Real-time monitoring management system for a machine-pump device

Abstract

The application relates to the technical field of management and supervision data processing, and discloses a real-time monitoring management system for machine pump equipment, which comprises the following steps: a data acquisition module acquires state data sets of each node in a machine pump group and stores the state data sets in a ring buffer area; an analysis and processing module constructs a directed acyclic topology model according to a physical connection relationship of a pipe network, determines time bias weights between nodes by using pipeline parameters, further determines a traceability window, calculates a correlation coefficient to lock a root fault node, and implements alarm suppression on derivative alarms; and a management output module calculates maintenance scheduling weights according to a downstream out-degree of the root node and a state degradation degree, and outputs a work order. The application suppresses derivative alarms by using a topology and time bias causality traceability mechanism, eliminates alarm storms, and realizes precise scheduling of operation and maintenance resources.

Description

Technical Field

[0001] This invention relates to a real-time monitoring and management system for pump equipment, belonging to the field of management and supervision data processing technology. Background Technology

[0002] Currently, the mainstream solutions for online monitoring and management of industrial equipment collect real-time operating data through vibration sensors and temperature sensors deployed at pump measuring points. The controller compares the real-time operating data with preset thresholds to determine the operating status of the equipment. In large-scale coupled pipeline network environments, pumps and equipment are closely physically linked through fluid media. When an upstream pump experiences abnormal fluctuations, the disturbance energy is rapidly transmitted along the pipeline, causing multiple pumps in the downstream area to exceed the associated parameters. For example, Chinese invention patent application CN117666528A discloses a fault diagnosis method, device, equipment, and medium based on a directed acyclic graph. It uses reinforcement learning and neural networks to learn the equipment logic strategy from the observation data and construct a directed acyclic graph.

[0003] Existing management methods typically treat each pump as an isolated data node. When faced with such systemic disturbances, the management platform generates a large number of concurrent redundant alarms. If a linear improvement method of simply raising the alarm threshold or enhancing signal filtering is adopted, it will lead to a delay in the system's detection and response to real, weak faults. Simply increasing the density of physical measuring points not only generates massive amounts of redundant data, but also still fails to identify the evolution path of the fault from a logical perspective. The existing monitoring system lacks a logical mapping of the physical space of the pipeline network, resulting in maintenance resources being blindly dispersed in the alarm storm, making it difficult to perform precise scheduling for root cause faults.

[0004] Therefore, how to construct a directed acyclic model that can characterize the process path and combine it with time bias weights to achieve causal tracing of abnormal root nodes and improve the scheduling efficiency of maintenance resources has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A real-time monitoring and management system for pump equipment, the system comprising:

[0006] The data acquisition module is used to acquire the operating status dataset of each node in the pump equipment group, and store the operating status dataset into the circular data buffer corresponding to each node according to the sampling time sequence. The operating status dataset includes the vibration amplitude, frequency distribution and process pressure parameters collected at each monitoring point.

[0007] The analysis and processing module is used to construct a directed acyclic topology model reflecting the flow direction of the process medium based on the physical connection topology of the industrial pipeline network. Specifically, the module extracts physical pipeline length and velocity parameters to determine the theoretical propagation delay between adjacent nodes in the directed acyclic topology model, and maps this theoretical propagation delay to various directed time bias weights in the model, forming a bias weight matrix to characterize the lag in the propagation of disturbances in physical space. Furthermore, when the characteristic value fluctuation of the first node in the directed acyclic topology model exceeds a preset threshold, the module determines a source tracing time window based on the bias weight matrix. Within this window, it performs correlation matching of the fluctuation sequences between the first node and upstream candidate nodes, thereby identifying the root cause fault node causing the physical chain reaction at the initial stage of alarm generation. It also implements alarm suppression logic for derivative alarm signals generated by associated nodes other than the root cause fault node.

[0008] The management output module is used to calculate the maintenance scheduling weight based on the out-degree of the downstream nodes of the root cause fault node in the directed acyclic topology model determined by the analysis and processing module, combined with the state degradation degree of the root cause fault node, and output a quantitative maintenance work order generated by prioritizing the maintenance scheduling weight.

[0009] Preferably, when performing correlation matching of fluctuation sequences, the analysis and processing module obtains the characteristic fluctuation sequence of the first node and the running data fluctuation sequence of the upstream node in the directed acyclic topology model within the tracing time window, calculates the Pearson correlation coefficient between the characteristic fluctuation sequence and the running data fluctuation sequence, and determines the upstream node as the disturbance source corresponding to the root cause fault node when the Pearson correlation coefficient is greater than the preset correlation threshold.

[0010] Preferably, nodes in the directed acyclic topology model represent independent pumps and machinery, and directed edges between nodes represent the flow direction of industrial fluid process media. The analysis and processing module performs sliding window dimensionality reduction processing on the time-series data of each node in the pump and machinery group, converting the original high-dimensional operating status data of each node into a multi-dimensional state feature vector for characterizing the health trend of the equipment.

[0011] Preferably, the analysis and processing module is also used to prioritize the generated quantitative maintenance work orders according to the numerical value of the maintenance scheduling weight, and push the ranking result to the management output module in real time. The management output module generates resource scheduling instructions including spare parts requirements, maintenance types and suggested maintenance durations based on the ranking results, so as to guide maintenance resources to concentrate on the root cause fault nodes.

[0012] Preferably, the sampling period of the data acquisition module is set to 100ms; the analysis and processing module is also used to determine the first node that generates the characteristic value fluctuation as the root cause fault node causing the alarm storm when the correlation coefficient is less than the preset correlation threshold, or when there is no upstream adjacent node that meets the preset conditions in the directed acyclic topology model, and to generate an independent source tracing and diagnosis report for the first node.

[0013] Preferably, the management output module is also used to maintain the equipment management ledger module, which records the physical connection topology, historical maintenance cycle, component life distribution data, and real-time operating status level of the pump equipment. The management output module provides macro-level maintenance planning suggestions for the pump equipment group to the management decision-making level by associating the real-time operating status level with the maintenance scheduling weight.

[0014] Preferably, after identifying the root cause of the fault, the analysis and processing module retrieves the corresponding spare parts inventory data from the equipment management ledger module based on the equipment model of the root cause of the fault. When it is determined that the current inventory value of the spare parts is lower than the preset safety stock threshold, the module automatically sends a spare parts procurement warning signal to the management output module, or the management output module directly triggers the automatic procurement application process instruction for the corresponding spare parts.

[0015] Preferably, the analysis and processing module is also used to predict the remaining lifespan of the equipment based on the long-term trend changes of the operating status dataset, and calculate the deviation between the remaining lifespan of the equipment and the preset maintenance cycle. When the deviation exceeds the preset alarm threshold, the state degradation benchmark of the corresponding node in the directed acyclic topology model is automatically increased, thereby inducing the management output module to generate maintenance decision suggestions for the corresponding node in advance through weight offset.

[0016] Preferably, the management output module also includes an energy-saving management unit, which is used to acquire real-time energy consumption monitoring data of each node in the pump group, and calculate the energy consumption distribution entropy on the process path according to the load distribution status of each node in the directed acyclic topology model, thereby outputting a load balancing scheduling instruction aimed at optimizing the overall operating load distribution of the pump group and reducing the total energy consumption level.

[0017] Compared with the prior art, the beneficial effects of the present invention are:

[0018] 1. In the real-time monitoring and management of pump equipment, a directed acyclic topology map based on the flow direction of the process medium is constructed, and the physical transmission delay between adjacent devices is converted into the time bias weight of the data processing layer. This changes the traditional monitoring and management approach of treating equipment as isolated individuals. When a single disturbance source triggers a chain reaction, the system controller performs cross-node causal tracing according to a preset time window. This enables logical suppression of derivative alarm signals at the initial stage of alarm generation, avoiding the management platform from outputting a large number of repetitive or misleading maintenance work orders and ensuring the uniqueness and accuracy of maintenance resource scheduling instructions.

[0019] 2. This solution establishes a dynamic weight allocation mechanism based on state degradation degree and topological influence. By calculating the failure degree of the root node equipment and its out-degree of the downstream nodes in the process path, it outputs quantified maintenance scheduling weights, enabling management decisions to prioritize equipment failures based on their impact on the overall production system. This approach prioritizes on-site maintenance resources to key impact sources, improves the efficiency of operation and maintenance organization for large-scale pump groups, and realizes the transformation from passively responding to threshold alarms to actively controlling the root causes of system failures.

[0020] 3. This solution utilizes dimensionality reduction mapping of multidimensional state feature vectors and ring buffer sliding storage technology to reduce the computational load on the engineer diagnostic platform under the premise of ensuring data timeliness. When multiple pumps in the pipeline network experience parameter fluctuations simultaneously, the system quickly locks the physical path of disturbance propagation by calculating the Pearson correlation coefficient between characteristic fluctuation sequences. This transforms complex industrial field conditions into intuitive logical dependencies, reduces the system's reliance on the physical modeling experience of diagnostic personnel, and enhances the objectivity of the monitoring and management system in handling complex coupled conditions. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the data processing and workflow of the real-time monitoring and management system for pump equipment of the present invention.

[0022] Figure 2 This is a diagram showing the physical deployment and network architecture of the real-time monitoring and management system for pumps and motors of the present invention.

[0023] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0025] A real-time monitoring and management system for pump equipment, the system comprising:

[0026] The data acquisition module is used to acquire the operating status dataset of each node in the pump equipment group, and store the operating status dataset into the circular data buffer corresponding to each node according to the sampling time sequence. The operating status dataset includes the vibration amplitude, frequency distribution and process pressure parameters collected at each monitoring point.

[0027] The analysis and processing module is used to construct a directed acyclic topology model reflecting the flow direction of the process medium based on the physical connection topology of the industrial pipeline network. Specifically, the module extracts physical pipeline length and velocity parameters to determine the theoretical propagation delay between adjacent nodes in the directed acyclic topology model, and maps this theoretical propagation delay to various directed time bias weights in the model, forming a bias weight matrix to characterize the lag in the propagation of disturbances in physical space. Furthermore, when the characteristic value fluctuation of the first node in the directed acyclic topology model exceeds a preset threshold, the module determines a source tracing time window based on the bias weight matrix. Within this window, it performs correlation matching of the fluctuation sequences between the first node and upstream candidate nodes, thereby identifying the root cause fault node causing the physical chain reaction at the initial stage of alarm generation. It also implements alarm suppression logic for derivative alarm signals generated by associated nodes other than the root cause fault node.

[0028] The management output module is used to calculate the maintenance scheduling weight based on the out-degree of the downstream nodes of the root cause fault node in the directed acyclic topology model determined by the analysis and processing module, combined with the state degradation degree of the root cause fault node, and output a quantitative maintenance work order generated by prioritizing the maintenance scheduling weight.

[0029] Preferably, when performing correlation matching of fluctuation sequences, the analysis and processing module obtains the characteristic fluctuation sequence of the first node and the running data fluctuation sequence of the upstream node in the directed acyclic topology model within the tracing time window, calculates the Pearson correlation coefficient between the characteristic fluctuation sequence and the running data fluctuation sequence, and determines the upstream node as the disturbance source corresponding to the root cause fault node when the Pearson correlation coefficient is greater than the preset correlation threshold.

[0030] Preferably, nodes in the directed acyclic topology model represent independent pumps and machinery, and directed edges between nodes represent the flow direction of industrial fluid process media. The analysis and processing module performs sliding window dimensionality reduction processing on the time-series data of each node in the pump and machinery group, converting the original high-dimensional operating status data of each node into a multi-dimensional state feature vector for characterizing the health trend of the equipment.

[0031] Preferably, the management output module uses the following quantification rules when calculating maintenance scheduling weights: Where W is the maintenance scheduling weight; D is the state degradation degree of the root cause fault node determined by the analysis and processing module; For the root cause fault node in the directed acyclic topology model, the first... The out-degree value of each downstream path node; α and β are the preset fault impact coefficient and topology weight coefficient, respectively.

[0032] Preferably, the analysis and processing module is also used to prioritize the generated quantitative maintenance work orders according to the numerical value of the maintenance scheduling weight, and push the ranking result to the management output module in real time. The management output module generates resource scheduling instructions including spare parts requirements, maintenance types and suggested maintenance durations based on the ranking results, so as to guide maintenance resources to concentrate on the root cause fault nodes.

[0033] Preferably, the sampling period of the data acquisition module is set to 100ms; the analysis and processing module is also used to determine the first node that generates the characteristic value fluctuation as the root cause fault node causing the alarm storm when the correlation coefficient is less than the preset correlation threshold, or when there is no upstream adjacent node that meets the preset conditions in the directed acyclic topology model, and to generate an independent source tracing and diagnosis report for the first node.

[0034] Preferably, the management output module is also used to maintain the equipment management ledger module, which records the physical connection topology, historical maintenance cycle, component life distribution data, and real-time operating status level of the pump equipment. The management output module provides macro-level maintenance planning suggestions for the pump equipment group to the management decision-making level by associating the real-time operating status level with the maintenance scheduling weight.

[0035] Preferably, after identifying the root cause of the fault, the analysis and processing module retrieves the corresponding spare parts inventory data from the equipment management ledger module based on the equipment model of the root cause of the fault. When it is determined that the current inventory value of the spare parts is lower than the preset safety stock threshold, the module automatically sends a spare parts procurement warning signal to the management output module, or the management output module directly triggers the automatic procurement application process instruction for the corresponding spare parts.

[0036] Preferably, the analysis and processing module is also used to predict the remaining lifespan of the equipment based on the long-term trend changes of the operating status dataset, and calculate the deviation between the remaining lifespan of the equipment and the preset maintenance cycle. When the deviation exceeds the preset alarm threshold, the state degradation benchmark of the corresponding node in the directed acyclic topology model is automatically increased, thereby inducing the management output module to generate maintenance decision suggestions for the corresponding node in advance through weight offset.

[0037] Preferably, the management output module also includes an energy-saving management unit, which is used to acquire real-time energy consumption monitoring data of each node in the pump group, and calculate the energy consumption distribution entropy on the process path according to the load distribution status of each node in the directed acyclic topology model, thereby outputting a load balancing scheduling instruction aimed at optimizing the overall operating load distribution of the pump group and reducing the total energy consumption level.

[0038] Example 1: In the real-time monitoring and management system for pumps and equipment of this invention, deployed in the online monitoring environment of petrochemical pumps and equipment, the system faces a process pipeline network consisting of 543 pumps and equipment, including 418 high-risk pumps conveying high-temperature media and 125 critical pumps. When the fully lined slurry pump P-2210 / 2, located upstream in the process pipeline network, experiences a mechanical component malfunction causing flow fluctuations, the physical disturbance energy is transmitted along the flow direction of the process medium, inducing synchronous vibrations and excessive process pressure parameters in multiple downstream related equipment. If a single-point threshold alarm method is used, the monitoring platform will instantly receive a large number of concurrent alarm signals, causing the system to generate repetitive and mutually conflicting alarms. Sudden on-site verification work orders disperse maintenance resources across alarm storms, making it difficult to implement precise scheduling for root cause faults. For the aforementioned operating conditions with topological coupling characteristics, the analysis and processing module constructs a directed acyclic topology model reflecting the flow direction of the process medium based on the physical connection relationship of the on-site pipeline network. In this model, nodes represent independent pumps and turbines, and directed edges between nodes represent the flow direction of the medium. The analysis and processing module determines the theoretical propagation delay between adjacent nodes in the model based on physical pipeline length and flow velocity parameters, and converts this delay into directed time bias weights. This forms a bias weight matrix characterizing the lag in the propagation of disturbances in physical space. Simultaneously, the data acquisition module... The system collects the operating status dataset of each node in the pump and motor equipment group during the sampling period, and stores the time-series data containing vibration amplitude, frequency distribution and process pressure parameters into the corresponding circular data buffer of each node. In the specific calculation, the flow velocity parameter is adaptively matched according to the physical properties of the monitored parameters. For the transmission of slowly changing state parameters driven by medium transportation, the macroscopic average medium flow velocity of the pipeline network is adopted. For transient mechanical disturbance characteristics such as vibration amplitude and process pressure, the system automatically switches the flow velocity parameter to the sound velocity parameter of the process medium at that temperature and pressure, i.e., the medium wave velocity, so as to establish a spatiotemporal mapping relationship that conforms to the physical propagation law of pressure waves in fluid.

[0039] When the system detects that the value fluctuation of the feature quantity of the first node in the directed acyclic topology model exceeds the preset safety threshold, it does not immediately trigger an alarm response from the first node, but instead freezes the current timestamp. The processor determines the source tracing time window based on the bias weight matrix. The time window bias Δt is determined based on the delay weight corresponding to the directed edge. The processor retrieves the running data fluctuation sequence of the upstream candidate node that has a direct directed edge connection with the first node within the tracing time window, and calculates the Pearson correlation coefficient between the feature quantity fluctuation sequence of the first node and the running data fluctuation sequence of the upstream candidate node. When the Pearson correlation coefficient is greater than 0.85, the anomaly of the first node is determined to be a passively conducted disturbance. The system transfers the causal tracing pointer to the upstream candidate node to continue iterative backtracking until the root fault node of the correlation coefficient mutation is located. Before performing the above correlation coefficient calculation, in order to solve the dimensional conflict that multidimensional state feature vectors cannot directly solve one-dimensional correlation, the analysis and processing module pre-calculates the L2 norm of the feature vector of each node at each sampling moment within the tracing time window, transforming the original multidimensional vector space containing vibration and pressure into a representation of the whole. The one-dimensional scalar time series of volume energy fluctuation amplitude is used, and then two one-dimensional energy fluctuation sequences aligned end to end are substituted into the Pearson formula to complete the linear correlation measurement. In a pump group management system containing 12 branch nodes, the analysis and processing module reduces false alarm interference through logical judgment thresholds. After the processor locates the root cause fault node, it extracts the set of all disturbed nodes in the directed acyclic topology model except for the root cause fault node. For each node in this set, the analysis and processing module queries its corresponding circular data buffer. If the correlation coefficient between the characteristic quantity fluctuation of the node and the root cause fault node in the source tracing time window exceeds the preset correlation threshold η, the processor invalidates the alarm trigger signal of the node, blocks the transmission of derivative alarm signals to the management terminal, and the management terminal only displays the alarm icon of the root cause fault node, guiding maintenance personnel to focus on the initial failure source and improving the pertinence of response decisions.

[0040] After identifying the root cause fault node, the analysis and processing module suppresses derivative alarm signals generated by related nodes other than the root cause fault node. The management output module then calculates the out-degree of the root cause fault node's downstream nodes in the directed acyclic topology model. The maintenance scheduling weight W is calculated based on the current state degradation degree D of the node. The specific quantification rules are as follows: Where W is the maintenance scheduling weight; D is the state degradation degree of the root cause fault node determined by the analysis and processing module; Let α be the out-degree value of the i-th downstream path node of the root cause fault node in the directed acyclic topology model; α and β are the preset fault impact coefficient and topology weight coefficient, respectively. The management output module sorts the generated quantitative maintenance work orders according to the value of the maintenance scheduling weight W, and pushes the sorting results to the management terminal in real time. The generated resource scheduling instructions include spare parts requirements, maintenance types and suggested maintenance durations, guiding maintenance resources to concentrate on the root cause fault node, thereby suppressing redundant alarm signals caused by process transmission at the alarm generation stage, making each management output instruction have clear directionality, and improving the management efficiency in large-scale equipment online monitoring scenarios. This solution constructs a logical topology map of the equipment at the data processing layer and introduces spatiotemporal bias tracing calculation, breaking the physical constraints based on single-point fixed thresholds. The system strips away the transmission disturbances at the data logic level, realizing the transformation from representing alarms to root cause localization. This approach concentrates maintenance efforts on the source of influence, improves the maintenance organization efficiency of large-scale pump groups, and transforms equipment management from passively responding to threshold alarms to actively controlling the root cause of the system.

[0041] Example 2: In the process of verifying the management and supervision effectiveness of the real-time monitoring and management system for pump equipment of the present invention through a physical test platform, the test environment was constructed as 5 pumps with a flow rate of 45 Furthermore, a multi-stage series pipe network consisting of centrifugal pumps with a head of 50m was constructed. Each centrifugal pump was an independent node in a directed acyclic topology model, and the directed edges between nodes represented the flow direction of the process medium. The experimental data were acquired from physical sensors, with the pressure sensor measuring between 0 and 2.0 MPa with a measurement accuracy of 0.25%. The vibration sensor signal acquisition frequency was set to 20 kHz, and the sampling period was... The settings directly relate to the system's accuracy in capturing the transmission characteristics of physical disturbances. Since the medium wave velocity in the pipeline is approximately 1100 m / s and the physical distance between each pump group is 15 m, in order for the system to be able to identify pressure pulsation peaks across nodes, while simultaneously controlling the processor's instantaneous computational occupancy rate below 60%, the sampling period... The timeframe was set to 10ms. To simulate a real industrial electromagnetic environment, Gaussian white noise with a signal-to-noise ratio of 25dB was superimposed onto the acquired raw signal. When verifying the causal source tracing mechanism, impeller cavitation was induced in the centrifugal pump at the beginning of the pipeline network by manual adjustment, causing a transient pulsation of 0.15MPa in the process pressure. In the control group, due to the lack of topological logical association between devices, the exceeding characteristic values ​​fed back by the sensors at each node triggered concurrent alarms. The monitoring terminal received 5 alarm signals of the same level within 2 seconds and generated 5 duplicate on-site verification work orders. In the sample group scheme of this invention, after the analysis and processing module detected that the fluctuation of the characteristic quantity of the first node exceeded the preset safety threshold, it froze the trigger timestamp. The source tracing time window with a length of 45ms is determined based on the bias weight matrix, and the running data fluctuation sequence of the upstream candidate nodes is extracted within this window.

[0042] In the verification of the synergistic effect, a partially missing control group was constructed by removing the delay weight matrix. Experimental data showed that when the missing delay weight was corrected, the waveform phase misalignment caused by the physical conduction delay resulted in a Pearson correlation coefficient of only 0.42 between the characteristic sequence of the first-end pump and the downstream pump, indicating that the system could not establish a causal mapping. After loading the time-biased weight calculated from the physical pipeline length and flow velocity parameters, the Pearson correlation coefficient of the sample group of this invention increased to 0.93 under the same operating conditions. Based on this, the system determined that the abnormality of the downstream node was caused by physical conduction, suppressing the generation of derivative alarm signals. For the verification of the key parameter boundaries, the system adjusted the sampling period. Perform gradient comparison when sampling period When the sampling period is set to 100ms, which exceeds the upper limit of the operating range, the system cannot effectively sample the characteristic peaks of the medium pressure pulsation, resulting in significant fluctuations in the Pearson correlation coefficient and a decrease in the accuracy of root cause fault location from 98.2% to 64.5%. When the lower limit of 1ms, which is outside the working range, is set, the massive data stream generated by the data acquisition module causes the processor's computing load to continuously exceed 92%, resulting in a 1.5s response lag when the resource scheduling instructions generated by the management output module are pushed to the management terminal. The test data confirms that 10ms to 50ms constitutes the optimal working range that balances positioning accuracy and system response time.

[0043] In the verification of the calculation logic of the maintenance scheduling weight W, the fault impact coefficient α is set to 0.6 and the topology weight coefficient β to 0.4. When the state degradation degree D of the root cause fault node at the end of the branch path is 0.45 and the total downstream out-degree is 2, the calculated maintenance scheduling weight W is 1.07. When the state degradation degree D of the root cause fault node at the upstream of the main path increases to 0.88 and the total downstream out-degree is 6, the calculated maintenance scheduling weight W increases to 2.93. The management output module prioritizes maintenance work orders based on the difference in the value of W, so that the resource scheduling instructions pushed to the management terminal include... With clear spare parts requirements and maintenance priorities, the results of this experiment show that by introducing time bias weights based on physical pipeline parameters into the directed acyclic topology model, the system can extract stable causal correlation features from signals containing 25dB of interference noise. This scheme quantifies the topological contribution of downstream nodes of the root cause fault node, making the urgency of management decision outputs monotonically positively correlated with the actual damage risk of the process system. The experiment verifies the complete logical closed loop from original state monitoring to management scheduling command output, confirming that the system can convert the underlying physical fluctuations of the equipment into decision indicators with management orientation.

[0044] Example 3: In the real-time monitoring and management system for pumps and turbines of this invention, applied to the operation and maintenance scenario of a pump and turbine group in an oil refinery with multiple branch pipelines, the system faces challenges such as performance degradation and sensor benchmark drift caused by long-term equipment operation. This results in a lack of standardized and unified evaluation benchmarks for the initial values ​​of key management parameters such as condition deterioration across different pump and turbine specifications. When the heavy oil to oil pump located at the bottom of the atmospheric and vacuum distillation tower shows a tendency for seal failure, if the system sets management weights based on general experience values, the urgency of the resource scheduling instructions output by the management terminal deviates from the actual process risk, making the resource allocation scheme lack specificity. To address the uncertainties in the parameter quantification process, the analysis and processing module, after constructing a directed acyclic topology model, initiates a calibration process for time-biased weights. The system obtains the physical pipeline length L between adjacent pump and turbine nodes and the theoretical flow rate of the process medium under average operating conditions. The analysis and processing module determines the theoretical propagation delay τ by calculating the ratio of the two. The processor determines the correction factor γ based on the pipeline inner wall roughness and the medium viscosity coefficient. The quantization logic of the time bias weight is as follows: ,in, For time bias weighting; γ is a preset medium conduction correction factor; L is the physical pipeline length between nodes. For the process medium flow rate, the analysis and processing module transforms the discrete physical pipeline parameters into a bias weight matrix through the above procedures, providing a logical basis for determining the source tracing time window. Specifically, the medium conduction correction factor γ is not a fixed constant, but is calculated by the processor using the Darcy-Weisbach drag coefficient formula, inputting the absolute roughness equivalent of the pipeline inner wall and the kinematic viscosity coefficient at the current medium temperature, and mapping it to a dimensionless value between 1.05 and 1.35, thereby accurately compensating for the wave velocity hysteresis effect caused by pipeline fluid friction.

[0045] During the process of determining the degree of degradation D, the data acquisition module acquires the vibration acceleration time series of each node in the annular data buffer, the analysis and processing module extracts the spectral energy distribution of the sequence within a preset frequency range, and the processor retrieves the device's factory reference energy spectrum recorded in the memory. and fault trigger limit energy value Combined with the currently measured average characteristic frequency energy The state degradation degree D, which characterizes the degree of physical state degradation, is calculated, and the specific mathematical mapping relationship is as follows: Where D is the degree of state degradation; This represents the average energy of the currently measured characteristic frequency. This is the baseline energy value for equipment operation. The preset fault limit energy value; when Approaching When the value of D converges towards 1.0, it reflects the urgency of equipment maintenance. The analysis and processing module retrieves historical condition deterioration data sequences and uses an autoregressive integral moving average (ARIMA) model to extrapolate and predict the equipment deterioration trend over time. It calculates the time period required for the predicted condition deterioration to reach the 1.0 threshold and outputs this time period as the predicted remaining life of the equipment. It also compares the deviation with the fixed maintenance cycle preset by the system to realize the transformation from time-based planned maintenance to condition-based predictive maintenance. When the vibration characteristic of the heavy oil pump exceeds the safety threshold, the system uses the calibrated time-biased weights. The source tracing time window Δt is determined as follows: of The processor calculates the Pearson correlation coefficient between the pump and its upstream node's running data sequence within the alignment window. When the correlation coefficient is less than 0.35 and the state degradation degree D is greater than 0.75, the pump is determined to be the source of the anomaly, suppressing the derivative alarms caused by the pump's flow disturbance at downstream nodes and eliminating invalid work order interference.

[0046] In a resource scheduling example for a petrochemical pump group cluster, the management output module matches corresponding technical instructions from a pre-set resource library based on the value range of the maintenance scheduling weight W. When the calculated W is greater than or equal to 2.5, the system associates an emergency requisition list containing sealing components and bearing spare parts, and generates a real-time task order assigning senior fitters, with a suggested maintenance duration of 4.8 hours. If W is within the range of 1.0 to 2.5, a preventative maintenance work order is generated, assigning junior technicians to complete routine lubrication status checks. This method transforms management evaluation indicators into action instructions for on-site execution, enabling dynamic allocation of maintenance resources under different risk levels. The management output module updates the maintenance scheduling weight W based on the calculated D and the node topology out-degree. The management terminal displays the resource scheduling instructions generated by the system, which include... To address the spare parts requisition list for heavy oil to oil pump seal replacement and the specific scheduling needs of fitter trades, this system introduces a physical energy distribution-based state degradation calibration and a delay weight matrix calculation based on pipeline fluid dynamics during the data preprocessing stage. This transforms the underlying physical signal fluctuations into digital indicators with management decision-making guidance value. By publicly disclosing the execution path of parameter calibration and feature quantification, this solution transforms the previously experience-based threshold system into a management model supported by physical causal logic. While identifying the root cause of the fault, the system provides an objective quantitative basis for the generation of resource scheduling instructions, solving the problem of resource allocation imbalance in the operation and maintenance of pump groups in large-scale industrial sites. Through the precise mapping of physical parameters to management indicators, the system improves the operational reliability of the management and supervision system in complex topology environments.

[0047] Example 4: In the field deployment scenario of a real-time monitoring and management system for pumps and turbines in a newly built integrated refining and chemical plant, the system faces a cluster of equipment with different energy efficiency levels and management strategy requirements. The analysis and processing module determines the fault impact coefficient α and topology weight coefficient β through a pre-set offline data regression analysis process. The system retrieves a dataset of historical fault maintenance records for pumps and turbines under similar operating conditions over 24 natural months. The processor constructs a loss function reflecting the associated risks of the equipment based on the ratio of unplanned downtime caused by a single equipment failure to the downtime losses of downstream processes. A gradient descent algorithm is used to find the coefficient combination that converges the objective function of historical maintenance resource allocation rationality. To clarify the convergence path of coefficient optimization, the system internally solidifies the resource allocation model calibration procedure. Based on the least squares principle, the optimization benchmark is minimizing the deviation between the model's predicted weights and historical actual losses. The objective function J of historical maintenance resource allocation rationality is expressed as follows: ,parameter for and The difference, J, represents the global error convergence value of the optimization iteration process. This represents the calculation of the maintenance scheduling weight prediction value for the k-th historical fault sample using undetermined coefficients. The representative is a quantitative indicator of actual downtime economic loss extracted from the equipment management ledger and linearly mapped to the range of 0 to 10. The processor calculates the partial derivatives of the objective function J with respect to each unknown coefficient, and updates the coefficient values ​​successively along the negative gradient direction until the error change rate between two iterations is less than one-thousandth. The quantification rules for calculating the maintenance scheduling weight W are as follows: Where W is the maintenance scheduling weight; α is the fault impact coefficient; β is the topology weight coefficient; and D is the state degradation degree. The out-degree value of the i-th downstream path node in the directed acyclic topology model is given by the root cause fault node. For the oil pump group cluster, the calculated α value is 0.62 and the β value is 0.38. This process aligns the maintenance scheduling weight W generated by the management output module with the production safety evaluation benchmark of the specific unit, ensuring that the priority sequence of the quantitative maintenance work order output by the system meets the actual management needs on site. In the management of this oil pump group cluster, the energy-saving management unit extracts the active power data of the motors of each node in the topology model in real time, calculates the energy consumption distribution entropy on the process path using the formula, and calculates the sum of the negative logarithm of the ratio of the power of each node to the total power of the group and the product of the ratio. When the energy consumption distribution entropy deviates from the maximum entropy equilibrium state by more than the set 15% dead zone, the system substitutes the flow-head characteristic curve of the pump group and the pipeline resistance equation in reverse, and solves the target operating frequency of each pump inverter that minimizes the total power by using the interior point method. This is then output as a load balancing scheduling command to the distributed control system for automatic frequency adjustment.

[0048] When the system is deployed in a complex media transport network environment with dynamically changing topology or multiple branch adjustments, the analysis and processing module corrects the time bias weights through the calibration process of the equipment logical topology map. Technicians apply pulse pressure disturbance signals to the first node of the network, and the data acquisition module simultaneously acquires the response characteristic time points in the corresponding ring data buffers of each downstream adjacent node. The processor calculates the residual between the measured transmission delay and the theoretical transmission delay, and uses this residual as feedback to compensate for the media transmission correction factor γ in the delay weight matrix. The time bias weights... The calculation logic is as follows: ,in, The time bias weight is γ; the medium conduction correction factor is L; and the physical pipeline length between nodes is L. The process medium velocity is used; transient pressure disturbances are actually propagated in the fluid network in the form of waves. The propagation wave velocity exhibits nonlinear decay under cavitation-induced gas-liquid two-phase flow conditions due to drastic changes in medium compressibility. The analysis and processing module incorporates an online wave velocity calibration procedure. Based on the principles of fluid transient dynamics, the processor extracts characteristic pressure peak values ​​captured by two adjacent sensor nodes within the same 10-millisecond sampling period, calculates the peak arrival time difference, and divides this time difference by the physical pipeline length L to obtain the transient disturbance propagation wave velocity. The data acquisition module detected that the instantaneous coefficient of variation of the process pressure exceeded the preset steady-state extreme point. The processor initiated the variable replacement mechanism, adjusting the process medium flow rate within the time bias weight calculation logic. Replace with the propagation speed of transient disturbances The tracing time window judgment benchmark is changed from steady-state transport displacement to transient stress wave propagation velocity, correcting the physical spatiotemporal image deviation caused by phase change in multiphase media environment. When the Pearson correlation coefficient corresponding to all directed edges reaches a discrimination accuracy of 0.90 or above under simulated fault triggering, the parameter solidification is completed. Through this pre-debugging procedure, the system establishes a correlation with the physical characteristics of the pipeline network at the data processing layer, making the cause-source tracing mechanism adapt to the dynamic transmission lag caused by pipeline fouling or fluid viscosity fluctuations, and improving the reliability of management and supervision command generation in large-scale monitoring environment.

[0049] Example 5: In the parameter initialization procedure for the real-time monitoring and management system of pump groups in a newly built petrochemical park, the system determines the correlation threshold for judging disturbance propagation. The analysis and processing module eliminates the interference of on-site background electromagnetic noise and random process fluctuations on root cause fault location by collecting controlled background data. During the operation period when the pump equipment is under rated conditions and there are no external disturbances, technicians continuously acquire the characteristic quantity fluctuation sequence between adjacent node pairs according to the sampling period. The processor calculates the cross-correlation function distribution of no less than 1000 sampling points and extracts the background correlation coefficient value under steady-state conditions. A controlled pressure fluctuation with an amplitude not exceeding 20% ​​of the safety threshold is introduced at the pump node at the beginning of the pipeline network. The analysis and processing module records the response sequence of each downstream node to this fluctuation and calculates its transmission correlation coefficient with the source sequence. The system determines the association threshold η based on the background noise distribution and the controlled conduction distribution. The specific quantization formula is as follows: Where η is a preset association threshold; The transmission correlation coefficient value under controlled pressure fluctuations; As the background correlation coefficient value under steady-state operating conditions, this procedure establishes a discrimination boundary for specific process environments at the data processing layer, so that the cause-effect tracing mechanism can maintain a stable disturbance recognition rate when facing different levels of background noise.

[0050] When the system is deployed in an environment with pump groups containing complex pipelines, the data acquisition module adjusts the storage depth N of the annular data buffer to ensure the data integrity within the traceability time window, and the analysis and processing module extracts the maximum theoretical propagation delay between each node pair in the bias weight matrix. and combined with the sampling period To determine the storage amount required for feature alignment, and to avoid data sequence loss caused by fluctuations in pipeline medium flow rate, the storage depth N of the ring data buffer is determined as the product of the reciprocal of the sampling period and the delay redundancy coefficient. The specific formula is as follows: Where N is the storage depth of the circular data buffer; κ is the preset latency redundancy coefficient, and its value is selected as 1.5; This represents the maximum theoretical propagation delay between adjacent nodes in a directed acyclic topology model. The sampling period is defined as follows: This procedure converts the conduction characteristics of the physical pipeline into a buffer memory depth limit. After the processor obtains the abnormal trigger signal, it completely extracts the timing information containing the rising edge of the disturbance from the circular data buffer. This enables the resource scheduling instructions generated by the management output module to cover all equipment evaluation indicators affected by the disturbance, thereby suppressing the alarm signals derived from process conduction and improving the instruction output efficiency of the management terminal.

[0051] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A real-time monitoring management system for a machine-pump plant, characterized by, The system includes: The data acquisition module is used to acquire the operating status dataset of each node in the pump equipment group, and store the operating status dataset into the circular data buffer corresponding to each node according to the sampling time sequence. The operating status dataset includes the vibration amplitude, frequency distribution and process pressure parameters collected at each monitoring point. The analysis and processing module is used to construct a directed acyclic topology model reflecting the flow direction of the process medium based on the physical connection topology of the industrial pipeline network. Specifically, the module extracts physical pipeline length and velocity parameters to determine the theoretical propagation delay between adjacent nodes in the directed acyclic topology model, and maps this theoretical propagation delay to various directed time bias weights in the model, forming a bias weight matrix to characterize the lag in the propagation of disturbances in physical space. Furthermore, when the characteristic value fluctuation of the first node in the directed acyclic topology model exceeds a preset threshold, the module determines a source tracing time window based on the bias weight matrix. Within this window, it performs correlation matching of the fluctuation sequences between the first node and upstream candidate nodes, thereby identifying the root cause fault node causing the physical chain reaction at the initial stage of alarm generation. It also implements alarm suppression logic for derivative alarm signals generated by associated nodes other than the root cause fault node. The management output module is used to calculate the maintenance scheduling weight based on the out-degree of the downstream nodes of the root cause fault node in the directed acyclic topology model determined by the analysis and processing module, combined with the state degradation degree of the root cause fault node, and output a quantitative maintenance work order generated by prioritizing the maintenance scheduling weight.

2. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, When performing correlation matching of fluctuation sequences, the analysis and processing module obtains the characteristic fluctuation sequence of the first node and the running data fluctuation sequence of the upstream node in the directed acyclic topology model within the source tracing time window. It calculates the Pearson correlation coefficient between the characteristic fluctuation sequence and the running data fluctuation sequence, and when the Pearson correlation coefficient is greater than the preset correlation threshold, it determines the upstream node as the disturbance source corresponding to the root cause fault node.

3. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, In the directed acyclic topology model, nodes represent independent pumps and machinery, and directed edges between nodes represent the flow direction of industrial fluid process media. The analysis and processing module performs sliding window dimensionality reduction processing on the time-series data of each node in the pump and machinery group, transforming the original high-dimensional operating status data of each node into a multi-dimensional state feature vector that characterizes the health trend of the equipment.

4. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, The analysis and processing module is also used to prioritize the generated quantitative maintenance work orders according to the numerical values ​​of the maintenance scheduling weights, and push the ranking results to the management output module in real time. The management output module generates resource scheduling instructions based on the ranking results, including spare parts requirements, maintenance types, and suggested repair durations, to guide maintenance resources to concentrate on the root cause fault nodes.

5. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, The sampling period of the data acquisition module is set to 100ms; the analysis and processing module is also used to determine the first node that generates the characteristic value fluctuation as the root cause fault node causing the alarm storm when the correlation coefficient is less than the preset correlation threshold, or when there is no upstream adjacent node that meets the preset conditions in the directed acyclic topology model, and to generate an independent source tracing and diagnosis report for the first node.

6. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, The management output module is also used to maintain the equipment management ledger module, which records the physical connection topology, historical maintenance cycles, component life distribution data, and real-time operating status levels of pump equipment. The management output module provides macro-level maintenance planning suggestions for pump equipment groups to the management decision-making level by associating real-time operating status levels with maintenance scheduling weights.

7. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, After identifying the root cause of the fault, the analysis and processing module retrieves the corresponding spare parts inventory data from the equipment management ledger module based on the equipment model of the root cause fault. When it determines that the current spare parts inventory value is lower than the preset safety stock threshold, it automatically sends a spare parts procurement warning signal to the management output module, or the management output module directly triggers the automatic procurement application process instruction for the corresponding spare parts.

8. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, The analysis and processing module is also used to predict the remaining lifespan of the equipment based on the long-term trend changes of the operating status dataset, and to calculate the deviation between the remaining lifespan of the equipment and the preset maintenance cycle. When the deviation exceeds the preset alarm threshold, the baseline of the state degradation of the corresponding node in the directed acyclic topology model is automatically increased.

9. The real-time monitoring and management system for pump equipment according to claim 1, characterized in that, The management output module also includes an energy-saving management unit, which is used to acquire real-time energy consumption monitoring data of each node in the pump group and calculate the energy consumption distribution entropy on the process path based on the load distribution status of each node in the directed acyclic topology model. As a result, it outputs load balancing scheduling instructions aimed at optimizing the overall operating load distribution of the pump group and reducing the total energy consumption level.

Images