A method for failure recurrence risk assessment based on maintenance closed-loop data

Abstract

The application discloses a kind of based on maintenance closed-loop data's fault recurrence risk assessment method;The method obtains the fault diagnosis data of target equipment, maintenance action data, maintenance verification data, recovery operation data and subsequent monitoring data, constructs fault cause chain and maintenance intervention-verification response matrix, according to this, eliminate node and residual node are divided and form fault observation object;According to the operating exposure of fault action mechanism, after reaching observation start threshold, form node activation sequence according to the parameter overrun starting time of fault node, identify eliminated node recurrence event, residual node propagation recurrence event or independent new fault event;Combined with operating exposure, verification parameter rollback state and historical maintenance closed-loop event to determine recurrence risk, and generate corresponding maintenance management work order.The application can improve the pertinence of fault recurrence identification and maintenance disposal.

Description

Technical Field

[0001] This invention relates to the field of equipment operation and maintenance management and fault diagnosis, specifically to a fault recurrence risk assessment method based on maintenance closed-loop data. Background Technology

[0002] During long-term operation, industrial equipment is susceptible to malfunctions such as abnormal vibration, abnormal temperature rise, pressure fluctuations, and performance degradation due to factors such as load fluctuations, start-up and shutdown shocks, changes in lubrication conditions, media contamination, component wear, and environmental stress. After a malfunction occurs, the fault diagnosis system typically records the fault code and abnormal parameters. Maintenance personnel then perform repairs such as component replacement, adjustment, cleaning, or media replenishment according to the maintenance work order. After repairs are completed, trial runs or parameter retests are conducted. With the increasing informatization of equipment management, fault diagnosis data, maintenance work order data, post-repair verification data, and equipment operation monitoring data can now be stored separately, providing a data foundation for analyzing equipment fault recurrence.

[0003] Existing methods for assessing fault recurrence typically determine whether a fault has recurred based on the frequency of repeated occurrences of the same fault code, the time between adjacent fault occurrences, cumulative runtime, number of repairs, or changes in equipment health. Some methods also establish fault propagation maps based on the relationships between faulty nodes, or calculate the probability of subsequent equipment failures using maintenance history and operational monitoring parameters. However, the maintenance targets recorded in maintenance work orders may not cover all faulty nodes in the original fault propagation process, and normal trial operation after maintenance does not prove that all faulty nodes have been restored. If upstream faulty nodes that have not been repaired or lack effective verification continue to exist, their effects may propagate downstream again after the equipment resumes operation; nodes that have already been repaired may also fail again due to changes in maintenance quality, coordination status, or operating conditions. Relying solely on fault codes, number of repairs, or overall equipment health makes it difficult to distinguish between the recurrence of repaired nodes, the continued propagation of unresolved nodes, and independent new faults unrelated to the original fault.

[0004] Furthermore, different faults exhibit varying cumulative relationships with operating conditions in their formation and recurrence. For instance, wear-related faults are associated with load, rotational speed, and cumulative operating volume; thermal damage faults are related to over-temperature amplitude and duration; and fatigue-related faults are related to load cycle amplitude and cycle count. Existing methods often rely on natural operating time, uniform cumulative duration, or general risk indicators as observation criteria, failing to select corresponding operating parameters based on potential fault nodes and their mechanisms after maintenance. This can easily lead to the misinterpretation of equipment that did not experience faults during periods of insufficient operational exposure as successfully maintained. Simultaneously, existing methods underutilize the temporal sequence of abnormal occurrences at various fault nodes after maintenance and their relationship to the original fault propagation direction. Fault recurrence risk typically only corresponds to the overall equipment or fault codes, making it difficult to further identify specific fault nodes requiring re-inspection and verification items. This impacts the targeted nature of maintenance procedures and the accuracy of equipment operation management. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a fault recurrence risk assessment method based on maintenance closed-loop data, thereby resolving the technical problems existing in the prior art.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A method for assessing the risk of fault recurrence based on maintenance closed-loop data includes the following steps: S1: Obtain fault diagnosis data, maintenance action data, post-maintenance verification data, recovery operation data and subsequent monitoring data corresponding to a fault maintenance event of the target equipment. Based on the fault diagnosis data and the component connection relationship, energy transfer relationship and fault propagation relationship of the target equipment, construct a fault cause chain including the fault node and the directed connection relationship between the fault nodes. S2: Map the maintenance objects and maintenance methods in the maintenance action data, as well as the verification items and verification parameters in the post-maintenance verification data, to the fault nodes respectively, and establish a maintenance intervention-verification response matrix; based on whether the maintenance action directly changes the fault state of the fault node and whether the post-maintenance verification parameters meet the recovery conditions determined by the fault-free operation data, divide the fault nodes into eliminated nodes and residual nodes, and form at least one fault observation object based on the eliminated nodes, residual nodes, and directed connection relationships; S3: Based on the fault mechanism of the fault node in the fault observation object, select the operating parameters that can promote the accumulation of the corresponding fault state from the recovery operation data, and determine the operation exposure of the fault observation object. S4: After the operational exposure reaches the observation start threshold determined by historical maintenance closed-loop events or reliability test data, a subsequent node activation sequence is formed based on the start time of the exceeding limit of the monitoring parameters corresponding to each fault node in the subsequent monitoring data. Based on the node status of the first active node in the subsequent node activation sequence and the relationship between the node activation order and the directed connection relationship, the eliminated node recurrence event, the residual node propagation recurrence event, or the independent new fault event is determined. S5: Based on the operational exposure, the rollback status of post-maintenance verification parameters, and recurrence events in historical maintenance closed-loop events, determine the recurrence risk of eliminated nodes and the recurrence risk of residual node propagation, respectively. Based on the recurrence risk that meets the maintenance disposal conditions determined by historical recurrence events, determine the faulty nodes to be re-inspected and the corresponding verification items, and generate maintenance management work orders.

[0007] Preferably, the data corresponding to a single fault repair event of the target equipment is acquired, including: The system obtains equipment identification, fault occurrence time, fault location, fault symptoms, monitoring parameter over-limit records, and fault cause records from the equipment fault diagnosis system; it also obtains maintenance work order identification, maintenance object, maintenance operation mode, maintenance start time, and maintenance end time from the maintenance work order system; it further obtains verification items, verification parameters, verification parameter acquisition location, verification completion time, and measurement error from the maintenance verification terminal; and it acquires operating parameters, operating status, and parameter acquisition time after maintenance is completed from the controller and operation monitoring terminal of the target equipment. Based on the equipment identifier, maintenance work order identifier, and parameter acquisition time, data belonging to the same target equipment and the same fault maintenance event are associated as maintenance closed-loop event units; Data acquired after maintenance is completed and before the operational exposure of the corresponding fault observation object reaches the observation start threshold is determined as recovery operation data, and data acquired after the operational exposure reaches the observation start threshold is determined as subsequent monitoring data.

[0008] Preferably, the fault cause chain is constructed, including: The equipment components or functional parts corresponding to the cause of the fault are identified as the causative nodes, the equipment components or functional parts that the fault passes through during the transmission process are identified as the transmission nodes, and the equipment components or functional parts that produce fault symptoms are identified as the manifestation nodes. Determine the physical transmission channel between two fault nodes based on the component connection relationship of the target equipment, and determine the direction of action transmission in the physical transmission channel based on the energy transmission relationship. When the monitoring parameters of the upstream fault node exceed the corresponding normal operating range, and the monitoring parameters of the downstream fault node exceed the corresponding normal operating range within the propagation delay range determined by historical fault data or step response test, and the direction of parameter change of the downstream fault node conforms to the direction of action transmission, it is determined that there is a fault propagation relationship between the two fault nodes. When there is a physical transmission channel, a direction of action transmission, and a fault propagation relationship between two faulty nodes, a directed connection relationship is established from the upstream faulty node to the downstream faulty node.

[0009] Preferably, a maintenance intervention-verification response matrix is ​​established, including: The fault nodes are used as the rows of the matrix, and the maintenance actions and verification items are used as the columns of the matrix; When the maintenance object is the equipment component or functional part corresponding to the fault node, and the maintenance action changes the structural state, coordination state, power supply state or medium state of the fault node, a direct intervention relationship is established between the maintenance action and the fault node. When the interval between the parameter ranges of the verification parameters in the fault state sample and the fault-free state sample is greater than the measurement error of the verification parameters, and the acquisition location of the verification parameters is located at the fault node or the direct output location of the fault node, a direct verification relationship is established between the verification item and the fault node. The direct intervention relationship and direct verification relationship are written into the maintenance intervention-verification response matrix, and the measurement error is extended to both sides of the basic normal interval formed by the fault-free operation data to obtain the recovery judgment interval corresponding to the recovery condition.

[0010] Preferably, the process of dividing eliminated nodes and residual nodes into fault observation objects includes: For fault nodes that have both direct intervention and direct verification relationships, obtain the continuous collection values ​​of the verification parameters corresponding to the fault node after the verification is completed. When the continuously acquired values ​​are all within the recovery judgment range during the verification period, and the distance between the average value of the continuously acquired values ​​and the fault direction boundary is greater than the measurement error, the corresponding fault node is determined as the eliminated node; the remaining fault nodes are determined as residual nodes. When one residual node can reach another residual node along a directed connection, the two residual nodes are assigned to the same residual cause object; residual nodes that are not reachable from other residual nodes in any directed direction are identified as single-node residual cause objects. Each residual causative object and its fault nodes and directed connections to the manifest node are formed into a fault observation object, and each eliminated node is formed into an eliminated node fault observation object based on the directed connection path from each eliminated node to the manifest node. When there are no residual nodes, only observation objects of faults in eliminated nodes are formed.

[0011] Preferably, determining the operational exposure of the fault observation object includes: Based on the failure mechanism of each fault node in the fault observation object, select the operating parameters that can cause wear accumulation, heat accumulation, cyclic impact accumulation, media erosion accumulation or load fatigue accumulation of the corresponding fault node to form an exposure parameter set. For continuously changing operating parameters, the cumulative exposure is determined based on the deviation amplitude and duration of the operating parameter from the baseline operating range along the fault-promoting direction; for cyclically changing operating parameters, the cumulative exposure is determined based on the portion of the actual change amplitude of each operating cycle that exceeds the baseline change amplitude and the number of cycles. The cumulative exposure of each operating parameter is normalized relative to the cumulative exposure when a stable fault symptom is first formed in a similar historical maintenance closed-loop event or reliability test. The normalized sensitivity is determined based on the relationship between the relative changes in operating parameters and the relative changes in monitoring parameters of the corresponding fault nodes. The propagation location coefficient is determined based on the number of directed connections between the fault node and the manifest node. The product of the normalized cumulative exposure, the normalized sensitivity, and the propagation location coefficient is accumulated to obtain the operating exposure.

[0012] Preferably, the observation start threshold and historical maintenance closed-loop events are determined in the following manner: Select historical maintenance closed-loop events from the historical maintenance closed-loop events that have the same fault node composition, the same directed connection order and the same maintenance intervention-verification response relationship as the current fault observation object. Obtain the operational exposure when the selected historical maintenance closed-loop events have recurrence events of eliminated nodes or recurrence events of residual node propagation, and form a set of historical recurrence exposures. The minimum value in the set of historical recurrent exposures is determined as the observation initiation threshold, and the maximum value is determined as the effective observation threshold. When there are no historical maintenance closed-loop events that meet the aforementioned conditions, the observation initiation threshold and effective observation threshold are determined sequentially by using historical maintenance closed-loop events of the same type of equipment with the same first causative node and the same fault mechanism, as well as the reliability test data of the target equipment or the same type of equipment. Historical maintenance closed-loop events that have resulted in recurrence of eliminated nodes or recurrence of residual node propagation are identified as recurrence samples. Historical maintenance closed-loop events that have not resulted in the aforementioned recurrence events and whose operational exposure reaches the effective observation threshold are identified as effective non-recurrence samples. Historical maintenance closed-loop events that have not resulted in the aforementioned recurrence events and whose operational exposure does not reach the effective observation threshold are identified as samples to be observed.

[0013] Preferably, forming a subsequent node activation sequence and identifying recurrence events includes: The continuous acquisition segment in which the monitored parameters continuously exceed the corresponding normal working range, the excess amplitude is greater than the measurement error, and the duration reaches the minimum effective response time is determined as the effective over-limit acquisition segment. The first acquisition time of the effective over-limit acquisition segment is determined as the parameter over-limit start time, and the corresponding fault node is determined as the activation node. Activated nodes are arranged according to the start time of parameter exceeding the limit to form a subsequent node activation sequence, and the fault node with the earliest start time of parameter exceeding the limit among the causal nodes and transit nodes included in the subsequent node activation sequence is determined as the first active node. When the first active node belongs to an eliminated node, and the activation order from the first active node to the active node is consistent with the directed connection direction of the fault cause chain, it is determined to be an eliminated node recurrence event. When the first acting node belongs to the residual cause object, and the activation order from the first acting node to the manifest node is consistent with the direction of the residual cause object and its downstream directed connections, it is determined to be a residual node propagation recurrence event. When the first active node does not belong to the eliminated node or any residual cause object, and there is no directed reachability relationship between the first active node and any residual node in any direction, it is determined as an independent new fault event; When the first active node or the node activation order cannot be determined, it is identified as an event to be confirmed, and supplementary monitoring, disassembly, sampling, or offline detection tasks are generated based on the missing faulty node.

[0014] Preferably, determining the risk of recurrence of eliminated nodes and the risk of recurrence of residual node propagation includes: Excluding the samples to be observed, and selecting relapsed samples and valid non-relapsed samples based on the current operational exposure level, the final observed exposure level is not less than the current operational exposure level. Calculate the proportion of the number of samples that have eliminated node recurrence events and residual node propagation recurrence events when the current operating exposure level is not greater than the total number of selected samples, to obtain the basic recurrence risk of eliminated nodes and the basic recurrence risk of residual node propagation. Based on the initial average value, current sliding average value, and fault direction boundary of the verification parameters after maintenance verification, determine the degree of regression of the verification parameters along the fault direction; Based on the correspondence between the degree of parameter rollback and the increment of operational exposure in historical maintenance closed-loop events, the current rollback degree is converted into additional operational exposure. The additional operational exposure is added to the current operational exposure to obtain the corrected operational exposure. Based on the corrected operational exposure, the risk of recurrence of eliminated nodes and the risk of propagation and recurrence of residual nodes are re-determined. When there are no valid samples available for calculating the risk of recurrence, the corresponding fault observation object is determined to be in a state of insufficient samples, and the risk of recurrence is not assigned to zero.

[0015] Preferably, a maintenance management work order is generated, including: The minimum value among the historical recurrence event risk values ​​calculated using the method of claim 9 is determined as the maintenance and treatment threshold corresponding to the maintenance and treatment conditions. When at least one of the risks of recurrence of eliminated nodes and the risk of recurrence of propagation of residual nodes reaches the maintenance and disposal threshold, the first eliminated node or the first residual node corresponding to the risk that reaches the maintenance and disposal threshold will be identified as the fault node to be re-inspected. When the fault node to be re-inspected has a direct verification relationship, the corresponding verification item is extracted from the maintenance intervention-verification response matrix; when the fault node to be re-inspected does not have a direct verification relationship, disassembly inspection items, sampling items or offline testing items are generated for the fault node to be re-inspected. From the set of exposure parameters corresponding to the fault node to be re-inspected, select controllable operating parameters that can be directly adjusted or indirectly changed by the controller of the target device. Among them, the controllable operating parameters that contribute the most to the current operating exposure are determined as limiting parameters. Based on the limit parameters, generate operating limit instructions along the boundary of the reference working interval in the fault promotion direction and send them to the controller. Obtain maintenance action data and re-inspection data after the maintenance management work order is executed. After confirming that the maintenance action directly changes the fault status of the fault node to be re-inspected and the re-inspection result meets the recovery conditions, update the fault node to be re-inspected to the eliminated node, and re-form the fault observation object and re-accumulate the operation exposure.

[0016] In summary, the present invention has the following main beneficial effects: By mapping fault diagnosis data, maintenance action data, and post-maintenance verification data to fault nodes in the fault cause chain, a maintenance intervention-verification response relationship is established. Based on whether the maintenance action directly changes the fault state of the fault node and whether the verification parameters meet the recovery conditions, eliminated nodes and residual nodes are distinguished. This avoids judging the maintenance completion status solely based on the maintenance work order closure status, maintenance count, or identical fault codes. It enables the accurate identification of residual fault states after maintenance and provides a data foundation consistent with the actual maintenance results for subsequent fault recurrence risk assessment.

[0017] By forming fault observation objects based on the directed connection relationship of eliminated nodes, residual nodes, and fault cause chains, and selecting operating parameters corresponding to the fault action mechanism for different fault observation objects, the operating exposure after equipment maintenance is calculated. After the operating exposure reaches the observation start condition, the recurrence event of eliminated nodes, the recurrence event of residual nodes propagation, or the independent new fault event is determined according to the parameter over-limit start time and node activation sequence of each fault node. This eliminates the interference caused by different operating loads, operating durations, and environmental conditions on recurrence judgment and reduces the possibility of misjudging related faults or independent new faults as recurrences of the original fault.

[0018] By combining operational exposure, the rollback status of post-maintenance verification parameters, and historical maintenance closed-loop events, the risk of fault recurrence is determined. Based on the risk type that meets the maintenance handling conditions, the fault nodes to be re-inspected and the verification items are determined in reverse. At the same time, operational restriction measures are generated for controllable operational parameters that promote the accumulation of fault status. This achieves the effect of implementing the risk assessment results to specific fault nodes, re-inspection items, and equipment operation control processes. This allows the maintenance handling results to update the fault node status and fault observation objects, forming a closed-loop processing process that connects fault diagnosis, maintenance verification, recurrence assessment, equipment control, and maintenance result write-back. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] 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.

[0021] Example 1 refer to Figure 1 A method for assessing the risk of fault recurrence based on maintenance closed-loop data includes the following steps: S1: Obtain fault diagnosis data, maintenance action data, post-maintenance verification data, recovery operation data and subsequent monitoring data corresponding to a fault maintenance event of the target equipment. Based on the fault diagnosis data and the component connection relationship, energy transfer relationship and fault propagation relationship of the target equipment, construct a fault cause chain including the fault node and the directed connection relationship between the fault nodes. S2: Map the maintenance objects and maintenance methods in the maintenance action data, as well as the verification items and verification parameters in the post-maintenance verification data, to the fault nodes respectively, and establish a maintenance intervention-verification response matrix; based on whether the maintenance action directly changes the fault state of the fault node and whether the post-maintenance verification parameters meet the recovery conditions determined by the fault-free operation data, divide the fault nodes into eliminated nodes and residual nodes, and form at least one fault observation object based on the eliminated nodes, residual nodes, and directed connection relationships; S3: Based on the fault mechanism of the fault node in the fault observation object, select the operating parameters that can promote the accumulation of the corresponding fault state from the recovery operation data, and determine the operation exposure of the fault observation object. S4: After the operational exposure reaches the observation start threshold determined by historical maintenance closed-loop events or reliability test data, a subsequent node activation sequence is formed based on the start time of the exceeding limit of the monitoring parameters corresponding to each fault node in the subsequent monitoring data. Based on the node status of the first active node in the subsequent node activation sequence and the relationship between the node activation order and the directed connection relationship, the eliminated node recurrence event, the residual node propagation recurrence event, or the independent new fault event is determined. S5: Based on the operational exposure, the rollback status of post-maintenance verification parameters, and recurrence events in historical maintenance closed-loop events, determine the recurrence risk of eliminated nodes and the recurrence risk of residual node propagation, respectively. Based on the recurrence risk that meets the maintenance disposal conditions determined by historical recurrence events, determine the faulty nodes to be re-inspected and the corresponding verification items, and generate maintenance management work orders.

[0022] This implementation method is applicable to industrial equipment with equipment controller, operation monitoring terminal, equipment fault diagnosis system, maintenance work order system and maintenance verification terminal. The industrial equipment includes pump units, fans, compressors, machine tool spindles, conveying equipment and other equipment that can collect operating parameters and record the maintenance process.

[0023] This implementation method does not use the closure of the maintenance work order or the maintenance personnel filling in "maintenance completed" as the basis for determining that the fault node has been eliminated, nor does it directly identify the recurrence of the fault as the reappearance of the same fault code. Instead, it makes a comprehensive judgment based on the actual target of the maintenance action, the verification parameters after maintenance, the fault-promoting effect experienced by the equipment during the resumption of operation, and the activation order of each fault node after maintenance.

[0024] In this embodiment, the normal operating range, recovery judgment range, propagation delay range, verification duration, observation start threshold, effective observation threshold, and maintenance handling threshold are all determined based on the target equipment's technical documents, sensor verification or calibration records, confirmed fault-free operation data, historical maintenance closed-loop events, or reliability test data. They do not adopt uniform fixed values ​​that are unrelated to the equipment structure, fault mechanism, and operating conditions.

[0025] The equipment fault diagnosis system, maintenance work order system, maintenance verification terminal, target equipment controller, and operation monitoring terminal all store the equipment identifier of the target equipment. The maintenance work order system assigns a maintenance work order identifier to each fault repair event.

[0026] The system obtains equipment identification, fault occurrence time, fault location, fault symptoms, monitoring parameter over-limit records, and fault cause records from the equipment fault diagnosis system; it also obtains maintenance work order identification, maintenance object, maintenance operation mode, maintenance start time, and maintenance end time from the maintenance work order system; it retrieves verification items, verification parameters, verification parameter acquisition location, verification completion time, and measurement error of verification parameters from the maintenance verification terminal; and it obtains operating parameters, operating status, control commands, and parameter acquisition time after maintenance is completed from the target equipment controller and operation monitoring terminal.

[0027] Each data source uses the system time of the target device controller as a unified time base. For data sources that cannot be directly synchronized with the target device controller, the data timestamps are corrected based on the clock calibration records of that data source.

[0028] Based on the equipment identifier and maintenance work order identifier, fault diagnosis data, maintenance action data, and post-maintenance verification data are associated with the same maintenance closed-loop event unit. For operational data without a recorded maintenance work order identifier, based on the equipment identifier and parameter acquisition time, it is associated with the maintenance closed-loop event unit with the most recent maintenance end time that has not yet concluded subsequent observation.

[0029] Each maintenance closed-loop event unit includes at least a fault diagnosis data segment, a maintenance action data segment, a post-maintenance verification data segment, a recovery operation data segment, and a subsequent monitoring data segment.

[0030] Data obtained after maintenance is completed but before the operational exposure of the observed object reaches the observation start threshold is determined as recovery operation data. Data obtained after the operational exposure reaches the observation start threshold is determined as subsequent monitoring data.

[0031] When a single fault maintenance event generates two or more fault observation objects, the time when each fault observation object reaches the observation start threshold is recorded separately. For fault observation objects that have not yet reached the observation start threshold, the corresponding operating data is still used as recovery operating data; for fault observation objects that have already reached the observation start threshold, the corresponding data is used as subsequent monitoring data.

[0032] When data acquisition is interrupted, no cross-period interpolation is performed on the data during the interruption period. If the target device is confirmed to be in a stopped state during the interruption period based on the operating status of the target device controller, then this period is not included in the operational exposure. If the operating status during the interruption cannot be confirmed, then this period is marked as a data missing period. When data missing results in the inability to determine the node activation time or node activation order, the corresponding event is designated as an event to be confirmed.

[0033] Based on the fault cause, fault location, and fault symptoms in the fault diagnosis data, the equipment components or functional parts involved in this fault repair event are determined. Each fault node corresponds to a piece of equipment or functional part that can be directly affected by maintenance actions or whose status can be characterized by monitoring parameters.

[0034] The equipment components or functional parts corresponding to the cause of the fault are identified as the causative nodes, the equipment components or functional parts through which the fault effect is transmitted from the causative nodes to the fault symptoms are identified as the transmission nodes, and the equipment components or functional parts that produce the fault symptoms are identified as the manifestation nodes.

[0035] Whether a directed connection is established between two faulty nodes is determined by the physical transmission channel, the direction of action transmission, and the relationship between changes in monitoring parameters.

[0036] Physical transmission channels include mechanical connection channels, fluid connection channels, electrical connection channels, heat transfer channels, or control action channels. The direction of action transmission is determined according to the energy flow direction, medium flow direction, force transmission direction, or control signal action direction recorded in the equipment design documents.

[0037] The propagation delay range is determined based on confirmed historical fault data. The start times of exceeding limits for monitoring parameters at both the upstream and downstream fault nodes are extracted, and the time difference between these two times is taken as the historical propagation delay. After excluding samples from cases of sensor clock asynchrony, sampling interruption, maintenance intervention, control mode switching, or fault path changes, the minimum and maximum values ​​of the effective historical propagation delays are used to form the propagation delay range.

[0038] When historical fault data is insufficient, the propagation delay range is determined through a step response test of the target device. Without damaging the target device, the controllable input corresponding to the upstream node is changed, and the moment when the downstream node's monitoring parameters first show a stable change greater than the measurement error is recorded. The propagation delay range is then determined based on the results of multiple valid tests.

[0039] When the monitoring parameters of the upstream fault node exceed the normal operating range, and the monitoring parameters of the downstream fault node exceed the normal operating range within the propagation delay range, and the direction of change of the downstream monitoring parameters conforms to the corresponding direction of energy, medium, mechanical action or control action transmission, it is determined that there is a fault propagation relationship between the two fault nodes.

[0040] A directed connection from the upstream fault node to the downstream fault node is established only when there is a physical transmission channel, a direction of action transmission, and a fault propagation relationship between the two fault nodes.

[0041] For example, in a centrifugal pump unit using rolling bearings, when the disassembly and inspection results confirm that lubricating medium contamination causes rolling bearing wear and further increases bearing housing vibration, the lubrication circuit is identified as the causative node, the rolling bearing as the transmission node, and the bearing housing as the manifestation node, thus establishing a fault causative chain of "lubrication circuit → rolling bearing → bearing housing".

[0042] A maintenance intervention-verification response matrix is ​​established by using the fault nodes in the fault cause chain as matrix rows and the current maintenance action and post-maintenance verification items as matrix columns.

[0043] When a maintenance action targets a faulty node's equipment component or functional part, and the maintenance action can directly change the node's structural state, coordination state, power supply state, or medium state, a direct intervention relationship is established between the maintenance action and the faulty node.

[0044] For example, replacing a rolling bearing can change the structural state of the rolling bearing, thus establishing a direct intervention relationship with the rolling bearing nodes; cleaning the lubrication pipeline and changing the lubrication medium can change the medium state of the lubrication circuit, thus establishing a direct intervention relationship with the lubrication circuit nodes; cleaning only the outer surface of the bearing housing cannot change the internal wear state of the rolling bearing, thus not establishing a direct intervention relationship with the rolling bearing nodes.

[0045] The direct verification relationship between the verification project and the faulty node is determined by the ability to distinguish the status of the verification parameters and the collection location.

[0046] Fault-free status samples are derived from the operational phase where the target equipment or equipment of the same model has completed acceptance testing, operates within the specified range, has no corresponding fault alarms, and has not experienced a corresponding fault even after subsequent effective observation conditions are met. Fault status samples are derived from the operational phase where the corresponding fault node has been confirmed to have a fault through disassembly and inspection, offline testing, protection action records, or maintenance results.

[0047] Determine the parameter ranges for the verification parameters in both the fault-free and fault-prone state samples. When the interval between the two parameter ranges is greater than the measurement error of the verification parameter, and the acquisition location of the verification parameter is located at the fault node or the direct output location of the fault node, establish a direct verification relationship between the corresponding verification item and the fault node.

[0048] When a verification parameter is affected by two or more fault nodes at the same time, and its state source cannot be distinguished based on frequency band, acquisition location, phase, medium composition or other characteristics, the verification parameter shall not be used as a direct verification parameter of the upstream fault node, but may be used as a direct verification parameter of the manifest node or an indirect verification parameter of the upstream fault node.

[0049] The maintenance intervention-verification response matrix records direct intervention relationships, direct verification relationships, and indirect verification relationships, respectively.

[0050] For each verification parameter, the normal operating range given in the target equipment's technical documentation or by the manufacturer should be used first. If the technical documentation does not provide a normal operating range, or if the given range cannot distinguish specific operating conditions, a basic normal range should be established based on fault-free operating data.

[0051] Fault-free operating data are grouped according to operating load, speed, environmental conditions and control mode, and only data identical to the current post-maintenance verification conditions are used.

[0052] Before establishing a basic normal range, exclude sensor offline data, calibration failure data, transient data during equipment start-up and shutdown, control mode switching data, and data during periods with other fault alarms.

[0053] The confirmed fault-free operating data are sorted numerically, with the minimum and maximum values ​​retained as the lower and upper boundaries of the basic normal range, respectively. If the equipment technical documents specify the normal operating range, the intersection of the specified range and the fault-free operating data range is used as the basic normal range. If there is no intersection, the sensor calibration status, equipment operating condition, and the applicability of the technical documents are re-verified; a recovery judgment range is not directly established.

[0054] The measurement error of the verification parameters is extended downwards at the lower boundary of the baseline normal range, and upwards at the upper boundary of the baseline normal range to obtain the recovery judgment range.

[0055] When the fault condition increases the verification parameter, the upper boundary of the recovery decision interval is determined as the fault direction boundary; when the fault condition decreases the verification parameter, the lower boundary of the recovery decision interval is determined as the fault direction boundary.

[0056] When the fault state may change in two directions, two fault direction boundaries are established respectively, and the corresponding boundary is selected according to the actual change direction of the current verification parameters relative to the initial average value after maintenance.

[0057] For equipment operating in cycles, the verification duration must cover at least one complete operating cycle and the maximum response time of the verification parameters. A complete operating cycle refers to the process from startup, entering a specified operating state, completing one major operating procedure, to returning to the initial state or a stopped state.

[0058] For continuously operating equipment, the verification period begins after the equipment enters a steady state. A steady state is defined as the speed, load, flow rate, power, or key control parameters remaining within the target set range.

[0059] The duration of the verification period for continuously operating equipment is the greater of the maximum historical duration from equipment startup to a stable state and the maximum response time of the verification parameter. The maximum historical duration is derived from equipment debugging records or effective startup records; the maximum response time is derived from sensor technical documents, step response tests, or historical fault response data.

[0060] When no valid historical data is available, the main operating parameters and the time required for the verification parameters to enter a stable range are determined through equipment startup tests, and the maximum value among the valid test results is taken.

[0061] When a faulty node has a direct intervention relationship and its direct verification parameters are within the recovery judgment range throughout the entire verification period, the faulty node is determined to be a node that has been eliminated.

[0062] Post-repair verification parameters should also establish an identifiable recovery margin. When the distance between the initial average value after repair and the fault direction boundary is not greater than the measurement error of the verification parameters, it cannot be confirmed that the verification parameters have escaped the fault boundary, and the corresponding fault node is identified as a residual node.

[0063] A faulty node is identified as a residual node if any of the following conditions exist: There is no direct intervention relationship between the faulty node and the faulty node; There is no direct verification relationship between the faulty nodes; The direct verification parameter exceeds the recovery judgment range during the verification period; Post-repair verification data is missing or the collection time is insufficient; The distance between the initial average value after maintenance and the boundary of the fault direction is no greater than the measurement error.

[0064] For any two residual nodes, if one residual node can reach the other residual node along the directed connection direction of the fault cause chain, the two residual nodes are assigned to the same residual cause object, and the directed connection relationship between them is preserved.

[0065] Residual nodes that are not reachable from other residual nodes in any directed direction are identified as single-node residual cause objects.

[0066] When the fault cause chain has branches, residual cause objects are formed with the manifestation node as the grouping endpoint. When an upstream residual node points to two manifestation nodes, the upstream residual node can be recorded as a shared residual node in two residual cause objects, and the operational exposure and recurrence risk are calculated for the two residual cause objects respectively.

[0067] For each residual cause object, along with the faulty node and directed connection relationship between the residual cause object and the manifest node, a fault observation object is formed.

[0068] For eliminated nodes, establish an observation object for eliminated node faults based on the directed connection path from the eliminated node to the manifest node.

[0069] When there are no residual nodes, no residual cause object is created. Instead, a fault observation object is created based on the directed connection paths from each eliminated node to the manifest node, and the recurrence risk of eliminated nodes is calculated.

[0070] When there are two or more residual causative objects, an exposure parameter set is established for each residual causative object, the operational exposure amount is calculated, the observation initiation threshold and the effective observation threshold are determined, and the risk of residual node propagation and recurrence is calculated separately.

[0071] Based on the fault mechanism of each fault node in the fault observation object, determine the operating parameters that can promote the accumulation of the corresponding fault state.

[0072] For wear accumulation faults, select load, speed, cumulative revolutions, and lubrication status parameters; for heat accumulation faults, select temperature deviation amplitude, over-temperature duration, and thermal cycling parameters; for cyclic impact faults, select start-stop times, impact amplitude, and load change amplitude; for media erosion faults, select media concentration, water content, particulate contamination, media flow rate, and contact time; for load fatigue faults, select stress amplitude, number of cycles, and average load.

[0073] Whether an operating parameter is included in the exposed parameter set is determined based on whether there is an identifiable response relationship between its changes and the monitoring parameters of the fault node. The change in the operating parameter should be greater than the measurement error of the operating parameter, the change in the monitoring parameter of the fault node should be greater than the measurement error of the corresponding monitoring parameter, and the direction of change of both should conform to the specific fault mechanism.

[0074] Operating parameters that only have statistical correlations but lack the basis for equipment structure, energy transfer, media transfer, or control effects are not included in the exposure parameter set.

[0075] For continuously varying operating parameters, the continuous cumulative exposure is determined based on the magnitude and duration of the deviation of the operating parameter from the baseline operating range along the fault-promoting direction. For cyclically varying operating parameters, the cyclic cumulative exposure is determined based on the portion of the actual change magnitude in each operating cycle that exceeds the fault-free baseline change magnitude, as well as the number of cycles.

[0076] When the operating parameters are within the baseline operating range, or when their change direction is opposite to the direction of fault promotion, the corresponding time period or cycle will not be included in the cumulative exposure.

[0077] To eliminate the influence of different operating parameter units and accumulation methods, the cumulative exposure of each operating parameter is divided by the corresponding reference cumulative exposure to obtain the normalized cumulative exposure. The reference cumulative exposure is preferentially adopted from the cumulative exposure when the corresponding fault node first produces a stable anomaly in a similar historical maintenance closed-loop event; if no valid historical event exists, the cumulative exposure when the corresponding fault node first produces a stable anomaly in a reliability test is adopted.

[0078] A stable anomaly refers to a situation where the corresponding monitoring parameter continuously exceeds the normal operating range and the duration reaches the minimum effective response time.

[0079] The intensity of the effect of each operating parameter on different fault nodes is determined based on the normalized sensitivity. The normalized sensitivity is determined based on the relative change of the operating parameter relative to its baseline operating range, and the relative change of the fault node monitoring parameter relative to its normal operating range.

[0080] The propagation position coefficient is determined based on the minimum number of directed edges between the faulty node and the performance node. The closer the faulty node is to the performance node, the larger the propagation position coefficient is.

[0081] Operational exposure of the fault observation object Determine according to the following formula: ; In the formula: This indicates the operational exposure level of the currently monitored fault object; Indicates the number of faulty nodes in the fault observation object; Indicates the first The number of operating parameters corresponding to each fault node; Indicates the first The propagation location coefficient of each fault node; Indicates the first The running parameters for the first Normalized sensitivity of each fault node; Indicates the first The running parameters are for the first Normalized cumulative exposure of each faulty node.

[0082] When the object of fault observation changes, the fault node, exposure parameter set, normalized sensitivity and propagation location coefficient are redefined, and the operational exposure of the original object of fault observation is no longer used.

[0083] For the current fault observation object, select historical maintenance closed-loop events from historical maintenance closed-loop events that have the same fault node composition, the same directed connection sequence, and the same maintenance intervention-verification response relationship.

[0084] For single-node residual cause objects, only the faulty node and the maintenance intervention-verification response relationship are compared. For branch-type fault observation objects, historical events are selected according to the specific branch node composition.

[0085] Only historical maintenance closed-loop events with complete equipment identification, operating conditions, maintenance actions, verification results, node activation sequences, and recurrence types can be used to form the historical recurrence exposure set. Recurrence events caused by sensor failure, abnormal equipment downtime, external accidents, sudden changes in operating conditions, or missing data are not included in the historical recurrence exposure set.

[0086] The runtime exposure values ​​at the time of historical events, including recurrence events of eliminated nodes or recurrence events propagated by residual nodes, are obtained to form a set of historical recurrence exposure values. The minimum value in this set is determined as the observation initiation threshold, and the maximum value is determined as the effective observation threshold.

[0087] When the set of historical recurrent exposures includes only one valid value, the observation initiation threshold and the valid observation threshold use the same value.

[0088] If no historical events with the same node composition and connection order exist, the scope is expanded sequentially to historical events of the same type of equipment with the same first causative node and the same fault mechanism.

[0089] If no valid recurrence data is available for the same model of equipment, the observation initiation threshold and effective observation threshold shall be determined based on the reliability tests of the target equipment or the same model of equipment. The operational exposure level when the corresponding monitoring parameter first forms a stable anomaly shall be used as the observation initiation threshold, and the operational exposure level when the failure limit specified in the equipment technical document is triggered, the protection action occurs, or the fault is confirmed by disassembly and inspection shall be used as the effective observation threshold.

[0090] Historical events in which recurrence events of eliminated nodes or recurrence events of residual node propagation were identified as recurrence samples; historical events in which no recurrence occurred and the final operational exposure reached the effective observation threshold were identified as effective non-recurrence samples; and historical events in which no recurrence occurred but the final operational exposure did not reach the effective observation threshold were identified as samples to be observed.

[0091] The samples to be observed are not included in the recurrence risk calculation.

[0092] Minimum effective response time is used to eliminate single-point noise, sensor anti-interference processes, and equipment control transients.

[0093] The anti-interference time of the sensor corresponding to the monitoring parameter, one control cycle of the target device, and the shortest duration of continuous anomaly in the valid historical fault samples are obtained respectively. The maximum value among the three is determined as the minimum effective response time. If no valid historical fault samples exist, the larger value between the sensor anti-interference time and the target device control cycle is used.

[0094] Once the operational exposure reaches the observation initiation threshold, identify monitoring parameters in subsequent monitoring data that continuously exceed the normal operating range, have an amplitude greater than the measurement error, and a duration that reaches the minimum effective response time.

[0095] The first acquisition time of the continuous out-of-limit acquisition segment is determined as the parameter out-of-limit start time, and the corresponding fault node is determined as the active node. The active nodes are arranged from earliest to latest according to the parameter out-of-limit start time to form the subsequent node activation sequence.

[0096] If the difference in the start time of parameter exceeding limits between two faulty nodes is not greater than the maximum time synchronization error of the data source, they are identified as simultaneously activated nodes. If simultaneous activation makes it impossible to determine the direction of fault propagation, the corresponding event is identified as an event to be confirmed, and a high-frequency sampling or supplementary verification task is generated.

[0097] Among the causal nodes and propagating nodes included in the subsequent node activation sequence, the fault node with the earliest parameter over-limit start time is determined as the first active node.

[0098] When the first active node belongs to an eliminated node, and all activated nodes from the first active node to the manifest node belong to the original fault cause chain, and the activation order is consistent with the directed connection direction of the fault cause chain, it is determined to be an eliminated node recurrence event.

[0099] A residual node propagation recurrence event is defined as the first acting node belonging to a residual cause object, and the activation order from the first acting node to the manifest node is consistent with the direction of the residual cause object and its downstream directed connections.

[0100] When the first active node does not belong to either the eliminated node or any residual causative object, there is no directed reachability relationship between the first active node and any residual node in any direction, and the first active node is not located in the original fault propagation path from the eliminated node to the manifest node, it is determined as an independent new fault event.

[0101] When only node activation is observed, key node data is missing, simultaneously activated nodes cannot be sorted, the node activation order is inconsistent with the fault cause chain, or the first active node cannot be determined, the corresponding event will be identified as an event pending confirmation.

[0102] For events awaiting confirmation, extract direct verification items from the maintenance intervention-verification response matrix based on missing nodes or nodes whose status cannot be confirmed; if no direct verification items exist, generate disassembly inspection, sampling, offline testing, or indirect verification tasks.

[0103] When the faulty node to be re-inspected has a direct verification relationship, the corresponding direct verification item is extracted from the maintenance intervention-verification response matrix.

[0104] When the faulty node to be re-inspected does not have a direct verification relationship, a monitoring parameter that can respond to its state change is selected from the direct downstream node of the node as an indirect verification parameter. At the same time, disassembly inspection items, sampling items or offline detection items are generated for the faulty node to be re-inspected.

[0105] Indirect verification parameters are only used to determine whether a direct inspection of the faulty node to be re-inspected is necessary, and cannot be used alone as the basis for updating residual nodes to eliminated nodes.

[0106] A faulty node awaiting re-inspection is only updated to a eliminated node after a maintenance action that directly changes its fault status is performed, and its recovery is confirmed through direct verification, disassembly, sampling, or offline testing.

[0107] Calculate the recurrence risk for each fault observation object separately.

[0108] Let the operational exposure of the current fault observation object be... Select samples from relapsed and valid non-relapsed samples to determine the final observed exposure level, which should be no less than [amount missing]. Historical samples.

[0109] Let the total number of selected historical samples be... Among them, the exposure level during operation is no greater than The number of samples that have already eliminated node recurrence events is The number of samples that have experienced residual node propagation and recurrence events is This eliminates the risk of node-based recurrence. and the risk of recurrence based on the spread of residual nodes They are respectively: ; ; In the formula: This indicates the risk of node recurrence that has been eliminated under the current operational exposure level; This indicates the risk of residual node propagation and recurrence under the current operational exposure level; This indicates the total number of historical samples whose final observed exposure is not less than the current operational exposure. This indicates the number of samples that have experienced node recurrence events prior to the current operational exposure level; This indicates the number of samples where residual node propagation recurrence events occurred prior to the current operational exposure level. When If no valid samples of the same model of equipment are found, then reliability test data is used.

[0110] If neither historical samples nor reliability test data are available, the recurrence risk is not assigned a value of zero. Instead, the observed fault is marked as having insufficient samples, and a supplementary testing task is generated.

[0111] Obtain the initial average value and current moving average value of the verification parameters after the maintenance verification is completed.

[0112] The initial average value is determined by continuous sampling values ​​taken when the target device is in a stable state during the verification period. The current moving average value of the cyclically running device covers the most recent complete operating cycle; the current moving average value of the continuously running device covers the most recent continuous operating period that is not shorter than the maximum response time of the verification parameter and has no shutdown or control mode switching.

[0113] Based on the direction of change of the current sliding average value relative to the initial average value after maintenance, the corresponding fault direction boundary is selected. When the current sliding average value moves towards the fault direction boundary, the degree of regression of the verification parameters relative to the initial average value is calculated; when the current sliding average value moves away from the fault direction boundary, the degree of regression is zero.

[0114] Establish a correspondence between the degree of rollback of verification parameters and the increment of operational exposure based on historical maintenance closed-loop events. When the current rollback degree is between two adjacent historical rollback degrees, interpolation is performed between the corresponding two historical operational exposure increments to obtain the additional operational exposure.

[0115] When the current rollback level is lower than the minimum rollback level in the historical correspondence, the additional operational exposure is set to zero; when the current rollback level is higher than the maximum historical rollback level, the operational exposure increment corresponding to the maximum historical rollback level is used, and this verification parameter is marked as exceeding the historical range, and no extrapolation without data basis is performed.

[0116] The additional operational exposure is added to the current operational exposure to obtain the revised operational exposure, and the recurrence risk is then reassessed.

[0117] When there are two or more residual causative objects in the target device, calculate the operational exposure of each residual causative object and the risk of residual node propagation and recurrence.

[0118] Residual causative objects that have reached the maintenance threshold are identified as residual causative objects to be addressed. If only one residual causative object reaches the maintenance threshold, it is identified as the target residual causative object.

[0119] When two or more residual causative objects reach the maintenance and disposal threshold, they are arranged from high to low according to the risk of recurrence of residual node propagation. The nodes to be re-inspected and the verification items corresponding to each residual causative object can be recorded separately in the same maintenance management work order, or maintenance management work orders can be generated separately.

[0120] When the risk of recurrence at eliminated nodes is equal to the risk of recurrence from residual nodes, and both reach the maintenance threshold, the faulty nodes to be re-inspected for each type of risk are identified separately and jointly recorded in the maintenance management work order. If two faulty nodes to be re-inspected are the same, the duplicate verification items are merged.

[0121] The maintenance and handling threshold is determined based on historical risk data recalculated using the risk calculation method of this implementation method.

[0122] For the same fault observation object, obtain the risk value of the corresponding recurrence event in the subsequent complete operating cycle or continuous production cycle in the historical maintenance closed-loop event, and determine the minimum risk value as the maintenance handling threshold.

[0123] If no valid historical events exist for the target equipment, data from the same model of equipment shall be used; if no valid data exists for the same model of equipment, the risk value of the valid monitoring point before the first triggering of equipment protection action, performance exceeding tolerance, or failure confirmation during reliability testing shall be used.

[0124] When historical data and reliability test data are insufficient to determine the maintenance threshold, the threshold is not arbitrarily set. Instead, the corresponding fault observation object is kept in a state of insufficient sample, and a supplementary testing task is generated.

[0125] Once the risk of recurrence reaches the maintenance and handling threshold, the fault nodes to be re-inspected are determined according to the risk type, and the corresponding verification items are extracted from the maintenance intervention-verification response matrix.

[0126] From the set of exposure parameters corresponding to the target risk type, select controllable operating parameters that can be directly adjusted by the target equipment controller or indirectly changed by adjusting other control parameters. Among the controllable operating parameters, select the operating parameter that contributes the most to the current operational exposure as the limiting parameter.

[0127] When the limiting parameter promotes fault accumulation in the direction of increasing value, the upper boundary of its reference working range is determined as the operating limit; when the limiting parameter promotes fault accumulation in the direction of decreasing value, the lower boundary of its reference working range is determined as the operating limit.

[0128] The target equipment controller adjusts the speed, load, flow rate, pressure, temperature, or start-up / shutdown schedule according to the operating limit instructions to ensure that the limiting parameters do not exceed the operating limits.

[0129] If no controllable operating parameters are found in the exposed parameter set, a load reduction command, a stop adding new tasks command, or a shutdown command will be generated according to the safety handling method specified in the target equipment's technical documents.

[0130] The processing time for maintenance management work orders is determined based on the remaining difference between the effective observation threshold and the current corrected operational exposure, as well as the recent growth rate of operational exposure per unit time.

[0131] The recent growth rate of cyclically operating equipment is determined based on the most recent complete operating cycle; the recent growth rate of continuously operating equipment is determined based on the most recent continuous operating period that is not shorter than the maximum response time of the verification parameter and has no downtime or operating condition switching.

[0132] If the current adjusted operational exposure level has reached the effective observation threshold, the response timeframe will be set until the end of the current operational cycle. If the recent operational exposure rate per unit time is zero, no finite response timeframe will be generated based on the operational exposure level, but subsequent monitoring will continue.

[0133] After a maintenance management work order is executed, maintenance action data and re-inspection data for the faulty nodes to be re-inspected are obtained.

[0134] When the faulty node to be re-inspected has a direct verification relationship, its recovery status is determined based on the recovery judgment interval and the verification duration.

[0135] When a faulty node to be re-inspected does not have a direct verification relationship, the results of disassembly, sampling, offline testing, or the addition of direct verification items must be obtained. It cannot be updated as a eliminated node simply based on the recovery of downstream indirect verification parameters.

[0136] The node is only updated to a cleared node when the maintenance action directly changes the fault status of the node to be re-inspected, and the fault status is confirmed to have been eliminated by direct verification, disassembly, sampling, or offline testing results.

[0137] After the node status is updated, the residual cause objects are reclassified, the fault observation objects, exposure parameter sets and historical sample sets are re-established, and the observation initiation threshold and effective observation threshold are re-determined.

[0138] The operational exposure data of the original fault observation object will no longer be used. Starting from the time the current maintenance verification is completed, the operational exposure data will be re-accumulated for the updated fault observation object.

[0139] When the recurrence risk does not reach the maintenance threshold, the operating exposure and recurrence risk are updated once after each complete operating cycle of the cyclically operating equipment; the operating exposure and recurrence risk are updated once after each cumulative maximum response time of the verification parameter for continuously operating equipment.

[0140] When an event awaiting confirmation occurs, it is not subject to maintenance and handling threshold restrictions and directly generates supplementary monitoring, disassembly and inspection, sampling, or offline testing tasks.

[0141] Taking a centrifugal pump unit in a circulating water system as an example, the operation monitoring terminal collects data on lubricating medium particulate contamination, bearing temperature, bearing fault frequency band energy, bearing housing vibration, pump speed, pump load, start-up and shutdown status, and running time.

[0142] During a troubleshooting incident, the fault diagnosis system recorded bearing housing vibration exceeding the normal operating range. Based on increased lubricant particle contamination, increased bearing fault frequency band energy, and disassembly and inspection results, the fault causation chain was determined to be the lubrication circuit, the rolling bearing, and the bearing housing.

[0143] The maintenance work order recorded the replacement of the rolling bearing, but did not record the cleaning of the lubrication circuit or the replacement of the lubrication medium. Post-maintenance verification items included bearing clearance testing and bearing housing vibration testing, but did not include testing for particulate contamination in the lubrication medium.

[0144] According to the maintenance intervention-verification response matrix, the rolling bearing node has a direct intervention relationship and a direct verification relationship, and the verification parameters are in the recovery judgment range during the verification duration. Therefore, the rolling bearing node is identified as the eliminated node.

[0145] The lubrication circuit nodes do not have direct intervention or verification relationships, therefore they are identified as residual nodes, forming a single-node residual cause object.

[0146] For the residual causes in the lubrication circuit, a set of exposure parameters is formed by selecting the particle contamination level of the lubricating medium, bearing load, pump speed, and operating duration from the recovery operation data.

[0147] Once the operational exposure reaches the observation start threshold, if the lubricating medium particulate contamination level first exceeds the limit continuously, followed by the bearing fault frequency band energy exceeding the limit, and then the bearing housing vibration exceeds the limit, then the first active node belongs to the residual cause object of the lubrication circuit, and the activation sequence conforms to the directional connection direction of the lubrication circuit, rolling bearing, and bearing housing, and is determined to be a residual node propagation recurrence event.

[0148] If the particle contamination level of the lubricating medium remains within the normal operating range, and the energy of the fault frequency band corresponding to the rolling bearing first exceeds the limit continuously, followed by the bearing housing vibration exceeding the limit, then the first point of action belongs to the eliminated point and is determined to be a recurrence event of the eliminated point.

[0149] If the coupling alignment parameter first exceeds the limit, followed by an increase in bearing housing vibration, and there is no directional reachability between the coupling node and the residual node in the lubrication circuit in any direction, and the coupling node is not located in the original fault propagation path from the rolling bearing to the bearing housing, then it is determined to be an independent new fault event.

[0150] If only the bearing housing vibration exceeds the limit, but there is no effective monitoring data for the lubrication circuit and rolling bearing nodes, it is identified as an event to be confirmed, and tasks such as lubrication medium sampling, bearing fault frequency band detection, and coupling alignment inspection are generated.

[0151] When the risk of recurrence due to residual node propagation reaches the maintenance threshold, the lubrication circuit is identified as a fault node to be re-inspected. Since the lubrication circuit lacks directly online verification items, lubrication medium sampling, filter inspection, and lubrication pipeline inspection items are generated.

[0152] When the bearing load is the most contributing controllable operating parameter in the current set of exposed parameters, the boundary of the reference working range of the bearing load along the fault promotion direction is determined as the operating limit. The pump unit controller reduces the pump speed or flow rate set value to ensure that the bearing load does not exceed the operating limit.

[0153] After the maintenance personnel complete the cleaning of the lubrication circuit and the replacement of the lubrication medium, they confirm the recovery of particulate contamination by taking samples of the lubrication medium, and confirm that the relevant parameters are within the recovery judgment range during the verification period. Then, they update the lubrication circuit nodes to the nodes that have been eliminated.

[0154] After the node status is updated, the original single-node residual cause object is cancelled, the fault observation object and exposure parameter set are re-established, and the operational exposure quantity is re-accumulated from the time when this maintenance verification is completed.

[0155] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for assessing the risk of fault recurrence based on maintenance closed-loop data, characterized in that, Includes the following steps: S1: Obtain fault diagnosis data, maintenance action data, post-maintenance verification data, recovery operation data and subsequent monitoring data corresponding to a fault maintenance event of the target equipment. Based on the fault diagnosis data and the component connection relationship, energy transfer relationship and fault propagation relationship of the target equipment, construct a fault cause chain including the fault node and the directed connection relationship between the fault nodes. S2: Map the maintenance objects and maintenance methods in the maintenance action data, as well as the verification items and verification parameters in the post-maintenance verification data, to the fault nodes respectively, and establish a maintenance intervention-verification response matrix; based on whether the maintenance action directly changes the fault state of the fault node and whether the post-maintenance verification parameters meet the recovery conditions determined by the fault-free operation data, divide the fault nodes into eliminated nodes and residual nodes, and form at least one fault observation object based on the eliminated nodes, residual nodes, and directed connection relationships; S3: Based on the fault mechanism of the fault node in the fault observation object, select the operating parameters that can promote the accumulation of the corresponding fault state from the recovery operation data, and determine the operation exposure of the fault observation object. S4: After the operational exposure reaches the observation start threshold determined by historical maintenance closed-loop events or reliability test data, a subsequent node activation sequence is formed based on the start time of the exceeding limit of the monitoring parameters corresponding to each fault node in the subsequent monitoring data. Based on the node status of the first active node in the subsequent node activation sequence and the relationship between the node activation order and the directed connection relationship, the eliminated node recurrence event, the residual node propagation recurrence event, or the independent new fault event is determined. S5: Based on the operational exposure, the rollback status of post-maintenance verification parameters, and recurrence events in historical maintenance closed-loop events, determine the recurrence risk of eliminated nodes and the recurrence risk of residual node propagation, respectively. Based on the recurrence risk that meets the maintenance disposal conditions determined by historical recurrence events, determine the faulty nodes to be re-inspected and the corresponding verification items, and generate maintenance management work orders.

2. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 1, characterized in that, Obtain data corresponding to a single fault repair event of the target equipment, including: The system obtains equipment identification, fault occurrence time, fault location, fault symptoms, monitoring parameter over-limit records, and fault cause records from the equipment fault diagnosis system; it also obtains maintenance work order identification, maintenance object, maintenance operation mode, maintenance start time, and maintenance end time from the maintenance work order system; it further obtains verification items, verification parameters, verification parameter acquisition location, verification completion time, and measurement error from the maintenance verification terminal; and it acquires operating parameters, operating status, and parameter acquisition time after maintenance is completed from the controller and operation monitoring terminal of the target equipment. Based on the equipment identifier, maintenance work order identifier, and parameter acquisition time, data belonging to the same target equipment and the same fault maintenance event are associated as maintenance closed-loop event units; Data acquired after maintenance is completed and before the operational exposure of the corresponding fault observation object reaches the observation start threshold is determined as recovery operation data, and data acquired after the operational exposure reaches the observation start threshold is determined as subsequent monitoring data.

3. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 2, characterized in that, Constructing a fault cause chain includes: The equipment components or functional parts corresponding to the cause of the fault are identified as the causative nodes, the equipment components or functional parts that the fault passes through during the transmission process are identified as the transmission nodes, and the equipment components or functional parts that produce fault symptoms are identified as the manifestation nodes. Determine the physical transmission channel between the two fault nodes based on the component connection relationship of the target equipment, and determine the direction of action transmission in the physical transmission channel based on the energy transmission relationship. When the monitoring parameters of the upstream fault node exceed the corresponding normal operating range, and the monitoring parameters of the downstream fault node exceed the corresponding normal operating range within the propagation delay range determined by historical fault data or step response test, and the direction of parameter change of the downstream fault node conforms to the direction of action transmission, it is determined that there is a fault propagation relationship between the two fault nodes. When there is a physical transmission channel, a direction of action transmission, and a fault propagation relationship between two faulty nodes, a directed connection relationship is established from the upstream faulty node to the downstream faulty node.

4. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 3, characterized in that, Establish a maintenance intervention-validation response matrix, including: The fault nodes are used as the rows of the matrix, and the maintenance actions and verification items are used as the columns of the matrix; When the maintenance object is the equipment component or functional part corresponding to the fault node, and the maintenance action changes the structural state, coordination state, power supply state or medium state of the fault node, a direct intervention relationship is established between the maintenance action and the fault node. When the interval between the parameter ranges of the verification parameters in the fault state sample and the fault-free state sample is greater than the measurement error of the verification parameters, and the acquisition location of the verification parameters is located at the fault node or the direct output location of the fault node, a direct verification relationship is established between the verification item and the fault node. The direct intervention relationship and direct verification relationship are written into the maintenance intervention-verification response matrix, and the measurement error is extended to both sides of the basic normal interval formed by the fault-free operation data to obtain the recovery judgment interval corresponding to the recovery condition.

5. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 4, characterized in that, The eliminated nodes and residual nodes are divided into fault observation objects, including: For fault nodes that have both direct intervention and direct verification relationships, obtain the continuous collection values ​​of the verification parameters corresponding to the fault node after the verification is completed. When the continuously acquired values ​​are all within the recovery judgment range during the verification period, and the distance between the average value of the continuously acquired values ​​and the fault direction boundary is greater than the measurement error, the corresponding fault node is determined as the eliminated node; the remaining fault nodes are determined as residual nodes. When one residual node can reach another residual node along a directed connection, the two residual nodes are assigned to the same residual cause object; residual nodes that are not reachable from other residual nodes in any directed direction are identified as single-node residual cause objects. Each residual causative object and its fault nodes and directed connections to the manifest node are formed into a fault observation object, and each eliminated node is formed into an eliminated node fault observation object based on the directed connection path from each eliminated node to the manifest node. When there are no residual nodes, only observed objects of faults in eliminated nodes are formed.

6. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 5, characterized in that, Determine the operational exposure of the fault observation object, including: Based on the failure mechanism of each fault node in the fault observation object, select the operating parameters that can cause wear accumulation, heat accumulation, cyclic impact accumulation, media erosion accumulation or load fatigue accumulation of the corresponding fault node to form an exposure parameter set. For continuously changing operating parameters, the cumulative exposure is determined based on the deviation amplitude and duration of the operating parameter from the baseline operating range along the fault-promoting direction; for cyclically changing operating parameters, the cumulative exposure is determined based on the portion of the actual change amplitude of each operating cycle that exceeds the baseline change amplitude and the number of cycles. The cumulative exposure of each operating parameter is normalized relative to the cumulative exposure when a stable fault symptom is first formed in a similar historical maintenance closed-loop event or reliability test. The normalized sensitivity is determined based on the relationship between the relative changes in operating parameters and the relative changes in monitoring parameters of the corresponding fault nodes. The propagation location coefficient is determined based on the number of directed connections between the fault node and the manifest node. The product of the normalized cumulative exposure, the normalized sensitivity, and the propagation location coefficient is accumulated to obtain the operating exposure.

7. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 6, characterized in that, The activation threshold and historical maintenance closed-loop events are determined as follows: Select historical maintenance closed-loop events from the historical maintenance closed-loop events that have the same fault node composition, the same directed connection order and the same maintenance intervention-verification response relationship as the current fault observation object. Obtain the operational exposure when the selected historical maintenance closed-loop events have recurrence events of eliminated nodes or recurrence events of residual node propagation, and form a set of historical recurrence exposures. The minimum value in the set of historical recurrent exposures is determined as the observation initiation threshold, and the maximum value is determined as the effective observation threshold. When there are no historical maintenance closed-loop events that meet the aforementioned conditions, the observation initiation threshold and effective observation threshold are determined sequentially by using historical maintenance closed-loop events of the same type of equipment with the same first causative node and the same fault mechanism, as well as the reliability test data of the target equipment or the same type of equipment. Historical maintenance closed-loop events that have resulted in recurrence of eliminated nodes or recurrence of residual node propagation are identified as recurrence samples. Historical maintenance closed-loop events that have not resulted in the aforementioned recurrence events and whose operational exposure reaches the effective observation threshold are identified as effective non-recurrence samples. Historical maintenance closed-loop events that have not resulted in the aforementioned recurrence events and whose operational exposure does not reach the effective observation threshold are identified as samples to be observed.

8. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 7, characterized in that, Forming subsequent node activation sequences and identifying recurrence events, including: The continuous acquisition segment in which the monitored parameters continuously exceed the corresponding normal working range, the amplitude of the exceedance is greater than the measurement error, and the duration reaches the minimum effective response time is determined as the effective over-limit acquisition segment. The first acquisition time of the effective over-limit acquisition segment is determined as the parameter over-limit start time, and the corresponding fault node is determined as the activation node. Activated nodes are arranged according to the start time of parameter exceeding the limit to form a subsequent node activation sequence, and the fault node with the earliest start time of parameter exceeding the limit among the causal nodes and transit nodes included in the subsequent node activation sequence is determined as the first active node. When the first active node belongs to an eliminated node, and the activation order from the first active node to the active node is consistent with the directed connection direction of the fault cause chain, it is determined to be an eliminated node recurrence event. When the first acting node belongs to the residual cause object, and the activation order from the first acting node to the manifest node is consistent with the direction of the residual cause object and its downstream directed connections, it is determined to be a residual node propagation recurrence event. When the first active node does not belong to the eliminated node or any residual cause object, and there is no directed reachability relationship between the first active node and any residual node in any direction, it is determined as an independent new fault event; When the first active node or the node activation order cannot be determined, it is identified as an event to be confirmed, and supplementary monitoring, disassembly, sampling, or offline detection tasks are generated based on the missing faulty node.

9. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 8, characterized in that, Identify the risks of recurrence at eliminated nodes and the risks of recurrence from the spread of residual nodes, including: Excluding the samples to be observed, and selecting relapsed samples and valid non-relapsed samples based on the current operational exposure level, the final observed exposure level is not less than the current operational exposure level. Calculate the proportion of the number of samples that have eliminated node recurrence events and residual node propagation recurrence events when the current operating exposure level is not greater than the total number of selected samples, to obtain the basic recurrence risk of eliminated nodes and the basic recurrence risk of residual node propagation. Based on the initial average value, current sliding average value, and fault direction boundary of the verification parameters after maintenance verification, determine the degree of regression of the verification parameters along the fault direction; Based on the correspondence between the degree of parameter rollback and the increment of operational exposure in historical maintenance closed-loop events, the current rollback degree is converted into additional operational exposure. The additional operational exposure is added to the current operational exposure to obtain the corrected operational exposure. Based on the corrected operational exposure, the risk of recurrence of eliminated nodes and the risk of propagation and recurrence of residual nodes are re-determined. When there are no valid samples available for calculating the risk of recurrence, the corresponding fault observation object is determined to be in a state of insufficient samples, and the risk of recurrence is not assigned to zero.

10. The fault recurrence risk assessment method based on maintenance closed-loop data according to claim 9, characterized in that, Generate maintenance management work orders, including: The minimum value among the historical recurrence event risk values ​​calculated using the method of claim 9 is determined as the maintenance and treatment threshold corresponding to the maintenance and treatment conditions. When at least one of the risks of recurrence of eliminated nodes and the risk of recurrence of propagation of residual nodes reaches the maintenance and disposal threshold, the first eliminated node or the first residual node corresponding to the risk that reaches the maintenance and disposal threshold will be identified as the fault node to be re-inspected. When the fault node to be re-inspected has a direct verification relationship, the corresponding verification item is extracted from the maintenance intervention-verification response matrix; when the fault node to be re-inspected does not have a direct verification relationship, disassembly inspection items, sampling items or offline testing items are generated for the fault node to be re-inspected. From the set of exposure parameters corresponding to the fault node to be re-inspected, select controllable operating parameters that can be directly adjusted or indirectly changed by the controller of the target device. Among them, the controllable operating parameters that contribute the most to the current operating exposure are determined as limiting parameters. Based on the limit parameters, generate operating limit instructions along the boundary of the reference working interval in the fault promotion direction and send them to the controller. Obtain maintenance action data and re-inspection data after the maintenance management work order is executed. After confirming that the maintenance action directly changes the fault status of the fault node to be re-inspected and the re-inspection result meets the recovery conditions, update the fault node to be re-inspected to the eliminated node, and re-form the fault observation object and re-accumulate the operation exposure.

Images