Real-time data driven car dumper unhooking robot fault analysis method

Abstract

The invention discloses a car dumper unhooking robot fault analysis method driven by real-time data, and relates to the related field of car dumper unhooking robot fault analysis, and the method comprises the steps: determining a plurality of function units; establishing a normal cooperative relationship and an abnormal cooperative influence relationship set when any one or more units are abnormal; performing distinction degree analysis on the fault features, and establishing M functional unit combinations of which the distinction degree meets a preset distinction degree index; constructing a fault detection model according to the abnormal cooperative influence relationship and the normal cooperative relationship corresponding to each functional unit in the M functional unit combinations; and collecting real-time operation data, inputting the data into the fault detection model for fault analysis, and outputting a fault early warning signal. The technical problems that fault feature extraction is difficult, model construction is complex, and real-time early warning and quick response are difficult to achieve in existing fault analysis are solved, and the technical effects of improving feature extraction accuracy, reducing model construction complexity and achieving real-time early warning and quick response are achieved.

Description

Technical Field

[0001] The present application relates to the field related to fault analysis of a car tipping machine unhooking robot, and in particular to a real-time data-driven fault analysis method of a car tipping machine unhooking robot. Background Art

[0002] As a key device in the car tipper system, the unhooking robot is responsible for automatically unhooking the carriages during the tipping process in the transportation of bulk cargo such as coal and ore on railways, which greatly improves the operation efficiency and safety. Various faults may occur in the unhooking robot during operation, which will not only affect the normal operation of the car tipper, but also cause equipment damage or even safety accidents. In the field of fault analysis of the unhooking robot of the car tipper, the existing data-driven fault diagnosis requires preprocessing and feature extraction of a large amount of data, and the construction of complex mathematical models, resulting in the results of fault analysis often lagging behind the actual occurrence of faults, and unable to achieve real-time warning and rapid response to faults. This will not only miss the best time to handle the fault, but may also aggravate the severity of the fault and expand the scope of the fault.

[0003] In the current related technologies, the fault analysis of the dumper unhooking robot has the difficulties in extracting fault features and the complex model construction, which leads to the technical problem that it is difficult to achieve real-time fault warning and rapid response. Summary of the invention

[0004] The present application provides a real-time data-driven method for analyzing the faults of a car unhooking robot for a dumper. The method determines multiple functional units of the robot, establishes a set of normal collaborative relationships and abnormal collaborative influence relationships, performs a discrimination analysis of fault characteristics, and constructs a combination of M functional units whose discrimination meets preset indicators. Based on these functional unit combinations and their corresponding abnormal collaborative influence relationships and normal collaborative relationships, a fault detection model is constructed, real-time operation data is collected and input into the model for fault analysis, and a fault warning signal is output. Technical means are used to achieve the technical effect of improving the accuracy of fault feature extraction, reducing the complexity of model construction, and achieving real-time warning and rapid response.

[0005] The present application provides a real-time data-driven method for fault analysis of a car tipper unhooking robot, comprising: determining multiple functional units of the car tipper unhooking robot; establishing a normal collaborative relationship among the multiple functional units, and a set of abnormal collaborative influence relationships when any one or more units are abnormal; performing a discrimination analysis of fault characteristics using the set of abnormal collaborative influence relationships, and establishing a combination of M functional units whose discrimination meets a preset discrimination index, where M is an integer greater than 1; constructing a fault detection model using the abnormal collaborative influence relationships corresponding to each functional unit in the combination of the M functional units and the normal collaborative relationship; collecting real-time operating data of the multiple functional units and inputting the data into the fault detection model for fault analysis, and outputting a fault warning signal.

[0006] In a possible implementation, the abnormal collaborative influence relationship set is used to perform a discrimination analysis of fault characteristics, and M functional unit combinations whose discriminations meet a preset discrimination index are established, and the following processing is performed: extract the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship in the abnormal collaborative influence relationship set, as well as the corresponding first abnormal unit marking information and the second abnormal unit marking information, wherein the abnormal collaborative influence relationship refers to the collaborative relationship exhibited by two or more functional units under an abnormal state, and the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship refer to any two different abnormal collaborative influence relationships in the abnormal collaborative influence relationship set; perform a discrimination analysis of abnormal identification on the collaborative relationship data between the functional units in the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship, and generate a first discrimination; if the first discrimination is less than the preset discrimination index, merge the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information as a functional unit combination, and add them to the M functional unit combinations; if the first discrimination is greater than or equal to the preset discrimination index, merge the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information as a functional unit combination, and add them to the M functional unit combinations.

[0007] In a possible implementation, the preset discrimination index is determined by performing the following processing: determining a preset network structure for constructing a fault detection model; configuring a binary classifier with the preset network structure, and constructing several groups of collaborative relationship sample data corresponding to several data discriminations; testing the binary classifier with the several groups of collaborative relationship sample data, and obtaining the minimum discrimination with a test classification accuracy greater than a preset threshold, which is set as the preset discrimination index.

[0008] In a possible implementation, a fault detection model is constructed using the abnormal collaborative influence relationships and the normal collaborative relationships corresponding to each functional unit in the combination of M functional units, and the following processing is performed: topological transformation is performed on each abnormal collaborative influence relationship and the normal collaborative relationship corresponding to each functional unit in the combination of M functional units to generate each abnormal collaborative influence topology and normal collaborative relationship topology; each topological difference vector between the each abnormal collaborative influence topology and the normal collaborative relationship topology is calculated; and the fault detection model is trained using the each topological difference vector.

[0009] In a possible implementation, after the fault warning signal is output, the following processing is also performed: determine whether the fault warning signal contains warnings for multiple faulty functional units; if so, connect to the fault log record database to collect multiple historical fault record data sets of the multiple faulty functional units; perform fault probability analysis at the current moment using the multiple historical fault record data sets to generate multiple fault probabilities corresponding to the multiple faulty functional units; based on the multiple fault probabilities, arrange them in order from large to small to generate a recommended maintenance priority sequence, which is sent together with the fault warning signal to the control terminal of the dumper unhooking robot.

[0010] In a possible implementation, the multiple historical fault record data sets are used to perform a fault probability analysis at the current moment, generate multiple fault probabilities corresponding to the multiple fault functional units, and perform the following processing: based on the multiple historical fault record data sets, the multiple fault functional units are respectively analyzed for changes in fault frequency to generate multiple fault frequency characteristics; the multiple fault frequency characteristics are used to predict the difference between the next fault time and the current moment to generate multiple first fault probabilities; the real-time operation data of the dumper unhooking robot is collected to analyze the operation scene and load characteristics, analyze the failure probability of the multiple fault functional units under the current operation scene and load characteristics, compensate for the multiple first fault probabilities, and generate the multiple failure probabilities.

[0011] In a possible implementation, the failure probability of the multiple faulty functional units under the current operating scenario and load characteristics is analyzed, and the following processing is performed: based on the multiple historical fault record data sets, the frequency distribution characteristics of the faulty operating scenarios and faulty load characteristics of the multiple faulty functional units are analyzed; based on the frequency distribution characteristics, the proportion of the corresponding distribution frequency of the operating scenario and load characteristics to the total failure frequency in the frequency distribution characteristics is calculated, and the failure probability of the multiple faulty functional units under the current operating scenario and load characteristics is generated.

[0012] In a possible implementation, the following processing is performed: the multiple functional units include at least a mechanical arm, a gripping device, a visual system, a control system and a driving system of a dumper unhooking robot.

[0013] The real-time data-driven fault analysis method for the unhooking robot for a tipping machine proposed in this application first determines the multiple functional units of the unhooking robot for a tipping machine, then establishes the normal collaborative relationship of the multiple functional units, and the abnormal collaborative influence relationship set when any one or more units are abnormal, then performs a discrimination analysis of the fault characteristics with the abnormal collaborative influence relationship set, establishes a combination of M functional units whose discrimination meets the preset discrimination index, M is an integer greater than 1, and then constructs a fault detection model with the abnormal collaborative influence relationship corresponding to each functional unit in the combination of M functional units and the normal collaborative relationship, finally collects the real-time operation data of the multiple functional units and inputs it into the fault detection model for fault analysis, and outputs a fault warning signal. The technical effect of improving the accuracy of fault feature extraction, reducing the complexity of model construction, and realizing real-time warning and rapid response is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] 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.

[0015] Figure 1 A flow chart of a real-time data-driven method for fault analysis of a car tipping machine unhooking robot provided in an embodiment of the present application.

[0016] Figure 2 A schematic diagram of a flow chart for generating a recommended maintenance priority sequence in a real-time data-driven method for analyzing a car tipping machine unhooking robot fault provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] 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.

[0018] In order to make the objectives, 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.

[0019] 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.

[0020] The present application embodiment provides a real-time data driven method for analyzing the failure of a car dumper unhooking robot. Figure 1 As shown, the method includes: Step S100, determining multiple functional units of the dumper unhooking robot.

[0021] Specifically, the overall structure and working principle of the unhooking robot for dumpers are deeply understood, and multiple independent but interrelated functional units are automatically identified or manually specified by analyzing the robot's design drawings, instruction manuals or existing system architecture diagrams. These functional units refer to the components or systems in the robot that perform specific functions or tasks, which can be the robot's mechanical arm, sensors, control system, power system, etc.

[0022] In a possible implementation, step S100 further includes step S110, and the multiple functional units include at least a mechanical arm, a grasping device, a visual system, a control system and a driving system of the dumper unhooking robot.

[0023] Specifically, the robot's multiple functional units include at least a robotic arm, a gripping device, a visual system, a control system, and a drive system. Among them, the robotic arm is the main actuator of the dumper unhooking robot, responsible for performing the unhooking action. Its design includes multiple joints and drives to achieve flexible movement. To determine the robotic arm as one of the functional units, it is necessary to clarify its key parameters such as range of motion, speed, load capacity, etc. through simulation or actual testing, and analyze its motion characteristics under normal and abnormal conditions.

[0024] The gripping device is the end effector of the robot arm, which is used to grab and release goods. Its design includes different forms such as grippers and suction cups. To determine that the gripping device is a functional unit, it is necessary to conduct actual tests to evaluate its gripping force, gripping accuracy, adaptability and other performance, and analyze its performance under different working conditions.

[0025] The visual system is used to identify the location, shape, size and other information of the goods, and provide a basis for positioning and control of the robot, including cameras, image processing algorithms, etc. To determine the visual system as a functional unit, it is necessary to test the recognition accuracy, speed, stability and other performance of the visual system under different lighting, angles, occlusion and other conditions, and optimize the image processing algorithm of the visual system based on the test results to improve the recognition speed and accuracy.

[0026] The control system is the brain of the robot, responsible for receiving input signals, processing data, and issuing control instructions, including PLC (Programmable Logic Controller), computer, etc. To determine the control system as a functional unit, it is necessary to analyze its control algorithm, response time, stability and other performance, and analyze its control strategy under different working conditions.

[0027] The drive system provides power for the robot, including motors, reducers, transmissions, etc. To determine the drive system as a functional unit, it is necessary to test its driving force, efficiency, reliability and other performance, and analyze its performance under different loads and working conditions. This implementation method can better understand the working principle and failure mode of the robot by refining the functional units. When building a fault detection model, the input features and model structure can be selected more accurately, thereby improving the prediction performance and robustness of the model, and achieving more accurate and efficient fault detection and early warning.

[0028] Step S200: establishing a normal coordination relationship among the plurality of functional units and a set of abnormal coordination influence relationships when any one or more units are abnormal.

[0029] Specifically, after determining the functional units, analyze how these units cooperate with each other during normal operation, and how other units will be affected when one or some units are abnormal. This can be obtained through simulation and historical data analysis. For example, a collaborative relationship database can be established to store the normal collaborative rules and abnormal collaborative influence relationships between functional units. These rules are learned from historical data through machine learning algorithms.

[0030] Step S300, performing discrimination analysis of fault features based on the abnormal collaborative influence relationship set, and establishing M functional unit combinations whose discrimination meets a preset discrimination index, where M is an integer greater than 1.

[0031] Specifically, the data in the collaborative relationship database is mined to calculate the discrimination indexes of different functional unit combinations when a fault occurs, such as information gain, Gini coefficient, etc. Then, M functional unit combinations whose discrimination meets the preset index are selected. Among them, the discrimination index is an index used to measure the ability of the functional unit combination to distinguish the fault type when a fault occurs. The preset discrimination index is a threshold used to judge whether the abnormal collaborative influence relationship has sufficient discrimination.

[0032] In a possible implementation, the abnormal collaborative influence relationship set is used to perform a discrimination analysis of fault characteristics, and a combination of M functional units whose discrimination meets a preset discrimination index is established. Step S300 further includes step S310, extracting the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship in the abnormal collaborative influence relationship set, and the corresponding first abnormal unit marking information and the second abnormal unit marking information, wherein the abnormal collaborative influence relationship refers to the collaborative relationship exhibited by two or more functional units under an abnormal state, and the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship refer to any two different abnormal collaborative influence relationships in the abnormal collaborative influence relationship set. Specifically, the abnormal collaborative influence relationship set is traversed to extract specific abnormal collaborative influence relationship pairs therefrom. The abnormal collaborative influence relationship refers to the collaborative relationship exhibited by two or more functional units under an abnormal state, and this relationship reveals the propagation path or common cause of the fault. Each pair of abnormal collaborative influence relationships contains abnormal interaction information between at least two functional units. At the same time, first abnormal unit marking information and second abnormal unit marking information corresponding to these abnormal collaborative impact relationships are extracted. This information is used to identify which functional units are involved in the abnormal collaboration, and includes data for identifying the abnormal status of the functional units, such as abnormality type, timestamp, severity and other attributes.

[0033] Step S320, performing a discrimination analysis of abnormal identification on the coordination relationship data between the functional units in the first abnormal coordination influence relationship and the second abnormal coordination influence relationship to generate a first discrimination. Specifically, the extracted coordination relationship data is cleaned, normalized, and other preprocessing operations are performed to improve the accuracy of the analysis. The first discrimination, that is, the effectiveness of the abnormal coordination relationship in distinguishing different fault characteristics, is calculated by applying statistical indicators (such as mutual information, correlation coefficient) or machine learning models (such as classifiers, clustering algorithms).

[0034] Step S330: If the first discrimination is less than the preset discrimination index, the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information are merged as a functional unit combination and added to the M functional unit combinations. Specifically, when the first discrimination is less than the preset discrimination index, it is considered that the abnormal collaborative influence relationship currently analyzed is not effective enough in distinguishing fault characteristics. Therefore, according to the abnormal unit marking information, units with similar abnormal characteristics or close collaborative relationships are merged to form a new functional unit combination. The merged functional unit combination is added to the M functional unit combinations, and the relevant records are updated at the same time.

[0035] Step S340, if the first discrimination is greater than or equal to the preset discrimination index, the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information are respectively taken as a functional unit combination, and added into the M functional unit combinations. Specifically, when the first discrimination is greater than or equal to the preset discrimination index, it is considered that the abnormal collaborative influence relationship currently analyzed is sufficiently effective in distinguishing fault characteristics, so each abnormal unit is taken as an independent functional unit combination according to the abnormal unit marking information. These independent functional unit combinations are added to the M functional unit combinations. This implementation method optimizes the construction process of the functional unit combination through detailed discrimination analysis, thereby improving the accuracy and efficiency of the fault detection model.

[0036] In a possible implementation, the preset discrimination index is determined, and step S330 further includes step S331, determining a preset network structure for constructing a fault detection model. Specifically, according to the working characteristics and fault detection requirements of the dumper unhooking robot, a suitable neural network structure is selected, such as a convolutional neural network (CNN), a recurrent neural network (RNN) or a deep neural network (DNN). The parameters such as the number of layers, the number of nodes, and the activation function of the network are configured to ensure that the network can handle the scale and complexity of the input data.

[0037] Step S332, configure a binary classifier with the preset network structure, and construct several groups of collaborative relationship sample data corresponding to several data discriminations. Specifically, on the basis of the preset network structure, add an output layer and configure it as a binary classification task to distinguish whether the abnormal collaborative influence relationship is effective or invalid in distinguishing fault characteristics. According to the historical operation data and fault records of the unhooking robot of the dumper, construct several groups of collaborative relationship sample data. Each group of data includes collaborative relationship data between functional units (such as sensor readings, control signals, etc.) and corresponding labels (valid or invalid).

[0038] Step S333, test the binary classifier with the several groups of collaborative relationship sample data, obtain the minimum discrimination with a test classification accuracy greater than a preset threshold, and set it as the preset discrimination index. Specifically, the binary classifier is tested using the constructed collaborative relationship sample data to evaluate its classification performance. By adjusting the discrimination, the minimum discrimination with a test classification accuracy greater than a preset threshold (such as 90%, 95%, etc.) is found, and it is set as the preset discrimination index. This implementation method can find the feature threshold that is most capable of distinguishing between effective and invalid abnormal collaborative influence relationships by adjusting the discrimination and testing the classification performance, thereby improving the accuracy of the model.

[0039] Step S400: constructing a fault detection model based on the abnormal collaborative influence relationship corresponding to each functional unit in the combination of M functional units and the normal collaborative relationship.

[0040] Specifically, after determining the functional unit combinations with significant differentiation, these combinations can be used to build a fault detection model. This model can be rule-based, statistical, or machine learning-based. According to the functional unit combinations, abnormal collaborative influence relationships, and normal collaborative relationships, a machine learning model is trained. The goal of the model is to determine whether there is a fault based on real-time operation data and predict the type of fault.

[0041] In a possible implementation, a fault detection model is constructed using the abnormal collaborative influence relationship corresponding to each functional unit in the M functional unit combination and the normal collaborative relationship, and step S400 further includes step S410, performing topological transformation on each abnormal collaborative influence relationship and the normal collaborative relationship corresponding to each functional unit in the M functional unit combination, respectively, to generate each abnormal collaborative influence topology and normal collaborative relationship topology. Specifically, a topological representation method in graph theory is adopted to define each functional unit in the M functional unit combination as a node, and a unique identifier is assigned to it, and the connection (edge) between the nodes is defined according to the collaborative relationship between the functional units. Normal collaborative relationships are represented by solid lines, and abnormal collaborative influence relationships are represented by dotted lines or lines of different colors. According to the definitions of nodes and edges, normal collaborative relationship topologies and each abnormal collaborative influence topology are constructed. These topological structures can be undirected graphs or directed graphs, depending on the nature of the collaborative relationship.

[0042] Step S420, calculate each topology difference vector between each abnormal collaborative impact topology and the normal collaborative relationship topology. Specifically, a topology difference analysis method is used to calculate the difference between each abnormal collaborative impact topology and the normal collaborative relationship topology by comparing the positions of nodes, the number of edges, the weights of edges and other attributes in the two topologies. The calculated difference is expressed in vector form, the magnitude of the vector represents the degree of the difference, and the direction of the vector represents the type or direction of the difference (such as increase, decrease, change, etc.).

[0043] Step S430, training the fault detection model with the various topology difference vectors. Specifically, a machine learning or deep learning method is used to construct a training set using the calculated various topology difference vectors as training data. The training set is used to train a model (such as a neural network) so that the model can learn the difference between normal collaborative relationships and abnormal collaborative effects. The performance of the model is evaluated through cross-validation or a test set to ensure that the model can accurately identify functional units that may have abnormalities. This implementation method quantifies the difference between normal collaborative relationships and abnormal collaborative effects by calculating topology difference vectors, provides a reliable basis for fault detection, and improves the accuracy and efficiency of fault analysis.

[0044] Step S500, collecting real-time operation data of the plurality of functional units and inputting the data into the fault detection model for fault analysis, and outputting a fault warning signal.

[0045] Specifically, in actual applications, the real-time operation data of the unhooking robot of the dumper (such as sensor readings, system logs, etc.) is continuously collected through sensor networks, data interfaces, etc., and input into the fault detection model for analysis after preprocessing. The model will output a prediction result, indicating whether there is a fault and the type of fault. If there is a fault, the early warning mechanism is triggered and a fault warning signal is output. The embodiment of the present application adopts the method of determining multiple functional units of the robot, establishing a set of normal collaborative relationships and abnormal collaborative influence relationships, performing a discrimination analysis of fault characteristics, and constructing a combination of M functional units whose discrimination meets the preset indicators. Based on these functional unit combinations and their corresponding abnormal collaborative influence relationships and normal collaborative relationships, a fault detection model is constructed, real-time operation data is collected and input into the model for fault analysis, and a fault warning signal is output. Technical means such as the above have achieved the technical effect of improving the accuracy of fault feature extraction, reducing the complexity of model construction, and realizing real-time warning and rapid response.

[0046] like Figure 2 As shown, in a possible implementation, after outputting the fault warning signal, the method further includes step S600, determining whether the fault warning signal includes warnings for multiple faulty functional units. Specifically, the warning signal is parsed to identify the specific functional unit it points to, and to determine whether there are multiple such units.

[0047] Step S700: If yes, connect to the fault log record database to collect multiple historical fault record data sets of the multiple faulty functional units. Specifically, once it is determined that the warning signal involves multiple functional units, the historical fault records of these functional units are obtained by connecting to the fault log record database (a database storing historical fault records of each functional unit of the dumper unhooking robot). For each faulty functional unit, its historical fault record is retrieved to form multiple historical fault record data sets.

[0048] Step S800, performing a fault probability analysis at the current moment with the multiple historical fault record data sets, and generating multiple fault probabilities corresponding to the multiple faulty functional units. Specifically, using machine learning or statistical methods, the historical fault record data sets are analyzed to estimate the probability of each faulty functional unit failing at the current moment. For example, a prediction model can be trained using information such as historical fault frequency, fault type, and environmental conditions in which the fault occurs, and a fault probability value is generated for each faulty functional unit through the prediction model.

[0049] Step S900, based on the multiple fault probabilities, arrange them in order from large to small, generate a recommended maintenance priority sequence, and send it to the control terminal of the tipper unhooking robot together with the fault warning signal. Specifically, the fault probability values ​​generated in step 800 are sorted in order from large to small to form a recommended maintenance priority sequence. This sequence indicates which functional units are most likely to fail and should therefore be given priority for maintenance. Finally, this priority sequence is sent together with the original fault warning signal to the control terminal of the tipper unhooking robot (the operating interface or control system of the tipper unhooking robot, used to receive information such as fault warnings and maintenance suggestions), so that the operator or automation system can take corresponding maintenance measures. This implementation method can more effectively allocate maintenance resources and reduce unnecessary downtime by providing a maintenance priority sequence based on fault probability.

[0050] In a possible implementation, the multiple historical fault record data sets are used to perform a fault probability analysis at the current moment to generate multiple fault probabilities corresponding to the multiple faulty functional units. Step S800 further includes step S810, which performs a fault frequency change analysis on the multiple faulty functional units based on the multiple historical fault record data sets to generate multiple fault frequency features. Specifically, statistical analysis and data mining methods are used to process multiple historical fault record data sets. For each faulty functional unit, the fault frequency (the number of faults occurring per unit time) in different time periods is calculated by drawing a time series graph, calculating a moving average, or applying a time series analysis model (such as an ARIMA model), and the changing trend of these frequencies over time is observed. The fault frequency features may include the mean value, standard deviation, maximum / minimum value, and changing trend (increasing / decreasing / stable) of the fault frequency.

[0051] Step S820, using the multiple fault frequency characteristics to predict the difference between the next fault time and the current time, and generate multiple first fault probabilities. Specifically, based on the fault frequency characteristics generated in step S810, a time series prediction model (such as exponential smoothing, an extension of the ARIMA model, etc.) is used to predict the next fault time of each faulty functional unit. Then, the difference between the predicted next fault time and the current time is calculated, and the first fault probability is generated based on this difference. The smaller the difference, the higher the first fault probability, indicating that the faulty functional unit is more likely to fail in a short period of time in the future.

[0052] Step S830, collect the real-time operation data of the unhooking robot of the tipping machine, analyze the operation scene and load characteristics, analyze the failure probability of the multiple faulty functional units under the current operation scene and load characteristics, compensate the multiple first failure probabilities, and generate the multiple failure probabilities. Specifically, the operation data of the unhooking robot of the tipping machine is collected in real time by using sensors and monitoring systems, including operation scenes (such as working environment, temperature, humidity, etc.) and load characteristics (such as load size, load type, load distribution, etc.). By constructing a conditional probability model (a model that describes the probability of an event occurring under certain given conditions), applying a Bayesian network (a model based on probability theory and graph theory, used to represent the dependency relationship between variables, and to perform reasoning and prediction), etc., these data are used to analyze the impact of the current operation scene and load characteristics on the failure probability of each faulty functional unit. According to the analysis results, the first failure probability generated in step S820 is compensated to generate a more accurate failure probability. This implementation method comprehensively analyzes and predicts the failure probability of the faulty functional unit by combining the historical fault record data set and the real-time operation data, thereby improving the accuracy of the prediction.

[0053] In one possible implementation, the failure probability of the multiple faulty functional units under the current operating scenario and load characteristics is analyzed, and step S830 further includes step S831, analyzing the frequency distribution characteristics of the faulty operating scenarios and faulty load characteristics of the multiple faulty functional units based on the multiple historical fault record data sets. Specifically, data mining and statistical analysis methods are used to process multiple historical fault record data sets. For each faulty functional unit, the operating scenario and load characteristic information in its historical fault records are extracted, and the frequency of occurrence of these characteristics in different fault events is counted. For example, a frequency distribution table or a frequency distribution graph can be constructed, in which each feature combination corresponds to a frequency value, indicating the number of times the feature combination appears in the fault event.

[0054] Step S832, based on the frequency distribution characteristics, calculate the proportion of the corresponding distribution frequency of the operation scene and load characteristics to the total fault frequency in the frequency distribution characteristics, and generate the fault probability of the multiple fault function units under the current operation scene and load characteristics. Specifically, on the basis of step S831, calculate the proportion of the frequency of each operation scene and load characteristic to the total fault frequency. This proportion is an estimated value of the probability of the fault function unit failing under the current operation scene and load characteristics. That is, for each fault function unit, the frequency of each feature combination in its historical fault record is divided by the total fault frequency (the total number of faults), and the corresponding fault probability of the feature combination is obtained. Then, for the operation scene and load characteristics in the current real-time operation data, find the feature combination that matches it in the frequency distribution characteristics, and obtain its corresponding fault probability. This probability value reflects the possibility of the fault function unit failing under the current operation scene and load characteristics. This implementation method comprehensively considers the operating scenario and load characteristic information in historical fault records as well as this information in current real-time operating data, accurately estimates the probability of failure of the faulty functional unit under the current operating scenario and load characteristics, and improves the accuracy of fault analysis.

[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 real-time data-driven method for analyzing the failure of a car dumper unhooking robot, characterized in that: include: Determine the multiple functional units of the dumper unhooking robot; Establishing a normal coordination relationship among the plurality of functional units and a set of abnormal coordination influence relationships when any one or more units are abnormal; Performing a discrimination analysis of fault features based on the abnormal collaborative influence relationship set, and establishing a combination of M functional units whose discrimination meets a preset discrimination index, where M is an integer greater than 1; Constructing a fault detection model based on the abnormal collaborative influence relationship corresponding to each functional unit in the combination of M functional units and the normal collaborative relationship; The real-time operation data of the multiple functional units are collected and input into the fault detection model for fault analysis, and a fault warning signal is output.

2. The real-time data-driven unhooking robot fault analysis method for a dumper according to claim 1, characterized in that: The abnormal collaborative influence relationship set is used to perform a discrimination analysis of the fault characteristics, and a combination of M functional units whose discrimination meets the preset discrimination index is established, including: Extracting a first abnormal collaborative influence relationship and a second abnormal collaborative influence relationship from the abnormal collaborative influence relationship set, as well as corresponding first abnormal unit marking information and second abnormal unit marking information, wherein an abnormal collaborative influence relationship refers to a collaborative relationship exhibited by two or more functional units under an abnormal state, and the first abnormal collaborative influence relationship and the second abnormal collaborative influence relationship refer to any two different abnormal collaborative influence relationships in the abnormal collaborative influence relationship set; Performing a discrimination degree analysis of abnormality identification on the coordination relationship data between the functional units in the first abnormal coordination influence relationship and the second abnormal coordination influence relationship to generate a first discrimination degree; If the first discrimination is less than the preset discrimination index, merging the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information as a functional unit combination, and adding the combination to the M functional units; If the first discrimination is greater than or equal to the preset discrimination index, the marked abnormal units in the first abnormal unit marking information and the second abnormal unit marking information are respectively taken as a functional unit combination and added into the M functional unit combinations.

3. The real-time data-driven method for analyzing the failure of a car dumper unhooking robot according to claim 2, characterized in that: The determination of the preset discrimination index includes: Determine a preset network structure for building a fault detection model; A binary classifier is configured with the preset network structure, and several groups of collaborative relationship sample data corresponding to several data discrimination degrees are constructed; The binary classifier is tested with the several groups of collaborative relationship sample data to obtain a minimum discrimination index with a test classification accuracy greater than a preset threshold, which is set as the preset discrimination index.

4. The real-time data driven method for analyzing the failure of a car dumper unhooking robot according to claim 1, characterized in that: Constructing a fault detection model based on the abnormal collaborative influence relationship corresponding to each functional unit in the combination of M functional units and the normal collaborative relationship, including: Performing topological transformation on each abnormal collaborative influence relationship and the normal collaborative relationship corresponding to each functional unit in the combination of the M functional units, respectively, to generate each abnormal collaborative influence topology and a normal collaborative relationship topology; Calculating each topology difference vector between each abnormal collaborative influence topology and the normal collaborative relationship topology; The fault detection model is trained with the respective topology difference vectors.

5. The real-time data driven method for analyzing the failure of a car dumper unhooking robot according to claim 1, characterized in that: After outputting the fault warning signal, it also includes: Determining whether the fault warning signal includes warnings for multiple faulty functional units; If yes, connect to the fault log record database to collect multiple historical fault record data sets of the multiple faulty functional units; Performing a fault probability analysis at a current moment using the multiple historical fault record data sets to generate multiple fault probabilities corresponding to the multiple faulty functional units; Based on the multiple fault probabilities, they are arranged in descending order to generate a recommended maintenance priority sequence, which is sent to the control terminal of the dumper unhooking robot together with the fault warning signal.

6. The real-time data driven method for analyzing the failure of a car dumper unhooking robot according to claim 5, characterized in that: Performing a fault probability analysis at the current moment using the multiple historical fault record data sets to generate multiple fault probabilities corresponding to the multiple faulty functional units includes: Based on the multiple historical fault record data sets, respectively analyze the changes in the fault frequencies of the multiple fault function units to generate multiple fault frequency features; Predicting the difference between the next fault time and the current time using the multiple fault frequency characteristics to generate multiple first fault probabilities; Collect real-time operation data of the dumper unhooking robot, analyze the operation scene and load characteristics, analyze the failure probability of the multiple faulty functional units under the current operation scene and load characteristics, compensate for the multiple first failure probabilities, and generate the multiple failure probabilities.

7. The real-time data driven method for analyzing the failure of a car dumper unhooking robot according to claim 6, characterized in that: Analyzing the failure probability of the multiple faulty functional units under the current operation scenario and load characteristics, including: Analyzing frequency distribution characteristics of fault operation scenarios and fault load characteristics of the multiple fault function units based on the multiple historical fault record data sets; Based on the frequency distribution characteristics, the proportion of the corresponding distribution frequency of the operation scenario and load characteristics to the total fault frequency in the frequency distribution characteristics is calculated to generate the failure probability of the multiple fault functional units under the current operation scenario and load characteristics.

8. The real-time data driven method for analyzing the failure of a car dumper unhooking robot according to claim 1, characterized in that: The multiple functional units include at least a mechanical arm, a grasping device, a visual system, a control system and a driving system of the dumper unhooking robot.

Images