Predictive maintenance method and device for brake-by-wire system, equipment and medium

Abstract

The invention discloses a predictive maintenance method and device for a brake-by-wire system, equipment and a medium, and relates to the technical field of intelligent networking, and the method comprises the steps: obtaining the operation data of the brake-by-wire system from a plurality of vehicles in a motorcade; based on a preset diagnosis rule, performing anomaly detection on the operation data of the brake-by-wire system of each vehicle in the motorcade, and determining an abnormal data fragment of each vehicle; the integrated abnormal data fragments are analyzed based on an unsupervised learning algorithm, a potential fault mode is identified, and the integrated abnormal data fragments comprise abnormal data fragments of different vehicles in the motorcade; aiming at the potential fault mode, adopting an interpretable machine learning method to generate a corresponding diagnosis rule; and sending the diagnosis rule to all the vehicles in the motorcade to update the preset diagnosis rule, thereby realizing predictive maintenance of the potential fault mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent network connection, and in particular to a predictive maintenance method and device for a brake-by-wire system, equipment and a medium. BACKGROUND

[0002] As a key safety component of intelligent network connected vehicles, the reliability of the brake-by-wire system is directly related to driving safety. At present, the maintenance of such systems mostly adopts diagnostic rules based on preset thresholds or fixed models trained on historical data, and performs abnormal monitoring and protection at the single vehicle level. Although this method can identify known types of faults, it is heavily dependent on prior knowledge and lacks the ability to discover new potential fault patterns that have not appeared in the training data. In addition, since the judgment is based only on single vehicle data, the system cannot use more extensive data samples to improve the comprehensiveness of fault recognition and the accuracy of early warning. Therefore, there is an urgent need for a predictive maintenance method for a brake-by-wire system to solve the problems mentioned above. SUMMARY

[0003] A series of simplified concepts are introduced in the summary section, which will be further described in detail in the specific embodiments section. The summary section of the present application does not mean to attempt to limit the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0004] The present application aims to solve the problems of existing brake-by-wire system maintenance strategies, such as relying on preset rules, being unable to discover unknown faults, and lacking vehicle fleet coordination. By obtaining multi-vehicle operation data from a vehicle fleet, automatically identifying potential fault patterns based on unsupervised learning, and generating diagnostic rules using interpretable machine learning, the vehicle fleet's diagnostic capabilities can be coordinated through dynamic distribution, thereby changing the maintenance mode from post-processing to pre-prediction, and improving system reliability and safety.

[0005] In a first aspect, the present application provides a predictive maintenance method for a brake-by-wire system, comprising: obtaining brake-by-wire system operation data from multiple vehicles in a vehicle fleet; performing abnormal detection on the brake-by-wire system operation data of each vehicle in the vehicle fleet based on a preset diagnostic rule, to determine an abnormal data segment of each vehicle; analyzing the integrated abnormal data segments based on an unsupervised learning algorithm to identify potential fault patterns, wherein the integrated abnormal data segments include the abnormal data segments of different vehicles in the vehicle fleet; generating corresponding diagnostic rules for the potential fault patterns using an interpretable machine learning method; sending the diagnostic rule to all vehicles in the vehicle fleet to update the preset diagnostic rule, and achieving predictive maintenance on the potential failure mode.

[0006] In some embodiments, the abnormality detection on the brake-by-wire system operation data of each vehicle in the vehicle fleet based on the preset diagnostic rule comprises: calculating the brake-by-wire system operation data of each vehicle in the vehicle fleet based on the preset diagnostic rule to determine a warning time point of each vehicle; For each vehicle, determining the abnormal data segment of the vehicle from the brake-by-wire system operation data based on the warning time point corresponding to the vehicle and a preset time window.

[0007] In some embodiments, the determination of the abnormal data segment of each vehicle from the brake-by-wire system operation data based on the warning time point corresponding to the vehicle and a preset time window comprises: For each vehicle, determining a data interception start time and a data interception end time based on the warning time point corresponding to the vehicle and the preset time window; and determining a target data sequence from the brake-by-wire system operation data based on the data interception start time and the data interception end time, and taking the target data sequence as the abnormal data segment of the vehicle.

[0008] In some embodiments, the analysis of the integrated abnormal data segments based on the unsupervised learning algorithm to identify the potential failure mode comprises: performing feature extraction on the integrated abnormal data segments to generate a feature vector set; performing clustering analysis on the feature vector set based on an unsupervised clustering algorithm to generate a plurality of data clusters; identifying the potential failure mode from the plurality of data clusters based on the plurality of data clusters and a preset failure mode feature library.

[0009] In some embodiments, the clustering analysis on the feature vector set based on the unsupervised clustering algorithm to generate a plurality of data clusters comprises: performing density clustering calculation on the feature vector set based on the unsupervised clustering algorithm to generate the plurality of data clusters, wherein the unsupervised clustering algorithm is determined based on a preset neighborhood radius parameter and a minimum sample number parameter of the density clustering algorithm.

[0010] In some embodiments, the preset fault mode feature library comprises a plurality of known fault modes and a feature representation corresponding to each known fault mode, and the identifying the potential fault mode from the plurality of data clusters based on the plurality of data clusters and the preset fault mode feature library comprises: calculating a center feature vector of each data cluster in the plurality of data clusters; calculating a feature distance between each center feature vector and a feature representation of each known fault mode in the preset fault mode feature library; determining an unmatched cluster from the plurality of data clusters based on the feature distance and a preset distance threshold, and identifying the unmatched cluster as the potential fault mode.

[0011] In some embodiments, the generating a corresponding diagnostic rule for the potential fault mode by using an interpretable machine learning method comprises: constructing a positive sample dataset based on all abnormal data segments corresponding to the potential fault mode; constructing a negative sample dataset based on normal operation data segments and other abnormal data segments not corresponding to the potential fault mode; training an interpretable classification model based on the positive sample dataset and the negative sample dataset; determining a decision path for distinguishing the positive sample dataset from the negative sample dataset from the trained interpretable classification model; determining a judgment condition sequence based on the decision path; generating the diagnostic rule based on the judgment condition sequence.

[0012] In a second aspect, the present application provides a predictive maintenance device for a brake-by-wire system, comprising: an operation data acquisition unit configured to acquire brake-by-wire system operation data from a plurality of vehicles in a vehicle fleet; an abnormal data determination unit configured to perform abnormality detection on the brake-by-wire system operation data of each vehicle in the vehicle fleet based on a preset diagnostic rule, and determine abnormal data segments of each vehicle; a potential fault analysis unit configured to analyze integrated abnormal data segments based on an unsupervised learning algorithm, and identify a potential fault mode, wherein the integrated abnormal data segments comprise the abnormal data segments of different vehicles in the vehicle fleet; a diagnostic rule generation unit configured to generate a corresponding diagnostic rule for the potential fault mode by using an interpretable machine learning method; a diagnostic rule updating unit configured to send the diagnostic rule to all vehicles in the vehicle fleet to update the preset diagnostic rule, and achieve predictive maintenance for the potential fault mode.

[0013] In a third aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the predictive maintenance method of the brake-by-wire system according to any one of the first aspect when executing the computer program.

[0014] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, wherein the computer program is executable on a processor to implement the steps of the predictive maintenance method of the brake-by-wire system according to any one of the first aspect.

[0015] In summary, the present application provides a predictive maintenance method of a brake-by-wire system, which effectively collects scattered fault symptom information by obtaining running data of multiple vehicles in a vehicle fleet and performing preliminary anomaly detection based on preset rules, and then analyzes the integrated abnormal data using unsupervised learning to automatically identify potential fault patterns that have not been recognized from the collective running experience of the vehicle fleet, thereby overcoming the limitations of single-vehicle intelligent systems in discovering unknown faults. Subsequently, the discovered fault patterns are converted into executable diagnostic rules using an interpretable machine learning method, and the rules are updated to all vehicles in the vehicle fleet, thereby synchronously updating the diagnostic capabilities of all vehicles. This process enables the vehicle fleet as a whole to continuously learn and evolve from running data, enabling early detection and warning of potential faults, and changing the maintenance mode from passive post-treatment to proactive prediction, thereby improving the reliability and safety of the brake-by-wire system. BRIEF DESCRIPTION OF DRAWINGS

[0016] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments, and are not intended to limit the present application. Moreover, like reference numerals are intended to represent like parts throughout the various drawings. In the drawings: Figure 1 Flowchart of the predictive maintenance method of the brake-by-wire system according to an embodiment of the present application; Figure 2 Structure diagram of the predictive maintenance device of the brake-by-wire system according to an embodiment of the present application; Figure 3 Structure diagram of the predictive maintenance device of the brake-by-wire system according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0018] Please see Figure 1 This is a schematic flowchart of a predictive maintenance method for a brake-by-wire system provided in an embodiment of this application, which may specifically include: S110. Obtain brake-by-wire system operation data from multiple vehicles in the fleet; For example, in step S110, runtime sequence data is continuously acquired from the brake-by-wire systems of multiple vehicles in the fleet via the onboard data acquisition unit. This data covers key parameters such as brake pedal opening, motor current and voltage, brake clamping force, and system temperature, and is acquired via the vehicle bus and temporarily stored on the vehicle side. This step aims to provide a raw data foundation covering various operating conditions and potential abnormal states for subsequent analysis, and its core lies in aggregating the operating information distributed across each vehicle in the fleet.

[0019] S120. Based on preset diagnostic rules, perform anomaly detection on the brake-by-wire system operating data of each vehicle in the fleet and identify the abnormal data segments of each vehicle. For example, in step S120, the vehicle-side performs online calculations and monitoring on the real-time collected operational data based on the currently loaded preset diagnostic rule set. When the monitored data meets the warning conditions in the preset diagnostic rules, a warning time point is determined, and the corresponding data sequence is automatically extracted based on this time point and a preset time window. This data sequence is the abnormal data segment, which contains the complete temporal changes of key parameters before and after the anomaly occurs, providing structured data for subsequent cloud analysis.

[0020] S130. Analyze the integrated abnormal data segments based on the unsupervised learning algorithm to identify potential fault modes. The integrated abnormal data segments include abnormal data segments from different vehicles in the fleet. Exemplarily, in step S130, the cloud integrates the abnormal data pieces from different vehicles in the vehicle fleet to form an integrated dataset covering multiple operation scenarios. By performing feature extraction and unsupervised clustering analysis on the dataset, the algorithm can automatically discover the underlying, pre-defined data distribution patterns. This process does not rely on pre-labeled fault samples, but identifies data communities with similar features through algorithm, thereby identifying potential unknown fault patterns and providing target objects for subsequent generation of diagnostic knowledge.

[0021] S140, for the potential fault pattern, an interpretable machine learning method is used to generate corresponding diagnostic rules; Exemplarily, in step S140, for the identified potential fault pattern, an interpretable machine learning method is used to automatically generate corresponding diagnostic rules. This method first constructs a positive sample set based on all abnormal data corresponding to the fault pattern, and constructs a negative sample set combining normal operation data and abnormal data of other irrelevant faults, to train an interpretable classification model. Then the key decision path is extracted from the trained model and converted into a sequence composed of judgment conditions, finally forming a structured diagnostic rule that can be loaded and executed by the vehicle end, thereby completing the conversion from fault pattern to specific diagnostic knowledge.

[0022] S150, the diagnostic rules are sent to all vehicles in the vehicle fleet to update the preset diagnostic rules, realizing predictive maintenance of the potential fault pattern.

[0023] Exemplarily, in step S150, the newly generated diagnostic rules are safely distributed to all vehicles in the vehicle fleet through over-the-air technology, and the vehicle end system integrates them into the local diagnostic rule library to complete dynamic updating. This process ensures the consistency and timeliness of the diagnostic capability within the vehicle fleet, so that any new potential fault pattern identified by a vehicle can be quickly converted into diagnostic knowledge shared by all vehicles. The vehicle can use the latest diagnostic rules to monitor the brake-by-wire system in real time, thereby issuing an early warning at the early stage of potential failure and improving the efficiency of predictive maintenance of the entire vehicle fleet.

[0024] In summary, the embodiments of the present application effectively collect scattered fault symptom information by acquiring the running data of multiple vehicles in a vehicle fleet and performing preliminary anomaly detection based on preset rules. Then, the integrated abnormal data is analyzed using unsupervised learning, which can automatically identify potential fault patterns that have not been recognized from the collective running experience of the vehicle fleet, overcoming the limitations of single-vehicle intelligent systems in discovering unknown faults. Subsequently, the discovered fault patterns are converted into executable diagnostic rules using interpretable machine learning methods, and the rules are updated to all vehicles in the vehicle fleet, thereby synchronously updating the diagnostic capabilities of all vehicles. This process enables the vehicle fleet as a whole to continuously learn and evolve from running data, achieving early detection and warning of potential faults, and transforming the maintenance mode from passive post-processing to proactive prediction, thereby improving the reliability and safety of the brake-by-wire system.

[0025] In some examples, based on the preset diagnostic rules, the running data of the brake-by-wire system of each vehicle in the vehicle fleet is subjected to anomaly detection, and abnormal data segments of each vehicle are determined, including: Based on the preset diagnostic rules, the running data of the brake-by-wire system of each vehicle in the vehicle fleet is calculated to determine the warning time point of each vehicle. For each vehicle, based on the warning time point corresponding to the vehicle and the preset time window, the abnormal data segment of the vehicle is determined from the running data of the brake-by-wire system.

[0026] For example, in the process of determining the abnormal data segment of each vehicle, the running data of the brake-by-wire system of each vehicle needs to be calculated based on the preset diagnostic rules first. The preset diagnostic rules are a set of logical judgments defined in advance and stored at the vehicle end, which include but are not limited to rule sets in the form of IF-THEN or micro machine learning models. These rules are designed based on the domain knowledge of the brake-by-wire system and are used to identify known abnormal conditions, such as motor current over-limit, response delay anomaly, or temperature over-limit, etc. The system matches and calculates the real-time collected running data, such as brake pedal opening, motor current, system temperature, etc. with the conditions in the preset diagnostic rules. When the running data meets all the logical conditions of a specific rule, the system determines that a warning is triggered, and records the time as the warning time point. This warning time point marks the identification of an abnormal event.

[0027] After determining the early warning time point, the system extracts the abnormal data segment for each vehicle. This process is based on the early warning time point corresponding to the vehicle and a preset time window. The preset time window is a pre-configured time interval, which is usually centered on the early warning time point and extends forward and backward, for example, covering a complete period of several seconds before and several seconds after the early warning time. The system calculates the specific data interception start time and data interception end time according to the boundaries of the early warning time point and the preset time window. Then, the system intercepts all time series data from the start time to the end time from the continuous running data stream of the vehicle to form a continuous data sequence, and the target data sequence is the abnormal data segment of the vehicle for this abnormal event. This segment completely retains the dynamic change process of key operating parameters before and after the abnormal event.

[0028] In summary, the embodiments of the present application can locate and extract the data segment strongly related to the abnormal event from the massive vehicle running data, effectively filter the redundant data under normal working conditions, improve the data quality of subsequent uploading and analysis, and provide a data basis for deep pattern discovery and knowledge mining in the cloud.

[0029] In some examples, for each vehicle, based on the early warning time point corresponding to the vehicle and the preset time window, the abnormal data segment of the vehicle is determined from the brake-by-wire system running data, including: For each vehicle, based on the early warning time point corresponding to the vehicle and the preset time window, the data interception start time and the data interception end time are determined. Based on the data interception start time and the data interception end time, the target data sequence is determined from the brake-by-wire system running data, and the target data sequence is taken as the abnormal data segment of the vehicle.

[0030] For example, in identifying anomalous data segments, a standardized data extraction process is executed for each vehicle that triggered a warning. This process begins with a predetermined warning time and a predefined time window, typically set as a fixed duration interval extending forward and backward from the warning time, such as covering a complete period from a specific number of seconds before to a specific number of seconds after the warning. Based on the configuration parameters of the warning time and the preset time window, calculations are performed to determine two time boundaries: a data extraction start time and a data extraction end time. Specifically, the data extraction start time is determined by shifting the warning time forward by the initial offset of the time window, while the data extraction end time is determined by shifting the warning time backward by the final offset of the time window. After defining the time boundaries, all time-series data points between the data extraction start time and the data extraction end time are located and extracted from the continuously collected brake-by-wire system operation data of the vehicle. These data points together constitute a continuous and complete target data sequence. Finally, the extracted target data sequence is labeled and output as the anomalous data segment corresponding to this warning event for the vehicle.

[0031] In summary, the embodiments of this application, through the standardized data extraction process described above, can ensure the location and extraction of the most relevant data segments from massive amounts of vehicle operation data. These segments fully preserve the dynamic evolution of key operating parameters before and after the anomaly. This method effectively filters out a large amount of redundant data generated under normal operating conditions, improves the efficiency of subsequent data transmission and storage, and provides structured data input for in-depth fault mode mining and diagnostic knowledge generation in the cloud, laying a data foundation for the reliable implementation of fleet-level predictive maintenance.

[0032] In some instances, unsupervised learning algorithms are used to analyze integrated anomalous data fragments to identify potential failure modes, including: Feature extraction is performed on the integrated abnormal data fragments to generate a feature vector set; Cluster analysis of feature vector sets is performed based on unsupervised clustering algorithms to generate multiple data clusters; Based on multiple data clusters and a preset fault mode feature library, potential fault modes are identified from multiple data clusters.

[0033] Exemplarily, in the process of identifying potential failure modes, the integrated abnormal data segments are first subjected to feature extraction to generate a set of feature vectors. The integrated abnormal data segments aggregate data uploaded from different vehicles in the vehicle fleet at the time of triggering a warning, containing multi-dimensional runtime information of the braking system in abnormal states. The feature extraction operation calculates, for each abnormal data segment, statistical and dynamic features of its internal key parameters, including but not limited to the mean and variance of motor current fluctuations, system response delay, change gradient of brake clamping force, and temperature rise trend, etc. Through this operation, each originally time-series-form abnormal data segment is converted into a numerical vector containing multiple feature values, i.e., a feature vector. The feature vectors corresponding to all abnormal data segments collectively constitute a set of feature vectors, which characterizes the failure characteristics implied in the original data in a structured mathematical form.

[0034] After generating the set of feature vectors, unsupervised clustering algorithm is used to perform clustering analysis on the set of feature vectors to generate a plurality of data clusters. The unsupervised clustering algorithm automatically explores the inherent data distribution structure in the set of feature vectors without relying on prior sample labels. The algorithm merges vectors in the feature space that are adjacent in position and similar in features into the same group, i.e., a data cluster, by calculating the similarity or distance between the feature vectors. Each data cluster represents a class of abnormal data with common feature patterns, corresponding to a specific system behavior state or failure type. Through this clustering analysis, the massive and mixed abnormal data is divided into several data groups, thereby revealing different failure modes that may exist in the data.

[0035] On the basis of generating a plurality of data clusters, the plurality of data clusters are compared with a preset failure mode feature library to identify potential failure modes therefrom. The preset failure mode feature library is a pre-constructed knowledge base that stores a plurality of known failure modes and a standard feature representation corresponding to each known failure mode. The identification process includes calculating the center feature vector of each data cluster and calculating the feature distance between the center feature vector and the feature representation of each known failure mode in the feature library. By comparing the feature distance with a preset distance threshold, it can be determined whether the data cluster is similar to a certain known failure mode. If the feature distance of a data cluster to all known failure modes is greater than the preset distance threshold, the data cluster is determined to be an unmatched cluster, i.e., a new, unrecognized potential failure mode.

[0036] In summary, the embodiments of the present application convert heterogeneous time series data into analyzable feature vectors through standardized feature extraction, automatically discover the internal structure of the data using unsupervised clustering, and finally identify new potential failure modes through intelligent comparison with known knowledge bases. This method realizes the transformation from raw abnormal data to new failure knowledge discovery, overcomes the limitations of relying on pre-set failure types, enables the predictive maintenance system to actively adapt and discover unknown risks, and improves the overall cognitive ability and forward-looking maintenance level of the system.

[0037] In some examples, the feature vector set is subjected to clustering analysis based on an unsupervised clustering algorithm, to generate a plurality of data clusters, including: The feature vector set is subjected to density clustering calculation based on an unsupervised clustering algorithm, to generate a plurality of data clusters, wherein the unsupervised clustering algorithm is determined based on a preset neighborhood radius parameter and a minimum sample number parameter of the density clustering algorithm.

[0038] For example, in the process of clustering analysis of the feature vector set based on the unsupervised clustering algorithm to generate a plurality of data clusters, a density-based clustering algorithm is specifically used for density clustering calculation. This process first initializes the configuration of the feature vector set according to the preset neighborhood radius parameter and the minimum sample number parameter. The neighborhood radius parameter defines the neighborhood range of individual sample points in the feature space, and the minimum sample number parameter sets the minimum number of samples that must be included in an effective cluster. The algorithm then iterates through each sample point in the feature vector set, identifies the density-connected sample group by calculating the spatial distance between the sample point and other sample points in the neighborhood; if the number of samples contained in the neighborhood of a sample point reaches or exceeds the minimum sample number parameter, the point is marked as a core point. Based on the density reachability of the core point, the algorithm merges the mutually connected core points and the boundary points in their neighborhoods into the same data cluster, thereby forming a plurality of data groups that have high local density in the feature space and are obviously separated from other regions. For discrete sample points that cannot be grouped into any cluster, they are identified as noise data for separate processing. Through the above density clustering calculation, the abnormal patterns with similar statistical characteristics implied in the feature vector set are identified and divided into different data clusters, and each data cluster represents a potential failure type with common characteristic performance.

[0039] In summary, the embodiments of the present application can automatically form groups according to the spatial distribution characteristics of the data itself by using the density clustering-based algorithm to analyze the feature vector set, without pre-specifying the number of clusters, effectively avoiding clustering bias caused by improper pre-set number of categories. This method has good adaptability to any shape of distribution pattern in the feature space, and can accurately identify the fault data community with high local density but irregular global distribution, thereby ensuring comprehensive coverage and effective discovery of various potential fault patterns. The density clustering algorithm is robust to noise data and can maintain the stability of the clustering results even in the presence of data outliers, improving the accuracy and reliability of fault pattern discovery and providing a high-quality data basis for the generation of subsequent diagnostic rules.

[0040] In some examples, the pre-set fault pattern feature library includes a plurality of known fault patterns and a feature representation corresponding to each known fault pattern, and the potential fault pattern is identified from the plurality of data clusters based on the plurality of data clusters and the pre-set fault pattern feature library, including: calculating a center feature vector of each data cluster in the plurality of data clusters; calculating a feature distance between each center feature vector and a feature representation of each known fault pattern in the pre-set fault pattern feature library; determining an unmatched cluster from the plurality of data clusters based on the feature distance and a pre-set distance threshold, and identifying the unmatched cluster as a potential fault pattern.

[0041] For example, in the process of identifying the potential fault pattern, a center feature vector of each data cluster in the plurality of data clusters needs to be calculated first. The center feature vector is obtained by calculating the mean value of each dimension feature value of all feature vectors in a data cluster, which represents the core position and typical feature of the data cluster in the feature space. Then, a feature distance between each center feature vector and a feature representation of each known fault pattern in the pre-set fault pattern feature library is calculated. The pre-set fault pattern feature library is a pre-constructed knowledge base that stores a plurality of known fault patterns and a standard feature representation corresponding to each known fault pattern; the feature distance is calculated using methods such as Euclidean distance or cosine distance to quantify the similarity of the current data cluster and the known fault pattern in the feature space. Finally, an unmatched cluster is determined from the plurality of data clusters based on the comparison result of the feature distance and the pre-set distance threshold. Specifically, if the minimum feature distance between the center feature vector of a data cluster and the feature representation of all known fault patterns in the feature library is still greater than the pre-set distance threshold, it is determined that the data cluster cannot be matched with any known pattern, and it is identified as a new potential fault pattern.

[0042] In summary, the embodiments of the present application can accurately distinguish newly discovered data patterns from existing fault patterns. The process avoids subjective judgment bias by comparing the quantified feature distance calculation with the preset threshold, ensuring the reliability of unknown fault pattern recognition. This method enables the predictive maintenance system to continuously learn from new operational data, enabling early detection and warning of potential risks.

[0043] In some examples, for a potential fault pattern, an interpretable machine learning method is used to generate a corresponding diagnostic rule, including: Based on all abnormal data segments corresponding to the potential fault pattern, a positive sample dataset is constructed; Based on normal operation data segments and other abnormal data segments corresponding to non-potential fault patterns, a negative sample dataset is constructed; Based on the positive sample dataset and the negative sample dataset, an interpretable classification model is trained; From the trained interpretable classification model, a decision path for distinguishing the positive sample dataset from the negative sample dataset is determined; Based on the decision path, a judgment condition sequence is determined; Based on the judgment condition sequence, a diagnostic rule is generated.

[0044] For example, in the process of generating a corresponding diagnostic rule for a potential fault pattern, a positive sample dataset is constructed based on all abnormal data segments corresponding to the potential fault pattern. This step aims to convert the new fault patterns discovered by unsupervised learning into labeled data required by supervised learning. The positive sample dataset contains all data instances identified as belonging to the potential fault pattern.

[0045] Based on normal operation data segments and other abnormal data segments corresponding to non-potential fault patterns, a negative sample dataset is constructed. Normal operation data segments represent the behavior patterns of the system in a healthy state, while other abnormal data segments corresponding to non-potential fault patterns cover data corresponding to other known fault types. The construction of the negative sample dataset ensures that the classification model not only learns the characteristics of the target fault, but also effectively distinguishes it from the normal state and other known faults.

[0046] After the construction of the positive and negative sample datasets, an interpretable classification model is trained based on the two datasets. The model usually selects machine learning algorithms with transparent decision-making processes such as decision trees. The training process enables the model to learn to distinguish between positive and negative samples based on input features, and its internal structure can clearly reflect the contribution of different features to the classification decision.

[0047] From the trained explainable classification model, a target decision path for distinguishing the positive sample dataset from the negative sample dataset is determined. The decision path refers to a sequence of nodes and branches from the root node of the model to a leaf node representing a positive class prediction.

[0048] Based on the determined target decision path, a sequence of judgment conditions contained therein is extracted. The sequence is composed of feature judgment conditions corresponding to each node on the path in order, such as whether feature one is greater than threshold one, whether feature two is less than threshold two, and the like. These conditions directly correspond to specific logical judgments of the brake-by-wire system operating parameters, constituting the core content of the diagnostic rules.

[0049] Based on the extracted sequence of judgment conditions, a final diagnostic rule is generated. The rule is usually expressed as a logical sentence that if all conditions are sequentially satisfied, it is determined that the potential failure mode exists. This diagnostic rule is lightweight, readable, and can be directly loaded and executed by the vehicle-side rule engine, thereby converting cloud-side discovered knowledge into vehicle-side usable diagnostic capabilities.

[0050] The embodiments of the present application realize an automated process from the identification of potential failure modes to the generation of specific diagnostic rules through the above steps. This method can convert new fault knowledge discovered by unsupervised learning into deployable and interpretable diagnostic logic, reducing the dependence on manual definition of rules by domain experts, improving the learning of unknown faults, and providing technical support for the evolution of fleet predictive maintenance capabilities.

[0051] In some examples, the diagnostic rule is sent to all vehicles in the fleet to update the preset diagnostic rule, realizing predictive maintenance of the potential failure mode, including: Exemplarily, in the process of sending the diagnostic rules to all vehicles in the vehicle fleet to update the preset diagnostic rules and realize predictive maintenance of potential failure modes, a secure communication link with the vehicles in the vehicle fleet is established by an over-the-air technology which uses an encrypted transmission protocol to ensure the confidentiality of the diagnostic rules during the distribution process; the rule management and delivery module packages the newly generated diagnostic rules that have passed verification into a rule update package, which contains the diagnostic logic expressed in the form of IF-THEN, a rule version identifier, and applicable vehicle configuration information; during the delivery process, the target vehicle group is matched based on the vehicle model and system version number, and a gray release strategy is used to first push the rule update package to a selected small range of test vehicle set for operation verification, and after confirming that the false positive rate and false negative rate meet the preset performance indicators, the rule update package is distributed to all remaining vehicles in the vehicle fleet; after the communication unit of the vehicle end receives the rule update package, the local processing unit decrypts and analyzes it, and integrates the diagnostic rules obtained by analysis into the embedded rule engine, and the integration operation includes adding new rules to the existing rule library and performing conflict detection to eliminate logical contradictions; the rule engine immediately performs online calculation and monitoring on the real-time collected brake-by-wire system operation data based on the new rule set after the update is completed, and when the operation data meets the logical conditions defined by the new rules, the system triggers an early warning and performs a preset prompt or protection action; through the process, any new potential failure mode identified by a single vehicle can be quickly converted into diagnostic knowledge shared by the entire vehicle fleet, thereby realizing early detection and early warning of potential failures.

[0052] In summary, the embodiments of the present application ensure the consistency of diagnostic capabilities within the vehicle fleet through an automated rule distribution and update mechanism. Once a diagnostic rule for a new potential failure is discovered and generated in the cloud, it can be quickly covered to the entire vehicle fleet, so that all vehicles can simultaneously have the ability to identify the failure. This process reduces the dependence on human intervention, improves the efficiency of maintenance response, and at the same time, through the immediate application of the updated rules at the vehicle end, early detection and early warning of potential failure risks are realized, the reliability of the brake-by-wire system operation is enhanced, and the vehicle fleet maintenance mode is promoted from passive response to active prediction.

[0053] The technical solutions of the present application will be further described in detail through specific embodiments.

[0054] In a specific embodiment, the pattern discovery process begins with cloud-side integration of a large number of abnormal data segments from the vehicle fleet. By performing feature extraction on these segments, such as calculating key indicators like motor current mean, fluctuation variance, and system response delay, each data segment is transformed into a high-dimensional feature vector. To facilitate analysis, principal component analysis is used to reduce the dimensionality of the feature vectors, mapping them to a two-dimensional plane to form a visual data point distribution. Subsequently, the DBSCAN clustering algorithm is applied for pattern mining, with a neighborhood radius parameter of 0.5 and a minimum sample size parameter of 10. After running the algorithm, three significant clusters are formed: two clusters are identified as known fault patterns A (representing sensor failure) and B (representing power supply interference) through comparison, and the other cluster is sufficiently dense and distinct from existing clusters, thus being identified as a new potential fault pattern and assigned code #47.

[0055] In the rule generation phase, a machine learning training dataset is constructed for the newly discovered fault pattern #47. All abnormal data labeled with this pattern are used as positive samples, while normal operation state data and abnormal data from other unrelated fault patterns are selected as negative samples. Based on this dataset, a CART decision tree classification model is trained, and during the model learning process, a hierarchical judgment logic is formed: First, check if the motor current exceeds the 5.2A threshold, if not, it is classified as normal; if it meets the condition, further judge if the response delay is greater than 120ms, if it does not reach the threshold, it is still classified as normal; when both conditions are met, continue to verify if the brake temperature exceeds 80°C, if all three conditions are finally met, it is determined as a potential motor drive failure, the sample confidence of the leaf node reaches 92%. Based on this decision path, the system automatically extracts key judgment conditions and combines to generate standardized diagnostic rules, that is, "when motor current > 5.2A, response delay > 120ms, and brake temperature > 80°C are met simultaneously, trigger a potential motor drive failure warning (fault code #47, confidence 92%)". After the rule is verified for effectiveness, it can be deployed to all vehicles in the fleet through over-the-air technology, completing the collaborative upgrade of diagnostic capabilities.

[0056] Please refer to Figure 2 A predictive maintenance device structure diagram of a brake-by-wire system is provided for the embodiments of the present application, comprising: An operating data acquisition unit 21 is configured to acquire brake-by-wire system operating data from multiple vehicles in a vehicle fleet; An abnormal data determination unit 22 is configured to perform abnormality detection on the brake-by-wire system operating data of each vehicle in the vehicle fleet based on a preset diagnostic rule, and determine abnormal data segments of each vehicle; The potential fault analysis unit 23 is configured to analyze the integrated abnormal data segments based on an unsupervised learning algorithm to identify a potential fault mode, wherein the integrated abnormal data segments include abnormal data segments of different vehicles in the vehicle fleet. The diagnostic rule generation unit 24 is configured to generate a corresponding diagnostic rule for the potential fault mode by using an interpretable machine learning method. The diagnostic rule updating unit 25 is configured to send the diagnostic rule to all vehicles in the vehicle fleet to update a preset diagnostic rule and achieve predictive maintenance of the potential fault mode.

[0057] Please refer to Figure 3 The embodiments of the present application further provide an electronic device 300, which comprises a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and capable of running on the processor. When the processor 320 executes the computer program 311, the steps of the predictive maintenance method of the brake-by-wire system are implemented.

[0058] Since the electronic device described in the embodiments is the device used to implement the predictive maintenance device of the brake-by-wire system in the embodiments of the present application, the specific implementation of the electronic device and its various forms can be understood by those skilled in the art based on the method described in the embodiments of the present application. Therefore, how the electronic device implements the method in the embodiments of the present application will not be described in detail, as long as the device used by those skilled in the art to implement the method in the embodiments of the present application belongs to the scope of the present application.

[0059] In the implementation process, the computer program 311 can implement any of the embodiments of the first aspect when executed by the processor.

[0060] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0061] Those skilled in the art should understand that the embodiments of the present application can provide a method, a system or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer readable storage media containing computer readable program code.

[0062] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0063] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0064] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0065] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0066] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired or wireless manner. The computer-readable storage medium can be any available medium that can be stored by the computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium, an optical medium or a semiconductor medium, etc.

[0067] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0068] In several embodiments provided in the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the apparatus embodiments described above are only schematic, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other form.

[0069] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0070] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be in the form of hardware and / or software functional units.

[0071] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, and various program code storage media.

[0072] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

[0073] Although the preferred embodiments of the present specification have been described, those skilled in the art can make additional changes and modifications to the embodiments once they know the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present specification.

[0074] Obviously, those skilled in the art can make various modifications to the present specification without departing from the spirit and scope of the present specification. Thus, if these modifications of the present specification fall within the scope of the claims of the present specification and their equivalents, the present specification also intends to include these modifications.

Claims

1. A predictive maintenance method for a brake-by-wire system, characterized in that, include: Acquire operational data from the brake-by-wire systems of multiple vehicles in the fleet; Based on preset diagnostic rules, anomaly detection is performed on the brake-by-wire system operating data of each vehicle in the fleet to identify abnormal data segments for each vehicle. The integrated abnormal data fragments are analyzed based on an unsupervised learning algorithm to identify potential fault modes. The integrated abnormal data fragments include abnormal data fragments from different vehicles in the fleet. For the aforementioned potential failure modes, interpretable machine learning methods are used to generate corresponding diagnostic rules; The diagnostic rules are sent to all vehicles in the fleet to update the preset diagnostic rules, enabling predictive maintenance of the potential failure modes.

2. The method according to claim 1, characterized in that, Based on preset diagnostic rules, anomaly detection is performed on the brake-by-wire system operating data of each vehicle in the fleet to identify abnormal data segments for each vehicle, including: Based on the preset diagnostic rules, the operating data of the brake-by-wire system of each vehicle in the fleet are calculated to determine the warning time point for each vehicle. For each vehicle, based on the warning time point and preset time window corresponding to that vehicle, the abnormal data segment of that vehicle is determined from the brake-by-wire system operation data.

3. The method according to claim 2, characterized in that, For each vehicle, based on the warning time point and preset time window corresponding to that vehicle, the abnormal data segment for that vehicle is determined from the brake-by-wire system operating data, including: For each vehicle, based on the warning time point and the preset time window corresponding to that vehicle, the start time and end time of data interception are determined; based on the start time and end time of data interception, a target data sequence is determined from the brake-by-wire system operation data, and the target data sequence is used as the abnormal data segment of that vehicle.

4. The method according to claim 1, characterized in that, The analysis of the integrated abnormal data fragments based on the unsupervised learning algorithm to identify potential fault modes includes: Feature extraction is performed on the integrated abnormal data fragments to generate a feature vector set; The feature vector set is clustered using an unsupervised clustering algorithm to generate multiple data clusters; Based on the multiple data clusters and the preset fault mode feature library, the potential fault modes are identified from the multiple data clusters.

5. The method according to claim 4, characterized in that, The unsupervised clustering algorithm is used to perform cluster analysis on the feature vector set to generate multiple data clusters, including: Based on the unsupervised clustering algorithm, density clustering calculation is performed on the feature vector set to generate the multiple data clusters, wherein the unsupervised clustering algorithm is determined based on the preset neighborhood radius parameter and minimum sample number parameter of the density clustering algorithm.

6. The method according to claim 4, characterized in that, The preset fault mode feature library includes multiple known fault modes and feature representations corresponding to each known fault mode. The step of identifying the potential fault modes from the multiple data clusters based on the multiple data clusters and the preset fault mode feature library includes: Calculate the central feature vector of each data cluster in the plurality of data clusters; Calculate the feature distance between each of the central feature vectors and the feature representations of each known fault mode in the preset fault mode feature library; Based on the feature distance and the preset distance threshold, unmatched clusters are determined from the multiple data clusters, and the unmatched clusters are identified as the potential fault modes.

7. The method according to claim 1, characterized in that, The step of generating corresponding diagnostic rules for the potential failure modes using interpretable machine learning methods includes: Based on all the abnormal data fragments corresponding to the potential failure modes, construct a positive sample dataset; A negative sample dataset is constructed based on normal operating data segments and other abnormal data segments that do not correspond to the potential failure modes. Based on the positive sample dataset and the negative sample dataset, an interpretable classification model is trained; From the trained interpretable classification model, determine the decision path used to distinguish between the positive sample dataset and the negative sample dataset; Based on the decision path, determine the sequence of judgment conditions; The diagnostic rules are generated based on the judgment condition sequence.

8. A predictive maintenance device for a brake-by-wire system, characterized in that, include: The operation data acquisition unit is used to acquire the operation data of the brake-by-wire system from multiple vehicles in the fleet; An abnormal data determination unit is used to perform abnormal detection on the brake-by-wire system operation data of each vehicle in the fleet based on preset diagnostic rules, and to determine the abnormal data segments of each vehicle. The potential fault analysis unit is used to analyze the integrated abnormal data fragments based on an unsupervised learning algorithm to identify potential fault modes, wherein the integrated abnormal data fragments include the abnormal data fragments of different vehicles in the fleet. The diagnostic rule generation unit is used to generate corresponding diagnostic rules for the potential fault modes using an interpretable machine learning method. The diagnostic rule update unit is used to send the diagnostic rules to all vehicles in the fleet to update the preset diagnostic rules and realize predictive maintenance of the potential fault modes.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program stored in the memory, implements the steps of the predictive maintenance method for a brake-by-wire system as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the predictive maintenance method for a brake-by-wire system as described in any one of claims 1 to 7.

Images