An intelligent monitoring method and device for device state, an electronic device, and a storage medium

Abstract

The present application relates to the technical field of data processing, and particularly relates to a device state intelligent monitoring method and device, electronic equipment and storage medium, the method comprises the following steps: periodically determining a real-time observation scale value corresponding to a monitoring image, extracting a plurality of preset image features at a moment corresponding to each real-time observation scale value from the monitoring image, obtaining a current dynamic weight corresponding to each preset image feature under any real-time observation scale value according to a pre-constructed observation scale-feature weight relationship database, inputting the current dynamic weight of each preset image feature and the weighted sum of a plurality of preset image features under the corresponding real-time observation scale value into a preset fault prediction model after calculation, and obtaining a fault probability of a current state of a target dynamic device; the present application improves the accuracy and reliability of the fault prediction result by analyzing the dynamic weight of each preset image feature under different observation scales, and is beneficial to timely taking device maintenance measures.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, and storage medium for intelligent monitoring of equipment status. Background Technology

[0002] In modern port bulk cargo transport systems, the continuous and efficient transfer of materials such as iron ore and coal relies on the coordinated operation of a series of large-scale mechanical equipment, including tippers, belt conveyors, and ship loaders. Therefore, the health status of these equipment directly determines the smooth operation and safety of the entire port logistics chain. Real-time and accurate monitoring of equipment operating status, and the subsequent fault prediction and coordinated scheduling based on this data, are crucial.

[0003] Currently, the mainstream monitoring methods in this field mainly rely on physical sensors installed on the equipment, such as vibration sensors and temperature sensors. These sensors collect data and display it in real-time on a web page, triggering alarms. However, due to the insufficient reliability of sensor monitoring, the continuous vibration of the equipment itself, coupled with the open-air environment of ports, the long-term stability and accuracy of precision instruments such as vibration sensors are severely affected, leading to significant errors or even omissions in the monitoring data. Furthermore, sensor monitoring cannot effectively capture some visually visible faults in the equipment, such as macroscopic cracks in key structural components or deformation of parts, requiring the use of visual analysis models for monitoring. For equipment that continuously changes dynamically during operation, existing visual analysis models typically use fixed feature weights, which cannot adapt to such scale changes. This causes features that are effective at close range to become noise at longer distances, severely impacting the accuracy and reliability of fault prediction. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent monitoring method, device, electronic equipment, and storage medium for equipment status. By analyzing the dynamic weights of each preset image feature under different observation scales, the accuracy and reliability of fault prediction results are improved, which is beneficial for timely equipment maintenance measures.

[0005] According to a first aspect of the present invention, a method for intelligent monitoring of device status is provided, comprising the following steps:

[0006] The system acquires real-time monitoring images of the target dynamic device and periodically determines the real-time observation scale value corresponding to the monitoring image; the real-time observation scale value is the value of the target dynamic device under a preset observation dimension calculated based on the monitoring image.

[0007] Extract several preset image features corresponding to each real-time observation scale value from the monitoring image; the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values.

[0008] For any real-time observation scale value, the current dynamic weight corresponding to each preset image feature is obtained according to the pre-constructed observation scale-feature weight relation library; the observation scale-feature weight relation library stores the mapping functions between the pre-fitted observation scale value and the weight of each preset image feature.

[0009] The current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value are weighted and the calculation result is input into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic device.

[0010] According to a second aspect of the present invention, an intelligent monitoring device for equipment status is provided, the device comprising:

[0011] The scale determination module is used to acquire monitoring images of the target dynamic device in real time and periodically determine the real-time observation scale value corresponding to the monitoring image; the real-time observation scale value is the value of the target dynamic device under the preset observation dimension calculated based on the monitoring image.

[0012] The feature extraction module is used to extract several preset image features corresponding to each real-time observation scale value from the monitoring image; the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values.

[0013] The weight acquisition module is used to obtain the current dynamic weight corresponding to each preset image feature for any real-time observation scale value according to the pre-constructed observation scale-feature weight relationship library; the observation scale-feature weight relationship library stores the mapping functions between the pre-fitted observation scale value and the weight of each preset image feature.

[0014] The calculation module is used to calculate the weighted sum of the current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value, and input the calculation result into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic device.

[0015] According to a third aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or the at least one program being loaded and executed by a processor to implement the above-described intelligent monitoring method for device status.

[0016] According to a fourth aspect of the present invention, an electronic device is provided, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0017] The present invention has at least the following beneficial effects:

[0018] This invention provides an intelligent monitoring method for equipment status. First, it periodically determines the real-time observation scale value corresponding to the monitoring image and extracts several preset image features corresponding to each real-time observation scale value from the monitoring image. This method uses image acquisition to obtain monitoring features that are difficult for sensors to obtain, enabling timely prediction of the failure probability of the target dynamic equipment based on these acquired feature data. Then, it obtains the current dynamic weight corresponding to each preset image feature under any real-time observation scale value according to a pre-built observation scale-feature weight relationship library. The current dynamic weight of each preset image feature and several preset image features under the corresponding real-time observation scale value are weighted and input into a preset fault prediction model to obtain the failure probability of the target dynamic equipment's current state. Different dynamic feature weights are set for different observation scale values, and by analyzing the dynamic weights of each preset image feature under different observation scales, the accuracy and reliability of the fault prediction results are improved, facilitating timely equipment maintenance measures. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of the intelligent device status monitoring method provided in the embodiments of the present invention;

[0021] Figure 2 This is a schematic diagram of the intelligent monitoring device for device status provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention provides a method for intelligent monitoring of equipment status, such as... Figure 1 As shown, the method includes the following steps:

[0024] S100 acquires real-time monitoring images of the target dynamic equipment and periodically determines the real-time observation scale value corresponding to the monitoring images. In specific implementations, several monitoring cameras are installed and deployed within the target area as image acquisition devices. These cameras monitor the target dynamic equipment in real time. The target dynamic equipment can be equipment such as ship loaders, tippers, and belt conveyors. For example, a ship loader operates in motion through cantilever extension drive, cantilever pitch drive, and cantilever belt drive. Furthermore, the determination period for the real-time observation scale value corresponding to the monitoring images is on the order of seconds, for example, 3-5 seconds.

[0025] Specifically, the real-time observation scale value is the value of the target dynamic device under the preset observation dimension, calculated based on the monitoring image.

[0026] Furthermore, the preset observation dimension can be any one of the following: the imaging scale of the target dynamic device in the monitoring image, the physical distance between the target dynamic device and the image acquisition device corresponding to the monitoring image, or the movement amplitude of the target dynamic device in the monitoring image. The preset observation dimension is determined based on the activity type of the target dynamic device. For example, when the activity type of the target dynamic device is a mobile device, the preset observation dimension uses either the imaging scale or the physical distance; when the target dynamic device is a device with dynamically operating local components, the preset observation dimension uses the movement amplitude of the dynamic components. In specific implementations, when the preset observation dimension is the imaging scale, it can be calculated based on the ratio between a known-sized reference object in the image and the imaging pixel size. When the preset observation dimension is the physical distance or the movement amplitude, it can be obtained from a positioning and ranging sensor mounted on the target dynamic device or the image acquisition device. Those skilled in the art are familiar with the specific implementation methods for obtaining observation scale values ​​under different preset observation dimensions, and these will not be elaborated upon here.

[0027] As mentioned above, considering that the sensors have low accuracy in monitoring dynamic devices, monitoring image acquisition is adopted as another monitoring method. By acquiring the activity of the target dynamic device in the monitoring image, the monitoring accuracy can be improved. Furthermore, by periodically acquiring the device's working status, the consumption of computing resources can be reduced.

[0028] S200 extracts several preset image features corresponding to each real-time observation scale value from the monitoring image; this can be understood as: periodically extracting preset image features, and the extraction time point and time interval are consistent with the acquisition time of the real-time observation scale value.

[0029] Specifically, the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values.

[0030] Furthermore, the plurality of preset image features include at least one of the following: texture features of the surface state of the target dynamic device, key point spatial features of the relative positional relationship between the components of the target dynamic device, and motion features of the motion state of the target dynamic device. For example, the texture feature may be a surface crack of the target dynamic device. Since the observation scale value changes with the clarity of the monitoring image, the feature value corresponding to the collected surface crack will change dynamically.

[0031] In a specific implementation, when the texture feature is a surface crack in the target dynamic device, the area to be detected on the device surface is located in the image, the gray-level co-occurrence matrix of the area to be detected is calculated, and the contrast of the area to be detected is calculated based on the gray-level co-occurrence matrix. A higher contrast value indicates a deeper texture or groove. When the preset image feature is a key point spatial feature, a key point detection algorithm is used to locate the connection points of key components of the target dynamic device, and the geometric relationship formed between the connection points of key components is calculated. For example, the angle formed between the cantilever and the ship's fuselage during cantilever pitch should be within a preset safety angle range. If the angle exceeds the preset safety angle range... This indicates a precursor to instability or mechanical failure. Motion characteristics can be obtained through optical flow methods. In continuous monitoring images, optical flow is used to calculate the motion vector of each pixel in a specific area, such as the size and direction of the cantilever end. Statistical analysis is performed on the motion vector information to obtain the standard deviation of the motion vector amplitude and the rate of change of the main motion direction. The standard deviation of the motion vector amplitude is used to describe the severity of the shaking. The standard deviation is very small during stable motion, but it increases sharply when abnormal shaking occurs. If irregular up-and-down shaking or swaying occurs, the change in motion direction is large, and the rate of change of the main motion direction will increase.

[0032] As mentioned above, when analyzing the faults of target dynamic equipment, for monitoring features that are not easily obtained by sensors, image acquisition is used, and the feature values ​​corresponding to these monitoring features are extracted from the monitoring images. Based on the acquired data, the probability of fault occurrence or even the type of fault of the target dynamic equipment can be predicted, which is conducive to the timely detection of equipment faults.

[0033] S300, for any real-time observation scale value, obtain the current dynamic weight corresponding to each preset image feature according to the pre-constructed observation scale-feature weight relationship library; the observation scale-feature weight relationship library stores the mapping functions of the pre-fitted observation scale value and the weight of each preset image feature.

[0034] In a specific embodiment, the mapping function between the observed scale value and the weight of each preset image feature is fitted by the following steps:

[0035] S10, acquire historical image data of the target dynamic device at several preset observation scale values, as well as the device status label at the corresponding time; the device status label includes normal status and fault status; it can be understood that: acquire several historical monitoring images at each preset observation scale value.

[0036] S20, extract several sets of preset image features corresponding to each preset observation scale value from historical image data; this can be understood as: each historical monitoring image corresponds to a set of preset image features.

[0037] S30: Using each preset observation scale value and several sets of preset image features corresponding to each preset observation scale value as joint input, and the corresponding equipment status label as the supervision target, a fault analysis model is trained, and the weight of each preset image feature corresponding to each preset observation scale value is obtained based on the trained fault analysis model.

[0038] Furthermore, the step of obtaining the weight of each preset image feature corresponding to each preset observation scale value based on the trained fault analysis model includes the following steps:

[0039] S31, based on the trained fault analysis model, adopts a gradient-based feature importance calculation method to calculate the degree of influence of the change of each preset image feature on the output result of the fault analysis model when the input of the fault analysis model is each preset observation scale value and several sets of preset image features corresponding to each preset observation scale value; it can be understood as: for each preset observation scale value, the weight of a set of preset image features corresponding to the preset observation scale value itself is obtained.

[0040] Specifically, step S31 includes the following steps:

[0041] S311, for any preset observation scale value, based on several sets of preset image features corresponding to the preset observation scale value, obtain the feature mean value corresponding to each preset image feature.

[0042] S312, using the feature mean as a benchmark, calculates the absolute value of the gradient of the fault analysis model's output with respect to each preset image feature. In specific implementations, the partial derivative of the fault analysis model's output with respect to each preset image feature is used as the gradient value with respect to each preset image feature.

[0043] S313, normalize the absolute value of the gradient corresponding to each preset image feature, and use the normalized value corresponding to each preset image feature as the degree of influence of the preset image feature itself on the output result of the fault analysis model.

[0044] S32, quantify the influence of the change of each preset image feature on the output of the fault analysis model into the weight of each preset image feature, so as to obtain the weight of each preset image feature corresponding to each preset observation scale value.

[0045] In a specific embodiment, the current dynamic weight corresponding to each preset image feature is obtained through the following steps:

[0046] S301, for any real-time observation scale value, obtain the weights of several preset image features corresponding to the real-time observation scale value based on the mapping function between the observation scale value and the weights of each preset image feature.

[0047] S302, normalize the weights of several preset image features corresponding to the real-time observation scale value to obtain the current dynamic weight corresponding to each preset image feature; this can be understood as: normalizing the weights of several preset image features corresponding to the real-time observation scale value to between 0 and 1, and the sum of the weights of several preset image features is 1.

[0048] S40, for any preset image feature, and for different preset observation scale values ​​corresponding to different weights of the preset image feature, fit a mapping function between the observation scale value and the weight of the preset image feature.

[0049] As mentioned above, the distance between the target dynamic device and the monitoring camera varies at different observation scales, resulting in different clarity of the monitoring images. Consequently, the preset image feature values ​​extracted from different monitoring images are inconsistent. Inconsistent feature values ​​can affect fault judgment. For example, for cracks in equipment, when the equipment is far away, the collected texture feature values ​​are smaller, but they need to be given a larger weight. Therefore, by using the above method, different dynamic feature weights are set for different observation scale values, making the final fault analysis results more accurate and facilitating timely maintenance measures.

[0050] S400 calculates the weighted sum of the current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value, and inputs the calculation result into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic device.

[0051] Preferably, the preset fault prediction model embeds a Sigmoid function; this can be understood as using the Sigmoid function as the fault probability prediction function in the trained preset fault prediction model.

[0052] As mentioned above, the real-time operating status of the target dynamic device can be obtained from the monitoring images. Based on the current dynamic weight, the fault situation can be analyzed in real time to obtain the fault probability, which is conducive to timely detection of faults and taking maintenance measures. In addition, it can also play a verification role when the sensor detects abnormal conditions, ensuring the accuracy and reliability of fault warning.

[0053] In another embodiment, the method further includes the following steps:

[0054] P100 is based on extracting several preset image features corresponding to each real-time observation scale value from the monitoring image, and simultaneously obtaining at least one auxiliary feature; the auxiliary feature refers to a preset feature whose corresponding feature value is independent of different observation scale values.

[0055] Furthermore, the auxiliary features include at least one of the following: equipment temperature characteristics obtained through infrared images or temperature sensors; equipment vibration frequency characteristics obtained through vibration sensors; and cumulative operating time or number of operating cycles of the equipment obtained from a preset equipment control system. For example, when the moving parts of the equipment vibrate abnormally, the vibration sensor can detect abnormal vibration frequency; when the equipment experiences mechanical or drive failure, the local temperature of the equipment will rise abnormally.

[0056] P200 adds auxiliary features as new preset image features to several preset image features; this can be understood as follows: when using preset image features to obtain the mapping function between the observation scale value and the weight of each preset image feature, and when training the fault analysis model, auxiliary features are used.

[0057] As mentioned above, auxiliary features are added to the preset image features. For example, when a drive failure occurs in the cantilever of a ship loader, it may be a normal operation adjustment or a sign of mechanical failure. The model's judgment result may be inaccurate. However, after introducing auxiliary features, if the temperature also rises abnormally at the same time, the accuracy of fault judgment is greatly improved. That is, by introducing auxiliary features that cannot be obtained from the monitoring images, a complementary and mutually verifying effect is formed with the preset image features, thereby reducing false alarms and false negatives in fault prediction and improving the accuracy of fault prediction.

[0058] Embodiments of the present invention also provide an intelligent device for monitoring device status, such as... Figure 2 As shown, the device includes:

[0059] The scale determination module 100 is used to acquire monitoring images of the target dynamic device in real time and periodically determine the real-time observation scale value corresponding to the monitoring image; the real-time observation scale value is the value of the target dynamic device under a preset observation dimension calculated based on the monitoring image.

[0060] The feature extraction module 200 is used to extract several preset image features corresponding to each real-time observation scale value from the monitoring image; the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values.

[0061] The weight acquisition module 300 is used to acquire the current dynamic weight corresponding to each preset image feature for any real-time observation scale value according to a pre-constructed observation scale-feature weight relationship library; the observation scale-feature weight relationship library stores the mapping functions between the pre-fitted observation scale value and the weight of each preset image feature.

[0062] The calculation module 400 is used to calculate the weighted sum of the current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value, and input the calculation result into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic device.

[0063] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0064] Embodiments of the present invention also provide a non-transitory computer-readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a method in the method embodiments, wherein the at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided in the above embodiments.

[0065] Embodiments of the present invention also provide an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0066] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. It should also be understood that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.

Claims

1. A method for intelligent monitoring of equipment status, characterized in that, The method includes the following steps: The system acquires real-time monitoring images of a target dynamic device and periodically determines the real-time observation scale value corresponding to the monitoring image. The real-time observation scale value is the value of the target dynamic device calculated based on the monitoring image under a preset observation dimension. The preset observation dimension is any one of the following: the imaging scale of the target dynamic device in the monitoring image, the physical distance between the target dynamic device and the image acquisition device corresponding to the monitoring image, and the movement amplitude of the target dynamic device in the monitoring image. The preset observation dimension is determined according to the activity type of the target dynamic device. Extract several preset image features corresponding to each real-time observation scale value from the monitoring image; the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values. For any real-time observation scale value, the current dynamic weight corresponding to each preset image feature is obtained according to the pre-constructed observation scale-feature weight relationship library; the observation scale-feature weight relationship library stores the mapping functions between the pre-fitted observation scale value and the weight of each preset image feature. The current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value are weighted and the calculation result is input into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic equipment. The current dynamic weights corresponding to each preset image feature are obtained through the following steps: For any real-time observation scale value, based on the mapping function between the observation scale value and the weight of each preset image feature, the weights of several preset image features corresponding to the real-time observation scale value are obtained. The weights of several preset image features corresponding to the real-time observation scale value are normalized to obtain the current dynamic weights corresponding to each preset image feature.

2. The intelligent equipment status monitoring method according to claim 1, characterized in that, The plurality of preset image features include at least one of the following: texture features of the surface state of the target dynamic device, key point spatial features of the relative positional relationship between the components of the target dynamic device, and motion features of the motion state of the target dynamic device.

3. The intelligent equipment status monitoring method according to claim 1, characterized in that, The following steps are used to fit a mapping function between the observed scale values ​​and the weights of each preset image feature: Acquire historical image data of the target dynamic device at several preset observation scale values, as well as the device status labels at corresponding times; the device status labels include normal status and fault status. Extract several sets of preset image features corresponding to each preset observation scale value from historical image data; Using each preset observation scale value and several sets of preset image features corresponding to each preset observation scale value as joint input, and the corresponding equipment status label as supervision target, a fault analysis model is trained, and the weight of each preset image feature corresponding to each preset observation scale value is obtained based on the trained fault analysis model. For any preset image feature, and for different preset observation scale values ​​corresponding to different weights of the preset image feature, a mapping function between the observation scale value and the weight of the preset image feature is fitted.

4. The intelligent equipment status monitoring method according to claim 3, characterized in that, The process of obtaining the weights of each preset image feature corresponding to each preset observation scale value based on the trained fault analysis model includes the following steps: Based on the trained fault analysis model, a gradient-based feature importance calculation method is used to calculate the degree of influence of the change of each preset image feature on the output of the fault analysis model when the input of the fault analysis model is each preset observation scale value and several sets of preset image features corresponding to each preset observation scale value. The influence of changes in each preset image feature on the output of the fault analysis model is quantified into the weight of each preset image feature, so as to obtain the weight of each preset image feature corresponding to each preset observation scale value.

5. The intelligent equipment status monitoring method according to claim 1, characterized in that, The method further includes the following steps: Based on extracting several preset image features corresponding to each real-time observation scale value from the monitoring image, at least one auxiliary feature is also obtained; the auxiliary feature refers to a preset feature whose corresponding feature value is independent of different observation scale values. The auxiliary features are added as new preset image features to several preset image features.

6. The intelligent equipment status monitoring method according to claim 5, characterized in that, The auxiliary features include at least one of the following: Device temperature characteristics obtained through infrared images or temperature sensors; Vibration frequency characteristics of the equipment obtained through vibration sensors; The cumulative working time or number of working cycles of the equipment obtained from the preset equipment control system.

7. The intelligent equipment status monitoring method according to claim 1, characterized in that, The preset fault prediction model incorporates a Sigmoid function.

8. An intelligent monitoring device for equipment status, characterized in that, The device includes: The scale determination module is used to acquire monitoring images of the target dynamic device in real time and periodically determine the real-time observation scale value corresponding to the monitoring image. The real-time observation scale value is the value of the target dynamic device under a preset observation dimension calculated based on the monitoring image. The preset observation dimension is any one of the following: the imaging scale of the target dynamic device in the monitoring image, the physical distance between the target dynamic device and the image acquisition device corresponding to the monitoring image, and the movement amplitude of the target dynamic device in the monitoring image. The preset observation dimension is determined according to the activity type of the target dynamic device. The feature extraction module is used to extract several preset image features corresponding to each real-time observation scale value from the monitoring image; the preset image features refer to preset features whose corresponding feature values ​​can dynamically change with different observation scale values. The weight acquisition module is used to obtain the current dynamic weight corresponding to each preset image feature for any real-time observation scale value according to the pre-constructed observation scale-feature weight relationship library; the observation scale-feature weight relationship library stores the mapping functions between the pre-fitted observation scale value and the weight of each preset image feature. The calculation module is used to calculate the weighted sum of the current dynamic weight corresponding to each preset image feature and several preset image features under the corresponding real-time observation scale value, and input the calculation result into the preset fault prediction model to obtain the fault probability of the current state of the target dynamic device. The weight acquisition module is further used for: For any real-time observation scale value, based on the mapping function between the observation scale value and the weight of each preset image feature, the weights of several preset image features corresponding to the real-time observation scale value are obtained. The weights of several preset image features corresponding to the real-time observation scale value are normalized to obtain the current dynamic weights corresponding to each preset image feature.

9. The intelligent equipment status monitoring device according to claim 8, characterized in that, The device further includes a function fitting module, which includes: The first acquisition unit is used to acquire historical image data of the target dynamic device at several preset observation scale values, as well as the device status labels at corresponding times; the device status labels include normal status and fault status. The feature extraction unit is used to extract several sets of preset image features corresponding to each preset observation scale value from historical image data; The weight acquisition unit is used to train the fault analysis model with each preset observation scale value and several sets of preset image features corresponding to each preset observation scale value as joint inputs and the corresponding equipment status label as the supervision target, and to obtain the weight of each preset image feature corresponding to each preset observation scale value based on the trained fault analysis model. The function fitting unit is used to fit a mapping function between the observation scale value and the weight of the preset image feature for any preset image feature and for different preset observation scale values ​​corresponding to different weights of the preset image feature.

10. A non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the intelligent monitoring method for device status as described in any one of claims 1-7.

11. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 10.

Images