Method and system for monitoring operation of heavy load multistage centrifugal pump
By constructing a nonlinear coupling model and a multi-factor fusion scoring function, the problem of fault early warning for multi-stage centrifugal pumps in complex environments was solved, achieving high-precision fault identification and early warning, and improving the stability and robustness of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN TANE OCEAN PUMP CO LTD
- Filing Date
- 2025-06-27
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies are insufficient for effectively monitoring and early warning of faults in multi-stage centrifugal pumps under complex operating environments, especially under long-term high pressure and high flow conditions. Traditional methods lack modeling of the collaborative operation mechanism between multi-stage pumps, resulting in low accuracy of anomaly identification, high false alarm rate, and poor robustness.
By constructing a nonlinear coupling model, utilizing timestamp alignment and unified sampling frequency, operating parameters of multi-stage centrifugal pumps are collected synchronously, deviation analysis is performed, sensitive operating parameters are identified, and adaptive dynamic thresholds are generated. Combined with a multi-factor fusion scoring function, fault early warning is provided.
It improves the accuracy and response efficiency of multi-stage centrifugal pump fault early warning, enhances the ability to identify system coupling anomalies, reduces the false alarm rate, and improves the system's stability and intelligent perception level.
Smart Images
Figure CN120537747B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of operational safety technology, and in particular to a method and system for monitoring the operation of a heavy-duty multistage centrifugal pump. Background Technology
[0002] With the rapid development of intelligent monitoring and predictive maintenance technologies for industrial equipment, multistage centrifugal pumps, as core fluid transport equipment, are crucial for the stable operation of industrial systems, particularly under heavy-load conditions. Their operational safety and fault early warning capabilities are critical guarantees for stable operation. Especially under long-term high-pressure, high-flow operating conditions, complex dynamic coupling relationships form between the pump stages within a multistage centrifugal pump. Even minor anomalies in a single pump stage can be amplified through interstage coupling, ultimately leading to decreased overall efficiency or even structural failure. Therefore, developing an operational monitoring method capable of comprehensively sensing the operating status of multistage centrifugal pumps, identifying potential anomalies, and providing early warnings is of great significance for improving equipment stability and reducing maintenance costs.
[0003] In the prior art, patent CN117909621B discloses a method and system for monitoring the operating status of a submersible electric pump for sewage and sludge. This method collects current data during the pump's operation, extracts the PR component to construct a pumping blockage index, and combines historical time-period data with the blockage degree to construct a sewage viscosity interference disorder index. Finally, a neural network is used to assess the current operating status. This method achieves pump monitoring based on multi-index fusion under specific operating conditions and has certain engineering applicability. However, this method is mainly applied to single pump units and is not directly applicable to multi-stage centrifugal pumps with complex equipment structures and strong parameter coupling.
[0004] Furthermore, traditional operation monitoring methods mostly focus only on changes in individual signals (such as current or vibration), neglecting the collaborative operation mechanism between multi-stage pumps and lacking modeling research on the nonlinear coupling relationship between operating parameters, making it difficult to effectively reveal systemic operational deviations caused by coupling anomalies. On the other hand, traditional threshold methods or simple neural network methods have problems such as poor adaptability, low anomaly identification accuracy, and high false alarm rate. Especially when facing complex operating environments or unknown disturbances, their robustness and generalization ability are significantly limited.
[0005] Therefore, there is an urgent need in this field for a new method to monitor the operating status of multi-stage centrifugal pumps in complex operating environments, so as to improve the accuracy and response efficiency of pump system fault early warning. Summary of the Invention
[0006] In view of this, the present invention provides an operation monitoring scheme for a heavy-duty multistage centrifugal pump. By using time-series synchronization and nonlinear coupling modeling of operation data, abnormally sensitive operation parameters are identified and disturbance sources are located. Then, an adaptive dynamic threshold for the disturbance operation parameters is generated. Combined with a multi-factor fusion scoring function, the system can achieve graded judgment and early warning push for potential operation faults.
[0007] To achieve the above objectives, the present invention provides a method for monitoring the operation of a heavy-duty multistage centrifugal pump, comprising the following steps: S1: Use timestamp alignment and uniformly adjust the sampling frequency to the target frequency to synchronously collect the operating data of each stage of the multi-stage centrifugal pump; The multistage centrifugal pump consists of multiple centrifugal pump stages arranged in series, each of which has an independent working chamber, impeller system and pump shaft structure. S2: Construct a nonlinear coupling model between operating parameters at all levels, use the nonlinear coupling model to perform deviation analysis on the operating data, obtain the residual sequence of operating parameters at all levels, and extract sensitive operating parameters that deviate significantly from normal coupling behavior based on the residual deviation score, thus forming a set of sensitive operating parameters; S3: Perform trend analysis and coupled mutation analysis on sensitive operating parameters, and combine the results of deviation analysis to extract the disturbance operating parameters as disturbance sources from the set of sensitive operating parameters, thus forming a set of disturbance operating parameters; S4: Based on the adaptive dynamic threshold of the disturbance operating parameters and the residual deviation score, a multi-factor fusion early warning scoring function is constructed. The multi-factor fusion early warning scoring function is used to monitor and warn of potential operating faults of multi-stage centrifugal pumps.
[0008] Preferably, step S1 includes: Sensor modules for sensing the operating parameters of the centrifugal pump stage are arranged in the centrifugal pump stage. A master clock synchronization signal generator is set in the control system. A sampling trigger command is sent to the sensor modules of each centrifugal pump stage through the CAN bus to drive each sensor module to start the sampling task synchronously. Each sensor module collects the operating data of the centrifugal pump stage and performs normalization processing. The operating data consists of the sequence data of various operating parameters. Set a target frequency, adjust the sampling frequency of the normalized operating data of each centrifugal pump stage, upsample the sequence data of low-frequency operating parameters to the target frequency using linear interpolation, and downsample the sequence data of high-frequency operating parameters to the target frequency using a mean sliding window method, to obtain the operating data after sampling frequency adjustment, where the sequence data of each operating parameter has a length of N.
[0009] It should be noted that the operating parameters include head, flow rate, pressure, power, operating efficiency, rotational speed, temperature, and axial thrust; where head is the height to which the liquid is lifted by the centrifugal pump stage, flow rate is the volume of liquid delivered by the centrifugal pump stage per unit time, pressure is the position pressure of the liquid in the centrifugal pump stage, power is the energy consumed by the centrifugal pump stage during operation, operating efficiency is the ratio between water power and shaft power in the centrifugal pump stage, water power is the power obtained by the liquid itself when the centrifugal pump stage lifts the liquid to the target height or delivers it to the target pressure, shaft power is the input power required to drive the pump shaft to rotate, rotational speed is the rotational speed of the impeller in the centrifugal pump stage, temperature is the temperature of the liquid in the centrifugal pump stage, and axial thrust is the force exerted on the pump shaft along the axial direction by the unbalanced force generated by the liquid on the impeller in the centrifugal pump stage.
[0010] Preferably, in step S2, the nonlinear coupling model includes a time-series slice layer, an inter-level parameter joint layer, a nonlinear coupling modeling layer, and a prediction residual output layer; and The time-series slicing layer is used to slice the operating data of each centrifugal pump stage using a time sliding window to obtain sliced operating data of each centrifugal pump stage. The interstage parameter joint layer is used to capture the sensitivity of each centrifugal pump stage to changes in the overall operating state and generate dynamic sensitivity weights for each centrifugal pump stage. Based on the dynamic sensitivity weights, the sliced operating data is weighted to obtain a joint operating state vector. The nonlinear coupling modeling layer is a hierarchical gated recurrent neural network structure, used to receive the joint operating state vector and generate prediction sequences for each centrifugal pump stage under different operating parameters. The prediction residual output layer is used to perform residual calculation on the prediction sequences of each operating parameter to obtain the residual sequence of each operating parameter.
[0011] Preferably, in step S2, deviation analysis of the operating data is performed using the nonlinear coupling model, including: The operating data of each centrifugal pump stage is sliced to obtain sliced operating data containing multiple slices. Each slice consists of... The sequence data vector is composed of the sequence data values of each operating parameter at the sampling time. Indicates the window size of the time-sliding window; The dynamic sensitivity weights of each centrifugal pump stage are calculated using the following expression: , ; in, Indicates the first Dynamic sensitivity weights of each centrifugal pump stage This indicates the total number of stages in a centrifugal pump. Indicates the weighting factor. Indicates the first The cosine similarity between the operating data of each centrifugal pump stage and the overall operating data, wherein the overall operating data is... The sum of the operating data of each centrifugal pump stage, Indicates the first The mean of the standard deviations of the sequence data of each operating parameter in the operating data of a centrifugal pump stage; The slice operation data is weighted based on dynamic sensitive weights to obtain a joint operation state vector; The nonlinear coupling modeling layer decomposes the input joint operating state vector into slice data sequences for each centrifugal pump stage under different operating parameters. The slice data sequences are then input into a gated recurrent unit to obtain feature codes for the slice data. An attention mechanism is used to generate attention weights for the feature codes of all slice data in the sequence. These feature codes are then attention-weighted, and a fully connected layer is used to predict the attention-weighted feature codes, resulting in a prediction sequence for each centrifugal pump stage under different operating parameters. The length of the prediction sequence is... , This indicates the number of slices in the slice execution data. Indicates the step size of the time sliding window; Calculate the mean value of each slice in the sliced operation data under different operating parameters, and sort all the means according to the slice order to obtain the mean value sequence of different operating parameters in the centrifugal pump stage. Calculate the sequence difference between the mean value sequence and the predicted sequence as the residual sequence of the centrifugal pump stage under different operating parameters.
[0012] Preferably, in step S2, the extraction of sensitive operating parameters that significantly deviate from normal coupling behavior based on the residual deviation score includes: The residual deviation score for different operating parameters in each centrifugal pump stage is calculated using the following expression: , , ; in, Indicates the first The r-th operating parameter in a centrifugal pump stage The residual deviation score, Indicates running parameters The residual sequence, Representing the residual sequence The j-th sequence value in the sequence, Indicates the preset first The standard deviation of the residual sequence of the r-th operating parameter in a centrifugal pump stage, where R represents the number of types of operating parameters; Set a deviation threshold, and use the operating parameters whose residual deviation scores are higher than the deviation threshold as sensitive operating parameters that deviate significantly from the normal coupling behavior under this centrifugal pump stage.
[0013] It should be noted that the sequence values in the residual sequence represent the difference between the actual observed values and the normal expected values of the system. The normal expected values are the predicted sequence, which corresponds to the sequence that should exist under the normal coupling mechanism. When a certain parameter of the pump group is disturbed (such as increased vibration caused by bearing failure), its residual will deviate. That is, the higher the residual deviation score, the more the operating parameter deviates from its historical coupling model, and the more likely it is to be a source of disturbance or fault.
[0014] This invention utilizes time-series slicing and a dynamic sensitive weighting mechanism to divide continuous operating parameter data into multiple slices within a sliding time window, enabling modeling of the operating states of multi-stage centrifugal pumps across different time segments. Each pump stage's slice sequence not only retains time-series characteristics but also undergoes a dynamic sensitive weighting mechanism to adjust its contribution weight based on the pump stage's response to changes in the global operating state. This weighting mechanism effectively addresses modeling bias caused by varying coupling strengths between pump stages, enhancing the model's responsiveness to "coupling anomalies." The nonlinear coupling modeling layer constructed in this invention combines gated recurrent units and an attention mechanism, enabling deep modeling and feature extraction of slices for each operating parameter. The intention mechanism can automatically determine the contribution of each time slice to the overall coupling state and assign different weights to its feature encoding, further improving the accuracy and robustness of the model in nonlinear coupling feature modeling. After obtaining the predicted sequence of each operating parameter, this invention constructs a residual sequence with the actual sampled data and performs standardized residual evaluation to calculate the residual deviation score of different operating parameters. This effectively identifies sensitive operating parameters that deviate significantly from normal coupling behavior in different operating parameter dimensions. The residual deviation scoring mechanism comprehensively considers the standardized influence of historical residual fluctuations (i.e., residual standard deviation) and current residuals (i.e., sequence values in the residual sequence), and has adaptive capability and sensitivity, enabling dynamic adjustment of the anomaly detection response to different pump stages and parameters.
[0015] Preferably, step S3 includes: The trend score for each sensitive operating parameter is calculated using the following expression: ; in, Indicates sensitive operating parameters Trend ratings Indicates sensitive operating parameters residual sequence The j-th sequence value in the sequence, Indicates sensitive operating parameters residual sequence The mean; The coupling mutation index of each sensitive operating parameter is calculated as follows: ; in, Indicates sensitive operating parameters The coupling mutation index, Indicates sensitive operating parameters The attention weight of the j-th slice in the corresponding slice data sequence; The residual deviation score, trend score, and coupling mutation index of each sensitive operating parameter are weighted to obtain the disturbance intensity score of the corresponding sensitive operating parameter. The top G sensitive operating parameters with the highest disturbance intensity scores are selected from the set of sensitive operating parameters as the disturbance operating parameters.
[0016] It should be noted that, The higher the value, the more sensitive the operating parameter. Significant fluctuations in the level of attention within the system's attention mechanism indicate a sudden change in the coupling structure. The higher the value, the more sensitive the operating parameter. There is a clear upward trend.
[0017] Preferably, step S4 includes: The adaptive dynamic thresholds for each disturbance operating parameter are calculated using the following expression: ; ; ; in, Indicates the first Disturbance operating parameters Adaptive dynamic threshold, Indicates the first Disturbance operating parameters The mean baseline threshold, Indicates the first Disturbance operating parameters The residual deviation threshold, Indicates the first The mean of the residual sequence of the perturbation operating parameters. Indicates the first The standard deviation of the residual sequence of the perturbation operating parameters. Indicates the first The j-th sequence value of the residual sequence of the perturbation operating parameters. Both represent adjustment factors; A multi-factor fusion early warning scoring function is constructed, which takes the adaptive dynamic threshold of the disturbance operation parameters of each centrifugal pump stage and the residual deviation score as input, and the early warning score of each centrifugal pump stage as output. If the early warning score is higher than the preset early warning threshold, it indicates that there is a potential operation failure of the centrifugal pump stage, and an early warning notification is issued.
[0018] It should be noted that, The mean absolute deviation of the residuals reflects the overall deviation trend of the residual sequence. This part is used to assess the tolerance of abnormal evolution trends of disturbed operating parameters. When the residual sequence deviates significantly from the mean, even if the standard deviation is not high, the threshold can be adjusted through this part to enhance the early identification capability of potential operating faults. This represents the mean baseline, which takes into account the current average residual level and adds a safety margin to cover residual deviations from the scoring range during normal operation.
[0019] Preferably, in the multi-factor fusion early warning scoring function, the expression for the early warning score is as follows: , ; in, Indicates the first Early warning score for a centrifugal pump level. Indicates the first A set of disturbance operating parameters for a centrifugal pump stage. Represents the set of disturbance operating parameters Any disturbance operating parameter in, Indicates disturbance operating parameters The residual deviation score, Indicates disturbance operating parameters Adaptive dynamic threshold, Denotes the discriminant function, if If the result is positive, the value of the discriminant function is 1; otherwise, the value of the discriminant function is 0.
[0020] Compared with the prior art, the present invention has at least the following beneficial effects: This approach models continuously running data slices using a time sliding window and introduces a dynamic sensitive weighting mechanism to assign differentiated weights to time segments of different pump stages, enhancing the ability to identify system coupling anomalies. It employs a nonlinear modeling method combining gated cyclic units and an attention mechanism, extracting deep features of each parameter sequence and automatically assessing the impact of different time segments on the overall system state. Subsequently, through residual analysis between the predicted sequence and actual data, a standardized residual deviation score is calculated, effectively identifying anomalous factors deviating from normal coupling behavior across different parameter dimensions. This scoring mechanism is adaptive and can dynamically respond to changes in the operation of multi-stage centrifugal pumps.
[0021] This invention addresses the core technical challenge of accurately identifying potential disturbance sources and providing early warnings from multi-source, high-dimensional operating parameters. It innovatively proposes a judgment mechanism integrating trend scoring, mutation indicators, adaptive dynamic thresholds, and a multi-factor fusion early warning function. First, by introducing trend scoring and a coupled mutation index, the mean shift and disturbance amplitude of the residual sequence are characterized. This accurately captures non-stationary trends and short-term mutation behaviors caused by potential disturbance factors during system operation. A higher trend score indicates a significant upward trend in the abnormality of sensitive operating parameters, while a higher coupled mutation index reflects instantaneous jumps. This trend-mutation dual-dimensional indicator system systematically improves the perception accuracy of different types of disturbances. Then, an adaptive dynamic threshold is constructed. Through the weighted superposition of the mean, standard deviation, and fluctuation intensity residual terms, personalized thresholds are obtained for different disturbance parameters under the dual regulation of historical statistical regularity and short-term volatility. This effectively solves the problem of false alarms or missed alarms in traditional static threshold strategies under nonlinear, strongly coupled scenarios, improving the continuity, stability, and intelligent perception level of the overall system operation. Attached Figure Description
[0022] Figure 1 This is a flowchart of an operation monitoring method for a heavy-duty multistage centrifugal pump provided in Embodiment 1 of the present invention.
[0023] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0024] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0025] This application provides a method for monitoring the operation of a heavy-duty multistage centrifugal pump. The execution entity of this method includes, but is not limited to, at least one electronic device configured to execute the method provided in this application, such as a server or a terminal. In other words, the method for monitoring the operation of a heavy-duty multistage centrifugal pump can be executed by software or hardware installed on a terminal device or a server device, and the software may be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.
[0026] Example 1 like Figure 1 As shown, a method for monitoring the operation of a heavy-duty multistage centrifugal pump includes the following steps: S1: Using timestamp alignment and uniformly adjusting the sampling frequency to the target frequency, synchronously collect operating data of each stage of operation parameters in a multi-stage centrifugal pump; including: The multistage centrifugal pump consists of multiple centrifugal pump stages arranged in series, each of which has an independent working chamber, impeller system and pump shaft structure. Sensor modules for sensing the operating parameters of the centrifugal pump stage are arranged in the centrifugal pump stage. A master clock synchronization signal generator is set in the control system. A sampling trigger command is sent to the sensor modules of each centrifugal pump stage through the CAN bus to drive each sensor module to start the sampling task synchronously. Each sensor module collects the operating data of the centrifugal pump stage and performs normalization processing. The operating data consists of the sequence data of various operating parameters. Set the target frequency, adjust the sampling frequency of the normalized centrifugal pump stage operating data, upsample the sequence data of low-frequency sampling operating parameters to the target frequency using linear interpolation, and downsample the sequence data of high-frequency sampling operating parameters to the target frequency using a mean sliding window method, to obtain the operating data after sampling frequency adjustment; The sequence data length of each operating parameter in the operating data after the sampling frequency adjustment is N.
[0027] S2: Construct a nonlinear coupling model between operating parameters at each level, use the nonlinear coupling model to perform deviation analysis on the operating data, obtain the residual sequence of operating parameters at each level, and extract sensitive operating parameters that deviate significantly from normal coupling behavior based on the residual deviation score, forming a set of sensitive operating parameters; including: The nonlinear coupling model includes a time-slice layer, an inter-level parameter joint layer, a nonlinear coupling modeling layer, and a prediction residual output layer. The time-series slicing layer is used to slice the operating data of each centrifugal pump stage using a time sliding window, obtaining sliced operating data for each centrifugal pump stage; specifically: The operating data of each centrifugal pump stage is sliced to obtain sliced operating data containing multiple slices. Then, the... The slice operation data for each centrifugal pump stage is as follows: , , This indicates the total number of stages in a centrifugal pump. This indicates the window size of the time-sliding window. This indicates the step size of the time sliding window. Indicates the number of slices, slice execution data. The first in Each slice is , Slice In Each slice consists of a sequence of data vectors. The sequence data vector is composed of a sequence of data vectors, which are composed of the sequence data of each operating parameter at the sampling time corresponding to the sequence data vector; The interstage parameter joint layer is used to capture the sensitivity of each centrifugal pump stage to changes in the overall operating state and generate dynamic sensitivity weights for each centrifugal pump stage. Based on these dynamic sensitivity weights, the sliced operating data is weighted to obtain a joint operating state vector; specifically: The dynamic sensitivity weights of each centrifugal pump stage are calculated using the following expression: , ; in, Indicates the first Dynamic sensitivity weights of each centrifugal pump stage Indicates the weighting factor. Indicates the first The cosine similarity between the operating data of each centrifugal pump stage and the overall operating data is used to describe the coupling strength of the centrifugal pump stages within a preset time range, where the overall operating data is... The sum of the operating data of each centrifugal pump stage, Indicates the first The mean of the standard deviation of the sequence data of each operating parameter in the operating data of a centrifugal pump stage is used to describe the disturbance amplitude of the centrifugal pump stage within the preset time range. The nonlinear coupling modeling layer is a hierarchical gated recurrent neural network structure, used to receive the joint operating state vector and generate prediction sequences for each centrifugal pump stage under different operating parameters; specifically: The nonlinear coupling modeling layer decomposes the input joint operating state vector into slice data sequences of each centrifugal pump stage under different operating parameters. The slice data sequence of a centrifugal pump stage under the r-th operating parameter is as follows: Sliced data sequence The first in Each slice of data is , Indicates from slice Extract the data sequence of the r-th running parameter. R represents the number of types of operating parameters. The slice data from the slice data sequence is input into a gated recurrent unit to obtain the feature codes of the slice data. An attention mechanism is used to generate attention weights for the feature codes of all slice data in the slice data sequence. The feature codes are then attention-weighted, and a fully connected layer is used to predict the attention-weighted feature codes, resulting in a predicted sequence for each centrifugal pump stage under different operating parameters. The length of the predicted sequence is [missing information]. ; The prediction residual output layer is used to calculate the residuals of the prediction sequences of each operating parameter, thereby obtaining the residual sequences of each operating parameter; specifically: Calculate the mean value of each slice in the sliced operation data under different operating parameters, and sort all the means according to the slice order to obtain the mean value sequence of different operating parameters in the centrifugal pump stage. Calculate the sequence difference between the mean value sequence and the predicted sequence as the residual sequence of the centrifugal pump stage under different operating parameters. The residual deviation score for the residual sequence of different operating parameters in each centrifugal pump stage is calculated as follows: , , ; in, Indicates the first The r-th operating parameter in a centrifugal pump stage The residual deviation score, Indicates running parameters The residual sequence, Representing the residual sequence The j-th sequence value in the sequence, Indicates the preset first The standard deviation of the residual sequence of the rth operating parameter in a centrifugal pump stage is used in this embodiment. The standard deviation of the residual sequence under the historical normal state of the multi-stage centrifugal pump is used. Set a deviation threshold, and use the operating parameters whose residual deviation scores are higher than the deviation threshold as sensitive operating parameters that deviate significantly from the normal coupling behavior under this centrifugal pump stage.
[0028] S3: Perform trend analysis and coupled mutation analysis on sensitive operating parameters, and combine the results of deviation analysis to extract disturbance operating parameters as disturbance sources from the sensitive operating parameter set, forming a disturbance operating parameter set; including: The trend score for each sensitive operating parameter is calculated using the following expression: ; in, Indicates sensitive operating parameters Trend ratings Indicates sensitive operating parameters residual sequence The j-th sequence value in the sequence, Indicates sensitive operating parameters residual sequence The mean; The coupling mutation index of each sensitive operating parameter is calculated as follows: ; in, Indicates sensitive operating parameters The coupling mutation index, Indicates sensitive operating parameters The attention weight of the j-th slice in the corresponding slice data sequence; The residual deviation score, trend score, and coupling mutation index of each sensitive operating parameter are weighted to obtain the disturbance intensity score of the corresponding sensitive operating parameter. The top G sensitive operating parameters with the highest disturbance intensity scores are selected from the set of sensitive operating parameters as the disturbance operating parameters.
[0029] S4: Based on the adaptive dynamic threshold of disturbance operating parameters and residual deviation score, a multi-factor fusion early warning scoring function is constructed. This function is then used to monitor and warn of potential operational faults in multi-stage centrifugal pumps; including: The adaptive dynamic thresholds for each disturbance operating parameter are calculated using the following expression: ; ; ; in, Indicates the first Disturbance operating parameters Adaptive dynamic threshold, Indicates the first Disturbance operating parameters The mean baseline threshold, Indicates the first Disturbance operating parameters The residual deviation threshold, Indicates the first The mean of the residual sequence of the perturbation operating parameters. Indicates the first The standard deviation of the residual sequence of the perturbation operating parameters. Indicates the first The j-th sequence value of the residual sequence of the perturbation operating parameters. Both represent adjustment factors; A multi-factor fusion early warning scoring function is constructed, taking the adaptive dynamic thresholds of the disturbance operating parameters of each centrifugal pump stage and the residual deviation scores as inputs, and the early warning scores of each centrifugal pump stage as outputs. The expression is as follows: , ; in, Indicates the first Early warning score for a centrifugal pump level. Indicates the first A set of disturbance operating parameters for a centrifugal pump stage. Represents the set of disturbance operating parameters Any disturbance operating parameter in, Indicates disturbance operating parameters The residual deviation score, Indicates disturbance operating parameters Adaptive dynamic threshold, Denotes the discriminant function, if If the result is positive, the value of the discriminant function is 1; otherwise, the value of the discriminant function is 0. If the warning score is higher than the preset warning threshold, it indicates that there is a potential operational failure in the centrifugal pump stage, and a warning notification will be issued.
[0030] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.
[0031] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.
[0032] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0033] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A method of monitoring the operation of a heavy-duty multistage centrifugal pump, characterized in that, Includes the following steps: S1: Use timestamp alignment and uniformly adjust the sampling frequency to the target frequency to synchronously collect the operating data of each stage of the multi-stage centrifugal pump; The multistage centrifugal pump consists of multiple centrifugal pump stages arranged in series, each of which has an independent working chamber, impeller system and pump shaft structure. S2: Construct a nonlinear coupling model between operating parameters at all levels, use the nonlinear coupling model to perform deviation analysis on the operating data, obtain the residual sequence of operating parameters at all levels, and extract sensitive operating parameters that deviate significantly from normal coupling behavior based on the residual deviation score, thus forming a set of sensitive operating parameters; S3: Perform trend analysis and coupled mutation analysis on sensitive operating parameters, and combine the results of deviation analysis to extract the disturbance operating parameters as disturbance sources from the set of sensitive operating parameters, thus forming a set of disturbance operating parameters; S4: Based on the adaptive dynamic threshold of the disturbance operating parameters and the residual deviation score, a multi-factor fusion early warning scoring function is constructed. The multi-factor fusion early warning scoring function is used to monitor and warn of potential operating faults of multi-stage centrifugal pumps.
2. The operation monitoring method for a heavy-duty multistage centrifugal pump as described in claim 1, characterized in that, Step S1 includes: Sensor modules for sensing the operating parameters of the centrifugal pump stage are arranged in the centrifugal pump stage. A master clock synchronization signal generator is set in the control system. A sampling trigger command is sent to the sensor modules of each centrifugal pump stage through the CAN bus to drive each sensor module to start the sampling task synchronously. Each sensor module collects the operating data of the centrifugal pump stage and performs normalization processing. The operating data consists of the sequence data of various operating parameters. Set a target frequency, adjust the sampling frequency of the normalized operating data of each centrifugal pump stage, upsample the sequence data of low-frequency operating parameters to the target frequency using linear interpolation, and downsample the sequence data of high-frequency operating parameters to the target frequency using a mean sliding window method, to obtain the operating data after sampling frequency adjustment, where the sequence data of each operating parameter has a length of N.
3. The operation monitoring method for a heavy-duty multistage centrifugal pump as described in claim 1, characterized in that, In step S2, the nonlinear coupling model includes a time-slice layer, an inter-level parameter joint layer, a nonlinear coupling modeling layer, and a prediction residual output layer; and The time-series slicing layer is used to slice the operating data of each centrifugal pump stage using a time sliding window to obtain sliced operating data of each centrifugal pump stage. The interstage parameter joint layer is used to capture the sensitivity of each centrifugal pump stage to changes in the overall operating state and generate dynamic sensitivity weights for each centrifugal pump stage. Based on the dynamic sensitivity weights, the sliced operating data is weighted to obtain a joint operating state vector. The nonlinear coupling modeling layer is a hierarchical gated recurrent neural network structure, used to receive the joint operating state vector and generate prediction sequences for each centrifugal pump stage under different operating parameters. The prediction residual output layer is used to perform residual calculation on the prediction sequences of each operating parameter to obtain the residual sequence of each operating parameter.
4. The operation monitoring method for a heavy-duty multistage centrifugal pump as described in claim 3, characterized in that, In step S2, deviation analysis is performed on the running data using the nonlinear coupling model, including: The operating data of each centrifugal pump stage is sliced to obtain sliced operating data containing multiple slices. Each slice consists of... The sequence data vector is composed of the sequence data values of each operating parameter at the sampling time. Indicates the window size of the time-sliding window; The dynamic sensitivity weights of each centrifugal pump stage are calculated using the following expression: , ； in, Indicates the first Dynamic sensitivity weights of each centrifugal pump stage This indicates the total number of stages in a centrifugal pump. Indicates the weighting factor. Indicates the first The cosine similarity between the operating data of each centrifugal pump stage and the overall operating data, wherein the overall operating data is... The sum of the operating data of each centrifugal pump stage, Indicates the first The mean of the standard deviations of the sequence data of each operating parameter in the operating data of a centrifugal pump stage; The slice operation data is weighted based on dynamic sensitive weights to obtain a joint operation state vector; The nonlinear coupling modeling layer decomposes the input joint operating state vector into slice data sequences for each centrifugal pump stage under different operating parameters. The slice data sequences are then input into a gated recurrent unit to obtain feature codes for the slice data. An attention mechanism is used to generate attention weights for the feature codes of all slice data in the sequence. These feature codes are then attention-weighted, and a fully connected layer is used to predict the attention-weighted feature codes, resulting in a prediction sequence for each centrifugal pump stage under different operating parameters. The length of the prediction sequence is... , This indicates the number of slices in the slice execution data. Indicates the step size of the time sliding window; Calculate the mean value of each slice in the sliced operation data under different operating parameters, and sort all the means according to the slice order to obtain the mean value sequence of different operating parameters in the centrifugal pump stage. Calculate the sequence difference between the mean value sequence and the predicted sequence as the residual sequence of the centrifugal pump stage under different operating parameters.
5. The operation monitoring method for a heavy-duty multistage centrifugal pump as described in claim 4, characterized in that, In step S2, sensitive operating parameters that significantly deviate from normal coupling behavior are extracted based on the residual deviation score, including: The residual deviation score for different operating parameters in each centrifugal pump stage is calculated using the following expression: , , ； in, Indicates the first The r-th operating parameter in a centrifugal pump stage The residual deviation score, Indicates running parameters The residual sequence, Representing the residual sequence The j-th sequence value in the sequence, Indicates the preset first The standard deviation of the residual sequence of the r-th operating parameter in a centrifugal pump stage, where R represents the number of types of operating parameters; Set a deviation threshold, and use the operating parameters whose residual deviation scores are higher than the deviation threshold as sensitive operating parameters that deviate significantly from the normal coupling behavior under this centrifugal pump stage.
6. The method for monitoring the operation of a heavy-duty multistage centrifugal pump as described in claim 5, characterized in that, Step S3 includes: The trend score for each sensitive operating parameter is calculated using the following expression: ； in, Indicates sensitive operating parameters Trend ratings Indicates sensitive operating parameters residual sequence The j-th sequence value in the sequence, Indicates sensitive operating parameters residual sequence The mean; The coupling mutation index of each sensitive operating parameter is calculated as follows: ； in, Indicates sensitive operating parameters The coupling mutation index, Indicates sensitive operating parameters The attention weight of the j-th slice in the corresponding slice data sequence; The residual deviation score, trend score, and coupling mutation index of each sensitive operating parameter are weighted to obtain the disturbance intensity score of the corresponding sensitive operating parameter. The top G sensitive operating parameters with the highest disturbance intensity scores are selected from the set of sensitive operating parameters as the disturbance operating parameters.
7. The method for monitoring the operation of a heavy-duty multistage centrifugal pump as described in claim 6, characterized in that, Step S4 includes: The adaptive dynamic thresholds for each disturbance operating parameter are calculated using the following expression: ； ； ； in, Indicates the first Disturbance operating parameters Adaptive dynamic threshold, Indicates the first Disturbance operating parameters The mean baseline threshold, Indicates the first Disturbance operating parameters The residual deviation threshold, Indicates the first The mean of the residual sequence of the perturbation operating parameters. Indicates the first The standard deviation of the residual sequence of the perturbation operating parameters. Indicates the first The j-th sequence value of the residual sequence of the perturbation operating parameters. Both represent adjustment factors; A multi-factor fusion early warning scoring function is constructed, which takes the adaptive dynamic threshold of the disturbance operation parameters of each centrifugal pump stage and the residual deviation score as input, and the early warning score of each centrifugal pump stage as output. If the early warning score is higher than the preset early warning threshold, it indicates that there is a potential operation failure of the centrifugal pump stage, and an early warning notification is issued.
8. The method for monitoring the operation of a heavy-duty multistage centrifugal pump as described in claim 7, characterized in that, In the multi-factor fusion early warning scoring function, the expression for the early warning score is as follows: , ； in, Indicates the first Early warning score for a centrifugal pump level. Indicates the first A set of disturbance operating parameters for a centrifugal pump stage. Represents the set of disturbance operating parameters Any disturbance operating parameter in, Indicates disturbance operating parameters The residual deviation score, Indicates disturbance operating parameters Adaptive dynamic threshold, Denotes the discriminant function, if If the result is positive, the value of the discriminant function is 1; otherwise, the value of the discriminant function is 0.
9. A monitoring system for the operation of a heavy-duty multistage centrifugal pump, characterized in that, include: Data acquisition module: It adopts timestamp alignment and uniformly adjusts the sampling frequency to the target frequency to synchronously acquire the operating data of each stage of the multi-stage centrifugal pump; Deviation Analysis Module: Constructs a nonlinear coupling model between operating parameters at all levels, uses the nonlinear coupling model to perform deviation analysis on the operating data, obtains the residual sequence of operating parameters at all levels, and extracts sensitive operating parameters that deviate significantly from normal coupling behavior based on the residual deviation score, thus forming a set of sensitive operating parameters; Disturbance location module: Performs trend analysis and coupled mutation analysis on sensitive operating parameters, and combines the results of deviation analysis to extract the disturbance operating parameters as disturbance sources from the set of sensitive operating parameters, thus forming a set of disturbance operating parameters; Monitoring and early warning module: Based on the adaptive dynamic threshold of disturbance operating parameters and residual deviation score, a multi-factor fusion early warning scoring function is constructed, and the multi-factor fusion early warning scoring function is used to monitor and warn of potential operating faults of multi-stage centrifugal pumps; To achieve the operation monitoring method for heavy-duty multistage centrifugal pumps as described in any one of claims 1-8.