Search

CN-122016334-A - System and method for predicting and maintaining service life of windscreen wiper

CN122016334ACN 122016334 ACN122016334 ACN 122016334ACN-122016334-A

Abstract

The invention relates to a system and a method for predicting and maintaining the service life of a windscreen wiper, and belongs to the technical field of intelligent network automobiles. The system comprises a sensor network, a processing unit, a wear detection module, a life prediction module, a user reminding module and the like, wherein the system acquires current of a wiper motor, working pressure of the wiper blade, a visual image after wiping and operation audio in real time, the health index of the wiper blade is output by integrating the data through a multi-mode fusion model based on LSTM, then a GRU-based time sequence prediction model is adopted, environment, usage habit and adhesive tape material data are combined to predict the residual service life, when the life is too low or the health index is too low, a grading early warning is sent to a user through a vehicle-mounted terminal and mobile application, the driving safety is ensured by self-adapting to the speed limit, and the system supports continuous optimization of an algorithm model and system parameters through OTA upgrading. The invention solves the problem that the traditional wiper system lacks real-time wear monitoring and predictive maintenance capabilities, and obviously improves the driving safety and the system intelligence level.

Inventors

  • LI YONGTAO
  • CHEN HUI
  • DONG ZHENG
  • LIANG QUAN
  • CHEN ZHUSHENG
  • WANG HE
  • LI HANYAN

Assignees

  • 广西科技大学

Dates

Publication Date
20260512
Application Date
20260124

Claims (10)

  1. 1. The service life prediction and maintenance system of the windscreen wiper is characterized by comprising a sensor network, a vehicle-mounted ECU, a cloud server, a wear detection module, a service life prediction module, a user reminding module and an OTA upgrading module, wherein the wear detection module, the service life prediction module, the user reminding module and the OTA upgrading module are respectively arranged on the vehicle-mounted ECU; The sensor network comprises a current sensor, a pressure sensor, a visual sensor and an audio sensor, and is used for collecting working current data of a wiper motor, working pressure data of a wiper blade on a windshield, visual image data of the windshield after wiping and audio data of the wiper blade during operation in real time; The wear detection module adopts a multi-mode fusion model based on a long-short-term memory network, inputs a current characteristic vector, a pressure characteristic vector, a visual cleanliness characteristic vector and an audio characteristic vector which are acquired by the sensor network and are preprocessed, and outputs a health index HI (t) of the wiper blade; The service life prediction module adopts a time sequence prediction model based on a cyclic neural network, inputs a historical health index sequence output by the abrasion detection module, fuses an environmental factor vector, a using habit vector and an adhesive tape material characteristic vector, and outputs the residual service life RUL (t) of the wiper blade; The user reminding module monitors the health index HI (t) and the residual service life RUL (t) in real time, and when the preset early warning condition is met, the user is sent out a grading early warning through the vehicle-mounted terminal and/or the mobile application; The OTA upgrading module is respectively communicated with the vehicle-mounted ECU and the cloud server and is used for receiving an algorithm model updating packet, a system parameter configuration packet and a firmware updating packet issued by the cloud server, so as to realize remote optimization and upgrading of the system.
  2. 2. The system for predicting and maintaining the service life of the windscreen wiper according to claim 1, wherein the multi-mode fusion model based on the long-short-term memory network comprises a plurality of independent LSTM encoders and a feature fusion fully-connected network, each LSTM encoder is of a multi-layer stacked structure and is used for respectively processing a current feature vector, a pressure feature vector, a visual cleanliness feature vector and an audio feature vector, and a health index HI (t) is output through a fully-connected network nonlinear mapping after extracting high-level features, wherein the value range of the health index HI (t) is 0-1, wherein 0 represents complete failure, 1 represents a brand-new state, the environment factor vector E comprises historical average rainfall intensity, ultraviolet irradiation intensity, temperature fluctuation, pollutant type and rainfall PH value, the usage habit vector U comprises average daily use duration, high-speed wiping mode occupation ratio and pressure dynamic adjustment aggressive degree, and the adhesive tape material feature vector M comprises adhesive tape material type, material hardness reference value and material sensitivity aging coefficient.
  3. 3. The wiper life prediction and maintenance system according to claim 1, wherein the hierarchical early warning includes: A primary reminder of pushing mild maintenance advice text information when RUL (t) <30 days or HI (t) < 0.6; A secondary reminding, when RUL (t) is <7 days or HI (t) is <0.4, displaying a striking icon warning and accompanying a prompt tone, and providing nearby service network point information and a reservation function; And three-level reminding, namely calculating an environmental Risk index risk_env when RUL (t) =0 days or HI (t) <0.25, sending a speed limit instruction to a vehicle power system, displaying a red flashing warning icon, carrying out steering wheel vibration reminding and providing a one-key navigation to a service website function.
  4. 4. The system for predicting and maintaining the service life of the windscreen wiper according to claim 1, wherein the current sensor is a hall effect sensor, the sampling frequency is not lower than 1kHz, the pressure sensor is a piezoelectric sensor or indirectly acquires pressure data through a motor torque model, the visual sensor is a vehicle-mounted camera with resolution not lower than 1080p, the audio sensor is a directional microphone and is used for collecting audio signals generated by friction between the windscreen wiper and the windscreen, and upgrading contents supported by the OTA upgrading module comprise abrasion detection model updating, service life prediction model updating, current threshold adjustment, pressure threshold adjustment, reminding strategy optimization, system firmware repairing and newly-added monitoring indexes or early warning functions.
  5. 5. A method for predicting and maintaining the service life of a wiper blade, characterized in that the system for predicting and maintaining the service life of a wiper blade according to any one of claims 1 to 4 is used, comprising the steps of: S1, data acquisition, namely acquiring wiper motor working current data, wiper blade working pressure data, wiped visual image data and wiper operation audio data in real time through a current sensor, a pressure sensor, a visual sensor and an audio sensor; S2, data preprocessing and feature extraction, namely filtering, time synchronization alignment and image processing are carried out on the collected original data, and a current feature vector V_I, a pressure feature vector V_P, a visual cleanliness feature vector V_S and an audio feature vector V_A are respectively extracted; s3, detecting the health state, namely inputting the four feature vectors into a multi-mode fusion model based on LSTM, and outputting the health index HI (t) of the wiper blade; S4, predicting the residual service life, namely inputting a historical health index sequence, an environmental factor vector E, a using habit vector U and an adhesive tape material characteristic vector M into a GRU-based time sequence prediction model, and outputting the residual service life RUL (t); s5, performing hierarchical early warning, namely triggering corresponding primary, secondary or tertiary early warning according to the monitoring results of HI (t) and RUL (t), sending reminding information to a user and executing corresponding safety control measures; s6, system upgrading, namely receiving an update package issued by the cloud through an OTA upgrading module, and realizing remote updating and optimizing of an algorithm model, system parameters and firmware.
  6. 6. The method for predicting and maintaining the life of a wiper blade according to claim 5, wherein the step S1 includes: S11, collecting motor current data, namely monitoring working current of a wiper motor in real time through a current sensor, recording current waveforms of each wiping period, wherein the sampling frequency is not lower than 1kHz, and the wiping period comprises starting, uniform speed and stopping stages; S12, collecting pressure data of the wiper blade, namely indirectly calculating the pressure value of the wiper blade on the windshield through a pressure sensor integrated on a wiper arm or based on a motor torque model, and synchronizing sampling frequency and current to ensure the time sequence alignment of the data; S13, acquiring visual data of the wiping effect, namely capturing an image of the windshield after each wiping by a visual sensor, wherein the image covers a main wiping area; S14, collecting wiping audio data, namely collecting audio signals generated by friction between a windscreen wiper and the windscreen when the windscreen wiper runs through a directional microphone, wherein the sampling frequency is not lower than 44.1kHz; the step S2 includes: S21, preprocessing a current signal, namely, applying a low-pass filter to an original current waveform to eliminate high-frequency noise and power supply interference, and extracting the following characteristics from each wiping period, namely, an average current I_average of a wiping uniform speed stage, a peak current I_peak during starting and reversing, a current fluctuation amplitude I_wave of current of the uniform speed stage and the current wiping total Energy-consumption condition to form a characteristic vector V_I; S22, preprocessing a pressure signal, namely synchronously aligning pressure data and current data through a time stamp, wherein the pressure signal is characterized in that a target pressure value P_target of a system instruction, an actual measured average pressure P_actual_average and a fluctuation variance P_variance in a pressure control process form a feature vector V_P; S23, performing visual preprocessing on the wiping effect, namely analyzing a camera image by using an image processing algorithm, calculating a cleanliness score, obtaining a comprehensive cleanliness percentage score S_score by comparing pixel differences of the wiped area and an ideal cleaning area, identifying a water mark line by an edge detection algorithm, calculating the ratio of the residual water film or the stain area to the total visual field area to form a feature vector V_S= [ S_score, streak _count and Residue _area_ratio ], wherein S_score is the comprehensive cleanliness percentage score, streak _count is the number of strip-shaped water marks or stains identified in the image, and Residue _area_ratio is the ratio of the residual water film or the stain area to the total visual field area; S24, preprocessing an audio signal, namely, eliminating environmental noise by applying a wavelet threshold denoising algorithm to the original audio signal, converting the audio signal into a frequency domain through Fourier transformation, and extracting the characteristics of an abnormal sound frequency band energy ratio A_ abnormal _ratio, a short-time energy entropy A_entropy and a zero-crossing rate A_zero_cross to form a characteristic vector V_A= [ A_ abnormal _ratio, A_entropy and A_zero_cross ], wherein the abnormal sound frequency band is 5kHz-7kHz.
  7. 7. The method according to claim 6, wherein the step S3 includes inputting the current feature vector V_I, the pressure feature vector V_P, the visual cleanliness feature vector V_S and the audio feature vector V_A into the LSTM-based multi-modal fusion model, and detecting the health status according to the following procedures: S31, judging whether the current I (t) tends to rise, if not, returning to the step S1 to restart, if so, entering the step S32; s32, judging whether the pressure P (t) rises synchronously, if not, proceeding to step S36, if so, proceeding to step S33; S33, judging whether the visual cleanliness S (t) is reduced, if not, entering a step S35, and if so, entering a step S34; s34, confirming the strong abrasion characteristic, determining that the wiper is in a strong abrasion state, and entering step S37; s35, judging that the environment change or the slight scraping degradation is possible, wherein the environment change or the slight scraping degradation belongs to a slight performance change condition; S36, judging that single interference is possibly ignored or weight is reduced, and the method belongs to temporary interference conditions; S37, inputting the results of the steps S34, S35 and S36 into a multi-mode fusion model to calculate a health index HI (t); s38, updating a health index HI (t) and quantifying the current health state of the wiper; S39, judging whether HI (t) is smaller than a threshold value, if not, returning to the step S1, and if so, entering the step S5; the step S37 includes: S371, inputting the feature vectors V_ I, V _ P, V _ S, V _A into corresponding LSTM encoders respectively, extracting high-level feature representation H_ I, H _ P, H _ S, H _A, splicing H_ I, H _ P, H _ S, H _A in series to form 256-dimensional fusion feature vectors, carrying out nonlinear mapping through a fully connected network, and finally outputting health index HI (t); S372, training a model in an offline mode by using a historical data set at a cloud server, wherein the historical data comprise brand new to invalid wiper blade data, optimizing a loss function, and ensuring model convergence in a training period of at least 1000 rounds; S373, performing online self-adaption, namely when a user executes manual scraping or replaces a wiper blade, performing fine adjustment on the model by using new data, updating LSTM and full-connection network weights, and uploading fine adjustment data to a cloud end through OTA for model optimization.
  8. 8. The method for predicting and maintaining the life of a wiper blade according to claim 5, wherein the step S4 includes: s41, data collection and pretreatment: S411, collecting a historical health index sequence [ HI (t), HI (t-1) ], wherein the historical health index sequence is output by the abrasion detection subunit, and the HI (t-n) is at least 50 scraping times; s412, collecting environmental data E, namely acquiring historical average rainfall intensity, ultraviolet intensity, temperature fluctuation, pollutant type and rainfall PH value through a vehicle-mounted sensor or an external API, and normalizing into vectors; s413, collecting usage habit data U, namely extracting average daily usage time length, high-speed wiping mode duty ratio, pressure dynamic regulation frequency and the like from a vehicle CAN bus, and normalizing the average daily usage time length, the high-speed wiping mode duty ratio, the pressure dynamic regulation frequency and the like into vectors; S414, collecting adhesive tape material data M: obtaining the material type, the material hardness reference value and the material aging sensitivity coefficient of the wiper adhesive tape, normalizing the vector into a vector; s42, predicting an RNN prediction model: The method comprises the steps of learning sequence dependency relations through a GRU network by using n past health index sequences, environment vectors E, using habit vectors U and adhesive tape material vectors M, and outputting hidden states, wherein a full-connection layer maps the hidden states into RUL (t), and mathematical expressions of RUL (t) =NN_model (HI (t), HI (t-1), HI (t-n) |E, U and M), wherein NN_model is a trained GRU Model, and triggering a user to remind if RUL (t) < 30 days; S43, updating a model: and periodically receiving updated model parameters through OTA to adapt to a new environment mode and a use habit, thereby improving the prediction precision.
  9. 9. The method for predicting and maintaining the life of a wiper blade according to claim 5, wherein the step S5 includes: S51, threshold monitoring Real-time monitoring of RUL (t) and HI (t): triggering a primary alert if RUL (t) <30 days or HI (t) < 0.6; triggering a secondary alert if RUL (t) <7 days or HI (t) < 0.4; if RUL (t) =0 days or HI (t) <0.25, triggering a three-level alert and initiating an environmental risk assessment and vehicle speed limit mechanism; s52, reminding generation and sending The primary reminding is 'suggested maintenance', text messages are displayed through a vehicle-mounted display screen, and the same information is pushed through a mobile phone application program; The second-level reminding is 'quick change', a striking icon and a warning sound are displayed on an instrument panel, a notification is pushed through a mobile phone application program, and a map of a nearby service network point and a reservation link are embedded; The three-level reminding is 'emergency replacement', a striking red flickering warning icon is displayed on an instrument panel, the current environment risk index is calculated, the vehicle speed is properly limited, a 'wiper blade emergency replacement-vehicle speed limitation' prompt is displayed on a central control screen, and the warning is enhanced through the vibration of a steering wheel and accompanied by a prompt tone; S53 user interaction The user confirms the reminding through the vehicle-mounted interface or the mobile phone App, and the system records the feedback data and uploads the feedback data to the cloud for subsequent model optimization and reminding strategy adjustment.
  10. 10. The method for predicting and maintaining the life of a wiper blade according to claim 5, wherein the step S6 includes: S61 upgrade detection and download The vehicle ECU periodically connects with the cloud server to check available update, downloads the update package through a secure HTTP protocol, and verifies the digital signature to ensure security; S62 upgrade execution Replacing a deep learning model of wear detection or life prediction; parameter updating, namely adjusting a current threshold value, a pressure threshold value or a reminding threshold value to adapt to environmental conditions of different areas; updating the system firmware to repair bug or add function; personalized optimization, namely pushing a custom algorithm package according to the vehicle use history; s63 data feedback and optimization The cloud uses the data to retrain the model, and realizes continuous optimization and self-adaptive improvement of the system algorithm.

Description

System and method for predicting and maintaining service life of windscreen wiper Technical Field The invention relates to the technical field of intelligent network automobiles, in particular to a system and a method for predicting and maintaining the service life of a windscreen wiper. Background The traditional wiper system lacks real-time monitoring and predictive maintenance functions for the wear state of the wiper blade, so that the driving safety is directly affected when the wiper blade is suddenly worn or fails. The actual service life of the wiper blade cannot be accurately estimated in the prior art, users are difficult to replace in time, and serious potential safety hazards exist. In addition, the algorithm of the traditional system is solidified, cannot be optimally updated according to the actual use effect and environmental change, lacks continuous improvement capability, and is difficult to adapt to the use habits and diversified driving environments of different users. The invention patent application with the application publication number of CN120724679A discloses a method for predicting the service life of a wiper, an ageing reminding method and equipment, and the scheme has the defects that (1) only the hardness, friction force, abrasion thickness and other mechanical and material parameters of a wiper blade adhesive tape are focused, multi-mode data such as audio frequency, visual cleaning effect and the like are not integrated, the simulation of the actual working state of the wiper is not comprehensive enough, the data dimension is single, and (2) the scheme relies on a manually derived mathematical formula and dynamic calibration logic, so that complex nonlinear relations in the ageing process of the wiper are difficult to capture, and the prediction precision is easily affected in the face of a multi-factor coupling scene. The invention patent application with the application publication number of CN114347952A discloses a method, a device and a system for predicting the residual life of a windscreen wiper, which have the following defects that (1) although the method relates to images, audios, environments and operation characteristics, the residual life is calculated by adopting simple weighting, the internal association among multi-mode data is not fully excavated, the data fusion depth is insufficient, the prediction precision is still insufficient, and (2) the method, the device and the system only depend on a similarity comparison and linear weighting model of a preset reference sequence, lack modeling capability of long-term dependence of time sequence data and are difficult to adapt to the dynamic process of progressive aging of the windscreen wiper. The application publication number CN120764066A discloses a vehicle wiper service life detection method, a device, a vehicle and a program product, wherein the scheme has the following defects that (1) the method adopts an end-to-end deep neural network to predict the residual service life, but does not establish a physical mapping relation between the hardness of a rubber strip and a friction coefficient, and lacks fine modeling on an aging mechanism of the rubber strip material, (2) environmental parameter characteristic acquisition of the scheme depends on an external meteorological data interface, when the vehicle is in a network signal weak or no network coverage area, the real-time performance and accuracy of environmental parameter acquisition are influenced, and (3) the scheme provides multi-mode state data fusion, but does not relate to a friction coefficient dynamic calibration mechanism based on motor current parameters, and can not sense friction characteristic mutation caused by instantaneous load changes such as glass surface foreign matters, ice crystals and the like in real time. Disclosure of Invention The invention aims to solve the technical problem of providing a system and a method for predicting and maintaining the service life of a wiper blade, which can monitor the wearing state of the wiper blade in real time, accurately predict the residual service life and have continuous optimization capability. In order to solve the technical problems, the invention adopts the following technical scheme that the wiper life prediction and maintenance system comprises a sensor network, a vehicle-mounted ECU, a cloud server, a wear detection module, a life prediction module, a user reminding module and an OTA upgrading module, wherein the wear detection module, the life prediction module and the user reminding module are respectively arranged on the vehicle-mounted ECU; The sensor network comprises a current sensor, a pressure sensor, a visual sensor and an audio sensor, and is used for collecting working current data of a wiper motor, working pressure data of a wiper blade on a windshield, visual image data of the windshield after wiping and audio data of the wiper blade during operation in real time; The wear detect