Device and method for monitoring equipment status and operation

Abstract

The present application discloses a device and method for monitoring the operation of equipment status. It first arranges the operation status data of the plurality of predetermined time points according to the time dimension into an operation status data input vector, and then inputs the data into the first convolutional neural network model and the second convolutional neural network model respectively to obtain a first-scale operation status feature vector and a second-scale operation status feature vector. Then, the first-scale operation status feature vector and the second-scale operation status feature vector are fused to obtain a multi-scale operation status feature vector. Then, based on the Gaussian density map, the multi-scale operation status feature vector is subjected to feature-level data enhancement to obtain a classification feature matrix. Finally, the classification feature matrix is ​​passed through a classifier to obtain a classification result for indicating whether the operation status of the equipment to be monitored is normal. In this way, the operation status of the equipment to be monitored can be monitored intelligently.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring technology, and more specifically, to a device and method for monitoring the operation of equipment status. Background Art

[0002] In electromechanical engineering projects, some electrical equipment is often used and operated in an unattended environment, such as GNSS receivers, video cameras, water level gauges, rain gauges and various sensors. These devices often have some faults during long-term operation. These faults include: freeze, power shortage, equipment line disconnection, and component damage.

[0003] Once these devices fail, equipment maintenance personnel are unable to accurately determine the cause of the failure and can only find the fault point through on-site observation and measurement, and then repair and eliminate the fault.

[0004] This approach to the problem has some disadvantages:

[0005] (1) Equipment maintenance personnel need to travel long distances to the equipment installation site, which not only wastes time and causes fatigue, but also costs a lot.

[0006] (2) Some faults can be easily solved in a few seconds with simple operations. For example, if the computer freezes, just disconnect the power supply and then reconnect it.

[0007] (3) Solving some faults requires a long process, which requires observing the equipment's operating status for a long time (several days or months) to discover the fault pattern and find a solution to the problem. For example, in the case of power shortage, it is necessary to obtain the energy consumption curve of the equipment over a period of time, and then reasonably design the equipment's operating cycle and battery power configuration. It is impossible to let equipment maintenance personnel stay in the wilderness for a long time.

[0008] (4) Equipment maintenance personnel need to bring tools and accessories to the site for repair. Since they do not know the specific cause of the equipment failure in advance, it is impossible for them to bring all the tools and accessories. This can easily lead to the problem that after arriving at the site, they find that the tools and accessories they brought are not useful, and the required tools and spare parts are not brought. For example, equipment line disconnection and component damage may cause the entire equipment to stop working, but these two types of failures require different parts to be replaced and the fault points to be repaired, so the tools and accessories required are different.

[0009] Therefore, an optimized equipment status operation monitoring solution is expected. Summary of the invention

[0010] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an equipment status operation monitoring device and method. It first arranges the operation status data of the multiple predetermined time points according to the time dimension as an operation status data input vector, and then inputs them into the first convolutional neural network model and the second convolutional neural network model respectively to obtain a first-scale operation status feature vector and a second-scale operation status feature vector. Then, the first-scale operation status feature vector and the second-scale operation status feature vector are fused to obtain a multi-scale operation status feature vector. Then, based on the Gaussian density map, the multi-scale operation status feature vector is subjected to feature-level data enhancement to obtain a classification feature matrix. Finally, the classification feature matrix is ​​passed through a classifier to obtain a classification result for indicating whether the operation status of the equipment to be monitored is normal. In this way, the operation status of the equipment to be monitored can be monitored intelligently.

[0011] According to one aspect of the present application, there is provided a device for monitoring the operation of an equipment state, comprising:

[0012] An operation data acquisition module is used to obtain the operation status data of the equipment to be monitored at multiple predetermined time points within a predetermined time period;

[0013] A vectorization module, used for arranging the operation status data of the plurality of predetermined time points into an operation status data input vector according to a time dimension;

[0014] A first scale feature extraction module, configured to obtain a first scale running state feature vector by extracting the running state data input vector using a first convolutional neural network model having a one-dimensional convolution kernel of a first scale;

[0015] A second scale feature extraction module, configured to extract the running state data input vector by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second scale running state feature vector;

[0016] a multi-scale fusion module, configured to fuse the first-scale running state feature vector and the second-scale running state feature vector to obtain a multi-scale running state feature vector;

[0017] a data enhancement module, configured to perform feature-level data enhancement on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix; and

[0018] The monitoring result generating module is used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operating status of the device to be monitored is normal.

[0019] In the above-mentioned equipment status operation monitoring device, the first-scale feature extraction module is further used to: by using each layer of the first convolutional neural network model with a one-dimensional convolution kernel of the first scale, respectively perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of the layer so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the first-scale operation status feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the operation status data input vector.

[0020] In the above-mentioned equipment status operation monitoring device, the second-scale feature extraction module is further used to: by using each layer of the second convolutional neural network model with a one-dimensional convolution kernel of the second scale, respectively perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of the layer so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the second-scale operation status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the operation status data input vector.

[0021] In the above-mentioned equipment status operation monitoring device, the multi-scale fusion module is further used to perform Hilbert space constraints of vector modulus basis on the first-scale operation status feature vector and the second-scale operation status feature vector according to the following formula to obtain the multi-scale operation status feature vector;

[0022] Wherein, the formula is:

[0023]

[0024] in, represents the first scale running state feature vector, represents the second scale running state feature vector, represents the transposed vector of the second scale running state feature vector, represents the multi-scale operating state feature vector, represents positional addition, represents the two-norm of a vector, Represents a one-dimensional convolution operation, that is, the convolution operator Pair Vector Perform a one-dimensional convolution.

[0025] In the above-mentioned equipment status operation monitoring device, the data enhancement module includes:

[0026] A Gaussian density map construction unit, used to construct a Gaussian density map of the multi-scale operating state feature vector, wherein the mean vector of the Gaussian density map is the multi-scale operating state feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of corresponding two positions in the multi-scale operating state feature vector;

[0027] The Gaussian discretization unit is used to randomly sample the Gaussian distribution at each position of the Gaussian density map to obtain a plurality of row vectors, and to arrange the plurality of row vectors in two dimensions to obtain the classification feature matrix.

[0028] In the above-mentioned equipment status operation monitoring device, the monitoring result generating module is further used to:

[0029] Expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors;

[0030] Performing full connection encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and

[0031] The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

[0032] In the above-mentioned device status operation monitoring apparatus, the operation status data is the operation power value of the device to be monitored.

[0033] According to another aspect of the present application, a method for monitoring the operation of an equipment state is provided, comprising:

[0034] Obtaining the operating status data of the equipment to be monitored at multiple predetermined time points within a predetermined time period;

[0035] Arranging the operation status data of the plurality of predetermined time points into an operation status data input vector according to a time dimension;

[0036] The operation status data input vector is passed through a first convolutional neural network model having a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector;

[0037] The operation status data input vector is passed through a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector;

[0038] fusing the first-scale running state feature vector and the second-scale running state feature vector to obtain a multi-scale running state feature vector;

[0039] Performing feature-level data enhancement on the multi-scale operating state feature vector based on the Gaussian density map to obtain a classification feature matrix; and

[0040] The classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the operating status of the device to be monitored is normal.

[0041] In the above-mentioned equipment status operation monitoring method, the operation status data input vector is obtained by using a first convolutional neural network model with a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector, including: by using each layer of the first convolutional neural network model with a one-dimensional convolution kernel of a first scale, in the forward transfer of the layer, respectively performing convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolution kernel of the first scale is the first-scale operation status feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolution kernel of the first scale is the operation status data input vector.

[0042] In the above-mentioned equipment status operation monitoring method, the operation status data input vector is obtained by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector, including: by using each layer of the second convolutional neural network model with a one-dimensional convolution kernel of the second scale, in the forward pass of the layer, respectively performing convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolution kernel of the second scale is the second-scale operation status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolution kernel of the second scale is the operation status data input vector.

[0043] Compared with the prior art, the present application provides a device and method for monitoring the operation of equipment status, which first arranges the operation status data of the plurality of predetermined time points according to the time dimension as an operation status data input vector and then respectively inputs the first convolutional neural network model and the second convolutional neural network model to obtain a first-scale operation status feature vector and a second-scale operation status feature vector, then fuses the first-scale operation status feature vector and the second-scale operation status feature vector to obtain a multi-scale operation status feature vector, then performs feature-level data enhancement on the multi-scale operation status feature vector based on a Gaussian density map to obtain a classification feature matrix, and finally passes the classification feature matrix through a classifier to obtain a classification result for indicating whether the operation status of the equipment to be monitored is normal. In this way, the operation status of the equipment to be monitored can be intelligently monitored. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0045] Figure 1 This is an application scenario diagram of the equipment status operation monitoring device according to an embodiment of the present application.

[0046] Figure 2 It is a block diagram of a device status operation monitoring apparatus according to an embodiment of the present application.

[0047] Figure 3 It is a block diagram of the data enhancement module in the equipment status operation monitoring device according to an embodiment of the present application.

[0048] Figure 4 The present invention is a flowchart of a method for monitoring the operation of a device according to an embodiment of the present application.

[0049] Figure 5 Schematic diagram of the system architecture of the device status operation monitoring method according to an embodiment of the present application. DETAILED DESCRIPTION

[0050] 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.

[0051] Scenario Overview

[0052] The present application provides a device for monitoring the operation status of an equipment based on CAT.1 technology, which solves the problem that when the equipment fails, the equipment maintenance personnel cannot accurately determine the cause of the failure, and can only find the fault point through on-site observation and measurement, and then repair and troubleshoot the fault, which consumes too much manpower and material resources. It is based on CAT.1 equipment communication technology and equipment operation status monitoring technology. Specifically, CAT.1 is designed with a single antenna and low storage solution, with a simpler hardware architecture, using CAT.1 modules for communication, relying on operator base stations, with a wide signal coverage range, not limited by the installation environment, and no need to lay network cables. It has the characteristics of high transmission stability, high integration, and low power consumption. Among them, the design of the device operation status monitoring device based on CAT.1 technology: including smart street lights, main power cabinets and cloud platforms, the main power cabinet is equipped with smart energy meters and current and voltage acquisition modules, and the switch expansion module is electrically connected to the smart energy meter, the smart street light monitor and the current and voltage acquisition module, which is convenient for detecting current, voltage and collecting data for use. The status monitoring sensor group collects one or more operating status data when the equipment is running. The memory stores the operating status data collected by the monitoring sensor group and forwards it to the processor. The processor forwards the operating status data to the wireless data transmitter, which transmits it to the external data analysis terminal. The wireless data transmitter also receives control commands sent by the external data analysis terminal. The processor is equivalent to the storage controller in the memory, which is already a prior art. The processor deletes or retains the data in the memory according to the control command, releases the storage space of the memory, and ensures that the memory has sufficient storage space, which is conducive to subsequent monitoring.

[0053] The present application further provides a device positioning and installation structure design, in which a magnet is installed on the bottom surface of the device's installation base for magnetic attraction on the monitored device parts. The plane geometric shape of the magnet is a trapezoid, and a rectangular coordinate system is established with the centroid of the magnet as the origin. The shape is asymmetric in the Y-axis direction, i.e., the length direction, to avoid errors in the installation direction of the sensor.

[0054] The device operation status monitoring device based on CAT.1 technology provided in this application is convenient for monitoring the status of the device, and adopts CAT.1 device communication technology to achieve high transmission stability, high integration, and low power consumption. It is also convenient for transmitting data over long distances, and the transmission is faster and more convenient, reducing the cost of human labor, making it more intelligent and flexible, and realizing the core functions of smart city management such as smart lighting, smart transportation, and wireless cities in a big data interactive environment. Its response delay is less than 5s, dielectric strength is greater than 1.5kvac, power consumption is less than 50W, and transmission power is less than 50mW.

[0055] Accordingly, in the technical solution of the present application, the operating status data of the device to be monitored during operation is collected by the status monitoring sensor, the memory stores the operating status data collected by the monitoring sensor group and forwards it to the processor, the processor forwards the operating status data to the wireless data transmitter, and the wireless data transmitter transmits it to the external data analysis terminal. In particular, in the technical solution of the present application, the data analysis terminal is deployed with an equipment status operation monitoring algorithm, which can process the operating status data of the device to be monitored based on the equipment status operation monitoring algorithm to obtain a monitoring result indicating whether the operating status of the device to be monitored is normal.

[0056] Specifically, in the technical solution of the present application, the operating status data of the monitored device at multiple predetermined time points within a predetermined time period are first obtained, and then the operating status data at the multiple predetermined time points are arranged according to the time dimension into an operating status data input vector.

[0057] Next, the operation status data input vector is passed through a first convolutional neural network model with a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector. At the same time, the operation status data input vector is passed through a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector. Furthermore, the first-scale operation status feature vector and the second-scale operation status feature vector are fused to obtain a multi-scale operation status feature vector. That is, in the technical solution of the present application, a neural network model based on deep learning is used as a feature extractor to capture the high-dimensional implicit correlation features contained in the operation status data.

[0058] Compared with traditional feature engineering, the neural network model based on deep learning can simulate or approximate any nonlinear function through learnable neural network parameters to fit a functional module with specific functions. That is, in the technical solution of the present application, the first convolutional neural network model and the second convolutional neural network model that have been trained perform one-dimensional convolution encoding of different scales on the operating status data input vector through a learnable neural network parameter model to capture the high-dimensional implicit pattern characteristics of the distribution of the operating status data in different time spans in the operating status data input vector, so as to obtain the first-scale operating status feature vector and the second-scale operating status feature vector. And further, the first-scale operating status feature vector and the second-scale operating status feature vector are fused to obtain the multi-scale operating status feature vector.

[0059] Considering that the multi-scale operating state feature vector is obtained from the operating state data of the multiple predetermined time points, although the amount of data can be expanded by extending the time length, the sparse data on the source domain end will affect the accuracy of the subsequent feature distribution expression. Therefore, in the technical solution of the present application, the multi-scale operating state feature vector is subjected to feature-level data enhancement based on the Gaussian density map to obtain a classification feature matrix. A person of ordinary skill in the art should know that the Gaussian density map is the target distribution of deep learning. Therefore, the multi-scale operating state feature vector can be subjected to feature-level data enhancement using the Gaussian density map as the data form of the posterior distribution to obtain the classification feature matrix. Specifically, a Gaussian density map of the multi-scale operating state feature vector is first constructed, the mean vector of the Gaussian density map is the multi-scale operating state feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale operating state feature vector; then, the Gaussian distribution of each position of the Gaussian density map is randomly sampled to obtain multiple row vectors, and the multiple row vectors are arranged in two dimensions to obtain the classification feature matrix.

[0060] Then, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the operating status of the device to be monitored is normal. That is, the classifier is used to determine the class probability label to which the classification feature matrix belongs, wherein the class probability label includes the normal operating status of the device to be monitored (first label) and the abnormal operating status of the device to be monitored (second label). Accordingly, after obtaining the classification result, it can be indicated whether the operating status of the device to be monitored is normal based on the classification result, so that the device dimension and fault warning can be performed based on the classification result.

[0061] In particular, in the technical solution of the present application, when fusing the first-scale operating state feature vector and the second-scale operating state feature vector to obtain the multi-scale operating state feature vector, considering that the first-scale operating state feature vector and the second-scale operating state feature vector respectively express the temporal correlation characteristics of the operating state data at different scales, if they are directly fused in a manner such as point addition, there may be a problem of poor convergence of the overall feature distribution of the multi-scale operating state feature vector, which will lead to poor fitting effect of the classifier. On the other hand, when classifying the multi-scale operating state feature vector, if the correlation between the eigenvalues ​​of the multi-scale operating state feature vector is high, the classification accuracy will be reduced. Since feature-level data enhancement of the multi-scale operating state feature vector based on the Gaussian density map cannot solve the above problem, it is necessary to optimize the multi-scale operating state feature vector.

[0062] Therefore, the first scale running state feature vector and the second scale operating state feature vector The Hilbert space constraint of the vector modulus basis is performed to obtain the multi-scale operating state feature vector, which is expressed as:

[0063]

[0064] Represents a one-dimensional convolution operation, that is, the convolution operator Pair Vector Perform a one-dimensional convolution.

[0065] That is, the fused multi-scale running state feature vector is constrained by using a convolution operator in a Hilbert space that defines the inner product of the vector sum modulus and the vector, so as to convert the fused multi-scale running state feature vector The characteristic distribution of is limited to a finite closed domain in the Hilbert space based on the modulus of the vector, and the fused multi-scale running state characteristic vector is improved. The orthogonality between the basis dimensions of the high-dimensional manifold of the feature distribution is improved, thereby achieving sparse correlation between eigenvalues ​​while maintaining the overall convergence of the feature distribution. In this way, when the multi-scale operating state feature vector is further enhanced at the feature level based on the Gaussian density map to obtain the classification feature matrix, the fitting effect and classification accuracy of the classification feature matrix through the classifier can be improved.

[0066] Based on this, the present application provides an equipment status operation monitoring device, which includes: an operation data acquisition module, used to obtain the operation status data of the equipment to be monitored at multiple predetermined time points within a predetermined time period; a vectorization module, used to arrange the operation status data of the multiple predetermined time points into an operation status data input vector according to the time dimension; a first scale feature extraction module, used to obtain a first scale operation status feature vector by using a first convolutional neural network model with a one-dimensional convolution kernel of a first scale; a second scale feature extraction module, used to obtain a second scale operation status feature vector by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale; a multi-scale fusion module, used to fuse the first scale operation status feature vector and the second scale operation status feature vector to obtain a multi-scale operation status feature vector; a data enhancement module, used to perform feature-level data enhancement on the multi-scale operation status feature vector based on a Gaussian density map to obtain a classification feature matrix; and a monitoring result generation module, used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation status of the equipment to be monitored is normal.

[0067] Figure 1 FIG. 1 is an application scenario diagram of the device status operation monitoring device according to an embodiment of the present application. Figure 1 As shown, in this application scenario, the operating status data of the device to be monitored at multiple predetermined time points within a predetermined time period is obtained (for example, Figure 1 Then, the operation status data of the plurality of predetermined time points are input into a server deployed with an equipment status operation monitoring algorithm (for example, Figure 1 S), wherein the server can use the device status operation monitoring algorithm to process the operation status data of the multiple predetermined time points to obtain a classification result indicating whether the operation status of the device to be monitored is normal.

[0068] 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.

[0069] Exemplary Devices

[0070] Figure 2 FIG. 1 is a block diagram of a device status operation monitoring apparatus according to an embodiment of the present application. Figure 2 As shown, according to the embodiment of the present application, the device state operation monitoring device 100 includes: an operation data acquisition module 110, which is used to obtain the operation state data of the device to be monitored at multiple predetermined time points within a predetermined time period; a vectorization module 120, which is used to arrange the operation state data of the multiple predetermined time points into an operation state data input vector according to the time dimension; a first scale feature extraction module 130, which is used to obtain a first scale operation state feature vector by using a first convolution neural network model with a one-dimensional convolution kernel of a first scale; a second scale feature extraction module 140, which is used to extract the operation state data input vector The vector is obtained by using a second convolution neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operating state feature vector; a multi-scale fusion module 150 is used to fuse the first-scale operating state feature vector and the second-scale operating state feature vector to obtain a multi-scale operating state feature vector; a data enhancement module 160 is used to perform feature-level data enhancement on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix; and a monitoring result generation module 170 is used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operating state of the device to be monitored is normal.

[0071] More specifically, in the embodiment of the present application, the operation data acquisition module 110 is used to obtain the operation status data of the monitored equipment at multiple predetermined time points within a predetermined time period.

[0072] Accordingly, in a specific example, the operating status data is the operating power value of the device to be monitored.

[0073] More specifically, in the embodiment of the present application, the vectorization module 120 is used to arrange the operating status data of the multiple predetermined time points into an operating status data input vector according to the time dimension.

[0074] Compared with traditional feature engineering, the neural network model based on deep learning can simulate or approximate any nonlinear function through learnable neural network parameters to fit a functional module with specific functions. That is, in the technical solution of the present application, the first convolutional neural network model and the second convolutional neural network model that have been trained perform one-dimensional convolution encoding of different scales on the operating status data input vector through a learnable neural network parameter model to capture the high-dimensional implicit pattern characteristics of the distribution of the operating status data in different time spans in the operating status data input vector, so as to obtain the first-scale operating status feature vector and the second-scale operating status feature vector. That is, in the technical solution of the present application, a neural network model based on deep learning is used as a feature extractor to capture the high-dimensional implicit correlation features contained in the operating status data.

[0075] More specifically, in an embodiment of the present application, the first-scale feature extraction module 130 is used to obtain a first-scale running state feature vector by converting the running state data input vector into a first convolutional neural network model using a one-dimensional convolution kernel having a first scale.

[0076] Accordingly, in a specific example, the first-scale feature extraction module 130 is further used to: by using each layer of the first convolutional neural network model with a one-dimensional convolution kernel of the first scale to perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of the layer, so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the first-scale operating state feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the operating state data input vector.

[0077] More specifically, in an embodiment of the present application, the second-scale feature extraction module 140 is used to obtain a second-scale operating status feature vector by converting the operating status data input vector into a second convolutional neural network model using a one-dimensional convolution kernel with a second scale.

[0078] Accordingly, in a specific example, the second-scale feature extraction module 140 is further used to: by using each layer of the second convolutional neural network model with a one-dimensional convolution kernel of the second scale, perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of the layer so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the second-scale operating status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the operating status data input vector.

[0079] More specifically, in the embodiment of the present application, the multi-scale fusion module 150 is used to fuse the first-scale operating state feature vector and the second-scale operating state feature vector to obtain a multi-scale operating state feature vector.

[0080] In particular, in the technical solution of the present application, when fusing the first-scale operating state feature vector and the second-scale operating state feature vector to obtain the multi-scale operating state feature vector, considering that the first-scale operating state feature vector and the second-scale operating state feature vector respectively express the temporal correlation characteristics of operating state data at different scales, if they are directly fused in a manner such as point addition, there may be a problem of poor convergence of the overall feature distribution of the multi-scale operating state feature vector, which will lead to poor fitting effect of the classifier. On the other hand, when classifying the multi-scale operating state feature vector, if the correlation between the eigenvalues ​​of the multi-scale operating state feature vector is high, the classification accuracy will be reduced. Since feature-level data enhancement of the multi-scale operating state feature vector based on the Gaussian density map cannot solve the above problems, it is necessary to optimize the multi-scale operating state feature vector. Therefore, the first-scale operating state feature vector and the second scale operating state feature vector A Hilbert space constraint of a vector modulus basis is performed to obtain the multi-scale operating state feature vector.

[0081] Accordingly, in a specific example, the multi-scale fusion module 150 is further used to perform a vector modulus basis Hilbert space constraint on the first-scale operating state feature vector and the second-scale operating state feature vector using the following formula to obtain the multi-scale operating state feature vector; wherein the formula is:

[0082]

[0083] in, represents the first scale running state feature vector, represents the second scale running state feature vector, represents the transposed vector of the second scale running state feature vector, represents the multi-scale operating state feature vector, represents positional addition, represents the two-norm of a vector, Represents a one-dimensional convolution operation, that is, the convolution operator Pair Vector Perform a one-dimensional convolution.

[0084] That is, the fused multi-scale running state feature vector is constrained by using a convolution operator in a Hilbert space that defines the inner product of the vector sum modulus and the vector, so as to convert the fused multi-scale running state feature vector The characteristic distribution of is limited to a finite closed domain in the Hilbert space based on the modulus of the vector, and the fused multi-scale running state characteristic vector is improved. The orthogonality between the basis dimensions of the high-dimensional manifold of the feature distribution is improved, thereby achieving sparse correlation between eigenvalues ​​while maintaining the overall convergence of the feature distribution. In this way, when the multi-scale operating state feature vector is further enhanced at the feature level based on the Gaussian density map to obtain the classification feature matrix, the fitting effect and classification accuracy of the classification feature matrix through the classifier can be improved.

[0085] More specifically, in an embodiment of the present application, the data enhancement module 160 is used to perform feature-level data enhancement on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix. Considering that the multi-scale operating state feature vector is obtained from the operating state data of the multiple predetermined time points, although the amount of data can be expanded by extending the time length, the sparse data on the source domain side will affect the accuracy of subsequent feature distribution expression. Therefore, in the technical solution of the present application, feature-level data enhancement is performed on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix. Those of ordinary skill in the art should know that the Gaussian density map is the target distribution of deep learning. Therefore, the multi-scale operating state feature vector can be subjected to feature-level data enhancement using the Gaussian density map as the data form of the posterior distribution to obtain the classification feature matrix.

[0086] Accordingly, in a specific example, Figure 3As shown, the data enhancement module 160 includes: a Gaussian density map construction unit 161, which is used to construct a Gaussian density map of the multi-scale operating state feature vector, the mean vector of the Gaussian density map is the multi-scale operating state feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale operating state feature vector; a Gaussian discretization unit 162, which is used to randomly sample the Gaussian distribution of each position of the Gaussian density map to obtain a plurality of row vectors, and arrange the plurality of row vectors in two dimensions to obtain the classification feature matrix.

[0087] More specifically, in an embodiment of the present application, the monitoring result generation module 170 is used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operating status of the device to be monitored is normal. That is, the classifier is used to determine the class probability label to which the classification feature matrix belongs, wherein the class probability label includes the normal operating status of the device to be monitored (first label) and the abnormal operating status of the device to be monitored (second label). Accordingly, after obtaining the classification result, it can be indicated whether the operating status of the device to be monitored is normal based on the classification result, so that the device dimension and fault warning can be performed based on the classification result.

[0088] Accordingly, in a specific example, the monitoring result generation module 170 is further used to: expand the classification feature matrix into a classification feature vector according to a row vector or a column vector; use multiple fully connected layers of the classifier to fully connect the classification feature vector to obtain an encoded classification feature vector; and pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0089] In summary, the device state operation monitoring device 100 based on the embodiment of the present application is explained, which first arranges the operation state data of the multiple predetermined time points according to the time dimension as the operation state data input vector and then inputs them into the first convolutional neural network model and the second convolutional neural network model respectively to obtain the first scale operation state feature vector and the second scale operation state feature vector, then fuses the first scale operation state feature vector and the second scale operation state feature vector to obtain a multi-scale operation state feature vector, then, based on the Gaussian density map, the multi-scale operation state feature vector is subjected to feature-level data enhancement to obtain a classification feature matrix, and finally, the classification feature matrix is ​​passed through a classifier to obtain a classification result for indicating whether the operation state of the device to be monitored is normal. In this way, the operation state of the device to be monitored can be monitored intelligently.

[0090] As described above, the device state operation monitoring device 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server with a device state operation monitoring algorithm. In one example, the device state operation monitoring device 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the device state operation monitoring device 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the device state operation monitoring device 100 can also be one of the many hardware modules of the terminal device.

[0091] Alternatively, in another example, the device status operation monitoring apparatus 100 and the terminal device may also be separate devices, and the device status operation monitoring apparatus 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0092] Exemplary Methods

[0093] Figure 4 FIG. 1 is a flow chart of a method for monitoring the operation of a device according to an embodiment of the present application. Figure 4 As shown, according to the device status operation monitoring method of the embodiment of the present application, it includes: S110, obtaining the operation status data of the device to be monitored at multiple predetermined time points within a predetermined time period; S120, arranging the operation status data of the multiple predetermined time points into an operation status data input vector according to the time dimension; S130, obtaining a first-scale operation status feature vector by using a first convolutional neural network model with a one-dimensional convolution kernel of a first scale; S140, obtaining a second-scale operation status feature vector by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale; S150, fusing the first-scale operation status feature vector and the second-scale operation status feature vector to obtain a multi-scale operation status feature vector; S160, performing feature-level data enhancement on the multi-scale operation status feature vector based on a Gaussian density map to obtain a classification feature matrix; and, S170, passing the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation status of the device to be monitored is normal.

[0094] Figure 5 Schematic diagram of the system architecture of the device status operation monitoring method according to an embodiment of the present application. Figure 5As shown, in the system architecture of the device status operation monitoring method, first, the operation status data of the device to be monitored at multiple predetermined time points within a predetermined time period are obtained; then, the operation status data of the multiple predetermined time points are arranged according to the time dimension as an operation status data input vector; then, the operation status data input vector is passed through a first convolutional neural network model with a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector; then, the operation status data input vector is passed through a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector; then, the first-scale operation status feature vector and the second-scale operation status feature vector are fused to obtain a multi-scale operation status feature vector; then, feature-level data enhancement is performed on the multi-scale operation status feature vector based on a Gaussian density map to obtain a classification feature matrix; finally, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation status of the device to be monitored is normal.

[0095] In a specific example, in the above-mentioned device status operation monitoring method, the operation status data is the operating power value of the device to be monitored.

[0096] In a specific example, in the above-mentioned equipment status operation monitoring method, the operation status data input vector is obtained by using a first convolutional neural network model with a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector, including: by using each layer of the first convolutional neural network model with a one-dimensional convolution kernel of a first scale, in the forward pass of the layer, respectively performing convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the first-scale operation status feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the operation status data input vector.

[0097] In a specific example, in the above-mentioned equipment status operation monitoring method, the operation status data input vector is obtained by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector, including: by using each layer of the second convolutional neural network model with a one-dimensional convolution kernel of the second scale, in the forward pass of the layer, respectively performing convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the second-scale operation status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolution kernel with the second scale is the operation status data input vector.

[0098] In a specific example, in the above-mentioned device status operation monitoring method, the fusing the first-scale operation status feature vector and the second-scale operation status feature vector to obtain a multi-scale operation status feature vector further includes: performing a Hilbert space constraint of a vector modulus basis on the first-scale operation status feature vector and the second-scale operation status feature vector according to the following formula to obtain the multi-scale operation status feature vector; wherein the formula is:

[0099]

[0100] in, represents the first scale running state feature vector, represents the second scale running state feature vector, represents the transposed vector of the second scale running state feature vector, represents the multi-scale operating state feature vector, represents positional addition, represents the two-norm of a vector, Represents a one-dimensional convolution operation, that is, the convolution operator Pair Vector Perform a one-dimensional convolution.

[0101] In a specific example, in the above-mentioned equipment status operation monitoring method, the feature-level data enhancement of the multi-scale operation status feature vector based on the Gaussian density map to obtain a classification feature matrix includes: constructing a Gaussian density map of the multi-scale operation status feature vector, the mean vector of the Gaussian density map is the multi-scale operation status feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale operation status feature vector; randomly sampling the Gaussian distribution of each position of the Gaussian density map to obtain multiple row vectors, and arranging the multiple row vectors in two dimensions to obtain the classification feature matrix.

[0102] In a specific example, in the above-mentioned device status operation monitoring method, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation status of the device to be monitored is normal. It further includes: expanding the classification feature matrix into a classification feature vector according to a row vector or a column vector; using multiple fully connected layers of the classifier to fully connect encode the classification feature vector to obtain an encoded classification feature vector; and passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0103] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned device status operation monitoring method have been referred to above. Figures 1 to 3 The description of the equipment status operation monitoring system has been described in detail, and therefore, its repeated description will be omitted.

[0104] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0105] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0106] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0107] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0108] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A device for monitoring equipment status operation, characterized in that: include: An operation data acquisition module is used to obtain the operation status data of the equipment to be monitored at multiple predetermined time points within a predetermined time period; A vectorization module, used for arranging the operation status data of the plurality of predetermined time points into an operation status data input vector according to a time dimension; A first scale feature extraction module, configured to obtain a first scale running state feature vector by extracting the running state data input vector using a first convolutional neural network model having a one-dimensional convolution kernel of a first scale; A second scale feature extraction module, configured to extract the running state data input vector by using a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second scale running state feature vector; a multi-scale fusion module, configured to fuse the first-scale running state feature vector and the second-scale running state feature vector to obtain a multi-scale running state feature vector; A data enhancement module, used for performing feature-level data enhancement on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix; as well as A monitoring result generating module, used for passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the operating status of the device to be monitored is normal; The data enhancement module comprises: A Gaussian density map construction unit, used to construct a Gaussian density map of the multi-scale operating state feature vector, wherein the mean vector of the Gaussian density map is the multi-scale operating state feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of corresponding two positions in the multi-scale operating state feature vector; The Gaussian discretization unit is used to randomly sample the Gaussian distribution at each position of the Gaussian density map to obtain a plurality of row vectors, and to arrange the plurality of row vectors in two dimensions to obtain the classification feature matrix.

2. The equipment status operation monitoring device according to claim 1, characterized in that: The first-scale feature extraction module is further used to: perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of the layer by using each layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale, so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the first-scale operating status feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolution kernel with the first scale is the operating status data input vector.

3. The equipment status operation monitoring device according to claim 2, characterized in that: The second-scale feature extraction module is further used to: perform convolution processing based on the one-dimensional convolution kernel, mean pooling processing based on the feature matrix and activation processing on the input data in the forward pass of each layer of the second convolutional neural network model using the one-dimensional convolution kernel of the second scale, so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolution kernel of the second scale is the second-scale operating status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolution kernel of the second scale is the operating status data input vector.

4. The equipment status operation monitoring device according to claim 3 is characterized in that: The multi-scale fusion module is further used to perform a vector modulus basis Hilbert space constraint on the first-scale running state feature vector and the second-scale running state feature vector according to the following formula to obtain the multi-scale running state feature vector; Wherein, the formula is: in, represents the first scale running state feature vector, represents the second scale running state feature vector, represents the transposed vector of the second scale running state feature vector, represents the multi-scale operating state feature vector, represents positional addition, represents the two-norm of a vector, Represents a one-dimensional convolution operation, that is, the convolution operator Pair Vector Perform a one-dimensional convolution.

5. The equipment status operation monitoring device according to claim 4, characterized in that: The monitoring result generating module is further used for: Expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors; Performing full connection encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; as well as The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

6. The equipment status operation monitoring device according to claim 5, characterized in that: The operating status data is the operating power value of the device to be monitored.

7. A method for monitoring equipment status operation, characterized in that: include: Obtaining the operating status data of the equipment to be monitored at multiple predetermined time points within a predetermined time period; Arranging the operation status data of the plurality of predetermined time points into an operation status data input vector according to a time dimension; The operation status data input vector is passed through a first convolutional neural network model having a one-dimensional convolution kernel of a first scale to obtain a first-scale operation status feature vector; The operation status data input vector is passed through a second convolutional neural network model with a one-dimensional convolution kernel of a second scale to obtain a second-scale operation status feature vector; fusing the first-scale running state feature vector and the second-scale running state feature vector to obtain a multi-scale running state feature vector; Performing feature-level data enhancement on the multi-scale operating state feature vector based on a Gaussian density map to obtain a classification feature matrix; as well as Passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the operating status of the device to be monitored is normal; The performing feature-level data enhancement on the multi-scale operating state feature vector based on the Gaussian density map to obtain a classification feature matrix includes: Constructing a Gaussian density map of the multi-scale operating state feature vector, wherein the mean vector of the Gaussian density map is the multi-scale operating state feature vector, and the variance of each position in the covariance matrix of the Gaussian density map is the variance between the eigenvalues ​​of corresponding two positions in the multi-scale operating state feature vector; The Gaussian distribution at each position of the Gaussian density map is randomly sampled to obtain a plurality of row vectors, and the plurality of row vectors are arranged in two dimensions to obtain the classification feature matrix.

8. The device status operation monitoring method according to claim 7, characterized in that: The step of passing the running status data input vector through a first convolutional neural network model using a one-dimensional convolutional kernel having a first scale to obtain a first-scale running status feature vector comprises: performing convolution processing based on the one-dimensional convolutional kernel, mean pooling processing based on a feature matrix, and activation processing on the input data in the forward transfer of each layer of the first convolutional neural network model using the one-dimensional convolutional kernel having the first scale, so that the output of the last layer of the first convolutional neural network model using the one-dimensional convolutional kernel having the first scale is the first-scale running status feature vector, wherein the input of the first layer of the first convolutional neural network model using the one-dimensional convolutional kernel having the first scale is the running status data input vector.

9. The device status operation monitoring method according to claim 8, characterized in that: The step of passing the operating status input vector through a second convolutional neural network model using a one-dimensional convolutional kernel with a second scale to obtain a second-scale operating status feature vector comprises: performing convolution processing based on the one-dimensional convolutional kernel, mean pooling processing based on a feature matrix and activation processing on the input data in the forward transfer of each layer of the second convolutional neural network model using the one-dimensional convolutional kernel with the second scale, so that the output of the last layer of the second convolutional neural network model using the one-dimensional convolutional kernel with the second scale is the second-scale operating status feature vector, wherein the input of the first layer of the second convolutional neural network model using the one-dimensional convolutional kernel with the second scale is the operating status data input vector.

Images