Intelligent inspection method for secondary water supply pump station based on expert system

Abstract

The application discloses a secondary water supply pump station intelligent inspection method based on an expert system, relates to the technical field of intelligent inspection, and comprises the following steps: generating an augmented case package, updating an expert system knowledge base, and then issuing an interface link to each edge side inspection device; constructing a working condition portrait of each device; collecting current operation data of each device in the pump station by the inspection device, performing simple rule identification and edge side data processing on the current operation data; performing logical reasoning on the expert system knowledge base, generating a fault tree diagram and a fault reasoning result, and associating historical similar cases according to the fault tree diagram; verifying and detecting the working condition portrait, returning to logical reasoning again if the verification fails, generating a final treatment scheme if the verification succeeds; constructing a health degree index of each device, and generating a predictive maintenance scheme according to the health degree index. The application has the effects of improving the intelligence, comprehensiveness and foresight of the secondary water supply pump station inspection.

Description

Technical Field

[0001] This application relates to the technical field of intelligent inspection, and in particular to an intelligent inspection method for secondary water supply pumping stations based on expert systems. Background Technology

[0002] Secondary water supply pumping stations are a crucial component of urban public water supply systems. They are primarily used to pressurize tap water a second time and deliver it to high-rise buildings or homes of users located far away.

[0003] In existing technologies, the inspection of secondary water supply pumping stations mainly relies on periodic manual checks or automated monitoring based on fixed thresholds. Traditional inspection methods are problematic. Firstly, manual inspections consume significant manpower and are prone to errors, leading to resource waste. Furthermore, for complex situations, they cannot promptly link to other pumping stations for verification, resulting in low efficiency and accuracy. Secondly, automated monitoring based on fixed thresholds is rigid, relying on single-point detection, which can cause delays in responding to fault modes and lacks predictive fault reporting for equipment within the pumping station, resulting in a passive response and limited effectiveness. Therefore, improving the intelligence, comprehensiveness, and foresight of secondary water supply pumping station inspections is a pressing issue. Summary of the Invention

[0004] The purpose of this invention is to provide a smart inspection method for secondary water supply pumping stations based on an expert system, so as to solve the problems mentioned in the background art.

[0005] This application provides a smart inspection method for secondary water supply pumping stations based on an expert system, the method comprising: The system acquires the original fault data, handling measures data, and manual correction data for each fault handling, generates an augmented case package, and pushes the augmented case package to the preset expert system knowledge base. After updating the expert system knowledge base, it sends an interface link to each edge-side inspection device. Collect sufficient multi-dimensional data for each device in the pumping station within a preset time period, and construct a working condition profile for each device based on the multi-dimensional data. The inspection equipment collects the current operating data of each device in the pumping station, performs simple rule recognition and edge data processing on the current operating data to obtain simple fault information and clean feature data, processes the simple fault information, and uploads the clean feature data through the interface link; The expert system knowledge base performs logical reasoning based on the clean feature data, records the reasoning rules and data used in each step of the reasoning, generates a fault tree diagram and fault reasoning results, and associates historical similar cases based on the fault tree diagram. Based on the fault reasoning results and the historical similar cases, the working condition profile is verified and tested. If the verification fails, the logical reasoning is restarted. If the verification is successful, similar data from neighboring pumping stations within the same time period are retrieved for horizontal comparison to obtain comparison results. Based on the comparison results, common or individual abnormal results are generated, and a final processing plan is generated. Based on the operating condition profile and the current operating data, a health index is constructed for each device, and a predictive maintenance plan is generated based on the health index.

[0006] Preferably, the steps of acquiring the original fault data, handling measure data, and manual correction data for each fault handling, generating an augmented case package, pushing the augmented case package to a preset expert system knowledge base, updating the expert system knowledge base, and then issuing an interface link to each edge-side inspection device are as follows: Obtain the original fault data, handling measures data, and manual correction data for each fault handling at each pumping station, and generate an augmented case package; The augmented case package is pushed to a preset expert system knowledge base, and the expert system knowledge base compares the augmented case package with local cases to obtain a similarity value between the augmented case package and the local cases. Based on the similarity value, the data in the augmented case package is extracted to obtain the target effective data; Based on the target valid data, new inference rules are generated or the confidence of existing inference rules is supplemented, and then the expert system knowledge base is updated. To generate the updated expert system knowledge base interface link, the interface link is distributed to the inspection equipment on each edge side.

[0007] Preferably, the step of collecting sufficient multi-dimensional data for each device in the pumping station within a preset time period, and constructing a working condition profile for each device based on the multi-dimensional data, specifically includes: Collect sufficient multi-dimensional data for each device in the pumping station within a preset time period. Most of the multi-dimensional data includes equipment pressure parameters, equipment flow parameters, equipment current parameters, and equipment temperature parameters. Based on the equipment pressure parameters and the equipment flow parameters, a graph showing the changes in water flow parameters during equipment operation is obtained; Based on the device current parameters and the device temperature parameters, a diagram showing the changes in device status during operation is obtained. By combining the water flow parameter variation diagram and the equipment status variation diagram, a normal operating parameter range and a typical operating condition mode library for the equipment are established. Based on the normal operating parameter range and the typical operating condition mode library, a working condition profile of each device during operation is constructed.

[0008] Preferably, the steps of performing simple rule recognition and edge-side data processing on the current running data to obtain simple fault information and clean feature data, processing the simple fault information, and uploading the clean feature data through the interface link are as follows: The current operating data is identified using simplified rule recognition based on the preset simplified identification rules in the local inspection equipment to obtain simplified fault information. The complexity of the simple fault information is identified to obtain the complexity of the simple fault information, and it is determined whether the complexity exceeds a preset complexity threshold. If it is determined that the complexity does not exceed the complexity threshold, a work order is generated and sent to the maintenance personnel; if it is determined that the complexity exceeds the complexity threshold, an unprocessed tag is generated. The current running data is cleaned, and features are extracted from the cleaned current running data to obtain clean feature data; The unprocessed tags are added to the clean feature data before uploading.

[0009] Preferably, the expert system knowledge base performs logical reasoning based on the clean feature data, records the reasoning rules and data used at each step, generates a fault tree diagram and fault reasoning results, and associates historical similar cases according to the fault tree diagram, specifically as follows: The expert system knowledge base extracts fault information from the clean feature data to obtain explicit fault information, and uses the explicit fault information as the logical root for logical reasoning. Record the reasoning rules and data used at each step of the logical reasoning process, and generate logic tree nodes based on the reasoning rules and the data used. By combining the logical root and the logical tree node, a fault tree diagram is generated, and multiple terminal result information in the fault tree diagram is collected; A confidence level assessment is performed on each of the terminal result information to obtain a confidence value for each of the terminal result information, and the results are filtered based on the confidence values ​​to obtain the fault reasoning results; Based on the fault reasoning results, a similarity query is performed in the expert system knowledge base to obtain historical similar cases associated with the fault tree diagram.

[0010] Preferably, based on the fault reasoning results and the historical similar cases, the working condition profile is verified and tested. If the verification fails, the process returns to the step of re-performing logical reasoning, specifically as follows: Based on the fault reasoning results and the historical similar cases, the inferred fault causes and historical fault phenomena are obtained; The inferred fault causes and the historical fault phenomena are substituted into the working condition profile to conduct a simulation exercise, and the exercise results are obtained. The similarity between the exercise results and the current running data is compared to obtain the target similarity value, and it is determined whether the target similarity value exceeds the preset standard value. If it is determined that the similarity value exceeds the standard value, then the reasoning failure cause and the historical similar cases are used as target result information and passed to the next step; If it is determined that the similarity value does not exceed the standard value, then the differential data between the exercise result and the current running data is identified, and the previous step is returned to perform logical reasoning again with the differential data as reference information.

[0011] Preferably, the steps of retrieving similar data from neighboring pumping stations within the same time period for horizontal comparison, obtaining comparison results, generating common or individual anomaly results based on the comparison results, and generating a final processing solution are as follows: Based on the reasoning failure cause and the historical similar cases, basic information is extracted from the reasoning failure cause and the historical similar cases to obtain the current basic failure information and failure time period; Based on the current basic fault information and the fault time period, retrieve similar data from neighboring pumping stations during the fault time period; The basic fault information is compared horizontally with the similar data to obtain the comparison results, and the fault reasoning results are determined to be either common abnormal results or individual abnormal results based on the comparison results. If the fault reasoning result is a common abnormal result, then the final processing solution for the entire region is generated based on the fault reasoning result; If the fault reasoning result is an individual anomaly, then a final processing plan for the region is generated based on the fault reasoning result.

[0012] Preferably, the step of constructing a health index for each device based on the operating condition profile and the current operating data, and generating a predictive maintenance plan based on the health index, specifically includes: Based on the described operating condition profile, the health status of the equipment is assessed to obtain the first health parameter of the equipment at the current time point; Based on the current operating data, the health status of the device is assessed to obtain the second health parameter of the device at the current time point; By combining the first health parameter and the second health parameter, a health index is constructed for each device, and the target component of the health index is identified. Based on the health index and the target component, a predictive maintenance plan is generated for each device.

[0013] In summary, this application includes at least one of the following beneficial technical effects: By acquiring raw fault data, handling measures data, and manually corrected data from each fault handling process, an augmented case package is generated. This augmented case package is used to continuously update the pre-set expert system knowledge base, and after updates, an interface link is generated and distributed to each edge-side inspection device. First, sufficient multi-dimensional data is collected from each device in the pumping station within a preset time period, and a working condition profile for each device is constructed based on this data. Then, the inspection devices collect current operating data from each device in the pumping station, performing simple rule recognition and edge-side data processing on this data to obtain simple fault information and clean feature data. After processing the simple fault information, the clean feature data is uploaded via the interface link. The expert system knowledge base performs logical reasoning based on the clean feature data, recording each step of the reasoning process, generating a fault tree diagram and fault reasoning results, and associating them with historical similar cases. The fault reasoning results are verified against the working condition profile. If verification fails, the reasoning is repeated. If verification succeeds, similar data from neighboring pumping stations within the same time period is retrieved for comparison to identify whether the fault reasoning result is a common or unique anomaly, and a final handling solution is generated. Finally, based on the operational profile and current running data, a health index is constructed for each piece of equipment, and predictive maintenance plans are generated based on the health index. This improves the intelligence, comprehensiveness, and foresight of the inspection of secondary water supply pumping stations. Attached Figure Description

[0014] Figure 1 This is the step flow of the intelligent inspection method for secondary water supply pumping stations based on an expert system provided in the embodiments of this application. Detailed Implementation

[0015] The following combination Figure 1 This application will be described in further detail, but the embodiments of the present invention are not limited thereto. This application discloses a smart inspection method for secondary water supply pumping stations based on an expert system.

[0016] In this embodiment, a smart inspection method for secondary water supply pumping stations based on an expert system is described, which includes: S100: Obtain the original fault data, handling measures data, and manual correction data for each fault handling, generate an augmented case package, and push the augmented case package to the preset expert system knowledge base. After updating the expert system knowledge base, send the interface link to each edge-side inspection device. S200: Collect sufficient multi-dimensional data of each device in the pumping station within a preset time period, and construct a working condition profile of each device based on the multi-dimensional data; S300: The inspection equipment collects the current operating data of each device in the pumping station, performs simple rule recognition and edge data processing on the current operating data to obtain simple fault information and clean feature data, processes the simple fault information, and uploads the clean feature data through the interface link; S400: The expert system knowledge base performs logical reasoning based on clean feature data, records the reasoning rules and data used in each step of reasoning, generates fault tree diagrams and fault reasoning results, and associates historical similar cases based on the fault tree diagrams; S500: Based on the fault reasoning results and similar historical cases, the working condition profile is verified and tested. If the verification fails, the logical reasoning is restarted. S600: If the verification is successful, retrieve similar data from neighboring pumping stations within the same time period for horizontal comparison, obtain the comparison results, generate common or individual abnormal results based on the comparison results, and generate the final processing solution. S700: Based on the operating condition profile and current operating data, it constructs a health index for each device and generates predictive maintenance plans based on the health index.

[0017] The steps for acquiring raw fault data, handling measures data, and manual correction data for each fault handling process, generating augmented case packages, pushing these packages to a pre-set expert system knowledge base, updating the expert system knowledge base, and then issuing interface links to each edge-side inspection device are as follows: Obtain the original fault data, handling measures data, and manual correction data for each fault handling at each pumping station, and generate an augmented case package; The augmented case package is pushed to the preset expert system knowledge base. The expert system knowledge base compares the augmented case package with the local cases to obtain the similarity value between the augmented case package and the local cases. The data in the augmented case package is extracted based on the similarity value to obtain the target effective data; Based on the target's valid data, new inference rules are generated or the confidence levels of existing inference rules are supplemented, and then the expert system's knowledge base is updated. To generate the updated expert system knowledge base interface link, the interface link is distributed to the inspection equipment on each edge side.

[0018] In application, taking a secondary water supply pumping station in District A of a certain city as an example, during the handling of a pump bearing overheating fault, the system acquired the original fault data (including a sudden 15% increase in pump current, bearing temperature reaching 85℃, and abnormally increased vibration amplitude at the time of the fault), handling measures data (maintenance personnel replaced the bearing and adjusted the lubrication system), and manual correction data (the engineer added in the record that excessively high ambient temperature was also one of the contributing factors). This data was packaged into an augmented case package. This case package was pushed to a pre-set expert system knowledge base. The knowledge base first compared this new case with hundreds of existing local cases in the database one by one, calculating the similarity through an algorithm. Assuming the comparison found that the similarity with a historical case about "insufficient lubrication leading to bearing overheating" reached 75%, while the similarity with another case about "sudden load change leading to increased current" was only 30%, the system extracted "high bearing temperature," "sudden current increase," and "effective lubrication system intervention" from the new case package as target effective data based on the 75% similarity value. Based on this data, the system did not generate entirely new inference rules. Instead, it supplemented and strengthened the confidence level of the existing inference rule "abnormal bearing lubrication - temperature rise - current fluctuation" in the knowledge base, increasing its confidence level from 82% to 89%. After completing this incremental update of the knowledge base, the system generated an interface link containing the latest rules and cases, and automatically distributed it to the edge-side inspection equipment deployed at the pump station in Area A and ten other similar pump stations throughout the city, enabling all equipment to conduct subsequent analysis based on the updated knowledge.

[0019] The steps for collecting sufficient multi-dimensional data from each piece of equipment in the pumping station within a preset time period, and constructing a working condition profile for each piece of equipment based on the multi-dimensional data, are as follows: Collect sufficient multi-dimensional data for each device in the pumping station within a preset time period. Most of the multi-dimensional data includes equipment pressure parameters, equipment flow parameters, equipment current parameters, and equipment temperature parameters. Based on the equipment pressure parameters and equipment flow parameters, a graph showing the changes in water flow parameters during equipment operation is obtained; Based on the equipment current parameters and equipment temperature parameters, a diagram showing the changes in equipment status during equipment operation is obtained. By combining the water flow parameter variation diagram and the equipment status variation diagram, a normal operating parameter range and typical operating condition mode library for the equipment are established; Based on the normal operating parameter range and typical operating condition mode library, construct the operating condition profile of each device during operation.

[0020] In application, taking a secondary water supply pumping station in District A of a certain city as an example, in order to construct a working condition profile of the core equipment (such as the main water supply pump) within the pumping station, the system collected sufficient multi-dimensional operational data of the equipment over a preset period of 30 consecutive days. This data specifically includes: equipment pressure parameters (outlet pressure value, sampling frequency 1 time / minute, totaling 43,200 data points), equipment flow parameters (instantaneous flow value, sampling frequency 1 time / minute), equipment current parameters (three-phase current value, sampling frequency 10 times / second), and equipment temperature parameters (pump body and bearing temperature, sampling frequency 1 time / 30 seconds). The system first plots the water flow parameter changes of the pump over 30 days based on the pressure and flow parameters. This plot shows that during the peak water usage periods in the morning and evening, the pressure curve exhibits a regular double-peak fluctuation, with a corresponding increase in flow rate. Simultaneously, based on the current and temperature parameters, it plots the equipment status change, showing that the current value changes synchronously with pressure fluctuations, while the bearing temperature stabilizes at around 65℃ after 8 hours of continuous operation. Combining these two variation graphs, the system, through machine learning algorithms, established normal operating parameter ranges for the pump under various typical operating conditions, including "normal water supply mode," "peak pressure boosting mode," and "nighttime low load mode" (for example, in peak mode, the pressure should be between 0.35-0.42 MPa, and the bearing temperature should be below 70℃). Finally, the system integrated all normal operating ranges and identified typical operating modes to construct a digital operating condition profile that comprehensively reflects the healthy operating status of the main water supply pump under different times and loads, serving as a benchmark model for subsequent fault diagnosis.

[0021] The steps involve performing simple rule recognition and edge data processing on the current running data to obtain simple fault information and clean feature data, processing the simple fault information, and uploading the clean feature data via an interface link. Specifically: Based on the simplified identification rules preset in the local inspection equipment, the current operating data is identified using simplified rules to obtain simplified fault information; The complexity of simple fault information is identified, the complexity of simple fault information is obtained, and it is determined whether the complexity exceeds a preset complexity threshold. If the complexity is determined to be less than or equal to the complexity threshold, a work order is generated and sent to the operations and maintenance personnel; if the complexity is determined to be greater than or equal to the complexity threshold, an unprocessed tag is generated. The current running data is cleaned, and features are extracted from the cleaned current running data to obtain clean feature data; After adding clean feature data to the unprocessed tags, upload them.

[0022] In operation, the inspection equipment at the secondary water supply pump station in District A of a certain city collected real-time operating data of the main water supply pump, including pressure, flow rate, current, and temperature. First, the equipment's built-in simple rule recognition engine started working. According to preset rules, such as "if the instantaneous bearing temperature reading > 75℃, then alarm," the engine scanned the data and found the current bearing temperature to be 72℃, thus not triggering an alarm. However, another rule, "if the three-phase current imbalance > 10%, then alarm," was triggered because the current current imbalance was detected to have reached 12%. The system determined this to be a simple fault. Next, the system assessed the complexity of this simple fault. It analyzed that the fault only involved a single parameter, current, and that most similar imbalance cases in the past were caused by external voltage dips, making them simple to handle. Therefore, it judged the complexity to be low, not exceeding the preset complexity threshold (for example, a threshold set to require the association of more than three parameters or involve internal mechanical damage). Thus, the system automatically generated a maintenance work order stating "inspect the external power supply line and calibrate the current sensor," and sent it directly to the mobile app of Mr. Wang, the maintenance personnel responsible for the area. Meanwhile, the system performs data cleaning on all collected raw current operating data in the background, removing null values ​​caused by brief interruptions in sensor communication and smoothing out individual sampling noise. After cleaning, the system extracts key features, such as the fundamental current amplitude, harmonic distortion rate, and temperature gradient, forming a clean and well-organized set of feature data. Finally, the system uploads this clean feature data to the cloud-based expert system knowledge base via the previously provided interface link, awaiting further in-depth analysis.

[0023] The expert system's knowledge base performs logical reasoning based on clean feature data, records the reasoning rules and data used at each step, generates a fault tree diagram and fault reasoning results, and associates the fault tree diagram with historical similar cases, specifically as follows: The expert system knowledge base extracts fault information from clean feature data to obtain explicit fault information, and uses the explicit fault information as the logical root for logical reasoning. Record the reasoning rules and data used at each step of the logical reasoning process, and generate logic tree nodes based on the reasoning rules and the data used. By combining logical roots and logical tree nodes, a fault tree diagram is generated, and multiple terminal result information in the fault tree diagram is collected. The confidence level of each terminal result is evaluated to obtain a confidence value for each terminal result. The results are then filtered based on the confidence values ​​to obtain the fault reasoning results. Based on the fault reasoning results, a similarity query is performed in the expert system knowledge base to obtain historical similar cases associated with the fault tree diagram.

[0024] In practice, the cloud-based expert system knowledge base received a clean feature dataset of the main water supply pump uploaded by a pumping station in District A of a certain city. The knowledge base first performed in-depth analysis of this feature data to extract fault information. It identified "current harmonic distortion rate significantly increased to 8%" (normally less than 5%) and "abnormally increased bearing temperature gradient" as the most prominent external fault information. The system used "abnormally increased current harmonic distortion rate" as the starting point (logic root) for logical reasoning. The reasoning process began: First, the rule engine called the rule "Increased harmonic distortion rate may originate from power quality or motor winding problems," recording this reasoning rule and the data "harmonic distortion rate 8%" used, generating the first logic tree node. Second, combining the feature "abnormal bearing temperature gradient," it called another rule "If accompanied by excessively rapid bearing temperature rise, the possibility of motor bearing failure or load mechanical friction increases," generating the second node. The system continued reasoning, considering branches such as "whether voltage parameters are stable" and "whether there are recent maintenance records," recording the rules and data used at each step. Finally, a fault tree diagram is generated, rooted at "current harmonic distortion" and containing multiple branches of possible causes (such as power supply problems, motor bearing damage, load jamming, etc.). The knowledge base collects information from several terminal results in this tree diagram, such as "severe wear of motor bearings" and "third harmonic injection in power supply lines." The system evaluates the confidence level of each terminal result. Based on historical case matching and feature similarity, the confidence level of "severe wear of motor bearings" is evaluated as 70%, while the confidence level of "power supply line problems" is 25%. According to a preset screening threshold (e.g., 60%), the system uses "severe wear of motor bearings" as the fault reasoning result for this time. Subsequently, the knowledge base uses this result and the fault tree diagram as a basis to perform similar queries in the case database, successfully linking three historical similar cases about "increased current harmonics due to wear of water pump motor bearings" for reference in subsequent steps.

[0025] Based on the fault reasoning results and similar historical cases, the working condition profile is verified and tested. If the verification fails, the process returns to the logical reasoning step and is repeated. Specifically: Based on the fault reasoning results and similar historical cases, the causes of faults and historical fault phenomena are obtained. The causes of the faults and historical fault phenomena are substituted into the working condition profile for simulation exercises to obtain the exercise results. The similarity between the exercise results and the current running data is compared to obtain the target similarity value, and it is determined whether the target similarity value exceeds the preset standard value. If the similarity value is determined to exceed the standard value, the reasoning for the failure and historical similar cases will be used as the target result information and passed to the next step. If the similarity value is determined to be no more than the standard value, then the differential data between the exercise results and the current running data is identified, and the previous step is returned to re-perform logical reasoning with the differential data as reference information.

[0026] In application, based on the fault reasoning result ("severe wear of motor bearings") obtained from the expert system's knowledge base and three related historical similar cases, the system begins to verify and detect the previously established operating condition profile of the main water supply pump. The system extracts the reasoned fault cause (bearing wear) and common fault phenomena from the historical cases (increased current harmonics, accelerated temperature rise, and subsequent abnormal noise) from the reasoning result and historical cases. Next, the system "injects" these fault causes and phenomena into the pump's digital operating condition profile for simulation. The simulation process, based on the equipment's physical model and operating logic, deduces how parameters such as pressure, flow rate, current, and temperature should change under the condition of bearing wear, thus obtaining a simulated "fault state" data sequence, i.e., the simulation result. Then, the system performs a detailed similarity comparison between this simulated simulation result and the current operating data actually collected by the inspection equipment. The calculated overall waveform and feature matching degree (target similarity value) is 85%. The system's preset standard verification value is 80%. Since 85% exceeds 80%, the system determines that the verification is successful. This means that the current actual fault data closely matches the data simulated based on the inference of "bearing wear," thus confirming the reliability of the reasoning result. Therefore, the system uses this inferred fault cause (bearing wear) and related historical cases as valid target result information, passing it to the next step for broader comparative analysis. If the similarity value does not exceed the standard value, for example, only 60%, the system will identify a significant difference between the simulated data and the actual data in the "degree of traffic decline," and feed this "differential data" back, requiring the knowledge base to use "traffic anomaly" as a new reference point to re-perform logical reasoning.

[0027] The steps for retrieving similar data from neighboring pumping stations within the same time period for horizontal comparison, obtaining comparison results, generating common or unique anomalies based on the comparison results, and finally generating a final processing plan are as follows: Based on the reasoning of fault causes and historical similar cases, basic information is extracted from the reasoning of fault causes and historical similar cases to obtain current basic fault information and fault time period. Based on the current basic fault information and fault time period, retrieve similar data from neighboring pumping stations during the fault time period; The basic fault information is compared horizontally with similar data to obtain the comparison results, and the fault inference results are determined to be either common abnormal results or individual abnormal results based on the comparison results. If the fault reasoning result is a common abnormal result, then the final processing plan for the entire region is generated based on the fault reasoning result; If the fault reasoning result is an individual anomaly, then a final processing plan for the region is generated based on the fault reasoning result.

[0028] In application, after the fault reasoning result (main pump motor bearing wear in pumping station A) is verified through the operating condition profile, the system further analyzes the scope of the fault's impact. The system first extracts basic fault information ("motor bearing wear") and the time period of the fault occurrence (e.g., within the last 24 hours) from the reasoning result and related historical cases. Then, based on this fault information and time period, the system automatically retrieves similar operating data (mainly current harmonics and bearing temperature data) from two geographically adjacent secondary water supply pumping stations (areas B and C, with the same equipment model) within the same 24 hours. The system then compares the basic fault characteristics of pumping station A (such as harmonic distortion rate increasing from 5% to 8%, and abnormal bearing temperature gradient) with similar data from pumping stations B and C. The comparison reveals that the current harmonic distortion rate of pumping station B only slightly increased from 4.8% to 5.1% during the same period, and the bearing temperature change was normal; the data for pumping station C also remained stable. The comparison results indicate that the fault characteristics of pumping station A are unique and not commonly observed in neighboring pumping stations of the same type. Therefore, the system determined that this "motor bearing wear" fault was an isolated anomaly, likely stemming from specific usage wear or an occasional mechanical problem at the pump in Area A, rather than a regional issue (such as widespread deterioration of power grid quality). Based on this assessment, the system's final solution will be targeted, potentially including: "Immediately shut down and inspect the No. 1 main water supply pump at pumping station A, focusing on the condition of the motor bearing, and prepare spare parts for replacement. Strengthen monitoring of similar parameters at pumping stations in Areas B and C, and temporarily refrain from large-scale preventative maintenance." This avoids unnecessary regional shutdowns and achieves precise maintenance.

[0029] The steps for constructing a health index for each device based on its operating condition profile and current operational data, and then generating a predictive maintenance plan based on that health index, are as follows: Based on the operating condition profile, the health status of the equipment is assessed to obtain the first health parameter of the equipment at the current point in time; Based on the current operating data, the health status of the equipment is assessed to obtain the second health parameter of the equipment at the current point in time; By combining the first and second health parameters, a health index is constructed for each device, and the target component of the health index is identified. Based on the health index and the target parts, a predictive maintenance plan is generated for each device.

[0030] In application, to perform predictive maintenance on equipment in a secondary water supply pumping station in District A of a certain city, the system comprehensively utilizes a pre-constructed equipment condition profile and real-time collected current operating data. First, the system evaluates the equipment based on its condition profile. The profile defines parameter ranges for the equipment under various healthy operating conditions. The system compares the theoretical parameter curves under the "healthy state" in the profile with the actual average state curves calculated from long-term operation statistics, assesses the degree of deviation, and calculates a first health parameter representing its long-term health trend, assumed to be 88 (out of 100). Second, the system evaluates the equipment based on current (or recent) operating data. It analyzes the immediate deviations of key characteristics (such as vibration amplitude, efficiency, and sealing performance) from the "healthy" baseline in the current data, calculating a second health parameter reflecting the immediate state, assumed to be 76. Then, the system combines the first health parameter (long-term trend 88) and the second health parameter (immediate state 76) through weighted fusion (e.g., long-term trend weight 40%, immediate state weight 60%) to construct the current comprehensive health index of the equipment, with a calculated result of 81.2. Further system analysis revealed that the main contributors to the decline in the health index were increased vibration and a slight decrease in efficiency. Through feature tracing, the system identified the primary components affected by these features as the "motor drive bearing" and the "pump impeller." Finally, based on the health index of 81.2 (within the "concern" range but not yet at the "alarm" level) and the identified components, the system generated a predictive maintenance plan, which might include: Prioritizing a condition check (such as vibration spectrum analysis) and lubrication maintenance of the motor drive bearing of the No. 1 main pump in Pump Station A over the next two weeks, and testing the pump efficiency. Simultaneously, the impeller condition check should be included in the maintenance plan for the next month.

[0031] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A smart inspection method for secondary water supply pumping stations based on expert systems, characterized in that, include: The system acquires the original fault data, handling measures data, and manual correction data for each fault handling, generates an augmented case package, and pushes the augmented case package to the preset expert system knowledge base. After updating the expert system knowledge base, it sends an interface link to each edge-side inspection device. Collect sufficient multi-dimensional data for each device in the pumping station within a preset time period, and construct a working condition profile for each device based on the multi-dimensional data. The inspection equipment collects the current operating data of each device in the pumping station, performs simple rule recognition and edge data processing on the current operating data to obtain simple fault information and clean feature data, processes the simple fault information, and uploads the clean feature data through the interface link; The expert system knowledge base performs logical reasoning based on the clean feature data, records the reasoning rules and data used in each step of the reasoning, generates a fault tree diagram and fault reasoning results, and associates historical similar cases based on the fault tree diagram. Based on the fault reasoning results and the historical similar cases, the working condition profile is verified and tested. If the verification fails, the logical reasoning is restarted. If the verification is successful, similar data from neighboring pumping stations within the same time period are retrieved for horizontal comparison to obtain comparison results. Based on the comparison results, common or individual abnormal results are generated, and a final processing plan is generated. Based on the operating condition profile and the current operating data, a health index is constructed for each device, and a predictive maintenance plan is generated based on the health index.

2. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 1, characterized in that, The steps of acquiring raw fault data, handling measure data, and manual correction data for each fault handling, generating an augmented case package, pushing the augmented case package to a preset expert system knowledge base, updating the expert system knowledge base, and then issuing interface links to each edge-side inspection device are as follows: Obtain the original fault data, handling measures data, and manual correction data for each fault handling at each pumping station, and generate an augmented case package; The augmented case package is pushed to a preset expert system knowledge base, and the expert system knowledge base compares the augmented case package with local cases to obtain a similarity value between the augmented case package and the local cases. Based on the similarity value, the data in the augmented case package is extracted to obtain the target effective data; Based on the target valid data, new inference rules are generated or the confidence of existing inference rules is supplemented, and then the expert system knowledge base is updated. To generate the updated expert system knowledge base interface link, the interface link is distributed to the inspection equipment on each edge side.

3. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 2, characterized in that, The steps for collecting sufficient multi-dimensional data from each device in the pumping station within a preset time period, and constructing a working condition profile for each device based on the multi-dimensional data, are as follows: Collect sufficient multi-dimensional data for each device in the pumping station within a preset time period. Most of the multi-dimensional data includes equipment pressure parameters, equipment flow parameters, equipment current parameters, and equipment temperature parameters. Based on the equipment pressure parameters and the equipment flow parameters, a graph showing the changes in water flow parameters during equipment operation is obtained; Based on the device current parameters and the device temperature parameters, a diagram showing the changes in device status during operation is obtained. By combining the water flow parameter variation diagram and the equipment status variation diagram, a normal operating parameter range and a typical operating condition mode library for the equipment are established. Based on the normal operating parameter range and the typical operating condition mode library, a working condition profile of each device during operation is constructed.

4. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 3, characterized in that, The steps of performing simple rule recognition and edge-side data processing on the current running data to obtain simple fault information and clean feature data, processing the simple fault information, and uploading the clean feature data through the interface link are as follows: The current operating data is identified using simplified rule recognition based on the preset simplified identification rules in the local inspection equipment to obtain simplified fault information. The complexity of the simple fault information is identified to obtain the complexity of the simple fault information, and it is determined whether the complexity exceeds a preset complexity threshold. If it is determined that the complexity does not exceed the complexity threshold, a work order is generated and sent to the maintenance personnel; if it is determined that the complexity exceeds the complexity threshold, an unprocessed tag is generated. The current running data is cleaned, and features are extracted from the cleaned current running data to obtain clean feature data; The unprocessed tags are added to the clean feature data before uploading.

5. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 4, characterized in that, The expert system knowledge base performs logical reasoning based on the clean feature data, records the reasoning rules and data used at each step, generates a fault tree diagram and fault reasoning results, and associates historical similar cases based on the fault tree diagram, specifically as follows: The expert system knowledge base extracts fault information from the clean feature data to obtain fault explicit information, and uses the fault explicit information as the logical root for logical reasoning. Record the reasoning rules and data used at each step of the logical reasoning process, and generate logic tree nodes based on the reasoning rules and the data used. By combining the logical root and the logical tree node, a fault tree diagram is generated, and multiple terminal result information in the fault tree diagram is collected; A confidence level assessment is performed on each of the terminal result information to obtain a confidence value for each of the terminal result information, and the results are filtered based on the confidence values ​​to obtain the fault reasoning results; Based on the fault reasoning results, a similarity query is performed in the expert system knowledge base to obtain historical similar cases associated with the fault tree diagram.

6. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 5, characterized in that, Based on the fault reasoning results and the historical similar cases, the working condition profile is verified and tested. If the verification fails, the process returns to the step of re-performing logical reasoning, specifically as follows: Based on the fault reasoning results and the historical similar cases, the inferred fault causes and historical fault phenomena are obtained; The inferred fault causes and the historical fault phenomena are substituted into the working condition profile to conduct a simulation exercise, and the exercise results are obtained. The similarity between the exercise results and the current running data is compared to obtain the target similarity value, and it is determined whether the target similarity value exceeds the preset standard value. If it is determined that the similarity value exceeds the standard value, then the reasoning failure cause and the historical similar cases are used as target result information and passed to the next step; If it is determined that the similarity value does not exceed the standard value, then the differential data between the exercise result and the current running data is identified, and the previous step is returned to perform logical reasoning again with the differential data as reference information.

7. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 6, characterized in that, The steps of retrieving similar data from neighboring pumping stations within the same time period for horizontal comparison, obtaining comparison results, generating common or individual anomaly results based on the comparison results, and generating a final processing plan are as follows: Based on the reasoning failure cause and the historical similar cases, basic information is extracted from the reasoning failure cause and the historical similar cases to obtain the current basic failure information and failure time period; Based on the current basic fault information and the fault time period, retrieve similar data from neighboring pumping stations during the fault time period; The basic fault information is compared horizontally with the similar data to obtain the comparison results, and the fault reasoning results are determined to be either common abnormal results or individual abnormal results based on the comparison results. If the fault reasoning result is a common abnormal result, then the final processing solution for the entire region is generated based on the fault reasoning result; If the fault reasoning result is an individual anomaly, then a final processing plan for the region is generated based on the fault reasoning result.

8. The intelligent inspection method for secondary water supply pumping stations based on expert systems according to claim 7, characterized in that, The steps for constructing a health index for each device based on the described operating condition profile and the current operating data, and generating a predictive maintenance plan based on the health index, are as follows: Based on the described operating condition profile, the health status of the equipment is assessed to obtain the first health parameter of the equipment at the current time point; Based on the current operating data, the health status of the device is assessed to obtain the second health parameter of the device at the current time point; By combining the first health parameter and the second health parameter, a health index is constructed for each device, and the target component of the health index is identified. Based on the health index and the target component, a predictive maintenance plan is generated for each device.

Images