Water wiper motor health state early warning system based on big data

Abstract

The invention discloses a wiper motor health state early warning system based on big data, and particularly relates to the technical field of electrical equipment state monitoring and fault diagnosis, which comprises a data acquisition module, a state interface module, a data preprocessing module, a data analysis module, an early warning decision module and an early warning information output module, electrical parameters such as current, voltage, temperature and rotating speed are collected through a sensor, and a feature set is generated through preprocessing in combination with vehicle bus working condition data; comparing the real-time electrical characteristics with a pre-stored baseline model and a fault mode library, calculating a health score based on a mahalanobis distance algorithm, identifying a fault and predicting the life; and generating graded early warning according to a multi-level threshold strategy, and outputting the graded early warning through a vehicle-mounted interface and a remote monitoring center. According to the invention, early, accurate and predictive early warning of the health state of the motor is realized, and the accuracy of state evaluation and the reliability of the system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment state monitoring and fault diagnosis, more particularly, the present application relates to a wiper motor health state early warning system based on big data. BACKGROUND

[0002] At present, in the field of motor state monitoring and fault warning, especially for small driving devices such as wiper motors, the common prior art mainly relies on limited electrical measurement and simple threshold judgment. One way is to alarm based on the static threshold of single or few electrical parameters, which only monitors whether the motor operating current exceeds a certain fixed upper limit value to determine whether the locked rotor occurs, or integrates a thermal protection switch in the motor to cut off the circuit when the winding temperature is too high. Another is the planned maintenance following a fixed cycle, that is, the maintenance personnel performs routine inspection and replacement on the equipment during maintenance, without considering the actual health status of the motor.

[0003] However, in actual use, there are still some shortcomings, such as the wiper motor working environment is complex and changeable, and the load is greatly affected by factors such as rainfall, wind resistance, and wiper blade aging, so the electrical parameters show strong working condition dependence; the existing technology lacks decoupling analysis of the specific working conditions of the wiper motor, and cannot distinguish between normal load fluctuations and early fault characteristics, resulting in high false alarm rate and insensitive early degradation detection. In addition, the wiper motor is a safety component of the vehicle, and its failure directly affects the driving safety, so a more accurate and adaptive early warning system is needed. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a wiper motor health state early warning system based on big data, which solves the problems raised in the background art by the following scheme.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a wiper motor health state early warning system based on big data, comprising:

[0006] The data acquisition module is used for real-time acquisition of electrical operating parameters of the target motor, and the electrical operating parameters include motor operating current, motor driving voltage, motor winding temperature and motor speed;

[0007] The state interface module is used for obtaining working condition data related to the load of the target motor from the communication bus, and the working condition data includes wiper switch state signal and speed signal;

[0008] The working condition data is used to decouple the electrical parameter changes of the wiper motor under different working modes (such as low speed, high speed, and intermittent) and different vehicle speeds, and to provide context for accurate state evaluation;

[0009] A data preprocessing module is configured to preprocess the received electrical operation parameters and working condition data to generate a motor state feature data set;

[0010] The preprocessing includes extracting time domain features and working condition association features related to the wiper motor load characteristics, such as current increment per unit vehicle speed, temperature change rate at each gear, etc., which are specifically designed for the dynamic load and common failure modes of the wiper motor;

[0011] A data analysis module includes a storage unit and a diagnostic analysis unit;

[0012] The storage unit is configured to store an electrical performance baseline model constructed based on the electrical data of a group of wiper motors of the same type, and a failure mode library constructed based on historical failure data of the wiper motor;

[0013] The diagnostic analysis unit is configured to compare and analyze the received motor state feature data set with the electrical performance baseline model and the failure mode library in real time, calculate a real-time health status score of the target motor based on a machine learning algorithm, and predict the remaining service life based on the electrical parameter degradation trend;

[0014] A warning decision module is configured to compare the real-time health status score with a preset electrical performance threshold, and generate corresponding warning information when an electrical performance anomaly is determined;

[0015] The warning threshold is set in multiple levels based on typical degradation behavior and failure modes of the wiper motor, covering early degradation, intermediate maintenance and emergency failure;

[0016] A warning information output module is configured to send the warning information to a human-machine interface and a remote monitoring center.

[0017] Preferably, the electrical operation parameters include:

[0018] The motor operating current is collected by a Hall effect current sensor connected in series to the driving circuit of the wiper motor;

[0019] The motor driving voltage is collected by a high-impedance differential operational amplifier circuit connected in parallel to the motor driving power supply;

[0020] The motor winding temperature is collected by a digital temperature sensor attached to the motor winding shell or embedded in the motor;

[0021] The motor speed is collected by a Hall sensor or an optical encoder installed on the output shaft or transmission mechanism of the wiper motor.

[0022] Preferably, the working condition data comprises: acquiring the working condition data from a vehicle communication bus; and the working condition data comprises a wiper switch state signal and a vehicle speed signal.

[0023] Preferably, the preprocessing comprises:

[0024] The data cleaning and synchronization unit is configured to perform timestamp alignment, outlier elimination and digital filtering processing on the collected electrical operation parameters and working condition data.

[0025] The feature extraction unit is configured to extract time domain statistical features and working condition associated features from the preprocessed data to generate a motor state feature dataset.

[0026] The data normalization unit is configured to perform standardization processing on the feature dataset to adapt to subsequent analysis models.

[0027] Preferably, the motor state feature dataset comprises: time domain features extracted from electrical operation parameters within a fixed time window, the time domain features comprising mean value, standard deviation, peak value, current or voltage waveform factor, and temperature change rate.

[0028] Preferably, the electrical performance baseline model comprises: being generated based on historical data of a large number of healthy motors of the same type under multiple working conditions, and being generated by clustering and modeling through a Gaussian mixture model or a multivariate Gaussian distribution.

[0029] Preferably, the fault mode library comprises: being constructed based on electrical data of historical fault cases, each fault mode corresponding to one or more feature "fingerprints", and the library being a queryable database or lookup table, each record containing a fault code, a feature mode description, a disposal suggestion and a warning level.

[0030] Preferably, the warning decision module comprises:

[0031] The first-level warning is triggered when the real-time health status score is lower than a first threshold but higher than or equal to a second threshold.

[0032] The second-level warning is triggered when the real-time health status score is lower than the second threshold but higher than or equal to a third threshold.

[0033] The third-level warning is triggered when the real-time health status score is lower than the third threshold.

[0034] Preferably, the warning information comprises: warning trigger time, warning level, current health score, identified fault code, predicted remaining life, specific disposal suggestion text, and warning trigger source.

[0035] The technical effects and advantages of the present application are as follows:

[0036] 1. Early, accurate and predictive warning for wiper motor: By fusing multi-source electrical parameters and real-time operating data, and using machine learning algorithms based on Gaussian mixture model and Mahalanobis distance, the system can sensitively capture early signs of degradation, minor waveform distortion of wiper motor under complex operating conditions, and realize early detection of faults.

[0037] 2. Reduced false alarm rate of wiper motor health monitoring: The system obtains operating condition information such as wiper gear position and vehicle speed through a state interface, and performs operating condition correlation analysis during feature extraction, so that the state assessment can be decoupled from the specific working mode, avoiding false judgments caused by normal changes in wiper load, and maintaining high reliability in various scenarios such as rainy days and high-speed driving.

[0038] 3. Provides three-dimensional, multi-level state assessment and decision support: The system not only outputs a single health score, but also integrates wiper motor specific fault mode recognition and remaining useful life prediction; the warning decision module integrates information in three dimensions of real-time state, fault type and future trend through multi-level threshold and rule fusion strategy, generates graded and accurate warning information and disposal suggestions, provides immediate warning for the driver, and provides prospective planning basis for motor management and maintenance service.

[0039] 4. Complete data closed loop and system iteration capability: Through the remote monitoring center path, the system can continuously upload warning data and operating data to the cloud for optimizing wiper motor baseline model, enriching fault mode library, and providing big data support for motor design improvement. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a schematic diagram of the overall structure of the present application.

[0041] Figure 2 is a schematic diagram of the health score analysis structure of the present application.

[0042] Figure 3 is a schematic diagram of the data analysis module structure of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0044] Reference Figure 1The illustrated big data-based wiper motor health state early warning system comprises a data acquisition module, a state interface module, a data preprocessing module, a data analysis module, an early warning decision module and an early warning information output module; the data acquisition module and the state interface module are connected with the data preprocessing module respectively, the data preprocessing module is connected with the data analysis module, the data analysis module is connected with the early warning decision module, and the early warning decision module is connected with the early warning information output module.

[0045] The system is specially designed for wiper motors, and is deeply adapted in various modules in view of the characteristics of the wiper motors, such as complex working environment, large load fluctuation, influence of vehicle speed and wiper gear position, etc. The modules will be described in detail below in combination with the actual operation scene of the wiper motor.

[0046] The data acquisition module is used for collecting the electrical operation parameters of the target motor in real time, and the electrical operation parameters include motor working current, motor driving voltage, motor winding temperature and motor rotating speed.

[0047] It is necessary to specifically describe the sensor selection and signal processing process in this embodiment as follows:

[0048] The data acquisition module is composed of a series of high-precision sensors, signal conditioning circuits and analog-to-digital conversion units, and communicates with the system main controller through a bus.

[0049] In view of the characteristics of the wiper motor, such as frequent dynamic changes of the load, and wet and vibration working environment, the module selects sensors with high reliability, wide temperature range and good anti-interference ability.

[0050] The current acquisition unit adopts a Hall effect current sensor for non-contact measurement. The sensor is connected in series in the driving circuit of the wiper motor, and converts the working current of the motor into a proportional voltage signal. Its wideband characteristic (DC~100kHz) can accurately capture the current change of the wiper motor in the transient process such as starting and locked rotor.

[0051] The voltage acquisition unit adopts a high-impedance differential operational amplifier circuit directly connected in parallel to the two ends of the motor driving power supply to measure the driving voltage of the motor. The circuit design contains overvoltage protection and low-pass filtering to suppress the common surge voltage and high-frequency noise in the power supply system, and to ensure the stability of the voltage measurement. Especially for the possible power voltage drop caused by the start and stop of the wiper motor, the measurement is optimized.

[0052] The temperature acquisition unit adopts a negative temperature coefficient thermistor or a digital temperature sensor attached to the motor winding shell or embedded in the motor. In this embodiment, the digital temperature sensor is preferred, because it has a solid package, a simple interface, and is more suitable for the engine compartment high temperature and high humidity environment of the wiper motor, and can accurately monitor the early phenomenon of motor overheating caused by the increase of wiper resistance.

[0053] Rotation speed acquisition unit: Hall sensor or optical encoder. A magnetic target or grating disc is installed on the output shaft or transmission mechanism of the wiper motor. When the motor rotates, the Hall sensor senses the change in magnetic flux and outputs a pulse signal; the optical encoder generates a pulse through the on-off of the grating disc. By counting the number of pulses in a unit of time, the real-time rotation speed of the motor can be accurately calculated. This rotation speed signal is a key parameter for determining whether the mechanical transmission of the wiper is stuck or whether the wiper blade is aging, causing an increase in load.

[0054] Signal processing and data transmission:

[0055] Signal conditioning circuit: The raw signals (voltage or current signals) output by each sensor first pass through the signal conditioning circuit. This circuit usually includes:

[0056] Filtering circuit: RC low-pass filter is used to filter out electromagnetic interference and high-frequency noise from the environment.

[0057] Amplification circuit: Weak sensor signals are amplified to match the input range of the analog-to-digital converter.

[0058] Analog-to-digital conversion unit: The conditioned analog signals are sent to a high-precision analog-to-digital converter (ADC). The ADC discretizes the analog signals into digital signals at a fixed sampling frequency (1 kHz). This sampling frequency is much higher than the actual frequency of the motor electrical parameters, satisfying the Nyquist sampling theorem and allowing the signal characteristics to be restored without distortion.

[0059] Microcontroller and communication interface: The converted digital signals are processed by an embedded microcontroller. The controller synchronizes the data from different sensors with timestamps and packages them into a specific data frame format. Finally, the data is sent in real-time to the system's data preprocessing module through the Controller Area Network (CAN) bus or Local Interconnect Network (LIN) bus.

[0060] The electrical operating parameters specifically include:

[0061] Motor operating current: RMS or instantaneous value, used to analyze motor load and stall faults.

[0062] Motor drive voltage: RMS value, used to determine whether the power supply is stable.

[0063] Motor winding temperature: Real-time temperature value, used to monitor the risk of motor overheating.

[0064] Motor rotation speed: Real-time rotation speed value, used to determine whether the mechanical transmission is normal.

[0065] The sampling strategy adopts a combination of timing triggering and event triggering. When the wiper is working normally, fixed frequency (10 Hz) is used for sampling; when the current or speed changes dramatically, it automatically switches to high-speed sampling mode (100 Hz) to capture more detailed fault transient characteristics.

[0066] State interface module: for obtaining working condition data related to the target motor load from the communication bus, the working condition data including wiper switch state signal and speed signal;

[0067] The embodiment needs to be specifically described that the working condition data acquisition and analysis process is as follows:

[0068] In the wiper motor application scenario, the motor load is directly related to the wiper working gear (off, intermittent, low speed, high speed), and at the same time, the wind resistance has a great impact on the load of the wiper motor during high-speed driving. By accurately obtaining these working condition data, the system can decouple the changes of electrical parameters with specific working modes and environmental factors, which is the core of realizing accurate early warning and reducing false alarms.

[0069] The hardware core of the state interface module is a microcontroller (MCU) with a CAN / LIN controller, and is equipped with a corresponding physical layer transceiver chip.

[0070] CAN bus access: the module is connected to the CAN bus through the CAN transceiver. CAN bus is a neural network, carrying data shared between ECUs (Electronic Control Unit). The module accesses the CAN bus or power CAN bus through the physical wire harness.

[0071] Gateway function: the module acts as an independent ECU node with simple gateway function. It can listen to various data frames on the bus and filter out the information related to the system according to the predefined identifier.

[0072] The module obtains the working condition data dedicated to wiper motor health status analysis by listening to and analyzing specific data frames broadcast on the bus.

[0073] Wiper switch state signal:

[0074] Data source: this signal is usually published on the CAN bus by the control module (BCM) or the steering column control unit.

[0075] Data analysis: the module continuously listens to the CAN frame containing wiper control information. There are usually specific data bits (bit) in this data frame to represent the state of the wiper, for example:

[0076] 00: off

[0077] 01: intermittent work

[0078] 10: low speed

[0079] 11: high speed

[0080] The signal is pre-stored in the firmware program of the module with a parsing dictionary (DBC file mapping) that can real-time parse the original CAN data into switch state enumeration values with clear engineering significance. The state directly defines the basic load working condition of the wiper motor and is the basis for subsequent working condition segmentation statistics and feature extraction.

[0081] Speed signal:

[0082] Data source: The signal is usually provided by the anti-lock braking system (ABS) ECU or the combination instrument ECU on the CAN bus.

[0083] Data parsing: The module listens to the CAN frame containing speed information. The speed is usually in the form of a 16-bit unsigned integer in the data frame. The module obtains the real-time vehicle speed in km / h by parsing and multiplying the conversion coefficient. Vehicle speed is a key factor affecting the wind resistance load of the wiper motor. High current at high speed is easily misjudged as a fault without vehicle speed information. The system effectively avoids such false positives by introducing the vehicle speed signal.

[0084] Data synchronization and timestamp: After obtaining the switch state and speed signal, the module will stamp these two sets of data with high-precision timestamps synchronized with the data acquisition module.

[0085] Signal validity verification: The module performs cyclic redundancy check (CRC) and reasonableness check on the received bus data. If the data is found to be invalid or timeout, the module will output the last valid value or a specific "invalid" state flag to notify the subsequent module that the data is not reliable.

[0086] Data encapsulation and forwarding: The working condition data (switch state, speed) after verification and synchronization is encapsulated into a unified data structure and transmitted in real time to the data preprocessing module through the module's internal SPI or UART communication interface.

[0087] It needs to be further explained that the signal validity verification mechanism implemented by the module is crucial. It can effectively identify and handle error data caused by high bus load, node failure or electromagnetic interference, preventing invalid or erroneous working condition information from polluting the subsequent analysis process, thereby ensuring the robustness of the system decision. In addition, the high-precision timestamp synchronization technology ensures the strict consistency of electrical parameters and working condition data in time sequence, which is the premise for accurate working condition segmentation and correlation analysis, avoiding misjudgment due to asynchronous data.

[0088] The data preprocessing module is used for preprocessing the received electrical operation parameters and the working condition data to generate motor state feature data sets.

[0089] It is necessary for the embodiment to specify that the preprocessing procedure and the feature engineering process are as follows:

[0090] The data preprocessing module runs on the main control microprocessor of the system. The processor receives the electrical parameter data packet from the data acquisition module through SPI, UART or internal parallel bus, and receives the working condition data from the state interface module through the CAN controller or internal memory sharing mode. Adequate processing capacity and memory resources ensure the real-time execution of complex preprocessing algorithms.

[0091] The preprocessing procedure mainly includes three core steps: data cleaning and synchronization, feature engineering for wiper motor, and data normalization and formatting.

[0092] The feature engineering is specially designed for the dynamic load characteristics and common failure modes of the wiper motor, aiming to extract effective features sensitive to early degradation and decoupled from working conditions from the original data. For example, the "current increment per unit speed" feature is extracted for the feature that the wind resistance load of the wiper motor increases at high speed; the "temperature change rate" and "current waveform factor" features are extracted for the slow increase in load caused by blade aging. These features can effectively distinguish between normal working condition fluctuations and abnormal degradation.

[0093] It is necessary for the embodiment to specify that the data cleaning and synchronization process is as follows:

[0094] Data cleaning and synchronization:

[0095] Timestamp alignment: the module first gives all input data (current, voltage, temperature, speed, switch state, vehicle speed) a microsecond-precision timestamp based on the system global clock. Then, a unified time sequence (100 milliseconds) is used as a reference to linearly interpolate or nearest neighbor interpolation all data, resample them to the same time point, and form a strictly synchronized multi-dimensional time series.

[0096] Outlier processing:

[0097] 1. Transient pulse rejection: sliding window method based on threshold. For example, calculate the mean and standard deviation of the current data in a window (such as 1 second), any data point outside the range of mean ± 3 times standard deviation will be considered as transient pulse noise, and the window median or linear interpolation result will be used to replace it;

[0098] 2. Sensor failure determination: For data continuously beyond the physically reasonable range (e.g. temperature -40℃ or 200℃), or data without updates for a long time, the module will mark it as "sensor failure" and trigger the system degradation processing strategy.

[0099] Signal filtering: Digital filtering of the original signal with large fluctuations to suppress high-frequency noise.

[0100] The module not only outputs the instantaneous value of the original parameter, but more importantly, calculates a series of statistical and dynamic characteristics with physical meaning and diagnostic value within a fixed time window (10 seconds), thereby generating a motor state feature dataset.

[0101] The feature extraction process of the present embodiment is as follows:

[0102] Time domain feature extraction:

[0103] Basic statistics: Calculate the mean (representing the average level), standard deviation (representing the fluctuation degree), and peak value (representing the extreme case) of each electrical parameter in the window.

[0104] Advanced features:

[0105] 1. Current / voltage waveform factor: Calculate the ratio of the root mean square value (RMS) to the absolute average value, which is used to reflect the distortion degree of the waveform, and is particularly sensitive to carbon brush wear and winding short circuit faults of the wiper motor:

[0106] Carbon brush wear sensitivity: When the carbon brush wears out, it will cause the contact resistance of the motor commutator to increase, causing the current waveform to appear intermittent and burr, thereby increasing the waveform factor. Specifically, under healthy conditions, the waveform factor is usually stable in the range of 1.0-1.1; when the carbon brush wear reaches the middle stage, the waveform factor may continuously increase to above 1.2, accompanied by an increase in current fluctuation standard deviation.

[0107] Winding short circuit sensitivity: When the winding has inter-turn or inter-layer short circuit, the change of motor impedance will cause the current waveform to distort, the harmonic component to increase, and the waveform factor to rise significantly. For example, in the early stage of winding short circuit, the waveform factor may rise from 1.1 to above 1.3, and cooperate with the temperature change rate feature to indicate the fault.

[0108] 2. Temperature change rate: Calculate the change of temperature per unit time ( ), which is used to monitor the overheating trend of the motor, and is particularly suitable for identifying continuous overload caused by wiper blade aging and transmission jamming.

[0109] Working condition related feature extraction:

[0110] Segmented statistics: According to the wiper switch state, the data window is divided into different working intervals (such as "off", "low speed", "high speed"). Then the electrical parameter characteristics in each working interval are calculated respectively. This step binds the wiper motor state evaluation with the specific working gear, establishing an analysis mode of "one working condition one baseline".

[0111] Load-related features and their effectiveness verification: To accurately decouple the influence of wind resistance on motor load, the system constructs "unit speed current increment" as the key discriminant feature, which is defined as Δ motor operating current / Δ vehicle speed. The effectiveness of this feature is established through the following multi-dimensional verification:

[0112] Theoretical self-consistency verification: Based on the principle of aerodynamics, the additional load of the wiper motor at high speed is mainly due to the wind resistance which is proportional to the square of the vehicle speed. The design of this feature directly corresponds to this physical law, and its dimension (current / speed) essentially quantifies the sensitivity of wind resistance load. Theoretical analysis shows that the increase of normal wind resistance is manifested as a smooth and slow change of the feature value; while mechanical stall, which is independent of vehicle speed, will cause the feature value to jump or diverge, ensuring the separability of the two from a mechanistic point of view.

[0113] Historical data backtracking verification: Statistical analysis of massive historical operation data provides empirical support. The data shows that in the confirmed normal high-speed working condition, the feature value is always stably distributed in a narrow low value interval (such as 0.02-0.05 A / (km / h)); while before all recorded stall events, the feature value jumps to an abnormally high value several times higher than the normal interval. The clustering separation degree of the two in the feature space is high, with no overlap, proving the working condition distinguishing ability.

[0114] Model discriminant efficiency verification: In subsequent machine learning models, through feature importance ranking analysis, this feature is consistently identified as the Top key factor to distinguish "high speed normal" and "stall fault". Its introduction directly improves the model classification accuracy of such working conditions by more than 15%, and significantly reduces more than 90% of false alarms caused by high-speed wind resistance, verifying its engineering practical value from the system output level.

[0115] The data normalization and formatting process in this embodiment needs to be specifically explained as follows:

[0116] Data normalization and formatting:

[0117] Normalization: All extracted features are normalized to eliminate the influence of different dimensions and orders of magnitude. This embodiment adopts Z-score standardization or Min-Max scaling to scale the features to the [0, 1] interval. The mean, standard deviation, maximum value, and minimum value parameters used for normalization are derived from the electrical performance baseline model, ensuring consistency in processing between the training (baseline) and inference (real-time) stages.

[0118] Formatting and caching: All cleaned, synchronized, extracted, and normalized features are combined into a fixed-dimension feature vector. This feature vector is packaged together with the corresponding timestamp and operating condition label into a structured data object, i.e., a "motor state feature dataset." This dataset is stored in a circular data buffer for real-time reading by the data analysis module.

[0119] Through the above fine preprocessing process, this module achieves:

[0120] Data quality improvement: Through cleaning and filtering, noise and outliers are effectively suppressed, improving the signal-to-noise ratio of the data.

[0121] Information density enhancement: Through feature engineering, a large amount of deep information that can represent the health state of the motor is extracted from limited raw parameters.

[0122] Analysis robustness: Through normalization and operating condition association, subsequent machine learning models can more stably and accurately perform cross-condition and cross-time period health state evaluation.

[0123] Data analysis module: including a storage unit and a diagnostic analysis unit;

[0124] The storage unit is used to store an electrical performance baseline model constructed based on the electrical data of the same type of wiper motor population, and a fault mode library constructed based on historical fault data of the wiper motor;

[0125] The diagnostic analysis unit is used to perform real-time comparison and analysis of the received motor state feature dataset with the electrical performance baseline model and the fault mode library, calculate the real-time health state score of the target motor based on machine learning algorithms, and predict the remaining useful life based on the electrical parameter degradation trend;

[0126] The storage unit construction process of this embodiment is as follows:

[0127] The storage unit is the knowledge base of the system, responsible for storing the benchmark models and fault cases for comparison and analysis. It is usually implemented by non-volatile memory and high-speed static random access memory, with the former for persistent storage and the latter for fast access during model runtime.

[0128] Baseline model of electrical performance: This model is generated based on historical data of a large number of same type of healthy wiper motors running under multiple typical working conditions (such as different vehicle speeds, different wiper gears), through Gaussian Mixture Model (GMM) or multivariate Gaussian distribution for clustering and modeling.

[0129] The baseline model construction process of the present embodiment is as follows:

[0130] The model construction method includes: first, data collection, collecting current, voltage, temperature, and speed time series data of the same type of motors judged as healthy under different wiper gears and different vehicle speed intervals in strict tests and actual use. Then, feature extraction and working condition labeling, pre-processing and feature extraction of the original data, forming feature vectors, and labeling each vector with its corresponding working condition label. Finally, clustering modeling, for each set of feature vectors under a specific working condition label, a Gaussian mixture model is trained to describe the multivariate probability distribution of the feature data of the healthy motor under this working condition;

[0131] Clustering modeling: for each set of feature vectors under a specific working condition label, a Gaussian mixture model is used for unsupervised clustering modeling. The Gaussian mixture model assumes that the data in the healthy state forms several sub-clusters in the feature space that conform to the Gaussian distribution, which can accurately describe the multi-modal distribution characteristics of the wiper motor under different healthy working conditions.

[0132] The specific process of clustering modeling is as follows:

[0133] Model initialization: based on the Bayesian Information Criterion or Akaike Information Criterion, and through multiple initialization comparisons by the Expectation-Maximization algorithm, the optimal number of Gaussian components of the healthy data under the current working condition is determined. This number reflects the typical types of healthy states of the wiper motor under stable operation in this working condition;

[0134] Parameter estimation: the Expectation-Maximization algorithm is used for iterative optimization until the model parameters converge. This algorithm calculates the posterior probability of each data point belonging to each Gaussian component through the E-step (Expectation Step), and then updates the weight, mean vector, and covariance matrix of each Gaussian component through the M-step (Maximization Step), finally making the log-likelihood function of the model reach the maximum;

[0135] Model verification: the silhouette coefficient or cross-validation method is used to evaluate the clustering effect, to ensure that the data points in the same cluster have high similarity, while the data points between different clusters have good separation degree. At the same time, each generated Gaussian component is checked for physical reasonableness to ensure that its mean and variance conform to the normal operation characteristics of the wiper motor under this working condition;

[0136] Model storage: The model parameters obtained after training convergence, including the weights, mean vectors and covariance matrices of each Gaussian component, are stored in the storage unit after being bound with the working condition label.

[0137] Through this process, the electrical performance baseline model can accurately describe the complex probability distribution of the multi-dimensional electrical characteristics of the healthy wiper motor in a specific working condition in the feature space, providing a solid statistical benchmark for subsequent Mahalanobis distance-based anomaly detection.

[0138] In terms of data structure, the baseline model is represented as a queryable data structure in the storage unit, which stores a set of corresponding model parameters for each working condition combination. For example, for the working condition "low gear, vehicle speed < 30 km / h", the stored parameters include: feature mean vector, which can include current mean, voltage mean, temperature mean, etc.; feature covariance matrix, which describes the cooperative change relationship between these features; and a health threshold Mahalanobis distance calculated based on the distribution.

[0139] This model specifically depicts the normal distribution range of the electrical characteristics of the wiper motor under healthy state as the wiper gear and vehicle speed change, providing a wiper motor-specific benchmark for subsequent health score calculation based on Mahalanobis distance.

[0140] The specific construction process of the fault mode library is as follows:

[0141] Fault mode library: This library is a structured queryable database or lookup table built based on historical fault cases and big data analysis. Its construction is a closed-loop process of continuous learning and improvement, which specifically includes the following core links:

[0142] 1. Fault data collection and aggregation: The system collects wiper motor fault data through various channels to form an initial case library. The data sources mainly include:

[0143] Laboratory bench test: In a controlled environment, simulate various typical faults of the wiper motor, such as artificially setting the wiper blade to be stuck, accelerating the carbon brush wear, and creating a slight short circuit in the winding, and simultaneously collecting high-frequency electrical parameter data throughout the process.

[0144] Real vehicle road test verification: Conduct durability and reliability tests in real road environments, and record naturally occurring or induced fault data.

[0145] After-sales and maintenance channel feedback: Collect warning data and final confirmed fault information reported by on-site vehicles through the remote monitoring center, as well as maintenance records and fault part analysis reports from maintenance sites.

[0146] 2. Fault feature extraction and fingerprint construction: For each confirmed fault case, the system extracts deep features from its corresponding raw electrical data to construct a unique "feature fingerprint" of the fault:

[0147] Multi-dimensional feature correlation: Not only extract single current or temperature value, but also focus on extracting feature combination and variation trend that can represent fault mechanism. For example, for 'carbon brush wear' fault, its fingerprint is specifically described as: current waveform factor continuously increases by more than 15% (e.g. from baseline 1.0 to 1.15 or above), while current standard deviation increases by more than 20%, and this pattern repeatedly occurs in multiple working cycles. For 'winding short circuit' fault, its fingerprint is described as: current waveform factor rapidly rises by more than 20% in a short time (10 seconds), accompanied by synchronous increase of drive voltage waveform factor and abnormal rise of winding temperature (temperature change rate exceeds 0.5℃ / s). These feature fingerprints are statistically derived based on historical fault data and have been verified through experiments.

[0148] Working condition decoupling: When constructing feature fingerprints, working condition information at the time of fault occurrence is associated to ensure that the extracted pattern is a stable representation that holds true under specific or all working conditions.

[0149] 3. Pattern storage and knowledge: Standardize and knowledgeize the extracted fault feature fingerprints to form a record in the library that can be automatically queried and matched by the system. Each record contains the following fields:

[0150] Fault code: A unique code that identifies the fault type, such as F001.

[0151] Fault mode description: A textual description of the fault type, such as "stator winding short circuit".

[0152] Feature fingerprint description: A detailed and quantitative description of the core electrical feature pattern of the fault.

[0153] Disposal suggestion: According to the fault type and severity, give specific and operable maintenance guidance text.

[0154] Warning level: According to the impact of the fault on the safety and function of the motor, the preset warning level (level 1 / level 2 / level 3).

[0155] 4. Verification and iteration of the library: The fault mode library is not static and unchangeable, and its accuracy needs to be verified through historical data backtracking, that is, using the patterns in the library to match historical data to test whether it can correctly identify known faults. At the same time, the system has online learning ability: when new and confirmed fault cases and their data are collected through the remote monitoring center, the system will automatically or under the guidance of semi-automatic start the model updating process, extract the feature fingerprint of the new fault, and supplement it as a new record to the fault mode library, so as to realize the continuous enrichment and continuous improvement of the accuracy of the library.

[0156] The embodiment needs to be specifically described as follows:

[0157] Reference Figure 3 As described above, the diagnostic analysis unit is the reasoning engine of the system, which is usually realized by an embedded machine learning reasoning framework running on the host processor. It receives the feature vector from the preprocessing module and performs the following core tasks:

[0158] 1. Real-time health state score calculation: This task is completed by comparing the real-time feature vector with the baseline model dedicated to the wiper motor.

[0159] Abnormality detection based on Mahalanobis distance: The diagnostic unit calculates the Mahalanobis distance between the real-time feature vector and the health cluster corresponding to the current wiper operating position and vehicle speed. Mahalanobis distance considers the correlation between features, and can effectively deal with the characteristics of coordinated changes of multiple parameters of the wiper motor, and is more true than Euclidean distance to reflect the deviation from the health state.

[0160] Score mapping: The calculated Mahalanobis distance is mapped to a real-time health state score between 0 and 100 through a predefined nonlinear function:

[0161] Score = 100: Indicates that the current state is completely in the center of the health cluster.

[0162] Score ≈ 70 (preset threshold): Indicates that the state has been at the edge of the health cluster, and early degradation such as slight wear of carbon brush has begun to appear.

[0163] Score → 0: Indicates that the state deviates seriously from the health benchmark, and may soon occur an emergency fault such as serious stall or winding short circuit.

[0164] 2. Fault mode identification: This task is completed by matching the real-time feature with the fault mode library.

[0165] Pattern matching algorithm: using nearest neighbor classifier or rule-based method. The diagnosis unit calculates the similarity between the real-time feature pattern and the feature pattern of each record in the fault pattern library. When the similarity exceeds a certain threshold, it is determined that the fault corresponding to the record may occur, and the corresponding fault code is output.

[0166] 3. Remaining useful life prediction: This task is to predict the future health state using trend data.

[0167] Trend feature extraction: The diagnosis unit maintains a time series database to record long-term historical data of health state scores and key features (such as average current).

[0168] Degradation trend modeling: Time series prediction models are used, such as: Autoregressive Integrated Moving Average Model (ARIMA): suitable for linear data with obvious trends and seasonality; Long Short-Term Memory Network (LSTM): more suitable for capturing non-linear, complex degradation trends, with higher accuracy.

[0169] RUL calculation: The model extrapolates the future health score decline trajectory based on historical data. The remaining useful life is defined as the predicted time it takes for the health score to drop from the current value to a pre-set failure threshold (such as 20). This prediction value is dynamically updated as new data arrives.

[0170] It needs to be further explained that the electrical performance baseline model uses a Gaussian mixture model probability modeling method, which has the advantage of being able to describe the multiple clusters that the data in the health state may form in the feature space. Compared with a single threshold range, it can more accurately depict the normal performance of the motor under different health conditions. The application of Mahalanobis distance is another major feature of this module. It is different from Euclidean distance and is a distance measure that takes into account the correlation between features, which can more truly reflect the degree of deviation of a data point from a data cluster, making the health state score more scientific and reliable. At the same time, the dynamic updating mechanism of the remaining useful life prediction module ensures that the prediction model can continuously correct the prediction results as the actual degradation process of the motor progresses, achieving adaptive learning ability.

[0171] Early warning decision module: used to compare the real-time health state score with the pre-set electrical performance threshold, and generate corresponding early warning information when the electrical performance is abnormal;

[0172] The multi-level threshold and rule fusion decision process of this embodiment needs to be specifically explained as follows:

[0173] The early warning decision module uses the system main control microprocessor as the hardware platform, and its internal implementation is a high-reliability rule-based reasoning engine. The engine reads the structured diagnostic results directly from the result buffer of the data analysis module through the internal data bus, including:

[0174] : Real-time health score (0-100).

[0175] : Recognized fault code (may be null).

[0176] : Predicted remaining useful life (hours / days).

[0177] Reference Figure 2 The module adopts a multi-level threshold and rule fusion decision-making strategy, and its core logic flow is as follows:

[0178] 1. Hierarchical warning based on health score: The module compares the real-time health score with a set of pre-set electrical performance thresholds for wiper motor degradation behavior. These thresholds are calibrated through a large number of experiments and field data for wiper motors.

[0179] Primary warning (mild / attention level): Triggering condition: Indicates early signs of performance degradation, such as slight current fluctuations caused by initial carbon brush wear, but it can still be safely operated for a certain period of time. At this time, the system focuses on recording the degradation trend and prompting the user to pay attention.

[0180] Secondary warning (moderate / maintenance level): Triggering condition: Indicates that the performance degradation has been relatively obvious, possibly accompanied by increased wiper resistance or moderate carbon brush wear, increasing the risk of failure. The system suggests planned maintenance during the next maintenance cycle.

[0181] Tertiary warning (severe / emergency level): Triggering condition: Indicates that the motor is in a serious unhealthy state, with high risks such as severe stall, winding short circuit, etc. The system requires immediate repair.

[0182] 2. Precise warning based on fault mode library: When the data analysis module identifies a specific , the module will start the precise warning logic.

[0183] Logic flow: The module queries the fault mode library according to to obtain the pre-set warning level and exclusive disposal suggestion for this fault. This logic takes precedence over simple score classification.

[0184] Example: Health score is 65, which should trigger a secondary warning, but if the fault code is identified as "F002: Winding Inter-turn Short Circuit", which is defined as "emergency" level in the library, then the final warning will be triggered. Level three, and generate precise information including "suspected motor winding short circuit, please stop immediately and check".

[0185] 3. Proactive Warning Based on Remaining Useful Life: This logic serves as an important supplement to the above two logics, focusing on long-term planning.

[0186] Triggering Condition: When the predicted remaining useful life is below a certain time threshold (e.g., 30 days); indicating that even if the current health score is below 70, the model predicts that its life is about to be exhausted, the module will trigger an early secondary warning, and the information will clearly indicate "the motor is predicted to fail within 30 days, it is recommended to plan ahead for replacement", achieving truly proactive maintenance.

[0187] Warning Information Generation and Output: When any of the warning triggering conditions are met, this module will generate a structured warning information package and immediately send it to the warning information output module. The information contains the following fields:

[0188] : Warning Trigger Time.

[0189] : Warning Level (Primary / Secondary / Tertiary).

[0190] : Current Health Score.

[0191] : Identified Fault Code.

[0192] : Predicted Remaining Life.

[0193] : Specific Disposal Suggestion Text Combined with the Characteristics of the Wiper Motor Fault, Including "Early Carbon Brush Wear, Continue to Observe" "Increased Wiper Blade Resistance, Recommend Cleaning or Replacing the Wiper Blade" "Motor Severe Stalling, Immediately Stop and Check the Transmission Mechanism".

[0194] : Record the main source of triggering the warning, such as "low health score" "specific failure mode" "short remaining life", for subsequent analysis.

[0195] Warning Information Output Module, used to send the warning information to the human-machine interface and remote monitoring center.

[0196] The specific description of this embodiment is as follows:

[0197] The hardware basis of the warning information output module is the system's main microprocessor and its integrated multiple communication peripherals. It connects with external systems through the following interfaces:

[0198] ​CAN Bus Controller and Transceiver: Used for communication with the in-vehicle network, it is one of the main channels for information output.

[0199] Wireless Communication Module: As the core of remote transmission, it is usually connected with the host processor through UART or USB interface, and has a built-in SIM card to realize internet access.

[0200] This module selects the optimal path for parallel or priority output according to the level and content of the warning information, and its core workflow is as follows:

[0201] 1. Output to the human-machine interface: This path aims to provide real-time and intuitive warning information for the driver.

[0202] Output protocol: Through the CAN bus, the warning information is sent to the combination instrument (IC) or information entertainment main screen (IVI) according to the diagnostic protocol or vehicle information protocol defined by the manufacturer.

[0203] Information presentation form:

[0204] Primary presentation: Trigger standard warning lights / icons and scroll short text information on the small display screen of the instrument panel, such as "wiper motor failure, please drive carefully".

[0205] Advanced presentation: Pop up a more detailed graphical warning window on the central large screen. This window clearly displays:

[0206] Warning level: Intuitively distinguish with color (green / yellow / red).

[0207] Specific information: Such as "health degree: 45%, motor performance has seriously degraded".

[0208] Suggested measures: Such as "predicted remaining life: 15 days, recommend immediate repair".

[0209] Fault code: Facilitate quick problem positioning for maintenance personnel.

[0210] 2. Output to the remote monitoring center: This path aims to provide data support for motor management, manufacturers or third-party service providers to realize remote monitoring and proactive service.

[0211] Data transmission protocol: The module encapsulates the warning information into lightweight data formats such as JSON or Protocol Buffers, and uploads the data to the remote monitoring center in the cloud through the integrated 4G / 5G wireless module using MQTT or HTTP / HTTPS protocol.

[0212] Transmission content: In addition to all the information generated by the early warning decision module, the identification code (VIN), GPS positioning information and data timestamp are usually attached to form a complete remote report.

[0213] Application scenarios:

[0214] Motor management: The large screen of the management center displays the real-time health status of the motor, and the abnormality is tracked and managed.

[0215] Active service: The manufacturer or service center can actively contact after receiving the third-level early warning or short remaining life early warning, remind the risk and provide one-key repair service, etc.

[0216] Big data analysis: All uploaded early warning data is gathered in the cloud, which can be used to optimize motor design, improve baseline model and enrich fault mode library.

[0217] Communication guarantee and degradation strategy: To ensure the reliable transmission of information, the following guarantee mechanisms are built into this module:

[0218] Sending confirmation and retry mechanism: For remote sending, if no confirmation response is received from the server, the module will perform multiple retries in the background according to the exponential backoff strategy until the sending is successful.

[0219] Information priority management: The third-level early warning information enjoys the highest sending priority and will interrupt the current low-level data communication and immediately attempt to send.

[0220] Network degradation processing: When no cellular network signal is detected, the early warning information will be stored in the local non-volatile memory. When the network is restored, the temporarily stored information will be automatically re-sent to the remote center. At the same time, the human-machine interface alarm will continue to be effective, ensuring that the warning is not lost.

[0221] Secondly: The drawings of the embodiments of the present application only involve the structures involved in the embodiments of the present disclosure, other structures can refer to the usual design, and in the case of no conflict, the same embodiment and different embodiments of the present application can be combined with each other;

[0222] Finally: The above only describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A wiper motor health status early warning system based on big data, characterized in that, include: Data acquisition module: used to collect the electrical operating parameters of the target motor in real time, including motor operating current, motor drive voltage, motor winding temperature and motor speed; Status interface module: used to acquire operating condition data related to the target motor load from the communication bus, the operating condition data including wiper switch status signal and speed signal; Data preprocessing module: used to preprocess the received electrical operating parameters and operating condition data to generate a motor state feature dataset; Data analysis module: includes storage unit and diagnostic analysis unit; The storage unit is used to store an electrical performance baseline model constructed based on the electrical data of a group of wiper motors of the same model, and a fault mode library constructed based on historical fault data of wiper motors. The diagnostic analysis unit is used to perform real-time comparative analysis of the received motor state feature dataset with the electrical performance baseline model and the fault mode library, calculate the real-time health status score of the target motor based on machine learning algorithms, and predict the remaining service life based on the degradation trend of electrical parameters. Early warning decision module: used to compare the real-time health status score with the preset electrical performance threshold, and generate corresponding early warning information when the electrical performance is determined to be abnormal; The early warning information output module is used to send the early warning information to the human-machine interface and the remote monitoring center.

2. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The electrical operating parameters include: The motor operating current is acquired by a Hall effect current sensor, which is connected in series in the drive circuit of the wiper motor. The motor drive voltage is acquired through a high-impedance differential operational amplifier circuit, which is connected in parallel across the motor drive power supply. The temperature of the motor windings is acquired by a digital temperature sensor, which is attached to the outer shell of the motor windings or embedded inside the motor. The motor speed is acquired by a Hall sensor or photoelectric encoder, which is installed on the output shaft or transmission mechanism of the wiper motor.

3. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The operating condition data includes: the operating condition data obtained from the vehicle communication bus; the operating condition data includes wiper switch status signal and vehicle speed signal.

4. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The preprocessing includes: The data cleaning and synchronization unit is used to perform timestamp alignment, outlier removal, and digital filtering on the collected electrical operating parameters and operating condition data. The feature extraction unit is used to extract time-domain statistical features and operating condition-related features from the preprocessed data to generate a motor state feature dataset. The data normalization unit is used to standardize the feature dataset to adapt it to the subsequent analysis model.

5. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The motor state feature dataset includes: time-domain features extracted from the electrical operating parameters that are sensitive to brush wear and winding short-circuit faults in the wiper motor, wherein the time-domain features include waveform factors of current or voltage.

6. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The electrical performance baseline model includes models generated by clustering and modeling a large amount of historical electrical data of the same type of healthy wiper motor under multiple operating conditions using a Gaussian mixture model.

7. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The fault mode library includes: wiper blade jamming and stalling, carbon brush wear, winding short circuit and transmission mechanism jamming; each fault record in the library contains a characteristic fingerprint corresponding to the wiper motor, handling suggestions and warning levels.

8. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The early warning decision module includes: Level 1 warning: The trigger condition is that the real-time health status score is lower than the first threshold but higher than or equal to the second threshold; Level 2 warning: The trigger condition is that the real-time health status score is lower than the second threshold but higher than or equal to the third threshold; Level 3 warning: The trigger condition is that the real-time health status score is lower than the third threshold.

9. The wiper motor health status early warning system based on big data according to claim 1, characterized in that, The warning information includes: warning trigger time, warning level, current health status score, identified fault code, predicted remaining service life, specific handling suggestion text, and warning trigger source; the handling suggestion text includes "initial carbon brush wear, continue to observe", "increased scraper resistance, it is recommended to clean or replace the scraper", and "motor severely stalled, stop immediately and check the transmission mechanism".

Images