Charging robot fault diagnosis method and system based on deep learning

Abstract

The invention discloses a charging robot fault diagnosis method and system based on deep learning, and relates to the related field of robots, and the method comprises the steps: obtaining application attribute information, carrying out the element extraction and classification, and obtaining an application attribute element system parameter set; obtaining a working fault data set, and carrying out clustering integration to obtain a multi-series parameter working fault data set; fault feature extraction and feature cascade analysis are carried out, and a robot work fault cascade feature forest is constructed; to-be-applied scene information is obtained, and a target robot work fault cascade feature tree is obtained; performing associated feature extraction and data mapping search to obtain a robot working fault feature data set; and a deep learning network structure is adopted to carry out training verification tuning, a target robot fault diagnosis network is obtained, and working fault diagnosis is carried out. The technical problem that a diagnosis result is inaccurate in fault diagnosis of an existing charging robot is solved, and the technical effect of improving the accuracy and reliability of fault diagnosis is achieved.

Description

Technical Field

[0001] The present application relates to the field of robots, and in particular to a charging robot fault diagnosis method and system based on deep learning. Background Art

[0002] With the rapid development of electric vehicles and new energy technologies, charging robots, as a new type of charging equipment, have been widely used in charging stations, parking lots and other places. However, charging robots are bound to have various faults in a long-term, high-load working environment. These faults will not only affect the working efficiency of the charging robot, but may also pose a threat to the safety of equipment and personnel. In the field of charging robot fault diagnosis, it is currently mainly based on expert experience and pattern recognition technology. Expert experience relies on manual observation and analysis of fault phenomena, which is easily affected by human subjectivity and inefficient; pattern recognition is based on the statistics and induction of historical data, but in the face of the complex fault modes and uncertainty problems of charging robots, this method is often difficult to accurately capture fault characteristics and laws, resulting in inaccurate or missed diagnostic results.

[0003] In the current related technologies, there is a technical problem of inaccurate diagnostic results in charging robot fault diagnosis. Summary of the invention

[0004] The present application provides a charging robot fault diagnosis method and system based on deep learning, which obtains the application attribute information of the charging robot, clusters and integrates the working fault data set and extracts fault features, constructs a robot working fault cascade feature forest, matches the target robot working fault cascade feature tree according to the application scenario information of the target charging robot, uses a deep learning network structure to train, verify and tune the robot working fault feature data set to obtain a target robot fault diagnosis network, and performs working fault diagnosis on the target charging robot based on the network, etc., and learns and trains a large amount of fault data through a deep learning network, thereby achieving the technical effect of improving the accuracy and reliability of fault diagnosis.

[0005] The present application provides a charging robot fault diagnosis method based on deep learning, comprising:

[0006] Acquire the application attribute information of the charging robot, perform element extraction and classification on the application attribute information of the charging robot, and obtain the application attribute element series parameter set; collect and acquire the charging robot working fault data set, and respectively cluster and integrate the charging robot working fault data set through the application attribute element series parameter set to obtain a multi-series parameter working fault data set; respectively perform fault feature extraction and feature cascade analysis on the multi-series parameter working fault data set to construct a robot working fault cascade feature forest; acquire the application scenario information of the target charging robot, match the application scenario information with the robot working fault cascade feature forest to obtain a target robot working fault cascade feature tree; perform associated feature extraction and data mapping search based on the target robot working fault cascade feature tree to obtain a robot working fault feature data set; use a deep learning network structure to train, verify and optimize the robot working fault feature data set to obtain a target robot fault diagnosis network, and perform working fault diagnosis on the target charging robot based on the target robot fault diagnosis network.

[0007] In a possible implementation, the obtaining of the application attribute element series parameter set performs the following processing:

[0008] Perform element extraction on the charging robot application attribute information to obtain robot application element information, wherein the robot application element information includes scene environment elements, equipment attribute elements and charging application object elements; determine an application element parameter classification system based on the robot application element information; crawl to obtain a charging robot application data set, classify and identify the charging robot application data set based on the application element parameter classification system, and obtain a robot application element classification parameter set; classify and number the robot application element classification parameter set according to the robot application element information to obtain the application attribute element system parameter set.

[0009] In a possible implementation, the construction of the robot work failure cascade feature forest performs the following processing:

[0010] Fault feature extraction is performed on the multi-system parameter working fault data set respectively to obtain a multi-system parameter working fault feature set; feature interaction correlation analysis is performed in sequence based on the multi-system parameter working fault feature set to obtain a multi-system parameter fault feature correlation coefficient set; feature cascade analysis is performed on the multi-system parameter working fault feature set according to the multi-system parameter fault feature correlation coefficient set to obtain a robot working fault cascade feature tree set; the robot working fault cascade feature tree set is integrated and combined according to the application attribute element system parameter set to construct the robot working fault cascade feature forest.

[0011] In a possible implementation, the robot working fault cascade feature tree set is obtained, and the following processing is performed:

[0012] Performing first-level feature extraction on the multi-system parameter working fault feature set to obtain a first-level fault feature set of multi-system parameters; performing lower-level feature association screening on the first-level fault feature set of multi-system parameters according to the multi-system parameter fault feature association coefficient set in the multi-system parameter working fault feature set to obtain a second-level fault feature set of multi-system parameters; performing feature cascade analysis in sequence based on the second-level fault feature set of multi-system parameters until an N-level fault feature set of multi-system parameters is obtained; obtaining the robot working fault cascade feature tree set based on the first-level fault feature set of multi-system parameters, the second-level fault feature set of multi-system parameters and up to the N-level fault feature set of multi-system parameters.

[0013] In a possible implementation, the robot working fault feature data set is obtained and the following processing is performed:

[0014] Acquire a fault diagnosis target, analyze the target robot working fault cascade feature tree according to the fault diagnosis target, and determine the feature tree mining depth; extract associated features of the target robot working fault cascade feature tree based on the feature tree mining depth, and determine an associated working fault feature set; perform data mapping search on the charging robot working fault data set based on the associated working fault feature set to obtain the robot working fault feature data set.

[0015] In a possible implementation, the target robot fault diagnosis network is obtained and the following processing is performed:

[0016] The deep learning network structure is architecturally designed according to the fault diagnosis target to obtain network structure parameter information; the robot working fault feature data set is diagnostically trained based on the network structure parameter information to obtain a basic robot fault diagnosis network; the basic robot fault diagnosis network is performance verified and evaluated to obtain fault diagnosis network performance evaluation parameters; the basic robot fault diagnosis network is tuned and updated based on the fault diagnosis network performance evaluation parameters to obtain the target robot fault diagnosis network.

[0017] In a possible implementation, the obtaining of the target robot fault diagnosis network performs the following processing:

[0018] A deep network optimization strategy library is constructed, and a matching analysis is performed based on the fault diagnosis network performance evaluation parameters and the deep network optimization strategy library to obtain a target network optimization strategy; the network structure parameter information is iteratively tuned and performance evaluated according to the target network optimization strategy to obtain the network structure optimization parameter information; the basic robot fault diagnosis network is configured and updated based on the network structure optimization parameter information to obtain the target robot fault diagnosis network.

[0019] The present application also provides a charging robot fault diagnosis system based on deep learning, comprising:

[0020] An application attribute element extraction and classification module, the application attribute element extraction and classification module is used to obtain the application attribute information of the charging robot, extract and classify the application attribute information of the charging robot, and obtain the application attribute element system parameter set; a working fault data set clustering module, the working fault data set clustering module is used to collect and obtain the charging robot working fault data set, cluster and integrate the charging robot working fault data set through the application attribute element system parameter set, and obtain a multi-system parameter working fault data set; a robot working fault cascade feature forest construction module, the robot working fault cascade feature forest construction module is used to extract fault features and perform feature cascade analysis on the multi-system parameter working fault data set, and construct a robot working fault cascade feature forest; the target robot working fault cascade feature tree is obtained Module, the target robot working fault cascade feature tree acquisition module is used to obtain the target application scenario information of the target charging robot, and match the target application scenario information with the robot working fault cascade feature forest to obtain the target robot working fault cascade feature tree; robot working fault feature data set acquisition module, the robot working fault feature data set acquisition module is used to perform associated feature extraction and data mapping search based on the target robot working fault cascade feature tree to obtain a robot working fault feature data set; working fault diagnosis module, the working fault diagnosis module is used to use a deep learning network structure to train, verify and tune the robot working fault feature data set to obtain a target robot fault diagnosis network, and perform working fault diagnosis on the target charging robot based on the target robot fault diagnosis network.

[0021] A charging robot fault diagnosis method and system based on deep learning proposed in this application first obtains the charging robot application attribute information, extracts and classifies the elements of the charging robot application attribute information, obtains the application attribute element series parameter set, then collects and obtains the charging robot working fault data set, clusters and integrates the charging robot working fault data set through the application attribute element series parameter set, obtains a multi-series parameter working fault data set, then performs fault feature extraction and feature cascade analysis on the multi-series parameter working fault data set, constructs a robot working fault cascade feature forest, then obtains the target charging robot's application scenario information, matches the target application scenario information with the robot working fault cascade feature forest, obtains the target robot working fault cascade feature tree, then extracts associated features and performs data mapping search based on the target robot working fault cascade feature tree, obtains the robot working fault feature data set, then uses a deep learning network structure to train, verify and optimize the robot working fault feature data set, obtains the target robot fault diagnosis network, and finally performs working fault diagnosis on the target charging robot based on the target robot fault diagnosis network, thereby achieving the technical effect of improving the accuracy and reliability of fault diagnosis. 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 system 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 flowchart of a charging robot fault diagnosis method based on deep learning provided in an embodiment of the present application.

[0024] Figure 2 A schematic diagram of the structure of a charging robot fault diagnosis system based on deep learning provided in an embodiment of the present application.

[0025] Explanation of the accompanying drawings: application attribute element extraction and classification module 10, working fault data set clustering module 20, robot working fault cascade feature forest construction module 30, target robot working fault cascade feature tree acquisition module 40, robot working fault feature data set acquisition module 50, working fault diagnosis module 60. DETAILED DESCRIPTION

[0026] 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.

[0027] 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.

[0028] 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. 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 commonly 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.

[0029] The present application embodiment provides a charging robot fault diagnosis method based on deep learning, such as Figure 1 As shown, the method includes:

[0030] Step S100, obtain the application attribute information of the charging robot, extract and classify the elements of the application attribute information of the charging robot, and obtain the application attribute element system parameter set. Specifically, collect information about its application attributes (information on various attributes and characteristics of the charging robot in actual application) from sources such as the design documents, instruction manuals, and historical data of the charging robot. Extract key elements from this information, including the robot's physical properties (such as size, weight, and material), functional attributes (such as charging speed and charging efficiency), and operating environment (such as temperature, humidity, and electromagnetic interference). Classify the extracted elements according to their properties, such as scene environment elements (temperature, humidity, etc.), equipment attribute elements (size, material, etc.), and charging application object elements (battery type, charging interface, etc.), to form an application attribute element system parameter set.

[0031] In a possible implementation, the step of obtaining the application attribute element system parameter set, step S100 further includes step S110, extracting elements from the charging robot application attribute information, and obtaining the robot application element information, wherein the robot application element information includes scene environment elements, equipment attribute elements, and charging application object elements. Specifically, the application attribute information of the charging robot is comprehensively collected from the design documents, user manuals, technical specifications, historical application records, and other materials of the charging robot. From the collected information, the key robot application elements are identified and extracted. The extracted elements are preliminarily classified according to their properties to form a preliminary framework of the robot application element information (information that can reflect the key attributes and characteristics of the charging robot in actual applications), including three categories: scene environment elements, equipment attribute elements, and charging application object elements. Among them, the scene environment elements refer to the working environment characteristics of the charging robot, such as the site type, lighting conditions, temperature and humidity conditions, etc., which affect the operation efficiency and stability of the robot. The equipment attribute elements refer to the physical and functional characteristics of the charging robot itself, such as the robot type, load capacity, navigation method, etc., which determine the scope of application and performance level of the robot. The charging application object factors refer to the characteristics of the charging object served by the charging robot, such as the charging object type, charging interface type, charging power requirements, etc. These factors determine the compatibility and matching degree between the robot and the charging object.

[0032] Step S120, according to the robot application element information, determine the application element parameter classification system. Specifically, on the basis of the preliminary classification, the application elements of each category are divided more finely to form a specific parameter classification system. For example, the scene environment element can be further subdivided into site type (indoor / outdoor, plane / complex terrain), lighting conditions (brightness range, lighting stability), temperature and humidity conditions (temperature range, humidity range), etc.; the equipment attribute element can be subdivided into charging robot type (fixed / mobile), load capacity (weight range), navigation mode (laser navigation / visual navigation / inertial navigation), etc.; the charging application object element can be subdivided into charging object type (electric car / electric bicycle / mobile phone, etc.), charging interface type (Type-C / Micro USB / Lightning, etc.), charging power requirement (fast charging / slow charging / constant voltage and current, etc.). The subdivided parameter types are arranged and combined according to logical relationships and importance to form a clear and systematic application element parameter classification system. Step S130, crawl and obtain the charging robot application data set, classify and identify the charging robot application data set based on the application element parameter classification system, and obtain the robot application element classification parameter set. Specifically, crawler technology or data interface is used to crawl historical application data sets of charging robots from relevant websites, databases or log files. These data sets contain information such as robot usage, fault records, performance parameters, etc. According to the application factor parameter classification system established in step S120, the crawled data sets are classified and identified in detail, and corresponding category labels, parameter values ​​and necessary annotation information are assigned to each data set. Step S140, the robot application factor classification parameter set is classified and numbered according to the robot application factor information to obtain the application attribute element system parameter set. Specifically, the classified and identified data sets are further sorted and classified, and each parameter is assigned a unique number or identifier so that it can be quickly located and referenced in subsequent data processing and analysis. The classified and numbered data sets are summarized and integrated to form a complete application attribute element system parameter set, which contains all key application element information and corresponding parameter values ​​of the charging robot in actual application. This implementation method ensures the accuracy and completeness of the application attribute element series parameter set through steps such as element extraction, classification and organization, data crawling and classification identification, and constructs a comprehensive, systematic and easy-to-process charging robot application attribute element parameter set, thereby improving the efficiency and convenience of data processing.

[0033] Step S200, collect and obtain the charging robot working fault data set, cluster and integrate the charging robot working fault data set respectively through the application attribute element series parameter set, and obtain a multi-series parameter working fault data set. Specifically, the fault data of the charging robot at work is collected through sensors, log records, etc. The application attribute element series parameter set is used to perform cluster analysis on the fault data set, and the fault data with similar characteristics are classified into one category to form a multi-series parameter working fault data set.

[0034] Step S300, respectively extract fault features and perform feature cascade analysis on the multi-system parameter work fault data set to construct a robot work fault cascade feature forest. Specifically, extract fault features from the multi-system parameter work fault data set, analyze the correlation between different fault features, and construct a cascade relationship of fault features based on the correlation between features to form a robot work fault cascade feature tree set. Integrate the fault cascade feature trees of different systems according to the application attribute element system parameter set to construct a robot work fault cascade feature forest.

[0035] In a possible implementation, the construction of the robot working fault cascade feature forest, step S300 further includes step S310, respectively extracting fault features from the multi-system parameter working fault data set to obtain a multi-system parameter working fault feature set. Specifically, the multi-system parameter working fault data set is preprocessed, including data cleaning (removing duplicates, missing or outliers), data conversion (normalization, standardization), etc., to ensure the accuracy and consistency of the data. From the preprocessed data set, key features that can reflect the robot working fault are selected, including environmental parameters (such as temperature, humidity, light, etc.), equipment state parameters (such as power, voltage, current, etc.) and fault manifestation parameters (such as fault type, fault frequency, fault duration, etc.). The selected features are further processed using a feature extraction algorithm (such as principal component analysis) to extract feature vectors that can characterize the robot working fault, and these feature vectors constitute a multi-system parameter working fault feature set. Step S320, based on the multi-system parameter working fault feature set, feature interaction correlation analysis is performed in sequence to obtain a multi-system parameter fault feature correlation coefficient set. Specifically, a correlation analysis algorithm (such as the Pearson correlation coefficient) is used to calculate the correlation coefficient between each feature in the multi-system parameter working fault feature set to evaluate the degree of correlation between them. The calculated correlation coefficients are organized into a correlation matrix, and each element in the matrix represents the correlation coefficient between two features. According to the correlation matrix, feature pairs with high correlation coefficients (for example, a threshold is set) are screened out, and these feature pairs represent the key interactive correlations of robot working faults. Step S330, according to the multi-system parameter fault feature correlation coefficient set, the multi-system parameter working fault feature set is subjected to feature cascade analysis to obtain a robot working fault cascade feature tree set. Specifically, based on the screened key correlation feature pairs, a robot working fault cascade feature tree is constructed, each node of the feature tree represents a feature, and the connection between the nodes represents the correlation relationship between the features. Repeat the above steps to construct multiple feature trees for different fault types or fault scenarios to form a robot working fault cascade feature tree set. Step S340, the robot working fault cascade feature tree set is integrated and combined according to the application attribute element system parameter set to construct the robot working fault cascade feature forest. Specifically, multiple feature trees in the robot working fault cascade feature tree set are integrated and combined according to the application attribute element series parameter set to form a unified feature forest. The integrated feature forest is optimized, including removing redundant features, adjusting the feature tree structure, etc., to improve the accuracy and robustness of the feature forest. This implementation method constructs a comprehensive, accurate and efficient robot working fault cascade feature forest through the steps of fault feature extraction, feature interaction correlation analysis and feature cascade analysis, which is conducive to quickly and accurately locating and analyzing the cause of the robot's fault, thereby improving the efficiency and accuracy of fault diagnosis.

[0036] In a possible implementation, the step S330 of obtaining a set of cascade feature trees of robot working faults further includes step S331, extracting first-level features from the set of multi-system parameter working fault features to obtain a set of multi-system parameter first-level fault features. Specifically, from the set of multi-system parameter working fault features, the features most directly related to the robot working faults are selected as first-level fault features based on experience or preset rules. These features are parameters that have a significant impact on the robot fault state, such as fault type, fault occurrence time, fault duration, etc. The first-level fault features are the starting point for constructing a fault cascade feature tree. The selected first-level fault features are named and classified to ensure that each feature has a clear identification and attribution. For example, "robot fault" is used as the general category of the first-level fault features, and second-level features such as "main fault type" can be subdivided under it. The first-level fault features that have been selected and named are integrated into a set to form a first-level fault feature set, which provides a basis for subsequent feature association screening and cascade analysis. Step S332, according to the multi-system parameter fault feature correlation coefficient set in the multi-system parameter working fault feature set, the multi-system parameter first-level fault feature set is subjected to lower-level feature correlation screening to obtain the multi-system parameter second-level fault feature set. Specifically, the multi-system parameter fault feature correlation coefficient set is used to analyze the degree of correlation between the first-level fault feature and other features. According to the correlation analysis results, the features with a high degree of correlation with the first-level fault feature are screened out as the second-level fault features. These features are parameters that have a further explanation or supplementary effect on the first-level fault features, such as fault mode, subdivided fault type, etc. The screened second-level fault features are integrated into a set to form a second-level fault feature set, which provides a basis for subsequent feature cascade analysis. Step S333, based on the multi-system parameter second-level fault feature set, feature cascade analysis is performed in sequence until the multi-system parameter N-level fault feature set is obtained. Specifically, based on the second-level fault feature set, the process of correlation screening and feature cascade is repeated. That is, the second-level fault feature is used as a new starting point, the degree of correlation between it and the remaining features is analyzed, and the features with a higher degree of correlation are selected as the third-level fault features, and so on, until the preset cascade depth N is reached. In each cascade analysis, the selected features are integrated into the corresponding fault feature set, and finally a complete fault feature set system from level 1 to level N is formed.

[0037] Step S334, based on the first-level fault feature set of the multi-system parameter, the second-level fault feature set of the multi-system parameter, and the N-level fault feature set of the multi-system parameter, obtain the robot work fault cascade feature tree set. Specifically, with the first-level fault feature as the root node, according to the association relationship between the fault features at each level, the second-level, third-level, and N-level fault features are sequentially connected to the corresponding parent node as child nodes to form a robot work fault cascade feature tree. The robot work fault cascade feature tree is a tree structure formed by association screening and feature cascade, which is used to represent the hierarchy and association relationship between fault features. For different fault types or fault scenarios, the above process of constructing feature trees is repeated to form multiple robot work fault cascade feature trees, and finally integrated into a feature tree set. This implementation method ensures that the robot work fault cascade feature tree can accurately reflect the feature hierarchy and association relationship of the robot work fault by gradually performing feature extraction, association screening, and feature cascade analysis, and constructs a robot work fault cascade feature tree set with clear hierarchy and close association.

[0038] Step S400, obtaining the target charging robot's application scenario information, matching the target charging robot's work fault cascade feature forest with the target charging robot's application scenario information, and obtaining the target robot's work fault cascade feature tree. Specifically, collect the target charging robot's application scenario information, and match and analyze the target charging robot's application scenario information with the robot's work fault cascade feature forest. According to the matching results, select the fault cascade feature tree that is most relevant to the target scenario from the feature forest.

[0039] Step S500: extract associated features and perform data mapping search based on the target robot working fault cascade feature tree to obtain a robot working fault feature data set. Specifically, based on the target robot working fault cascade feature tree, extract features associated with the target fault. Search the charging robot working fault data set for data records matching these associated features.

[0040] In a possible implementation, the robot working fault feature data set is obtained, and step S500 further includes step S510, obtaining a fault diagnosis target, analyzing the target robot working fault cascade feature tree according to the fault diagnosis target, and determining the feature tree mining depth. Specifically, the specific target of fault diagnosis is clarified, that is, the specific purpose that is expected to be achieved through fault diagnosis, such as determining the robot's fault type, fault cause, or both. The target robot working fault cascade feature tree is a hierarchical structure that describes various features of robot working faults and their relationships. Each node represents a feature, and the path from the root node to the leaf node represents a path of a fault mode or cause. According to the fault diagnosis target, the depth to be mined in the feature tree is determined, that is, the hierarchical depth to be explored in the feature tree to obtain sufficient information to meet the fault diagnosis target. For example, if the goal is to determine the fault type, the mining depth only needs to reach the node that can distinguish different fault types; if the goal is to determine the fault type and cause, the mining depth needs to reach the node that can explain the fault cause. Step S520, based on the feature tree mining depth, the associated features of the target robot working fault cascade feature tree are extracted to determine the associated working fault feature set. Specifically, according to the determined mining depth, all possible paths starting from the root node of the feature tree to satisfy the mining depth are selected. Along the selected path, all relevant features are extracted to form an associated working fault feature set, which represents all possible information related to the fault diagnosis target. Step S530, based on the associated working fault feature set, a data mapping search is performed on the charging robot working fault data set to obtain the robot working fault feature data set. Specifically, in the charging robot working fault data set, a mapping relationship with a known fault instance is established for each associated feature. Using the associated working fault feature set as a query, a matching fault instance is searched in the data set. The features of all matching fault instances are collected to form a robot working fault feature data set, that is, the robot working fault feature data set is a data set containing the features of all fault instances related to the fault diagnosis target. This implementation method determines the mining depth according to different fault diagnosis targets, making the diagnosis process more flexible and customizable, and helps to quickly find fault instances related to the diagnosis target, thereby improving the accuracy of fault diagnosis.

[0041] Step S600, using a deep learning network structure to train, verify and tune the robot work fault feature data set, obtain a target robot fault diagnosis network, and perform work fault diagnosis on the target charging robot based on the target robot fault diagnosis network. Specifically, the structure of the deep learning network (neural network) is designed according to the fault diagnosis target, the network is trained using the robot work fault feature data set, the network is evaluated for performance, and the network is tuned and updated according to the evaluation results, and the tuned deep learning network is used to perform fault diagnosis on the target charging robot. The embodiment of the present application adopts the method of obtaining the application attribute information of the charging robot, clustering and integrating the work fault data set and extracting fault features, constructing a robot work fault cascade feature forest, matching the target robot work fault cascade feature tree according to the target charging robot's to-be-applied scenario information, using a deep learning network structure to train, verify and tune the robot work fault feature data set, obtain a target robot fault diagnosis network, and perform work fault diagnosis on the target charging robot based on the network and other technical means, and a large amount of fault data is learned and trained through a deep learning network, so as to achieve the technical effect of improving the accuracy and reliability of fault diagnosis.

[0042] In a possible implementation, the target robot fault diagnosis network is obtained, and step S600 further includes step S610, designing the architecture of the deep learning network structure according to the fault diagnosis target, and obtaining network structure parameter information. Specifically, the specific goal of fault diagnosis is clarified, such as identifying the fault type of the robot, predicting the fault development trend, or diagnosing the specific cause of the fault. According to the complexity of fault diagnosis and data characteristics, a deep learning model, such as a convolutional neural network (CNN), is selected. According to the fault diagnosis target and model, the structure of the network is designed, including the input layer, hidden layer, output layer, etc., and the number of neurons, activation function, connection mode, etc. of each layer are determined. The designed network structure is expressed in a parameterized form, such as a weight initialization method, a learning rate, an optimization algorithm, etc. Step S620, based on the network structure parameter information, the robot working fault feature data set is diagnosed and trained to obtain a basic robot fault diagnosis network. Specifically, the robot working fault feature data set is divided into a training set, a validation set, and a test set. According to the network structure parameter information, the weights and biases of the deep learning network are initialized. The deep learning network is trained using the training set data, the loss is calculated by forward propagation, the network parameters are updated by back propagation, and the iteration is repeated many times until the preset stop condition is reached. After the training is completed, a basic diagnostic network capable of preliminarily identifying robot faults is obtained. Step S630, the basic robot fault diagnosis network is subjected to performance verification and evaluation to obtain the performance evaluation parameters of the fault diagnosis network. Specifically, the evaluation index, such as accuracy, is determined according to the fault diagnosis target. The verification set data is input into the basic diagnostic network to obtain the prediction result, and the performance parameters are calculated according to the evaluation index. Step S640, the basic robot fault diagnosis network is tuned and updated based on the performance evaluation parameters of the fault diagnosis network to obtain the target robot fault diagnosis network. Specifically, according to the obtained performance evaluation parameters, the deficiencies of the basic diagnostic network are analyzed. According to the analysis results, the network structure (such as increasing the number of layers, changing the number of neurons, etc.) or parameters (such as learning rate, optimization algorithm, etc.) are adjusted to improve the performance of the network. The training set data is retrained using the adjusted network structure and parameters to obtain the updated diagnostic network. The updated diagnostic network is verified using the verification set data to evaluate whether its performance has been improved. If the performance of the updated diagnostic network meets the requirements, it will be used as the target robot fault diagnosis network. This implementation method gradually optimizes the performance of the deep learning network by designing the network architecture according to the fault diagnosis target, conducting diagnostic training, performance verification and evaluation, and tuning and updating, so that it can better adapt to the needs of fault diagnosis tasks and improve the accuracy and efficiency of fault diagnosis.

[0043] In a possible implementation, the target robot fault diagnosis network is obtained, and step S640 further includes step S641, constructing a deep network optimization strategy library, performing matching analysis based on the fault diagnosis network performance evaluation parameters and the deep network optimization strategy library, and obtaining the target network optimization strategy. Specifically, various strategies for deep network optimization are collected from literature, experiments, experience, etc., including network structure adjustment (such as increasing or decreasing the number of layers, changing the number of neurons), parameter adjustment (such as learning rate, weight decay, batch size, etc.), regularization methods (such as Dropout, L2 regularization, etc.), optimization algorithm selection (such as SGD, Adam, RMSprop, etc.) and other aspects. The collected optimization strategies are sorted and constructed into a strategy library, and each strategy in the strategy library has a clear description and applicable scenario. According to the fault diagnosis network performance evaluation parameters, the performance bottlenecks and potential improvement points of the current network are analyzed, and the optimization strategies matching these bottleneck points are found in the strategy library. Step S642, the network structure parameter information is iteratively tuned and performance evaluated according to the target network optimization strategy, and the network structure optimization parameter information is optimized. Specifically, one or more initial optimization strategies are selected according to the matching analysis results. The selected strategies are applied to the basic robot fault diagnosis network, the network structure parameter information is adjusted, and then the network is retrained. The adjusted network is evaluated for performance using the validation set data to obtain new performance evaluation parameters. According to the new performance evaluation parameters, it is determined whether the adjusted network has improved. If it has improved, the strategy is retained; if it has not improved or the improvement is not obvious, other strategies or combined strategies are tried. This process is iterated multiple times until the best network structure optimization parameter information is found. Step S643, based on the network structure optimization parameter information, the basic robot fault diagnosis network is configured and updated to obtain the target robot fault diagnosis network. Specifically, the network structure optimization parameter information is applied to the basic robot fault diagnosis network, and the network is configured and updated. The updated network is verified and tested using the validation set and test set data to ensure that its performance on different data sets is stable and meets the requirements. If the verification and test results both show that the network performance is good, the network is used as the target robot fault diagnosis network. This implementation method provides a rich reference set for network tuning by building a deep network optimization strategy library, making the tuning process more systematic and efficient, improving the efficiency of tuning, reducing human intervention and uncertainty in the tuning process, and making network optimization more scientific and controllable.

[0044] In the above, refer to Figure 1 A charging robot fault diagnosis method based on deep learning according to an embodiment of the present invention is described in detail. Figure 2A charging robot fault diagnosis system based on deep learning according to an embodiment of the present invention is described.

[0045] A charging robot fault diagnosis system based on deep learning according to an embodiment of the present invention is used to solve the technical problem of inaccurate diagnosis results in the existing charging robot fault diagnosis, and achieve the technical effect of improving the accuracy and reliability of fault diagnosis. A charging robot fault diagnosis system based on deep learning includes: an application attribute element extraction and classification module 10, a working fault data set clustering module 20, a robot working fault cascade feature forest construction module 30, a target robot working fault cascade feature tree acquisition module 40, a robot working fault feature data set acquisition module 50, and a working fault diagnosis module 60.

[0046] The application attribute element extraction and classification module 10 is used to obtain the application attribute information of the charging robot, extract and classify the application attribute information of the charging robot, and obtain the application attribute element system parameter set; the working fault data set clustering module 20 is used to collect and obtain the charging robot working fault data set, and cluster and integrate the charging robot working fault data set through the application attribute element system parameter set to obtain a multi-system parameter working fault data set; the robot working fault cascade feature forest construction module 30 is used to extract fault features and perform feature cascade analysis on the multi-system parameter working fault data set, and construct a robot working fault cascade feature forest; the target robot working fault cascade feature tree is obtained. The acquisition module 40 is used to obtain the application scenario information of the target charging robot, and to match the application scenario information with the robot working fault cascade feature forest to obtain the target robot working fault cascade feature tree; the robot working fault feature data set acquisition module 50 is used to perform associated feature extraction and data mapping search based on the target robot working fault cascade feature tree to obtain the robot working fault feature data set; the working fault diagnosis module 60 is used to use a deep learning network structure to train, verify and tune the robot working fault feature data set to obtain the target robot fault diagnosis network, and perform working fault diagnosis on the target charging robot based on the target robot fault diagnosis network.

[0047] The specific configuration of the application attribute element extraction and classification module 10 will be described in detail below. As described above, to obtain the application attribute element system parameter set, the application attribute element extraction and classification module 10 may further include: a robot application element information acquisition unit is used to extract elements from the charging robot application attribute information to obtain robot application element information, wherein the robot application element information includes scene environment elements, equipment attribute elements and charging application object elements; an application element parameter classification system determination unit is used to determine the application element parameter classification system according to the robot application element information; a robot application element classification parameter set acquisition unit is used to crawl and obtain the charging robot application data set, classify and identify the charging robot application data set based on the application element parameter classification system, and obtain the robot application element classification parameter set; an application attribute element system parameter set acquisition unit is used to classify and number the robot application element classification parameter set according to the robot application element information to obtain the application attribute element system parameter set.

[0048] The specific configuration of the robot working fault cascade feature forest construction module 30 will be described in detail below. As described above, to construct the robot working fault cascade feature forest, the robot working fault cascade feature forest construction module 30 may further include: a fault feature extraction unit for extracting fault features from the multi-system parameter working fault data set to obtain a multi-system parameter working fault feature set; a feature interaction correlation analysis unit for sequentially performing feature interaction correlation analysis based on the multi-system parameter working fault feature set to obtain a multi-system parameter fault feature correlation coefficient set; a feature cascade analysis unit for performing feature cascade analysis on the multi-system parameter working fault feature set according to the multi-system parameter fault feature correlation coefficient set to obtain a robot working fault cascade feature tree set; and an integration and combination unit for integrating and combining the robot working fault cascade feature tree set according to the application attribute element system parameter set to construct the robot working fault cascade feature forest.

[0049] Among them, a robot working fault cascade feature tree set is obtained, and the feature cascade analysis unit may further include: a first-level feature extraction subunit is used to perform first-level feature extraction on the multi-system parameter working fault feature set to obtain a multi-system parameter first-level fault feature set; a lower-level feature association screening subunit is used to perform lower-level feature association screening on the multi-system parameter first-level fault feature set in the multi-system parameter working fault feature set according to the multi-system parameter fault feature association coefficient set in the multi-system parameter working fault feature set to obtain a multi-system parameter second-level fault feature set; a feature sequential cascade analysis subunit is used to perform feature cascade analysis in sequence based on the multi-system parameter second-level fault feature set until a multi-system parameter N-level fault feature set is obtained; a robot working fault cascade feature tree set acquisition subunit is used to obtain the robot working fault cascade feature tree set based on the multi-system parameter first-level fault feature set, the multi-system parameter second-level fault feature set and up to the multi-system parameter N-level fault feature set.

[0050] The specific configuration of the robot work fault feature data set acquisition module 50 will be described in detail below. As described above, the robot work fault feature data set is obtained, and the robot work fault feature data set acquisition module 50 may further include: a feature tree mining depth determination unit is used to obtain a fault diagnosis target, analyze the target robot work fault cascade feature tree according to the fault diagnosis target, and determine the feature tree mining depth; an associated feature extraction unit is used to extract associated features from the target robot work fault cascade feature tree based on the feature tree mining depth, and determine an associated work fault feature set; a data mapping search unit is used to perform a data mapping search on the charging robot work fault data set based on the associated work fault feature set to obtain the robot work fault feature data set.

[0051] The specific configuration of the working fault diagnosis module 60 will be described in detail below. As described above, the target robot fault diagnosis network is obtained, and the working fault diagnosis module 60 may further include: an architecture design unit for performing architecture design on the deep learning network structure according to the fault diagnosis target to obtain network structure parameter information; a diagnosis training unit for performing diagnosis training on the robot working fault feature data set based on the network structure parameter information to obtain a basic robot fault diagnosis network; a performance verification and evaluation unit for performing performance verification and evaluation on the basic robot fault diagnosis network to obtain a fault diagnosis network performance evaluation parameter; and a tuning and updating unit for tuning and updating the basic robot fault diagnosis network based on the fault diagnosis network performance evaluation parameter to obtain the target robot fault diagnosis network.

[0052] Among them, to obtain the target robot fault diagnosis network, the tuning and updating unit may further include: a target network optimization strategy acquisition subunit is used to construct a deep network optimization strategy library, and a matching analysis is performed based on the fault diagnosis network performance evaluation parameters and the deep network optimization strategy library to obtain the target network optimization strategy; a network structure optimization parameter information acquisition subunit is used to iteratively tune and perform performance evaluation on the network structure parameter information according to the target network optimization strategy, and optimize the network structure optimization parameter information; a configuration update subunit is used to perform configuration update on the basic robot fault diagnosis network based on the network structure optimization parameter information to obtain the target robot fault diagnosis network.

[0053] A charging robot fault diagnosis system based on deep learning provided in an embodiment of the present invention can execute a charging robot fault diagnosis method based on deep learning provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0054] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0055] 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 charging robot fault diagnosis method based on deep learning, characterized in that: The method comprises: Acquire application attribute information of the charging robot, extract and classify elements of the application attribute information of the charging robot, and obtain an application attribute element system parameter set; Acquire a charging robot working fault data set, and cluster and integrate the charging robot working fault data set respectively through the application attribute element series parameter set to obtain a multi-series parameter working fault data set; Performing fault feature extraction and feature cascade analysis on the multi-system parameter working fault data set to construct a robot working fault cascade feature forest; Acquire the application scenario information of the target charging robot, and match the application scenario information with the robot working fault cascade feature forest to obtain the target robot working fault cascade feature tree; Based on the target robot working fault cascade feature tree, associated feature extraction and data mapping search are performed to obtain a robot working fault feature data set; A deep learning network structure is used to train, verify and tune the robot working fault feature data set to obtain a target robot fault diagnosis network, and working fault diagnosis of the target charging robot is performed based on the target robot fault diagnosis network.

2. A charging robot fault diagnosis method based on deep learning as claimed in claim 1, characterized in that: The step of obtaining the application attribute element series parameter set includes: Extracting elements from the charging robot application attribute information to obtain robot application element information, wherein the robot application element information includes scene environment elements, device attribute elements, and charging application object elements; Determine an application factor parameter classification system according to the robot application factor information; Crawling and obtaining a charging robot application data set, classifying and identifying the charging robot application data set based on the application factor parameter classification system, and obtaining a robot application factor classification parameter set; The robot application element classification parameter set is classified and numbered according to the robot application element information to obtain the application attribute element series parameter set.

3. A charging robot fault diagnosis method based on deep learning as claimed in claim 1, characterized in that: The method of constructing a robot work failure cascade feature forest includes: Extracting fault features from the multi-parameter working fault data sets respectively to obtain a multi-parameter working fault feature set; Based on the multi-system parameter working fault feature set, feature interaction correlation analysis is performed in sequence to obtain a multi-system parameter fault feature correlation coefficient set; Performing feature cascade analysis on the multi-system parameter working fault feature set according to the multi-system parameter fault feature correlation coefficient set to obtain a robot working fault cascade feature tree set; The robot working failure cascade feature tree set is integrated and combined according to the application attribute element series parameter set to construct the robot working failure cascade feature forest.

4. A charging robot fault diagnosis method based on deep learning as claimed in claim 3, characterized in that: The obtaining of the robot working fault cascade feature tree set comprises: Performing primary feature extraction on the multi-system parameter working fault feature set to obtain a multi-system parameter primary fault feature set; According to the multi-system parameter fault feature correlation coefficient set in the multi-system parameter working fault feature set, the multi-system parameter first-level fault feature set is subjected to lower-level feature correlation screening to obtain a multi-system parameter second-level fault feature set; Performing feature cascade analysis in sequence based on the multi-system parameter second-level fault feature set until a multi-system parameter N-level fault feature set is obtained; Based on the multi-system parameter first-level fault feature set, the multi-system parameter second-level fault feature set and up to the multi-system parameter N-level fault feature set, the robot working fault cascade feature tree set is obtained.

5. A charging robot fault diagnosis method based on deep learning as claimed in claim 1, characterized in that: The robot working fault feature data set is obtained, including: Acquire a fault diagnosis target, analyze the target robot working fault cascade feature tree according to the fault diagnosis target, and determine the feature tree mining depth; Extracting associated features from the target robot working fault cascade feature tree based on the feature tree mining depth to determine an associated working fault feature set; A data mapping search is performed on the charging robot working fault data set based on the associated working fault feature set to obtain the robot working fault feature data set.

6. A charging robot fault diagnosis method based on deep learning as claimed in claim 5, characterized in that: The obtaining of the target robot fault diagnosis network comprises: Designing the deep learning network structure according to the fault diagnosis target to obtain network structure parameter information; Performing diagnostic training on the robot working fault feature data set based on the network structure parameter information to obtain a basic robot fault diagnosis network; Performing performance verification and evaluation on the basic robot fault diagnosis network to obtain performance evaluation parameters of the fault diagnosis network; The basic robot fault diagnosis network is tuned and updated based on the fault diagnosis network performance evaluation parameters to obtain the target robot fault diagnosis network.

7. A charging robot fault diagnosis method based on deep learning as claimed in claim 6, characterized in that: The step of obtaining the target robot fault diagnosis network comprises: Constructing a deep network optimization strategy library, performing matching analysis based on the fault diagnosis network performance evaluation parameters and the deep network optimization strategy library, and obtaining a target network optimization strategy; Iteratively tune and evaluate the performance of the network structure parameter information according to the target network optimization strategy to obtain the network structure optimization parameter information; The basic robot fault diagnosis network is configured and updated based on the network structure optimization parameter information to obtain the target robot fault diagnosis network.

8. A charging robot fault diagnosis system based on deep learning, characterized in that: The system is used to implement a charging robot fault diagnosis method based on deep learning according to any one of claims 1 to 7, and the system comprises: An application attribute element extraction and classification module, wherein the application attribute element extraction and classification module is used to obtain application attribute information of the charging robot, extract and classify elements of the application attribute information of the charging robot, and obtain an application attribute element system parameter set; A working fault data set clustering module, wherein the working fault data set clustering module is used to collect and obtain a working fault data set of a charging robot, and cluster and integrate the working fault data set of the charging robot respectively through the application attribute element series parameter set to obtain a multi-series parameter working fault data set; A robot working fault cascade feature forest construction module, wherein the robot working fault cascade feature forest construction module is used to perform fault feature extraction and feature cascade analysis on the multi-system parameter working fault data set to construct a robot working fault cascade feature forest; A target robot working fault cascade feature tree acquisition module, wherein the target robot working fault cascade feature tree acquisition module is used to acquire the to-be-applied scenario information of the target charging robot, and to obtain the target robot working fault cascade feature tree by matching the to-be-applied scenario information with the robot working fault cascade feature forest; A robot work fault feature data set acquisition module, wherein the robot work fault feature data set acquisition module is used to perform associated feature extraction and data mapping search based on the target robot work fault cascade feature tree to obtain a robot work fault feature data set; A working fault diagnosis module, wherein the working fault diagnosis module is used to train, verify and tune the robot working fault feature data set using a deep learning network structure, obtain a target robot fault diagnosis network, and perform working fault diagnosis on the target charging robot based on the target robot fault diagnosis network.

Images