Method and device for determining fault cause of industrial system and electronic equipment

Abstract

The invention discloses a fault cause determination method and device of an industrial system and electronic equipment. The method comprises the following steps: acquiring a plurality of industrial devices in an industrial system and historical state data of the plurality of industrial devices; determining a causal map based on the plurality of industrial devices and the historical state data, the causal map being used for indicating causal association between the plurality of industrial devices and the historical state data; obtaining a current detection result of the industrial system; under the condition that the current detection result indicates that the abnormal phenomenon exists, determining a target fault propagation path causing the abnormal phenomenon based on a causal map; and determining a fault cause causing the abnormal phenomenon based on the fault propagation path target fault propagation path. According to the method and the device, the technical problem that the analysis accuracy of the fault reason of the industrial system is insufficient in the face of a complex industrial system environment is solved.

Description

Technical Field

[0001] This invention relates to the field of industrial internet, and more particularly to the field of industrial internet and energy management technology. Specifically, it relates to a method, apparatus, and electronic equipment for determining the causes of failures in industrial systems. Background Technology

[0002] With the development of industrial and IoT technologies, industrial systems can collect and process massive amounts of equipment status, operating parameters, and environmental data in real time. This data is crucial for fault location, but it also presents challenges in data processing and analysis, including large data volumes, diverse data types, and data noise. Effective fault diagnosis methods need to be able to quickly filter out fault-related signals from massive amounts of data while handling data noise to ensure the reliability of the analysis results. In industrial systems, fault cause analysis is a complex and critical task, and its complexity and importance increase significantly with the scale and interconnectivity of the production environment. The significant shortcomings of related technologies in determining the causes of faults in industrial systems are mainly manifested in the following aspects:

[0003] Most related technologies rely on fixed logical relationships or data correlation analysis, lacking accurate capture of the dynamic causal links between equipment, parameters, and operating conditions in industrial systems. They are unable to adapt to new failure modes brought about by equipment parameter adjustments or modifications. In scenarios with noisy data or concurrent anomalies, these technologies are easily misled by superficial correlations, leading to decreased root cause accuracy and increasing the difficulty and cost of troubleshooting. Furthermore, most fault diagnosis models in these technologies are difficult to adjust once established, failing to self-optimize according to real-time changes in industrial systems, thus limiting diagnostic effectiveness and applicability. Although related industrial anomaly diagnosis technologies can provide preliminary fault detection and partial cause analysis to a certain extent, their accuracy in analyzing the causes of failures in complex industrial system environments is insufficient due to the failure to effectively establish dynamic causal relationships between equipment.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method, apparatus, and electronic device for determining the causes of failures in industrial systems, in order to at least solve the technical problem of insufficient accuracy in analyzing the causes of failures in industrial systems when facing complex industrial system environments.

[0006] According to one aspect of the present invention, a method for determining the cause of a fault in an industrial system is provided, comprising: acquiring multiple industrial devices in the industrial system and historical state data of the multiple industrial devices, wherein the historical state data includes historical operating parameter data and historical operating condition data; determining a causal graph based on the multiple industrial devices and the historical state data, wherein the causal graph is used to indicate the causal relationship between the multiple industrial devices and the historical state data; acquiring the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an abnormal phenomenon in the industrial system at the current moment; if the current detection result indicates that there is an abnormal phenomenon, determining a target fault propagation path leading to the abnormal phenomenon based on the causal graph; and determining the cause of the fault leading to the abnormal phenomenon based on the target fault propagation path.

[0007] According to another aspect of the present invention, an industrial system fault cause determination apparatus is also provided, comprising: an industrial equipment and historical state data acquisition module, configured to acquire multiple industrial devices in the industrial system and historical state data of the multiple industrial devices, wherein the historical state data includes historical operating parameter data and historical operating condition data; a causal graph determination module, configured to determine a causal graph based on the multiple industrial devices and historical state data, wherein the causal graph is used to indicate the causal relationship between the multiple industrial devices and the historical state data; a current detection result acquisition module, configured to acquire the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an abnormal phenomenon in the industrial system at the current moment; a target fault propagation path determination module, configured to determine the target fault propagation path leading to the abnormal phenomenon based on the causal graph when the current detection result indicates the existence of an abnormal phenomenon; and a fault cause determination module, configured to determine the fault cause leading to the abnormal phenomenon based on the target fault propagation path.

[0008] According to another aspect of the present invention, a non-volatile storage medium is also provided, the non-volatile storage medium storing a plurality of instructions adapted for loading by a processor and executing any one of the fault cause determination methods for industrial systems.

[0009] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the fault cause determination method of any one of the industrial systems described above.

[0010] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the method for determining the cause of failure in any of the industrial systems described above.

[0011] In this embodiment of the invention, multiple industrial devices in an industrial system and their historical status data are acquired, including historical operating parameter data and historical operating condition data. Based on these data, a causal graph is determined, indicating the causal relationships between the devices and the historical status data. The current detection result of the industrial system is obtained, indicating whether an anomaly exists at the current moment. If the current detection result indicates an anomaly, the target fault propagation path leading to the anomaly is determined based on the causal graph. Based on the target fault propagation path, the cause of the anomaly is determined. This achieves the goal of accurately determining the cause of the fault by integrating multiple industrial devices and their historical status data to determine the causal graph, predicting the target fault propagation path based on the causal graph when an anomaly is detected, and thus improving the accuracy of fault cause analysis in industrial systems. This solves the technical problem of insufficient accuracy in fault cause analysis in complex industrial system environments. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0013] Figure 1 This is a flowchart of a method for determining the cause of failure in an industrial system according to an embodiment of the present invention;

[0014] Figure 2 This is a flowchart of an optional method for determining the cause of failure in an industrial system according to an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of a fault cause determination device for an industrial system according to an embodiment of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] According to an embodiment of the present invention, a method embodiment for determining the cause of failure in an industrial system is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0019] Figure 1 This is a flowchart of a method for determining the cause of failure in an industrial system according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0020] Step S102: Obtain multiple industrial devices in the industrial system, as well as historical status data of the multiple industrial devices, wherein the historical status data includes historical operating parameter data and historical operating condition data.

[0021] Optionally, in an industrial environment, an industrial system consists of multiple industrial devices, each with unique functions and operating parameters. These devices include, but are not limited to, pumps, valves, sensors, controllers, reactors, and air compressors. Historical status data is acquired through sensors (including but not limited to temperature sensors, pressure sensors, and vibration sensors) and monitoring systems deployed on these industrial devices. This historical status data refers to various data records accumulated during the historical operation of the industrial devices, documenting their operating parameters and conditions at different points in time. Historical operating parameter data represents the key performance indicators of the industrial devices, including but not limited to temperature, pressure, flow rate, and energy consumption. Historical operating condition data represents the operating environment and conditions of the industrial devices, including but not limited to production load, operating mode, ambient temperature, and humidity. Historical status data is a crucial source of information for analyzing the operating status of industrial devices, identifying failure modes, and understanding the performance trends of industrial systems. Collecting and analyzing historical status data of industrial devices is a key means of improving the operating efficiency and safety of industrial systems.

[0022] Step S104: Based on multiple industrial devices and historical status data, determine a causal graph, wherein the causal graph is used to indicate the causal relationships between multiple industrial devices and historical status data.

[0023] Optionally, in industrial systems, determining a causal graph based on the historical status data of multiple industrial devices refers to constructing a graph representing the causal relationships between multiple industrial devices and their historical operating parameters and conditions by analyzing historical operating parameter data and historical condition data. By analyzing historical status data, the causal graph can reveal the long-term trends of equipment parameters and the dynamic impact of changes in operating conditions, thereby predicting possible failure modes and maintenance needs of industrial equipment. The causal graph clearly shows the causal path between a failure event and its potential causes. By adjusting operating parameters, such as production load and operating temperature, the operating status of the equipment can be optimized. When an industrial system detects an anomaly, it can trace back along the causal logic in the causal graph to directly locate the root cause of the anomaly. This precise location method can significantly shorten troubleshooting time and improve production efficiency.

[0024] In one optional embodiment, a causal graph is determined based on multiple industrial devices and historical state data, including: standardizing the historical state data to obtain standard historical state data; dividing the multiple industrial devices and standard historical state data into three types of nodes, wherein the three types of nodes include a first type of node, a second type of node, and a third type of node, the first type of node including multiple industrial devices, the second type of node including the historical operating parameters corresponding to each of the multiple industrial devices, and the third type of node including the historical operating condition data corresponding to each of the multiple industrial devices; and obtaining multiple causal edges based on the three types of nodes, wherein the causal edges are used to quantify the causal relationship between any two nodes of different categories, and the causal edges include causal direction, causal strength, and delay time, the causal direction indicating the starting and ending nodes of the corresponding causal edge, and the causal... Strength is used to quantify the degree of correlation between the starting and ending nodes of a corresponding causal edge, and delay time represents the time difference between the starting and ending nodes of the corresponding causal edge. Based on multiple causal edges, three types of nodes are classified into multiple levels, including anomaly layer, parameter anomaly layer, equipment status anomaly layer, and component failure layer. The anomaly layer represents the set of nodes directly detected that are associated with historical anomalies in the industrial system. The parameter anomaly layer represents the set of nodes associated with historical abnormal changes in the operating parameters of multiple industrial devices. The equipment status anomaly layer represents the set of nodes associated with historical abnormal states of multiple industrial devices. The component failure layer represents the set of nodes associated with historical failures of internal components of multiple industrial devices. Based on multiple causal edges and multiple levels, a causal graph is obtained.

[0025] Optionally, the collected historical state data is first preprocessed to eliminate the effects of dimensional differences, data noise, and missing values, ensuring that all historical state data can be compared and analyzed on the same basis. This process includes, but is not limited to, steps involving data cleaning, outlier removal, missing value imputation, and unit standardization, resulting in more reliable standard historical state data. Next, industrial equipment, historical operating parameters, and historical operating condition data are transformed into nodes in a causal graph, forming three types of nodes. The first type of node represents the industrial equipment itself, including but not limited to compressors, pumps, and reactors; the second type of node covers the historical operating parameters of multiple industrial equipment, including but not limited to temperature, pressure, speed, and energy consumption; the third type of node covers the historical operating condition data of multiple industrial equipment, including but not limited to production load, operating mode, and ambient temperature. This division helps to clearly define the relationship between equipment, parameters, and operating conditions, facilitating the subsequent construction of causal relationships. Then, by analyzing the dynamic interactions between equipment, parameters, and operating conditions, multiple causal edges are identified. Each causal edge contains three elements: causal direction, causal strength, and delay time. The causal direction is used to clarify the causal relationship between two nodes, such as "temperature increase (parameter anomaly) → equipment failure (equipment status anomaly)," and indicates the direction of the impact. Causal strength uses numerical quantification to determine the degree of correlation between two nodes, reflecting the credibility of the causal relationship and helping to prioritize among numerous potential correlations. Delay time is used to identify the time difference in the transmission of influence from one node to another, considering that many causal effects in industrial systems do not occur instantly but have a certain lag. Furthermore, to more accurately track and understand the root causes of faults in industrial systems, fault diagnosis nodes are classified according to their role and level in the fault chain, constructing a multi-level fault diagnosis framework. This framework is specifically divided into four levels: the anomaly layer, which reflects directly observed abnormal results in industrial systems, such as "increased carbon emissions from air compressors" or "excessive energy consumption." These anomalies are often the starting point for fault diagnosis, directly triggering subsequent fault tracking and analysis processes. Moving down to the next level, the parameter anomaly layer focuses on changes in intermediate process parameters that lead to macroscopic anomalies, such as "extended loading time" or "decreased exhaust volume." While changes in these process parameters are not easily observed directly, they are crucial links connecting macroscopic phenomena with the specific state of equipment, providing clues for further fault localization. Continuing to delve deeper, we reach the equipment condition anomaly layer, indicating abnormal operating conditions of the equipment itself, such as "air compressor pressure drop" or "abnormal motor speed." Anomalies at this level are usually directly related to specific internal components or operating conditions and are core elements of fault diagnosis and localization. The component failure layer is the deepest level, directly touching upon specific faults in internal components, including "seale aging," "filter blockage," or "sensor error." These component-level faults are the root causes of equipment malfunctions and are the direct targets for maintenance and repair.This diagnostic approach, moving from phenomenon to essence, effectively breaks down complex fault chains into manageable analytical steps, significantly reducing troubleshooting time and costs, and improving the operational efficiency and maintenance level of industrial systems. Finally, all causal nodes and edges are integrated to form a causal graph. This graph visually displays the complex causal network between equipment, parameters, and operating conditions, as well as the potential paths of fault propagation. By regularly updating and optimizing the graph, its adaptability to the current state of the industrial system can be maintained, improving its diagnostic accuracy.

[0026] In one optional embodiment, multiple causal edges are obtained based on three types of nodes, including: obtaining the causal strength of any causal edge based on the three types of nodes by: detecting whether any causal edge meets a predetermined causal association standard; if any causal edge meets the predetermined causal association standard, determining the mechanism weight and data weight of any causal edge based on the three types of nodes, wherein the mechanism weight is used to quantify the reliability of any causal edge relative to the predetermined causal association standard, and the data weight is used to quantify the degree of association between the starting node and the ending node of any causal edge in the historical operation of the industrial system; obtaining the causal strength of any causal edge based on the mechanism weight and the data weight; and obtaining the causal strength of each of the multiple causal edges using the same method as obtaining the causal strength of any causal edge.

[0027] Optionally, determining the causal strength of causal edges is a crucial step in constructing a causal graph, directly impacting the accuracy and reliability of fault diagnosis. First, it's necessary to determine whether the association between two nodes meets predetermined causal association criteria. These criteria include, but are not limited to, identification of node pairs theoretically linked by mechanistic knowledge from the industrial field, equipment manuals, and expert experience. For example, theoretically, it's known that "air compressor pressure drop" may be caused by "seal aging," thus, the association between these two is considered a candidate causal edge. Next, for causal edges that meet the predetermined criteria, the underlying physical or chemical principles are analyzed to determine the mechanistic weight, reflecting the theoretical strength of the causal relationship. For instance, according to fluid mechanics principles, "seal aging" has a high probability of directly causing "air compressor pressure drop," therefore, this causal edge has a high mechanistic weight. Further analysis of historical operating data, using data mining methods (including but not limited to Pearson correlation coefficient, mutual information, and causal discovery algorithms), calculates the actual degree of association between the starting and ending nodes during historical operation. For example, if historical data shows that "air compressor pressure drops" frequently occur whenever "seal aging" occurs, then the data weight for this causal edge will be relatively high. Finally, combining the mechanism weight and data weight, a weighted formula is used to calculate the causal strength of each causal edge. The causal strength of any causal edge = first adjustment coefficient. Mechanism weight + second adjustment coefficient Data weights, where the first and second adjustment coefficients are used to quantify the importance between the mechanism weights and the data weights. In this way, the credibility of each causal edge can be quantified, providing data support for subsequent fault diagnosis.

[0028] In an optional embodiment, after detecting whether any causal edge meets a predetermined causal association criterion, the method further includes: if any causal edge does not meet the predetermined causal association criterion, determining a target causal edge, wherein the target causal edge represents a causal edge whose mutual information is greater than a preset mutual information threshold, and the mutual information is used to quantify the degree of mutual dependence between the starting node and the ending node of the corresponding causal edge; obtaining a preset mechanism weight and a preset data weight of the target causal edge; and determining the causal strength of the target causal edge based on the preset mechanism weight and the preset data weight.

[0029] Optionally, in the process of constructing the causal graph, in addition to the causal edges determined based on the preset causal association criteria, there are also some causal edges (target causal edges) that are not covered by the preset causal association criteria but show significant correlations in historical data. The discovery and analysis of these target causal edge pairs are extremely important for improving the causal graph and enhancing the comprehensiveness and accuracy of fault diagnosis. For causal edges that do not meet the preset causal association criteria, mutual information is used to quantify the degree of interdependence of the data distribution between the two nodes corresponding to the causal edge. Mutual information can reveal the nonlinear correlation between two nodes and has high sensitivity and descriptive ability for complex nonlinear relationships in industrial systems. For causal edges with mutual information greater than the preset mutual information threshold, these causal edge pairs constitute target causal edges. For the selected target causal edges, due to the lack of sufficient mechanistic support, a preset mechanistic weight is used to reflect their low theoretical reliability. Simultaneously, a preset data weight is calculated based on the mutual information to reflect the correlation strength shown in historical data. Combining the preset mechanistic weight and data weight of the target causal edge, a weighted formula is also used to determine the causal strength of the target causal edge. The introduction of target causal edges can reduce misdiagnosis or missed diagnosis to some extent and improve the reliability of diagnostic results.

[0030] Step S106: Obtain the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an abnormality in the industrial system at the current moment.

[0031] Optionally, real-time status data from multiple industrial devices in the industrial system is collected. This real-time data is compared with historical data to identify anomalies that deviate from normal operating ranges or behavioral patterns. These anomalies include, but are not limited to, parameter fluctuations, equipment performance degradation, and abnormal increases in energy consumption. The three-standard-deviation rule can be used to determine whether current status data exceeds the normal operating range, thus indicating the presence of anomalies in the industrial system. The three-standard-deviation rule is based on the statistical properties of a normal distribution; if any parameter value of the industrial system exceeds […], it indicates an anomaly. -3 , +3 ] interval, where, The mean of any parameter value. If the value is the standard deviation of any parameter, then that parameter value is considered abnormal. Once an anomaly is detected, the industrial system immediately generates a current detection result indicating an anomaly. By acquiring the current detection results from the industrial system, operators can understand the real-time operating status of industrial equipment and promptly identify and address anomalies.

[0032] Step S108: If the current detection result indicates an abnormal phenomenon, determine the target fault propagation path that caused the abnormal phenomenon based on the causal graph.

[0033] Optionally, in industrial systems, once an anomaly is detected, it is crucial to quickly locate the source of the fault and understand how it propagates within the system. Fault diagnosis methods based on causal graphs can provide a precise solution, capable of extracting propagation paths leading to anomalies from complex interactions. Upon detecting an anomaly, a pre-constructed causal graph is used to traverse the causal relationships within the graph backwards from the anomaly node. This backward search mechanism gradually forms a propagation path from the anomaly to the potential fault source based on the causal strength and delay time of each causal edge in the graph, ultimately yielding the target fault propagation path.

[0034] In an optional embodiment, when the current detection result indicates the presence of an anomaly, a target fault propagation path leading to the anomaly is determined based on a causal graph. This includes: identifying a target anomaly node, where the target anomaly node represents a node matching the anomaly in the anomaly layer of the causal graph, and the anomaly layer represents the set of nodes directly detected that are associated with historical anomalies in the industrial system; traversing the causal graph based on the target anomaly node to obtain an initial fault propagation path, where the initial fault propagation path represents a path composed of nodes associated with the target anomaly node in the causal graph; filtering the initial fault propagation path to obtain a filtered fault propagation path; and performing intervention verification processing on the filtered fault propagation path to obtain the target fault propagation path, where the intervention verification processing means verifying the authenticity of the causal relationship in the filtered fault propagation path by simulating the normal operating state of the industrial system.

[0035] Optionally, once an anomaly is confirmed, the node directly associated with it, i.e., the target anomaly node, will be located in the causal graph. This node, located at the anomaly layer, represents the directly observed anomaly state. Starting from the target anomaly node, the causal graph is traversed to find all associated upstream nodes—those that influence the state of the target anomaly node. This traversal generates an initial fault propagation path, which contains a series of causal chains tracing back from the target anomaly node to the potential source of the fault in the system. This path consists of a series of associated nodes connected by causal edges. However, not all causal edges in the initial fault propagation path have equal reliability. Therefore, the initial fault propagation path needs to be screened and validated to obtain the target fault propagation path, which accurately determines the precise location of the fault. The target fault propagation path not only displays the entire causal path from the source to the anomaly, helping operations personnel understand the full picture of the fault and providing detailed analytical basis for fault diagnosis and handling.

[0036] In one optional embodiment, the initial fault propagation path is filtered to obtain a filtered fault propagation path, including: filtering multiple causal edges in the initial fault propagation path based on a preset causal strength threshold to obtain multiple filtered causal edges, wherein the filtered causal edges represent causal edges in the initial fault propagation path whose causal strength is greater than the preset causal strength threshold; and obtaining the filtered fault propagation path based on the multiple filtered causal edges and the nodes connected to each of the multiple filtered causal edges.

[0037] Optionally, each causal edge in the initial fault propagation path is evaluated to compare its causal strength with a preset causal strength threshold. This threshold is set based on a deep understanding of the industrial system, historical data analysis, and expectations for diagnostic accuracy. If the causal strength threshold is exceeded, the corresponding causal edge is considered a strong causal relationship and will be retained; if the threshold is not met, the corresponding causal edge is considered a weak causal relationship or an incorrect association and will be removed from the initial fault propagation path. The retained causal edges form the filtered fault propagation path. By filtering, edges with low causal strength can be eliminated, reducing the possibility of false diagnoses. This means that maintenance personnel can focus more on highly reliable fault propagation paths, avoiding wasted resources and unnecessary checks, thereby significantly improving diagnostic accuracy.

[0038] In an optional embodiment, when the filtered fault propagation path is obtained based on multiple filtered causal edges, the filtered fault propagation path is subjected to intervention verification processing to obtain the target fault propagation path. This includes: performing intervention verification processing on any filtered causal edge to obtain any intervened causal edge in the following manner: performing an intervention operation on the endpoint node of any filtered causal edge to obtain the endpoint running state of any filtered causal edge, wherein the endpoint running state represents the running state of the endpoint node of the corresponding causal edge; if the endpoint running state of any filtered causal edge is normal, obtaining the starting running state of any filtered causal edge, wherein the starting running state represents the running state of the starting node of the corresponding causal edge; if the starting running state of any filtered causal edge is normal, determining any filtered causal edge as any intervened causal edge; obtaining multiple intervened causal edges using the method of obtaining any intervened causal edge; and obtaining the target fault propagation path based on the multiple intervened causal edges and the nodes connected to each of the multiple intervened causal edges.

[0039] Optionally, although the filtered causal edges already possess high credibility, intervention verification is a necessary step to further confirm the authenticity of their causal relationships. For each filtered causal edge, intervention is applied to its endpoint node. This intervention includes, but is not limited to, changing parameter values ​​in the software model and actually adjusting equipment parameters, observing whether the endpoint node's operating status returns to normal after the intervention. After intervening in the endpoint node, its operating status is continuously monitored to determine if it has returned to normal operation. If the endpoint node's operating status returns to normal, the starting node's operating status is further checked. If the starting node's operating status also returns to normal, it indicates that the causal relationship represented by the causal edge is real. At this point, the causal edge is confirmed as an intervened causal edge. This intervention verification process is performed on each filtered causal edge, ultimately resulting in multiple intervened causal edges. These intervened causal edges constitute a more accurate and verified fault propagation path. Based on these multiple intervened causal edges and their connected nodes, a target fault propagation path is constructed. Through intervention verification, the fault propagation path that truly causes the anomaly can be verified and identified.

[0040] Step S110: Based on the target fault propagation path, determine the cause of the fault that led to the abnormal phenomenon.

[0041] Optionally, based on the target fault chain, operations and maintenance personnel can locate the deepest fault point in the chain, which is the direct source of the anomaly. For example, if the path indicates "seal aging" leading to a series of chain reactions ultimately causing "abnormal carbon emissions," then the seal is the direct fault point. Once the direct fault point is identified, operations and maintenance personnel will implement corresponding repair operations to eliminate the source of the fault. After handling the fault, operations and maintenance personnel can determine the true cause of the anomaly. After obtaining the true cause, they will report it back and compare it with the target fault propagation path obtained based on the causal graph, calculating the inference error of any causal edge. If the inference error of any causal edge exceeds a preset inference error threshold, it indicates that the target fault propagation path obtained based on that causal edge has a large deviation, and the weights of the causal graph need to be adjusted. The updated mechanism weight of any causal edge can be obtained as follows: W m_new =W m_old + ×(1-E), where W m_new W represents the updated mechanistic weight of any causal edge. m_old This represents the mechanistic weight of any causal edge before it is updated. W represents the first correction coefficient, used to control the magnitude of weight adjustment; E represents the inference error of any causal edge; the updated data weight of any causal edge can be obtained as follows: d_new =Wd_old + ×(1-E), where W d_new W represents the updated data weight of any causal edge. d_old This represents the data weight of any causal edge before the update. The second correction coefficient, based on the updated data weight and the updated mechanism weight of any causal edge, determines the updated causal weight of each causal edge. After weight adjustment, the correction is retested using similar past anomaly data to verify its effectiveness. If the difference in accuracy improvement between the updated and unupdated versions exceeds a preset accuracy improvement threshold, the correction mechanism is considered effective and significantly improves fault diagnosis accuracy. If the expected results are not achieved, the first and second correction coefficients are readjusted to ensure continuous optimization and adaptability of the causal graph. Furthermore, the latest power system state data is fed back into the causal graph, and the data weight of each causal edge is recalculated to obtain an updated causal graph. Based on the latest state data, previously undiscovered causal relationships can be identified, and new causal edges can be added according to predetermined causal association criteria and data mining results. Simultaneously, if the data weight of a new causal edge is lower than a preset data weight threshold, it indicates that the causal edge has a weak impact on the anomaly, and these edges are deleted to avoid introducing interference factors into fault diagnosis. By integrating real-time updates with self-learning corrections, the causal graph can accurately reflect the current state of the industrial system, thereby more accurately predicting and locating the causes of failures and improving the efficiency and accuracy of fault response.

[0042] Through the above steps S102 to S110, the goal is to determine the causal graph by integrating the historical status data of multiple industrial devices in the industrial system. When an anomaly is detected in the industrial system, prediction is made based on the causal graph to determine the target fault propagation path, thereby accurately determining the cause of the fault. This achieves the technical effect of improving the accuracy of fault cause analysis in industrial systems, and solves the technical problem of insufficient accuracy in fault cause analysis in industrial systems when facing complex industrial system environments.

[0043] Based on the above embodiments and optional embodiments, taking the diagnosis of abnormal energy and carbon emissions from air compressors in the chemical industry as an example, this invention proposes an optional system for determining the causes of abnormal energy and carbon emissions. The system relies on a hardware architecture of "industrial data acquisition layer - edge computing inference layer - cloud optimization layer," coupled with software modules of "dynamic cause-effect graph construction - cause-effect chain inference - dynamic optimization," to achieve full automation from data input to root cause output. First, the system hardware architecture and environment configuration: First, clarify the hardware foundation supporting the technical solution to ensure that the method can be implemented on physical devices. The hardware architecture must correspond one-to-one with the software modules. The specific configuration is as follows: (1) Industrial data acquisition layer, responsible for acquiring multi-source signals, including three types of hardware devices (equipment sensors, working condition data acquisition device, and environmental monitoring device). Among them, equipment sensors refer to pressure sensors, temperature sensors, and energy consumption sensors deployed on air compressors; working condition data acquisition device is used to connect to the Manufacturing Execution System (MES) to acquire data such as production load and production shifts. It adopts a gateway and supports the Modbus-TCP protocol; environmental monitoring device refers to temperature and humidity sensors deployed in the workshop, with a collection frequency of 5 minutes / time, used to eliminate environmental interference factors. (2) Edge computing inference layer, responsible for the real-time calculation task of causal inference, adopts an edge server to ensure the real-time performance of the industrial site (inference delay ≤ 1 second). (3) Cloud optimization layer, responsible for dynamic updating of causal graphs and storage of historical data, supporting distributed computing.

[0044] Second, the implementation steps of the core module (combining physical processes and program flow):

[0045] Step 1: Implementation of the dynamic cause-effect graph construction module. The core of this module is to transform the "physical association" of the industrial system into a "structured cause-effect graph". It requires four physical steps: "data preprocessing - node definition - cause-effect edge construction - hierarchical division". Each step corresponds to a clear program implementation logic: (1) Data preprocessing (physical signal → standardized digital data) physical process: the acquisition layer hardware converts physical signals such as pressure and temperature (such as the current signal of the sensor) into digital data, and then eliminates noise and missing values ​​through the edge server; program implementation process: using the Pandas library of Python, first through the "three standard deviation rule" (3 The process involves removing outliers (e.g., a sudden pressure jump to 2 MPa, exceeding the equipment's range, which is considered sensor malfunction data) and then using linear interpolation to fill in missing values ​​(e.g., if no energy consumption data is collected within 5 minutes, interpolation is performed based on 10 data points before and after the data point). The output results are then processed into a standardized dataset with fields including "equipment identifier - parameter name - acquisition time - parameter value - operating condition label". The data format is comma-separated values ​​(CSV), and the dataset is stored locally on the edge server. (2) Node definition (physical entity → graph node), including: physical entities based on the industrial site “equipment-parameter-operating condition”, which are abstracted into three types of nodes in the cause-effect graph to ensure that each node corresponds to a unique physical object; the equipment entity node is defined as “equipment type + number” node_id, such as “air compressor-01” and “valve-03”, storing attributes such as equipment model and installation location; the parameter node is defined as “equipment identifier + parameter type” node_name, such as “air compressor-01_pressure” and “air compressor-01_energy consumption”, with associated parameter unit unit (MPa, kWh) and normal threshold normal_range (such as normal pressure range 0.8-1.2 MPa); the operating condition node is defined as “operating condition type + value”, such as “production load-80%” and “shift-day shift”, with associated operating condition effective time (such as 8:00-16:00 is day shift); output results. Node list (JSON format), example is {"node_id":"P001","node_type":"parameter node","node_name":"air compressor-01_pressure","unit":"MPa","normal_range":[0.8,1.2]}. (3) Construction of causal edges (physical association → directed edge), adopting the dual approach of "mechanism rules + data mining" to ensure that the causal edges conform to industrial laws and fit the actual operating data. The specific implementation method is as follows: Mechanism rule construction (based on physical laws), calling the preset standard and equipment manual (such as the association of "leakage source_node → pressure drop target_node" in the air compressor operation and maintenance manual), preset mechanism rules, such as "if the air compressor seal leaks (equipment status), the pressure drops within 10 minutes (parameter change)"; for causal edges that meet the predetermined causal association standard, determine the mechanism weight W. m and data weights W dA data mining optimization method (based on historical data) is employed. Normal operation data for a predetermined period is selected from the past. The mutual information and Pearson correlation coefficient between parameters are calculated using Python's scipy library to determine the data weights of corresponding causal edges. Based on the mechanistic weights and data weights, the causal weights are ultimately determined. For associations that do not meet the predetermined causal association criteria (e.g., "ambient temperature → air compressor heat dissipation efficiency"), potential causal edges (target causal edges) are discovered using a causal discovery algorithm. Causal edges are retained when the mutual information exceeds a preset mutual information threshold. Causal edge attributes are improved; each causal edge is given a "delay time" (based on historical data statistics, e.g., "pressure drops on average 5 minutes after a leak, so delay_time 5min"). The final format of the causal edge is {"edge_id":"E001","source_node":"air compressor-01_leak","target_node":"air compressor-01_pressure","W m ":0.9,"W d ":0.88,"delay_time":5min,"strength":0.892}(strength=0.6W m +0.4W d (4) Hierarchical Division (Physical Logic → Hierarchical Structure) Based on the physical logic of "abnormal phenomena → process changes → equipment status → component failure", the nodes are divided into 4 levels to ensure that the causal path conforms to the tracing logic of industrial problems. Storage and Visualization: The causal graph is constructed using Python's networkx library and stored in GraphML format. The hierarchical structure is displayed on the web interface of the edge server, supporting maintenance personnel to manually adjust node / edge attributes.

[0046] Step 2: Implementation of the causal chain reasoning module. This module starts from the "phenomenal anomaly" and outputs an explainable root cause chain through reverse search, intensity calculation, and intervention verification. The core is to combine "mathematical calculation" with "industrial physical operation" to avoid pure software reasoning. Taking the "increased carbon emissions from air compressors" anomaly as an example, the specific steps are as follows: (1) Anomaly triggering (physical anomaly → program startup). The edge server compares the collected data with the normal threshold in real time and uses the three-standard-deviation rule to determine the anomaly. If the parameter value x exceeds [ -3 , +3 ](in The mean, If the value is less than the standard deviation, it is considered an anomaly. For example, the average carbon emissions from air compressors. =45kgCO2 / t =1.5, when x=52kgCO2 / t is detected, the inference program is triggered; anomaly location, the anomaly node is determined to be "Air Compressor-01_Carbon Emission Increase" in the phenomenon layer, and the time of anomaly occurrence (e.g., 2024-05-20 14:30:00) is recorded as the starting point of inference. (2) Reverse search (tracing the physical association from the phenomenon to the equipment) Parent node screening: Extract all parent nodes of "Air compressor-01_Increased carbon emissions" (i.e., the direct cause that may lead to the anomaly) from the causal graph. Nodes with mismatched time are excluded based on the "delay time" of the causal edge. For example, the parent nodes include "Air compressor-01_Increased energy consumption" (delay time 2 min, there is a record of increased energy consumption 2 min before the anomaly occurs) and "Increased ambient temperature" (delay time 10 min, the ambient temperature did not change 10 min before the anomaly, so it is excluded). Initial fault propagation path generation: Starting from "Increased carbon emissions → Increased energy consumption", continue to search in reverse for the parent nodes of "Increased energy consumption" (such as "Extended loading time" and "Decreased motor efficiency"). Nodes with causal strength exceeding the preset causal strength threshold are selected to form the initial fault propagation path: "Increased carbon emissions → Increased energy consumption → Extended loading time → Decreased pressure". (3) Causal strength calculation (quantifying the degree of physical association): Calculate the causal strength of each causal edge to ensure that the causal strength value reflects the "mechanism reliability" and "data correlation". All causal edges in the initial fault propagation path are sorted in descending order of causal strength, and causal edges with a preset causal strength threshold are retained. (4) Intervention calculation verification (simulated physical operation to exclude false associations), the authenticity of the causal relationship is verified by "simulated physical intervention" to avoid misjudging "correlation" as "causation". The specific implementation method is as follows: intervention design, for nodes in the initial fault propagation path (such as "pressure drop"), simulate "physical repair operation" in the edge server to disconnect the input signal of the "pressure drop" node (equivalent to replacing the seal on site to repair the pressure); after the simulation intervention, calculate the parameter change of the "carbon emission increase" node. If the simulated value returns to the normal operating range (such as from 52kgC) / Loss of 44kgC / t), then verify the causal relationship of "pressure drop → carbon emission increase" is established. If there is no change after simulation (such as "ambient temperature → carbon emission increase"), it is judged as a false association and removed from the initial fault propagation path; chain pruning, based on the intervention results, the final target fault propagation path is "seal aging (component layer) → air compressor pressure drop (equipment layer) → loading time extension (process layer) → energy consumption increase (process layer) → carbon emission increase (phenomenon layer)". (5) Result output (target fault propagation path → operation and maintenance guidance), display the visual causal chain on the web page interface of the edge server (mark the causal strength of each of the multiple causal edges with different colors), and generate a text report, including "abnormal node, root cause node, impact path, and suggested operation (such as replacing air compressor seals)"; data push, push the target to the factory operation and maintenance system through the MQTT protocol, trigger the maintenance work order, and ensure that the operation and maintenance personnel can directly obtain guidance.

[0047] Step 3: Implementation of the dynamic optimization module. This module solves the problem of "static solidification" of cause-effect graphs. Through "real-time data updates" and "feedback corrections", it adapts to changes in the working conditions of industrial systems (such as equipment modification and process adjustment). Specific implementation: (1) Time window setting: The update program is started at a predetermined time every day (during the production trough) to select the day's operating data (including normal and abnormal data) to avoid affecting production; Data preprocessing: The new data is denoised (including but not limited to wavelet transform denoising algorithm) and completed (including but not limited to predicting missing values ​​based on long short-term memory network model) to ensure data quality; Cause-effect edge updates New features include, but are not limited to, recalculating the data weights of causal edges using causal discovery algorithms (e.g., the data weight of "loading time → energy consumption" changes from 0.92 to 0.95 due to enhanced correlation caused by equipment aging); adjusting delay times, such as changing the delay time of "seal aging → pressure drop" from 5 minutes to 4 minutes (due to accelerated seal aging, the fault manifests faster); adding / deleting edges, such as adding an edge based on data mining if a new correlation of "inverter fault → abnormal motor speed" (a newly emerging fault type) is added; and deleting an edge if the data weight of "ambient temperature → pressure" does not exceed the preset data weight threshold. (2) Self-learning correction (weight adjustment based on operation and maintenance feedback): After handling the fault, the operation and maintenance personnel input the "actual root cause" (such as inferred as "aging of seals" but actually "sensor error") through the operation and maintenance system to form feedback data; if the error calculation exceeds the preset inference error threshold, the weight adjustment is triggered; correction verification: after adjustment, the historical abnormal data (such as the pressure abnormality caused by "sensor error" in the past 10 times) is used to re-infer. If the difference in accuracy improvement before and after adjustment exceeds the preset accuracy improvement difference threshold, the correction is confirmed to be effective; if there is no improvement, the weight adjustment process is repeated.

[0048] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method. Figure 2 This is a flowchart of an optional method for determining the cause of failure in an industrial system according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes:

[0049] S1: Collect data on multiple power devices in the power system, as well as historical status data (equipment operation data) and historical anomaly information of the multiple power devices. The specific implementation process is the same as the aforementioned embodiment, and will not be repeated here.

[0050] S2: Based on historical state data and historical anomaly information, construct a causal graph (i.e., a causal graph). The specific implementation process is the same as the aforementioned embodiments, and will not be repeated here.

[0051] S3: Obtain the current detection results of the power system. If the current detection results indicate the presence of an anomaly, use a cause-effect graph to perform reasoning to obtain the target fault propagation path, i.e., output the cause chain corresponding to the fault information. The specific implementation process is the same as the aforementioned embodiment, and will not be repeated here.

[0052] S4: Based on the target fault propagation path, determine the target fault cause of the power system. The specific implementation process is the same as the previous embodiment, and will not be repeated here.

[0053] S5: Based on the target fault propagation path and the target fault cause, send an optimization signal to the cause-effect graph construction module to improve the cause-effect graph construction process. The specific implementation process is the same as the previous embodiment, and will not be repeated here.

[0054] This embodiment can achieve at least one of the following effects: (1) Improved root cause location accuracy. The structured design of the cause-effect graph and the intervention calculation verification mechanism improve the root cause location accuracy; (2) Explainability of the cause chain. The output of multi-level cause-effect chains (average length 3-5 levels) replaces the traditional single cause output, improving the efficiency of operation and maintenance personnel's understanding; (3) Enhanced anti-interference capability. Accuracy is guaranteed in noisy data (signal-to-noise ratio 1:1) or concurrent abnormal scenarios; (4) Dynamic adaptability. The dynamically learned cause-effect graph can be automatically updated with the adjustment of production process and equipment aging, which has stronger robustness and adaptability than fixed expert rules; (5) Upgrading from "notifying abnormality" to "locating root cause" greatly shortens the fault investigation time, reduces the enterprise's dependence on senior experts, and reduces operation and maintenance costs; (6) Improved intelligent level of industrial abnormality handling, helping to accurately reduce emissions (such as quickly locating the root cause of carbon emission abnormalities), while reducing resource waste caused by blind maintenance.

[0055] This embodiment also provides a fault cause determination device for an industrial system. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0056] According to an embodiment of the present invention, an apparatus embodiment for implementing the above-described method for determining the cause of failure in an industrial system is also provided. Figure 3 This is a schematic diagram of a fault cause determination device for an industrial system according to an embodiment of the present invention, as shown below. Figure 3 As shown, the above-mentioned industrial system fault cause determination device includes: an industrial equipment and historical status data acquisition module 300, a cause-effect graph determination module 302, a current detection result acquisition module 304, a target fault propagation path determination module 306, and a fault cause determination module 308, wherein:

[0057] The industrial equipment and historical status data acquisition module 300 is used to acquire multiple industrial devices in the industrial system, as well as the historical status data of the multiple industrial devices. The historical status data includes historical operating parameter data and historical operating condition data.

[0058] The causal graph determination module 302 is connected to the industrial equipment and historical state data acquisition module 300. It is used to determine the causal graph based on multiple industrial equipment and historical state data. The causal graph is used to indicate the causal relationship between multiple industrial equipment and historical state data.

[0059] The current detection result acquisition module 304 is connected to the cause-effect graph determination module 302 and is used to acquire the current detection result of the industrial system. The current detection result is used to indicate whether there is an abnormal phenomenon in the industrial system at the current moment.

[0060] The target fault propagation path determination module 306 is connected to the current detection result acquisition module 304. It is used to determine the target fault propagation path that causes the abnormality based on the causal graph when the current detection result indicates that there is an abnormality.

[0061] The fault cause determination module 308 is connected to the target fault propagation path determination module 306 and is used to determine the fault cause that leads to the abnormal phenomenon based on the target fault propagation path.

[0062] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0063] It should be noted that the aforementioned industrial equipment and historical status data acquisition module 300, cause-effect graph determination module 302, current detection result acquisition module 304, target fault propagation path determination module 306, and fault cause determination module 308 correspond to steps S102 to S110 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run on a computer terminal.

[0064] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0065] The aforementioned industrial system fault cause determination device may further include a processor and a memory. The aforementioned industrial equipment and historical status data acquisition module 300, cause-effect graph determination module 302, current detection result acquisition module 304, target fault propagation path determination module 306, fault cause determination module 308, etc., are all stored in the memory as program modules, and the processor executes the aforementioned program modules stored in the memory to realize the corresponding functions.

[0066] The processor contains a core that retrieves the corresponding program modules from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0067] According to an embodiment of this application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program runs, it controls the device where the non-volatile storage medium is located to execute any of the aforementioned industrial system fault cause determination methods.

[0068] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the non-volatile storage medium includes stored programs.

[0069] Optionally, during program execution, the device containing the non-volatile storage medium performs the following functions: acquiring multiple industrial devices in the industrial system, and historical status data of the multiple industrial devices, wherein the historical status data includes historical operating parameter data and historical operating condition data; determining a causal graph based on the multiple industrial devices and historical status data, wherein the causal graph is used to indicate the causal relationship between the multiple industrial devices and historical status data; acquiring the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an anomaly in the industrial system at the current moment; if the current detection result indicates that there is an anomaly, determining the target fault propagation path leading to the anomaly based on the causal graph; and determining the fault cause leading to the anomaly based on the target fault propagation path.

[0070] According to an embodiment of this application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any of the above-described methods for determining the cause of failure in an industrial system.

[0071] According to an embodiment of this application, an embodiment of a computer program product is also provided. Optionally, in this embodiment, the computer program product includes a computer program that, when executed by a processor, implements the steps of the method for determining the cause of failure in an industrial system as described above.

[0072] Optionally, when the aforementioned computer program product is executed on a data processing device, it is suitable to execute an initialization program with the following method steps: acquiring multiple industrial devices in an industrial system, and historical state data of the multiple industrial devices, wherein the historical state data includes historical operating parameter data and historical operating condition data; determining a causal graph based on the multiple industrial devices and the historical state data, wherein the causal graph is used to indicate the causal relationship between the multiple industrial devices and the historical state data; acquiring the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an anomaly in the industrial system at the current moment; if the current detection result indicates the existence of an anomaly, determining the target fault propagation path leading to the anomaly based on the causal graph; and determining the fault cause leading to the anomaly based on the target fault propagation path.

[0073] This invention provides an electronic device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring multiple industrial devices in an industrial system, and historical state data of the multiple industrial devices, wherein the historical state data includes historical operating parameter data and historical operating condition data; determining a causal graph based on the multiple industrial devices and the historical state data, wherein the causal graph is used to indicate the causal relationship between the multiple industrial devices and the historical state data; acquiring the current detection result of the industrial system, wherein the current detection result is used to indicate whether there is an anomaly in the industrial system at the current moment; if the current detection result indicates the existence of an anomaly, determining a target fault propagation path leading to the anomaly based on the causal graph; and determining the fault cause leading to the anomaly based on the target fault propagation path.

[0074] The order of the above embodiments of the present invention is merely for description and does not represent the superiority or inferiority of the embodiments.

[0075] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0076] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of modules described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules, and may be electrical or other forms.

[0077] The modules described above as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0078] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0079] If the aforementioned integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0080] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A failure cause determination method of an industrial system, characterized by, The method comprises the following steps: obtaining a plurality of industrial devices in an industrial system and historical state data of the plurality of industrial devices, wherein the historical state data comprises historical operating parameter data and historical working condition data; determining a causal graph based on the plurality of industrial devices and the historical state data, wherein the causal graph is used to indicate the causal correlation between the plurality of industrial devices and the historical state data; obtaining a current detection result of the industrial system, wherein the current detection result is used to indicate whether an abnormal phenomenon exists in the industrial system at a current time; in a case where the current detection result indicates that the abnormal phenomenon exists, determining a target fault propagation path causing the abnormal phenomenon based on the causal graph; determining a fault cause causing the abnormal phenomenon based on the target fault propagation path.

2. The method of claim 1, wherein, The method further comprises the following steps: standardizing the historical state data to obtain standard historical state data; dividing the plurality of industrial devices and the standard historical state data to obtain three types of nodes, wherein the three types of nodes comprise a first type of node, a second type of node and a third type of node, the first type of node comprises the plurality of industrial devices, the second type of node comprises historical operating parameters corresponding to the plurality of industrial devices respectively, and the third type of node comprises historical working condition data corresponding to the plurality of industrial devices respectively; obtaining a plurality of causal edges based on the three types of nodes, wherein the causal edge is used to quantify the causal correlation between any two nodes of different types, the causal edge comprises a causal direction, a causal strength and a delay time, the causal direction is used to indicate a starting node and an ending node of the corresponding causal edge, the causal strength is used to quantify the correlation degree between the starting node and the ending node of the corresponding causal edge, and the delay time represents a time difference between the starting node and the ending node of the corresponding causal edge; classifying the three types of nodes based on the plurality of causal edges to obtain a plurality of levels, wherein the plurality of levels comprise an abnormal phenomenon layer, a parameter abnormality layer, a device state abnormality layer and a component fault layer, the abnormal phenomenon layer represents a set of nodes directly detected to be associated with historical abnormal phenomena of the industrial system, the parameter abnormality layer represents a set of nodes associated with historical operating parameter change abnormalities of the plurality of industrial devices, the device state abnormality layer represents a set of nodes associated with historical state abnormalities of the plurality of industrial devices, and the component fault layer represents a set of nodes associated with historical faults of internal components of the plurality of industrial devices; obtaining the causal graph based on the plurality of causal edges and the plurality of levels.

3. The method of claim 2, wherein, The method further comprises the following steps: obtaining the causal strength of any causal edge based on the three types of nodes by the following way: detecting whether the any causal edge meets a predetermined causal correlation standard; In a case where the any causal edge meets the predetermined causal correlation standard, a mechanism weight and a data weight of the any causal edge are determined based on the three types of nodes, where the mechanism weight is used to quantify a reliability degree of the any causal edge relative to the predetermined causal correlation standard, and the data weight is used to quantify a correlation degree between the start node and the end node of the any causal edge in a historical operation process of the industrial system. A causal strength of the any causal edge is obtained based on the mechanism weight and the data weight. The causal strength of the any causal edge is obtained in a manner that the causal strength of the any causal edge is obtained.

4. The method of claim 3, wherein, After the detection of whether the any causal edge meets the predetermined causal correlation standard, the method further includes: In a case where the any causal edge does not meet the predetermined causal correlation standard, a target causal edge is determined, where the target causal edge represents a causal edge with mutual information greater than a preset mutual information threshold, and the mutual information is used to quantify a mutual dependence degree between the start node and the end node of the corresponding causal edge; A preset mechanism weight and a preset data weight of the target causal edge are obtained; A causal strength of the target causal edge is determined based on the preset mechanism weight and the preset data weight.

5. The method of claim 1, wherein, In a case where the current detection result indicates that the abnormal phenomenon exists, a target fault propagation path causing the abnormal phenomenon is determined based on the causal graph, including: In a case where the current detection result indicates that the abnormal phenomenon exists, a target abnormal node is determined, where the target abnormal node represents a node matching the abnormal phenomenon in an abnormal phenomenon layer of the causal graph, and the abnormal phenomenon layer represents a set of nodes directly detected as being associated with a historical abnormal phenomenon of the industrial system; The causal graph is traversed based on the target abnormal node to obtain an initial fault propagation path, where the initial fault propagation path represents a path composed of nodes associated with the target abnormal node in the causal graph; The initial fault propagation path is screened to obtain a screened fault propagation path; The screened fault propagation path is subjected to intervention verification processing to obtain the target fault propagation path, where the intervention verification processing represents a process of verifying a real degree of causal relationships in the screened fault propagation path by simulating a normal operation state of the industrial system.

6. The method of claim 5, wherein, The screening of the initial fault propagation path to obtain the screened fault propagation path includes: Based on a preset causal strength threshold, a plurality of causal edges in the initial fault propagation path are screened to obtain a plurality of screened causal edges, where the screened causal edge represents a causal edge with a causal strength greater than the preset causal strength threshold in the initial fault propagation path; The screened fault propagation path is obtained based on the plurality of screened causal edges and nodes connected by the plurality of screened causal edges.

7. The method of claim 5, wherein, In a case where the screened fault propagation path is obtained based on a plurality of screened causal edges, the intervention verification processing on the screened fault propagation path is performed to obtain the target fault propagation path, comprising: The intervention verification processing on any screened causal edge is performed to obtain any intervened causal edge by the following way: An intervention operation is performed on the terminal node of the any screened causal edge to obtain a terminal running state of the any screened causal edge, wherein the terminal running state represents a running state of the terminal node of the corresponding causal edge; In a case where the terminal running state of the any screened causal edge is normal running, an initial running state of the any screened causal edge is obtained, wherein the initial running state represents a running state of the initial node of the corresponding causal edge; In a case where the initial running state of the any screened causal edge is normal running, the any screened causal edge is determined as the any intervened causal edge; The plurality of intervened causal edges are obtained by the way of obtaining the any intervened causal edge; The target fault propagation path is obtained based on the plurality of intervened causal edges and nodes connected by the plurality of intervened causal edges.

8. An apparatus for determining a cause of failure of an industrial system, characterized by comprising: Comprise: An industrial equipment and historical state data acquisition module is configured to acquire a plurality of industrial equipments in an industrial system and historical state data of the plurality of industrial equipments, wherein the historical state data comprises historical running parameter data and historical working condition data; A causal graph determination module is configured to determine a causal graph based on the plurality of industrial equipments and the historical state data, wherein the causal graph is used to indicate a causal correlation between the plurality of industrial equipments and the historical state data; A current detection result acquisition module is configured to acquire a current detection result of the industrial system, wherein the current detection result is used to indicate whether an abnormal phenomenon exists in the industrial system at a current time; A target fault propagation path determination module is configured to determine a target fault propagation path causing the abnormal phenomenon based on the causal graph in a case where the current detection result indicates that the abnormal phenomenon exists; A fault cause determination module is configured to determine a fault cause causing the abnormal phenomenon based on the target fault propagation path.

9. A non-volatile storage medium, comprising: The non-volatile storage medium stores a plurality of instructions, and the instructions are adapted to be loaded and executed by a processor to implement the fault cause determination method of the industrial system in any one of claims 1 to 7.

10. An electronic device, comprising: Comprise one or more processors and memories, and the memories store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the fault cause determination method of the industrial system in any one of claims 1 to 7.

Images