Multi-module manifold system for two-wheelers

Abstract

The present invention discloses a multi-module joint management system for two-wheeled vehicles, which relates to the field of data processing technology. The system comprises: reading real-time monitoring data of each module in the two-wheeled vehicle; obtaining multiple module monitoring data areas; extracting the first module monitoring data area; performing abnormal detection feature learning based on abnormal detection feature learning function and abnormal detection feature learning particle group, and constructing the first module abnormal detection channel; performing abnormal analysis in combination with the embedded abnormal feature coupling detection model; when the first module abnormal feature coupling detection index is greater than / equal to the abnormal feature coupling detection threshold, generating the first module abnormal warning signal. The system solves the technical problem that the existing two-wheeled vehicle multi-module joint management system lacks efficient data exchange and coordination among the modules, and it is difficult to achieve accurate control of various performances, resulting in poor overall system operation efficiency, and achieves the technical effect of improving the intelligence level and overall operation efficiency of the two-wheeled vehicle multi-module joint management.

Description

Technical Field

[0001] The present application relates to data processing technology-related fields, and in particular to a multi-module joint pipe system for a two-wheeled vehicle. Background Art

[0002] With the increasing density of urban population and the increasingly serious traffic congestion problem, driven by the growing demand for environmental protection and travel efficiency, two-wheeled vehicles, as an important part of urban transportation, have become the first choice for more and more people to travel with their light, flexible and environmentally friendly characteristics. The central control system of two-wheeled vehicles integrates multiple modules such as battery management, motor control, vehicle monitoring, intelligent navigation, and safety protection. However, the various modules of the traditional two-wheeled vehicle central control system have many deficiencies in data exchange and collaborative work, and the intelligence level is limited. It is difficult to achieve precise control and optimization of multiple functions of two-wheeled vehicles, which makes the overall performance and reliability of the system low and cannot meet the modern people's convenient, efficient and safe travel needs.

[0003] Therefore, in the current technologies related to the multi-module joint management system of two-wheeled vehicles, there is a technical problem that the modules lack efficient data exchange and coordination, making it difficult to achieve precise control of various performances, resulting in poor overall operating efficiency of the system. Summary of the invention

[0004] The present application provides a multi-module joint management system for two-wheeled vehicles, and adopts technical means such as parameter decomposition operation, establishment of mapping relationship, backtracking search optimization, etc., to solve the technical problems of the existing multi-module joint management system for two-wheeled vehicles, such as the lack of efficient data exchange and coordination among modules, difficulty in achieving precise control of various performances, and poor overall operating efficiency of the system, thereby achieving the technical effect of improving the intelligence level and overall operating efficiency of the multi-module joint management of two-wheeled vehicles.

[0005] The present application provides a multi-module joint management system for a two-wheeled vehicle, the system comprising: a data interaction component, the data interaction component is used to read the real-time monitoring data of each module in the two-wheeled vehicle to obtain a plurality of module monitoring data sets; a data preprocessing component, the data preprocessing component is used to perform data cleaning on the plurality of module monitoring data sets to obtain a plurality of module monitoring data areas; a data extraction component, the data extraction component is used to extract a first module monitoring data area corresponding to a first module according to the plurality of module monitoring data areas, wherein the first module includes the respective modules; an abnormality detection feature learning component, the abnormality detection feature learning component performs abnormality detection feature learning on the first module based on an embedded abnormality detection feature learning function and an abnormality detection feature learning particle swarm to construct an abnormality detection channel for the first module; an abnormality analysis component, the abnormality analysis component performs abnormality analysis on the monitoring data area of ​​the first module based on the abnormality detection channel of the first module in combination with an embedded abnormality feature coupling detection model to generate an abnormality feature coupling detection index for the first module; an abnormality warning component, the abnormality warning component is used to generate an abnormality warning signal for the first module when the abnormality feature coupling detection index of the first module is greater than / equal to an abnormality feature coupling detection threshold.

[0006] In a possible implementation, the anomaly detection feature learning component further performs the following processing: connecting to the data interaction component, reading the anomaly detection record data of the first module, and obtaining multiple module anomaly detection record sets, wherein each module anomaly detection record set includes a module monitoring data area record set and a module anomaly identification data record set; performing supervised learning on the multiple module anomaly detection record sets according to the anomaly detection feature learning function and the anomaly detection feature learning particle swarm, and constructing multiple module anomaly detection models that meet the anomaly detection feature learning constraints; merging the multiple module anomaly detection models as parallel nodes to generate the first module anomaly detection channel.

[0007] In a possible implementation, the anomaly detection feature learning component further performs the following processing: extracting a first module anomaly detection record set according to the multiple module anomaly detection record sets, wherein the first module anomaly detection record set includes any one module anomaly detection record set among the multiple module anomaly detection record sets; dividing the first module anomaly detection record set according to a preset data partitioning condition to generate a first module anomaly detection training set and a first module anomaly detection verification set; randomly extracting a first anomaly detection feature learning particle according to the anomaly detection feature learning particle group, wherein the anomaly detection feature learning particle group includes multiple anomaly detection feature learning particles; cross-supervising the first module anomaly detection training set according to the first anomaly detection feature learning particle. Learning, generating a first module anomaly detector; verifying the first module anomaly detector according to the first module anomaly detection verification set to obtain a first detector verification result, wherein the first detector verification result includes a first anomaly detection output accuracy coefficient and a first anomaly detection output error coefficient; inputting the first detector verification result into the anomaly detection feature learning function to obtain a first anomaly detection feature learning coefficient; judging whether the first anomaly detection feature learning coefficient satisfies the anomaly detection feature learning constraint; if the first anomaly detection feature learning coefficient satisfies the anomaly detection feature learning constraint, setting the first module anomaly detector as a first module anomaly detection model, and adding the first module anomaly detection model to the multiple module anomaly detection models.

[0008] In a possible implementation, the anomaly detection feature learning component further performs the following processing: the anomaly detection feature learning function is:

[0009]

[0010] Among them, FICFAD represents the feature learning coefficient of anomaly detection, FADX represents the output accuracy coefficient of anomaly detection, ADXO represents the output accuracy coefficient of preset anomaly detection, FADY represents the output error coefficient of anomaly detection, and ADYO represents the output error coefficient of preset anomaly detection.

[0011] In a possible implementation, the anomaly detection feature learning component also performs the following processing: if the first anomaly detection feature learning coefficient does not satisfy the anomaly detection feature learning constraint, activating the incremental learning model within the anomaly detection feature learning component; obtaining a verification loss data set corresponding to the first module anomaly detector; taking the anomaly detection feature learning constraint as an optimization learning target; optimizing the learning of the first module anomaly detector according to the incremental learning model and the verification loss data set until the optimization learning target is met, thereby generating the first module anomaly detection model.

[0012] In a possible implementation, the anomaly analysis component also performs the following processing: inputting the monitoring data area of ​​the first module into the anomaly detection channel of the first module to obtain the anomaly recognition result of the first module; the anomaly feature coupling detection model includes an anomaly feature depth detection network, an anomaly feature breadth detection network and an anomaly feature coupling weighted network; performing an anomaly depth evaluation on the anomaly recognition result of the first module according to the anomaly feature depth detection network to obtain the anomaly depth detection index of the first module; performing an anomaly risk prediction on the anomaly recognition result of the first module according to the anomaly feature breadth detection network to obtain the anomaly risk detection index of the first module; performing a weighted calculation on the anomaly depth detection index of the first module and the anomaly risk detection index of the first module according to the anomaly feature coupling weighted network to obtain the anomaly feature coupling detection index of the first module.

[0013] In a possible implementation, the anomaly analysis component further performs the following processing: inputting the first module monitoring data area into the first module anomaly detection channel, and performing anomaly identification on the first module monitoring data area according to each module anomaly detection model in the first module anomaly detection channel, to obtain multiple groups of first module anomaly identification data; performing data fusion according to the multiple groups of first module anomaly identification data, to generate the first module anomaly identification result.

[0014] In a possible implementation, the multi-module joint management system also includes an abnormal feature operation and maintenance component, and the abnormal warning component also performs the following processing: activating the two-wheeled vehicle fault prediction model and the two-wheeled vehicle fault operation and maintenance model in the abnormal feature operation and maintenance component according to the abnormal warning signal of the first module; inputting the abnormal identification result of the first module into the two-wheeled vehicle fault prediction model to obtain the first module fault prediction result; inputting the first module fault prediction result into the two-wheeled vehicle fault operation and maintenance model to obtain the first module fault operation and maintenance plan; and performing operation and maintenance management on the first module according to the first module fault operation and maintenance plan.

[0015] Through the multi-module joint management system for two-wheeled vehicles proposed in this application, the real-time monitoring data of each module in the two-wheeled vehicle is read to obtain multiple module monitoring data sets; the multiple module monitoring data sets are cleaned to obtain multiple module monitoring data areas; according to the multiple module monitoring data areas, the first module monitoring data area corresponding to the first module is extracted, wherein the first module includes the various modules; based on the embedded abnormal detection feature learning function and the abnormal detection feature learning particle swarm, the first module is subjected to abnormal detection feature learning, and the first module abnormal detection channel is constructed; based on the first module abnormal detection channel, the first module monitoring data area is subjected to abnormal analysis in combination with the embedded abnormal feature coupling detection model, and the first module abnormal feature coupling detection index is generated; when the first module abnormal feature coupling detection index is greater than / equal to the abnormal feature coupling detection threshold, the first module abnormal warning signal is generated. The technical problem that the existing two-wheeled vehicle multi-module joint management system lacks efficient data exchange and coordination among the modules, and it is difficult to achieve accurate control of various performances, resulting in poor overall system operation efficiency is solved, and the technical effect of improving the intelligence level and overall operation efficiency of the two-wheeled vehicle multi-module joint management is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure 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.

[0017] Figure 1 A schematic diagram of the structure of a multi-module joint pipe system for a two-wheeled vehicle provided in an embodiment of the present application;

[0018] Figure 2 A schematic diagram of the execution process of an anomaly detection feature learning component in a multi-module joint management system for a two-wheeled vehicle provided in an embodiment of the present application.

[0019] Explanation of the reference numerals: data interaction component 10 , data preprocessing component 20 , data extraction component 30 , anomaly detection feature learning component 40 , anomaly analysis component 50 , anomaly warning component 60 . DETAILED DESCRIPTION

[0020] 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.

[0021] 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.

[0022] 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 thereof are intended to cover non-exclusive inclusions, for example, a process, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, 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.

[0023] The present application embodiment provides a multi-module joint pipe system for a two-wheeled vehicle, such as Figure 1 As shown, the system comprises:

[0024] Data interaction component 1, the data interaction component 1 is used to read the real-time monitoring data of each module in the two-wheeled vehicle to obtain multiple module monitoring data sets. The real-time monitoring data of each module of the two-wheeled vehicle constitutes a monitoring data set, and finally multiple module monitoring data sets are obtained. Among them, each module of the two-wheeled vehicle includes a power module, a main control module, a Bluetooth module, an NFC module, a PKE module, an RF remote control sending and receiving module, a horn drive module, an electric switch, an ignition device control module, a turn signal, a horn control module, an electronic steering column lock, a seat bucket lock, a fuel tank lock control module, a LIN bus communication module, and an acceleration sensor module. Specifically, the power module is responsible for providing a stable power supply for the entire vehicle, and can convert the input AC or DC power into the specific voltage required by the equipment to meet the voltage requirements of different components or parts. At the same time, the power supply The module also has an overload protection function. When the device is overloaded or short-circuited, it can detect and cut off the power supply in time to protect the safety of the device. The main control module is the core of the vehicle control system. It is responsible for receiving and processing signals from each module and controlling the working status of other modules according to the preset program and logic. The main control module can coordinate the work between each module to ensure the normal operation of the vehicle. The Bluetooth module is used to realize wireless communication between the vehicle and other Bluetooth devices. Through the Bluetooth module, the vehicle can be remotely controlled using a mobile phone or other Bluetooth device, such as remote start, locking, unlocking, etc. The Bluetooth OTA upgrade is mainly used for the firmware upgrade of the device, that is, updating the software program of the device. The Bluetooth module can also Used for transmission and sharing of vehicle data; the NFC module uses near-field communication technology to realize data exchange between the vehicle and devices that support NFC functions. The NFC module is often used for vehicle startup authentication, anti-theft reminders, remote unlocking and other functions, which improves the safety and convenience of the vehicle; the PKE module is a keyless entry module that uses RFID technology to automatically unlock and lock the vehicle. When the owner approaches the vehicle with a legally authorized key, the PKE module can sense the presence of the key and automatically unlock the door. When the owner leaves the vehicle a certain distance, the PKE module can automatically lock the door; the RF remote control sending and receiving module is used to realize the remote control function of the vehicle. The sending module is responsible for the user's control The command is sent to the vehicle via wireless signals, and the receiving module is responsible for receiving the signal from the vehicle and converting it into information that the user can understand, such as using the WeChat applet to control the vehicle or carrying a mobile phone or remote control to achieve sensorless start, automatic unlocking and powering on when getting in the vehicle, and automatic power off and defense when leaving; the horn drive module controls the sound of the vehicle's horn; the ignition switch and ignition device control module is responsible for controlling the vehicle's ignition switch and ignition device. When the user operates the ignition switch or ignition device, the module will receive the corresponding signal and control the corresponding actuator action; the turn signal and horn control module controls the vehicle's turn signal and horn. When the user operates the turn signal or horn, the module receives the corresponding signal and controls the corresponding light or sound.The electronic steering column lock, seat bucket lock, and fuel tank lock control modules control the electronic steering column lock, seat bucket lock, and fuel tank lock of the vehicle respectively; the LIN bus communication module is used to realize data communication between various modules inside the vehicle. The LIN bus communication module adopts a low-cost serial communication protocol and can build an efficient and reliable data transmission network inside the vehicle; the acceleration sensor module is used to detect the acceleration change of the vehicle. The acceleration sensor module can send the detected acceleration data to the main control module or other modules that need the data to realize the corresponding function or control. ;

[0025] Data preprocessing component 2, the data preprocessing component 2 is used to perform data cleaning on the multiple module monitoring data sets to obtain multiple module monitoring data areas. Data cleaning is performed on the acquired multiple module monitoring data sets to remove noise, errors and other data in the data, so as to obtain more accurate and reliable multiple module monitoring data areas, each data area corresponds to the monitoring data of a module. Specifically, data cleaning may include missing value processing (check whether there are missing values ​​in the data set, that is, some fields or records have no data, and choose to fill them according to the specific situation (such as using the average, median, mode, etc.)), abnormal value processing (identify and process abnormal values ​​in the data, that is, those values ​​that obviously deviate from the normal range, for abnormal values, you can choose to delete, replace or correct), duplicate value processing (detect and process duplicate records in the data, for duplicate records, you can choose to delete or merge into a unique record), error value processing (check the error records in the data, such as format errors, logical errors, etc., for error values, choose to correct or delete according to the specific situation) and format standardization (unify the format of the data for subsequent analysis and processing), etc.

[0026] The data extraction component 3 is used to extract the first module monitoring data area corresponding to the first module according to the multiple module monitoring data areas, wherein the first module includes the various modules. The first module refers to any one of all the modules of the two-wheeled vehicle. Specifically, the data extraction component extracts the monitoring data areas corresponding to all the modules of the two-wheeled vehicle according to the acquired multiple module monitoring data areas. For example, when the first module is the power module of the two-wheeled vehicle, extracting its monitoring data area means obtaining data on the current state and operation of the power module, which may include voltage data (total battery pack voltage and single cell voltage, etc.), current data (current size during charging or discharge current during driving), temperature data (battery pack temperature and module internal temperature), status data (charging and discharging status or fault status), battery health status and energy consumption data, etc.

[0027] Anomaly detection feature learning component 4, the anomaly detection feature learning component 4 performs anomaly detection feature learning on the first module based on the embedded anomaly detection feature learning function and the anomaly detection feature learning particle group, and constructs the first module anomaly detection channel. Based on the embedded anomaly detection feature learning function and the anomaly detection feature learning particle group, the first module (for example, the power module of the two-wheeled vehicle) is subjected to anomaly detection feature learning, and the first module anomaly detection channel is constructed, that is, the anomaly detection channel of all modules of the two-wheeled vehicle is constructed. Specifically, the anomaly detection feature learning function uses a supervised learning method (such as cluster analysis, isolation forest, classifier, etc.) to learn or extract features related to the abnormal state from the real-time monitoring data of the first module. The anomaly detection feature learning particle group is used to optimize the parameters or structure of the anomaly detection feature learning function to find the feature set that can best distinguish between normal and abnormal states. Each particle represents a possible feature learning function configuration (such as parameter setting, model structure, etc.). After determining the effective feature set, an anomaly detection channel for the first module is constructed to receive the real-time monitoring data as input and output whether the data is abnormal.

[0028] Next, the specific configuration of the anomaly detection feature learning component 4 will be described in detail. Figure 2As shown, the abnormal detection feature learning component 4 may further include: connecting the data interaction component, reading the abnormal detection record data of the first module, and obtaining multiple module abnormal detection record sets, wherein each module abnormal detection record set includes a module monitoring data area record set and a module abnormal identification data record set. The abnormal detection record data refers to the abnormality-related record data of the first module monitored when the two-wheeled vehicle is running. Specifically, when the data interaction component is connected and triggered to read the abnormal detection record data, the abnormal data is collected from each module of the two-wheeled vehicle respectively to form multiple module abnormal detection record sets, each record set corresponds to a specific module, and contains all abnormal detection records generated during the operation of the module, including a module monitoring data area record set and a module abnormal identification data record set. The module monitoring data area record set records the monitoring data when the module is operating normally, which is used to establish a benchmark or threshold for normal operation; the module abnormal identification data record set records the abnormal data detected and recorded by the system when an abnormal situation occurs in the module. According to the abnormal detection feature learning function and the abnormal detection feature learning particle swarm, the multiple module abnormal detection record sets are supervised and learned, and multiple module abnormal detection models that meet the abnormal detection feature learning constraints are constructed. The anomaly detection feature learning function and the anomaly detection feature learning particle swarm are used to perform supervised learning on multiple module anomaly detection record sets. By iteratively updating the position and velocity of the particles, the particle swarm can gradually converge to the optimal model parameters or configuration to meet the constraints of the anomaly detection feature learning. The constraints may include the performance indicators of the model on the validation set (such as accuracy, recall rate, etc.), the complexity of the model (to avoid overfitting), and the real-time performance of the model (to ensure that new abnormal data can be processed in a timely manner), etc., to build multiple module anomaly detection models that meet the anomaly detection feature learning constraints. Specifically, the anomaly detection feature learning function will extract key features from multiple module anomaly detection record sets, and use these features and corresponding labels to train the anomaly detection model, that is, to obtain multiple module anomaly detection models. The multiple module anomaly detection models are merged as parallel nodes to generate the first module anomaly detection channel. The anomaly detection models trained for different modules of two-wheeled vehicles are integrated together to form a unified anomaly detection channel for the first module. Specifically, the node merger is to integrate the anomaly detection information from different modules, which is responsible for real-time monitoring of the operating status of the first module and triggering corresponding alarms when an anomaly is found, so as to perform more comprehensive and accurate anomaly detection on the first module.

[0029] The specific configuration of the anomaly detection feature learning component 4 will be described in detail below. The anomaly detection feature learning component 4 may further include: extracting a first module anomaly detection record set according to the multiple module anomaly detection record sets, wherein the first module anomaly detection record set includes any one module anomaly detection record set in the multiple module anomaly detection record sets. Divide the first module anomaly detection record set according to the preset data partitioning condition to generate a first module anomaly detection training set and a first module anomaly detection verification set. Divide the first module anomaly detection record set according to the preset data partitioning condition, for example, a common partitioning ratio is 70% training set and 30% verification set, that is, 70% of the data is used to train the model, and the remaining 30% is used to verify the performance of the model. Randomly extract a first anomaly detection feature learning particle according to the anomaly detection feature learning particle group, wherein the anomaly detection feature learning particle group includes multiple anomaly detection feature learning particles. Each particle of the anomaly detection feature learning particle swarm represents a possible anomaly detection feature learning solution or parameter configuration, and each particle has certain position and velocity attributes, indicating a specific point or configuration in the feature learning parameter space. Specifically, in the iterative process of particle swarm optimization, a first anomaly detection feature learning particle is randomly selected from multiple anomaly detection feature learning particle swarms for evaluation or update. The first module anomaly detection training set is cross-supervised learned according to the first anomaly detection feature learning particle to generate a first module anomaly detector. The first module anomaly detection training set is cross-trained using the first anomaly detection feature learning particle (i.e., the parameter configuration or feature set represented by the particle), which means that the training set is divided into multiple subsets (such as k-fold cross validation) for repeated training and verification to obtain an anomaly detector for the module, i.e., the first module anomaly detector. The first module anomaly detector is verified according to the first module anomaly detection verification set to obtain a first detector verification result, wherein the first detector verification result includes a first anomaly detection output accuracy coefficient and a first anomaly detection output error coefficient. By inputting the data of the validation set into the first module anomaly detector, the prediction result of the detector for each sample can be obtained, and compared with the true label of the sample in the validation set, the first detector verification result is calculated, including the first anomaly detection output accuracy coefficient and the first anomaly detection output error coefficient. Specifically, the first anomaly detection output accuracy coefficient is the ratio of the number of correctly predicted samples (including correctly predicted normal and abnormal) to the total number of samples, which measures the accuracy of the overall prediction of the detector. The higher the accuracy, the better the performance of the detector; the first anomaly detection output error coefficient indicates the degree of prediction error of the detector. The lower the error coefficient, the less prediction error of the detector and the better the performance. The first detector verification result is input into the anomaly detection feature learning function to obtain the first anomaly detection feature learning coefficient.When the verification result of the first detector is input into the anomaly detection feature learning function, the anomaly detection feature learning function adjusts and optimizes the features used by the anomaly detector according to these evaluation indicators, and the function outputs a set of new coefficients or parameters, namely the first anomaly detection feature learning coefficients, which represent the optimized feature selection or model parameters, and may include feature weights, thresholds, model parameters, etc. It is determined whether the first anomaly detection feature learning coefficients meet the anomaly detection feature learning constraints. If the first anomaly detection feature learning coefficients meet the anomaly detection feature learning constraints, the first module anomaly detector is set as the first module anomaly detection model, and the first module anomaly detection model is added to the multiple module anomaly detection models. If the first anomaly detection feature learning coefficients meet the anomaly detection feature learning constraints, the current first module anomaly detector is set as the first module anomaly detection model, that is, the detector has passed the verification and evaluation, and is added to the set of multiple module anomaly detection models, including the anomaly detection models of all modules of the two-wheeled vehicle, so that the anomalies in each module can be accurately identified.

[0030] The specific configuration of the anomaly detection feature learning component 4 will be described in detail below. The anomaly detection feature learning component 4 may further include: the anomaly detection feature learning function is:

[0031]

[0032] Among them, FICFAD represents the feature learning coefficient of anomaly detection, FADX represents the output accuracy coefficient of anomaly detection, ADXO represents the output accuracy coefficient of preset anomaly detection, FADY represents the output error coefficient of anomaly detection, and ADYO represents the output error coefficient of preset anomaly detection.

[0033] The specific configuration of the anomaly detection feature learning component 4 will be described in detail below. The anomaly detection feature learning component 4 may further include: if the first anomaly detection feature learning coefficient does not satisfy the anomaly detection feature learning constraint, activating the incremental learning model in the anomaly detection feature learning component. When the first anomaly detection feature learning coefficient does not satisfy the constraint, it means that the current anomaly detector has poor performance on the validation set, which may be due to the fact that some features are not correctly selected or optimized, then activating the incremental learning model in the anomaly detection feature learning component, wherein the incremental learning model will fine-tune and iterate the current model parameters or features according to new data or feedback information (such as error cases on the validation set) to improve the performance of the model on subsequent data. Obtain the validation loss data set corresponding to the first module anomaly detector. When training the first module anomaly detector, the training set is used to train the model, and the validation set is used to evaluate the performance of the model. In the validation stage, the data of the validation set is input into the model, and the difference between the prediction result of the model and the actual label, i.e., the loss value, is calculated. The validation loss data set is a collection of these loss values, which records the loss value of the model at each iteration on the validation set. Taking the anomaly detection feature learning constraint as the optimization learning target, the first module anomaly detector is optimized and learned according to the incremental learning model and the verification loss data set until the optimization learning target is met, and then the first module anomaly detection model is generated. The verification loss data set is used as input, and the performance of the model on the verification set is analyzed. According to the anomaly detection feature learning constraint and the current performance, the direction to be optimized is determined, and the incremental learning model is used to update and optimize the parameters or features of the first module anomaly detector. This process is repeated until the performance of the model on the verification set meets the preset optimization learning target, that is, when the performance of the first module anomaly detector meets the optimization learning target, the first module anomaly detection model is generated.

[0034] Anomaly analysis component 5, based on the anomaly detection channel of the first module, combined with the embedded abnormal feature coupling detection model, performs anomaly analysis on the monitoring data area of ​​the first module to generate an abnormal feature coupling detection index of the first module. Based on the anomaly detection channel of the first module, combined with the embedded abnormal feature coupling detection model, anomaly analysis is performed on the monitoring data area of ​​the first module to generate an abnormal feature coupling detection index of the first module, aiming to deeply analyze and quantify the abnormal state of the first module (i.e., all modules of the two-wheeled vehicle), wherein the abnormal feature coupling detection model is a model that specifically uses graph models, neural networks, association rule learning and other technologies to detect the coupling relationship between abnormal features. The anomaly analysis uses the output of the anomaly detection channel of the first module (i.e., whether there is an abnormality) and the abnormal feature coupling detection model to deeply analyze the monitoring data area of ​​the first module, for example, considering data at multiple time points, the relationship between multiple abnormal features, and the changes of these features over time. Finally, an abnormal feature coupling detection index of the first module is generated. The abnormal feature coupling detection index combines multiple abnormal features in the monitoring data area of ​​the first module and the coupling relationship between these features, and is used to evaluate the abnormal state and severity of the first module. Specifically, the real-time monitoring data of the first module is input into the abnormal detection channel to determine whether there is an abnormality. If an abnormality is detected, it is input into the abnormal feature coupling detection model for a more in-depth abnormal feature analysis. The abnormal feature coupling detection model considers the relationship between multiple abnormal features and the changing trend of these features over time, and generates one or more abnormal feature coupling detection results. Based on the detection results, the abnormal feature coupling detection index of the first module is generated.

[0035] Below, the specific configuration of the anomaly analysis component 5 will be described in detail. The anomaly analysis component 5 may further include: inputting the monitoring data area of ​​the first module into the anomaly detection channel of the first module to obtain the anomaly identification result of the first module. The data of the monitoring data area of ​​the first module is input into the anomaly detection channel of the first module. The anomaly detection channel of the first module uses a built-in algorithm to perform anomaly detection on the monitoring data to determine whether there are anomalies in the monitoring data of each module, and finally obtains the anomaly identification result of the first module, which is usually a binary classification (normal / abnormal) or a more complex classification (such as different types of anomalies) to indicate whether there are anomalies in the input data, as well as the type and degree of the anomaly. The anomaly feature coupling detection model includes an anomaly feature depth detection network, an anomaly feature breadth detection network and an anomaly feature coupling weighted network. Specifically, the abnormal feature depth detection network uses deep learning (such as neural network) to focus on the abnormal features in the deep analysis data and identify the abnormal features; the abnormal feature breadth detection network pays more attention to the breadth and diversity of the data, and may use statistical methods, machine learning algorithms, etc. to detect various abnormal patterns in the data set, with the goal of ensuring that as many abnormal types as possible can be covered; the abnormal feature coupling weighted network is the result of combining the deep detection network and the breadth detection network, and weighting them to generate the final abnormal detection result according to the importance of each detection result. According to the abnormal feature deep detection network, the abnormal depth evaluation of the first module abnormal recognition result is performed to obtain the first module abnormal depth detection index. The first module abnormal depth detection index is the result output by the abnormal feature deep detection network, which is used to quantify the depth or severity of the abnormal features in the first module. According to the abnormal feature breadth detection network, the abnormal risk prediction of the first module abnormal recognition result is performed to obtain the first module abnormal risk detection index. The first module abnormal risk detection index is the result output by the abnormal feature breadth detection network, which is used to quantify the risk level that the abnormal features in the first module may bring. The first module abnormal depth detection index and the first module abnormal risk detection index are weightedly calculated according to the abnormal feature coupling weighted network to obtain the first module abnormal feature coupling detection index.

[0036] The specific configuration of the abnormal analysis component 5 will be described in detail below. The abnormal analysis component 5 may further include: inputting the first module monitoring data area into the first module abnormal detection channel, and according to the first module abnormal detection channel, each module abnormal detection model respectively identifies the abnormality of the first module monitoring data area, and obtains multiple groups of first module abnormal identification data. Among them, in the first module abnormal detection channel, there are usually multiple independent abnormal detection models, which detect different abnormal types or data features respectively. When the data of the first module monitoring data area is input into the first module abnormal detection channel, each abnormal detection model in the channel will process and analyze these data respectively, determine whether the input data is abnormal, and generate corresponding abnormal identification results, and obtain multiple groups of first module abnormal identification data. Data fusion is performed according to the multiple groups of first module abnormal identification data to generate the first module abnormal identification results. The abnormal identification data detected by the multiple abnormal detection models of the first module abnormal detection channel are fused to generate the first module abnormal identification data. Similarly, each module can generate the abnormal identification data corresponding to the module.

[0037] Abnormal warning component 6, the abnormal warning component 6 is used to generate an abnormal warning signal for the first module when the abnormal feature coupling detection index of the first module is greater than / equal to the abnormal feature coupling detection threshold. The abnormal feature coupling detection threshold is a pre-set value or standard used to determine whether the abnormal feature coupling detection index has reached the level of triggering the warning signal. It is usually set based on historical data to ensure that the warning is triggered when a problem that really needs attention occurs, while avoiding false alarms. When the abnormal feature coupling detection index is greater than or equal to the abnormal feature coupling detection threshold, an abnormal warning signal for the first module is generated, which can be a simple alarm prompt or detailed information of the abnormality.

[0038] Below, the specific configuration of the abnormal feature operation and maintenance component will be described in detail. The multi-module joint management system also includes an abnormal feature operation and maintenance component, which is used to: activate the two-wheeled vehicle fault prediction model and the two-wheeled vehicle fault operation and maintenance model in the abnormal feature operation and maintenance component according to the abnormal warning signal of the first module. The abnormal feature operation and maintenance component is used to handle abnormalities and faults in the two-wheeled vehicle system, including a two-wheeled vehicle fault prediction model and a two-wheeled vehicle fault operation and maintenance model. The abnormal identification result of the first module is input into the two-wheeled vehicle fault prediction model to obtain the fault prediction result of the first module. When the abnormal warning signal of the first module is triggered, the abnormal feature operation and maintenance component starts to run, the two-wheeled vehicle fault prediction model is activated, and based on the current abnormal features (such as voltage instability, current abnormality, etc.) and historical data, it attempts to predict the possible fault type, occurrence time and the impact of the fault on the system performance. The fault prediction result of the first module is input into the two-wheeled vehicle fault operation and maintenance model to obtain the first module fault operation and maintenance plan. After the fault prediction model outputs the fault prediction result, the fault operation and maintenance model will be activated. The fault operation and maintenance model formulates corresponding operation and maintenance strategies according to the fault prediction result, such as replacing key components in advance, adjusting system parameters, issuing maintenance notices, etc., to generate a first module fault operation and maintenance plan. The first module is operated and maintained according to the first module fault operation and maintenance plan. The first module is operated and maintained according to the first module fault operation and maintenance plan to ensure that when a fault or potential fault occurs in the first module, intervention and recovery can be carried out quickly and effectively to maintain the stability and availability of the system.

[0039] Although the present application makes various references to certain components in the system according to the embodiments of the present application, any number of different components may be used and run on a user terminal and / or server, and the various units and components 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.

[0040] The above specific implementations do not constitute a limitation on the protection scope of this 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 this application should be included in the protection scope of this application.

Claims

1. A multi-module joint pipe system for a two-wheeled vehicle, characterized in that: The multi-module joint pipe system comprises: A data interaction component, which is used to read the real-time monitoring data of each module in the two-wheeled vehicle to obtain multiple module monitoring data sets; A data preprocessing component, the data preprocessing component is used to perform data cleaning on the plurality of module monitoring data sets to obtain a plurality of module monitoring data areas; A data extraction component, the data extraction component is used to extract a first module monitoring data area corresponding to a first module according to the plurality of module monitoring data areas, wherein the first module includes the respective modules; An anomaly detection feature learning component, wherein the anomaly detection feature learning component performs anomaly detection feature learning on the first module based on an embedded anomaly detection feature learning function and anomaly detection feature learning particle swarm to construct an anomaly detection channel for the first module; An anomaly analysis component, which performs an anomaly analysis on the monitoring data area of ​​the first module based on the anomaly detection channel of the first module and in combination with an embedded anomaly feature coupling detection model to generate an anomaly feature coupling detection index of the first module; An abnormal warning component, the abnormal warning component is used to generate an abnormal warning signal of the first module when the abnormal feature coupling detection index of the first module is greater than / equal to the abnormal feature coupling detection threshold; The anomaly detection feature learning component performs anomaly detection feature learning on the first module based on the embedded anomaly detection feature learning function and the anomaly detection feature learning particle swarm to construct an anomaly detection channel for the first module, including: Connecting the data interaction component, reading the anomaly detection record data of the first module, and obtaining a plurality of module anomaly detection record sets, wherein each module anomaly detection record set includes a module monitoring data area record set and a module anomaly identification data record set; Performing supervised learning on the plurality of module anomaly detection record sets according to the anomaly detection feature learning function and the anomaly detection feature learning particle swarm, and constructing a plurality of module anomaly detection models that meet anomaly detection feature learning constraints; Merging the multiple module anomaly detection models as parallel nodes to generate the first module anomaly detection channel; Among them, according to the anomaly detection feature learning function and the anomaly detection feature learning particle swarm, supervised learning is performed on the multiple module anomaly detection record sets to construct multiple module anomaly detection models that meet the anomaly detection feature learning constraints, including: Extracting a first module anomaly detection record set according to the multiple module anomaly detection record sets, wherein the first module anomaly detection record set includes any one module anomaly detection record set among the multiple module anomaly detection record sets; Divide the first module anomaly detection record set according to a preset data division condition to generate a first module anomaly detection training set and a first module anomaly detection verification set; randomly extracting a first anomaly detection feature learning particle according to the anomaly detection feature learning particle group, wherein the anomaly detection feature learning particle group includes a plurality of anomaly detection feature learning particles; Performing cross-supervised learning on the first module anomaly detection training set according to the first anomaly detection feature learning particle to generate a first module anomaly detector; Verifying the first module anomaly detector according to the first module anomaly detection verification set to obtain a first detector verification result, wherein the first detector verification result includes a first anomaly detection output accuracy coefficient and a first anomaly detection output error coefficient; Inputting the first detector verification result into the anomaly detection feature learning function to obtain a first anomaly detection feature learning coefficient; Determining whether the first anomaly detection feature learning coefficient satisfies the anomaly detection feature learning constraint; If the first anomaly detection feature learning coefficient satisfies the anomaly detection feature learning constraint, setting the first module anomaly detector as a first module anomaly detection model, and adding the first module anomaly detection model to the multiple module anomaly detection models; Among them, the anomaly detection feature learning function is: Among them, FICFAD represents the anomaly detection feature learning coefficient, FADX represents the anomaly detection output accuracy coefficient, ADXO represents the preset anomaly detection output accuracy coefficient, FADY represents the anomaly detection output error coefficient, and ADYO represents the preset anomaly detection output error coefficient; Wherein, judging whether the first anomaly detection feature learning coefficient satisfies the anomaly detection feature learning constraint includes: If the first anomaly detection feature learning coefficient does not satisfy the anomaly detection feature learning constraint, activating an incremental learning model in the anomaly detection feature learning component; Obtain a validation loss data set corresponding to the anomaly detector of the first module; Taking the anomaly detection feature learning constraint as the optimization learning objective; The first module anomaly detector is optimized and learned according to the incremental learning model and the verification loss data set until the optimization learning goal is met, thereby generating the first module anomaly detection model.

2. The multi-module joint pipe system according to claim 1, characterized in that: The anomaly analysis component performs an anomaly analysis on the monitoring data area of ​​the first module based on the anomaly detection channel of the first module in combination with the embedded anomaly feature coupling detection model to generate an anomaly feature coupling detection index of the first module, including: Inputting the first module monitoring data area into the first module abnormality detection channel to obtain the first module abnormality identification result; The abnormal feature coupling detection model includes an abnormal feature depth detection network, an abnormal feature breadth detection network and an abnormal feature coupling weighted network; Performing an abnormal depth evaluation on the abnormal recognition result of the first module according to the abnormal feature depth detection network to obtain an abnormal depth detection index of the first module; Performing an abnormal risk prediction on the abnormal recognition result of the first module according to the abnormal feature breadth detection network to obtain an abnormal risk detection index of the first module; The first module abnormal depth detection index and the first module abnormal risk detection index are weightedly calculated according to the abnormal feature coupling weighted network to obtain the first module abnormal feature coupling detection index.

3. The multi-module joint pipe system according to claim 2, characterized in that: Inputting the first module monitoring data area into the first module abnormality detection channel to obtain the first module abnormality identification result, including: Inputting the first module monitoring data area into the first module abnormality detection channel, and performing abnormality identification on the first module monitoring data area according to each module abnormality detection model in the first module abnormality detection channel, to obtain multiple groups of first module abnormality identification data; Data fusion is performed based on the multiple groups of first module abnormality identification data to generate the first module abnormality identification result.

4. The multi-module joint pipe system according to claim 2, characterized in that: The multi-module joint management system further includes an abnormal feature operation and maintenance component, which is used to: activating a two-wheeled vehicle fault prediction model and a two-wheeled vehicle fault operation and maintenance model in the abnormal feature operation and maintenance component according to the abnormal warning signal of the first module; Inputting the first module abnormality recognition result into the two-wheeled vehicle fault prediction model to obtain the first module fault prediction result; Inputting the first module fault prediction result into the two-wheeled vehicle fault operation and maintenance model to obtain the first module fault operation and maintenance plan; Perform operation and maintenance management on the first module according to the first module fault operation and maintenance plan.

Images