Sewage pump station fault diagnosis system
By arranging a multi-sensor array in the sewage pumping station and performing signal processing and feature extraction, combined with dynamic comparison and intelligent diagnosis, the problems of insufficient monitoring and insufficient intelligence in the sewage pumping station fault diagnosis system in the existing technology are solved, and efficient and accurate fault diagnosis and health assessment are achieved.
Patent Information
- Application Number
- CN202511042024.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-28
- Publication Date
- 2025-10-17
AI Technical Summary
The existing sewage pump station fault diagnosis system has problems such as limited monitoring dimensions, insufficient data processing, incomplete feature analysis, inaccurate fault comparison, low intelligence level, and insufficient health assessment, resulting in low diagnostic efficiency and reliability, and difficulty in dealing with new fault types.
A multi-sensor array is arranged at key locations of the sewage pumping station to perform signal enhancement and adaptive filtering noise reduction, extract multi-dimensional feature vectors, and compare them with the dynamic threshold matrix and time series state space model. Fault diagnosis is performed through rule reasoning and Bayesian network, a health assessment model is constructed, and the fault feature library is dynamically updated.
It realizes all-round monitoring of the equipment's operating status, improves the accuracy of fault feature identification and diagnosis, supports rapid response and prediction of equipment health status, and enhances the system's adaptability and intelligence level.
Smart Images
Figure CN120804735A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of sewage treatment equipment monitoring, in particular to a sewage pump station fault diagnosis system. BACKGROUND
[0002] The sewage pump station is a key facility in the municipal drainage system and sewage treatment process. The stable operation of the pump body, motor, control cabinet and other equipment inside the sewage pump station directly affects the drainage efficiency and sewage treatment effect. Equipment failure may cause poor drainage, sewage overflow and other problems. Therefore, timely and accurate fault diagnosis of the sewage pump station equipment is an important technical link to ensure its continuous and reliable operation, involving equipment state monitoring, data intelligent analysis and fault warning and other technical fields.
[0003] In the prior art, the sewage pump station fault diagnosis has the problem of limited monitoring dimension. The sensor arrangement is often only for a few core parameters, which is difficult to fully capture various characteristic information of equipment operation, and is easy to miss fault signals. The data processing link lacks effective enhancement and noise reduction means. There are many interference factors in the original data, which affects the accuracy of subsequent feature extraction. Feature analysis is mostly concentrated in a single time domain or frequency domain, which is difficult to form a feature vector that can fully reflect the equipment state, leading to difficulty in fault feature recognition. Fault comparison relies on a single threshold or static model, which is difficult to adapt to changes in working conditions and is easy to misjudge or miss. The diagnosis process is not intelligent enough to accurately locate the fault position, type and quantitative confidence. The alarm mechanism lacks hierarchical design, and the fault tracing information is incomplete, which is not conducive to rapid response. At the same time, there is a lack of systematic evaluation of the health status of the equipment, which cannot predict the remaining life and develop a reasonable maintenance priority. Moreover, the fault feature library and diagnosis rules are updated slowly, which is difficult to cope with new fault types, and the overall diagnosis efficiency and reliability are low. SUMMARY
[0004] The present application provides the following technical scheme: a sewage pump station fault diagnosis system, comprising: a sensor deployment unit: A multi-element sensor array is arranged at a key position of the sewage pump station equipment, the multi-element sensor array comprising: a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor housing temperature monitoring sensor, a catchment tank liquid level monitoring sensor and a water outlet pipeline flow monitoring sensor; a data preprocessing unit: Receive the original operating parameters transmitted by the sensor array, and sequentially perform signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation to form a standardized parameter sequence; a feature extraction unit: The pre-processed parameter sequence is subjected to multi-dimensional feature analysis to extract a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features and time-frequency joint features; The double comparison unit: The multi-dimensional feature vector is compared with a dynamic threshold matrix and a time sequence state space model constructed based on historical data to generate a difference feature map; The intelligent diagnosis unit: Based on the difference feature map, the rule reasoning engine matches the fault feature library, and the Bayesian network is used for fault probability calculation to output a diagnosis result containing fault location, type and confidence; The hierarchical alarm unit: According to the fault level of the diagnosis result, the corresponding level of sound and light alarm signal is triggered, and an alarm data packet containing the fault tracing path is generated and pushed to the preset terminal through the secure transmission protocol; The health assessment unit: Based on the diagnosis result, a device health index model is constructed to generate a device remaining life prediction curve and a maintenance priority ranking; The knowledge updating unit: The actual fault handling cases are extracted to form new fault feature templates, and the fault feature library and diagnosis rules are dynamically updated.
[0005] A multi-element sensor array is arranged at a key position of the sewage pump station equipment, which includes a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor shell temperature monitoring sensor, a water collection tank liquid level monitoring sensor and an outlet pipeline flow monitoring sensor. The multi-element sensor array is used to collect the operating parameters of the sewage pump station equipment in real time; The original operating parameters transmitted by the sensor deployment unit are received, and the original operating parameters are sequentially subjected to signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation to form a standardized parameter sequence; The standardized parameter sequence output by the data preprocessing unit is received, and multi-dimensional feature analysis is performed to extract a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features and time-frequency joint features; The multi-dimensional feature vector output by the feature extraction unit is received, and the multi-dimensional feature vector is compared with a dynamic threshold matrix and a time sequence state space model constructed based on historical data to generate a difference feature map; The difference feature map output by the double comparison unit is received, and based on the difference feature map, the rule reasoning engine matches the fault feature library, and the Bayesian network is used for fault probability calculation to output a diagnosis result containing fault location, type and confidence; The diagnostic result output by the intelligent diagnosis unit is received, and according to the fault level of the diagnostic result, an alarm signal of a corresponding level is triggered, and an alarm data packet containing a fault tracing path is generated and pushed to a preset terminal through a secure transmission protocol; The diagnostic result output by the intelligent diagnosis unit is received, and a device health index model is constructed based on the diagnostic result, and a device remaining life prediction curve and a maintenance priority ranking are generated. Actual fault handling cases are collected, knowledge extraction is performed thereon, new fault feature templates are formed, and the fault feature library and the diagnostic rules are dynamically updated using the new fault feature templates.
[0006] Preferably, the water collecting pool liquid level monitoring sensor in the sensor deployment unit adopts a distributed arrangement, and a ring-shaped sensing array is arranged at different height planes of the water collecting pool.
[0007] In the sensor deployment unit, for the water collecting pool liquid level monitoring sensor, first, different height planes of the water collecting pool are determined, and then a ring-shaped sensing array is arranged on each height plane in a distributed arrangement, so as to ensure that the sensing array can cover the monitoring area of the corresponding height plane of the water collecting pool.
[0008] Preferably, the adaptive filtering and noise reduction of the data preprocessing unit adopts a hybrid filtering algorithm combining wavelet packet transform and empirical mode decomposition.
[0009] When the data preprocessing unit performs adaptive filtering and noise reduction, first, wavelet packet transform is used to perform multi-scale decomposition on the collected original data to obtain signal components of different frequency bands; then, the decomposed signal components are further processed through empirical mode decomposition to obtain a series of intrinsic mode functions; finally, the processing results of the two kinds of transformations are combined to effectively filter noise.
[0010] Preferably, the time-frequency joint feature extraction of the feature extraction unit adopts a fusion analysis method combining short-time Fourier transform and wavelet transform.
[0011] When the feature extraction unit performs time-frequency joint feature extraction, first, short-time Fourier transform is used to perform time-frequency analysis on the signal to obtain frequency features of the signal in different time periods; at the same time, wavelet transform is used to perform multi-resolution analysis on the signal to extract local features of the signal; then, the analysis results obtained by the two kinds of transformations are fused to form a complete time-frequency joint feature set.
[0012] Preferably, the time series state space model of the dual comparison unit adopts a dynamic modeling method combining an autoregressive integrated moving average model and a Kalman filter.
[0013] In the process of constructing the time sequence state space model by the double comparison unit, firstly, an autoregressive integrated moving average model is established to describe the trend and fluctuation characteristics of the time sequence data; then, a Kalman filter is combined with the model, and the Kalman filter is used to dynamically estimate and correct the state of the model, so that the dynamic modeling of the time sequence state space model is realized.
[0014] Preferably, the Bayesian network construction of the intelligent diagnosis unit includes three progressive steps of fault prior probability matrix generation, conditional probability table updating and posterior probability reasoning.
[0015] In the process of constructing the Bayesian network by the intelligent diagnosis unit, firstly, a fault prior probability matrix is generated to provide an initial probability basis for the network; then, the conditional probability table is updated according to new monitoring data and diagnosis results; finally, the posterior probability reasoning is performed based on the updated probability information, and the fault diagnosis process is completed.
[0016] Preferably, the security transmission protocol of the hierarchical alarm unit adopts a hybrid encryption mechanism based on elliptic curve encryption.
[0017] In the process of adopting the security transmission protocol by the hierarchical alarm unit, a hybrid encryption mechanism is constructed based on elliptic curve encryption, the key is generated and exchanged by using the elliptic curve encryption algorithm, and then the alarm information is encrypted by combining the symmetric encryption algorithm, so that the security transmission of the alarm information is realized.
[0018] Preferably, the residual life prediction curve generation of the health assessment unit adopts a prediction algorithm combining particle filtering and neural network.
[0019] In the process of generating the residual life prediction curve by the health assessment unit, firstly, the state parameters of the equipment are estimated and updated by using particle filtering, so that the dynamic change information of the equipment state is obtained; then, the information is input into the neural network, and the nonlinear mapping and prediction are performed by the neural network; finally, the results of particle filtering and neural network are fused to generate the residual life prediction curve.
[0020] Preferably, the knowledge extraction process of the knowledge updating unit includes three core links of case feature reduction, rule generalization and knowledge base consistency maintenance.
[0021] In the process of knowledge extraction by the knowledge updating unit, firstly, the case is subjected to feature reduction, and the redundant features are removed and the key features are retained; then, the extracted rules are subjected to generalization processing, and the application range of the rules is expanded; finally, the knowledge base consistency maintenance is performed, so that the newly added knowledge does not conflict with the original knowledge, and the consistency of the knowledge base is maintained.
[0022] Preferably, the sensor deployment unit further includes a torque monitoring sensor arranged at the pump body coupling position and a displacement monitoring sensor arranged at the valve actuator.
[0023] The sensor deployment unit installs a torque monitoring sensor at the pump body coupling position and a displacement monitoring sensor at the valve actuator position based on setting the catchment tank liquid level monitoring sensor, and connects these sensors with the monitoring system to realize real-time monitoring of the corresponding parameters.
[0024] Compared with the prior art, the sewage pump station fault diagnosis system provided by the present application has the following beneficial effects: The system arranges a multi-element sensor array at key positions of the sewage pump station equipment, comprehensively captures multiple types of key operating parameters such as pump motor current, control cabinet voltage, and pump body bearing seat vibration, realizes all-around monitoring of the equipment operating state, and reduces missing of fault information; signal enhancement, adaptive filtering and noise reduction, and feature scale transformation in the data preprocessing link effectively improve the quality of the original data and reduce the influence of interference factors on subsequent analysis; multi-dimensional feature analysis extracts time domain, frequency domain and time-frequency joint features, and the multi-dimensional feature vector formed can more comprehensively reflect the equipment state characteristics and enhance the recognizability of fault features; the intelligent diagnosis matches the fault feature library through the rule reasoning engine and calculates the fault probability by combining the Bayesian network, so that the diagnosis result can clearly indicate the fault position, type and confidence, and the diagnosis accuracy is improved; the graded alarm triggers corresponding sound and light alarms according to the fault grade, and pushes a data packet containing a fault tracing path, which helps to quickly respond and troubleshoot; the health assessment constructs an equipment health index model, generates a residual life prediction curve and a maintenance priority ranking, facilitates advance planning of maintenance, and reduces sudden failures; the knowledge updating unit dynamically updates the fault feature library and the diagnosis rules by extracting actual cases, so that the system can continuously adapt to new fault types and maintain long-term diagnosis capability. Overall, the system improves the comprehensiveness, accuracy and timeliness of the sewage pump station fault diagnosis, and helps to improve the stable operation and maintenance efficiency of the equipment. BRIEF DESCRIPTION OF DRAWINGS
[0025] Figure 1 is a schematic block diagram of the system of the present application. DETAILED DESCRIPTION
[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0027] Please refer to Figure 1 The present application provides a technical solution, a sewage pump station fault diagnosis system, which comprises: sensor deployment unit: a plurality of sensor arrays are arranged at key positions of the sewage pump station equipment, and the plurality of sensor arrays comprise: a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor shell temperature monitoring sensor, a sump liquid level monitoring sensor, and a water outlet pipeline flow monitoring sensor; data preprocessing unit: The original operation parameters transmitted by the sensor array are received, and signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation are sequentially performed to form a standardized parameter sequence; feature extraction unit: The preprocessed parameter sequence is subjected to multi-dimensional feature analysis, and a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features, and time-frequency joint features is extracted; double comparison unit: The multi-dimensional feature vector is compared with a dynamic threshold matrix and a time sequence state space model based on historical data to generate a difference feature map; intelligent diagnosis unit: Based on the difference feature map, the rule reasoning engine matches the fault feature library, and the fault probability is calculated by combining the Bayesian network to output a diagnosis result containing the fault location, type and confidence; graded alarm unit: According to the fault level of the diagnosis result, the corresponding level of sound and light alarm signal is triggered, and an alarm data packet containing the fault tracing path is generated and pushed to the preset terminal through the security transmission protocol; health assessment unit: Based on the diagnosis result, a device health index model is constructed to generate a device remaining life prediction curve and a maintenance priority ranking; knowledge updating unit: The actual fault handling cases are knowledge extracted to form new fault feature templates, and the fault feature library and diagnosis rules are dynamically updated.
[0028] A plurality of sensor arrays are arranged at key positions of the sewage pump station equipment, and the plurality of sensor arrays comprise: a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor shell temperature monitoring sensor, a sump liquid level monitoring sensor, and a water outlet pipeline flow monitoring sensor, and the operation parameters of the sewage pump station equipment are collected in real time through the array; The original operation parameters transmitted by the sensor deployment unit are received, and the original operation parameters are sequentially subjected to signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation to form a standardized parameter sequence; The standardized parameter sequence output by the data preprocessing unit is received, multi-dimensional feature analysis is performed thereon, and a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features and time-frequency joint features is extracted; The multi-dimensional feature vector output by the feature extraction unit is received, and the multi-dimensional feature vector is compared with a dynamic threshold matrix and a time sequence state space model constructed based on historical data at the same time, to generate a difference feature map; The difference feature map output by the double comparison unit is received, and based on the difference feature map, a rule-based reasoning engine is used to match a fault feature library, a Bayesian network is used for fault probability calculation, and a diagnosis result containing a fault location, a fault type and a confidence level is output; The diagnosis result output by the intelligent diagnosis unit is received, and according to the fault level of the diagnosis result, an audible and visual alarm signal of the corresponding level is triggered, and an alarm data packet containing a fault tracing path is generated and pushed to a preset terminal through a secure transmission protocol; The diagnosis result output by the intelligent diagnosis unit is received, and based on the diagnosis result, a device health index model is constructed, a device remaining life prediction curve and a maintenance priority ranking are generated; Actual fault handling cases are collected, knowledge extraction is performed thereon, new fault feature templates are formed, and the new fault feature templates are used to dynamically update the fault feature library and the diagnosis rules; A multi-element sensor array is arranged through the sensor deployment unit, which can comprehensively collect operation parameters of key positions of the sewage pump station equipment, providing rich and comprehensive raw data basis for subsequent fault diagnosis, and ensuring the comprehensiveness and pertinence of the diagnosis data; The data preprocessing unit sequentially performs signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation on the original operation parameters to form a standardized parameter sequence, effectively improving the quality of the original data, reducing the influence of noise and interference on subsequent processing, and providing reliable data support for feature extraction and fault diagnosis; The feature extraction unit performs multi-dimensional feature analysis on the preprocessed parameter sequence, extracts a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features and time-frequency joint features, comprehensively captures the device operation state features, and improves the accuracy of fault feature recognition; The double comparison unit compares the multi-dimensional feature vector with a dynamic threshold matrix and a time sequence state space model constructed based on historical data at the same time, to generate a difference feature map, through a double comparison mechanism, the sensitivity and recognition ability to the abnormal state of the device are enhanced, and the reliability of fault diagnosis is improved; The intelligent diagnosis unit matches a fault feature library through a rule-based reasoning engine based on the difference feature map, calculates the fault probability through a Bayesian network, and outputs a diagnosis result containing a fault location, a fault type and a confidence level, which combines the advantages of rule-based reasoning and probability calculation, improves the accuracy and reliability of fault diagnosis; The hierarchical alarm unit triggers an audible and visual alarm signal of a corresponding level according to the fault level of the diagnosis result, generates an alarm data packet containing a fault tracing path and pushes it to a preset terminal, realizes timely early warning and accurate tracing of the fault, and facilitates quick response and processing of the fault by relevant personnel; The health assessment unit constructs a device health index model based on the diagnosis result, generates a device remaining life prediction curve and a maintenance priority ranking, can predict the health condition and remaining life of the device in advance, provides a scientific basis for the maintenance of the device, and reduces the probability of fault occurrence; The knowledge updating unit extracts knowledge from actual fault handling cases to form new fault feature templates, dynamically updates the fault feature library and the diagnosis rules, so that the system can continuously accumulate experience and continuously optimize the diagnosis capability, and the adaptability and intelligent level of the system are improved.
[0029] The water collecting pool liquid level monitoring sensor in the sensor deployment unit adopts a distributed arrangement mode, and a ring-shaped sensing array is arranged at different height planes of the water collecting pool.
[0030] In the sensor deployment unit, for the water collecting pool liquid level monitoring sensor, first, different height planes of the water collecting pool are determined, and then a ring-shaped sensing array is arranged on each height plane in a distributed arrangement mode, so as to ensure that the sensing array can cover the monitoring area of the corresponding height plane of the water collecting pool; By arranging the ring-shaped sensing array in a distributed arrangement mode at different height planes of the water collecting pool, the liquid level information at different heights in the water collecting pool can be fully captured, the spatial coverage of the liquid level monitoring is improved, the monitoring blind area is reduced, and the liquid level monitoring data is more representative and comprehensive.
[0031] The adaptive filtering and noise reduction of the data preprocessing unit adopts a hybrid filtering algorithm combining wavelet packet transform and empirical mode decomposition.
[0032] When performing adaptive filtering and noise reduction, the data preprocessing unit first performs multi-scale decomposition on the collected original data by wavelet packet transform to obtain signal components of different frequency bands; then the decomposed signal components are further processed by empirical mode decomposition to obtain a series of intrinsic mode functions; finally, the processing results of the two transformations are combined to effectively filter noise; The hybrid filtering algorithm combining wavelet packet transform and empirical mode decomposition can fully utilize the advantages of the two algorithms, effectively process non-stationary signals, accurately separate signals and noise, improve the adaptive ability of filtering and noise reduction, retain effective feature information in the original data, and improve the quality of data preprocessing.
[0033] The time-frequency joint feature extraction of the feature extraction unit adopts a fusion analysis method of short-time Fourier transform and wavelet transform.
[0034] When the feature extraction unit performs time-frequency joint feature extraction, first, short-time Fourier transform is adopted to perform time-frequency analysis on the signal to obtain frequency features of the signal in different time periods; meanwhile, wavelet transform is adopted to perform multi-resolution analysis on the signal to extract local features of the signal; subsequently, analysis results obtained through the two kinds of transform are fused to form a complete time-frequency joint feature set; The analysis method of fusing short-time Fourier transform and wavelet transform can comprehensively utilize advantages of the two kinds of transform in time-frequency analysis, can clearly reflect time local features of the signal, can accurately capture frequency details of the signal, and thus can extract more comprehensive and accurate time-frequency joint features, thereby providing reliable basis for subsequent diagnosis and evaluation.
[0035] The time series state space model of the dual comparison unit adopts a dynamic modeling method combining an autoregressive integrated moving average model and a Kalman filter.
[0036] When the dual comparison unit constructs the time series state space model, first, an autoregressive integrated moving average model is established to describe trend and fluctuation characteristics of the time series data; then, the Kalman filter is combined with the model, and the Kalman filter is used to dynamically estimate and correct states of the model, thereby realizing dynamic modeling of the time series state space model. The dynamic modeling method combining the autoregressive integrated moving average model and the Kalman filter can effectively capture dynamic change rules of the time series data, through real-time correction of the Kalman filter, adaptability and tracking capability of the model to dynamic changes are improved, and accuracy and timeliness of the dual comparison are enhanced.
[0037] The Bayesian network construction of the intelligent diagnosis unit includes three progressive steps of fault prior probability matrix generation, conditional probability table updating and posterior probability reasoning.
[0038] When the intelligent diagnosis unit constructs the Bayesian network, first, a fault prior probability matrix is generated to provide an initial probability basis for the network; then, the conditional probability table is updated according to new monitoring data and diagnosis results; finally, posterior probability reasoning is performed based on updated probability information, thereby completing the fault diagnosis process. The three progressive steps of the Bayesian network construction enable the fault diagnosis process to have rigorous probability theory support. The fault prior probability matrix provides an initial basis for diagnosis, updating of the conditional probability table guarantees timeliness of the probability information, and the posterior probability reasoning can accurately obtain the diagnosis result, thereby improving accuracy and reliability of the intelligent diagnosis.
[0039] The security transmission protocol of the hierarchical alarm unit adopts a hybrid encryption mechanism based on elliptic curve encryption.
[0040] The hierarchical alarm unit adopts a security transmission protocol, and a hybrid encryption mechanism is constructed based on elliptic curve encryption, that is, an elliptic curve encryption algorithm is used to generate and exchange keys, and then a symmetric encryption algorithm is combined to encrypt the alarm information, so that the security transmission of the alarm information is realized. The hybrid encryption mechanism based on elliptic curve encryption not only has the high efficiency and security of elliptic curve encryption in key management, but also has the advantage of symmetric encryption algorithm in data encryption speed, so that the transmission efficiency is improved while ensuring the security of the transmission of the hierarchical alarm information, and the safe and reliable transmission of the alarm information is ensured.
[0041] The health assessment unit adopts a prediction algorithm of fusion of particle filtering and neural network to generate the residual life prediction curve.
[0042] When the health assessment unit generates the residual life prediction curve, the state parameters of the equipment are estimated and updated by using particle filtering, so that the dynamic change information of the equipment state is obtained; then the information is input into the neural network, and nonlinear mapping and prediction are performed by using the neural network; finally, the results of particle filtering and neural network are fused to generate the residual life prediction curve. The prediction algorithm of fusion of particle filtering and neural network can comprehensively utilize the advantages of particle filtering in processing nonlinear and non-Gaussian problems and the powerful nonlinear fitting capability of neural network, improve the accuracy of residual life prediction, and make the generated prediction curve more accurately reflect the change trend of the residual life of the equipment, thereby providing effective support for the maintenance decision of the equipment.
[0043] The knowledge extraction process of the knowledge updating unit includes three core links of case feature reduction, rule generalization and knowledge base consistency maintenance.
[0044] When the knowledge updating unit performs knowledge extraction, first, the case is subjected to feature reduction, redundant features are removed, and key features are retained; then, the extracted rules are subjected to generalization processing, and the application range of the rules is expanded; finally, the knowledge base consistency maintenance is performed, so that the newly added knowledge does not conflict with the original knowledge, and the consistency of the knowledge base is maintained. The three core links of the knowledge extraction process can effectively improve the quality and usability of the knowledge. The case feature reduction reduces data redundancy, the rule generalization enhances the applicability of the knowledge, and the knowledge base consistency maintenance ensures the reliability of the knowledge, thereby realizing the effective updating and optimization of the knowledge base.
[0045] The sensor deployment unit further includes a torque monitoring sensor arranged at the pump body coupling position and a displacement monitoring sensor arranged at the valve actuator.
[0046] The sensor deployment unit installs a torque monitoring sensor at the position of the pump body coupling and a displacement monitoring sensor at the position of the valve actuator, and connects these sensors with the monitoring system to realize real-time monitoring of the corresponding parameters. The torque monitoring sensor installed at the position of the pump body coupling and the displacement monitoring sensor installed at the position of the valve actuator can supplement the monitoring of the key operating parameters of the pump body and the valve, realize multi-dimensional and all-around state monitoring of the equipment, and provide more abundant data sources for fault diagnosis and health assessment of the equipment.
[0047] The sensor deployment unit arranges a multi-element sensor array at the key positions of the sewage pump station equipment, which includes a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor shell temperature monitoring sensor, a sump level monitoring sensor, and a water outlet pipeline flow monitoring sensor. The sump level monitoring sensor is arranged in a distributed manner, with a ring-shaped sensor array arranged at different height planes of the sump. The sensor deployment unit also includes a torque monitoring sensor installed at the position of the pump body coupling and a displacement monitoring sensor installed at the position of the valve actuator. These sensors are connected to the monitoring system to collect the operating parameters of the sewage pump station equipment in real time through the array. The data preprocessing unit receives the original operating parameters transmitted by the sensor deployment unit and sequentially performs signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation on the original operating parameters. The adaptive filtering and noise reduction uses a hybrid filtering algorithm combining wavelet packet transformation and empirical mode decomposition. First, the original data collected is decomposed into different frequency bands using wavelet packet transformation to obtain signal components of different frequency bands. Then, the signal components after decomposition are further processed using empirical mode decomposition to obtain a series of intrinsic mode functions. Finally, the processing results of the two transformations are combined to effectively filter noise and form a standardized parameter sequence. The feature extraction unit receives the standardized parameter sequence output by the data preprocessing unit and performs multi-dimensional feature analysis to extract a multi-dimensional feature vector containing time domain waveform features, frequency energy distribution features, and time-frequency joint features. The time-frequency joint feature extraction uses a fusion analysis method combining short-time Fourier transform and wavelet transform. First, the signal is analyzed in time and frequency using short-time Fourier transform to obtain the frequency characteristics of the signal at different time intervals. Then, the signal is analyzed in multiple resolutions using wavelet transform to extract local features of the signal. Finally, the analysis results of the two transformations are fused to form a complete time-frequency joint feature set. The dual comparison unit receives the multi-dimensional feature vector output by the feature extraction unit, and simultaneously compares the multi-dimensional feature vector with a dynamic threshold matrix and a time series state space model constructed based on historical data to generate a difference feature map, wherein the time series state space model adopts a dynamic modeling method combining an autoregressive integrated moving average model and a Kalman filter, an autoregressive integrated moving average model is first established to describe the trend and fluctuation characteristics of time series data, and then a Kalman filter is combined with the model to dynamically estimate and correct the state of the model, thereby realizing dynamic modeling of the time series state space model. The intelligent diagnosis unit receives the difference feature map output by the dual comparison unit, matches a fault feature library through a rule reasoning engine based on the difference feature map, calculates fault probabilities by combining a Bayesian network, and outputs a diagnosis result containing a fault location, type, and confidence level, wherein the construction of the Bayesian network includes three progressive steps of fault prior probability matrix generation, conditional probability table updating, and posterior probability reasoning, a fault prior probability matrix is first generated to provide an initial probability basis for the network, then the conditional probability table is updated according to new monitoring data and diagnosis results, and finally, posterior probability reasoning is performed based on the updated probability information to complete the fault diagnosis process. The hierarchical alarm unit receives the diagnosis result output by the intelligent diagnosis unit, triggers an audible and visual alarm signal of a corresponding level according to the fault level of the diagnosis result, generates an alarm data packet containing a fault tracing path, and pushes the alarm data packet to a preset terminal through a secure transmission protocol, wherein the secure transmission protocol adopts a hybrid encryption mechanism based on elliptic curve encryption, first generates and exchanges keys using an elliptic curve encryption algorithm, and then combines a symmetric encryption algorithm to encrypt the alarm information, thereby realizing secure transmission of the alarm information. The health assessment unit receives the diagnosis result output by the intelligent diagnosis unit, constructs a device health index model based on the diagnosis result, generates a device remaining life prediction curve, and sorts maintenance priorities, wherein the generation of the remaining life prediction curve adopts a prediction algorithm combining particle filtering and a neural network, first estimates and updates the state parameters of the device using particle filtering to obtain dynamic change information of the device state, then inputs this information into a neural network for nonlinear mapping and prediction, and finally fuses the results of particle filtering and the neural network to generate the remaining life prediction curve. The knowledge updating unit collects actual fault handling cases, extracts knowledge therefrom, forms new fault feature templates, and dynamically updates the fault feature library and the diagnosis rules by using the new fault feature templates. The knowledge extraction process includes three core links, i.e., case feature reduction, rule generalization and knowledge base consistency maintenance. Firstly, the case is subjected to feature reduction to remove redundant features and retain key features. Then, the extracted rules are subjected to generalization processing to expand the application scope of the rules. Finally, the knowledge base consistency maintenance is performed to ensure that the newly added knowledge does not conflict with the original knowledge and to maintain the consistency of the knowledge base.
[0048] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.
[0049] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.
Claims
1. A sewage pump station fault diagnosis system, characterized in that: include: Sensor deployment unit: Arrange a multi-sensor array at key locations of the sewage pump station equipment, including: a pump motor current monitoring sensor, a control cabinet voltage monitoring sensor, a pump body bearing seat vibration monitoring sensor, a pump motor housing temperature monitoring sensor, a water collection tank level monitoring sensor, and a water outlet pipe flow monitoring sensor; Data preprocessing unit: Receive the original operating parameters transmitted by the sensor array, perform signal enhancement processing, adaptive filtering and noise reduction, and feature scale transformation in sequence to form a standardized parameter sequence; Feature extraction unit: Perform multi-dimensional feature analysis on the pre-processed parameter sequence to extract a multi-dimensional feature vector containing time domain waveform features, frequency domain energy distribution features, and time-frequency joint features; Double comparison unit: The multi-dimensional feature vector is compared with the dynamic threshold matrix and the time series state space model constructed based on historical data to generate a difference feature map; Intelligent diagnostic unit: Based on the difference feature map, the fault feature library is matched through the rule inference engine, and the fault probability is calculated by combining with the Bayesian network to output the diagnosis result including the fault location, type and confidence level; Hierarchical alarm unit: According to the fault level of the diagnosis result, the corresponding level of sound and light alarm signal is triggered, and an alarm data packet containing the fault tracing path is generated and pushed to the preset terminal through a secure transmission protocol; Health Assessment Unit: Build an equipment health index model based on the diagnosis results to generate equipment remaining life prediction curves and maintenance priority rankings; Knowledge update unit: Extract knowledge from actual fault handling cases to form new fault feature templates, and dynamically update the fault feature library and diagnosis rules.
2. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The water collection tank liquid level monitoring sensors in the sensor deployment unit are arranged in a distributed manner, and annular sensor arrays are set at different height planes of the water collection tank.
3. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The adaptive filtering and denoising of the data preprocessing unit adopts a hybrid filtering algorithm combining wavelet packet transform and empirical mode decomposition.
4. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The time-frequency joint feature extraction of the feature extraction unit adopts an analysis method that combines short-time Fourier transform with wavelet transform.
5. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The time series state space model of the double comparison unit adopts a dynamic modeling method combining an autoregressive integral moving average model with a Kalman filter.
6. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The Bayesian network construction of the intelligent diagnosis unit includes three progressive steps: generating a priori fault probability matrix, updating the conditional probability table, and reasoning about the posterior probability.
7. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The secure transmission protocol of the hierarchical alarm unit adopts a hybrid encryption mechanism based on elliptic curve encryption.
8. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The remaining life prediction curve of the health assessment unit is generated by using a prediction algorithm that integrates particle filtering and neural network.
9. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The knowledge extraction process of the knowledge updating unit includes three core links: case feature simplification, rule generalization and knowledge base consistency maintenance.
10. A sewage pump station fault diagnosis system according to claim 1, characterized in that: The sensor deployment unit also includes a torque monitoring sensor arranged at the pump body coupling position and a displacement monitoring sensor arranged at the valve actuator.