A motor cooling control system with a fault diagnosis function

Abstract

The present invention discloses a motor cooling control system with a fault diagnosis function, which relates to the field of motor cooling control. The system includes a water pump, a water pump controller, a pipeline, a plurality of sensors, and a fault diagnosis module. The present invention realizes early warning of faults such as pipeline blockage, pipeline leakage, water pump degradation, and water pump failure, and has an adaptive compensation function. When the performance of some components degrades, the control parameters can be automatically adjusted to maintain the best working state of the system and improve the overall reliability. To achieve the accuracy of the prediction results, the present invention performs blind source separation on the water pump vibration signal and the pipeline vibration signal based on the independent component analysis method, and performs clustering sampling on the sample data through a clustering algorithm to reduce sample redundancy. At the same time, when predicting the water pump degradation degree, the periodic change of the motor temperature is considered, which improves the accuracy of the prediction.

Description

Technical Field

[0001] The present invention relates to the field of motor cooling control, and in particular to a motor cooling control system with a fault diagnosis function. Background Art

[0002] During the operation of a motor, a large amount of heat is generated. If cooling control is not carried out in time, it will cause the motor temperature to be too high, thereby affecting its performance and lifespan. Therefore, a reasonable cooling control method is an essential part of motor design. Water cooling is one of the commonly used motor cooling methods. It is achieved by setting water cooling pipes inside or outside the motor and using circulating water flow to absorb and carry away the heat generated by the motor. The water cooling method has excellent heat dissipation effect and is suitable for high-power and high-heat-generating motors. However, the water cooling system is relatively complex and prone to failures. If the failures cannot be eliminated in advance, serious consequences such as motor overheating will occur.

[0003] In the prior art, the motor cooling system is usually inspected manually at regular intervals. This method is inefficient and cannot detect faults before or in the initial stage of their occurrence, let alone predict and handle faults. Summary of the Invention

[0004] To solve the technical problem of fault prediction in the motor cooling control system in the prior art, the present invention provides a motor cooling control system with a fault diagnosis function.

[0005] The present invention is realized through the following technical solutions:

[0006] A motor cooling control system with a fault diagnosis function includes: a water pump, a water pump controller, pipes, several vibration sensors, a flow sensor, a differential pressure sensor, a temperature sensor, and a fault diagnosis module;

[0007] The vibration sensors are arranged on the water pump and the pipes; the flow sensor and the differential pressure sensor are arranged in the pipes; the temperature sensor is arranged on the motor to be cooled; the water pump controller controls the water pump;

[0008] The fault diagnosis module realizes the fault prediction of the motor cooling control system through the following steps:

[0009] S1: Signal acquisition, obtaining the water pump current, rotation speed, vibration signal, motor temperature signal, pipe vibration, flow rate, and differential pressure;

[0010] S2: Processing the collected signals and extracting features;

[0011] S3: Classifying the state of the motor cooling control system based on the K-means clustering algorithm combined with the LSTM model;

[0012] S4: Determine the status of the motor cooling control system; if the pump performance of the current motor cooling control system degrades, but there is no pump failure, pipeline blockage, or pipeline leakage, go to step S5; otherwise, go to S6;

[0013] S5: Calculate the pump degradation degree and perform adaptive compensation;

[0014] S6: Predict the system failure probability based on the competing failure model and give an early warning.

[0015] Furthermore, the processing of the collected signals includes blind source separation of the pump vibration signal and the pipeline vibration signal based on the independent component analysis method.

[0016] Furthermore, the classification results of the status of the motor cooling control system include normal, pump performance degradation, pump failure, pipeline blockage, and pipeline leakage; among them, pump failure, pipeline blockage, and pipeline leakage are respectively divided into three levels: minor, medium, and severe according to the severity.

[0017] Furthermore, step S3 also includes:

[0018] S31: Cluster the sample data based on the K-means clustering algorithm;

[0019] S32: Determine the training and validation samples according to the clustering results;

[0020] S33: Establish an LSTM model and train it;

[0021] S34: Classify the input data based on the trained LSTM model to obtain the classification results of the status of the motor cooling control system.

[0022] Furthermore, determining the training and validation samples according to the clustering results includes performing mixed sampling on the clustering results to determine the training and validation samples, and the mixed sampling combines stratified sampling and random sampling;

[0023] Stratified sampling respectively obtains the sampling results for N clustering classification samples and discrete samples, and the expression is as follows:

[0024] ;

[0025] Among them, is the sample set of the i-th clustering center, is the extraction ratio, and L is the discrete sample set.

[0026] Furthermore, is determined by the following method:

[0027] ;

[0028] Among them, is the adjustment coefficient, is the Davies-Bouldin index of the i-th cluster, is the set threshold of the Davies-Bouldin index, is the set constant.

[0029] Furthermore, the step S5 includes:

[0030] S51: Calculate the degradation degree of the water pump;

[0031] S52: When the degradation degree of the water pump reaches a certain threshold, calculate the water pump control parameters based on the adaptive control algorithm, and send the water pump control parameters to the water pump controller to achieve adaptive adjustment control.

[0032] Furthermore, the calculation of the degradation degree of the water pump includes:

[0033] Obtain the time series of the water pump current, pressure, and flow rate; predict the time series of the water pump current, pressure, and flow rate based on the autoregressive integrated moving average model with periodic correlation; calculate the degradation degree of the water pump according to the time series prediction results.

[0034] Furthermore, the adaptive adjustment control includes:

[0035] Construct an adaptive adjustment model based on the FNN network, obtain the degradation degree of the water pump, and use it as the input parameter of the adaptive adjustment model; input the degradation degree prediction result into the adaptive adjustment model to obtain the water pump motor control parameters output by the adaptive adjustment model.

[0036] Furthermore, the step S6 includes: S61: Calculate the system reliability

[0037] ;

[0038] Among them, is the system reliability when only the water pump degradation failure exists and there is no sudden failure factor acting within time t, is the system reliability when there is a sudden failure factor acting within time t;

[0039] S62: Calculate the predicted system failure probability:

[0040] ;

[0041] Among them, is the system density function, and the expression is as follows:

[0042] .

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] The present invention realizes early warning of faults such as pipeline blockage, pipeline leakage, pump degradation, and pump failure, and has an adaptive compensation function. When the performance of some components degrades, it can automatically adjust control parameters to maintain the optimal working state of the system and improve the overall reliability. To achieve the accuracy of prediction results, the present invention performs blind source separation on the pump vibration signal and the pipeline vibration signal based on the independent component analysis method, and performs clustering sampling on the sample data through the clustering algorithm. Combining hybrid sampling ensures the comprehensiveness and accuracy of the training sample data, reduces sample redundancy. At the same time, in the prediction of pump degradation degree, an autoregressive integrated moving average model with periodic correlation is used for prediction, considering the periodic change of motor temperature, which improves the accuracy of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings:

[0046] Figure 1 is a schematic diagram of the fault prediction process of the motor cooling control system according to an embodiment of the present application;

[0047] Figure 2 is a schematic diagram of the method for classifying the state of the motor cooling control system based on the K-means clustering algorithm combined with the LSTM model according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The embodiments of the present invention will be described in detail below with reference to the drawings.

[0049] The following describes the embodiments of the present invention through specific specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0050] It should also be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. The diagrams only show the components related to the present invention, rather than being drawn according to the number, shape, and size of the components in actual implementation. The form, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the layout form of its components may also be more complex.

[0051] A motor cooling control system with a fault diagnosis function, comprising a water pump, a water pump controller, a pipeline, a plurality of vibration sensors, a flow sensor, a differential pressure sensor, a temperature sensor, and a fault diagnosis module;

[0052] The vibration sensors are arranged on the water pump and the pipeline; the flow sensor and the differential pressure sensor are arranged in the pipeline; the temperature sensor is arranged on the motor to be cooled; the water pump controller controls the water pump;

[0053] As Figure 1 shown, the fault diagnosis module realizes the fault prediction of the motor cooling control system through the following steps:

[0054] S1: Signal acquisition, obtaining the water pump current, speed, vibration signal, motor temperature signal, pipeline vibration, flow, and differential pressure;

[0055] The water pump vibration signal is obtained by the vibration sensor arranged on the water pump; the pipeline vibration is obtained by the vibration sensor arranged on the pipeline; the motor temperature signal refers to the temperature of the motor to be cooled;

[0056] S2: Process the acquired signals and extract features, including:

[0057] S21: Signal preprocessing, including filtering, denoising, and normalization processing of the acquired signals;

[0058] S22: Perform blind source separation on the water pump vibration signal and the pipeline vibration signal;

[0059] Since the vibrations of the water pump and the pipeline affect each other, the vibration signals obtained by the vibration sensors arranged on the water pump and the pipeline are the result of their mutual influence. To more accurately implement subsequent signal analysis and processing, it is necessary to separate the two signals. The present invention restores the observed vibration signals received from the two vibration sensors to independent vibration source signals based on the independent component analysis method, including the following steps:

[0060] a. Perform preprocessing on the obtained vibration sensor signals, including centering processing and whitening processing;

[0061] b. Perform pre-separation on the preprocessed vibration sensor signals through the convolutional ICA algorithm to obtain pre-separated signals 、 , corresponding to the original signal of the pump vibration and the original signal of the pipeline vibration respectively;

[0062] c. Normalize and , and further perform short-time Fourier transform to the time-frequency domain to obtain and . Estimate the binary mask and perform cepstrum smoothing processing;

[0063] d. Strengthen and through the mask values after cepstrum smoothing processing to obtain and ;

[0064] e. Convert and to pure and separated vibration source signals and through inverse short-time Fourier transform, that is, the processed pump vibration signal and pipeline vibration signal.

[0065] S23: Feature extraction, including extracting time-domain and frequency-domain features from the pump current, pump vibration signal, and pipeline vibration signal; extracting time-domain features from the pump speed, motor temperature signal, pipeline flow rate, and pressure difference signal;

[0066] S3: Classify the state of the motor cooling control system based on the K-means clustering algorithm combined with the LSTM model. As Figure 2 shown, the classification results include normal, pump performance degradation, pump failure, pipeline blockage, and pipeline leakage. Among them, pump failure, pipeline blockage, and pipeline leakage are respectively divided into three levels: minor, medium, and severe according to the severity.

[0067] S31: Cluster the sample data based on the K-means clustering algorithm;

[0068] In a preferred embodiment, the clustering includes the following steps:

[0069] a. Import the sample data set into a list; the sample data set includes the collected signals and the extracted features;

[0070] b. Construct a set containing N clustering centers for the given data set. Use the improved k-means algorithm based on the maximum-minimum principle to initialize the clustering centers, and then perform k-means clustering on each category of data to finally obtain N optimal clustering centers;

[0071] c. Calculate the Euclidean distance between two sample vectors;

[0072] d. Create a matrix to store the cluster assignment results for each point, including the cluster index value and error. Calculate the distances from the current sample to each cluster center, find the nearest cluster center, and assign the sample to the nearest cluster.

[0073] e. After all samples are assigned, recalculate the centers of the N clusters.

[0074] f. Compare with the N cluster centers obtained in the previous calculation. If the cluster centers change, go to step c.

[0075] Otherwise, go to step g.

[0076] g. Stop when the cluster centers do not change and output the clustering results.

[0077] h. Merge the output clustering results and delete duplicate results to obtain the final set.

[0078] S32: Determine the training and validation samples according to the clustering results.

[0079] This includes determining the training and validation samples through mixed sampling of the clustering results. The clustering results include N cluster classification samples and discrete samples. The mixed sampling combines stratified sampling and random sampling, setting that a% of the training and validation samples are obtained through stratified sampling and b% are obtained through random sampling, where a% + b% = 1.

[0080] In a preferred embodiment, the stratified sampling separately obtains the sampling results for the N cluster classification samples and discrete samples, and the expression is as follows:

[0081] ;

[0082] Where, is the sample set of the i-th cluster center, is the sampling ratio, and L is the discrete sample set.

[0083] Optionally, is determined by the following method:

[0084] ;

[0085] Where, is the adjustment coefficient, which is determined according to the actual situation, is the Davies-Bouldin index of the i-th cluster, is the set Davies-Bouldin index threshold, is the set constant.

[0086] The above sampling method determines sampling parameters according to the performance indicators of clustering, realizes a more comprehensive sampling method, and reduces the redundancy of the sampling data volume.

[0087] S33: Establish an LSTM model and train it, including:

[0088] Encapsulate the sample data set and classification categories, construct a training set, establish a classification model, and determine the loss function; use the training set to train the classification model, observe the change of the loss function value with the number of iterations until the classification accuracy requirement is met.

[0089] S34: Classify the input data based on the trained LSTM model to obtain the classification result of the motor cooling control system state. The classification results include normal, pump performance degradation, pump failure, pipeline blockage, and pipeline leakage. Among them, pump failure, pipeline blockage, and pipeline leakage are respectively divided into three levels of minor, medium, and severe according to the severity.

[0090] S4: Judge the state of the motor cooling control system; if the pump performance of the current motor cooling control system degrades, but there are no pump failures, pipeline blockages, and pipeline leakages, then go to step S5, otherwise go to S6.

[0091] S5: Calculate the pump degradation degree and perform adaptive compensation;

[0092] S51: Calculate the pump degradation degree, which specifically includes the following steps:

[0093] S511: Obtain the pump current, pressure, and flow time series;

[0094] S512: Predict the pump current, pressure, and flow time series based on the autoregressive integrated moving average model with periodic correlation; since the temperature change of the motor during operation is periodic, correspondingly, the pump control parameters and flow rate of the cooling system are also periodically changed. Therefore, the present invention performs time series prediction of parameters based on the autoregressive integrated moving average model with periodic correlation, and the model is expressed as follows:

[0095] ;

[0096] In the formula: is the time series value at time t; is the white noise sequence value at time t; is the constant term; is the lag operator; is to shift back s times in time, that is and are the p-order autoregressive polynomial and the q-order moving average polynomial respectively; and are the periodic autoregressive polynomial of order P with period s and the periodic moving average polynomial of order Q, respectively. The model contains 7 parameters: p, d, q, s, P, D, and Q. The process of determining these parameters is called model order determination. p, d, and q represent the orders of autoregression, differencing, and moving average, respectively, while P, D, and Q represent the orders of periodic autoregression, differencing, and moving average with period s.

[0097] The process of model establishment and prediction includes: data processing: observing whether the sequence is stationary according to the time series graph of the original data. If it is a non-stationary sequence, perform differencing transformation to make it stationary; model parameter identification: identifying the model type through the autocorrelation coefficient and initially determining the order values of p, d, q, and P, D, Q; model identification and model diagnosis: selecting the most suitable model through the Akaike information criterion; model verification: verifying the optimal model using the sample data of the verification set; model prediction: predicting the time series of pump current, pressure, and flow rate using the optimal model.

[0098] S513: Calculate the pump degradation degree according to the time series prediction result;

[0099] Optionally, calculate the pump degradation degree based on a supervised learning classifier.

[0100] S52: When the pump degradation degree reaches a certain threshold, calculate the pump control parameters based on the adaptive control algorithm and send the pump control parameters to the pump controller to achieve adaptive adjustment control. The specific steps are as follows:

[0101] S521: Construct an adaptive adjustment model

[0102] The model is constructed by an FNN network. The network structure includes five layers, namely the input layer, fuzzification layer, rule inference layer, synthesis layer, defuzzification layer, and output layer. The network membership function is expressed as follows:

[0103] ;

[0104] where m is the center point, a is the width, and b is the slope.

[0105] S522: Obtain the pump degradation degree as the input parameter of the adaptive adjustment model;

[0106] S523: Input the degradation degree prediction result into the adaptive adjustment model to obtain the pump motor control parameters output by the adaptive adjustment model, including the modulation ratio, modulation degree, duty cycle, phase offset, and overmodulation of SVPWM.

[0107] S6: Predict the system failure probability based on the competing failure model and give an early warning; including the following steps:

[0108] S61: Calculate the reliability of the computing system

[0109] ;

[0110] wherein, is the system reliability when only the pump degradation failure exists within time t and there is no sudden failure factor acting is the system reliability when there is a sudden failure factor acting within time t;

[0111] S62: Calculate the predicted system failure probability:

[0112] ;

[0113] wherein, is the system density function, and the expression is as follows:

[0114] 。

[0115] In this embodiment, early warnings for faults such as pipeline blockage, pipeline leakage, pump degradation, and pump failure are realized, and an adaptive compensation function is provided. When the performance of some components degrades, the control parameters can be automatically adjusted to maintain the best working state of the system and improve the overall reliability. To achieve the accuracy of the prediction results, the present invention performs blind source separation on the pump vibration signal and the pipeline vibration signal based on the independent component analysis method, and performs clustering sampling on the sample data through a clustering algorithm, combines mixed sampling to ensure the comprehensiveness and accuracy of the training sample data, reduces sample redundancy. At the same time, in the prediction of the pump degradation degree, an autoregressive integrated moving average model with periodic correlation is used for prediction, considering the periodic change of the motor temperature, which improves the accuracy of the prediction.

[0116] The above-described embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A motor cooling control system with a fault diagnosis function, characterized in that, Including: A water pump, a water pump controller, pipes, several vibration sensors, a flow sensor, a differential pressure sensor, a temperature sensor, and a fault diagnosis module; The vibration sensors are arranged on the water pump and the pipes; the flow sensor and the differential pressure sensor are arranged in the pipes; the temperature sensor is arranged on the motor to be cooled; the water pump controller controls the water pump; The fault diagnosis module realizes the fault prediction of the motor cooling control system through the following steps: S1: Signal acquisition, obtaining the water pump current, rotation speed, vibration signal, motor temperature signal, pipe vibration, flow, and differential pressure; processing the acquired signals includes blind source separation of the water pump vibration signal and the pipeline vibration signal based on the independent component analysis method; S2: Processing the acquired signals and extracting features; S3: Classifying the state of the motor cooling control system based on the K-means clustering algorithm combined with the LSTM model; clustering the sample data, and determining the training and validation samples according to the clustering results; the determining the training and validation samples according to the clustering results includes performing mixed sampling on the clustering results to determine the training and validation samples, and the mixed sampling combines stratified sampling and random sampling; Stratified sampling respectively performs sampling on N clustering classification samples and discrete samples, and the expression is as follows: Among them, D i is the sample set of the i-th clustering center, r i is the extraction ratio, and L is the discrete sample set; S4: Judging the state of the motor cooling control system; if the water pump performance of the current motor cooling control system degrades, but there are no water pump failures, pipeline blockages, and pipeline leakages, then enter step S5, otherwise enter S6; S5: Calculating the water pump degradation degree and performing adaptive compensation, including: S51: Calculating the water pump degradation degree; S52: When the water pump degradation degree reaches a certain threshold, calculating the water pump control parameters based on the adaptive control algorithm, and sending the water pump control parameters to the water pump controller to achieve adaptive adjustment control; The step S52 includes: Constructing an adaptive adjustment model The model is constructed by an FNN network, and the network membership function is expressed as follows: Where m is the center point, a is the width, and b is the slope; The water pump control parameters include the modulation ratio, modulation degree, duty cycle, phase offset, and overmodulation of SVPWM; S6: Predicting the system fault probability based on the competing failure model and giving an early warning.

2. The motor cooling control system with a fault diagnosis function according to claim 1, characterized in that The classification results of classifying the state of the motor cooling control system include normal, water pump performance degradation, water pump failure, pipeline blockage, and pipeline leakage; among them, water pump failure, pipeline blockage, and pipeline leakage are respectively divided into three levels: minor, medium, and severe according to the severity; 3. The motor cooling control system with a fault diagnosis function according to claim 2, characterized in that, The step S3 further includes: S31: Clustering the sample data based on the K-means clustering algorithm; S32: Determining the training and validation samples according to the clustering results; S33: Establishing and training an LSTM model; S34: Classifying the input data based on the trained LSTM model to obtain the classification results of the state of the motor cooling control system.

4. The motor cooling control system with a fault diagnosis function according to claim 3, characterized in that, r i is determined by the following method: where k i is the adjustment coefficient, DB i is the Davies-Bouldin index of the i-th cluster, q0 is the set threshold of the Davies-Bouldin index, and r0 is the set constant.

5. The motor cooling control system with a fault diagnosis function according to claim 4, characterized in that, The calculating the water pump degradation degree includes: Obtaining the water pump current, pressure, and flow time series; predicting the water pump current, pressure, and flow time series based on the autoregressive integrated moving average model with periodic correlation; calculating the water pump degradation degree according to the time series prediction results.

6. The motor cooling control system with a fault diagnosis function according to claim 4, wherein The adaptive adjustment control includes: Construct an adaptive adjustment model based on the FNN network to obtain the degradation degree of the water pump, which is used as the input parameter of the adaptive adjustment model; input the degradation degree prediction result into the adaptive adjustment model to obtain the control parameter of the water pump motor output by the adaptive adjustment model.

7. The motor cooling control system with a fault diagnosis function according to claim 1, characterized in that The step S6 includes: S61: Calculate the system reliability R(t) R(t) = R1(t) + R2(t) where R1(t) is the system reliability when only the water pump degradation failure exists within time t and there is no sudden failure factor, and R2(t) is the system reliability when there is a sudden failure factor within time t; S62: Calculate the predicted system failure probability: where f(t) is the system density function, and the expression is as follows:

Images