Fault detection system and method for electro-tricycle

Abstract

The invention discloses a fault detection system and method for an electric tricycle, and relates to the related field of fault detection, and the system comprises an electrical monitoring data collection module which is used for collecting the monitoring data of an electrical layer in real time, transmitting the monitoring data back to a cloud platform, and activating a sensing array of a mechanical layer if there is an abnormality; the data coupling verification module is used for monitoring and acquiring mechanical monitoring data, transmitting the mechanical monitoring data back to the cloud platform, performing time synchronization alignment and data coupling verification on the electrical and mechanical monitoring data, and determining an abnormal coupling feature set and a feature abnormality degree; the fault probability prediction module is used for calling a health evaluation mechanism, performing shutdown fault probability prediction and outputting a fault prediction probability; and the fault early warning module is used for generating a fault early warning signal for early warning prompt. The technical problems that existing electric tricycle fault detection is low in prediction accuracy and high in shutdown fault occurrence rate are solved, and the technical effects of improving the prediction accuracy of shutdown faults and reducing the shutdown fault occurrence rate are achieved.

Description

Technical Field

[0001] The present application relates to the field of fault detection, and in particular to a fault detection system and method for an electric tricycle. Background Art

[0002] With the widespread use of electric tricycles in logistics, short-distance transportation and other fields, their safety and reliability have become the focus of users. In order to ensure the normal operation of electric tricycles, it is very important to detect and handle potential downtime failures in a timely manner. Traditional fault detection methods mostly rely on manual inspections and regular maintenance. This method is difficult to detect potential downtime failures in a timely manner. It is often not discovered until the fault has occurred or is about to cause downtime, missing the best time for maintenance. In addition, due to the lack of intelligent monitoring and early warning methods, the prediction accuracy of downtime failures is not high, and downtime events cannot be effectively prevented.

[0003] In the current related technologies, the fault detection of electric tricycles has technical problems such as low prediction accuracy and high occurrence rate of shutdown failures. Summary of the invention

[0004] The present application provides a fault detection system and method for an electric tricycle, which collects electrical layer data of a target tricycle in real time, activates a sensor array at the mechanical layer in the event of an abnormality, collects mechanical monitoring data, performs time synchronization alignment and data coupling verification on the two types of data, identifies abnormal coupling feature sets and their feature abnormality that may cause shutdown failures, calls a health assessment mechanism based on feature abnormality matching, uses a machine learning algorithm to analyze real-time vehicle condition features and abnormal coupling feature sets, predicts the probability of shutdown failures, and immediately generates a warning signal when the predicted probability exceeds a predetermined risk threshold, and uses a client to remind the user to handle the problem, thereby achieving the technical effect of improving the prediction accuracy of shutdown failures and reducing the occurrence rate of shutdown failures.

[0005] The present application provides a fault detection system for an electric tricycle, including: an electrical monitoring data acquisition module, which is used to collect monitoring data of the electrical layer of a target tricycle in real time and transmit it back to a cloud platform. If there is an abnormality in the electrical monitoring data, the sensor array of the mechanical layer is activated; a data coupling verification module, which is used to obtain mechanical monitoring data through the sensor array monitoring and transmit it back to the cloud platform, perform time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, determine an abnormal coupling feature set, and a feature abnormality degree; a fault probability prediction module, which is used to call a health evaluation mechanism according to the feature abnormality degree matching, and use the health evaluation mechanism to predict the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and output the fault prediction probability; a fault warning module, which is used to generate a fault warning signal if the fault prediction probability exceeds a predetermined risk threshold, and transmit it back to the client of the target tricycle for a warning prompt.

[0006] In a possible implementation, the data coupling verification module is also used for: the electrical monitoring data includes at least battery data, motor data and electrical control data, and the mechanical monitoring data includes at least part vibration data and driving status data; obtaining electrical monitoring data sets and mechanical monitoring data sets under multiple monitoring timestamps, and performing time synchronization alignment on the electrical monitoring data sets and the mechanical monitoring data sets to obtain multiple groups of monitoring data sets; performing data coupling verification on the multiple groups of monitoring data sets based on a long short-term memory network, and outputting an abnormal coupling feature set, wherein the abnormal coupling feature includes an abnormal type, and the abnormal type is identified by an abnormal intensity and an abnormal frequency; performing a feature abnormality degree analysis based on the abnormal coupling feature set, and outputting the feature abnormality degree.

[0007] In a possible implementation, the data coupling verification module is also used to: obtain characteristic benchmark indicators of the abnormal type, wherein the characteristic benchmark indicators include benchmark strength and benchmark frequency; perform deviation analysis on the abnormal strength and abnormal frequency of multiple abnormal types in the abnormal coupling feature set according to the benchmark strength and benchmark frequency, and weightedly obtain multiple abnormal deviation coefficients, wherein the abnormal deviation coefficient is positively correlated with the abnormal strength and abnormal frequency; configure multiple abnormal influence degrees according to the multiple abnormal types, perform weighted fusion on the multiple abnormal deviation coefficients, and output the characteristic abnormality degree.

[0008] In a possible implementation, the fault probability prediction module is also used to: calculate the ratio of the characteristic abnormality to the historical maximum characteristic abnormality, and set it as the abnormality scale coefficient; if the abnormality scale coefficient is less than 0.6, call the first health evaluation channel to predict the shutdown fault probability, and output the fault prediction probability; if the abnormality scale coefficient is greater than or equal to 0.6, call the first health evaluation channel and the second health evaluation mechanism to predict the shutdown fault probability, and output the fault prediction probability.

[0009] In a possible implementation, the fault probability prediction module is also used to: take the vehicle attribute characteristics of the target tricycle as a constraint, call the operation log of the same tricycle in the cloud platform, collect a sample vehicle condition feature set and multiple sample abnormal coupling feature sets, and count the historical vehicle downtime proportions of different sample vehicle condition characteristics and different sample abnormal coupling feature sets in the historical time window, set them as the sample downtime failure probability, and obtain the sample downtime failure probability set; use the sample vehicle condition feature set, multiple sample abnormal coupling feature sets and sample downtime failure probability set to supervise the BP neural network training until convergence, and obtain the first health evaluation channel.

[0010] In a possible implementation, the fault probability prediction module is also used for: if the abnormal scale coefficient is greater than or equal to 0.6, calling the first health evaluation channel, predicting the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and outputting the first fault prediction probability; calling the second health evaluation mechanism to perform big data retrieval, and outputting the second fault prediction probability, wherein the big data retrieval includes the sample matching data volume; determining the big data retrieval accuracy based on the sample matching data volume analysis, and configuring the trusted weight according to the big data retrieval accuracy and the prediction accuracy of the first health evaluation channel; performing weighted calculation on the first fault prediction probability and the second fault prediction probability according to the trusted weight, and outputting the fault prediction probability.

[0011] In a possible implementation, the fault probability prediction module is also used to: use the vehicle attribute characteristics as equipment retrieval constraints, use real-time vehicle condition characteristics and abnormal coupling feature sets as feature retrieval constraints, use electric tricycle parking failures as a guide, perform positive sample retrieval based on big data technology, and obtain a retrieval data set that meets a predetermined similarity threshold; and count the proportion of shutdown failures in the retrieval data set, and set it as the second fault prediction probability.

[0012] The present application also provides a fault detection method for an electric tricycle, comprising: collecting monitoring data of the electrical layer of the target tricycle in real time and transmitting it back to a cloud platform, and if there is an abnormality in the electrical monitoring data, activating a sensor array of the mechanical layer; acquiring mechanical monitoring data through the sensor array monitoring and transmitting it back to the cloud platform, performing time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, and determining an abnormal coupling feature set and a feature abnormality degree; calling a health evaluation mechanism according to the feature abnormality degree matching, and using the health evaluation mechanism to predict the probability of shutdown failure according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and outputting a fault prediction probability; if the fault prediction probability exceeds a predetermined risk threshold, generating a fault warning signal and transmitting it back to the client of the target tricycle for a warning prompt.

[0013] The electric tricycle fault detection system and method proposed in this application is intended to collect the monitoring data of the electrical layer of the target tricycle in real time through the electrical monitoring data acquisition module and transmit it back to the cloud platform. If there is an abnormality in the electrical monitoring data, the sensor array of the mechanical layer is activated, and the mechanical monitoring data is acquired through the sensor array monitoring and transmitted back to the cloud platform. The electrical monitoring data and the mechanical monitoring data are time-synchronized and data-coupled verified through the data coupling verification module to determine the abnormal coupling feature set and the feature abnormality. The health evaluation mechanism is called according to the feature abnormality matching through the fault probability prediction module, and the health evaluation mechanism is used to predict the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and the fault prediction probability is output. If the fault prediction probability exceeds the predetermined risk threshold, a fault warning signal is generated through the fault warning module and transmitted back to the client of the target tricycle for a warning prompt, thereby achieving the technical effect of improving the prediction accuracy of shutdown faults and reducing the occurrence rate of shutdown faults. 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 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.

[0015] Figure 1 A schematic structural diagram of a fault detection system for an electric tricycle provided in an embodiment of the present application.

[0016] Figure 2 A flow chart of a fault detection method for an electric tricycle provided in an embodiment of the present application.

[0017] Explanation of the reference numerals: electrical monitoring data acquisition module 10 , data coupling verification module 20 , fault probability prediction module 30 , fault warning module 40 . DETAILED DESCRIPTION

[0018] 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.

[0019] 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.

[0020] 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.

[0021] The present application embodiment provides a fault detection system for an electric tricycle, such as Figure 1 As shown, the system comprises: The electrical monitoring data acquisition module 10 is used to collect the monitoring data of the electrical layer of the target tricycle in real time and transmit it back to the cloud platform. If there is an abnormality in the electrical monitoring data, the sensor array of the mechanical layer is activated.

[0022] Specifically, various sensors (such as voltage sensors, current sensors, temperature sensors, etc.) installed on the target tricycle are used to collect monitoring data of the electrical layer (electrical-related parts) such as batteries, motors, and control systems in real time. The collected data (battery voltage, current, temperature, internal resistance, etc.; motor speed, current, voltage, temperature, load, etc.; current control signals, drive system status, fault codes, etc.) are transmitted to the cloud platform via wireless or wired means. On the cloud platform side, the preset threshold or algorithm is used to determine whether there is an abnormality in the electrical monitoring data. If there is an abnormality, the sensor array of the mechanical layer is activated for further monitoring. The sensor array refers to a set of sensors used to monitor the status of the mechanical layer.

[0023] The data coupling verification module 20 is used to obtain mechanical monitoring data through the sensor array monitoring and transmit it back to the cloud platform, perform time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, and determine the abnormal coupling feature set and the feature abnormality degree.

[0024] Specifically, mechanical monitoring data (vehicle body vibration, vehicle speed, acceleration, etc.) is obtained through the sensor array of the mechanical layer (such as vibration sensors, vehicle speed sensors, accelerometers, etc.). On the cloud platform side, the electrical monitoring data and mechanical monitoring data are aligned according to the timestamp to ensure the temporal consistency of the two. The algorithm is used to perform coupling analysis on the electrical and mechanical monitoring data, that is, to analyze the correlation between different data sources, identify potential failure modes, thereby determining the abnormal coupling feature set and calculating the feature abnormality, where the abnormal coupling feature refers to the abnormal relationship pattern between electrical and mechanical data, indicating potential failures. The feature abnormality is an indicator to measure the degree of influence of abnormal coupling features on the overall failure probability.

[0025] In a possible implementation, the data coupling verification module 20 is also used for: the electrical monitoring data includes at least battery data, motor data and electrical control data, and the mechanical monitoring data includes at least part vibration data and driving status data; obtaining electrical monitoring data sets and mechanical monitoring data sets under multiple monitoring timestamps, and performing time synchronization alignment on the electrical monitoring data sets and the mechanical monitoring data sets to obtain multiple groups of monitoring data sets; performing data coupling verification on the multiple groups of monitoring data sets based on a long short-term memory network, and outputting an abnormal coupling feature set, wherein the abnormal coupling feature includes an abnormal type, and the abnormal type is identified by an abnormal intensity and an abnormal frequency; performing a feature abnormality degree analysis based on the abnormal coupling feature set, and outputting the feature abnormality degree.

[0026] Specifically, the electrical monitoring data acquisition module 10 and the sensor array of the mechanical layer respectively collect the electrical monitoring data and mechanical monitoring data of the electric tricycle. These data include but are not limited to battery data, motor data, electrical control data (such as current, voltage, temperature, etc.), as well as part vibration data and driving status data (such as vehicle speed, acceleration, etc.).

[0027] Since the electrical and mechanical data come from different sensors and have different sampling frequencies, they need to be time-synchronized and aligned. This is achieved by matching timestamps to ensure the temporal consistency of the electrical and mechanical data.

[0028] The electrical monitoring dataset and the mechanical monitoring dataset at multiple monitoring timestamps are input into the long short-term memory network (LSTM). LSTM is a special type of recurrent neural network (RNN) that excels at processing time series data. LSTM is used to perform data coupling verification on multiple sets of monitoring datasets, identifying potential abnormal coupling features by analyzing the time dependency and correlation between electrical and mechanical data. During the training phase, the LSTM network learns the normal relationship patterns between electrical and mechanical data; during the verification phase, it compares these patterns with the input data to identify any anomalies. Once the LSTM network detects an anomaly, it outputs a set of abnormal coupling features, which include the type of anomaly (such as battery overheating, abnormal motor vibration, etc.), as well as the abnormal intensity and abnormal frequency of each anomaly type.

[0029] Finally, feature abnormality analysis is performed based on the abnormal coupling feature set, including calculating the weight or score of each abnormal feature to evaluate its impact on the overall failure probability and outputting the feature abnormality. This implementation method uses a long short-term memory network for data coupling verification. LSTM can capture complex relationships and patterns in the data, improving the accuracy of anomaly detection.

[0030] In a possible implementation, the data coupling verification module 20 is also used to: obtain characteristic benchmark indicators of the abnormal type, wherein the characteristic benchmark indicators include benchmark strength and benchmark frequency; perform deviation analysis on the abnormal strength and abnormal frequency of multiple abnormal types in the abnormal coupling feature set according to the benchmark strength and benchmark frequency, and weightedly obtain multiple abnormal deviation coefficients, wherein the abnormal deviation coefficient is positively correlated with the abnormal strength and abnormal frequency; configure multiple abnormal influence degrees according to the multiple abnormal types, perform weighted fusion on the multiple abnormal deviation coefficients, and output the characteristic abnormality degree.

[0031] Specifically, characteristic benchmark indicators are established, which are derived from a large amount of historical data and are used to measure the intensity and frequency range of electrical and mechanical data under normal conditions. The characteristic benchmark indicators include benchmark intensity and benchmark frequency, which represent the intensity threshold and frequency threshold of different abnormal types under normal conditions, respectively.

[0032] For the anomaly coupling feature set obtained from the data coupling verification, a deviation analysis is performed on it, that is, the anomaly intensity and frequency of each anomaly type are compared with the corresponding benchmark intensity and benchmark frequency, and the degree of deviation between them is calculated. In order to quantify the degree of deviation, an anomaly deviation coefficient is calculated for each anomaly type. This coefficient is the weighted result of the anomaly intensity and frequency relative to the benchmark intensity and benchmark frequency, and is positively correlated with the anomaly intensity and frequency. That is, the stronger or more frequent the anomaly, the greater the anomaly deviation coefficient.

[0033] Different anomaly types have different impacts on the overall failure probability. Based on historical data, an anomaly impact is configured for each anomaly type to measure the impact of different anomaly types on the overall failure probability. These anomaly deviation coefficients are weighted and fused according to their corresponding anomaly impacts to obtain the final characteristic anomaly degree. This implementation method more accurately evaluates the anomaly type and characteristic anomaly degree by introducing characteristic benchmark indicators and anomaly deviation coefficients, thereby improving the accuracy of fault detection.

[0034] The fault probability prediction module 30 is used to call the health evaluation mechanism according to the feature abnormality matching, and use the health evaluation mechanism to predict the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and output the fault prediction probability.

[0035] Specifically, according to the magnitude of the feature abnormality, the corresponding health evaluation mechanism is matched and called. The health evaluation mechanism is a model established based on historical data and is used to evaluate the health status of the target tricycle. The health evaluation mechanism is used to predict the probability of downtime failure by combining the real-time vehicle condition characteristics (the current operating status and parameters of the target tricycle) and the abnormal coupling feature set, and the prediction results are output in the form of probability for decision making.

[0036] In a possible implementation, the fault probability prediction module 30 is also used to: calculate the ratio of the characteristic abnormality to the historical maximum characteristic abnormality, and set it as the abnormality scale coefficient; if the abnormality scale coefficient is less than 0.6, call the first health evaluation channel to predict the shutdown fault probability, and output the fault prediction probability; if the abnormality scale coefficient is greater than or equal to 0.6, call the first health evaluation channel and the second health evaluation mechanism to predict the shutdown fault probability, and output the fault prediction probability.

[0037] Specifically, the currently monitored characteristic abnormality and the historical maximum characteristic abnormality output by the data coupling verification module 20 are obtained, and the historical maximum characteristic abnormality is the maximum value of the characteristic abnormality monitored in the past period of time, which is used for comparison with the current characteristic abnormality. The ratio of these two values ​​is calculated to obtain the abnormal scale coefficient, which is a dimensionless indicator used to measure the relative severity of the current abnormal situation.

[0038] According to the size of the abnormal scale coefficient, if the abnormal scale coefficient is less than 0.6, it means that the current abnormal situation is relatively mild, and the first health evaluation channel is selected to predict the shutdown failure probability; if the abnormal scale coefficient is greater than or equal to 0.6, it means that the current abnormal situation is more serious, and the first health evaluation channel and the second health evaluation mechanism are selected to predict at the same time. Finally, the fault prediction probability is output according to the selected health evaluation channel or mechanism. Among them, the first health evaluation channel is a relatively simple and fast prediction model, which is suitable for situations where the abnormal situation is not serious; the second health evaluation mechanism is a more complex and more accurate model, which is suitable for situations where the abnormal situation is more serious. This implementation method can improve the prediction efficiency while ensuring the accuracy of the prediction by selecting different health evaluation channels according to the size of the abnormal scale coefficient. For situations where the abnormal situation is not serious, using a simple prediction model can get results faster; for situations where the abnormal situation is more serious, using a more complex model can get more accurate prediction results.

[0039] In a possible implementation, the fault probability prediction module 30 is also used to: take the vehicle attribute characteristics of the target tricycle as a constraint, call the operation log of the same tricycle in the cloud platform, collect a sample vehicle condition feature set and multiple sample abnormal coupling feature sets, and count the historical vehicle downtime proportions of different sample vehicle condition characteristics and different sample abnormal coupling feature sets in the historical time window, set them as the sample downtime failure probability, and obtain the sample downtime failure probability set; use the sample vehicle condition feature set, multiple sample abnormal coupling feature sets and sample downtime failure probability set to supervise the BP neural network training until convergence, and obtain the first health evaluation channel.

[0040] Specifically, the vehicle attribute characteristics of the target tricycle include the model, configuration, and age of the tricycle, which are the basis for screening similar tricycles. In the cloud platform, based on the vehicle attribute characteristics of the target tricycle, similar tricycles are screened out, and the operation logs of these tricycles are obtained. The operation logs contain the vehicle condition characteristics (road, speed, load, etc.) and abnormal coupling feature sets at different time points.

[0041] The vehicle condition features similar to the current vehicle condition of the target tricycle are extracted from the operation log to form a sample vehicle condition feature set. At the same time, the abnormal coupling feature sets corresponding to these vehicle condition features are extracted, and each sample vehicle condition feature corresponds to one or more abnormal coupling feature sets.

[0042] Within the historical time window (such as the past year, half a year or a certain period of time), the proportion of vehicle downtime under different sample vehicle condition characteristics and different sample abnormal coupling feature sets is statistically calculated. This proportion is the sample downtime failure probability.

[0043] The BP neural network is supervised and trained using the above-extracted sample vehicle condition feature set, multiple sample abnormal coupling feature sets and corresponding sample shutdown fault probability sets. The BP neural network is a multi-layer feedforward neural network that continuously adjusts the weights and biases of the network through the back propagation algorithm so that the output value of the network gradually approaches the true value. During the training process, the sample vehicle condition feature set and the sample abnormal coupling feature set are used as input, and the sample shutdown fault probability is used as the target output. Through multiple iterations of training, until the network converges (that is, the training error reaches the preset threshold or the number of training times reaches the upper limit). After the training is completed, the BP neural network constitutes the first health evaluation channel, which can output the fault prediction probability based on the real-time vehicle condition characteristics and abnormal coupling feature set. This implementation method uses a supervised training method based on the BP neural network to construct the first health evaluation channel. The BP neural network has a strong nonlinear fitting ability and can handle complex nonlinear relationships, thereby improving the accuracy of the prediction of the shutdown fault probability of electric tricycles.

[0044] In a possible implementation, the fault probability prediction module 30 is also used for: if the abnormal scale coefficient is greater than or equal to 0.6, calling the first health evaluation channel, predicting the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and outputting the first fault prediction probability; calling the second health evaluation mechanism to perform big data retrieval, and outputting the second fault prediction probability, wherein the big data retrieval includes the sample matching data volume; determining the big data retrieval accuracy based on the sample matching data volume analysis, and configuring the trusted weight according to the big data retrieval accuracy and the prediction accuracy of the first health evaluation channel; performing weighted calculation on the first fault prediction probability and the second fault prediction probability according to the trusted weight, and outputting the fault prediction probability.

[0045] Specifically, when the abnormal scale coefficient is greater than or equal to 0.6, it indicates that the current fault is more serious or complex and requires a more comprehensive assessment. At this time, the first health evaluation channel (supervised training model based on BP neural network) is first called to predict the shutdown fault probability based on the real-time vehicle condition characteristics and abnormal coupling feature set, and the first fault prediction probability is output.

[0046] At the same time, the second health evaluation mechanism is called to perform big data retrieval. Big data retrieval includes searching the operation logs and fault records of vehicles similar to the target tricycle (based on vehicle attribute characteristics) in the cloud platform to find sample data that matches the current fault situation. Based on the sample data found, the historical downtime fault probability is statistically analyzed and the second fault prediction probability is output.

[0047] The accuracy of big data retrieval is determined based on the analysis of the amount of sample matching data. The larger the amount of sample matching data, the more similar fault conditions are found, and the higher the accuracy of big data retrieval. The trusted weight is configured according to the accuracy of big data retrieval and the prediction accuracy of the first health assessment channel. Among them, the trusted weight refers to the weight assigned according to the prediction accuracy of the model or method, which is used to determine the contribution of each model prediction result in multi-model fusion prediction. If the big data retrieval accuracy is high, it means that the similar fault conditions found are relatively consistent with the current fault condition, so the credibility of the second fault prediction probability is high and should be given a larger weight; if the prediction accuracy of the first health assessment channel is high, a larger weight should be given to the first health assessment channel.

[0048] According to the credible weight, the first fault prediction probability and the second fault prediction probability are weighted and calculated to output the final fault prediction probability. The weighted calculation ensures that both prediction results can contribute to the final fault prediction probability, and at the same time, a reasonable weight distribution is performed according to their respective accuracies. This implementation method combines the first health evaluation channel (model based on BP neural network) and the second health evaluation mechanism (method based on big data retrieval), making full use of the advantages of both and improving the accuracy of fault prediction. When the abnormal scale coefficient is large, it indicates that the current fault is more complex or serious. At this time, the use of multi-model fusion prediction enhances the robustness of the system and avoids false positives or negative positives caused by single model prediction.

[0049] In a possible implementation, the fault probability prediction module 30 is also used to: use the vehicle attribute characteristics as equipment retrieval constraints, use real-time vehicle condition characteristics and abnormal coupling feature sets as feature retrieval constraints, use the electric tricycle parking failure as a guide, perform positive sample retrieval based on big data technology, and obtain a retrieval data set that meets a predetermined similarity threshold; and count the proportion of shutdown failures in the retrieval data set, and set it as the second fault prediction probability.

[0050] Specifically, vehicle attribute features (such as model, brand, configuration, etc.) are used as device retrieval constraints to ensure that the retrieved data is similar to the target tricycle in physical characteristics. Real-time vehicle condition features and abnormal coupling feature sets are used as feature retrieval constraints to ensure that the retrieved data is relevant to the current fault condition of the target tricycle. The parking fault of the electric tricycle is used as a guide to ensure that the retrieved data is related to the parking fault, thereby improving the pertinence and accuracy of the retrieval.

[0051] Based on the above retrieval constraints, a positive sample retrieval operation is performed in the big data warehouse in the cloud platform. Among them, positive sample retrieval refers to retrieving positive samples similar to the target object (i.e. samples with the same or similar characteristics as the target object) in the big data warehouse according to specific retrieval constraints. By utilizing the powerful processing capabilities of big data technology, a retrieval data set that meets the predetermined similarity threshold (a threshold used to determine the degree of similarity between the retrieved data and the target tricycle, and only when the data meets this threshold is it considered a valid retrieval result) is quickly screened out. This data set contains parking fault records of electric tricycles that are similar to the target tricycle in terms of vehicle attributes, real-time vehicle conditions, and abnormal coupling characteristics.

[0052] The retrieved data sets are statistically analyzed to calculate the downtime failure ratio. This ratio represents the probability of an electric tricycle having a downtime failure under similar conditions, and this ratio is output as the second fault prediction probability. This implementation method uses big data technology to retrieve positive sample data sets that are highly similar to the target tricycle in terms of vehicle attributes, real-time vehicle conditions, and abnormal coupling characteristics. Therefore, the downtime failure ratio obtained based on these data sets has a high accuracy, which improves the accuracy of fault prediction.

[0053] The fault warning module 40 is used to generate a fault warning signal if the fault prediction probability exceeds a predetermined risk threshold, and transmit it back to the client of the target tricycle for early warning.

[0054] Specifically, if the fault prediction probability exceeds the predetermined risk threshold, that is, exceeds the set upper limit of the fault probability, a fault warning signal is generated. The warning signal is transmitted back to the client of the target tricycle (the device or interface used by the user to receive the warning signal, such as a mobile phone app, a vehicle display screen, etc.) to remind the user to pay attention and take corresponding measures. The embodiment of the present application adopts real-time acquisition of the electrical layer data of the target tricycle, and activates the sensor array of the mechanical layer in an abnormality, collects mechanical monitoring data, performs time synchronization alignment and data coupling verification on these two types of data, identifies the abnormal coupling feature set and its feature abnormality that may cause the shutdown failure, calls the health evaluation mechanism according to the feature abnormality matching, and uses the machine learning algorithm to analyze the real-time vehicle condition characteristics and the abnormal coupling feature set to predict the probability of the shutdown failure. When the predicted probability exceeds the predetermined risk threshold, a warning signal is immediately generated, and the user is reminded to handle the technical means through the client, so as to achieve the technical effect of improving the prediction accuracy of the shutdown failure and reducing the occurrence rate of the shutdown failure.

[0055] In the above, refer to Figure 1 The fault detection system of the electric tricycle according to the embodiment of the present invention is described in detail. Figure 2 A fault detection method for an electric tricycle according to an embodiment of the present invention is described.

[0056] The fault detection method for an electric tricycle according to an embodiment of the present invention is used to solve the technical problems of low prediction accuracy and high occurrence rate of shutdown failures in the existing electric tricycle fault detection, thereby achieving the technical effect of improving the prediction accuracy of shutdown failures and reducing the occurrence rate of shutdown failures.

[0057] The fault detection method for an electric tricycle includes: collecting monitoring data of the electrical layer of a target tricycle in real time and transmitting it back to a cloud platform; if there is an abnormality in the electrical monitoring data, activating a sensor array of a mechanical layer; acquiring mechanical monitoring data through the sensor array monitoring and transmitting it back to the cloud platform; performing time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, and determining an abnormal coupling feature set and a feature abnormality degree; calling a health evaluation mechanism according to the feature abnormality degree matching, and using the health evaluation mechanism to predict the shutdown failure probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and outputting a fault prediction probability; if the fault prediction probability exceeds a predetermined risk threshold, generating a fault warning signal and transmitting it back to the client of the target tricycle for a warning prompt.

[0058] Among them, the electrical monitoring data and the mechanical monitoring data are subjected to time synchronization alignment and data coupling verification, the abnormal coupling feature set and the feature abnormality degree are determined, and the following processing is performed: the electrical monitoring data at least includes battery data, motor data and electrical control data, and the mechanical monitoring data at least includes part vibration data and driving status data; the electrical monitoring data set and the mechanical monitoring data set under multiple monitoring timestamps are obtained, and the electrical monitoring data set and the mechanical monitoring data set are subjected to time synchronization alignment to obtain multiple groups of monitoring data sets; the multiple groups of monitoring data sets are subjected to data coupling verification based on the long short-term memory network, and the abnormal coupling feature set is output, wherein the abnormal coupling feature includes the abnormal type, and the abnormal type is identified by the abnormal intensity and the abnormal frequency; the feature abnormality degree is analyzed according to the abnormal coupling feature set, and the feature abnormality degree is output.

[0059] Among them, a feature abnormality degree analysis is performed according to the abnormal coupling feature set, and the feature abnormality degree is output, and the following processing is performed: a feature benchmark indicator of the abnormal type is obtained, wherein the feature benchmark indicator includes a benchmark strength and a benchmark frequency; according to the benchmark strength and the benchmark frequency, a deviation analysis is performed on the abnormal strength and abnormal frequency of multiple abnormal types in the abnormal coupling feature set, and multiple abnormal deviation coefficients are obtained by weighting, wherein the abnormal deviation coefficient is positively correlated with the abnormal strength and the abnormal frequency; multiple abnormal influence degrees are configured according to the multiple abnormal types, the multiple abnormal deviation coefficients are weightedly fused, and the feature abnormality degree is output.

[0060] Among them, the health assessment mechanism is called according to the feature abnormality matching, and the health assessment mechanism is used to predict the shutdown failure probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and the fault prediction probability is output, and the following processing is performed: the ratio of the feature abnormality to the historical maximum feature abnormality is calculated and set as the abnormal scale coefficient; if the abnormal scale coefficient is less than 0.6, the first health assessment channel is called to predict the shutdown failure probability, and the fault prediction probability is output; if the abnormal scale coefficient is greater than or equal to 0.6, the first health assessment channel and the second health assessment mechanism are called to predict the shutdown failure probability, and the fault prediction probability is output.

[0061] Among them, the first health evaluation channel is obtained, and the following processing is performed: taking the vehicle attribute characteristics of the target tricycle as a constraint, calling the operation log of the same tricycle in the cloud platform, collecting the sample vehicle condition feature set and multiple sample abnormal coupling feature sets, and counting the historical vehicle downtime proportions of different sample vehicle condition features and different sample abnormal coupling feature sets in the historical time window, setting them as the sample downtime failure probability, and obtaining the sample downtime failure probability set; using the sample vehicle condition feature set, multiple sample abnormal coupling feature sets and sample downtime failure probability set to supervise the BP neural network training until convergence, and obtaining the first health evaluation channel.

[0062] Among them, if the abnormal scale coefficient is greater than or equal to 0.6, call the first health assessment channel and the second health assessment mechanism to predict the shutdown failure probability, output the fault prediction probability, and perform the following processing: If the abnormal scale coefficient is greater than or equal to 0.6, call the first health assessment channel, and predict the shutdown failure probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and output the first fault prediction probability; call the second health assessment mechanism to perform big data retrieval, and output the second fault prediction probability, wherein the big data retrieval includes the sample matching data volume; determine the big data retrieval accuracy based on the sample matching data volume analysis, and configure the trusted weight according to the big data retrieval accuracy and the prediction accuracy of the first health assessment channel; according to the trusted weight, perform weighted calculation on the first fault prediction probability and the second fault prediction probability, and output the fault prediction probability.

[0063] Among them, the second health evaluation mechanism is called to perform big data retrieval, output the second fault prediction probability, and perform the following processing: take the vehicle attribute characteristics as the equipment retrieval constraints, take the real-time vehicle condition characteristics and the abnormal coupling feature set as the feature retrieval constraints, take the electric tricycle parking failure as the guide, perform positive sample retrieval based on big data technology, and obtain the retrieval data set that meets the predetermined similarity threshold; count the proportion of shutdown failures in the retrieval data set and set it as the second fault prediction probability.

[0064] The electric tricycle fault detection system provided in the embodiment of the present invention can execute the electric tricycle fault detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0065] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0066] 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 fault detection system for an electric tricycle, characterized in that: The fault detection system of the electric tricycle comprises: The electrical monitoring data acquisition module is used to collect the monitoring data of the electrical layer of the target tricycle in real time and transmit it back to the cloud platform. If there is an abnormality in the electrical monitoring data, the sensor array of the mechanical layer is activated; A data coupling verification module is used to obtain mechanical monitoring data through the sensor array monitoring and transmit it back to the cloud platform, perform time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, and determine an abnormal coupling feature set and a feature abnormality degree; A fault probability prediction module is used to call a health evaluation mechanism according to the feature abnormality matching, and use the health evaluation mechanism to predict the shutdown fault probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and output the fault prediction probability; The fault warning module is used to generate a fault warning signal if the fault prediction probability exceeds a predetermined risk threshold, and transmit it back to the client of the target tricycle for early warning prompt.

2. The fault detection system for an electric tricycle according to claim 1, characterized in that: The data coupling verification module is also used for: The electrical monitoring data includes at least battery data, motor data and electrical control data, and the mechanical monitoring data includes at least part vibration data and driving status data; Acquire electrical monitoring data sets and mechanical monitoring data sets at multiple monitoring timestamps, and perform time synchronization alignment on the electrical monitoring data sets and the mechanical monitoring data sets to obtain multiple groups of monitoring data sets; Based on the long short-term memory network, data coupling verification is performed on the multiple monitoring data sets, and an abnormal coupling feature set is output, wherein the abnormal coupling feature includes an abnormal type, and the abnormal type is identified by abnormal intensity and abnormal frequency; A feature abnormality degree analysis is performed according to the abnormal coupling feature set, and the feature abnormality degree is output.

3. The fault detection system for an electric tricycle according to claim 2, characterized in that: The data coupling verification module is also used for: Acquire characteristic benchmark indicators of the abnormal type, wherein the characteristic benchmark indicators include benchmark intensity and benchmark frequency; According to the reference strength and reference frequency, deviation analysis is performed on the abnormal strength and abnormal frequency of multiple abnormal types in the abnormal coupling feature set, and multiple abnormal deviation coefficients are obtained by weighting, wherein the abnormal deviation coefficient is positively correlated with the abnormal strength and abnormal frequency; A plurality of abnormal influence degrees are configured according to the plurality of abnormal types, the plurality of abnormal deviation coefficients are weightedly fused, and the characteristic abnormality degree is output.

4. The fault detection system for an electric tricycle according to claim 1, characterized in that: The fault probability prediction module is also used for: Calculate the ratio of the characteristic abnormality to the historical maximum characteristic abnormality, and set it as the abnormality scale coefficient; If the abnormal scale coefficient is less than 0.6, calling the first health evaluation channel to predict the shutdown failure probability and outputting the failure prediction probability; If the abnormal scale coefficient is greater than or equal to 0.6, the first health assessment channel and the second health assessment mechanism are called to predict the shutdown failure probability, and the failure prediction probability is output.

5. The fault detection system for an electric tricycle according to claim 4, characterized in that: The fault probability prediction module is also used for: Taking the vehicle attribute characteristics of the target tricycle as a constraint, the operation logs of similar tricycles are called in the cloud platform to collect sample vehicle condition feature sets and multiple sample abnormal coupling feature sets. The historical vehicle downtime proportions of different sample vehicle condition features and different sample abnormal coupling feature sets in the historical time window are counted and set as the sample downtime failure probability to obtain the sample downtime failure probability set. The BP neural network is supervised and trained using a sample vehicle condition feature set, multiple sample abnormal coupling feature sets, and a sample shutdown failure probability set until convergence, thereby obtaining the first health evaluation channel.

6. The fault detection system for an electric tricycle according to claim 5, characterized in that: The fault probability prediction module is also used for: If the abnormal scale coefficient is greater than or equal to 0.6, the first health evaluation channel is called to predict the shutdown failure probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and the first fault prediction probability is output; Calling the second health assessment mechanism to perform big data retrieval and output a second fault prediction probability, wherein the big data retrieval includes sample matching data volume; Determining the big data retrieval accuracy based on the sample matching data volume analysis, and configuring the trust weight according to the big data retrieval accuracy and the prediction accuracy of the first health evaluation channel; The first fault prediction probability and the second fault prediction probability are weightedly calculated according to the credible weight, and the fault prediction probability is output.

7. The fault detection system for an electric tricycle according to claim 6, characterized in that: The fault probability prediction module is also used for: Taking the vehicle attribute features as equipment retrieval constraints, taking the real-time vehicle condition features and the abnormal coupling feature set as feature retrieval constraints, taking the electric tricycle parking fault as a guide, performing positive sample retrieval based on big data technology, and obtaining a retrieval data set that meets a predetermined similarity threshold; The proportion of downtime failures in the retrieved data set is counted and set as the second fault prediction probability.

8. A fault detection method for an electric tricycle, characterized in that: The method is implemented by the fault detection system of the electric tricycle according to any one of claims 1 to 7, and the method comprises: The monitoring data of the electrical layer of the target tricycle is collected in real time and transmitted back to the cloud platform. If there is any abnormality in the electrical monitoring data, the sensor array of the mechanical layer is activated; Acquire mechanical monitoring data through the sensor array monitoring and transmit it back to the cloud platform, perform time synchronization alignment and data coupling verification on the electrical monitoring data and the mechanical monitoring data, and determine an abnormal coupling feature set and a feature abnormality degree; Calling a health evaluation mechanism according to the feature abnormality matching, and using the health evaluation mechanism to predict the shutdown failure probability according to the real-time vehicle condition characteristics and the abnormal coupling feature set, and outputting the failure prediction probability; If the fault prediction probability exceeds a predetermined risk threshold, a fault warning signal is generated and transmitted back to the client of the target tricycle for early warning.

Images