A method and system for detecting faults of water pump motor

Abstract

The present invention discloses a method and system for detecting faults in a water pump motor. First, the water flow velocity values ​​at multiple predetermined time points are processed along the time dimension to obtain a multi-dimensional time series input vector of the water flow, then the waveform diagram of the multi-dimensional time series input vector of the water flow and the vibration signal are cross-modally encoded to obtain a multi-modal fusion feature matrix, then the multi-modal fusion feature matrix is ​​subjected to feature distribution posterior optimization to obtain an optimized multi-modal fusion feature matrix, and finally, the optimized multi-modal fusion feature matrix is ​​passed through a classifier to obtain a classification result for indicating whether the detected water pump motor has a fault. In this way, it can be determined whether the water pump motor has a fault.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a water pump motor fault detection method and system. Background Art

[0002] With the development of science and technology, my country's pump industry has entered a new stage of development. Water pumps are common equipment in the mechanical equipment industry and are widely used in industrial production and urban construction. During the long-term use of water pumps, due to various reasons, the water pump motor is likely to have various faults, such as bearing wear, rotor imbalance, cable connector aging, winding short circuit, etc. Timely and accurate detection of these faults is of great significance to ensuring production safety, reducing accident risks, and saving maintenance costs.

[0003] At present, the traditional method of detecting water pump motor faults usually installs a vibration sensor on the water pump motor and obtains the waveform of the vibration sensor in real time to determine whether the current motor is faulty. However, judging whether the water pump motor is faulty based on a single condition has limitations, which will result in low accuracy of the results. For example, when the vibration sensor detects that the water pump motor is faulty, but the water flow is not abnormal, it cannot be determined that the water pump motor is faulty.

[0004] Therefore, an optimized water pump motor fault detection scheme is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and system for detecting faults in a water pump motor. First, the water flow velocity values ​​at multiple predetermined time points are processed along the time dimension to obtain a multi-dimensional time series input vector of the water flow. Then, the waveform diagram of the multi-dimensional time series input vector of the water flow and the vibration signal are cross-modally encoded to obtain a multi-modal fusion feature matrix. Then, the multi-modal fusion feature matrix is ​​subjected to a posteriori optimization of feature distribution to obtain an optimized multi-modal fusion feature matrix. Finally, the optimized multi-modal fusion feature matrix is ​​passed through a classifier to obtain a classification result for indicating whether the detected water pump motor has a fault. In this way, it can be determined whether the water pump motor has a fault.

[0006] According to one aspect of the present application, a method for detecting a water pump motor fault is provided, comprising:

[0007] Acquire a vibration signal of the detected water pump motor in a predetermined time period collected by a vibration sensor and water flow velocity values ​​at a plurality of predetermined time points in the predetermined time period collected by a water flow velocity sensor;

[0008] Processing the water flow velocity values ​​at the plurality of predetermined time points along the time dimension to obtain a water flow multi-dimensional time series input vector;

[0009] Cross-modal encoding is performed on the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix;

[0010] Performing feature distribution posterior optimization on the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix; and

[0011] The optimized multimodal fusion feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

[0012] In the above-mentioned water pump motor fault detection method, the water flow velocity values ​​at the plurality of predetermined time points are processed along the time dimension to obtain a multi-dimensional time series input vector of the water flow, including:

[0013] Arranging the water flow velocity values ​​at the plurality of predetermined time points into a water flow velocity time series input vector according to the time dimension;

[0014] Calculating the difference between the water flow velocity values ​​at two adjacent positions in the water flow velocity time series input vector to obtain a water flow velocity change time series input vector; and

[0015] The water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded to obtain the water flow multi-dimensional time series input vector.

[0016] In the above-mentioned water pump motor fault detection method, the water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded to obtain the water flow multi-dimensional time series input vector, including:

[0017] The water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded using the following cascade formula to obtain the water flow multi-dimensional time series input vector;

[0018] Wherein, the cascade formula is:

[0019] V c =Concat[ V 1 ， V 2 ]

[0020] in, and represent the water flow velocity time series input vector and the water flow velocity change time series input vector respectively, Concat[ · ] represents a cascade function, Represents the multi-dimensional time series input vector of the water flow.

[0021] In the above-mentioned water pump motor fault detection method, cross-modal encoding is performed on the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix, including:

[0022] The multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal are input into a CLIP model including an image encoder and a sequence encoder so as to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal through the CLIP model including an image encoder and a sequence encoder to obtain the multi-modal fusion feature matrix.

[0023] In the above-mentioned water pump motor fault detection method, the sequence encoder includes a first convolution layer, a second convolution layer and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer and the second convolution layer respectively use one-dimensional convolution kernels with different scales.

[0024] In the above-mentioned water pump motor fault detection method, the image encoder is a convolutional neural network model used as a filter.

[0025] In the above-mentioned water pump motor fault detection method, the water flow multi-dimensional time series input vector and the waveform of the vibration signal are input into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the water flow multi-dimensional time series input vector and the waveform of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix, including:

[0026] Inputting the water flow multi-dimensional time series input vector into the sequence encoder to obtain a water flow multi-dimensional time series feature vector;

[0027] Inputting the waveform diagram of the vibration signal into the image encoder to obtain a vibration waveform feature vector; and

[0028] The multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector are fused to obtain the multi-modal fusion feature matrix.

[0029] In the above-mentioned water pump motor fault detection method, the fusion of the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multi-modal fusion feature matrix includes:

[0030] The multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector are fused using the following fusion formula to obtain the multi-modal fusion feature matrix;

[0031] Among them, the fusion formula is:

[0032]

[0033] in, represents the multi-dimensional time series feature vector of the water flow, represents the transposed vector of the multi-dimensional time series feature vector of the water flow, represents the vibration waveform feature vector, represents the multimodal fusion feature matrix, Represents vector multiplication.

[0034] In the above-mentioned water pump motor fault detection method, the multimodal fusion feature matrix is ​​subjected to feature distribution posterior optimization to obtain an optimized multimodal fusion feature matrix, including:

[0035] Performing feature distribution posterior optimization on the multimodal fusion feature matrix using the following optimization formula to obtain the optimized multimodal fusion feature matrix;

[0036] Wherein, the optimization formula is:

[0037]

[0038]

[0039] in, is the eigenvalue of each position of the optimized multimodal fusion feature matrix, and Set hyperparameters for the neighborhood, and when or When it is less than or equal to zero or greater than the width or height of the multimodal fusion feature matrix, the eigenvalue Can be set to zero or one. Represents the base 2 logarithmic function.

[0040] According to another aspect of the present application, a water pump motor fault detection system is provided, comprising:

[0041] A data acquisition module, used to acquire the vibration signal of the detected water pump motor in a predetermined time period acquired by the vibration sensor and the water flow velocity values ​​at multiple predetermined time points in the predetermined time period acquired by the water flow velocity sensor;

[0042] An input vector arrangement module, used for processing the water flow velocity values ​​at the plurality of predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow;

[0043] A cross-modal encoding module, used for cross-modal encoding the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix;

[0044] a feature distribution posterior optimization module, used to perform feature distribution posterior optimization on the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix; and

[0045] The classification module is used to pass the optimized multimodal fusion feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

[0046] Compared with the prior art, the water pump motor fault detection method and system provided by the present application first processes the water flow velocity values ​​at multiple predetermined time points along the time dimension to obtain a water flow multi-dimensional time series input vector, then cross-modally encodes the water flow multi-dimensional time series input vector and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix, then performs feature distribution posterior optimization on the multi-modal fusion feature matrix to obtain an optimized multi-modal fusion feature matrix, and finally, passes the optimized multi-modal fusion feature matrix through a classifier to obtain a classification result for indicating whether the detected water pump motor has a fault. In this way, it can be determined whether the water pump motor has a fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The following drawings are not deliberately scaled according to the actual size, and the focus is on illustrating the main purpose of the present application.

[0048] Figure 1 This is an application scenario diagram of the water pump motor fault detection method according to an embodiment of the present application.

[0049] Figure 2 Flow chart of a water pump motor fault detection method according to an embodiment of the present application.

[0050] Figure 3 Schematic diagram of the architecture of a water pump motor fault detection method according to an embodiment of the present application.

[0051] Figure 4 This is a flowchart of sub-step S120 of the water pump motor fault detection method according to an embodiment of the present application.

[0052] Figure 5 This is a flowchart of sub-step S130 of the water pump motor fault detection method according to an embodiment of the present application.

[0053] Figure 64 is a block diagram of a water pump motor fault detection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0055] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0056] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0057] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0058] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0059] As mentioned above, the traditional method of detecting water pump motor faults usually installs a vibration sensor on the water pump motor, and determines whether the current motor is faulty by obtaining the waveform of the vibration sensor in real time. However, judging whether the water pump motor is faulty by a single condition has limitations, which will result in low accuracy of the results. For example, when the vibration sensor detects that the water pump motor is faulty, but the water flow rate is not abnormal, it cannot be determined that the water pump motor is faulty. Therefore, an optimized water pump motor fault detection scheme is desired.

[0060] Accordingly, considering that in the actual process of fault detection of the water pump motor, if there is a fault in the water pump motor, the vibration frequency and amplitude of the water pump motor will often change, and at the same time, the water flow rate will also be affected. Therefore, in the technical solution of the present application, it is expected to complete the fault detection of the water pump motor based on the timing analysis of the vibration signal and water flow rate value of the water pump motor, so as to improve the accuracy and efficiency of fault diagnosis. In addition, it is also considered that since the vibration signal of the water pump motor is a continuous signal in the time domain, and the water flow rate value is a discrete signal, how to correlate and fuse the timing characteristic information of the two in the time dimension is particularly critical. However, since there is a large amount of useless interference information in the vibration signal waveform of the water pump motor, and the change of the water flow rate value in the timing is relatively weak, it is a small-scale change characteristic information, which is difficult to fully capture. Therefore, in this process, the difficulty lies in how to fully express the implicit characteristic distribution information about vibration in the vibration signal waveform of the water pump motor and the time-series change characteristic information of the water flow velocity value, so as to encode and optimize the implicit vibration characteristics of the water pump motor based on the time-series change characteristics of the water flow velocity value, thereby facilitating timely and accurate discovery of abnormal information of the water pump motor during operation, so as to facilitate fault detection of the water pump motor and ensure the normal operation of the water pump motor.

[0061] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. The development of deep learning and neural networks provides new solutions and solutions for mining the implicit characteristic distribution information of vibration in the vibration signal waveform of the water pump motor and the time series change characteristic information of the water flow velocity value.

[0062] Specifically, in the technical solution of the present application, first, the vibration signal of the detected water pump motor in a predetermined time period is collected by a vibration sensor, and the water flow velocity value at multiple predetermined time points in the predetermined time period is collected by a water flow velocity sensor. Next, considering that the water flow velocity value has a dynamic change law in the time dimension, and the water flow velocity value is a discrete signal of a time series, therefore, before feature extraction, it is necessary to further process the water flow velocity values ​​of the multiple predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow. Specifically, in order to be able to extract the change feature information of the water flow velocity value in the time dimension, in the technical solution of the present application, first, it is necessary to arrange the water flow velocity values ​​of the multiple predetermined time points according to the time dimension as a water flow velocity time series input vector, so as to integrate the distribution information of the water flow velocity value in the time series.

[0063] Then, considering that the time-series change information of the water flow velocity value is weak, this weak change information is small-scale change information relative to the water flow velocity value. If the absolute change data of the water flow velocity value is used to extract the time-series dynamic change features of the water flow velocity value, it will not only result in a large amount of calculation and cause overfitting, but also make the small-scale weak change features of the water flow velocity value in the time dimension difficult to be detected, thereby affecting the accuracy of subsequent classification.

[0064] Based on this, in the technical solution of the present application, the time series relative change characteristics and absolute change characteristics of the water flow velocity value are used to comprehensively extract the time series dynamic characteristics of the water flow velocity value. Specifically, first, the difference between the water flow velocity values ​​at two adjacent positions in the water flow velocity time series input vector is calculated to obtain the water flow velocity change time series input vector. Next, considering that the time series relative change characteristics and the time series absolute change characteristics of the water flow velocity value have a dynamic change correlation relationship with respect to the water flow velocity. Therefore, in order to fully explore the dynamic change law of the water flow velocity value in the time dimension, so as to accurately perform fault detection of the water pump motor, in the technical solution of the present application, the water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded to obtain the multi-dimensional time series input vector of the water flow.

[0065] Furthermore, in the actual process of fault detection of the water pump motor, if there is a fault in the water pump motor, the vibration frequency and amplitude of the water pump motor will often change, and the water flow rate will also be affected. Therefore, the working state of the water pump motor can be evaluated by analyzing the timing characteristics of the vibration signal of the water pump motor and the timing change characteristics of the water flow rate respectively. In particular, considering that the vibration signal of the water pump motor and the water flow rate value are susceptible to abnormal changes due to interference from external environmental factors, in order to further improve the accuracy of fault detection of the water pump motor, in the technical solution of the present application, it is expected to use the timing change characteristic information of the water flow rate to encode and optimize the timing characteristic information of the vibration signal of the water pump motor, so as to more accurately monitor and evaluate the working state of the water pump motor, thereby discovering potential fault hazards and handling them in a timely manner.

[0066] Specifically, the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal are input into a CLIP model including an image encoder and a sequence encoder, so as to perform cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix. In particular, here, the waveform image encoder uses a convolutional neural network model as a filter to perform feature mining on the waveform of the vibration signal to extract the implicit feature distribution information about the working vibration of the water pump motor in the waveform of the vibration signal, thereby obtaining a vibration waveform feature vector. In addition, due to the volatility and uncertainty of the water flow velocity value in the time dimension, it presents different change rules under different time period spans. Therefore, here, the sequence encoder includes a first convolution layer, a second convolution layer, and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer. It is worth mentioning that the first convolution layer and the second convolution layer respectively use one-dimensional convolution kernels with different scales to extract the multi-scale time series dynamic change feature information of the water flow velocity value under different time spans.

[0067] Then, based on the joint encoding module of the CLIP model, the multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector are fused, so as to optimize the encoding of image attributes of the time series implicit feature distribution information of the vibration signal waveform based on the time series multi-scale dynamic change feature information of the water flow velocity, so as to obtain the multimodal fusion feature matrix.

[0068] Then, the multimodal fusion feature matrix is ​​used as a classification feature matrix for classification processing in a classifier to obtain a classification result for indicating whether the detected water pump motor has a fault. That is, in the technical solution of the present application, the label of the classifier includes the presence of a fault in the detected water pump motor (first label), and the presence of no fault in the detected water pump motor (second label), wherein the classifier determines which classification label the optimized multimodal fusion feature matrix belongs to by a soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether the detected water pump motor has a fault". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the detected water pump motor has a fault is actually converted into a binary class probability distribution that conforms to the laws of nature through the classification label. In essence, the physical meaning of the natural probability distribution of the label is used, rather than the linguistic text meaning of "whether the detected water pump motor has a fault". It should be understood that in the technical solution of the present application, the classification label of the classifier is a detection evaluation label for determining whether the detected water pump motor has a fault. Therefore, after obtaining the classification result, the water pump motor fault detection can be performed timely and accurately based on the classification result to ensure the normal operation of the water pump motor.

[0069] In particular, in the technical solution of the present application, when the multi-dimensional time series input vector of the water flow and the vibration signal are obtained by a CLIP model including an image encoder and a sequence encoder to obtain the multimodal fusion feature matrix, the sequence feature vector obtained by the multi-dimensional time series input vector of the water flow through the sequence encoder and the image feature vector obtained by the waveform diagram of the vibration signal through the image encoder are positionally associated to obtain the multimodal fusion feature matrix. Therefore, each row feature vector of the multimodal fusion feature matrix is ​​obtained by multiplying the corresponding eigenvalue of the sequence feature vector by the image feature vector. Therefore, the multimodal fusion feature matrix can be regarded as a combined feature set of local feature sets corresponding to each row feature vector, and the feature distribution of each row feature vector has a mutually correlated neighborhood distribution relationship along the column direction of the multimodal fusion feature matrix.

[0070] Moreover, since each eigenvalue of the sequence feature vector has a sequence feature distribution relationship of time series correlation of the absolute value and change value of the water flow velocity of the multi-dimensional time series input vector of the water flow, each row feature vector of the multimodal fusion feature matrix has, in addition to a mutually correlated neighborhood distribution relationship, a multi-source information correlation relationship corresponding to the time series correlation feature sequence distribution information of the water flow velocity.

[0071] Therefore, in order to improve the fusion expression effect of the multimodal fusion feature matrix on the time series correlation features of the water flow velocity parameter value and the local image semantic correlation features of the waveform of the vibration signal, the eigenvalue of each position of the multimodal fusion feature matrix is Perform spatial multi-source fusion prior information distribution optimization to obtain optimized eigenvalues , specifically expressed as:

[0072]

[0073]

[0074] in and Set hyperparameters for the neighborhood, and when or When it is less than or equal to zero or greater than the width or height of the multimodal fusion feature matrix, the eigenvalue Can be set to zero or one.

[0075] Here, the spatial multi-source fusion prior information distribution optimization can be based on the robustness-like maximum likelihood estimation of feature space distribution fusion to take the multimodal fusion feature matrix as a feature global set composed of feature local sets corresponding to multiple interrelated neighborhood parts, realize the effective conversion of the multi-source prior information of each feature local set to the feature global set, and obtain the standard expected optimization paradigm that can be used to evaluate the internal spatial correlation of the feature matrix and the spatial information fusion change relationship through the prior information distribution construction under the multi-source condition, so as to improve the fusion expression effect of the multimodal fusion feature matrix on the time series correlation feature of the water flow velocity parameter value and the local image semantic correlation feature of the waveform diagram of the vibration signal, thereby improving the accuracy of the classification result obtained by the classifier. In this way, the abnormal situation of the water pump motor during operation can be discovered in a timely and accurate manner, so as to facilitate the fault detection of the water pump motor and ensure the normal operation of the water pump motor.

[0076] Figure 1 FIG. 1 is an application scenario diagram of the water pump motor fault detection method according to an embodiment of the present application. Figure 1 As shown, in this application scenario, first, the vibration sensor (for example, Figure 1 The vibration signal of the detected water pump motor collected in a predetermined time period (for example, Figure 1 D1 as shown in FIG. 1 ) and the water flow velocity values ​​at multiple predetermined time points within the predetermined time period collected by the water flow velocity sensor (for example, Figure 1Then, the waveform of the vibration signal and the water flow rate values ​​at the plurality of predetermined time points are input into a server deployed with a water pump motor fault detection algorithm (for example, Figure 1 S), wherein the server is able to use the water pump motor fault detection algorithm to process the waveform of the vibration signal and the water flow rate values ​​at the multiple predetermined time points to obtain a classification result indicating whether the detected water pump motor has a fault.

[0077] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0078] Figure 2 FIG. 1 is a flow chart of a method for detecting a water pump motor fault according to an embodiment of the present application. Figure 2 As shown, the water pump motor fault detection method according to the embodiment of the present application includes the steps of: S110, acquiring the vibration signal of the detected water pump motor in a predetermined time period collected by a vibration sensor and the water flow velocity value at multiple predetermined time points in the predetermined time period collected by a water flow velocity sensor; S120, processing the water flow velocity values ​​at the multiple predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow; S130, cross-modally encoding the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix; S140, performing feature distribution a posteriori optimization on the multi-modal fusion feature matrix to obtain an optimized multi-modal fusion feature matrix; and, S150, passing the optimized multi-modal fusion feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

[0079] Figure 3 FIG. 1 is a schematic diagram of the structure of a pump motor fault detection method according to an embodiment of the present application. Figure 3 As shown, in the network architecture, first, the vibration signal of the detected water pump motor collected by the vibration sensor in a predetermined time period and the water flow velocity values ​​at multiple predetermined time points in the predetermined time period collected by the water flow velocity sensor are obtained; then, the water flow velocity values ​​at the multiple predetermined time points are processed along the time dimension to obtain a multi-dimensional time series input vector of the water flow; then, the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal are cross-modally encoded to obtain a multi-modal fusion feature matrix; then, the feature distribution of the multi-modal fusion feature matrix is ​​posteriorly optimized to obtain an optimized multi-modal fusion feature matrix; finally, the optimized multi-modal fusion feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

[0080] More specifically, in step S110, the vibration signal of the detected water pump motor in a predetermined time period collected by the vibration sensor and the water flow velocity values ​​at multiple predetermined time points in the predetermined time period collected by the water flow velocity sensor are obtained. In the actual process of fault detection of the water pump motor, if there is a fault in the water pump motor, the vibration frequency and amplitude of the water pump motor will often change, and the water flow velocity will also be affected. Therefore, the fault detection of the water pump motor can be completed based on the timing analysis of the vibration signal and the water flow velocity value of the water pump motor, so as to improve the accuracy and efficiency of fault diagnosis.

[0081] More specifically, in step S120, the water flow velocity values ​​at the plurality of predetermined time points are processed along the time dimension to obtain a water flow multi-dimensional time series input vector. Since the water flow velocity value has a dynamic change rule in the time dimension, and the water flow velocity value is a discrete signal of the time series, it is necessary to further process the water flow velocity values ​​at the plurality of predetermined time points along the time dimension before feature extraction to obtain a water flow multi-dimensional time series input vector.

[0082] Furthermore, since the time-series change information of the water flow velocity value is weak, this weak change information is a small-scale change information relative to the water flow velocity value. If the absolute change data of the water flow velocity value is used to extract the time-series dynamic change features of the water flow velocity value, it will not only cause a large amount of calculation and cause overfitting, but also make the small-scale weak change features of the water flow velocity value in the time dimension difficult to detect, thereby affecting the accuracy of subsequent classification. Therefore, the time-series relative change features and absolute change features of the water flow velocity value can be used to comprehensively extract the time-series dynamic features of the water flow velocity value.

[0083] Accordingly, in a specific example, Figure 4 As shown, the water flow velocity values ​​at the multiple predetermined time points are processed along the time dimension to obtain a water flow multi-dimensional timing input vector, including: S121, arranging the water flow velocity values ​​at the multiple predetermined time points into a water flow velocity timing input vector according to the time dimension; S122, calculating the difference between the water flow velocity values ​​at two adjacent positions in the water flow velocity timing input vector to obtain a water flow velocity change timing input vector; and, S123, cascading the water flow velocity timing input vector and the water flow velocity change timing input vector to obtain the water flow multi-dimensional timing input vector.

[0084] Accordingly, in a specific example, the water flow velocity timing input vector and the water flow velocity change timing input vector are cascaded to obtain the water flow multi-dimensional timing input vector, including: the water flow velocity timing input vector and the water flow velocity change timing input vector are cascaded according to the following cascade formula to obtain the water flow multi-dimensional timing input vector; wherein the cascade formula is:

[0085] V c =Concat[ V 1 ， V 2 ]

[0086] in, and represent the water flow velocity time series input vector and the water flow velocity change time series input vector respectively, Concat[ · ] represents a cascade function, Represents the multi-dimensional time series input vector of the water flow.

[0087] More specifically, in step S130, the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal are cross-modally encoded to obtain a multi-modal fusion feature matrix. The vibration signal of the water pump motor and the water flow velocity value are easily disturbed by external environmental factors and undergo abnormal changes. Therefore, in order to further improve the accuracy of fault detection of the water pump motor, the time series change characteristic information of the water flow velocity can be used to encode and optimize the time series characteristic information of the vibration signal of the water pump motor, so as to more accurately monitor and evaluate the working state of the water pump motor, thereby discovering potential fault hazards and handling them in a timely manner.

[0088] Accordingly, in a specific example, the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal are cross-modally encoded to obtain a multi-modal fusion feature matrix, including: inputting the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix.

[0089] Accordingly, in a specific example, the sequence encoder includes a first convolution layer, a second convolution layer, and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer and the second convolution layer respectively use one-dimensional convolution kernels with different scales. The image encoder is a convolutional neural network model as a filter.

[0090] It should be understood that Convolutional Neural Network (CNN) is an artificial neural network that is widely used in image recognition and other fields. Convolutional neural network may include input layer, hidden layer and output layer, wherein the hidden layer may include convolution layer, pooling layer, activation layer and fully connected layer, etc. The previous layer performs corresponding operations according to the input data and outputs the operation results to the next layer. The input initial data is processed through multiple layers of operations to obtain a final result.

[0091] Accordingly, in a specific example, Figure 5 As shown, the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal are input into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix, including: S131, inputting the multi-dimensional time series input vector of the water flow into the sequence encoder to obtain a multi-dimensional time series feature vector of the water flow; S132, inputting the waveform graph of the vibration signal into the image encoder to obtain a vibration waveform feature vector; and, S133, fusing the multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector to obtain the multi-modal fusion feature matrix.

[0092] Accordingly, in a specific example, fusing the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multimodal fusion feature matrix includes: fusing the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multimodal fusion feature matrix using the following fusion formula; wherein the fusion formula is:

[0093]

[0094] in, represents the multi-dimensional time series feature vector of the water flow, represents the transposed vector of the multi-dimensional time series feature vector of the water flow, represents the vibration waveform feature vector, represents the multimodal fusion feature matrix, Represents vector multiplication.

[0095] More specifically, in step S140, the feature distribution of the multimodal fusion feature matrix is ​​optimized a posteriori to obtain an optimized multimodal fusion feature matrix. Since each eigenvalue of the sequence feature vector has a sequence feature distribution relationship of the time series correlation of the absolute value and change value of the water flow velocity of the multi-dimensional time series input vector of the water flow, each row feature vector of the multimodal fusion feature matrix has a multi-source information correlation relationship corresponding to the time series correlation feature sequence distribution information of the water flow velocity in addition to the mutually correlated neighborhood distribution relationship. In order to improve the fusion expression effect of the multimodal fusion feature matrix for the time series correlation characteristics of the water flow velocity parameter value and the local image semantic correlation characteristics of the waveform diagram of the vibration signal, the spatial multi-source fusion a priori information distribution optimization is performed on the eigenvalues ​​of each position of the multimodal fusion feature matrix.

[0096] Accordingly, in a specific example, performing a posteriori optimization on the feature distribution of the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix includes: performing a posteriori optimization on the feature distribution of the multimodal fusion feature matrix according to the following optimization formula to obtain the optimized multimodal fusion feature matrix; wherein the optimization formula is:

[0097]

[0098]

[0099] in, is the eigenvalue of each position of the optimized multimodal fusion feature matrix, and Set hyperparameters for the neighborhood, and when or When it is less than or equal to zero or greater than the width or height of the multimodal fusion feature matrix, the eigenvalue Can be set to zero or one. Represents the base 2 logarithmic function.

[0100] The spatial multi-source fusion prior information distribution optimization can be based on the robust maximum likelihood estimation of the feature space distribution fusion to treat the multimodal fusion feature matrix as a feature global set composed of feature local sets corresponding to multiple interrelated neighborhood parts, so as to achieve effective conversion of the multi-source prior information of each feature local set to the feature global set, and through the construction of the prior information distribution in the multi-source case, to obtain the standard expected optimization paradigm that can be used to evaluate the internal spatial correlation of the feature matrix and the spatial information fusion change relationship, so as to improve the fusion expression effect of the multimodal fusion feature matrix on the time series correlation characteristics of the water flow velocity parameter value and the local image semantic correlation characteristics of the waveform diagram of the vibration signal, thereby improving the accuracy of the classification result obtained by the classifier.

[0101] More specifically, in step S150, the optimized multimodal fusion feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault. After obtaining the classification result, the fault detection of the water pump motor can be timely and accurately performed based on the classification result to ensure the normal operation of the water pump motor.

[0102] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

[0103] In summary, based on the water pump motor fault detection method of the embodiment of the present application, it first processes the water flow velocity values ​​at multiple predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow, then cross-modally encodes the waveform diagram of the multi-dimensional time series input vector of the water flow and the vibration signal to obtain a multi-modal fusion feature matrix, then performs feature distribution posterior optimization on the multi-modal fusion feature matrix to obtain an optimized multi-modal fusion feature matrix, and finally, passes the optimized multi-modal fusion feature matrix through a classifier to obtain a classification result for indicating whether the detected water pump motor has a fault. In this way, it can be determined whether the water pump motor has a fault.

[0104] Furthermore, the technical solution of the present application also provides a storage medium, in which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes any of the aforementioned water pump motor fault detection methods.

[0105] Figure 6 FIG. 1 is a block diagram of a water pump motor fault detection system 100 according to an embodiment of the present application. Figure 6 As shown, according to the water pump motor fault detection system 100 of the embodiment of the present application, it includes: a data acquisition module 110, which is used to obtain the vibration signal of the detected water pump motor in a predetermined time period collected by a vibration sensor and the water flow velocity value at multiple predetermined time points in the predetermined time period collected by a water flow velocity sensor; an input vector arrangement module 120, which is used to process the water flow velocity values ​​of the multiple predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow; a cross-modal encoding module 130, which is used to cross-modally encode the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix; a feature distribution a posteriori optimization module 140, which is used to perform feature distribution a posteriori optimization on the multi-modal fusion feature matrix to obtain an optimized multi-modal fusion feature matrix; and a classification module 150, which is used to pass the optimized multi-modal fusion feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

[0106] In one example, in the above-mentioned water pump motor fault detection system 100, the input vector arrangement module 120 is used to: arrange the water flow velocity values ​​at the multiple predetermined time points according to the time dimension into a water flow velocity timing input vector; calculate the difference between the water flow velocity values ​​at two adjacent positions in the water flow velocity timing input vector to obtain a water flow velocity change timing input vector; and cascade the water flow velocity timing input vector and the water flow velocity change timing input vector to obtain the water flow multi-dimensional timing input vector.

[0107] In one example, in the above-mentioned water pump motor fault detection system 100, the water flow velocity timing input vector and the water flow velocity change timing input vector are cascaded to obtain the water flow multi-dimensional timing input vector, including: the water flow velocity timing input vector and the water flow velocity change timing input vector are cascaded according to the following cascade formula to obtain the water flow multi-dimensional timing input vector; wherein the cascade formula is:

[0108] V c =Concat[ V 1 ， V 2 ]

[0109] in, and represent the water flow velocity time series input vector and the water flow velocity change time series input vector respectively, Concat[ · ] represents a cascade function, Represents the multi-dimensional time series input vector of the water flow.

[0110] In one example, in the above-mentioned water pump motor fault detection system 100, the cross-modal encoding module 130 is used to: input the water flow multi-dimensional time series input vector and the waveform of the vibration signal into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the water flow multi-dimensional time series input vector and the waveform of the vibration signal through the CLIP model including an image encoder and a sequence encoder to obtain the multi-modal fusion feature matrix.

[0111] In one example, in the above-mentioned water pump motor fault detection system 100, the sequence encoder includes a first convolution layer, a second convolution layer, and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer and the second convolution layer respectively use one-dimensional convolution kernels with different scales. The image encoder is a convolutional neural network model as a filter.

[0112] In one example, in the above-mentioned water pump motor fault detection system 100, the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal are input into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform graph of the vibration signal through the CLIP model including an image encoder and a sequence encoder to obtain the multi-modal fusion feature matrix, including: inputting the multi-dimensional time series input vector of the water flow into the sequence encoder to obtain a multi-dimensional time series feature vector of the water flow; inputting the waveform graph of the vibration signal into the image encoder to obtain a vibration waveform feature vector; and, fusing the multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector to obtain the multi-modal fusion feature matrix.

[0113] In one example, in the above-mentioned water pump motor fault detection system 100, fusing the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multimodal fusion feature matrix includes: fusing the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multimodal fusion feature matrix using the following fusion formula; wherein the fusion formula is:

[0114]

[0115] in, represents the multi-dimensional time series feature vector of the water flow, represents the transposed vector of the multi-dimensional time series feature vector of the water flow, represents the vibration waveform feature vector, represents the multimodal fusion feature matrix, Represents vector multiplication.

[0116] In one example, in the above-mentioned water pump motor fault detection system 100, the feature distribution a posteriori optimization module 140 is used to: perform feature distribution a posteriori optimization on the multimodal fusion feature matrix according to the following optimization formula to obtain the optimized multimodal fusion feature matrix; wherein the optimization formula is:

[0117]

[0118]

[0119] in, is the eigenvalue of each position of the optimized multimodal fusion feature matrix, and Set hyperparameters for the neighborhood, and when or When it is less than or equal to zero or greater than the width or height of the multimodal fusion feature matrix, the eigenvalue Can be set to zero or one. Represents the base 2 logarithmic function.

[0120] Here, those skilled in the art will appreciate that the specific functions and operations of each module in the above-mentioned water pump motor fault detection system 100 have been described in detail above. Figures 1 to 5 The description of the water pump motor fault detection method has been described in detail, and therefore, its repeated description will be omitted.

[0121] As described above, the water pump motor fault detection system 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a water pump motor fault detection algorithm. In one example, the water pump motor fault detection system 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the water pump motor fault detection system 100 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the water pump motor fault detection system 100 can also be one of the many hardware modules of the wireless terminal.

[0122] Alternatively, in another example, the water pump motor fault detection system 100 and the wireless terminal may also be separate devices, and the water pump motor fault detection system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0123] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

[0124] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0125] All or part of the software may sometimes communicate over a network, such as the Internet or other communications network. Such communications can load software from one computer device or processor to another. For example: from a server or host computer of a video target detection device to a hardware platform of a computer environment, or other computer environment that implements the system, or a system with similar functions related to providing information required for target detection. Therefore, another medium capable of transmitting software elements can also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., transmitted through cables, optical cables or air. Physical media used to carry carriers, such as cables, wireless connections or optical cables and the like, can also be considered as media that carry software. As used herein, unless limited to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media involved in the process of the processor executing any instructions.

[0126] The present application uses specific words to describe the embodiments of the present application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0127] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0128] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0129] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A method for detecting a water pump motor fault, characterized in that: include: Acquire a vibration signal of the detected water pump motor in a predetermined time period collected by a vibration sensor and water flow velocity values ​​at a plurality of predetermined time points in the predetermined time period collected by a water flow velocity sensor; Processing the water flow velocity values ​​at the plurality of predetermined time points along the time dimension to obtain a water flow multi-dimensional time series input vector; Cross-modal encoding is performed on the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix; Performing feature distribution posterior optimization on the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix; as well as Passing the optimized multimodal fusion feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the detected water pump motor has a fault; The water flow velocity values ​​at the plurality of predetermined time points are processed along the time dimension to obtain a multi-dimensional time series input vector of the water flow, including: Arranging the water flow velocity values ​​at the plurality of predetermined time points into a water flow velocity time series input vector according to the time dimension; Calculating the difference between the water flow velocity values ​​at two adjacent positions in the water flow velocity time series input vector to obtain a water flow velocity change time series input vector; and cascading the water flow velocity time series input vector and the water flow velocity change time series input vector to obtain the water flow multi-dimensional time series input vector; Cross-modal encoding is performed on the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix, including: Inputting the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix; Inputting the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal into a CLIP model including an image encoder and a sequence encoder to perform row cross-modal encoding on the multi-dimensional time series input vector of the water flow and the waveform of the vibration signal through the CLIP model including the image encoder and the sequence encoder to obtain the multi-modal fusion feature matrix, including: Inputting the water flow multi-dimensional time series input vector into the sequence encoder to obtain a water flow multi-dimensional time series feature vector; Inputting the waveform diagram of the vibration signal into the image encoder to obtain a vibration waveform feature vector; and fusing the water flow multi-dimensional time series feature vector and the vibration waveform feature vector to obtain the multimodal fusion feature matrix; The multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector are fused to obtain the multi-modal fusion feature matrix, including: The multi-dimensional time series feature vector of the water flow and the vibration waveform feature vector are fused using the following fusion formula to obtain the multi-modal fusion feature matrix; The fusion formula is: Among them, V a represents the multi-dimensional time series feature vector of the water flow, represents the transposed vector of the multi-dimensional time series feature vector of the water flow, V b represents the vibration waveform feature vector, M represents the multimodal fusion feature matrix, Represents vector multiplication; Performing a posteriori optimization on the feature distribution of the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix, comprising: performing a posteriori optimization on the feature distribution of the multimodal fusion feature matrix according to the following optimization formula to obtain the optimized multimodal fusion feature matrix; The optimization formula is: Among them, m i,j ′ is the eigenvalue of each position of the optimized multimodal fusion feature matrix, m and n are neighborhood setting hyperparameters, and when k or j is less than or equal to zero or greater than the width or height of the multimodal fusion feature matrix, the eigenvalue m k,l Can be set to zero or one, log represents the logarithmic function with base 2.

2. The water pump motor fault detection method according to claim 1, characterized in that: The water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded to obtain the water flow multi-dimensional time series input vector, including: The water flow velocity time series input vector and the water flow velocity change time series input vector are cascaded using the following cascade formula to obtain the water flow multi-dimensional time series input vector; Wherein, the cascade formula is: V c =Concat[V1, V2] Wherein, V1 and V2 represent the water flow velocity time series input vector and the water flow velocity change time series input vector respectively, Concat[·] represents the cascade function, V c Represents the multi-dimensional time series input vector of the water flow.

3. The water pump motor fault detection method according to claim 1, characterized in that: The sequence encoder includes a first convolution layer, a second convolution layer, and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer and the second convolution layer respectively use one-dimensional convolution kernels with different scales.

4. The water pump motor fault detection method according to claim 3, characterized in that: The image encoder is a convolutional neural network model as a filter.

5. A water pump motor fault detection system using the water pump motor fault detection method according to claim 1, characterized in that: include: A data acquisition module, used to acquire the vibration signal of the detected water pump motor in a predetermined time period acquired by the vibration sensor and the water flow velocity values ​​at multiple predetermined time points in the predetermined time period acquired by the water flow velocity sensor; An input vector arrangement module, used for processing the water flow velocity values ​​at the plurality of predetermined time points along the time dimension to obtain a multi-dimensional time series input vector of the water flow; A cross-modal encoding module, used for cross-modal encoding the multi-dimensional time series input vector of the water flow and the waveform diagram of the vibration signal to obtain a multi-modal fusion feature matrix; A feature distribution posterior optimization module, used for performing feature distribution posterior optimization on the multimodal fusion feature matrix to obtain an optimized multimodal fusion feature matrix; as well as The classification module is used to pass the optimized multimodal fusion feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the detected water pump motor has a fault.

Images