Internet equipment fault prediction system based on artificial intelligence

Abstract

The invention relates to the technical field of Internet equipment fault prediction, in particular to an Internet equipment fault prediction system based on artificial intelligence, which comprises a fault management center, an equipment data preparation module, a data sensing module, a fault propagation screening module, a prediction and correction module, a fault score calculation module and a fault early warning output module. According to the method, the association knowledge graph containing the element attributes and the association relationship weights is constructed, and the link search and validity screening mechanism of the fault propagation screening module is combined, so that the effective link in which the fault may be propagated can be accurately positioned from the initial degraded element; meanwhile, the predicted residual life and the fault probability of each element are corrected through a cooperative influence correction formula, prediction deviation caused by isolated analysis is effectively eliminated, operation and maintenance personnel can lock a fault source and clarify a risk level and a processing direction by constructing a fault prediction list, and the efficiency and pertinence of equipment operation and maintenance are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of internet equipment failure prediction, and particularly relates to an internet equipment failure prediction system based on artificial intelligence. BACKGROUND

[0002] With the rapid development of internet technology, internet equipment has become the core infrastructure of communication networks, data centers, industrial internet and other scenarios. Such equipment is usually composed of multiple core elements such as processors, memories, power modules, communication interfaces and cooling systems, and there is a close logical association and data interaction between the elements. Once a component performance degrades or fails, the fault will quickly spread through the connection link between the components, triggering a chain reaction, eventually causing the entire device to shut down, resulting in serious economic losses.

[0003] Currently, the failure detection of internet equipment mainly relies on two modes of post-maintenance or regular preventive maintenance. Some existing technologies attempt to collect component operating parameters through sensors and use machine learning models for failure prediction, but still have the following defects: on the one hand, only the parameter changes of a single component are analyzed in isolation, without considering the failure propagation relationship between components, which cannot identify the chain risk of failure propagation link, on the other hand, the existing failure prediction model lacks a collaborative correction mechanism for the prediction of component remaining life and failure probability. Since the state of the upstream associated components is not considered, the prediction result is prone to deviation, which cannot reflect the interaction of each component in the failure propagation process, thus leading to distortion of the failure score and affecting the accuracy of failure risk judgment.

[0004] Therefore, there is an urgent need for an internet equipment failure prediction system that can construct a global view based on component association, accurately analyze the failure propagation link, and improve the prediction accuracy through collaborative influence correction.

[0005] In view of the above technical defects, a solution is proposed. SUMMARY

[0006] The present application aims to provide an internet equipment failure prediction system based on artificial intelligence to solve the above technical defects. The present application breaks the traditional post-repair mode of failure management through real-time data collection, association analysis, prediction and correction, failure discrimination and early warning, and can predict potential risks before the device actually fails, leaving sufficient disposal time for maintenance personnel, avoiding blind operation, and significantly improving the efficiency and pertinence of device operation and maintenance.

[0007] The objective of this invention can be achieved through the following technical solution: an Internet device fault prediction system based on artificial intelligence, comprising a fault management center, a device data preparation module, a data perception module, a fault propagation screening module, a prediction and correction module, a fault scoring calculation module, and a fault early warning output module;

[0008] The device data preparation module is used to sort out the attribute information and relationship types of each component of the target Internet device, build a relational knowledge graph, and send the relational knowledge graph to the fault management center for storage;

[0009] The data sensing module is used to collect the operating status data of each component in real time and preprocess it to obtain standardized data;

[0010] The fault propagation screening module is used to compare standardized data, screen initially degraded components, identify candidate fault propagation links, and finally retain the effective fault propagation links among the candidate fault propagation links.

[0011] The prediction and correction module is used to perform lifetime prediction and failure probability prediction analysis on each component in each effective failure propagation link, correct the output predicted remaining lifetime and predicted failure probability, and finally output the corrected remaining lifetime and failure probability.

[0012] The fault score calculation module is used to perform overall fault score evaluation and analysis on the effective fault propagation link, output the final fault score SZ, and determine whether the target Internet device has a fault risk or is in a healthy state based on the final fault score SZ.

[0013] When there is a risk of failure, the fault warning output module is used to classify the fault risk level of the target Internet device, locate the fault, and perform fault management analysis, and output a fault prediction list.

[0014] Preferably, the analysis process of the device data preparation module is as follows:

[0015] Identify the components of the target Internet device and establish ontological attribute information for each component. The attribute information includes component model, performance parameter thresholds, normal operating range, historical fault records, design lifespan, and importance weight.

[0016] Based on the hardware connection logic and data interaction process of the target Internet device, define the types of relationships between components, including physical connection relationships, data interaction relationships, and dependency support relationships, and assign a relationship weight to each type of relationship.

[0017] Construct a knowledge graph of the target Internet devices. The knowledge graph is structured by integrating nodes with ontology attributes and edges with relational weights, with components as nodes and relations as edges.

[0018] Preferably, the analysis process of the fault propagation screening module is as follows:

[0019] By comparing the standardized data of each component collected in real time with the normal operating range of the component stored in the associated knowledge graph, the initial deterioration components whose standardized data deviate from the normal operating range are identified.

[0020] Starting with the initial deteriorated component, based on the component association relationship in the association knowledge graph, we search for downstream components that propagate from the initial deteriorated component and generate multiple candidate links for fault propagation.

[0021] Simultaneously, by combining the component association weights in the associated knowledge graph, the effectiveness of the fault propagation candidate links is screened, and the fault propagation candidate links with association weights lower than the preset threshold are eliminated, while the remaining fault propagation candidate links are set as valid fault propagation links.

[0022] Preferably, the analysis process of the prediction and correction module is as follows:

[0023] Retrieve the preset lifetime prediction sub-model and the preset failure probability prediction sub-model;

[0024] Real-time feature data is extracted from the standardized data of each component in each effective fault propagation link. The real-time feature data is then input into the preset lifetime prediction sub-model and the preset fault probability prediction sub-model, and the predicted remaining lifetime and predicted fault probability of each component are output.

[0025] Preferably, based on the correlation weights between components, a collaborative correction is made to the predicted remaining lifetime and the preset failure probability, specifically including:

[0026] Based on the association weights in the device association knowledge graph, a component association matrix is ​​constructed within the effective fault propagation link.

[0027] For a given target component, the correlation weighting coefficient of its upstream related components is calculated based on the correlation matrix. Combined with the predicted remaining lifetime of the upstream components, the predicted remaining lifetime of the target component is corrected. The correction formula is as follows: Where RSi is the corrected remaining lifetime of target component i, RCI is the initial predicted remaining lifetime of target component i, wji is the correlation weight between upstream component j and target component i, RCj is the initial predicted remaining lifetime of upstream component j, RRj is the design life of upstream component j, and n is the number of upstream related components of target component i, n≥0.

[0028] The preset failure probability of the target component is corrected using the following formula: , where PSi is the corrected failure probability of target component i, PCi is the initial preset failure probability of target component i, PCj is the initial preset failure probability of upstream component j, and α is the preset cooperative influence coefficient.

[0029] Preferably, the analysis process of the fault scoring calculation module is as follows:

[0030] For each effective fault propagation link, the overall fault score SZk of the link is calculated by weighted summation based on the corrected remaining lifetime and fault probability of each component in the link. The overall fault scores SZk of all effective fault propagation links are compared, and the maximum value is taken as the final fault score SZ of the device.

[0031] Obtain the preset equipment fault score threshold S, and compare and analyze the final fault score SZ with the preset equipment fault score threshold S. If the final fault score SZ ≥ the preset equipment fault score threshold S, the equipment is determined to have a fault risk. If the final fault score SZ < the preset equipment fault score threshold S, the equipment is determined to be in a healthy state.

[0032] Preferably, the analysis process of the fault early warning output module is as follows:

[0033] If the preset equipment fault score threshold S ≤ the final fault score SZ < the preset fault score threshold, it is judged as low fault risk; if the final fault score SZ ≥ the preset fault score threshold, it is judged as high fault risk.

[0034] The effective fault propagation link of the final fault score SZ is set as the fault propagation link, and the initial deteriorated element in the fault propagation link is selected as the root cause element.

[0035] The type of the root source component is obtained, including replaceable and non-replaceable types. If the root source component is a replaceable type, the basic information package of the root source component is obtained. If the root source component is a non-replaceable type, the preset processing decision of the root source component is obtained.

[0036] A fault prediction list is constructed based on the fault risk level (indicating low fault risk / high fault risk) or health status, fault propagation path, root cause component, and root cause component type.

[0037] The beneficial effects of this invention are as follows:

[0038] (1) By constructing a knowledge graph containing component attributes and associated weights, and combining the link search and validity screening mechanism of the fault propagation screening module, this invention can accurately locate the effective links that may propagate the fault from the initial deteriorated component, and can completely restore the fault propagation path, avoiding the failure to judge or misjudge the fault due to ignoring the component association. At the same time, based on the association weight in the knowledge graph, a component association matrix is ​​constructed, and the predicted remaining life and failure probability of each component are corrected by the collaborative influence correction formula, which effectively eliminates the prediction bias caused by isolated analysis, and makes the corrected remaining life and failure probability more in line with the actual operating logic of the equipment, providing accurate data support for fault scoring.

[0039] (2) The present invention adopts a link-level weighted summation scoring method, which can comprehensively and objectively reflect the true fault risk level of the equipment. Furthermore, the construction of a fault prediction list enables maintenance personnel to quickly identify the root cause of the fault, clarify the risk level and handling direction, avoid blind maintenance, effectively shorten the fault handling time, reduce maintenance costs, and significantly improve the efficiency and pertinence of equipment maintenance. Attached Figure Description

[0040] The invention will now be further described with reference to the accompanying drawings;

[0041] Fig. 1 This is a flowchart of the system of the present invention;

[0042] Fig. 2 This is an analysis reference diagram for the fault propagation screening module;

[0043] Fig. 3 This is an analysis reference diagram for the fault scoring calculation module. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0045] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments;

[0046] Example 1:

[0047] Please see Figs. 1 to 3 As shown, the present invention is an artificial intelligence-based Internet device fault prediction system, including a fault management center, a device data preparation module, a data perception module, a fault propagation screening module, a prediction and correction module, a fault score calculation module, and a fault early warning output module. The fault management center and the device data preparation module have a bidirectional communication connection, the data perception module and the fault management center have a unidirectional communication connection, the fault management center and the fault propagation screening module and the fault early warning output module have both unidirectional communication connections, the fault propagation screening module and the prediction and correction module have a unidirectional communication connection, the prediction and correction module and the fault score calculation module have a unidirectional communication connection, and the fault score calculation module and the fault management center have a unidirectional communication connection.

[0048] The device data preparation module is used to organize the attribute information and relationship types of each component of the target Internet device, and construct a relational knowledge graph, specifically including:

[0049] The components of the target Internet device are identified, including processors, memory, power modules, communication interfaces, cooling fans, hard drives, etc. On-site attribute information is established for each component, including component model, performance parameter thresholds, normal operating range, historical fault records, design lifespan, importance weight, etc.

[0050] Based on the hardware connection logic and data interaction process of the target Internet device, the types of relationships between components are defined, including physical connection relationships (such as power supply modules supplying power to the processor), data interaction relationships (such as read and write interaction between memory and processor), and dependency and support relationships (such as cooling fans cooling the processor). At the same time, each type of relationship is assigned a relationship weight, with the relationship weight ranging from (0, 1). The higher the weight value, the closer the relationship between the two components.

[0051] A knowledge graph of connections to target internet devices is constructed. This knowledge graph uses components as nodes and connections as edges. Nodes with ontology attributes are integrated with edges carrying connection weights to form a structured knowledge graph. The fault management center stores this knowledge graph.

[0052] The data sensing module is used to collect real-time operating status data of each component and preprocess it to obtain standardized data, specifically including:

[0053] Real-time acquisition of operating status data for each component, including component temperature, operating voltage, current, data transmission rate, load rate, etc.

[0054] The collected operational status data is preprocessed (e.g., cleaned, outlier removal) to obtain standardized data, which is then stored by the fault management center.

[0055] The fault propagation screening module is used to compare standardized data, screen initially degraded components, identify candidate fault propagation paths, and ultimately retain valid fault propagation paths from the candidate paths. Specifically, this includes:

[0056] By comparing the standardized data of each component collected in real time with the normal operating range of the component stored in the associated knowledge graph, the initial deterioration components whose standardized data deviate from the normal operating range are identified.

[0057] Starting with the initial deteriorated component, based on the component association relationship in the association knowledge graph, search for downstream components that propagate from the initial deteriorated component, and generate multiple candidate links for fault propagation, such as: power module deterioration - insufficient processor power supply - memory read / write error - communication interface data transmission interruption.

[0058] Simultaneously, by combining the component correlation weights in the related knowledge graph, the effectiveness of the fault propagation candidate links is screened, and the fault propagation candidate links with correlation weights lower than a preset threshold (such as 0.2) are eliminated, while the remaining fault propagation candidate links are set as valid fault propagation links.

[0059] Example 2:

[0060] The prediction and correction module is used to perform lifetime prediction and failure probability prediction analysis on each component in each effective failure propagation link, correct the output predicted remaining lifetime and predicted failure probability, and finally output the corrected remaining lifetime and failure probability, specifically including:

[0061] Retrieve the preset lifetime prediction sub-model and the preset failure probability prediction sub-model;

[0062] Extract real-time feature data from the standardized data of each component in each effective fault propagation link, input the real-time feature data into the preset lifetime prediction sub-model and the preset fault probability prediction sub-model, and output the predicted remaining lifetime and predicted fault probability of each component.

[0063] Among them, the predicted remaining lifetime represents the estimated time from the current moment until the component fails, and the preset failure probability represents the probability that the component will fail within a preset time window in the future.

[0064] Based on the correlation weights between components, a synergistic effect correction is applied to the predicted remaining lifetime and the preset failure probability to eliminate the bias of isolated analysis. Specifically, this includes:

[0065] Based on the association weights in the device association knowledge graph, a component association matrix is ​​constructed within the effective fault propagation link. The elements in the matrix represent the degree of association between two corresponding components.

[0066] For a given target component, the correlation weighting coefficient of its upstream related components is calculated based on the correlation matrix. Combined with the predicted remaining lifetime of the upstream components, the predicted remaining lifetime of the target component is corrected. The correction formula is as follows: Where RSi is the corrected remaining lifetime of target component i, RCI is the initial predicted remaining lifetime of target component i, wji is the correlation weight between upstream component j and target component i, RCj is the initial predicted remaining lifetime of upstream component j, RRj is the design life of upstream component j, n is the number of upstream related components of target component i, n≥0, j≥0.

[0067] The preset failure probability of the target component is corrected using the following formula: Where PSi is the corrected failure probability of target element i, PCi is the initial preset failure probability of target element i, wji is the correlation weight between upstream element j and target element i, PCj is the initial preset failure probability of upstream element j, α is the preset synergistic influence coefficient, α∈(0,0.5), and n is the number of upstream associated elements of target element i.

[0068] The fault score calculation module is used to perform overall fault score evaluation and analysis on effective fault propagation links, outputting a final fault score SZ, and determining whether the target Internet device has a fault risk or is in a healthy state based on the final fault score SZ, specifically including:

[0069] For each valid fault propagation link, the overall fault score SZk of the link is calculated using a weighted summation method based on the corrected remaining lifetime and fault probability of each component within the link, as shown in the following formula: Where SZk is the overall fault score of the k-th fault propagation link, m is the number of components in the k-th link, m is a natural number greater than zero, β is the preset proportional coefficient of the reciprocal of the corrected remaining lifetime of target component i, ε is the preset proportional coefficient of the corrected fault probability of target component i, β+ε=1, both β and ε are greater than zero, and ai is the importance weight of the i-th component in the link, which can be set according to the degree of influence of the component on the function of the equipment.

[0070] The overall fault scores SZk of all effective fault propagation links are compared, and the maximum value is taken as the final fault score SZ of the device. The reason for selecting the maximum value is that the higher the score of the fault propagation link, the higher the fault risk of the link. The overall fault risk of the device is determined by the link with the highest risk.

[0071] Obtain the preset equipment fault score threshold S, compare and analyze the final fault score SZ with the preset equipment fault score threshold S, and if the final fault score SZ ≥ the preset equipment fault score threshold S, then the equipment is determined to have a fault risk.

[0072] If the final fault score SZ < the preset device fault score threshold S, the device is determined to be in a healthy state. The fault management center responds to the fault risk or health status by marking the target Internet device. The marking includes marking fault risk as red and health status as green.

[0073] When a fault risk exists, the fault early warning output module is used to classify the fault risk level of the target Internet device, locate the fault, and perform fault management analysis, outputting a fault prediction list, which specifically includes:

[0074] If the preset equipment fault score threshold S ≤ the final fault score SZ < the preset fault score threshold, it is judged as low fault risk; if the final fault score SZ ≥ the preset fault score threshold, it is judged as high fault risk.

[0075] The effective fault propagation link of the final fault score SZ is set as the fault propagation link, and the initial deteriorated element in the fault propagation link is selected as the root cause element.

[0076] The type of the root cause component is obtained, including replaceable and non-replaceable types. If the root cause component is replaceable, the basic information package such as the model and parameters of the root cause component is obtained. If the root cause component is non-replaceable, the preset processing decision of the root cause component is obtained, including operation and maintenance measures and operation and maintenance periods.

[0077] Based on the fault risk level (representing low fault risk / high fault risk) or health status, fault propagation path, root cause component and root cause component type, a fault prediction list is constructed, the fault prediction list is output and displayed on the backend visual platform, so as to intuitively understand the fault status of the target Internet device, and thus facilitate targeted management.

[0078] In summary, this invention constructs a knowledge graph containing component attributes and associated weights, combined with the link search and validity screening mechanism of the fault propagation screening module. Starting from the initially deteriorated component, it can accurately locate the effective links through which faults may propagate, and completely reconstruct the fault propagation path, avoiding missed or incorrect fault detections due to neglecting component associations. Furthermore, based on the association weights in the knowledge graph, a component association matrix is ​​constructed, and the predicted remaining lifespan and failure probability of each component are corrected using a collaborative influence correction formula. This effectively eliminates prediction biases caused by isolated analysis, making the corrected remaining lifespan and failure probability more closely aligned with the actual operating logic of the equipment, providing accurate data support for fault scoring. The link-level weighted summation scoring method comprehensively and objectively reflects the true fault risk level of the equipment. The constructed fault prediction list enables maintenance personnel to quickly pinpoint the root cause of the fault, clarify the risk level and handling direction, avoid blind maintenance, effectively shorten fault handling time, reduce maintenance costs, and significantly improve the efficiency and targeting of equipment maintenance.

[0079] The threshold is set for comparative analysis of results to determine whether they are good or bad. The value of the threshold is determined by a combination of large-scale model analysis of sample data and human experience. It can also be adjusted appropriately based on seasonal or common-sense influencing factors.

[0080] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based internet device failure prediction system, characterized by, The system comprises a fault management center, a device data preparation module, a data perception module, a fault propagation screening module, a prediction and correction module, a fault score calculation module, and a fault early warning output module. The device data preparation module is used to sort out the attribute information and the correlation relationship type of each element of the target Internet device, construct a correlation knowledge graph, and send the correlation knowledge graph to the fault management center for storage. The data perception module is used to collect the running state data of each element in real time and obtain standardized data through preprocessing. The fault propagation screening module is used to compare the standardized data, screen the initial deteriorated elements, identify the candidate fault propagation links, and finally retain the effective fault propagation links in the candidate fault propagation links. The prediction and correction module is used to perform life prediction and fault probability prediction analysis on each element in each effective fault propagation link, correct the output predicted remaining life and predicted fault probability, and finally output the corrected remaining life and fault probability. The fault score calculation module is used to perform link overall fault score evaluation analysis on the effective fault propagation links, output the final fault score SZ, and determine whether the target Internet device has a fault risk or is in a healthy state based on the final fault score SZ. When there is a fault risk, the fault early warning output module is used to divide the target Internet device into fault risk levels, locate the fault, and perform fault management analysis, and output a fault prediction list.

2. The system as claimed in claim 1, wherein, The analysis process of the device data preparation module is as follows: Sort out each element of the target Internet device, establish the ontology attribute information for each element, and the attribute information includes element model, performance parameter threshold, normal working interval, historical fault record, designed service life, and importance weight. Based on the hardware connection logic and data interaction process of the target Internet device, define the correlation relationship type between elements, including physical connection relationship, data interaction relationship, and dependent support relationship, and assign a correlation degree weight to each type of correlation relationship. Construct the correlation knowledge graph of the target Internet device, which takes elements as nodes and correlation relationships as edges, integrates nodes with ontology attributes and correlation relationship edges with correlation degree weights, and forms a structured correlation knowledge graph. 3.The AI-based internet device failure prediction system of claim 1, wherein The analysis process of the fault propagation screening module is as follows: Compare the real-time collected standardized data of each element with the normal working interval of the element stored in the correlation knowledge graph, and identify the initial deteriorated elements whose standardized data deviate from the normal working interval. Take the initial deteriorated element as the starting point, search for the downstream elements propagated by the initial deteriorated element based on the element correlation relationship in the correlation knowledge graph, and generate multiple candidate fault propagation links. Meanwhile, in combination with the element correlation degree weight in the correlation knowledge graph, the candidate fault propagation links are screened for effectiveness, and the candidate fault propagation links with a correlation degree weight lower than a preset threshold are removed, and the remaining candidate fault propagation links are set as effective fault propagation links.

4. The system as claimed in claim 1, wherein, The analysis process of the prediction and correction module is as follows: Call the preset life prediction sub-model and the preset fault probability prediction sub-model. Real-time feature data in the standardized data of each element in each effective fault propagation link is extracted, and the real-time feature data is input into a preset service life prediction sub-model and a preset fault probability prediction sub-model to output a predicted residual service life and a predicted fault probability of each element.

5. The system as claimed in claim 4, wherein, The predicted residual service life and the preset fault probability are corrected based on the correlation degree weight between elements, and the correction specifically includes: An element correlation degree matrix in the effective fault propagation link is constructed based on the correlation degree weight in the device correlation knowledge graph. For a target element, the correlation degree weighting coefficients of the upstream associated elements are calculated according to the correlation degree matrix, and the predicted residual life of the target element is corrected in combination with the predicted residual life of the upstream elements, and the correction formula is as follows: Wherein, RSi is the corrected residual life of the target element i, RCi is the initial predicted residual life of the target element i, wji is the correlation degree weight of the upstream element j and the target element i, RCj is the initial predicted residual life of the upstream element j, RRj is the designed service life of the upstream element j, n is the number of the upstream associated elements of the target element i, and n≥0. The preset failure probability of the target element is corrected, and a correction formula is as follows: Wherein, PSi is the corrected failure probability of the target element i, PCi is the initial preset failure probability of the target element i, PCj is the initial preset failure probability of the upstream element j, and a is a preset synergistic influence coefficient. 6.The AI-based internet device failure prediction system of claim 1, wherein The analysis process of the fault score calculation module is as follows: For each effective fault propagation link, the residual service life and the fault probability of each element in the link are corrected, and a weighted summation method is used to calculate the overall fault score SZk of the link, and the overall fault scores SZk of all effective fault propagation links are compared, and the maximum value is taken as the final fault score SZ of the device. A preset device fault score threshold S is obtained, and the final fault score SZ is compared with the preset device fault score threshold S, if the final fault score SZ is greater than or equal to the preset device fault score threshold S, it is determined that the device has a fault risk, and if the final fault score SZ is less than the preset device fault score threshold S, it is determined that the device is currently in a healthy state.

7. The system as claimed in claim 1, wherein, The analysis process of the fault warning output module is as follows: If the preset device fault score threshold S is less than or equal to the final fault score SZ and greater than the preset fault score critical value, it is determined as a low fault risk, and if the final fault score SZ is greater than or equal to the preset fault score critical value, it is determined as a high fault risk; The effective fault propagation link of the final fault score SZ is set as a fault propagation link, and the initial deteriorated element in the fault propagation link is selected as a root element; The type of the root element is obtained, including a replaceable type and a non-replaceable type, if the root element is of the replaceable type, a basic information package of the root element is obtained, and if the root element is of the non-replaceable type, a preset processing decision of the root element is obtained; A fault prediction list is constructed based on the fault risk level or the health state, the fault propagation link, the root element, and the type of the root element.

Images