Fire pump station intelligent inspection and remote diagnosis expert method and system

Abstract

The application provides a fire pump station intelligent inspection and remote diagnosis expert method and system, relates to the technical field of intelligent inspection, and comprises the following steps: mapping real-time operation data and historical maintenance records of a fire pump station to a domain knowledge graph, constructing a fault diagnosis reasoning chain, analyzing an abnormal state propagation path, generating a maintenance suggestion, retrieving historical cases, evaluating the feasibility of a scheme, constructing an optimal maintenance execution sequence, and tracking maintenance effects. The application can realize accurate identification of abnormal states of a fire pump station, improve fault diagnosis efficiency, optimize maintenance resource allocation, reduce equipment failure rate, and ensure safe and reliable operation of the pump station.

Description

Technical Field

[0001] This invention relates to intelligent inspection technology, and more particularly to an expert method and system for intelligent inspection and remote diagnosis of fire pump stations. Background Technology

[0002] As core equipment in building fire protection systems, fire pump stations bear the crucial responsibility of ensuring water supply during fires. With the development of smart building technology, the operation and management of fire pump stations are gradually shifting towards digitalization and intelligence. Traditional fire pump station management mainly relies on manual periodic inspections and maintenance, achieving basic functional monitoring through on-site operation and simple monitoring systems. In recent years, emerging technologies such as the Internet of Things (IoT), big data analytics, and artificial intelligence have been widely applied in the field of industrial equipment monitoring and maintenance, providing technical support for intelligent inspection and remote diagnostics of fire pump stations. Current intelligent management systems for fire pump stations primarily collect operating parameters through sensor networks, use simple threshold judgment methods for fault alarms, and achieve basic equipment status visualization through remote monitoring platforms.

[0003] Existing technologies lack the ability to deeply integrate and analyze operational data and historical maintenance records of fire pump station equipment. This makes it difficult to build a complete knowledge system covering the entire equipment lifecycle, and to fully utilize historical experience to guide current maintenance decisions. Consequently, fault diagnosis accuracy is low, and maintenance recommendations lack specificity. Existing technologies are weak in analyzing the propagation paths and impact range of abnormal states, failing to accurately predict the chain reactions of equipment failures and their potential impact on the entire fire protection system. This makes it difficult to formulate globally optimal maintenance strategies, increasing the risk of system performance degradation. Furthermore, existing technologies lack a feasibility assessment mechanism for maintenance schemes based on multi-dimensional factors. They cannot dynamically adjust maintenance strategies according to actual conditions such as site conditions, resource allocation, and equipment status, leading to unreasonable allocation of maintenance resources, low maintenance efficiency, and difficulty in intelligently generating and dynamically adjusting the optimal maintenance execution sequence. Summary of the Invention

[0004] This invention provides an intelligent inspection and remote diagnostic expert method and system for fire pump stations, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides an expert method for intelligent inspection and remote diagnosis of fire pump stations, comprising:

[0006] The real-time operation data and historical maintenance records of the fire pump station are obtained, and the real-time operation data and historical maintenance records are mapped to the knowledge graph of the fire pump station domain. The equipment status nodes and abnormal operation characteristics are identified through entity alignment and relation reasoning, and a fault diagnosis reasoning chain is constructed.

[0007] Based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment status are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and handling measures are generated, and the priority of each handling measure is ranked.

[0008] Based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, the feasibility score of the maintenance plan is calculated through data mining and pattern matching techniques.

[0009] Based on the feasibility score, the maintenance plan is optimized. Combining the equipment operation level, fault development trend, and maintenance resource allocation, the optimal maintenance execution sequence is constructed and transformed into a fire pump station control instruction set. The parameter change curves and maintenance effect acceptance data during the maintenance operation are dynamically tracked. The standard compliance and handling effectiveness of the maintenance process are evaluated through the fault diagnosis reasoning chain to form equipment operation and maintenance optimization suggestions.

[0010] Mapping the real-time operational data and historical maintenance records to a knowledge graph in the fire pump station domain, and identifying equipment status nodes and operational anomaly characteristics through entity alignment and relational reasoning, a fault diagnosis reasoning chain is constructed, including:

[0011] The equipment status parameters in the real-time operation data and the equipment identifiers in the historical maintenance records are standardized. The standardized equipment identifiers are then matched with equipment entity nodes in the knowledge graph of the fire pump station domain to generate entity alignment results.

[0012] Based on the entity alignment results, state parameters that deviate from the normal operating range in the real-time operating data are extracted as abnormal features. Equipment status nodes associated with the abnormal features are retrieved in the knowledge graph of the fire pump station domain. Potential fault source nodes and affected equipment nodes are identified by tracing the upstream and downstream relationships of the equipment status nodes, and a node association graph is established.

[0013] Based on the node association graph, the causal relationship edges between the potential fault source node and the affected equipment node are analyzed. Combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the knowledge graph of the fire pump station domain to obtain a complete reasoning path from abnormal features to the root cause of the fault.

[0014] Arrange the node sequence and relationship type in the complete reasoning path according to the time evolution order, associate the fault probability and impact range attributes corresponding to each node, and construct a fault diagnosis reasoning chain that includes the fault propagation chain and diagnostic basis.

[0015] Based on the node association graph, the causal relationship edges between the potential fault source nodes and the affected equipment nodes are analyzed. Combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the fire pump station domain knowledge graph to obtain the complete reasoning path from abnormal features to the root cause of the fault, including:

[0016] Extract the causal relationship edges between potential fault source nodes and affected device nodes from the node association graph, and obtain the relationship type and propagation direction attributes associated with each causal relationship edge to form an initial causal relationship set;

[0017] Based on the relationship types in the initial causal relationship set, fault propagation rules matching the relationship types are retrieved in the knowledge graph of the fire pump station domain, and the fault types and associated maintenance measures corresponding to the potential fault source nodes in the historical maintenance records are extracted. The fault propagation rules are semantically matched with the fault types, and candidate inference starting point nodes that match the current abnormal characteristics are selected.

[0018] Starting with the candidate inference starting node, a multi-hop relationship traversal is performed in the knowledge graph of the fire pump station domain along the propagation direction of the causal relationship edge. At each jump, the validity of the jump path is verified according to the fault evolution sequence recorded in the historical maintenance record, and the fault probability and impact range attributes of each traversed node are recorded.

[0019] All traversal paths formed during the multi-hop relationship traversal are evaluated. The path confidence is calculated based on the product of the fault probabilities of the nodes in each traversal path and the path length. The traversal path with the highest confidence is selected as the complete inference path from the abnormal features to the root cause of the fault.

[0020] Based on the aforementioned knowledge graph of fire pump stations and the aforementioned fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment conditions are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and handling measures are generated, including:

[0021] By using the fault diagnosis reasoning chain to locate the root cause node of the fault, and by using the knowledge graph of the fire pump station domain to evaluate the diffusion effect of the root cause node along the relation edges, the affected downstream equipment nodes are tracked to construct the propagation path of the abnormal state of the equipment.

[0022] Based on the equipment type and operating status of each downstream equipment node in the propagation path, and combined with the fault impact coefficient and the number of associated equipment of each downstream equipment node in the knowledge graph of the fire pump station field, the impact range of abnormal equipment status is quantified.

[0023] For the equipment nodes within the affected area, historical maintenance records are compared with historical maintenance cases corresponding to the fault types of the equipment nodes. The maintenance operations and handling results in the historical maintenance cases are integrated and associated with the maintenance specifications and operational constraints of the equipment nodes in the equipment manual knowledge base.

[0024] A compatibility analysis is performed on the maintenance operations and the maintenance specifications. Maintenance measures that meet the requirements of the equipment manual and have been proven effective in the historical maintenance cases are selected. The execution priority of the maintenance measures is determined based on the failure impact coefficient of each equipment node in the scope of influence, and targeted maintenance suggestions and handling measures are formed.

[0025] Based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, the feasibility score of the maintenance plan is calculated using data mining and pattern matching techniques, including:

[0026] In the knowledge graph of the fire pump station domain, a feature vector of the current abnormal scenario is constructed, and the equipment type identifier, fault type identifier and operating parameters are extracted from the feature vector. Using the equipment type identifier, the fault type identifier and the operating parameters, the correlation degree between the historical maintenance cases in the knowledge graph of the fire pump station domain and the current abnormal scenario is calculated, and a historical maintenance case correlation degree sequence is generated.

[0027] Based on the correlation sequence of the historical maintenance cases, the historical maintenance case with the highest correlation is selected as the reference maintenance plan, and the maintenance steps and execution conditions in the reference maintenance plan are extracted; for the maintenance steps and execution conditions, the corresponding maintenance resource requirement data, on-site construction condition data, and equipment operation status data are collected;

[0028] The maintenance resource requirement data, the on-site construction condition data, and the equipment operating status data are input into a preset scoring function to generate scoring indicators that reflect the resource matching degree, construction feasibility, and equipment adaptability of the maintenance plan; the feasibility score of the maintenance plan is calculated based on the scoring indicators and preset indicator weights.

[0029] The system dynamically tracks parameter change curves and maintenance effectiveness acceptance data during maintenance operations. Through the fault diagnosis reasoning chain, it evaluates the standard compliance and effectiveness of the maintenance process, generating equipment operation and maintenance optimization suggestions, including:

[0030] During the maintenance operation, the equipment operating parameters are continuously sampled to construct a time-series data stream. Based on the time-series data stream, parameter change curves are plotted. At the same time, the equipment performance test results and fault elimination verification records are obtained after the maintenance operation is completed, forming maintenance effect acceptance data.

[0031] Based on the parameter change curve, operation timing constraints and parameter response feature patterns are extracted from the fault diagnosis inference chain. The parameter change curve is divided into multiple time periods. For the parameter change slope in each time period, pattern matching is performed on the parameter response feature patterns to identify abnormal time periods in the parameter change curve that deviate from the parameter response feature patterns. The maintenance operation identifiers corresponding to the abnormal time periods are correlated with the operation timing constraints to quantify the standard compliance deviation of the maintenance process.

[0032] Using the maintenance effect acceptance data, the equipment performance test results are compared with the preset performance recovery target value in the fault diagnosis reasoning chain to calculate the compliance rate. Based on the fault residual characteristics in the fault elimination verification record and the fault root cause node in the fault diagnosis reasoning chain, causal tracing is performed to form a quantitative indicator of the effectiveness of the treatment.

[0033] Based on the standard compliance deviation and the quantitative index of the effectiveness of the treatment, the corresponding implementation plan adjustment strategies and supplementary maintenance measures are retrieved from the knowledge graph of the fire pump station field. The implementation plan adjustment strategies and supplementary maintenance measures are then structured and organized to form equipment operation and maintenance optimization suggestions that include operation specification revisions and periodic maintenance plans.

[0034] Using the maintenance effectiveness acceptance data, the equipment performance test results are compared with the preset performance recovery target value in the fault diagnosis inference chain to calculate the compliance rate. Furthermore, based on the fault residual characteristics in the fault elimination verification record and the fault root cause node in the fault diagnosis inference chain, causal tracing is performed to form quantitative indicators of treatment effectiveness, including:

[0035] The equipment performance test results are analyzed from the maintenance effect acceptance data to obtain the measured performance parameter values ​​of multiple equipment performance dimensions, and the corresponding performance recovery target values ​​are obtained from the fault diagnosis inference chain.

[0036] The deviation between the measured performance parameter values ​​of each device performance dimension and the performance recovery target value is calculated and it is determined whether the standard is met. The percentage of the number of device performance dimensions that meet the standard is obtained to obtain the performance compliance rate. Based on the influence weight of each device performance dimension in the fault diagnosis inference chain on the overall operational reliability of the device, the influence weight of the non-compliant dimensions is accumulated to obtain the performance defect impact degree.

[0037] The fault residual features identified in the fault elimination verification record are matched with the fault feature descriptions associated with the fault root cause nodes in the fault diagnosis reasoning chain to determine the corresponding fault root cause nodes.

[0038] In the fault diagnosis reasoning chain, the fault root cause node is traced back to the fault triggering condition node. It is determined whether the associated equipment operating environment parameters are in the fault recurrence risk range. If so, the corresponding fault root cause node is marked as not eliminated. The proportion of fault root cause nodes in the not eliminated state is counted to obtain the root cause residue rate. The performance compliance rate, the performance defect impact degree and the root cause residue rate are comprehensively calculated to form a quantitative indicator of the effectiveness of the treatment.

[0039] A second aspect of the present invention provides an intelligent inspection and remote diagnostic expert system for fire pump stations, comprising:

[0040] The first unit is used to acquire real-time operation data and historical maintenance records of fire pump stations, map the real-time operation data and historical maintenance records to a knowledge graph in the field of fire pump stations, identify equipment status nodes and abnormal operation characteristics through entity alignment and relational reasoning, and construct a fault diagnosis reasoning chain.

[0041] The second unit is used to analyze the propagation path and impact range of abnormal equipment status based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, and generate targeted maintenance suggestions and handling measures by combining historical maintenance experience and equipment manuals, and to rank the priority of each handling measure.

[0042] The third unit is used to retrieve similar historical maintenance cases in the knowledge graph of the fire pump station field based on the current abnormal scenario, and to calculate the feasibility score of the maintenance plan by comprehensively considering the maintenance resource requirements, on-site construction conditions and equipment operating status through data mining and pattern matching technology.

[0043] The fourth unit is used to optimize the maintenance plan based on the feasibility score, construct the optimal maintenance execution sequence by combining the equipment operation level, fault development trend and maintenance resource configuration, and convert the optimal maintenance execution sequence into a fire pump station control instruction set; dynamically track the parameter change curve and maintenance effect acceptance data during the maintenance operation, evaluate the standard compliance and handling effectiveness of the maintenance process through the fault diagnosis reasoning chain, and form equipment operation and maintenance optimization suggestions.

[0044] A third aspect of the present invention provides an electronic device, comprising:

[0045] processor;

[0046] Memory used to store processor-executable instructions;

[0047] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0048] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0049] The beneficial effects of this application are as follows:

[0050] By mapping real-time operational data and historical maintenance records to a knowledge graph in the field of fire pump stations, automatic identification and correlation analysis of equipment status and abnormal characteristics are achieved, improving the accuracy and intelligence level of fault diagnosis and breaking through the limitations of traditional inspection methods that rely on human experience.

[0051] Based on the fault diagnosis reasoning chain analysis, the propagation path and impact range of abnormal equipment states can be accurately predicted, and targeted maintenance suggestions and handling measures can be generated. This realizes the transformation from passive maintenance to proactive prevention and significantly reduces the equipment failure rate.

[0052] By retrieving similar historical maintenance cases from a knowledge graph and combining data mining and pattern matching techniques, the feasibility score of maintenance plans is calculated, making maintenance decisions more scientific and rational and avoiding waste of maintenance resources. Based on the feasibility score, an optimal maintenance execution sequence is constructed and transformed into a control instruction set, achieving standardization and automation of the maintenance process, improving maintenance efficiency, and reducing human error. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the intelligent inspection and remote diagnostic expert method for fire pump stations according to an embodiment of the present invention.

[0054] Figure 2 This is a flowchart illustrating the evaluation and optimization of fire pump station maintenance effectiveness in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1This is a flowchart illustrating the intelligent inspection and remote diagnostic expert method for fire pump stations according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0058] The real-time operation data and historical maintenance records of the fire pump station are obtained, and the real-time operation data and historical maintenance records are mapped to the knowledge graph of the fire pump station domain. The equipment status nodes and abnormal operation characteristics are identified through entity alignment and relation reasoning, and a fault diagnosis reasoning chain is constructed.

[0059] Based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment status are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and handling measures are generated, and the priority of each handling measure is ranked.

[0060] Based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, the feasibility score of the maintenance plan is calculated through data mining and pattern matching techniques.

[0061] Based on the feasibility score, the maintenance plan is optimized. Combining the equipment operation level, fault development trend, and maintenance resource allocation, the optimal maintenance execution sequence is constructed and transformed into a fire pump station control instruction set. The parameter change curves and maintenance effect acceptance data during the maintenance operation are dynamically tracked. The standard compliance and handling effectiveness of the maintenance process are evaluated through the fault diagnosis reasoning chain to form equipment operation and maintenance optimization suggestions.

[0062] In one optional implementation, the real-time operational data and the historical maintenance records are mapped to a knowledge graph in the fire pump station domain. Equipment status nodes and operational anomaly characteristics are identified through entity alignment and relational reasoning. A fault diagnosis reasoning chain is constructed, including:

[0063] The equipment status parameters in the real-time operation data and the equipment identifiers in the historical maintenance records are standardized. The standardized equipment identifiers are then matched with equipment entity nodes in the knowledge graph of the fire pump station domain to generate entity alignment results.

[0064] Based on the entity alignment results, state parameters that deviate from the normal operating range in the real-time operating data are extracted as abnormal features. Equipment status nodes associated with the abnormal features are retrieved in the knowledge graph of the fire pump station domain. Potential fault source nodes and affected equipment nodes are identified by tracing the upstream and downstream relationships of the equipment status nodes, and a node association graph is established.

[0065] Based on the node association graph, the causal relationship edges between the potential fault source node and the affected equipment node are analyzed. Combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the knowledge graph of the fire pump station domain to obtain a complete reasoning path from abnormal features to the root cause of the fault.

[0066] Arrange the node sequence and relationship type in the complete reasoning path according to the time evolution order, associate the fault probability and impact range attributes corresponding to each node, and construct a fault diagnosis reasoning chain that includes the fault propagation chain and diagnostic basis.

[0067] In the implementation of the fire pump station fault diagnosis system, real-time operating data and historical maintenance records of the pump station are collected. Real-time operating data includes parameters such as pump speed, flow rate, pressure, current, and bearing temperature; historical maintenance records include information such as equipment identification, fault type, fault occurrence time, maintenance measures, and maintenance results. Simultaneously, a knowledge graph in the fire pump station domain is constructed as the knowledge foundation for fault diagnosis. This graph includes nodes such as equipment entities, status attributes, fault types, and maintenance measures, as well as edges representing physical connections between equipment and the correlation between status parameters and fault types.

[0068] When mapping real-time operational data and historical maintenance records to a knowledge graph for fire pump stations, the equipment status parameters in the real-time operational data and the equipment identifiers in the historical maintenance records are standardized. For equipment identifiers, different expressions are unified into standardized names in the knowledge graph; for example, "Main Pump 1," "Main Pump 1," and "Main-Pump-01" are unified into "MainPump01." For status parameters, measured values ​​in different units are converted to standard units, such as temperature being standardized to degrees Celsius and pressure to megapascals. After standardization, the equipment identifiers are matched against equipment entity nodes in the fire pump station knowledge graph. The matching process uses a combination of string similarity calculation and equipment hierarchy verification to ensure accuracy. For example, when "MainPump01-Bearing" is identified, not only is the main equipment "MainPump01" matched, but its sub-component "Bearing" is also matched, forming a complete entity identifier chain.

[0069] Based on entity alignment results, state parameters deviating from the normal operating range are extracted from real-time operational data as anomalous features. Anomaly identification can be achieved through the following methods: First, comparing real-time parameters with the normal range defined in the knowledge graph; for example, a water pump bearing temperature exceeding 75°C is marked as anomalous. Second, detecting parameter abrupt changes or abnormal trends through time-series analysis; for example, a pressure drop exceeding 30% in a short period. Third, judging complex anomalies based on multi-parameter combinations; for example, identifying pump blockage when flow is low but current is high. When retrieving equipment state nodes associated with anomalous features in the fire pump station domain knowledge graph, a graph traversal algorithm is used to find state nodes directly associated with the anomalous parameters. For example, when an abnormal bearing temperature is detected in "MainPump01", the state node "MainPump01-BearingTemperature-Abnormal" is located in the graph.

[0070] When identifying potential fault source nodes and affected equipment nodes by tracing the upstream and downstream relationships of equipment status nodes, a bidirectional graph traversal strategy is applied. Upstream, the node causing the current anomaly is traced; for example, high bearing temperature may be caused by insufficient lubrication, bearing wear, or pump shaft misalignment. Downstream, other equipment nodes affected by this anomaly are analyzed; for example, overheating bearings may lead to pump shaft deformation or seal damage, creating a chain reaction. During the construction of the node association graph, a confidence weight is assigned to each connection edge. The weight value is determined based on the predefined association strength in the knowledge graph and the statistical correlation in historical fault data.

[0071] When analyzing causal relationships between potential fault source nodes and affected equipment nodes based on node association graphs, a path importance scoring mechanism is employed. Path importance is determined by path length, edge weight, and node type, prioritizing short paths, high weights, and critical node types. Fault types and maintenance measures from historical maintenance records are incorporated into the reasoning process, and path importance scores are adjusted based on maintenance experience in similar fault scenarios. For example, when historical records show that excessively high bearing temperatures are often related to insufficient lubrication, the importance of the path "lubrication system failure → abnormal bearing temperature" is increased. When constructing a complete reasoning path through multi-hop relationship reasoning, a maximum hop limit is set, generally not exceeding 5 hops, to balance reasoning depth and efficiency.

[0072] When arranging the node sequence and relation types in the complete reasoning path according to the chronological evolution, the order of nodes is determined based on the physical characteristics of fault development and the temporal patterns in historical cases. When constructing the fault diagnosis reasoning chain, the fault probability and impact range attributes corresponding to each node are associated. The fault probability is defined based on historical fault statistics and expert knowledge; for example, the probability of "substandard lubricating oil quality → increased bearing temperature" is 0.85. The impact range attribute describes the equipment affected by the fault, the safety risk level, and the degree of functional impact; for example, "bearing failure" affects "the entire water pump operation," with a safety risk of "medium" and a functional impact of "the pump station pressure cannot be maintained."

[0073] The fault diagnosis reasoning chain ultimately forms a directed graph structure, containing the fault propagation chain and diagnostic criteria. The fault propagation chain displays the development process of the fault from its source to the manifestation of symptoms in a time series format. The diagnostic criteria include the type of evidence (such as physical laws, statistical correlations, or expert rules) and the strength of evidence for each reasoning step. For example, in the reasoning chain of "low lubricating oil pressure → insufficient bearing lubrication → increased bearing temperature → accelerated bearing wear → increased water pump vibration," the relationships between each node are accompanied by corresponding physical evidence and probability assessments, forming a complete diagnostic logic.

[0074] The above implementation methods map real-time data and historical records to a domain knowledge graph, and combine entity alignment and relational reasoning techniques to construct a complete reasoning path from abnormal features to the root cause of the fault, providing a systematic solution for fault diagnosis of fire pump stations.

[0075] In one optional implementation, based on the node association graph, the causal relationship edges between the potential fault source node and the affected equipment node are analyzed, and combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the knowledge graph of the fire pump station domain to obtain a complete reasoning path from abnormal features to the root cause of the fault, including:

[0076] Extract the causal relationship edges between potential fault source nodes and affected device nodes from the node association graph, and obtain the relationship type and propagation direction attributes associated with each causal relationship edge to form an initial causal relationship set;

[0077] Based on the relationship types in the initial causal relationship set, fault propagation rules matching the relationship types are retrieved in the knowledge graph of the fire pump station domain, and the fault types and associated maintenance measures corresponding to the potential fault source nodes in the historical maintenance records are extracted. The fault propagation rules are semantically matched with the fault types, and candidate inference starting point nodes that match the current abnormal characteristics are selected.

[0078] Starting with the candidate inference starting node, a multi-hop relationship traversal is performed in the knowledge graph of the fire pump station domain along the propagation direction of the causal relationship edge. At each jump, the validity of the jump path is verified according to the fault evolution sequence recorded in the historical maintenance record, and the fault probability and impact range attributes of each traversed node are recorded.

[0079] All traversal paths formed during the multi-hop relationship traversal are evaluated. The path confidence is calculated based on the product of the fault probabilities of the nodes in each traversal path and the path length. The traversal path with the highest confidence is selected as the complete inference path from the abnormal features to the root cause of the fault.

[0080] The causal relationship edges between potential fault source nodes and affected equipment nodes are extracted from the node association graph, and the relationship type and propagation direction attributes associated with each causal relationship edge are obtained. For example, for the association between pressure sensor anomaly and water pump failure in a fire pump station, a causal relationship edge of type "impact" is extracted, with its propagation direction from the water pump to the pressure sensor, indicating that the water pump failure caused the pressure sensor reading to be abnormal. At the same time, a causal relationship edge of type "control" from the control valve to the water pump indicates that the valve failure affects the normal operation of the water pump. In this way, an initial set of causal relationships is formed, containing all fault propagation paths.

[0081] Based on the relation types in the initial causal relation set, fault propagation rules matching the relation types are retrieved from the knowledge graph of the fire pump station domain. For example, the propagation rule "mechanical failure → abnormal vibration → sensor reading fluctuation" is retrieved, indicating that mechanical failure leads to abnormal equipment vibration, which in turn causes sensor reading fluctuation. Simultaneously, fault types and maintenance measures corresponding to potential fault source nodes are extracted from historical maintenance records, such as the "replace bearing" maintenance measure corresponding to the water pump bearing wear fault type. Semantic matching is performed between the fault propagation rules and fault types; for example, matching the "mechanical failure" rule with the "bearing wear" fault type yields a semantic similarity of 0.85, exceeding the preset threshold of 0.7. Therefore, the water pump node is identified as a candidate inference starting point node.

[0082] Starting from the candidate inference starting node, a multi-hop traversal of relationships is performed in the knowledge graph of the fire pump station domain along the propagation direction of the causal relationship edges. In the water pump example, starting from the water pump node, the first hop reaches the "vibration anomaly" node, the second hop reaches the "pressure fluctuation" node, and the third hop reaches the "flow anomaly" node. At each hop, the validity of the hop path is verified by referring to the fault evolution sequence recorded in the historical maintenance record. For example, the historical record shows that after the water pump bearing failure, the probability of vibration anomaly is 0.92, the probability of pressure fluctuation is 0.85, and the probability of flow anomaly is 0.73, which is consistent with the current traversal path, thus confirming the validity of the traversal path. At the same time, the fault probability and influence range attribute of each traversed node are recorded, such as the fault probability of the vibration anomaly node being 0.92 and the influence range attribute being "local equipment".

[0083] All traversal paths formed during the multi-hop relation traversal process are evaluated. For example, two fault propagation paths are formed: Path 1 "Water pump bearing wear → abnormal vibration → pressure fluctuation → abnormal flow", and Path 2 "Control valve failure → unstable water pressure → abnormal flow". The confidence level of Path 1 is calculated as: 0.92 × 0.85 × 0.73 × (1 / 4) = 0.143, and the confidence level of Path 2 is calculated as: 0.75 × 0.82 × (1 / 3) = 0.205. Since the confidence level of Path 2 is higher than that of Path 1, "Control valve failure → unstable water pressure → abnormal flow" is selected as the complete inference path from the abnormal feature to the root cause of the fault.

[0084] In practical applications, to improve the accuracy of reasoning, it can be verified by combining equipment operating status data. For example, when a fire pump station experiences an abnormal flow alarm, the opening data and feedback signal of the current control valve can be obtained. If the actual valve opening does not match the control command, it can be further confirmed that the control valve malfunction is the root cause of the abnormal flow. After completing the root cause reasoning, maintenance suggestions can be automatically generated based on the maintenance measures related to the control valve malfunction in the historical maintenance records. These suggestions include "inspect the valve actuator, replace the valve seal ring, calibrate the valve position feedback device," etc., and can be provided to maintenance personnel for reference.

[0085] This method effectively and quickly pinpoints the root cause of faults from the complex relationships within fire pump stations, shortening fault diagnosis time, improving equipment maintenance efficiency, and reducing safety risks caused by fire protection system malfunctions. Furthermore, as historical maintenance records accumulate, fault propagation rules and fault probability data are continuously optimized, leading to a sustained improvement in inference accuracy and ultimately enhancing the level of intelligent operation and maintenance of fire pump stations.

[0086] In one optional implementation, based on the knowledge graph of the fire pump station domain and the fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment conditions are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and remedial measures are generated, including:

[0087] By using the fault diagnosis reasoning chain to locate the root cause node of the fault, and by using the knowledge graph of the fire pump station domain to evaluate the diffusion effect of the root cause node along the relation edges, the affected downstream equipment nodes are tracked to construct the propagation path of abnormal equipment status.

[0088] Based on the equipment type and operating status of each downstream equipment node in the propagation path, and combined with the fault impact coefficient and the number of associated equipment of each downstream equipment node in the knowledge graph of the fire pump station field, the impact range of abnormal equipment status is quantified.

[0089] For the equipment nodes within the affected area, historical maintenance records are compared with historical maintenance cases corresponding to the fault types of the equipment nodes. The maintenance operations and handling results in the historical maintenance cases are integrated and associated with the maintenance specifications and operational constraints of the equipment nodes in the equipment manual knowledge base.

[0090] A compatibility analysis is performed on the maintenance operations and the maintenance specifications. Maintenance measures that meet the requirements of the equipment manual and have been proven effective in the historical maintenance cases are selected. The execution priority of the maintenance measures is determined based on the failure impact coefficient of each equipment node in the scope of influence, and targeted maintenance suggestions and handling measures are formed.

[0091] As a core component of building fire protection systems, the reliability of fire pump stations directly impacts the effectiveness of fire emergency response. This implementation provides a method for generating fault diagnosis and maintenance recommendations for fire pump stations based on domain knowledge graphs. By analyzing the propagation path and impact range of abnormal equipment conditions, and combining historical maintenance experience with equipment manuals, targeted maintenance recommendations and remedial measures are generated.

[0092] A knowledge graph for the fire pump station domain is constructed, including equipment nodes such as fire pumps, control cabinets, pipeline valves, and sensors, as well as relationship edges such as "control," "connection," and "monitoring." Each equipment node contains attributes such as equipment number, equipment type, operating status, and fault impact coefficient. The fault impact coefficient represents a quantitative indicator of the impact of equipment failure on the overall system, with a value ranging from 0 to 1, where a higher value indicates a more severe impact. Simultaneously, a fault diagnosis reasoning chain is established for each type of fault, including information such as fault symptoms, causes and their probability distribution, and propagation paths.

[0093] When an abnormal alarm occurs at a fire pump station, the root cause of the fault is located through a fault diagnosis reasoning chain. For example, if the system alarm displays "Fire pump No. 1 cannot start," by querying the fault diagnosis reasoning chain, the root causes are determined to include: power supply failure (0.3), control circuit failure (0.4), pump body mechanical failure (0.2), and valve failure (0.1), where the values ​​in parentheses represent the probability of each cause. Combined with current monitoring data, such as a zero current reading in the pump control cabinet and a power indicator light not being on, the root cause is inferred to be a power supply failure.

[0094] After identifying the root cause node, the fault propagation effect is assessed using a knowledge graph of the fire pump station domain. Starting from the root cause node, downstream equipment nodes are traced along the relationship edges. In the example above, starting from the "Power Supply for Fire Pump No. 1" node, the propagation path of the abnormal equipment state is constructed along the "Power Supply" relationship edge to the "Fire Pump No. 1" node, then along the "Pressure Supply" relationship edge to the "Main Pipeline Network" node, and along the "Monitoring" relationship edge to the "Control Cabinet" node. All affected downstream nodes are recursively traversed to construct the propagation path of the abnormal equipment state, such as: Power Failure → Fire Pump No. 1 Cannot Start → Insufficient Pressure in Main Pipeline Network → Fire System Pressure Alarm.

[0095] Based on the equipment type and operating status of each downstream device node in the propagation path, and combined with the fault impact coefficient and the number of associated devices for each node in the knowledge graph, the impact range of abnormal equipment status is quantified. Specifically, for each device node in the propagation path, its impact degree index is calculated: Impact degree = Device node fault impact coefficient × Number of associated downstream devices. For example, the fault impact coefficient of fire pump No. 1 is 0.8, and the associated downstream devices include the main pipeline, pressure sensors, and multiple zone valves, totaling 5 devices, so its impact degree is 0.8 × 5 = 4. By setting a threshold, device nodes with an impact degree greater than 3 are included in the maintenance priority processing scope.

[0096] For the affected equipment nodes, matching maintenance cases were retrieved from the historical maintenance record database. Taking power failure as an example, the search criteria included: equipment type = "fire pump power supply", and failure symptom = "no power supply". Multiple historical maintenance cases were retrieved, such as "Maintenance record 2022-03-15: Loose contactor of pump No. 1 caused power outage, resolved by tightening the contactor screws" and "Maintenance record 2021-11-02: Power fuse blew, restored normal operation after fuse replacement". Simultaneously, relevant maintenance specifications were retrieved from the equipment manual knowledge base, such as "Fire Pump Power Failure Handling Specifications", including inspection procedures, safety precautions, and necessary tools.

[0097] A compatibility analysis was conducted between historical maintenance procedures and the maintenance specifications in the equipment manual to assess whether each maintenance measure met the requirements of the equipment manual and whether it had proven effective in historical cases. For example, regarding the issue of loose power contactors, the "tightening contactor screws" operation in historical cases was consistent with the "regularly inspect and tighten electrical connections" specified in the equipment manual, and historically, it had proven effective; therefore, it was included as a recommended measure. Although the fuse replacement operation appeared in historical cases, the equipment manual stipulates that "after a fuse blows, the cause of a short circuit should be investigated before replacement," so short circuit inspection steps need to be added to the maintenance recommendations.

[0098] The execution priority of maintenance measures is determined based on the failure impact coefficient of each equipment node within the affected area. Maintenance measures for equipment nodes with a failure impact coefficient greater than 0.7 are marked as "highest priority," those with an impact coefficient between 0.4 and 0.7 are marked as "high priority," and those with an impact coefficient less than 0.4 are marked as "normal priority." The generated maintenance recommendations and handling measures include: problem description, root cause analysis, impact scope assessment, list of handling measures (including priority), required tools and materials, and safety precautions.

[0099] The above describes in detail the implementation process of the fault diagnosis and maintenance suggestion generation method based on the knowledge graph of the fire pump station field. By systematically analyzing the propagation path and impact range of abnormal equipment status, and combining historical maintenance experience and equipment specifications, targeted maintenance measures are effectively generated to improve the reliability and maintenance efficiency of the fire pump station system.

[0100] In one optional implementation, based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, a feasibility score for the maintenance plan is calculated using data mining and pattern matching techniques, including:

[0101] In the knowledge graph of the fire pump station domain, a feature vector of the current abnormal scenario is constructed, and the equipment type identifier, fault type identifier and operating parameters are extracted from the feature vector. Using the equipment type identifier, the fault type identifier and the operating parameters, the correlation degree between the historical maintenance cases in the knowledge graph of the fire pump station domain and the current abnormal scenario is calculated, and a historical maintenance case correlation degree sequence is generated.

[0102] Based on the correlation sequence of the historical maintenance cases, the historical maintenance case with the highest correlation is selected as the reference maintenance plan, and the maintenance steps and execution conditions in the reference maintenance plan are extracted; for the maintenance steps and execution conditions, the corresponding maintenance resource requirement data, on-site construction condition data, and equipment operation status data are collected;

[0103] The maintenance resource requirement data, the on-site construction condition data, and the equipment operating status data are input into a preset scoring function to generate scoring indicators that reflect the resource matching degree, construction feasibility, and equipment adaptability of the maintenance plan; the feasibility score of the maintenance plan is calculated based on the scoring indicators and preset indicator weights.

[0104] When an abnormal scenario occurs at a fire pump station, the knowledge graph construction module first acquires the monitoring data of the currently abnormal equipment, including operating parameters such as pump vibration frequency, motor current value, pipeline pressure reading, and flow count value. During feature vector construction, the equipment type identifier uses an 8-bit encoding method, where the first two bits represent the equipment category, the middle three bits represent the specific model, and the last three bits represent the installation location number. The fault type identifier uses a hierarchical encoding structure: the first-level code indicates the severity of the fault, the second-level code indicates the location of the fault, and the third-level code indicates the specific fault phenomenon. Operating parameters are normalized to values ​​between 0 and 1; pressure parameters are divided by the equipment's rated pressure, flow parameters are divided by the design flow rate, and temperature parameters are linearly mapped according to the operating temperature range.

[0105] Feature vector generation employs a sparse matrix representation method with a 512-dimensional dimension. A hash mapping is used to map device type, fault type, and operating parameters to different dimensional spaces. The device type identifier occupies the first 128 dimensions, the fault type identifier occupies the middle 192 dimensions, and the operating parameters occupy the last 192 dimensions. Non-zero elements in the vector represent the activation state of the corresponding feature, and their numerical values ​​reflect the importance of the feature. Feature weights are derived from training with historical cases, with the device type weight coefficient set to 0.3, the fault type weight coefficient set to 0.5, and the operating parameter weight coefficient set to 0.2.

[0106] The correlation calculation for historical maintenance cases employs a weighted cosine similarity algorithm, comparing the feature vector of the current abnormal scenario with the feature vectors of historical cases stored in the knowledge graph. During the calculation, the intersection of the non-zero dimensions of the two vectors is first extracted. Then, the sum of the products of the corresponding dimension values ​​is used as the numerator, and the product of the magnitudes of the two vectors is used as the denominator. The correlation value ranges from 0 to 1, with values ​​closer to 1 indicating higher similarity. To improve computational efficiency, a similarity threshold of 0.1 is set; historical cases below this threshold are directly filtered out to avoid invalid calculations.

[0107] The knowledge graph retrieval module maintains an inverted index structure, creating index tables based on device type and fault type to accelerate the search for similar cases. The index tables are implemented using hash tables, with keys representing a combined code of device type and fault type, and values ​​representing a list of identifiers for corresponding historical cases. During the retrieval process, the index is used to quickly locate the candidate case set, followed by precise similarity calculation. The candidate set size is limited to 1000 cases; if this limit is exceeded, cases are truncated in reverse chronological order.

[0108] When generating the historical maintenance case correlation sequence, all calculated correlation values ​​are arranged in descending order to form an ordered sequence. Each element in the sequence contains a case identifier, a correlation value, and basic case information. The correlation value is preserved to four decimal places, and cases with the same correlation are arranged in reverse timestamp order to ensure that the most recent maintenance experience is prioritized. The sequence length is limited to 100 cases to provide ample candidate space for subsequent solution selection.

[0109] The selection of reference maintenance solutions employs a multi-threshold strategy. First, the case with the highest correlation is selected as the primary reference, while other cases with a correlation of 90% or higher are selected as secondary references. The primary reference solution provides a complete sequence of maintenance steps, while the secondary reference solutions supplement handling measures for special circumstances. During the extraction of maintenance steps, each step includes five elements: operation description, required tools, expected time, preconditions, and acceptance criteria. The extraction of execution conditions includes constraints such as on-site environmental requirements, personnel skill requirements, safety protection requirements, and equipment downtime windows.

[0110] Maintenance resource requirements data collection covers four dimensions: human resources, material resources, tools and equipment, and time resources. Human resource requirements include the required number of personnel, skill level requirements, and estimated work hours. Material resource requirements include spare parts lists, consumable lists, and quantity and specification requirements. Tool and equipment requirements include the model and specifications of specialized tools, general-purpose tools, and testing equipment. Time resource requirements include the total maintenance duration, time allocation for each step, and downtime impact assessment. Resource requirement data is obtained by querying a resource management database, which stores standard resource configuration templates for various maintenance activities.

[0111] On-site construction condition data is collected in real time via an IoT sensor network, acquiring environmental parameters such as temperature, humidity, illuminance, noise level, space dimensions, and access conditions. Sensor data is collected every 5 minutes, and data transmission is conducted wirelessly using the LoRaWAN protocol to ensure reliable long-distance communication. An on-site video monitoring system provides real-time footage, analyzing personnel activities, equipment placement, and safety hazards using image recognition algorithms. A construction condition assessment model calculates a site suitability index based on the collected data, ranging from 0 to 100, with higher values ​​indicating more favorable construction conditions.

[0112] Equipment operating status data is continuously acquired through a SCADA system. Monitoring parameters include equipment operating mode, load rate, efficiency indicators, wear level, and historical fault records. The data acquisition cycle is set to 1 minute, increasing to once every 10 seconds under abnormal conditions. Operating status data undergoes filtering to remove noise interference using a 3rd-order Butterworth low-pass filter with a cutoff frequency of 0.1Hz. The equipment health assessment algorithm is based on a multiple regression model; it takes current operating parameters as input and outputs an equipment health score ranging from 0 to 100.

[0113] The scoring function is designed using a multilayer perceptron neural network structure, containing three hidden layers with 128, 64, and 32 neurons per layer, respectively. The input layer receives maintenance resource requirement data, on-site construction condition data, and equipment operating status data, all with a uniform 64-dimensional data dimension. The network activation function uses the ReLU function, and the output layer uses the Sigmoid function to ensure that the output value is between 0 and 1. The network is trained using the backpropagation algorithm, with a learning rate of 0.001, a batch size of 32, and 1000 training epochs.

[0114] The resource matching score calculation compares and analyzes currently available resources with those required by the maintenance plan. Human resource matching is calculated using a skills matching matrix, material resource matching is calculated by comparing inventory quantities, and tool and equipment matching is calculated by checking availability status. The matching score calculation results are synthesized using a weighted average method, with human resources weighted at 0.4, materials at 0.35, and tools at 0.25. The matching score output ranges from 0 to 100; a score below 60 triggers a resource allocation early warning mechanism.

[0115] The feasibility assessment comprehensively considers three factors: site environmental suitability, safety risk level, and workspace adequacy. Environmental suitability is determined through a comprehensive evaluation of temperature, humidity, illuminance, and noise levels, with suitability thresholds of 15-35 degrees Celsius, 30-80% relative humidity, illuminance not less than 200 lux, and noise not exceeding 85 decibels. The safety risk level is calculated using a risk assessment matrix, considering operational hazards, the completeness of protective measures, and the feasibility of emergency plans. Workspace adequacy is assessed through three-dimensional spatial modeling to ensure that the minimum operating space required for maintenance work is met.

[0116] The equipment adaptability score assesses the degree to which the current equipment status is suited to the maintenance plan, including the equipment's downtime window, the impact of maintenance operations on the equipment, and the expected performance recovery after maintenance. The downtime window is determined through a query in the production scheduling system, and the impact assessment considers the invasiveness of the maintenance operation and the complexity of the equipment structure. The expected performance recovery is predicted using a model built from historical maintenance performance data. The model employs a support vector regression algorithm with a radial basis function kernel and a regularization parameter set to 1.0.

[0117] The feasibility score calculation involves a weighted sum of three evaluation indicators according to preset weights: resource matching degree (0.4), construction feasibility degree (0.35), and equipment adaptability (0.25). These weight parameters are determined through expert experience and historical data analysis and can be adjusted based on the actual application scenario. The final feasibility score ranges from 0 to 100. A score greater than 80 allows for direct implementation of the maintenance plan; a score between 60 and 80 requires risk assessment and contingency plan preparation; and a score below 60 necessitates a revised maintenance plan or the allocation of additional resources.

[0118] In one optional implementation, the parameter change curves and maintenance effect acceptance data during the maintenance operation are dynamically tracked. The standard compliance and effectiveness of the maintenance process are evaluated through the fault diagnosis reasoning chain, resulting in equipment operation and maintenance optimization suggestions, including:

[0119] During the maintenance operation, the equipment operating parameters are continuously sampled to construct a time-series data stream. Based on the time-series data stream, parameter change curves are plotted. At the same time, the equipment performance test results and fault elimination verification records are obtained after the maintenance operation is completed, forming maintenance effect acceptance data.

[0120] Based on the parameter change curve, operation timing constraints and parameter response feature patterns are extracted from the fault diagnosis inference chain. The parameter change curve is divided into multiple time periods. For the parameter change slope in each time period, pattern matching is performed on the parameter response feature patterns to identify abnormal time periods in the parameter change curve that deviate from the parameter response feature patterns. The maintenance operation identifiers corresponding to the abnormal time periods are correlated with the operation timing constraints to quantify the standard compliance deviation of the maintenance process.

[0121] Using the maintenance effect acceptance data, the equipment performance test results are compared with the preset performance recovery target value in the fault diagnosis reasoning chain to calculate the compliance rate. Based on the fault residual characteristics in the fault elimination verification record and the fault root cause node in the fault diagnosis reasoning chain, causal tracing is performed to form a quantitative indicator of the effectiveness of the treatment.

[0122] Based on the standard compliance deviation and the quantitative index of the effectiveness of the treatment, the corresponding implementation plan adjustment strategies and supplementary maintenance measures are retrieved from the knowledge graph of the fire pump station field. The implementation plan adjustment strategies and supplementary maintenance measures are then structured and organized to form equipment operation and maintenance optimization suggestions that include operation specification revisions and periodic maintenance plans.

[0123] like Figure 2 As shown, the method includes:

[0124] During maintenance operations, a sensor network installed at key locations in the fire pump station collects equipment operating parameters, including but not limited to pump inlet and outlet pressures, flow rates, motor current, bearing temperature, and vibration levels. These sensors continuously sample at a frequency of 5 times per second, transmitting the data to the data acquisition unit via fieldbus. The data acquisition unit performs preliminary filtering of the raw data, removes outliers, constructs a time-series data stream, and stores this data in real-time in a time-series database.

[0125] Based on the acquired time-series data stream, a sliding window technique is used to plot parameter variation curves. The window size is set from 10 to 60 seconds depending on the characteristics of different parameters. For example, a 10-second window can clearly capture the transient characteristics of motor startup current variation during the startup process; while a 60-second window is used to smooth short-term fluctuations in bearing temperature variation. The parameter variation curves are plotted with time on the x-axis and the corresponding parameter values ​​on the y-axis to form a visual graph.

[0126] After maintenance is completed, standardized testing procedures are followed to obtain maintenance effectiveness acceptance data. This data includes: pump flow and pressure test values ​​under rated operating conditions, start-stop response time, noise and vibration measurement values, motor insulation resistance and other performance test results, as well as fault elimination verification records obtained through infrared thermal imagers, ultrasonic testing and other means.

[0127] The system extracts operational timing constraints and parameter response characteristic patterns relevant to the current maintenance task from a pre-established fault diagnosis reasoning chain knowledge base. Operational timing constraints describe the standard execution sequence, time interval requirements, and key control points of maintenance operations, such as "After replacing the pump bearing, it must first run at low speed for 30 minutes before gradually increasing to the rated speed." Parameter response characteristic patterns define the expected trends in parameter changes under standard maintenance operations, such as "After bearing replacement, the vibration value should show a decay curve of first high and then low, and stabilize within the standard value range within 2 hours."

[0128] The parameter change curves are divided into multiple time periods according to the key nodes of the maintenance operation. For example, the fire pump seal replacement operation can be divided into six time periods: "disassembly preparation - seal removal - seal cavity cleaning - new seal installation - tightening and adjustment - trial operation". The slope of the parameter change in each time period is calculated, and the calculation results are matched with the parameter response characteristic pattern.

[0129] Pattern matching employs a dynamic time warping algorithm to calculate the similarity between the parameter variation curve and the standard response feature pattern. When the similarity is lower than a preset threshold (usually set to 0.85), the time period is identified as an abnormal time period. For the identified abnormal time periods, the corresponding maintenance operation identifiers are further extracted and their correlation with the operation timing constraints is verified.

[0130] Correlation verification quantifies the deviation from standard compliance by calculating the edit distance between the actual operation sequence and the standard operation sequence. The larger the edit distance, the more serious the deviation from the standard specification. For example, if the "rotor balancing check" operation is found to be performed prematurely or omitted during the replacement of a fire pump impeller, it will result in a high deviation from the standard compliance.

[0131] The effectiveness of maintenance procedures is evaluated using maintenance performance acceptance data. Equipment performance test results are compared with preset performance recovery target values ​​in the fault diagnosis inference chain to calculate the compliance rate. The compliance rate is defined as the ratio of the number of performance indicators actually restored to the number of target indicators. For example, if 7 out of 8 performance indicators meet the target after fire pump maintenance, the compliance rate is 87.5%.

[0132] Causal tracing is performed based on residual fault characteristics (such as minor vibrations and small leaks) in the fault elimination verification record and the root cause nodes in the fault diagnosis inference chain. Causal tracing employs a Bayesian network model to calculate the conditional probability relationship between residual fault characteristics and each root cause node, quantifying the effectiveness index of the handling. This index reflects the degree to which maintenance work addresses the root cause of the fault.

[0133] Based on the quantitative indicators of standard compliance deviation and handling effectiveness, corresponding implementation plan adjustment strategies and supplementary maintenance measures are retrieved from the knowledge graph of the fire pump station domain. The knowledge graph contains multi-dimensional information such as equipment type, failure mode, maintenance measures, and expert experience, and the most suitable optimization suggestions are located through graph inference algorithms.

[0134] The retrieved implementation plan adjustment strategies and supplementary maintenance measures are structured and organized to form equipment operation and maintenance optimization suggestions. These suggestions consist of two parts: revised operating procedures and periodic maintenance plans. Revised operating procedures provide detailed guidelines for improving operating steps, addressing areas with significant deviations from standard compliance. Periodic maintenance plans, based on effectiveness analysis, provide recommendations for subsequent maintenance timelines, key inspection items, and preventative measures.

[0135] Through the above implementation methods, it is possible to achieve comprehensive and dynamic tracking and evaluation of the fire pump station maintenance process, effectively improve maintenance quality and equipment reliability, and provide stronger protection for fire safety.

[0136] In one optional implementation, the compliance rate is calculated by comparing the equipment performance test results with the preset performance recovery target value in the fault diagnosis inference chain using the maintenance effect acceptance data. Furthermore, causal tracing is performed based on the fault residual characteristics in the fault elimination verification record and the fault root cause nodes in the fault diagnosis inference chain to form quantitative indicators of treatment effectiveness, including:

[0137] The equipment performance test results are analyzed from the maintenance effect acceptance data to obtain the measured performance parameter values ​​of multiple equipment performance dimensions, and the corresponding performance recovery target values ​​are obtained from the fault diagnosis inference chain.

[0138] The deviation between the measured performance parameter values ​​of each device performance dimension and the performance recovery target value is calculated and it is determined whether the standard is met. The percentage of the number of device performance dimensions that meet the standard is obtained to obtain the performance compliance rate. Based on the influence weight of each device performance dimension in the fault diagnosis inference chain on the overall operational reliability of the device, the influence weight of the non-compliant dimensions is accumulated to obtain the performance defect impact degree.

[0139] The fault residual features identified in the fault elimination verification record are matched with the fault feature descriptions associated with the fault root cause nodes in the fault diagnosis reasoning chain to determine the corresponding fault root cause nodes.

[0140] In the fault diagnosis reasoning chain, the fault root cause node is traced back to the fault triggering condition node. It is determined whether the associated equipment operating environment parameters are in the fault recurrence risk range. If so, the corresponding fault root cause node is marked as not eliminated. The proportion of fault root cause nodes in the not eliminated state is counted to obtain the root cause residue rate. The performance compliance rate, the performance defect impact degree and the root cause residue rate are comprehensively calculated to form a quantitative indicator of the effectiveness of the treatment.

[0141] The equipment performance test results are analyzed from maintenance effectiveness acceptance data to obtain measured performance parameter values ​​for multiple equipment performance dimensions. Taking an industrial compressor as an example, parameters for multiple performance dimensions, including exhaust pressure, exhaust temperature, vibration value, bearing temperature, and sealing performance, are obtained through testing. Simultaneously, corresponding performance recovery target values ​​are extracted from the fault diagnosis reasoning chain. These target values ​​are pre-set based on normal equipment operating parameters and manufacturer specifications. For example, the performance recovery target value for vibration is set to no more than 3.5 mm / s, while the target value for bearing temperature is no more than 65℃.

[0142] For each performance dimension, the deviation between the measured performance parameter value and the performance recovery target value is calculated, and it is determined whether the standard is met. The judgment criteria are set according to the characteristics of different performance parameters. For example, the allowable deviation range for pressure parameters is ±5%, and the allowable deviation range for temperature parameters is ±2℃. In the above industrial compressor case, if the measured vibration value is 3.2mm / s, which is lower than the target value of 3.5mm / s, it is judged as meeting the standard; if the measured bearing temperature is 68℃, which is higher than the target value of 65℃ and exceeds the allowable deviation range, it is judged as not meeting the standard.

[0143] The number of all compliant performance dimensions is counted, and their proportion to the total number of performance dimensions is calculated to obtain the performance compliance rate. If 8 out of 10 performance dimensions are compliant, the performance compliance rate is 80%. Simultaneously, based on the pre-defined impact weights of each performance dimension on the overall operational reliability of the equipment in the fault diagnosis inference chain, the impact weights of the non-compliant dimensions are accumulated to calculate the performance defect impact degree. For example, if the impact weight of bearing temperature is 0.15 and the impact weight of sealing performance is 0.1, and both of these dimensions are non-compliant, then the performance defect impact degree is 0.25.

[0144] Analyze the residual fault features identified in the fault elimination verification records, such as abnormal sounds, localized overheating, or intermittent parameter fluctuations. Match these features with the fault feature descriptions associated with the root cause nodes in the fault diagnosis inference chain. Determine the corresponding root cause nodes through semantic similarity calculation or feature pattern matching. For example, if the verification record contains the residual feature of "slight metallic friction sound at startup," it matches the "bearing wear" root cause node in the fault diagnosis inference chain.

[0145] In the fault diagnosis reasoning chain, the identified root cause node is traced upwards to the fault triggering condition node, and the associated equipment operating environment parameters are checked to see if they are within the fault recurrence risk range. For the root cause of bearing wear, the triggering conditions include excessively low lubricating oil pressure or oil deterioration. If the current lubricating oil pressure is within the normal range but close to the lower threshold, the "bearing wear" root cause is marked as an unresolved state. The proportion of all unresolved root cause nodes to the total number of root cause nodes is counted to obtain the root cause residue rate. If two out of five root cause nodes are in an unresolved state, the root cause residue rate is 40%.

[0146] The performance compliance rate, the impact of performance defects, and the root cause retention rate are combined to form the final quantitative indicator of the effectiveness of the response. The calculation formula is: Effectiveness Index = Performance Compliance Rate × (1 - Impact of Performance Defects) × (1 - Root Cause Retention Rate). For the above case, the effectiveness index = 80% × (1 - 0.25) × (1 - 0.4) = 80% × 0.75 × 0.6 = 36%. The higher this index, the better the effectiveness of the fault response.

[0147] In practical applications, the selection of performance dimensions, weight allocation, and calculation methods of quantitative indicators can be adjusted for different types of equipment and failure scenarios to ensure that the evaluation results are more consistent with the actual situation. Quantitative indicators of handling effectiveness can not only be used to evaluate the effect of a single maintenance operation, but also serve as a basis for optimizing maintenance strategies and making predictive maintenance decisions, thereby improving the scientific nature and accuracy of equipment management.

[0148] When the effectiveness of remediation measures is low, targeted improvement measures can be developed based on specific values ​​in three areas: performance compliance rate, impact of performance defects, and root cause persistence rate. For example, for performance dimensions that fail to meet standards, the causes can be analyzed in depth and additional adjustment measures can be taken; for problems with high root cause persistence rates, monitoring of potential risk factors and preventative maintenance can be strengthened. This precise assessment and targeted improvement based on quantitative indicators helps to continuously improve the quality of equipment maintenance and operational reliability.

[0149] A second aspect of the present invention provides an intelligent inspection and remote diagnostic expert system for fire pump stations, comprising:

[0150] The first unit is used to acquire real-time operation data and historical maintenance records of fire pump stations, map the real-time operation data and historical maintenance records to a knowledge graph in the field of fire pump stations, identify equipment status nodes and abnormal operation characteristics through entity alignment and relational reasoning, and construct a fault diagnosis reasoning chain.

[0151] The second unit is used to analyze the propagation path and impact range of abnormal equipment status based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, and generate targeted maintenance suggestions and handling measures by combining historical maintenance experience and equipment manuals, and to rank the priority of each handling measure.

[0152] The third unit is used to retrieve similar historical maintenance cases in the knowledge graph of the fire pump station field based on the current abnormal scenario, and to calculate the feasibility score of the maintenance plan by comprehensively considering the maintenance resource requirements, on-site construction conditions and equipment operating status through data mining and pattern matching technology.

[0153] The fourth unit is used to optimize the maintenance plan based on the feasibility score, construct the optimal maintenance execution sequence by combining the equipment operation level, fault development trend and maintenance resource configuration, and convert the optimal maintenance execution sequence into a fire pump station control instruction set; dynamically track the parameter change curve and maintenance effect acceptance data during the maintenance operation, evaluate the standard compliance and handling effectiveness of the maintenance process through the fault diagnosis reasoning chain, and form equipment operation and maintenance optimization suggestions.

[0154] A third aspect of the present invention provides an electronic device, comprising:

[0155] processor;

[0156] Memory used to store processor-executable instructions;

[0157] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0158] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0159] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An expert method for intelligent inspection and remote diagnosis of fire pump stations, characterized in that, include: The real-time operation data and historical maintenance records of the fire pump station are obtained, and the real-time operation data and historical maintenance records are mapped to the knowledge graph of the fire pump station domain. The equipment status nodes and abnormal operation characteristics are identified through entity alignment and relation reasoning, and a fault diagnosis reasoning chain is constructed. Based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment status are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and handling measures are generated, and the priority of each handling measure is ranked. Based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, the feasibility score of the maintenance plan is calculated through data mining and pattern matching techniques. Based on the feasibility score, the maintenance plan is optimized. Combining equipment operating level, fault development trend, and maintenance resource allocation, an optimal maintenance execution sequence is constructed and converted into a fire pump station control command set. Parameter change curves and maintenance effect acceptance data are dynamically tracked during maintenance operations. The standard compliance and effectiveness of the maintenance process are evaluated through the fault diagnosis reasoning chain, resulting in equipment operation and maintenance optimization suggestions, including: During the maintenance operation, the equipment operating parameters are continuously sampled to construct a time-series data stream. Based on the time-series data stream, parameter change curves are plotted. At the same time, the equipment performance test results and fault elimination verification records are obtained after the maintenance operation is completed, forming maintenance effect acceptance data. Based on the parameter change curve, operation timing constraints and parameter response feature patterns are extracted from the fault diagnosis inference chain. The parameter change curve is divided into multiple time periods. For the parameter change slope in each time period, pattern matching is performed on the parameter response feature patterns to identify abnormal time periods in the parameter change curve that deviate from the parameter response feature patterns. The maintenance operation identifiers corresponding to the abnormal time periods are correlated with the operation timing constraints to quantify the standard compliance deviation of the maintenance process. Using the maintenance effect acceptance data, the equipment performance test results are compared with the preset performance recovery target value in the fault diagnosis reasoning chain to calculate the compliance rate. Based on the fault residual characteristics in the fault elimination verification record and the fault root cause node in the fault diagnosis reasoning chain, causal tracing is performed to form a quantitative indicator of the effectiveness of the treatment. Based on the standard compliance deviation and the quantitative index of the effectiveness of the treatment, the corresponding implementation plan adjustment strategies and supplementary maintenance measures are retrieved from the knowledge graph of the fire pump station field. The implementation plan adjustment strategies and supplementary maintenance measures are then structured and organized to form equipment operation and maintenance optimization suggestions that include operation specification revisions and periodic maintenance plans.

2. The method according to claim 1, characterized in that, Mapping the real-time operational data and historical maintenance records to a knowledge graph in the fire pump station domain, and identifying equipment status nodes and operational anomaly characteristics through entity alignment and relational reasoning, a fault diagnosis reasoning chain is constructed, including: The equipment status parameters in the real-time operation data and the equipment identifiers in the historical maintenance records are standardized. The standardized equipment identifiers are then matched with equipment entity nodes in the knowledge graph of the fire pump station domain to generate entity alignment results. Based on the entity alignment results, state parameters that deviate from the normal operating range in the real-time operating data are extracted as abnormal features. Equipment status nodes associated with the abnormal features are retrieved in the knowledge graph of the fire pump station domain. Potential fault source nodes and affected equipment nodes are identified by tracing the upstream and downstream relationships of the equipment status nodes, and a node association graph is established. Based on the node association graph, the causal relationship edges between the potential fault source node and the affected equipment node are analyzed. Combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the knowledge graph of the fire pump station domain to obtain a complete reasoning path from abnormal features to the root cause of the fault. Arrange the node sequence and relationship type in the complete reasoning path according to the time evolution order, associate the fault probability and impact range attributes corresponding to each node, and construct a fault diagnosis reasoning chain that includes the fault propagation chain and diagnostic basis.

3. The method according to claim 2, characterized in that, Based on the node association graph, the causal relationship edges between the potential fault source nodes and the affected equipment nodes are analyzed. Combined with the fault types and maintenance measures in the historical maintenance records, multi-hop relationship reasoning is performed in the fire pump station domain knowledge graph to obtain the complete reasoning path from abnormal features to the root cause of the fault, including: Extract the causal relationship edges between potential fault source nodes and affected device nodes from the node association graph, and obtain the relationship type and propagation direction attributes associated with each causal relationship edge to form an initial causal relationship set; Based on the relationship types in the initial causal relationship set, fault propagation rules matching the relationship types are retrieved in the knowledge graph of the fire pump station domain, and the fault types and associated maintenance measures corresponding to the potential fault source nodes in the historical maintenance records are extracted. The fault propagation rules are semantically matched with the fault types, and candidate inference starting point nodes that match the current abnormal characteristics are selected. Starting with the candidate inference starting node, a multi-hop relationship traversal is performed in the knowledge graph of the fire pump station domain along the propagation direction of the causal relationship edge. At each jump, the validity of the jump path is verified according to the fault evolution sequence recorded in the historical maintenance record, and the fault probability and impact range attributes of each traversed node are recorded. All traversal paths formed during the multi-hop relationship traversal are evaluated. The path confidence is calculated based on the product of the fault probabilities of the nodes in each traversal path and the path length. The traversal path with the highest confidence is selected as the complete inference path from the abnormal features to the root cause of the fault.

4. The method according to claim 1, characterized in that, Based on the aforementioned knowledge graph of fire pump stations and the aforementioned fault diagnosis reasoning chain, the propagation path and impact range of abnormal equipment conditions are analyzed. Combined with historical maintenance experience and equipment manuals, targeted maintenance suggestions and handling measures are generated, including: By using the fault diagnosis reasoning chain to locate the root cause node of the fault, and by using the knowledge graph of the fire pump station domain to evaluate the diffusion effect of the root cause node along the relation edges, the affected downstream equipment nodes are tracked to construct the propagation path of the abnormal state of the equipment. Based on the equipment type and operating status of each downstream equipment node in the propagation path, and combined with the fault impact coefficient and the number of associated equipment of each downstream equipment node in the knowledge graph of the fire pump station field, the impact range of abnormal equipment status is quantified. For the equipment nodes within the affected area, historical maintenance records are compared with historical maintenance cases corresponding to the fault types of the equipment nodes. The maintenance operations and handling results in the historical maintenance cases are integrated and associated with the maintenance specifications and operational constraints of the equipment nodes in the equipment manual knowledge base. A compatibility analysis is performed on the maintenance operations and the maintenance specifications. Maintenance measures that meet the requirements of the equipment manual and have been proven effective in the historical maintenance cases are selected. The execution priority of the maintenance measures is determined based on the failure impact coefficient of each equipment node in the scope of influence, and targeted maintenance suggestions and handling measures are formed.

5. The method according to claim 1, characterized in that, Based on the current abnormal scenario, similar historical maintenance cases are retrieved from the knowledge graph of the fire pump station domain. Taking into account maintenance resource requirements, on-site construction conditions, and equipment operating status, the feasibility score of the maintenance plan is calculated using data mining and pattern matching techniques, including: In the knowledge graph of the fire pump station domain, a feature vector of the current abnormal scenario is constructed, and the equipment type identifier, fault type identifier and operating parameters are extracted from the feature vector. Using the equipment type identifier, the fault type identifier and the operating parameters, the correlation degree between the historical maintenance cases in the knowledge graph of the fire pump station domain and the current abnormal scenario is calculated, and a historical maintenance case correlation degree sequence is generated. Based on the correlation sequence of the historical maintenance cases, the historical maintenance case with the highest correlation is selected as the reference maintenance plan, and the maintenance steps and execution conditions in the reference maintenance plan are extracted; for the maintenance steps and execution conditions, the corresponding maintenance resource requirement data, on-site construction condition data, and equipment operation status data are collected; The maintenance resource requirement data, the on-site construction condition data, and the equipment operating status data are input into a preset scoring function to generate scoring indicators that reflect the resource matching degree, construction feasibility, and equipment adaptability of the maintenance plan; the feasibility score of the maintenance plan is calculated based on the scoring indicators and preset indicator weights.

6. The method according to claim 1, characterized in that, Using the maintenance effectiveness acceptance data, the equipment performance test results are compared with the preset performance recovery target value in the fault diagnosis inference chain to calculate the compliance rate. Furthermore, based on the fault residual characteristics in the fault elimination verification record and the fault root cause node in the fault diagnosis inference chain, causal tracing is performed to form quantitative indicators of treatment effectiveness, including: The equipment performance test results are analyzed from the maintenance effect acceptance data to obtain the measured performance parameter values ​​of multiple equipment performance dimensions, and the corresponding performance recovery target values ​​are obtained from the fault diagnosis inference chain. The deviation between the measured performance parameter values ​​of each device performance dimension and the performance recovery target value is calculated and it is determined whether the standard is met. The percentage of the number of device performance dimensions that meet the standard is obtained to obtain the performance compliance rate. Based on the influence weight of each device performance dimension in the fault diagnosis inference chain on the overall operational reliability of the device, the influence weight of the non-compliant dimensions is accumulated to obtain the performance defect impact degree. The fault residual features identified in the fault elimination verification record are matched with the fault feature descriptions associated with the fault root cause nodes in the fault diagnosis reasoning chain to determine the corresponding fault root cause nodes. In the fault diagnosis reasoning chain, the fault root cause node is traced back to the fault triggering condition node. It is determined whether the associated equipment operating environment parameters are in the fault recurrence risk range. If so, the corresponding fault root cause node is marked as not eliminated. The proportion of fault root cause nodes in the not eliminated state is counted to obtain the root cause residue rate. The performance compliance rate, the performance defect impact degree and the root cause residue rate are comprehensively calculated to form a quantitative indicator of the effectiveness of the treatment.

7. A fire pump station intelligent inspection and remote diagnosis expert system, used to implement the method of any one of claims 1-6, characterized in that, include: The first unit is used to acquire real-time operation data and historical maintenance records of fire pump stations, map the real-time operation data and historical maintenance records to a knowledge graph in the field of fire pump stations, identify equipment status nodes and abnormal operation characteristics through entity alignment and relational reasoning, and construct a fault diagnosis reasoning chain. The second unit is used to analyze the propagation path and impact range of abnormal equipment status based on the knowledge graph of the fire pump station field and the fault diagnosis reasoning chain, and generate targeted maintenance suggestions and handling measures by combining historical maintenance experience and equipment manuals, and to rank the priority of each handling measure. The third unit is used to retrieve similar historical maintenance cases in the knowledge graph of the fire pump station field based on the current abnormal scenario, and to calculate the feasibility score of the maintenance plan by comprehensively considering the maintenance resource requirements, on-site construction conditions and equipment operating status through data mining and pattern matching technology. The fourth unit is used to optimize the maintenance plan based on the feasibility score, construct the optimal maintenance execution sequence by combining the equipment operation level, fault development trend and maintenance resource configuration, and convert the optimal maintenance execution sequence into a fire pump station control instruction set; dynamically track the parameter change curve and maintenance effect acceptance data during the maintenance operation, evaluate the standard compliance and handling effectiveness of the maintenance process through the fault diagnosis reasoning chain, and form equipment operation and maintenance optimization suggestions.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images