phenomenon node and proceed to steps 7.1) and 7.2) until a certain node is inferred, the activated fault cause node vector is empty or the activated fault cause node is not found in the fault phenomenon column, at which point the inference process ends.

[0150] 7.4) Output all fault cause nodes and propagation paths and corresponding normalized probability values.

[0151] Example 2

[0152] This embodiment provides a three-phase separator fault cause location system, including:

[0153] A sensor network building module is used to determine the key circuit information and parameter information of the three-phase separator according to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system;

[0154] An abnormal reasoning network construction module is used to construct an abnormal reasoning network structure based on key loop information and parameter information of the three-phase separator, calculate the possibility level between node variables in the abnormal reasoning network structure, and convert the abnormal reasoning network structure into a directed weighted graph in the form of query matrix storage;

[0155] A minimum activation value determination module is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure by using a particle swarm optimization algorithm;

[0156] A fault phenomenon node acquisition module is used to acquire the fault phenomenon nodes during the operation of the three-phase separator through the industrial field alarm system;

[0157] The fault cause node determination module is used to determine all the fault cause nodes of the industrial field alarm system based on the directed weighted graph stored in the query matrix form, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node.

[0158] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0159] Example 3

[0160] This embodiment provides a processing device corresponding to the method for locating the cause of a three-phase separator fault provided in Embodiment 1. The processing device can be applicable to a client processing device, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of Embodiment 1.

[0161] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processing device, and the processing device executes the method for locating the cause of the three-phase separator fault provided in this embodiment 1 when running the computer program.

[0162] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0163] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0164] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0165] Those skilled in the art will understand that the structure of the above-mentioned computing device is only a partial structure related to the solution of the present invention, and does not constitute a limitation on the computing device to which the solution of the present invention is applied. The specific computing device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0166] Example 4

[0167] This embodiment provides a computer program product corresponding to the method for locating the cause of a three-phase separator fault provided in Embodiment 1. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the method for locating the cause of a three-phase separator fault described in Embodiment 1 are loaded.

[0168] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0169] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0170] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0171] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0172] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0173] The above embodiments are only used to illustrate the present invention, wherein the structure, connection mode and manufacturing process of each component may be changed. Any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for locating the cause of a three-phase separator fault, characterized in that: include: According to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system, determine the key circuit information and parameter information of the three-phase separator; Based on the key loop information and parameter information of the three-phase separator, an abnormal reasoning network structure is constructed, the possibility level between node variables in the abnormal reasoning network structure is calculated, and the abnormal reasoning network structure is converted into a directed weighted graph in the form of query matrix storage; The particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure; Obtain the fault phenomenon nodes during the operation of the three-phase separator through the industrial field alarm system; Based on the directed weighted graph stored in the form of a query matrix, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon nodes, all the fault cause nodes of the industrial field alarm system are determined.

2. A method for locating the cause of a three-phase separator fault according to claim 1, characterized in that: The method of constructing an abnormal reasoning network structure based on key loop information and parameter information of the three-phase separator, calculating the possibility level between node variables in the abnormal reasoning network structure, and converting the abnormal reasoning network structure into a directed weighted graph in the form of query matrix storage includes: Based on the process data related to the three-phase separator and the key circuit information and parameter information of the three-phase separator, the corresponding relationship between the fault phenomenon node and the fault cause node is constructed to form an abnormal reasoning network structure; Using fuzzy theory knowledge, the possibility level between node variables in the abnormal reasoning network structure is calculated; The anomaly reasoning network structure is converted into a directed weighted graph in the form of query matrix storage.

3. A method for locating the cause of a three-phase separator fault as claimed in claim 2, characterized in that: The method of using fuzzy theory knowledge to calculate the likelihood level between node variables in the abnormal reasoning network structure includes: Classify the language of expert evaluation; Using trapezoidal and triangular fuzzy numbers, the fuzzy membership is determined; Based on the determined fuzzy membership, a comprehensive fuzzy number corresponding to the expert evaluation language is determined; The determined comprehensive fuzzy number is defuzzified to obtain the possibility level between node variables in the abnormal reasoning network structure.

4. A method for locating the cause of a three-phase separator fault as claimed in claim 1, characterized in that: The particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure, including: Determine the value range of the minimum activation value of the fault cause node in the abnormal reasoning network structure; Based on the probability level between node variables in the abnormal reasoning network structure, a test case set is constructed. The construction process of the test case set is to find out the most likely fault phenomenon node set corresponding to each fault cause node; Based on the constructed test case set and the determined range of the minimum activation value, the particle swarm optimization algorithm is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure.

5. A method for locating the cause of a three-phase separator fault as claimed in claim 4, characterized in that: The possibility level between node variables in the abnormal reasoning network structure is based on constructing a test case set, including: Determine the set of fault cause nodes in the abnormal reasoning network structure; Determine a set of fault phenomenon nodes corresponding to each fault cause node in the abnormal reasoning network structure; According to the process data, a very likely fault phenomenon node set is determined in the fault phenomenon node set, and the condition for the fault cause node to be activated is that all fault phenomenon nodes in the fault phenomenon node set are activated; An initial test case set is constructed based on a highly likely fault phenomenon node set, wherein each test case includes a fault cause node and a highly likely phenomenon node set.

6. A method for locating the cause of a three-phase separator fault as claimed in claim 5, characterized in that: The method of using a particle swarm optimization algorithm based on the constructed test case set and the determined range of the minimum activation value to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure includes: Initialize the minimum activation value for each fault cause node and use the calculated minimum activation value as input; The fault phenomenon nodes in the test case are input into the abnormal reasoning network structure as the input conditions of reasoning. The abnormal reasoning network structure activates the corresponding related fault cause nodes according to the input fault phenomenon nodes. Using the matching procedure, the cumulative fitness of all test cases in the current iteration is calculated; In this round of iteration, the current local optimal value and global optimal value are recorded; Update the particle's speed, position and weight coefficient according to the current local optimal value and global optimal value; Determine whether the current state meets the preset iteration termination condition. If the maximum number of iterations has been reached, the termination condition is met and the final result is output; if the maximum number of iterations has not been reached, continue to the next round of iteration.

7. A method for locating the cause of a three-phase separator fault according to claim 1, characterized in that: The directed weighted graph based on the query matrix storage form, the lowest activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node, determines all the fault cause nodes of the industrial field alarm system, including: Find the directed edge corresponding to the fault phenomenon node in the fault phenomenon column of the query matrix, and record the corresponding possibility level as the activation value of the fault cause node; Compare the activation value of the fault cause node with the minimum activation value. If the activation value is greater than the minimum activation value, the fault cause node is activated; otherwise, the fault cause node is not activated. The activated fault cause node is used as a fault phenomenon node, and is searched again in the fault phenomenon column of the query matrix until a certain node is inferred, and the activated fault cause node vector is empty or the activated fault cause node is not found in the fault phenomenon column, at which time the inference process ends; Output all fault cause nodes and propagation paths and corresponding normalized probability values.

8. A three-phase separator fault cause location system, characterized in that: include: A sensor network building module is used to determine the key circuit information and parameter information of the three-phase separator according to the structural characteristics and monitoring points of the three-phase separator in the industrial field alarm system; An abnormal reasoning network construction module is used to construct an abnormal reasoning network structure based on key loop information and parameter information of the three-phase separator, calculate the possibility level between node variables in the abnormal reasoning network structure, and convert the abnormal reasoning network structure into a directed weighted graph in the form of query matrix storage; A minimum activation value determination module is used to determine the minimum activation value of the fault cause node in the abnormal reasoning network structure based on the possibility level between the node variables in the abnormal reasoning network structure by using a particle swarm optimization algorithm; A fault phenomenon node acquisition module is used to acquire the fault phenomenon nodes during the operation of the three-phase separator through the industrial field alarm system; The fault cause node determination module is used to determine all the fault cause nodes of the industrial field alarm system based on the directed weighted graph stored in the query matrix form, the minimum activation value of the fault cause node in the abnormal reasoning network structure and the acquired fault phenomenon node.

9. A processing device, characterized in that: It comprises computer program instructions, wherein the computer program instructions, when executed by a processing device, are used to implement the steps corresponding to the method for locating the cause of a three-phase separator fault according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, wherein the computer program instructions, when executed by a processor, are used to implement steps corresponding to the method for locating the cause of a three-phase separator fault according to any one of claims 1 to 7.

Images