Steady-state evaluation method for telescopic lift fans based on operating status data analysis

By comprehensively using technologies such as data acquisition, data transition authentication, image monitoring and feature fusion recognition, real-time and accurate steady-state evaluation of telescopic lifting fans is achieved, solving the problem of insufficient evaluation accuracy in the existing technology, and improving the flying car safety performance.

CN118839289BActive Publication Date: 2024-11-29AISHIEN POWER TECHNOLOGY (JIANGSU) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411311618.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-11-29
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

In the prior art, the steady-state evaluation accuracy of telescopic lifting fans is limited and the lack of comprehensive evaluation is caused by inaccurate steady-state evaluation.

Method used

Using a method based on operational state data analysis, a comprehensive, real-time and accurate assessment of the fan operating status is achieved through technical means such as monitoring data acquisition, data transition authentication, image monitoring and feature fusion recognition, real-time operation data prediction and consistency authentication.

Benefits of technology

The steady-state evaluation capability and evaluation accuracy of telescopic lifting fans are improved, ensuring the overall safety performance of flying cars.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118839289B_ABST
    Figure CN118839289B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for evaluating the steady state of a telescopic lifting fan based on the analysis of operating status data, and relates to data processing related fields. The method includes: interactively monitoring a data acquisition system to retrieve fan operating status data; performing data transition authentication on a historical operating data set; obtaining a visible light image set and a thermal infrared image set; extracting a fan blade feature set to train a double-layer feature selection fusion device; obtaining real-time visible light images and real-time thermal infrared images to perform feature fusion recognition; collecting real-time operating data to predict fan operation; performing consistency authentication on predicted operating data according to flight mission requirements; performing steady-state evaluation to generate a target steady-state evaluation result. The method solves the technical problem of the lack of comprehensive evaluation in the existing telescopic lifting fan steady-state evaluation, which leads to limited steady-state evaluation accuracy, and achieves the technical effect of improving the steady-state evaluation capability and evaluation accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to data processing related fields, and in particular to a method for evaluating the steady state of a telescopic lifting fan based on operating status data analysis. Background Art

[0002] With the rapid development of aviation technology, flying cars, as a new type of transportation that integrates the functions of cars and airplanes, are gradually moving from concept to practical application. This type of vehicle places extremely high demands on the stability and reliability of each system in a complex and changeable flight environment. Among them, the telescopic lifting fan, as an important component of the flying car, is responsible for providing the necessary lift and attitude adjustment capabilities during takeoff, landing and hovering in the air. The stability of its operating state is directly related to the overall safety performance of the flying car. In the current flying car technology system, the steady-state evaluation of the telescopic lifting fan mainly relies on traditional sensor monitoring and data analysis methods. These methods usually rely on only a few physical quantities for state judgment, and the evaluation results are relatively one-sided.

[0003] In the current related technologies, there is a lack of comprehensive evaluation in the steady-state evaluation of telescopic lifting fans, which leads to the technical problem of limited accuracy of steady-state evaluation. Summary of the invention

[0004] The present application provides a steady-state assessment method for a telescopic lifting fan based on operating status data analysis, and adopts technical means such as comprehensive use of monitoring data collection, data transition authentication, image monitoring and feature fusion recognition, real-time operating data prediction and consistency authentication to achieve a comprehensive, real-time and accurate assessment of the fan's operating status, thereby achieving the technical effect of improving the steady-state assessment capability and assessment accuracy.

[0005] The present application provides a method for evaluating the steady state of a telescopic lifting fan based on operating status data analysis, including:

[0006] The monitoring data acquisition system of the interactive target flying car retrieves the fan operation status data of the target flying car within a preset time window to obtain a historical operation data set; performs data transition authentication on the historical operation data set to obtain a first steady-state coefficient; retrieves the image acquisition data of the telescopic lifting fan of the target flying car by the dual-target image monitoring module within the preset time window to obtain a visible light image set and a thermal infrared image set; extracts the fan blade feature set of the telescopic lifting fan, combines the visible light image set and the thermal infrared image set to train a double-layer feature selection fusion device, and obtains the trained double-layer feature selection fusion device; obtains the current The invention relates to a method for realizing a real-time visible light image and a real-time thermal infrared image of the telescopic lifting fan at the current moment, and using the double-layer feature selection fusion device to perform feature fusion recognition on the real-time visible light image and the real-time thermal infrared image to generate a real-time fan blade feature; collecting the real-time operation data of the telescopic lifting fan at the current moment, and performing fan operation prediction in combination with the real-time fan blade feature to obtain predicted operation data; according to the flight mission requirements of the target flying car, performing consistency authentication on the predicted operation data to obtain a second steady-state coefficient; performing steady-state evaluation based on the first steady-state coefficient and the second steady-state coefficient to generate a target steady-state evaluation result of the telescopic lifting fan.

[0007] In a possible implementation, the following processing is performed:

[0008] Each historical operation data in the historical operation data set includes fan speed, fan power, operation temperature, amplitude and vibration frequency.

[0009] In a possible implementation, data transition authentication is performed on the historical operation data set to obtain a first steady-state coefficient, and the following processing is performed:

[0010] The historical operation data set is clustered according to the first transition bandwidth to obtain multiple historical operation data clusters; intra-cluster stability identification is performed on the multiple historical operation data clusters respectively to generate multiple first historical stability factors; the multiple historical operation data clusters are traversed to perform mean calculation to obtain multiple historical operation data means; stability identification is performed on the multiple historical operation data means respectively to obtain a second historical stability factor; and the first steady-state coefficient is generated by analyzing the second historical stability factor and the multiple first historical stability factors.

[0011] In a possible implementation, the following processing is performed:

[0012] Randomly extract the first historical operation data from the historical operation data set; calculate the similarity between the historical operation data set and the first historical operation data to generate multiple first historical similarities; add the historical operation data less than or equal to the first transition bandwidth among the multiple first historical similarities into a first historical operation data cluster; remove the first historical operation data cluster from the historical operation data set to obtain an updated historical operation data set; perform iterative clustering based on the updated historical operation data set to obtain the multiple historical operation data clusters.

[0013] In a possible implementation, the following processing is performed:

[0014] Call the double-layer feature selection fusion device, wherein the double-layer feature selection fusion device includes a double embedding layer and a backbone network; input the real-time visible light image and the real-time thermal infrared image into the visible light extraction branch and the thermal infrared extraction branch of the double embedding layer to obtain a real-time visible light feature set and a real-time thermal infrared feature set; input the real-time visible light feature set and the real-time thermal infrared feature set into the backbone network for feature fusion to generate the real-time fan blade feature.

[0015] In a possible implementation, the following processing is performed:

[0016] Feature extraction is performed on the visible light image set and the thermal infrared image set respectively to obtain a visible light feature extraction set and a thermal infrared feature extraction set; the visible light extraction branch and the thermal infrared extraction branch are trained respectively using the visible light image set, the thermal infrared image set, the visible light feature extraction set and the thermal infrared feature extraction set; the backbone network is generated based on the training of the visible light feature extraction set, the thermal infrared feature extraction set and the fan blade feature set.

[0017] In a possible implementation, the following processing is performed:

[0018] The initial state of the Markov chain is constructed based on the real-time operation data and the real-time fan blade characteristics; the state transfer matrix of the Markov chain is constructed based on the analysis of the historical operation data set; the initial state is substituted into the state transfer matrix to determine the predicted operation data.

[0019] In a possible implementation, the following processing is performed:

[0020] Constructing a steady-state assessment weight distribution layer; inputting the first steady-state coefficient and the second steady-state coefficient into the steady-state assessment weight distribution layer for weighted calculation to generate the target steady-state assessment result.

[0021] The method for evaluating the steady-state of a telescopic lifting fan based on operating status data analysis proposed in the present application first interacts with the monitoring data acquisition system of the target flying car, retrieves the fan operating status data of the target flying car within a preset time window, and obtains a historical operating data set. Then, data transition authentication is performed on the historical operating data set to obtain a first steady-state coefficient. Then, the image acquisition data of the telescopic lifting fan of the target flying car performed by the dual-target image monitoring module within the preset time window is retrieved to obtain a visible light image set and a thermal infrared image set. Then, the fan blade feature set of the telescopic lifting fan is extracted, and a double-layer feature selection fusion device is trained in combination with the visible light image set and the thermal infrared image set to obtain a trained double-layer feature. A fusion device is selected, and then the real-time visible light image and real-time thermal infrared image of the telescopic lifting fan at the current moment are obtained. The real-time visible light image and the real-time thermal infrared image are subjected to feature fusion recognition by using a double-layer feature selection fusion device to generate real-time fan blade features. Then, the real-time operation data of the telescopic lifting fan at the current moment is collected, and the fan operation is predicted in combination with the real-time fan blade features to obtain predicted operation data. Then, according to the flight mission requirements of the target flying car, the predicted operation data is consistency certified to obtain the second steady-state coefficient. Finally, a steady-state evaluation is performed based on the first steady-state coefficient and the second steady-state coefficient to generate a target steady-state evaluation result of the telescopic lifting fan, thereby achieving the technical effect of improving the steady-state evaluation capability and evaluation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the method according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0023] Figure 1 A schematic flow chart of a method for evaluating the steady state of a telescopic lifting fan based on operating status data analysis provided in an embodiment of the present application.

[0024] Figure 2 A schematic diagram of a process for obtaining multiple historical operation data clusters in a telescopic lifting fan steady-state assessment method based on operation status data analysis provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0026] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0027] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0028] The present application embodiment provides a method for evaluating the steady state of a telescopic lifting fan based on analysis of operating status data, such as Figure 1 As shown, the method includes:

[0029] Step S100, the monitoring data acquisition system of the interactive target flying car retrieves the fan operation status data of the target flying car within the preset time window to obtain a historical operation data set. Specifically, the target flying car is identified based on the unique identifier of the flying car (such as serial number, IP address, etc.), and the target flying car refers to a specific flying car that needs to be evaluated for the steady state of the telescopic lifting fan. A communication connection is established with the monitoring data acquisition system of the target flying car through a preset communication protocol (such as TCP / IP, MQTT, etc.) and port number, wherein the monitoring data acquisition system is a system responsible for collecting sensor data of the operating status of various components of the flying car (including the telescopic lifting fan), and performing preliminary processing and storage. According to the evaluation requirements, the time range for retrieving data is set, that is, the preset time window (a specific time period). A data request is sent to the monitoring data acquisition system to request the operating status data of the telescopic lifting fan of the target flying car within a preset time window, wherein the request includes the type (such as voltage, current, speed, temperature, etc.) and format of the required data, and the acquired data is stored in a local or remote database to form a historical operating data set. The historical operating data set is a set that stores the fan operating status data within the preset time window, which is used for subsequent data analysis and steady-state evaluation.

[0030] In a possible implementation, step S100 further includes step S110, and each historical operation data in the historical operation data set includes fan speed, fan power, operating temperature, amplitude and frequency. Specifically, this implementation refines the dimensions of data collection, wherein fan speed refers to the number of rotations of the fan blades per minute, which is used to evaluate the operating status of the fan; fan power refers to the amount of electrical energy or mechanical energy consumed when the fan is running, which is used to reflect the working intensity and efficiency of the fan; operating temperature refers to the temperature of the fan and its surroundings, and too high or too low temperature will affect the normal operation of the fan; amplitude refers to the vibration amplitude generated by the fan during operation, and too large amplitude indicates that the fan has a fault or imbalance problem; frequency refers to the frequency of fan vibration, which is used together with amplitude to evaluate the vibration characteristics of the fan. This implementation can more accurately evaluate the steady-state performance of the fan by collecting data in multiple dimensions, thereby achieving the technical effect of improving the comprehensiveness of obtaining the fan's operating status.

[0031] Step S200, perform data transition authentication on the historical operation data set to obtain a first steady-state coefficient. Specifically, remove noise data, abnormal values ​​or repeated records in the historical operation data set to ensure the accuracy and consistency of the data. According to the design specifications, historical operation experience or industry standards of the telescopic lifting fan, determine the stable range of various operating parameters (such as voltage, current, speed, temperature, etc.) of the telescopic lifting fan. These stable ranges define the performance indicators that the telescopic lifting fan should meet under normal working conditions. Traverse the historical operation data set and use statistical methods (such as sliding window standard deviation, threshold detection, etc.) to perform transition detection on the data at each moment. The transition refers to a significant change in the data value in a short period of time, which exceeds the normal fluctuation range. For the detected data transition point, further analysis and authentication are performed to determine whether it is caused by the unstable operation of the telescopic lifting fan itself. According to the detection results of data transition, the first steady-state coefficient is calculated, such as calculating the proportion of stable data points to the total data points, calculating the ratio of the number of transition points to the total number of data points and taking the inverse, etc. The first steady-state coefficient reflects the proportion or degree of the operating state of the telescopic lifting fan within the stable range within the preset time window.

[0032] In a possible implementation, data transition authentication is performed on the historical operation data set to obtain a first steady-state coefficient, and step S200 further includes step S210, clustering the historical operation data set according to the first transition bandwidth to obtain multiple historical operation data clusters. Specifically, according to the characteristics of the fan operation data, a threshold for clustering analysis (i.e., the first transition bandwidth, which represents the minimum change required for a data point to transition from one state to another) is defined to distinguish data points in different operating states. The historical operation data set is clustered using a clustering algorithm (such as K-means, DBSCAN, etc.), wherein the basis for clustering is the similarity between data points, i.e., whether the difference between data points is less than the first transition bandwidth. After clustering is completed, multiple historical operation data clusters are obtained, each cluster representing a relatively stable operating state. Step S220, intra-cluster stability identification is performed on each of the multiple historical operation data clusters to generate multiple first historical stability factors. Specifically, a stability index, such as variance, is selected to measure the degree of dispersion of data points within a data cluster. The variance is calculated for the data points within each historical operation data cluster. The smaller the variance, the more concentrated the data points, that is, the higher the stability within the cluster. The calculated variance is used as the first historical stability factor, and each historical operation data cluster corresponds to a first historical stability factor. Step S230, traverse the multiple historical operation data clusters to perform mean calculation to obtain multiple historical operation data means. Specifically, the mean is calculated for the data points within each historical operation data cluster to obtain the mean of each historical operation data cluster. Step S240, stability identification is performed on the multiple historical operation data means respectively to obtain a second historical stability factor. Specifically, by comparing the differences and change trends between the means, the stability between the means of each historical operation data cluster is evaluated. According to the evaluation results, a second historical stability factor is generated, and the second historical stability factor reflects the stability between the means of multiple historical operation data clusters. Step S250, the first steady-state coefficient is generated based on the analysis of the second historical stability factor and the multiple first historical stability factors. Specifically, the second historical stability factor is combined with multiple first historical stability factors for comprehensive analysis, such as weight allocation, weighted averaging, etc., and a first steady-state coefficient is generated based on the analysis results. The first steady-state coefficient comprehensively reflects the steady-state performance of the historical operation data set. This implementation method divides the data into multiple relatively stable data clusters through cluster analysis, and evaluates the intra-cluster stability and inter-cluster stability respectively, and meticulously analyzes the steady-state performance of the historical operation data set, achieving the technical effect of improving the accuracy and reliability of steady-state evaluation.

[0033] like Figure 2As shown, in a possible implementation, the historical operation data set is clustered according to the first transition bandwidth to obtain multiple historical operation data clusters, and step S210 further includes step S211, randomly extracting the first historical operation data from the historical operation data set. Specifically, using a random number generator or a similar method, randomly select a data point from the historical operation data set as the first historical operation data, and the first historical operation data is a data point used for the clustering starting point. Step S212, calculate the similarity between the historical operation data set and the first historical operation data, and generate multiple first historical similarities. Specifically, for each data point in the historical operation data set, calculate its similarity with the first historical operation data. The similarity calculation can use methods such as Euclidean distance, Manhattan distance, cosine similarity, etc., and the similarity between each data point and the first historical operation data calculated is used as the first historical similarity. Step S213, add the historical operation data that is less than or equal to the first transition bandwidth among the multiple first historical similarities into the first historical operation data cluster. Specifically, all first historical similarities are traversed, and data points corresponding to similarities less than or equal to the first transition bandwidth are added to the first historical operation data cluster. With the addition of data points, the first historical operation data cluster is gradually formed. Among them, the first transition bandwidth is a threshold for clustering analysis, which indicates the upper limit of the similarity between data points. If the threshold is exceeded, it is considered that the data points do not belong to the same cluster. Step S214, remove the first historical operation data cluster from the historical operation data set to obtain an updated historical operation data set. Specifically, all data points that have been added to the first historical operation data cluster are removed from the original historical operation data set, and the historical operation data set is updated for the next round of clustering. Step S215, iterative clustering is performed based on the updated historical operation data set to obtain the multiple historical operation data clusters. Specifically, steps S211 to S214 are repeated, and in each iteration, new first historical operation data are randomly extracted from the updated historical operation data set as the clustering starting point. When the updated historical operation data set is empty (that is, all data points have been assigned to a data cluster) or reaches the preset number of iterations, the iteration is stopped, and multiple historical operation data clusters are finally obtained, each of which contains a group of data points with high similarity. This implementation method avoids the excessive dependence of clustering results on initial data points by randomly extracting the first historical operation data as the clustering starting point, and gradually assigns all data points to appropriate data clusters through iterative clustering, ensuring the comprehensiveness of clustering results, and achieving the technical effect of flexibly handling the clustering process and improving the accuracy and efficiency of clustering.

[0034] Step S300, retrieve the data of the image acquisition of the telescopic lifting fan of the target flying car by the dual-target image monitoring module in the preset time window, and obtain a set of visible light images and a set of thermal infrared images. Specifically, the dual-target image monitoring module refers to an image acquisition device that includes both a visible light camera and a thermal infrared camera, which can simultaneously capture the visible light image and thermal infrared image of the object, and provide information of different dimensions. Extract the data of the image acquisition of the telescopic lifting fan of the target flying car by the dual-target image monitoring module in the preset time window from the data processing center or the storage device, and obtain a visible light image and a thermal infrared image, wherein the visible light image is an image captured by the visible light camera, which reflects the color and brightness information of the object, and is used to observe the appearance and position change of the telescopic lifting fan blade; the thermal infrared image is an image captured by the thermal infrared camera, which reflects the temperature distribution information of the object, and is used to detect the temperature rise of the telescopic lifting fan blade, hot spots and other abnormal conditions. The collected image data is preprocessed, including denoising, enhancement, correction and other operations to improve the image quality and reduce the difficulty of subsequent processing. The preprocessed visible light images and thermal infrared images are classified and organized according to time sequence and camera identification to construct a visible light image set and a thermal infrared image set. These two image sets are the basic data for feature extraction and fusion.

[0035] Step S400, extracting the fan blade feature set of the telescopic lifting fan, combining the visible light image set and the thermal infrared image set to train a double-layer feature selection fusion device, and obtaining the trained double-layer feature selection fusion device; Step S500, obtaining the real-time visible light image and real-time thermal infrared image of the telescopic lifting fan at the current moment, and using the double-layer feature selection fusion device to perform feature fusion recognition on the real-time visible light image and the real-time thermal infrared image to generate real-time fan blade features.

[0036] Specifically, from the visible light image set, image processing techniques (such as edge detection, contour extraction, texture analysis, etc.) are used to extract the visible light features of the fan blades, including the shape, size, edge contour, texture details, etc. of the blades. Similarly, the thermal infrared features of the fan blades are extracted from the thermal infrared image set, including the temperature distribution, hot spot location, temperature rise trend, etc. of the blades. The features extracted from the visible light images and thermal infrared images are integrated to construct a comprehensive feature set of the fan blades.

[0037] A two-layer feature selection fusion is designed. The two-layer feature selection fusion includes two levels: feature extraction layer and feature fusion layer. The feature extraction layer is used to extract features that are valuable for fan steady-state assessment from visible light images and thermal infrared images; the feature fusion layer is used to fuse the extracted visible light features and thermal infrared features. The fusion method can be through feature cascading, feature weighted fusion, deep learning feature fusion, etc. The two-layer feature selection fusion is trained using historical data (fan blade feature set, visible light image set, and thermal infrared image set). During the training process, the parameters of the two-layer feature selection fusion are continuously adjusted to optimize the effect of feature extraction and fusion.

[0038] Obtain the real-time visible light image and real-time thermal infrared image of the telescopic lifting fan at the current moment. Preprocess the real-time image, including denoising, correction, cropping and other operations, to ensure that the image quality meets the requirements of feature extraction. Input the preprocessed real-time visible light image and real-time thermal infrared image into the trained double-layer feature selection fusion device, which automatically performs feature extraction and fusion operations to generate the real-time fan blade features at the current moment.

[0039] In one possible implementation, a real-time visible light image and a real-time thermal infrared image of the telescopic lifting fan at the current moment are obtained, and the real-time visible light image and the real-time thermal infrared image are subjected to feature fusion and recognition using the double-layer feature selection fusion device to generate real-time fan blade features. Step S500 further includes step S510, calling the double-layer feature selection fusion device, wherein the double-layer feature selection fusion device includes a double embedding layer and a backbone network. Specifically, the double-layer feature selection fusion device is a deep learning model that combines the feature extraction and fusion capabilities of visible light and thermal infrared images, processes images of different modalities respectively through a double embedding layer, and fuses these features through a backbone network. Step S520, inputs the real-time visible light image and the real-time thermal infrared image into the visible light extraction branch and the thermal infrared extraction branch of the double embedding layer to obtain a real-time visible light feature set and a real-time thermal infrared feature set. Specifically, the real-time visible light image and the real-time thermal infrared image are respectively input into the visible light extraction branch and the thermal infrared extraction branch of the dual embedding layer, and each branch uses a convolutional neural network (CNN) or other deep learning architecture to extract the features of the image to form a real-time visible light feature set and a real-time thermal infrared feature set. Among them, the dual embedding layer includes two independent feature extraction branches for processing the visible light image and the thermal infrared image respectively. Step S530, the real-time visible light feature set and the real-time thermal infrared feature set are input into the backbone network for feature fusion to generate the real-time fan blade feature. Specifically, the real-time visible light feature set and the real-time thermal infrared feature set are used as input and sent to the backbone network for feature fusion. The backbone network can include multiple layers, such as convolutional layers, pooling layers, fully connected layers, etc., for fusing features of different modalities. The two types of features are fused together through fusion strategies such as addition, splicing, and attention mechanisms, and finally a real-time fan blade feature containing rich information is generated. The real-time fan blade feature contains both texture and shape information in the visible light image and temperature distribution information in the thermal infrared image. This implementation method reduces the calculation time to a certain extent and achieves the technical effect of improving processing efficiency by processing visible light and thermal infrared images in parallel and fusing them in the subsequent stage.

[0040] In a possible implementation, the double-layer feature selection fusion device is called, wherein the double-layer feature selection fusion device includes a double embedding layer and a backbone network, and step S510 further includes step S511, performing feature extraction on the visible light image set and the thermal infrared image set respectively to obtain a visible light feature extraction set and a thermal infrared feature extraction set. Specifically, the visible light image set and the thermal infrared image set contain multiple images for training the double-layer feature selection fusion device. Using a predefined or randomly initialized convolutional neural network (CNN) or other deep learning architecture, feature extraction is performed on each image in the visible light image set to generate a visible light feature extraction set. Similarly, feature extraction is performed on each image in the thermal infrared image set to generate a thermal infrared feature extraction set. Each feature vector in the visible light feature extraction set and the thermal infrared feature extraction set corresponds to an image in the original image set. Step S512, using the visible light image set, the thermal infrared image set, the visible light feature extraction set, and the thermal infrared feature extraction set to train the visible light extraction branch and the thermal infrared extraction branch respectively. Specifically, for the visible light extraction branch, a visible light image set and a corresponding visible light feature extraction set are used for training, and the network parameters in the visible light extraction branch are adjusted by minimizing the loss function (such as cross entropy loss, mean square error, etc.), so that it can more accurately extract features from the visible light image. Similarly, for the thermal infrared extraction branch, a thermal infrared image set and a corresponding thermal infrared feature extraction set are used for training. The trained visible light extraction branch and thermal infrared extraction branch can accurately extract features from the images of the corresponding modalities. Step S513, based on the visible light feature extraction set, the thermal infrared feature extraction set and the fan blade feature set, the backbone network is generated by training. Specifically, the fan blade feature set is a standard feature representation of the fan blade obtained by manual annotation or other pre-training models. The visible light feature extraction set and the thermal infrared feature extraction set are used as inputs and sent to the backbone network for feature fusion. The fan blade feature set is used as supervision information, and the network parameters in the backbone network are adjusted by minimizing the loss function, so that it can fuse the features of different modalities and generate outputs matching the fan blade feature set. The trained backbone network can receive visible light and thermal infrared features, and generate feature representations of fan blades by fusing these features. This implementation method uses historical data to train and generate the visible light extraction branch, thermal infrared extraction branch and backbone network in the double-layer feature selection fusion device, making full use of the information in the historical data and achieving the technical effect of improving the generalization ability of the double-layer feature selection fusion device.

[0041] Step S600, collect the real-time operation data of the telescopic lifting fan at the current moment, combine the real-time fan blade characteristics to predict the fan operation, and obtain predicted operation data. Specifically, through the monitoring data acquisition system of the target flying car, the current operation status data of the telescopic lifting fan is obtained in real time, including but not limited to the fan speed, current, voltage, temperature, vibration and other parameters. The collected real-time operation data is preprocessed, including data cleaning (removing noise, outliers, etc.), data conversion (such as normalization, standardization) and data alignment operations to ensure the consistency and accuracy of the data. Features useful for fan operation prediction are extracted from the preprocessed real-time operation data. These features can be direct physical parameters or derived features calculated by algorithms or models. These features are constructed into a feature vector or matrix for prediction model input.

[0042] A prediction model is constructed based on a machine learning model (such as a regression model, a classification model, a time series prediction model, etc.) or a deep learning model (such as a neural network, a recurrent neural network, etc.). The prediction model can predict the future operating state of the fan based on historical data, real-time operating data, and real-time fan blade features. The prediction model is trained using the historical operating data set and the corresponding fan blade feature set. During the training process, the parameters and structure of the model are continuously adjusted to optimize the prediction performance. The real-time operating data and real-time fan blade features are used as inputs to the trained prediction model. The model calculates the prediction results based on the input data, that is, the predicted operating data, including the predicted values ​​of the speed, temperature, vibration and other parameters of the telescopic lifting fan for a period of time in the future.

[0043] In a possible implementation, the real-time operation data of the telescopic lifting fan at the current moment is collected, and the fan operation prediction is performed in combination with the real-time fan blade characteristics to obtain the predicted operation data. Step S600 further includes step S610, constructing the initial state of the Markov chain based on the real-time operation data and the real-time fan blade characteristics. Specifically, the real-time operation data of the telescopic lifting fan at the current moment includes the fan speed, current, voltage, temperature, etc., and the real-time fan blade characteristics include fan blade shape, position and other characteristics. The real-time operation data and the real-time fan blade characteristics are sorted to form a vector or set that can describe the current fan state. According to this vector or set, an initial state is defined in the state space of the Markov chain, and this initial state describes the specific operation status and characteristics of the telescopic lifting fan at the current moment. Step S620, based on the historical operation data set, analysis is performed to construct the state transfer matrix of the Markov chain. Specifically, the historical operation data set contains the operation data and corresponding operation states of the telescopic lifting fan at different time points. The historical operation data is analyzed to identify different fan operation states, and the frequency or probability of transitioning from one state to another in the historical data is counted. These probabilities are used to construct a state transfer matrix, and each element in the state transfer matrix represents the probability of transitioning from one state to another, describing the possibility of the telescopic lifting fan transitioning between different states. Step S630, the initial state is substituted into the state transfer matrix to determine the predicted operation data. Specifically, the initial state is substituted into the state transfer matrix, and the next possible state is calculated according to the probability distribution. This process is repeated multiple times (according to the time length and accuracy of the prediction required) to simulate the possible state changes of the telescopic lifting fan in the future. According to the simulation results, the predicted operation data is determined, and the predicted operation data describes the possible operation state of the telescopic lifting fan in the future, including the predicted values ​​of the speed, current, voltage and other parameters of the telescopic lifting fan at a certain time point in the future. This implementation method realizes the complex prediction of the future state of the telescopic lifting fan by constructing a Markov chain, including the interaction and dependency between multiple states, and the influence of different factors on the operation state of the telescopic lifting fan, so as to achieve the technical effect of improving the accuracy of the prediction.

[0044] Step S700, according to the flight mission requirements of the target flying car, the predicted operating data is consistency certified to obtain a second steady-state coefficient. Specifically, the current flight mission of the target flying car and its specific requirements for the telescopic lifting fan are determined, including the fan's speed range, temperature limit, vibration threshold, etc. According to the flight mission requirements, the consistency standards of the fan operating parameters are set, including the allowable range of parameters, change rate limit, stability requirements, etc. The predicted operating data is compared with the set consistency standards to evaluate whether the various operating parameters of the fan meet the flight mission requirements and whether the predicted operating data fluctuates within the allowable range. Based on the results of the comparison and evaluation, the second steady-state coefficient is calculated, and the second steady-state coefficient reflects the degree of consistency between the predicted operating data and the flight mission requirements. For example, the second steady-state coefficient can be a value between 0 and 1, where 1 represents complete consistency and 0 represents complete inconsistency.

[0045] Step S800, a steady-state evaluation is performed based on the first steady-state coefficient and the second steady-state coefficient to generate a target steady-state evaluation result of the telescopic lifting fan. Specifically, the first steady-state coefficient and the second steady-state coefficient represent the stability of the historical operation data of the telescopic lifting fan and the consistency between the predicted operation data and the flight mission requirements, respectively. According to the design specifications of the telescopic lifting fan, the safety requirements of the flying car and historical experience, a steady-state evaluation standard is defined, including a threshold value, a weight distribution, a scoring mechanism, etc. of the steady-state coefficient. Using the defined steady-state evaluation standard, the first steady-state coefficient and the second steady-state coefficient are weighted summed or other comprehensive calculation methods are adopted to obtain a comprehensive steady-state score, which reflects the overall steady-state performance of the telescopic lifting fan in the current state. The calculated comprehensive steady-state score is compared with the preset steady-state state threshold, and according to the comparison result, it is determined whether the telescopic lifting fan is in a steady-state state. For example, the steady-state state can be divided into "stable", "warning", "unstable", etc. Based on the result of the steady-state evaluation, an evaluation report is generated, which includes the specific value of the steady-state coefficient, the comprehensive steady-state score, the steady-state state judgment result and possible suggestions or measures. Feedback the evaluation report to the relevant flight control personnel or maintenance team so that corresponding measures can be taken, such as adjusting the flight plan, monitoring the fan status, arranging maintenance, etc. The embodiment of the present application adopts the comprehensive use of monitoring data collection, data transition authentication, image monitoring and feature fusion recognition, real-time operation data prediction and consistency authentication and other technical means to achieve a comprehensive, real-time and accurate evaluation of the fan operation status, achieving the technical effect of improving the steady-state evaluation capability and evaluation accuracy.

[0046] In a possible implementation, a steady-state evaluation is performed based on the first steady-state coefficient and the second steady-state coefficient to generate a target steady-state evaluation result of the telescopic lifting fan, and step S800 further includes step S810, constructing a steady-state evaluation weight distribution layer; step S820, inputting the first steady-state coefficient and the second steady-state coefficient into the steady-state evaluation weight distribution layer for weighted calculation to generate the target steady-state evaluation result. Specifically, a weight is assigned to each steady-state coefficient according to expert experience, historical data analysis or experimental verification, and the size of the weight reflects the importance of the steady-state evaluation result. The first steady-state coefficient and the second steady-state coefficient are used as input data and transmitted to the steady-state evaluation weight distribution layer. In the steady-state evaluation weight distribution layer, the first steady-state coefficient and the second steady-state coefficient are weightedly calculated, and the target steady-state evaluation result is generated according to the result of the weighted calculation, which is used to represent the performance of the telescopic lifting fan in maintaining a stable operating state under specific conditions. This implementation method more accurately evaluates the steady-state performance of the telescopic lifting fan by assigning weights, thereby achieving the technical effect of improving the accuracy of the steady-state evaluation.

[0047] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for evaluating the steady state of a telescopic lifting fan based on operating status data analysis, characterized in that: The method comprises: Interact with the monitoring data acquisition system of the target flying car, retrieve the fan operation status data of the target flying car within a preset time window, and obtain a historical operation data set; Performing data transition authentication on the historical operation data set to obtain a first steady-state coefficient; Retrieving data of image acquisition of the telescopic lifting fan of the target flying car by the dual-target image monitoring module in a preset time window to obtain a visible light image set and a thermal infrared image set; Extracting a fan blade feature set of the telescopic lifting fan, training a double-layer feature selection fusion device in combination with the visible light image set and the thermal infrared image set, and obtaining the trained double-layer feature selection fusion device; Acquire a real-time visible light image and a real-time thermal infrared image of the telescopic lifting fan at the current moment, and use the double-layer feature selection fusion device to perform feature fusion recognition on the real-time visible light image and the real-time thermal infrared image to generate real-time fan blade features; Collecting real-time operating data of the telescopic lifting fan at the current moment, and performing fan operation prediction based on the real-time fan blade characteristics to obtain predicted operating data; According to the flight mission requirements of the target flying car, consistency authentication is performed on the predicted operation data to obtain a second steady-state coefficient; A steady-state evaluation is performed based on the first steady-state coefficient and the second steady-state coefficient to generate a target steady-state evaluation result of the telescopic lifting fan.

2. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 1, characterized in that: Each historical operation data in the historical operation data set includes fan speed, fan power, operation temperature, amplitude and frequency.

3. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 1, characterized in that: Performing data transition authentication on the historical operation data set to obtain a first steady-state coefficient includes: Clustering the historical operation data set according to the first transition bandwidth to obtain a plurality of historical operation data clusters; Performing intra-cluster stability identification on the multiple historical operation data clusters respectively to generate multiple first historical stability factors; Traversing the plurality of historical operation data clusters to perform mean calculation to obtain a plurality of historical operation data means; Performing stability identification on the multiple historical operation data means respectively to obtain a second historical stability factor; The first steady-state coefficient is generated by analyzing the second historical stability factor and the plurality of first historical stability factors.

4. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 3, characterized in that: include: Randomly extracting first historical operation data from the historical operation data set; Calculating the similarity between the historical operation data set and the first historical operation data to generate a plurality of first historical similarities; Adding the historical operation data whose first historical similarities are less than or equal to the first transition bandwidth into the first historical operation data cluster; Eliminating the first historical operation data cluster from the historical operation data set to obtain an updated historical operation data set; Iterative clustering is performed based on the updated historical operation data set to obtain the multiple historical operation data clusters.

5. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 1, characterized in that: include: Calling the double-layer feature selection fusion device, wherein the double-layer feature selection fusion device includes a double embedding layer and a backbone network; Inputting the real-time visible light image and the real-time thermal infrared image into the visible light extraction branch and the thermal infrared extraction branch of the dual embedding layer to obtain a real-time visible light feature set and a real-time thermal infrared feature set; The real-time visible light feature set and the real-time thermal infrared feature set are input into the backbone network for feature fusion to generate the real-time fan blade feature.

6. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 5, characterized in that: include: Performing feature extraction on the visible light image set and the thermal infrared image set respectively to obtain a visible light feature extraction set and a thermal infrared feature extraction set; Using the visible light image set, the thermal infrared image set, the visible light feature extraction set, and the thermal infrared feature extraction set to respectively train the visible light extraction branch and the thermal infrared extraction branch; The backbone network is generated based on the training of the visible light feature extraction set, the thermal infrared feature extraction set and the fan blade feature set.

7. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 1, characterized in that: include: constructing an initial state of a Markov chain based on the real-time operating data and the real-time fan blade characteristics; Analyze the historical operation data set to construct a state transfer matrix of the Markov chain; Substitute the initial state into the state transfer matrix to determine the predicted operation data.

8. The method for evaluating the steady state of a telescopic lifting fan based on operating state data analysis according to claim 1, characterized in that: include: Construct a steady-state assessment weight distribution layer; The first steady-state coefficient and the second steady-state coefficient are input into the steady-state assessment weight distribution layer for weighted calculation to generate the target steady-state assessment result.

Citation Information

Patent Citations

  • Photovoltaic defect detection method based on Yolov4 and thermal infrared image

    CN113538391A

  • Convolutional neural network-based wheel-type inspection robot and method for substation

    CN117270545A