Centrifugal pump fault diagnosis system and method fusing multi-source heterogeneous data

By integrating multi-source heterogeneous data into a centrifugal pump fault diagnosis method, and utilizing environmental clustering and time-weighted prediction models, the accuracy problem of centrifugal pump fault prediction in existing technologies has been solved, achieving more accurate fault prediction.

CN118934658AActive Publication Date: 2024-11-12YANCHENG SHURONGZHISHENG TECH CO LTD

Patent Information

Application Number
CN202411235926.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-11-12
Estimated Expiration
2044-09-04

AI Technical Summary

Technical Problem

Existing technologies for centrifugal pump fault prediction rely on human experience or single operating parameters, lacking systematicity and accuracy, and failing to consider the impact of environmental factors, resulting in unstable prediction performance.

Method used

By collecting multi-source heterogeneous data, performing environmental clustering and constructing input feature vectors, training a time-weighted prediction model, and integrating environmental impact and operational health status, the failure time can be predicted.

Benefits of technology

It improves the accuracy of centrifugal pump failure prediction, comprehensively assesses the internal and external causes of failure, and takes into account the influence of environment and operating parameters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118934658A_ABST
    Figure CN118934658A_ABST
Patent Text Reader

Abstract

The invention discloses a centrifugal pump fault diagnosis system and method fused with multi-source heterogeneous data, and relates to the technical field of centrifugal pump fault diagnos.K environment clusters are obtained for each fault type of each application scene, and standard fault time and a standard operation parameter sequence are collected in each application scene; constructing an input feature vector and a time weight label, training a time weight prediction model, collecting an actual application scene, actual environment data and actual operation parameters of the centrifugal pump to be monitored in real time, and calculating a corresponding environment cluster as a target cluster based on the actual application scene and the actual environment data; based on the standard fault time and the standard operation parameter sequence of the target cluster, the time weight prediction model, the actual environment data and the actual operation parameters, obtaining fault time predicted by various fault types; and the accuracy of fault prediction and diagnosis of the centrifugal pump is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of centrifugal pump fault diagnosis, and in particular to a centrifugal pump fault diagnosis system and method that integrates multi-source heterogeneous data. Background Art

[0002] Centrifugal pumps are a type of fluid conveying equipment widely used in industrial production, playing an important role in many fields such as water supply, HVAC, chemical industry, power generation, etc. However, during long-term operation, centrifugal pumps may fail due to various reasons, such as bearing wear, impeller corrosion, shaft seal leakage, motor failure, blockage and scaling, etc. These failures not only affect the normal operation of centrifugal pumps, but may also cause equipment damage, energy waste, and even safety accidents. Therefore, it is very important to predict the failure and life of centrifugal pumps.

[0003] At present, centrifugal pump fault prediction mainly relies on manual experience or data-driven models based on a single operating parameter (such as vibration). Manual experience methods are too dependent on manual work and lack systematicity and accuracy. Although data-driven models use monitoring data, they ignore environmental impacts and physical mechanisms, resulting in uneven prediction performance and making it difficult to fundamentally solve the problem.

[0004] A Chinese patent application with publication number CN117846979A discloses a method and device for determining a centrifugal pump fault. The method comprises: obtaining real-time operation data of a centrifugal pump; extracting time-frequency domain features of the real-time operation data to obtain at least one feature matrix; inputting the feature matrix into a pre-trained centrifugal pump fault diagnosis model and / or a centrifugal pump operation status clustering model for processing to obtain at least one predictive diagnosis result; the centrifugal pump fault diagnosis model is trained based on the fusion result after fusion of multi-source historical information; and determining the fault type of the centrifugal pump based on the at least one predictive diagnosis result. However, this method fails to take into account the influence of environmental factors in the application scenario.

[0005] To this end, the present invention proposes a centrifugal pump fault diagnosis system and method that integrates multi-source heterogeneous data. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a centrifugal pump fault diagnosis system and method integrating multi-source heterogeneous data, which improves the accuracy of centrifugal pump fault prediction and diagnosis.

[0007] To achieve the above objectives, a centrifugal pump fault diagnosis method integrating multi-source heterogeneous data is proposed, which includes the following steps:

[0008] Step 1: Collect a set of application scenarios, and in each application scenario, use N test centrifugal pumps to perform a running experiment; N is the number of preset test centrifugal pumps;

[0009] Step 2: Collect the historical operating parameter sequence, historical environmental data, historical fault types and historical fault times of the centrifugal pump itself for each running experiment test in each application scenario;

[0010] Step 3: For each fault type in each application scenario, cluster the historical environment data to obtain K environment clusters; K is the number of selected environment clusters;

[0011] Step 4: Calculate the environmental data corresponding to the center point of each environmental cluster in each application scenario; and in each application scenario, select K new test centrifugal pumps to conduct new operation experiments, and collect standard failure time and standard operation parameter sequences;

[0012] Step 5: For each environmental cluster, construct an input feature vector based on historical environmental data, historical operating parameter sequence and standard operating parameter sequence, and construct a time weight label based on historical failure time and standard failure time;

[0013] Step 6: Take the input feature vector as input and the time weight label as output to train the time weight prediction model;

[0014] Step 7: Collect the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculate the corresponding environmental cluster as the target cluster based on the actual application scenarios and actual environmental data; obtain the predicted failure time of various fault types based on the standard failure time and standard operating parameter sequence of the target cluster, the time weight prediction model, the actual environmental data and the actual operating parameters;

[0015] The execution mode of the running experiment is:

[0016] In each application scenario, different environmental parameters are pre-set, and each test centrifugal pump is run until a failure of any type occurs in the test centrifugal pump. The operation process of each test centrifugal pump is regarded as one operation experiment;

[0017] The collection of the historical operating parameter sequence, historical environmental data, historical fault type and historical fault time of each running experiment in each application scenario includes the following steps:

[0018] Step 11: Collect the average values ​​of various environmental parameters of each running experiment during the entire running process, and normalize each average value. The normalized average values ​​constitute historical environmental data;

[0019] Step 12: During each operation experiment, collect data per unit time and test various operating parameters of the centrifugal pump corresponding to various fault types;

[0020] Step 13: For each running experiment, arrange each running parameter in time series order to obtain the historical running parameter sequence of the running experiment;

[0021] Step 14: Collect the fault type that occurs at the end of each running experiment as the historical fault type, and the time when the fault occurs as the historical fault time;

[0022] For each fault type in each application scenario, the historical environment data is clustered to obtain K environment clusters as follows:

[0023] Step 21: Group the running experiments of each application scenario according to the fault type to obtain M experimental groups, and the running experiments in each experimental group belong to the same fault type; M is the number of all fault types;

[0024] Step 22: taking the vector composed of each group of historical environmental data of the running experiment in each experimental group as a C-dimensional discrete point; wherein C is the number of environmental parameters in the historical environmental data, and the coordinate of the C-dimensional discrete point in each dimension corresponds to an environmental parameter in the historical environmental data;

[0025] Step 23: Randomly select K C-dimensional discrete points as initial cluster centers, and the remaining C-dimensional discrete points as non-initial cluster centers;

[0026] Step 24: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each C-dimensional discrete point into the cluster cluster where the nearest initial cluster center is located;

[0027] Step 25: Calculate the average value of the environmental parameters corresponding to each dimension of the C-dimensional discrete points in each cluster, form a new C-dimensional discrete point with the average value of the environmental parameters corresponding to each dimension, and use the new C-dimensional discrete point as the new cluster center of the cluster;

[0028] Step 26: Recalculate the Euclidean distance from each C-dimensional discrete point to each new cluster center, and re-divide each C-dimensional discrete point into the cluster where the nearest new cluster center is located;

[0029] Step 27: Repeat steps 25-26 until the C-dimensional discrete points in all divided clusters no longer change, mark the number of each cluster as k, k = 1, 2, 3, ..., K; obtain all C-dimensional discrete points contained in each cluster, and form all the C-dimensional discrete points into an environmental cluster;

[0030] The method of calculating the environmental data corresponding to the center point position of each environmental cluster in each application scenario, and selecting K new test centrifugal pumps for operation experiments in each application scenario, and collecting the standard failure time and standard operation parameter sequence is as follows:

[0031] Obtain the C-dimensional discrete point corresponding to the center point of each environmental cluster as the cluster center point;

[0032] Obtain each environmental parameter value in the C-dimensional discrete point corresponding to the cluster center point;

[0033] The way to select K new test centrifugal pumps for new running experiments is:

[0034] For each environment cluster in each application scenario:

[0035] Select a new test centrifugal pump and set its environmental parameters to those corresponding to the cluster center point;

[0036] Under the environmental data conditions, the new test centrifugal pump is operated until a corresponding fault type occurs in the new test centrifugal pump and the pump stops; the operation process of the new test centrifugal pump is regarded as a new operation experiment;

[0037] The failure time of each new running experiment is collected as the standard failure time, and the values ​​of various operating parameters of the newly tested centrifugal pump at each unit time are collected in chronological order to form a standard operating parameter sequence;

[0038] For each environmental cluster, the input feature vector is constructed based on the historical environmental data, the historical operating parameter sequence and the standard operating parameter sequence as follows:

[0039] For each run in each environment cluster:

[0040] Calculate the ratio of each environmental parameter in the historical environmental data of the running experiment to each environmental parameter corresponding to the center point of the environmental cluster to form an environmental ratio vector;

[0041] Obtaining each operating parameter in the historical operating parameter sequence and the standard operating parameter sequence at each unit time, respectively forming a historical parameter vector and a standard parameter vector, and calculating the ratio of each operating parameter in the historical parameter vector and the standard parameter vector to form an operating ratio vector;

[0042] The input feature vector is composed of the environment ratio vector and the operation ratio vector;

[0043] The method of constructing a time weight label based on the historical failure time and the standard failure time is:

[0044] For each environment cluster:

[0045] For each unit time of each running experiment, calculate the time from the unit time to the occurrence of the running experiment failure as the failure time difference;

[0046] For each unit time of the new running experiment, calculate the time from the unit time to the occurrence of the new running experiment failure as the standard failure time difference;

[0047] The ratio of the fault time difference to the standard fault time difference is used as the time weight label of the unit time;

[0048] The method of training the time weight prediction model is:

[0049] Each group of input feature vectors is used as the input of a time weight prediction model, and the time weight prediction model takes the predicted value of the fault time weight of each unit time as the output; the time weight prediction model takes the time weight label corresponding to each unit time as the prediction target, takes the difference between the predicted value of the fault time weight and the time weight label as the prediction error, and takes minimizing the sum of squares of the prediction errors of all unit times as the training target; the time weight prediction model is trained until the sum of squares of the prediction errors reaches convergence, and the training is stopped; the time weight prediction model is any one of the regression models;

[0050] The method for obtaining the predicted failure time of various failure types is as follows:

[0051] Calculate the ratio of each environmental parameter in the actual environmental data to each environmental parameter corresponding to the target cluster center point to form an actual environmental ratio vector;

[0052] Calculate the ratio of each operating parameter in the actual operating parameters to each operating parameter at the corresponding unit time in the standard operating parameter sequence of the target cluster to form an actual operating ratio vector;

[0053] The actual environment ratio vector and the actual operation ratio vector constitute an actual input vector, and the actual input vector is input into the time weight prediction model to obtain a predicted value of the predicted failure time weight;

[0054] The predicted value of the failure time weight is multiplied by the product of the unit time and the standard failure time difference to obtain the predicted time of the failure of this fault type.

[0055] A centrifugal pump fault diagnosis system integrating multi-source heterogeneous data is proposed, which includes a sample collection module, a clustering module, a model training module and a model application module; wherein each module is electrically connected;

[0056] The sample collection module collects a set of application scenarios. In each application scenario, N test centrifugal pumps are used to perform operation experiments. The historical operation parameter sequence, historical environmental data, historical fault type and historical fault time of each operation experiment in each application scenario are collected, and the historical operation parameter sequence and historical environmental data are sent to the clustering module, and the historical environmental data, historical fault type and historical fault time are sent to the model training module.

[0057] The clustering module clusters the historical environmental data for each fault type in each application scenario to obtain K environmental clusters, where K is the number of selected environmental clusters, and sends the environmental clusters to the model training module and the model application module;

[0058] The model training module calculates the environmental data corresponding to the center point position of each environmental cluster in each application scenario; and in each application scenario, selects K new test centrifugal pumps to conduct new operation experiments, collects standard failure time and standard operation parameter sequence, and constructs input feature vectors for each environmental cluster based on historical environmental data, historical operation parameter sequence and standard operation parameter sequence, and constructs time weight labels based on historical failure time and standard failure time. The input feature vector is used as input and the time weight label is used as output to train the time weight prediction model; and the standard failure time, standard operation parameter sequence and time weight prediction model are sent to the model application module;

[0059] The model application module collects the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculates the corresponding environmental clustering cluster as the target clustering cluster based on the actual application scenarios and actual environmental data; based on the standard failure time and standard operating parameter sequence of the target clustering cluster, the time weight prediction model, the actual environmental data, and the actual operating parameters, the predicted failure time of various fault types is obtained.

[0060] An electronic device is proposed, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0061] The processor executes the above-mentioned centrifugal pump fault diagnosis method integrating multi-source heterogeneous data by calling the computer program stored in the memory.

[0062] A computer-readable storage medium is provided, on which a rewritable computer program is stored;

[0063] When the computer program is executed on a computer device, the computer device is enabled to execute the above-mentioned centrifugal pump fault diagnosis method integrating multi-source heterogeneous data.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The present invention collects a set of application scenarios, and in each application scenario, uses N test centrifugal pumps to perform an operation experiment; N is the number of preset test centrifugal pumps, and collects historical operation parameter sequences, historical environmental data, historical fault types, and historical fault times of each operation experiment in each application scenario. For each fault type in each application scenario, the historical environmental data are clustered to obtain K environmental clustering clusters; K is the number of selected environmental clustering clusters, and the environmental data corresponding to the center point position of each environmental clustering cluster in each application scenario is calculated; and in each application scenario, K new test centrifugal pumps are selected to perform a new operation experiment, and standard fault times and standard operation parameter sequences are collected. For each environmental clustering cluster, an input feature vector is constructed based on the historical environmental data, the historical operation parameter sequence, and the standard operation parameter sequence. The time weight label is constructed based on the failure time and the standard failure time, the input feature vector is used as input and the time weight label is used as output, the time weight prediction model is trained, the actual application scenario, actual environmental data and actual operating parameters of the centrifugal pump to be monitored are collected in real time, and the corresponding environmental clustering cluster is calculated as the target clustering cluster based on the actual application scenario and the actual environmental data; the predicted failure time of various fault types is obtained based on the standard failure time and standard operating parameter sequence of the target clustering cluster, the time weight prediction model, the actual environmental data and the actual operating parameters; the present invention comprehensively considers the environmental parameters and the operating parameters of the centrifugal pump by integrating the environmental impact and the operating health status, uses the environmental clustering analysis to analyze the impact under different working conditions, and integrates the operating parameter sequence at the same time to comprehensively evaluate the internal and external causes of the failure, thereby improving the accuracy of the centrifugal pump fault prediction diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of a centrifugal pump fault diagnosis method integrating multi-source heterogeneous data in Example 1 of the present invention;

[0067] Figure 2 A module connection relationship diagram of a centrifugal pump fault diagnosis system integrating multi-source heterogeneous data in Example 2 of the present invention;

[0068] Figure 3 This is a schematic diagram of the structure of an electronic device in Embodiment 3 of the present invention;

[0069] Figure 4 This is a schematic diagram of the computer-readable storage medium structure in Example 4 of the present invention. DETAILED DESCRIPTION

[0070] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1 As shown, a centrifugal pump fault diagnosis method integrating multi-source heterogeneous data includes the following steps:

[0073] Step 1: Collect a set of application scenarios, and in each application scenario, use N test centrifugal pumps to perform a running experiment; N is the number of preset test centrifugal pumps;

[0074] Step 2: Collect the historical operating parameter sequence, historical environmental data, historical fault types and historical fault times of the centrifugal pump itself for each running experiment test in each application scenario;

[0075] Step 3: For each fault type in each application scenario, cluster the historical environment data to obtain K environment clusters; K is the number of selected environment clusters;

[0076] Step 4: Calculate the environmental data corresponding to the center point of each environmental cluster in each application scenario; and in each application scenario, select K new test centrifugal pumps to conduct new operation experiments, and collect standard failure time and standard operation parameter sequences;

[0077] Step 5: For each environmental cluster, construct an input feature vector based on historical environmental data, historical operating parameter sequence and standard operating parameter sequence, and construct a time weight label based on historical failure time and standard failure time;

[0078] Step 6: Take the input feature vector as input and the time weight label as output to train the time weight prediction model;

[0079] Step 7: Collect the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculate the corresponding environmental cluster as the target cluster based on the actual application scenarios and actual environmental data; obtain the predicted failure time of various fault types based on the standard failure time and standard operating parameter sequence of the target cluster, the time weight prediction model, the actual environmental data and the actual operating parameters;

[0080] The application scenario set includes several main application scenarios of centrifugal pumps, including but not limited to water supply systems, HVAC systems, fire protection systems, chemical processes, mining, sewage treatment, and power plants, etc.; it is understandable that in different application scenarios, the functions and environmental conditions of centrifugal pumps are different, and different working environments have different wear and tear on different parts of centrifugal pumps, so in different application scenarios, there are differences in the possible failures that may occur;

[0081] The execution mode of the running experiment is:

[0082] In each application scenario, different environmental parameters are pre-set, and each test centrifugal pump is run until a fault of any type occurs in the test centrifugal pump. The operation process of each test centrifugal pump is regarded as an operation experiment; the environmental parameters include but are not limited to temperature, humidity, corrosiveness, medium composition, pH value, viscosity, solid content concentration, etc.; the fault types include but are not limited to bearing fault, impeller fault, shaft seal fault, motor fault, blockage and scaling, etc.;

[0083] The collection of the historical operating parameter sequence, historical environmental data, historical fault type and historical fault time of the centrifugal pump itself in each application scenario and each running experiment test includes the following steps:

[0084] Step 11: Collect the average values ​​of various environmental parameters of each running experiment during the entire running process, and normalize each average value. The normalized average values ​​constitute historical environmental data;

[0085] Step 12: During each operation experiment, collect data per unit time and test various operating parameters of the centrifugal pump corresponding to various fault types;

[0086] It is understandable that for different fault types, the operating parameters of the centrifugal pump that need to be collected are different. Specifically, the corresponding relationship between the operating parameters corresponding to the various fault types can be as follows:

[0087] Bearing failure (wear, overheating, etc.): vibration data (axial, radial), temperature data (bearing temperature), speed data;

[0088] Impeller failure (wear, corrosion, blockage, etc.): vibration data, pressure data (inlet / outlet pressure), flow data, torque data;

[0089] Shaft seal failure (leakage, wear, etc.): pressure data, temperature data (shaft seal temperature), vibration data;

[0090] Motor failure (winding open circuit, overload, etc.): current data, voltage data, power data, temperature rise data;

[0091] Faults such as blockage and scaling: pressure data, temperature rise data, vibration data, flow data;

[0092] Step 13: For each running experiment, each running parameter is arranged in time series order to obtain a historical running parameter sequence of the centrifugal pump itself tested in the running experiment;

[0093] Step 14: Collect the fault type that occurs at the end of each running experiment as the historical fault type, and the time when the fault occurs as the historical fault time;

[0094] Furthermore, for each fault type in each application scenario, the historical environment data is clustered to obtain K environment clusters in the following manner:

[0095] Step 21: Group the running experiments of each application scenario according to the fault type to obtain M experimental groups, and the running experiments in each experimental group belong to the same fault type; M is the number of all fault types;

[0096] Step 22: taking the vector composed of each group of historical environmental data of the running experiment in each experimental group as a C-dimensional discrete point; wherein C is the number of environmental parameters in the historical environmental data, and the coordinate of the C-dimensional discrete point in each dimension corresponds to an environmental parameter in the historical environmental data;

[0097] For example, a vector of a set of historical environmental data corresponding to an operation experiment in a water supply system may be (0.5, 0.625, 0.25, 0.6, 0.692), where each value in the vector corresponds to the normalized results of medium temperature, pressure, pH value, conductivity, and ambient temperature.

[0098] Step 23: Randomly select K C-dimensional discrete points as initial cluster centers, and the remaining C-dimensional discrete points as non-initial cluster centers;

[0099] Step 24: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each C-dimensional discrete point into the cluster cluster where the nearest initial cluster center is located;

[0100] Step 25: Calculate the average value of the environmental parameters corresponding to each dimension of the C-dimensional discrete points in each cluster, form a new C-dimensional discrete point with the average value of the environmental parameters corresponding to each dimension, and use the new C-dimensional discrete point as the new cluster center of the cluster;

[0101] Step 26: Recalculate the Euclidean distance from each C-dimensional discrete point to each new cluster center, and re-divide each C-dimensional discrete point into the cluster where the nearest new cluster center is located;

[0102] Step 27: Repeat steps 25-26 until the C-dimensional discrete points in all divided clusters no longer change, mark the number of each cluster as k, k = 1, 2, 3, ..., K; obtain all C-dimensional discrete points contained in each cluster, and form all the C-dimensional discrete points into an environmental cluster;

[0103] Furthermore, the environmental data corresponding to the center point position of each environmental cluster in each application scenario is calculated, and in each application scenario, K new test centrifugal pumps are selected for operation experiments, and the standard failure time and standard operation parameter sequence are collected in the following manner:

[0104] Obtain the C-dimensional discrete point corresponding to the center point of each environmental cluster as the cluster center point;

[0105] Obtain each environmental parameter value in the C-dimensional discrete point corresponding to the cluster center point;

[0106] The way to select K new test centrifugal pumps for new running experiments is:

[0107] For each environment cluster in each application scenario:

[0108] Select a new test centrifugal pump and set its environmental parameters to those corresponding to the cluster center point;

[0109] Under the environmental data conditions, the new test centrifugal pump is operated until a corresponding fault type occurs in the new test centrifugal pump and the pump stops; the operation process of the new test centrifugal pump is regarded as a new operation experiment;

[0110] The failure time of each new running experiment is collected as the standard failure time, and the values ​​of various operating parameters of the newly tested centrifugal pump at each unit time are collected in chronological order to form a standard operating parameter sequence;

[0111] It can be understood that the standard failure time and the standard operating parameter sequence can be regarded as a standard comparison experiment performed in the corresponding environment cluster, and the rest of the operating experiments can use this operating experiment as a reference;

[0112] Furthermore, for each environmental cluster, based on the historical environmental data, the historical operating parameter sequence and the standard operating parameter sequence, the input feature vector is constructed in the following manner:

[0113] For each run in each environment cluster:

[0114] Calculate the ratio of each environmental parameter in the historical environmental data of the running experiment to each environmental parameter corresponding to the center point of the environmental cluster to form an environmental ratio vector; it can be understood that the environmental ratio vector measures the environmental difference between each running experiment and the standard running experiment, so as to achieve the purpose of evaluating the impact of environmental factors on failure time;

[0115] Obtain each operating parameter in the historical operating parameter sequence and the standard operating parameter sequence at each unit time, respectively form a historical parameter vector and a standard parameter vector, and calculate the ratio of each operating parameter in the historical parameter vector and the standard parameter vector to form an operating ratio vector; it can be understood that the operating ratio vector measures the difference in each operating health state between the test centrifugal pump and the new test centrifugal pump at each unit time, and obviously the greater the difference, the greater the impact on the failure time;

[0116] The input feature vector is composed of an environmental ratio vector and an operating ratio vector. It can be understood that the input feature vector includes both the influence of environmental parameters on the prediction of fault time and the influence of the real-time health status of the centrifugal pump on the prediction of fault time.

[0117] Furthermore, the method of constructing a time weight label based on the historical failure time and the standard failure time is:

[0118] For each environment cluster:

[0119] For each unit time of each running experiment, calculate the time from the unit time to the occurrence of the running experiment failure as the failure time difference;

[0120] For each unit time of the new running experiment, calculate the time from the unit time to the occurrence of the new running experiment failure as the standard failure time difference;

[0121] The ratio of the fault time difference to the standard fault time difference is used as the time weight label of the unit time;

[0122] It can be understood that the time weight label measures the difference in failure time results caused by the difference in environmental parameters and the difference in the health status of the centrifugal pump;

[0123] Furthermore, the method of training the time weight prediction model with the input feature vector as input and the time weight label as output is:

[0124] Each group of input feature vectors is used as the input of a time weight prediction model, and the time weight prediction model takes the predicted value of the fault time weight of each unit time as the output; the time weight prediction model takes the time weight label corresponding to each unit time as the prediction target, takes the difference between the predicted value of the fault time weight and the time weight label as the prediction error, and takes minimizing the sum of squares of the prediction errors of all unit times as the training target; the time weight prediction model is trained until the sum of squares of the prediction errors reaches convergence, and the training is stopped; the time weight prediction model is any one of the regression models, and the regression model includes but is not limited to the SVR model or the polynomial regression model, etc.;

[0125] Furthermore, the actual application scenario is the scenario in which the centrifugal pump to be monitored is actually applied, the actual environmental data is a vector composed of various environmental parameters of the environment in which the centrifugal pump to be monitored is located, and the actual operating parameters include operating parameters corresponding to various fault types of the centrifugal pump to be monitored;

[0126] Furthermore, the method of calculating the corresponding environment cluster as the target cluster based on the actual application scenario and the actual environment data is as follows:

[0127] For each fault type:

[0128] Calculate the Euclidean distance between the C-dimensional discrete points corresponding to the actual environmental data and the centers of each environmental cluster, and take the environmental cluster with the closest Euclidean distance as the target cluster;

[0129] The method of obtaining the predicted failure time of various failure types based on the standard failure time and standard operating parameter sequence of the target cluster, the time weight prediction model, the actual environment data, and the actual operating parameters is as follows:

[0130] Calculate the ratio of each environmental parameter in the actual environmental data to each environmental parameter corresponding to the target cluster center point to form an actual environmental ratio vector;

[0131] Calculate the ratio of each operating parameter in the actual operating parameters to each operating parameter at the corresponding unit time in the standard operating parameter sequence of the target cluster to form an actual operating ratio vector;

[0132] The actual environment ratio vector and the actual operation ratio vector constitute an actual input vector, and the actual input vector is input into the time weight prediction model to obtain a predicted value of the predicted failure time weight;

[0133] The predicted value of the failure time weight is multiplied by the product of the unit time and the standard failure time difference to obtain the predicted time of the failure of this fault type.

[0134] Example 2

[0135] like Figure 2 As shown, a centrifugal pump fault diagnosis system integrating multi-source heterogeneous data includes a sample collection module, a clustering module, a model training module and a model application module; wherein each module is electrically connected;

[0136] The sample collection module collects a set of application scenarios. In each application scenario, N test centrifugal pumps are used to perform operation experiments. The historical operation parameter sequence, historical environmental data, historical fault type and historical fault time of the centrifugal pump itself are collected in each application scenario and each operation experiment. The historical operation parameter sequence and historical environmental data are sent to the clustering module, and the historical environmental data, historical fault type and historical fault time are sent to the model training module.

[0137] The clustering module clusters the historical environmental data for each fault type in each application scenario to obtain K environmental clusters, where K is the number of selected environmental clusters, and sends the environmental clusters to the model training module and the model application module;

[0138] The model training module calculates the environmental data corresponding to the center point position of each environmental cluster in each application scenario; and in each application scenario, selects K new test centrifugal pumps to conduct new operation experiments, collects standard failure time and standard operation parameter sequence, and constructs input feature vectors for each environmental cluster based on historical environmental data, historical operation parameter sequence and standard operation parameter sequence, and constructs time weight labels based on historical failure time and standard failure time. The input feature vector is used as input and the time weight label is used as output to train the time weight prediction model; and the standard failure time, standard operation parameter sequence and time weight prediction model are sent to the model application module;

[0139] The model application module collects the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculates the corresponding environmental clustering cluster as the target clustering cluster based on the actual application scenarios and actual environmental data; based on the standard failure time and standard operating parameter sequence of the target clustering cluster, the time weight prediction model, the actual environmental data, and the actual operating parameters, the predicted failure time of various fault types is obtained.

[0140] Example 3

[0141] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3As shown, the present application also provides an electronic device 100. The electronic device 100 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, a centrifugal pump fault diagnosis method integrating multi-source heterogeneous data as described above may be executed.

[0142] The method or device according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3 As shown, the electronic device 100 may include a bus 101, one or more CPUs 102, a ROM 103, a RAM 104, a communication port 105 connected to a network, an input / output component 106, a hard disk 107, etc. A storage device in the electronic device 100, such as the ROM 103 or the hard disk 107, may store a centrifugal pump fault diagnosis method for integrating multi-source heterogeneous data provided in the present application.

[0143] Furthermore, the electronic device 100 may further include a user interface 108. Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0144] Example 4

[0145] Figure 4 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Figure 4 As shown, it is a computer-readable storage medium 200 disclosed in the present application. Computer-readable instructions are stored on the computer-readable storage medium 200. When the computer-readable instructions are executed by the processor, a centrifugal pump fault diagnosis method for integrating multi-source heterogeneous data according to an embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0146] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, and when the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0147] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0148] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0149] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0150] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0151] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data, characterized in that: The following steps are involved: Step 1: Collect a set of application scenarios, and in each application scenario, use N test centrifugal pumps to perform a running experiment; N is the number of preset test centrifugal pumps; Step 2: Collect the historical operating parameter sequence, historical environmental data, historical fault types and historical fault times of the centrifugal pump itself for each running experiment test in each application scenario; Step 3: For each fault type in each application scenario, cluster the historical environment data to obtain K environment clusters; K is the number of selected environment clusters; Step 4: Calculate the environmental data corresponding to the center point of each environmental cluster in each application scenario; and in each application scenario, select K new test centrifugal pumps to conduct new operation experiments, and collect standard failure time and standard operation parameter sequences; Step 5: For each environmental cluster, construct an input feature vector based on historical environmental data, historical operating parameter sequence and standard operating parameter sequence, and construct a time weight label based on historical failure time and standard failure time; Step 6: Take the input feature vector as input and the time weight label as output to train the time weight prediction model; Step 7: Collect the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculate the corresponding environmental clusters as the target clusters based on the actual application scenarios and actual environmental data; Based on the standard failure time and standard operating parameter sequence of the target cluster, the time weight prediction model, the actual environment data, and the actual operating parameters, the predicted failure time of various fault types is obtained.

2. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 1, characterized in that: The collection of the historical operating parameter sequence, historical environmental data, historical fault type and historical fault time of each running experiment in each application scenario includes the following steps: Step 11: Collect the average values ​​of various environmental parameters of each running experiment during the entire running process, and normalize each average value. The normalized average values ​​constitute historical environmental data; Step 12: During each operation experiment, collect data per unit time and test various operating parameters of the centrifugal pump corresponding to various fault types; Step 13: For each running experiment, arrange each running parameter in time series order to obtain the historical running parameter sequence of the running experiment; Step 14: Collect the fault type that occurs at the end of each running experiment as the historical fault type, and the time when the fault occurs as the historical fault time.

3. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 2, characterized in that: For each fault type in each application scenario, the historical environment data is clustered to obtain K environment clusters as follows: Step 21: Group the running experiments of each application scenario according to the fault type to obtain M experimental groups, and the running experiments in each experimental group belong to the same fault type; M is the number of all fault types; Step 22: taking the vector composed of each group of historical environmental data of the running experiment in each experimental group as a C-dimensional discrete point; wherein C is the number of environmental parameters in the historical environmental data, and the coordinate of the C-dimensional discrete point in each dimension corresponds to an environmental parameter in the historical environmental data; Step 23: Randomly select K C-dimensional discrete points as initial cluster centers, and the remaining C-dimensional discrete points as non-initial cluster centers; Step 24: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each C-dimensional discrete point into the cluster cluster where the nearest initial cluster center is located; Step 25: Calculate the average value of the environmental parameters corresponding to each dimension of the C-dimensional discrete points in each cluster, form a new C-dimensional discrete point with the average value of the environmental parameters corresponding to each dimension, and use the new C-dimensional discrete point as the new cluster center of the cluster; Step 26: Recalculate the Euclidean distance from each C-dimensional discrete point to each new cluster center, and re-divide each C-dimensional discrete point into the cluster where the nearest new cluster center is located; Step 27: Repeat steps 25-26 until the C-dimensional discrete points in all divided clusters no longer change, mark the number of each cluster as k, k = 1, 2, 3, ..., K; obtain all C-dimensional discrete points contained in each cluster, and form an environmental cluster with all the C-dimensional discrete points.

4. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 3, characterized in that: The method of calculating the environmental data corresponding to the center point position of each environmental cluster in each application scenario, and selecting K new test centrifugal pumps for operation experiments in each application scenario, and collecting the standard failure time and standard operation parameter sequence is as follows: Obtain the C-dimensional discrete point corresponding to the center point of each environmental cluster as the cluster center point; Obtain each environmental parameter value in the C-dimensional discrete point corresponding to the cluster center point; The way to select K new test centrifugal pumps for new running experiments is: For each environment cluster in each application scenario: Select a new test centrifugal pump and set its environmental parameters to those corresponding to the cluster center point; Under the environmental data conditions, the new test centrifugal pump is operated until a corresponding fault type occurs in the new test centrifugal pump and the pump stops; the operation process of the new test centrifugal pump is regarded as a new operation experiment; The failure time of each new running experiment is collected as the standard failure time, and the values ​​of various operating parameters of the newly tested centrifugal pump per unit time are collected in chronological order to form a standard operating parameter sequence.

5. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 4, characterized in that: For each environmental cluster, the input feature vector is constructed based on the historical environmental data, the historical operating parameter sequence and the standard operating parameter sequence as follows: For each run in each environment cluster: Calculate the ratio of each environmental parameter in the historical environmental data of the running experiment to each environmental parameter corresponding to the center point of the environmental cluster to form an environmental ratio vector; Obtaining each operating parameter in the historical operating parameter sequence and the standard operating parameter sequence at each unit time, respectively forming a historical parameter vector and a standard parameter vector, and calculating the ratio of each operating parameter in the historical parameter vector and the standard parameter vector to form an operating ratio vector; The input feature vector is composed of the environment ratio vector and the operation ratio vector.

6. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 5, characterized in that: The method of constructing a time weight label based on the historical failure time and the standard failure time is: For each environment cluster: For each unit time of each running experiment, calculate the time from the unit time to the occurrence of the running experiment failure as the failure time difference; For each unit time of the new running experiment, calculate the time from the unit time to the occurrence of the new running experiment failure as the standard failure time difference; The ratio of the fault time difference to the standard fault time difference is used as the time weight label of the unit moment.

7. A centrifugal pump fault diagnosis method integrating multi-source heterogeneous data according to claim 6, characterized in that: The method for obtaining the predicted failure time of various failure types is as follows: Calculate the ratio of each environmental parameter in the actual environmental data to each environmental parameter corresponding to the target cluster center point to form an actual environmental ratio vector; Calculate the ratio of each operating parameter in the actual operating parameters to each operating parameter at the corresponding unit time in the standard operating parameter sequence of the target cluster to form an actual operating ratio vector; The actual environment ratio vector and the actual operation ratio vector constitute an actual input vector, and the actual input vector is input into the time weight prediction model to obtain a predicted value of the predicted failure time weight; The predicted value of the failure time weight is multiplied by the product of the unit time and the standard failure time difference to obtain the predicted time of the failure of this fault type.

8. A centrifugal pump fault diagnosis system integrating multi-source heterogeneous data, which is used to implement the centrifugal pump fault diagnosis method integrating multi-source heterogeneous data as claimed in any one of claims 1 to 7, characterized in that: It includes a sample collection module, a clustering module, a model training module and a model application module; wherein each module is electrically connected; The sample collection module collects a set of application scenarios. In each application scenario, N test centrifugal pumps are used to perform operation experiments. The historical operation parameter sequence, historical environmental data, historical fault type and historical fault time of each operation experiment in each application scenario are collected, and the historical operation parameter sequence and historical environmental data are sent to the clustering module, and the historical environmental data, historical fault type and historical fault time are sent to the model training module. The clustering module clusters the historical environmental data for each fault type in each application scenario to obtain K environmental clusters, where K is the number of selected environmental clusters, and sends the environmental clusters to the model training module and the model application module; The model training module calculates the environmental data corresponding to the center point position of each environmental cluster in each application scenario; and in each application scenario, selects K new test centrifugal pumps to conduct new operation experiments, collects standard failure time and standard operation parameter sequence, and constructs input feature vectors for each environmental cluster based on historical environmental data, historical operation parameter sequence and standard operation parameter sequence, and constructs time weight labels based on historical failure time and standard failure time. The input feature vector is used as input and the time weight label is used as output to train the time weight prediction model; and the standard failure time, standard operation parameter sequence and time weight prediction model are sent to the model application module; The model application module collects the actual application scenarios, actual environmental data and actual operating parameters of the centrifugal pump to be monitored in real time, and calculates the corresponding environmental clustering cluster as the target clustering cluster based on the actual application scenarios and actual environmental data; based on the standard failure time and standard operating parameter sequence of the target clustering cluster, the time weight prediction model, the actual environmental data, and the actual operating parameters, the predicted failure time of various fault types is obtained.

9. An electronic device, characterized in that: include: processor and memory, wherein The memory stores a computer program that can be called by the processor; The processor executes the centrifugal pump fault diagnosis method for fusing multi-source heterogeneous data as described in any one of claims 1 to 7 in the background by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: A rewritable computer program is stored thereon; When the computer program is executed on a computer device, the computer device executes the centrifugal pump fault diagnosis method integrating multi-source heterogeneous data as described in any one of claims 1 to 7 in the background.

Citation Information

Patent Citations

  • Water pump multi-fault diagnosis method based on multi-source sensor data

    CN117553009A

  • Power grid equipment fault cost analysis method

    CN117788031A

  • Centrifugal pump fault determination method and equipment

    CN117846979A

  • Hydropower station intelligent fault diagnosis device and method based on multi-source information fusion

    CN118396132A

  • System and predictive modeling method for smelting process control based on multi-source information with heterogeneous relatedness

    US20180081339A1

Cited By

  • Intelligent fire pump fault diagnosis method based on BP neural network

    CN119377865A

  • Fault comprehensive diagnosis method and system for intermittent pumping unit

    CN119442067A

  • Fault diagnosis method and device for diaphragm type urea pump, electronic equipment and medium

    CN119554120A

  • Face recognition system and method based on multispectral image fusion, electronic equipment and storage medium

    CN119810896A

  • Face Recognition System, Method, Electronic Device and Storage Medium for Multispectral Image Fusion

    CN119810896B