Intelligent motorcycle monitoring system and method based on visual identification

Abstract

The invention discloses an intelligent motorcycle monitoring system and method based on visual identification, and relates to the technical field of fault monitoring analysis, and the intelligent motorcycle monitoring method specifically comprises the steps: deploying a sensor, collecting the data of the sensor, and carrying out the preprocessing of a vehicle-mounted calculation unit; extracting feature data from the preprocessed sensor data; taking the extracted feature data as a base, and training a running state evaluation model through data analysis and machine learning; potential faults are monitored in real time, evaluation is carried out through the operation state evaluation model, and when the faults occur, maintenance personnel check fault information, equipment states and historical records through the background system, and fault positioning and maintenance are rapidly carried out; and visually displaying the fault positioning and maintenance scheme. The intelligent operation state evaluation model is constructed in combination with data analysis and machine learning technologies, the vehicle state can be dynamically evaluated, and potential faults can be predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of fault monitoring and analysis, and specifically to an intelligent motorcycle monitoring system and method based on visual recognition. Background Art

[0002] The technical field of fault monitoring and analysis mainly focuses on real-time monitoring and fault diagnosis of the operating status of equipment through technical means such as sensor data acquisition, signal processing, feature extraction, machine learning, and big data analysis. This field is widely applied in industries such as industry, transportation, and energy, and realizes efficient and accurate fault location and prediction through non-contact monitoring, data-driven modeling, and intelligent algorithms. Its core goal is to improve the reliability of equipment operation, reduce downtime, optimize maintenance costs, and promote the development of intelligent and digital management.

[0003] With the development of intelligent transportation and intelligent manufacturing technologies, the fault monitoring and status assessment of vehicles have become important means to ensure operation efficiency and safety. The monitoring methods of traditional motorcycles mainly rely on regular inspections and manual maintenance of mechanical components, lacking real-time performance and intelligence. In addition, existing vehicle monitoring systems are more applied to automobiles or large transportation equipment, while the small-sized intelligent monitoring technology for motorcycles is relatively scarce. In recent years, with the development of technologies such as the Internet of Things, artificial intelligence, and big data analysis, it has become feasible to monitor the status and perform intelligent diagnosis of motorcycles. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent motorcycle monitoring system and method based on visual recognition to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An intelligent motorcycle monitoring method based on visual recognition, and the intelligent motorcycle monitoring method is specifically as follows: Deploy sensors, collect sensor data, and perform preprocessing through an in-vehicle computing unit; Extract feature data from the preprocessed sensor data; Using the extracted feature data as a basis, train an operating status evaluation model through data analysis and machine learning; Perform real-time monitoring on potential faults, evaluate through the operating status evaluation model, and when a fault occurs, maintenance personnel can view the fault information, equipment status, and historical records through the background system to quickly locate and repair the fault; Visually display the fault location and repair plan.

[0006] The deployment of sensors and the collection of sensor data are specifically as follows: Deploy sensors; Among them, different types of sensors are installed on key components of the motorcycle (such as the engine, brakes, transmission system, tires, etc.): The sensors include vibration sensors, temperature sensors, pressure sensors, and sound sensors; The vehicle-mounted computing unit (ECU) is used to collect sensor data in real time to capture the operating status information of each component.

[0007] The sensor data includes vibration data, temperature data, pressure data, and sound data.

[0008] The vibration data: Frequency components: The vibration signal is decomposed into multiple frequency components using Fourier transform to extract the main frequency and energy distribution; Amplitude: The difference between the maximum and minimum values of the signal, reflecting the intensity of vibration; Acceleration: The acceleration characteristics are calculated through second-order differentiation to capture rapidly changing vibration patterns; The temperature data: Average temperature: The average value within the sampling time, used to monitor the overall thermal balance; Temperature fluctuation range: The difference between the highest temperature and the lowest temperature, reflecting the operating stability; Instantaneous heating rate: The rate of change of temperature with time, used to monitor abnormal phenomena of rapid heating; The pressure data: Average pressure: Monitor the stable pressure level during component operation; Rate of change of pressure: The rate of change of pressure per unit time, used to capture dynamic loads; Peak pressure: The maximum pressure value within a short period of time, used to identify situations of high stress or impact loads; The sound data: Spectrum characteristics: The spectrum distribution of the sound signal is extracted using the fast Fourier transform, used to analyze the noise characteristics caused by specific faults; Sound intensity: The integral value (power) of the square of the amplitude of the signal, reflecting the energy level of the sound; Abnormal sound pattern: Identify non-periodic or sharp peak features present in the signal, usually related to friction, fracture, or impact.

[0009] Preprocessing is performed through the vehicle-mounted computing unit, specifically: Preprocess the collected data in the vehicle-mounted computing unit; The preprocessing includes data cleaning, denoising processing, and data normalization; Among them, the data cleaning: Remove outliers and invalid data; The denoising processing: Use a filtering algorithm (such as a low-pass filter) to eliminate noise interference; The data normalization: unify data of different dimensions into a comparable range; the method of data normalization can adopt the Max-Min method or the Z-score method; Set a timestamp; Align different sensor data according to the timestamp.

[0010] Among them, the timestamp is a time mark for different sensor data.

[0011] Extract feature data from the preprocessed sensor data, specifically: S4-1. Conduct preliminary screening through LASSO regression; S4-2. For the preliminary screening result, perform dimensionality reduction through principal component analysis to obtain feature data [R 1 、R 2 、…、R n ; R i ∈[R 1 、R 2 、…、R n ; where n represents the total number of feature data, n is a positive integer; R i represents the i-th feature data, i represents the quantity label of the feature data, i is a positive integer, and 1 ≤ i ≤ n.

[0012] With the extracted feature data as the basis, evaluate the model of the operation state through data analysis and machine learning training, specifically: The specific formula for evaluating the operation state is: ; Among them, S represents the current operation state evaluation value; w i represents the adjustable weight corresponding to the i-th feature data, and the sum is 1; φ i (R i ) represents the non-linear health scoring function of the i-th feature data; reflecting the normality degree of the operation state; η i (R i ) represents the sensitivity function of the i-th feature data; highlighting the influence of key anomalies; β represents the feature interaction factor; dynamically adjusting the influence of anomaly interaction on the evaluation result; δ(R) represents the cooperation degree between different feature data; used to capture potential joint failures; e is the natural logarithm; λ represents the time decay coefficient; S(t) represents the time state fitting function of the operation state evaluation value in history; T represents the operation time; Among them, the S(t) can be fitted according to the total operating time of a motorcycle component, or a period of time can be selected for fitting. Taking time as the coordinate axis, a one-to-one function mapping between time and the operating state evaluation value is established. Through a large amount of data in history, a specific function expression is fitted, which can represent the operating state of the component in history.

[0013] The fitting tool can select the data fitting library of Matlab or Python.

[0014] Among them, the time decay coefficient is different for different motorcycle components and is set based on the rated service life according to the actual situation; According to the operating state evaluation formula, training is carried out through machine learning.

[0015] Among them, the machine learning can select a convolutional neural network and train the model by minimizing the loss function.

[0016] Among them, a calculation method of the feature interaction factor is: .

[0017] Specifically: The non-linear health scoring function is specifically: ; Among them, R i 1 and R i 2 are the preset first scoring threshold and second scoring threshold; Among them, the first scoring threshold is the upper limit value of the health score and is set according to the nameplate information of each component of the motorcycle at the time of factory; the second scoring threshold is the lower limit value of the health score and is jointly determined according to the nameplate information of each component of the motorcycle at the time of factory and the historical state information of the motorcycle; it is different for different components.

[0018] The sensitivity function is specifically: ; Among them, k i represents the covariance coefficient between the i-th feature data and the operating state evaluation value; Among them, multiple groups of historical data can be selected, and the average value of the covariance is taken as the covariance coefficient k i .

[0019] Among them, when η i (R i ) > 1, that is, R i < R i 2When it indicates that the sensitivity of the feature data is abnormal, the health scoring function is directly set to 0, indicating that this feature is not referenceable.

[0020] The degree of coordination is specifically as follows: ; Among them, ξ ij represents the correlation coefficient between the feature data with quantity label i and the feature data with quantity label j; j represents the quantity label of the feature data, j is a positive integer, 1 ≤ j ≤ n; R j represents the j-th feature data.

[0021] Perform real-time monitoring on potential faults, and evaluate through the operating status evaluation model. When a fault occurs, maintenance personnel can view fault information, equipment status, and historical records through the background system to perform fault location and repair. Specifically: Set the first operating status threshold according to the nameplate information of different components; Set an early warning mechanism; Collect real-time information and evaluate through the operating status evaluation model. If the current operating status evaluation value is less than the first operating status threshold, the relevant components are determined to be faulty, and an early warning is given. Maintenance personnel can view fault information, equipment status, and historical records through the background system to perform fault location and repair.

[0022] It is characterized in that: visually display the fault location and repair plan, specifically: Construct a digital twin diagram and mark the location of the faulty component; Use color coding to highlight the status of the faulty component; Allow maintenance experts to view the fault situation through the remote system to guide on-site maintenance.

[0023] An intelligent motorcycle monitoring system based on visual recognition, the intelligent motorcycle monitoring system includes a data acquisition module, a preprocessing module, a feature extraction module, an operating status evaluation model module, a fault monitoring and alarm module, a visual display module, and a remote maintenance support module; The data acquisition module is used to deploy sensors and collect sensor data in real time to provide raw data; The preprocessing module is used to perform data preprocessing, including data cleaning, denoising processing, and data normalization; Provide timestamp alignment to ensure the consistency of different sensor data in the time dimension; The feature extraction module is used to extract key features from the preprocessed sensor data; it includes two steps: preliminary screening and dimensionality reduction; The operating status evaluation model module is used to evaluate the real-time operating status and provide a basis for fault monitoring; The fault monitoring and alarm module is used to detect potential faults in real time, provide early warnings and trigger alarms for maintenance by maintenance personnel; The visualization display module is used to intuitively present fault location and repair solutions through digital twin and data visualization technologies; The remote maintenance support module is used to store historical data, generate analysis reports, and provide remote support for maintenance personnel.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention realizes multi-dimensional real-time monitoring of key components of motorcycles by deploying multiple types of sensors (vibration, temperature, pressure, sound), constructs an intelligent operating status evaluation model by combining data analysis and machine learning technologies, and can dynamically evaluate the vehicle status and predict potential faults. The system has a real-time early warning mechanism to effectively avoid the occurrence of serious faults, and introduces digital twin technology to realize the visual display of fault information. Maintenance personnel can quickly locate problems and remotely guide maintenance. The overall design is miniaturized and highly integrated, adapting to different motorcycle environments, and significantly improving the intelligence, real-time performance and reliability of monitoring. Description of the Drawings

[0025] Figure 1 It is a schematic diagram of the steps of an intelligent motorcycle monitoring method based on visual recognition according to the present invention. Detailed Embodiments

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0027] Embodiment: As Figure 1 shown, the present invention provides a technical solution, an intelligent motorcycle monitoring method based on visual recognition. The specific intelligent motorcycle monitoring method is as follows: Deploy sensors, collect sensor data and preprocess it through an in-vehicle computing unit; Extract feature data from the preprocessed sensor data; Using the extracted feature data as a basis, train an operating status evaluation model through data analysis and machine learning; Perform real-time monitoring of potential faults, evaluate through the operating status evaluation model. When a fault occurs, maintenance personnel can view fault information, device status and historical records through the background system, and quickly locate and repair the fault. Visually display the fault location and repair plan.

[0028] Deploy the sensors and collect sensor data, specifically: Deploy sensors; Among them, install different types of sensors on the key components of the motorcycle (such as the engine, brakes, transmission system, tires, etc.): The sensors include vibration sensors, temperature sensors, pressure sensors, and sound sensors; Collect sensor data in real time through the in-vehicle computing unit (ECU) to capture the operating status information of each component.

[0029] The sensor data includes vibration data, temperature data, pressure data, and sound data.

[0030] The vibration data: Frequency components: Use Fourier transform to decompose the vibration signal into multiple frequency components, and extract the main frequency and energy distribution; Amplitude: The difference between the maximum and minimum values of the signal, reflecting the intensity of vibration; Acceleration: Calculate the acceleration characteristics through second-order differentiation to capture rapidly changing vibration patterns; The temperature data: Average temperature: The average value within the sampling time, used to monitor the overall thermal balance; Temperature fluctuation range: The difference between the highest temperature and the lowest temperature, reflecting the operating stability; Instantaneous temperature rise rate: The rate of change of temperature over time, used to monitor abnormal phenomena of rapid heating; The pressure data: Average pressure: Monitor the stable pressure level during component operation; Pressure change rate: The rate of change of pressure per unit time, used to capture dynamic loads; Pressure peak: The maximum pressure value within a short period of time, used to identify situations of high stress or impact loads; The sound data: Spectrum characteristics: Use the fast Fourier transform to extract the spectrum distribution of the sound signal, used to analyze the noise characteristics caused by specific faults; Sound intensity: The integral value (power) of the square of the signal amplitude, reflecting the energy level of the sound; Abnormal sound pattern: Identify the non-periodic or sharp peak characteristics present in the signal, usually related to friction, fracture, or impact.

[0031] Example 1, for a motorcycle, deploy sensors on its key components: Engine: Install vibration sensors, temperature sensors, and pressure sensors.

[0032] Brake: Install pressure sensors and sound sensors.

[0033] Transmission system: Install vibration sensors and temperature sensors.

[0034] Tire: Install pressure sensors; Example 2, vibration sensor: Installed at the crankshaft position, collecting vibration frequency and amplitude, with a frequency range of 0 - 10 kHz and a sensitivity of ±0.5%; Temperature sensor: Installed at the cylinder head and exhaust pipe, monitoring the temperature in real - time, with a measurement range of - 40°C to 200°C and an accuracy of ±0.2°C; Pressure sensor: Arranged in the fuel pipeline, monitoring the fuel pressure, with a measurement range of 0 - 10 MPa and an accuracy of ±1%; Sound sensor: Installed in the engine cavity, recording the operating noise, with a frequency range of 20 Hz - 20 kHz and a sensitivity of ±3 dB; The vehicle - mounted computing unit (ECU) collects sensor data in real - time, for example: Vibration data: When the engine is running, the vibration sensor collects vibration signals, which are decomposed into multiple frequency components using Fourier transform. The main frequencies are 50 Hz and 100 Hz, and the energy is mainly distributed in the 50 - 150 Hz frequency band; the amplitude range is 0.1 - 0.5 mm; the acceleration characteristics are calculated through second - order differentiation to capture fast - changing vibration patterns.

[0035] Temperature data: The engine temperature sensor collects the average temperature within a period of time, which is 80°C, the temperature fluctuation range is 10°C, and the instantaneous temperature rise rate reaches 5°C / s during hard acceleration.

[0036] Pressure data: The brake pressure sensor monitors the average pressure of 5 MPa, the pressure change rate is 0.5 MPa / s during braking, and the pressure peak reaches 8 MPa during emergency braking.

[0037] Sound data: The brake sound sensor collects sound signals, and its frequency spectrum distribution is extracted using fast Fourier transform, and abnormal sharp peak features are found; the sound intensity is calculated to be 0.5 W.

[0038] Pre - processing is performed through the vehicle - mounted computing unit, specifically: Pre - process the collected data in the vehicle - mounted computing unit; The pre - processing includes data cleaning, noise reduction processing, and data normalization; Among them, the data cleaning: Remove outliers and invalid data; The denoising process: Use a filtering algorithm (such as a low-pass filter) to eliminate noise interference; The data normalization: Unify data of different dimensions into a comparable range (such as [0, 1]); The method of data normalization can adopt the Max-Min method or the Z-score method; Set a timestamp; Align different sensor data according to the timestamp.

[0039] Among them, the timestamp is a time marker for different sensor data.

[0040] Extract feature data from the preprocessed sensor data, specifically: S4-1: Conduct a preliminary screening through LASSO regression; S4-2: For the preliminary screening results, perform dimensionality reduction through principal component analysis to obtain feature data [R 1 、R 2 、…、R n ; R i ∈[R 1 、R 2 、…、R n ; Among them, n represents the total number of feature data, and n is a positive integer; R i represents the i-th feature data, i represents the quantity label of the feature data, i is a positive integer, and 1 ≤ i ≤ n.

[0041] Taking the extracted feature data as the basis, evaluate the model of the operation state through data analysis and machine learning training, specifically: The specific formula for evaluating the operation state is: ; Among them, S represents the current operation state evaluation value; w i represents the adjustable weight corresponding to the i-th feature data, and the sum is 1; φ i (R i ) represents the non-linear health scoring function of the i-th feature data; reflecting the normality degree of the operation state; η i (R i ) represents the sensitivity function of the i-th feature data; highlighting the influence of key anomalies; β represents the feature interaction factor; dynamically adjusting the influence of anomaly interaction on the evaluation result; δ(R) represents the cooperation degree between different feature data; used to capture potential joint failures; e is the natural logarithm; λ represents the time decay coefficient; S(t) represents the time state fitting function of the operation state evaluation value in history; T represents the operation time; Among them, the S(t) can be fitted according to the total running time of a motorcycle component, or a period of time can be selected for fitting. Taking time as the coordinate axis, a one-to-one function mapping between time and the running state evaluation value is established. Through a large amount of data in history, a specific function expression is fitted, which can represent the running state of the component in history.

[0042] The fitting tool can select the data fitting library of Matlab or Python.

[0043] Among them, the time decay coefficient is different for different motorcycle components and is set based on the rated service life according to the actual situation; Train through machine learning according to the running state evaluation formula.

[0044] Among them, the machine learning can select a convolutional neural network and train the model by minimizing the loss function.

[0045] Among them, one calculation method of the feature interaction factor is: ; Specifically: The non-linear health scoring function is specifically: ; Among them, R i 1 and R i 2 are the preset first scoring threshold and second scoring threshold; Among them, the first scoring threshold is the upper limit value of the health score and is set according to the nameplate information of each component of the motorcycle at the time of factory; the second scoring threshold is the lower limit value of the health score and is jointly determined according to the nameplate information of each component of the motorcycle at the time of factory and the historical state information of the motorcycle; it is different for different components.

[0046] The sensitivity function is specifically: ; Among them, k i represents the covariance coefficient between the i-th feature data and the running state evaluation value; Among them, multiple groups of historical data can be selected, and the average value of the covariance is taken as the covariance coefficient k i .

[0047] Among them, when η i (R i ) > 1, that is, R i < R i 2When it indicates that the sensitivity of the feature data is abnormal, the health score function is directly set to 0, indicating that the feature has no reference value.

[0048] The degree of coordination is specifically: ; Among them, ξ ij represents the correlation coefficient between the feature data with quantity label i and the feature data with quantity label j; j represents the quantity label of the feature data, j is a positive integer, 1 ≤ j ≤ n; R j represents the j-th feature data.

[0049] Set according to the engine factory nameplate information: Example 1, fuel pressure R 1 1 = 3.5 MPa, R 1 2 = 2.0 MPa; The time decay coefficient λ = 0.1, set based on the engine's rated life of 5 years.

[0050] Covariance coefficient k i : Calculated from historical data, the covariance coefficient of fuel pressure data k 1 = 0.8, and the covariance coefficient of temperature data k 2 = 0.1; Example 2, data of a motorcycle engine, multiple features such as vibration, temperature, and pressure are collected. The calculation process of the operating state evaluation formula is as follows: Vibration feature (R 1 ): The collected vibration amplitude R 1 = 8 g, and R 1 1 = 10 g, R 1 2 = 5 g; Health score φ 1 (R 1 ) = (8 - 5) / (10 - 5) = 0.6; Sensitivity function η 1 (R 1 ) = 1 + k 1 ⋅ max(0, 5 - 8) = 1 + 0.2 ⋅ 0 = 1.

[0051] Temperature feature (R 2 ): The collected temperature R 2 = 95 °C, and R 2 1 = 100 °C, R 22 = 50 °C.

[0052] Health score φ 2 (R 2 ) = (95 - 50) / (100 - 50) = 0.9; Sensitivity function η 2 (R 2 ) = 1 + k 2 ⋅ max(0, 50 - 95) = 1 + 0.1 ⋅ 0 = 1; Degree of synergy δ(R): The correlation coefficient between vibration and temperature is ξ 12 = 0.8; Calculate the degree of synergy: δ(R) ≈ 55; Finally, by calculating all the characteristic data in turn and summing them up, the operation status evaluation value S is obtained, and the health status of the engine is judged according to the threshold.

[0053] Among them, for different components, the parameters are different and need to be adjusted according to the actual situation.

[0054] Real-time monitoring of potential faults is carried out, and evaluation is carried out through the operation status evaluation model. When a fault occurs, maintenance personnel can view the fault information, equipment status and historical records through the background system to locate and repair the fault. Specifically: Set the first operation status threshold according to the nameplate information of different components; Set up an early warning mechanism; Collect real-time information and evaluate it through the operation status evaluation model. If the current operation status evaluation value is less than the first operation status threshold, the relevant components are judged to be faulty, and an early warning is given. Maintenance personnel can view the fault information, equipment status and historical records through the background system to locate and repair the fault.

[0055] It is characterized in that: visually display the fault location and repair plan. Specifically: Construct a digital twin diagram and mark the location of the faulty component; Use color coding to highlight the status of the faulty component; Allow maintenance experts to view the fault situation through the remote system to guide on-site maintenance.

[0056] An intelligent motorcycle monitoring system based on visual recognition, the intelligent motorcycle monitoring system includes a data acquisition module, a preprocessing module, a feature extraction module, an operation status evaluation model module, a fault monitoring and alarm module, a visual display module and a remote maintenance support module; The data acquisition module includes: Sensor Deployment: The system deploys multiple sensors, such as vibration sensors, temperature sensors, pressure sensors, sound sensors, etc., to monitor the operating status of each key component of the engine in real time.

[0057] Multi-channel Data Acquisition: Collect data from different sensors simultaneously and ensure their synchronization in time and space.

[0058] Edge Computing Support: Deploy edge computing modules to perform preliminary data analysis at the sensor site, reduce data transmission latency and improve real-time performance. By local computing, the dependence on the central processor can be reduced, which is suitable for environments with low bandwidth and high latency.

[0059] Data Storage and Caching: Locally cache the collected data to ensure that the data will not be lost in case of network interruption.

[0060] The preprocessing module includes: Data Cleaning: Remove incomplete, inaccurate, and duplicate data. Fill in missing data through methods such as data interpolation and anomaly detection.

[0061] Denoising Processing: Remove noise from sensor data to improve data quality.

[0062] Data Normalization: Normalize sensor data to unify signals from different sources to a standard scale for subsequent analysis and model processing. Timestamp Alignment: Align the collected data of different sensors using timestamps to ensure the temporal consistency of the data. Within the time window, use interpolation or resampling techniques to synchronize the data to a unified time axis.

[0063] The feature extraction module includes: Initial Screening: Screen out key features from the preprocessed data and select indicators with higher fault correlation (such as temperature, pressure, vibration amplitude, etc.) based on prior knowledge.

[0064] Dimensionality Reduction Processing: Use techniques such as principal component analysis (PCA) and linear discriminant analysis (LDA) for feature dimensionality reduction to reduce redundant data and improve computational efficiency.

[0065] The operating status evaluation model module includes: State Evaluation Model Training: Train the operating status evaluation model through historical data to learn the complex relationship between features and faults. Use supervised learning or reinforcement learning methods to continuously optimize model parameters. Real-time Evaluation: Obtain new data in real time and calculate the current health score through the evaluation model, and generate a health evaluation result according to the set threshold. For example, when the health score is lower than a certain threshold, trigger an alarm.

[0066] The fault monitoring and alarm module includes: Anomaly Detection: Monitor sensor data and operating status evaluation values in real time, and detect potential faults through methods such as threshold judgment and machine learning models. When the operating status evaluation value is lower than the set threshold, the system immediately triggers an alarm. Fault Prediction: Based on historical data and real-time data, combined with machine learning models, predict the fault trend and identify impending faults in advance. For example, predict possible wear, temperature increase, pressure fluctuation, etc. of components. Alarm Strategy: Support flexible alarm strategy settings, and issue alarms at different levels according to the severity of the faults. For example, set multiple alarm levels corresponding to different fault types and emergency levels.

[0067] The visualization display module includes: Digital Twin Technology: By constructing a digital twin model of the engine, synchronously display the real-time state of the physical system and the virtual model. Digital twins can help maintenance personnel intuitively understand the operating state of the system and the location of potential faults.

[0068] Data Visualization: Through visualization tools such as charts, heat maps, and trend charts, display various sensor data and fault detection results. Support historical data playback and trend analysis to help users better understand the operating state of the equipment.

[0069] Interactive Dashboard: Provide a dynamic dashboard to display real-time equipment status, health scores, alarm information, etc. Users can view detailed data of different modules through interactive operations.

[0070] Fault Location and Repair Plan: Based on the fault diagnosis results, the system can automatically recommend corresponding repair plans and present detailed repair steps and suggestions on the visualization interface.

[0071] The remote maintenance support module includes: Historical Data Storage: Store historical data such as collected sensor data, evaluation results, and alarm information in the cloud or local database for subsequent query and analysis.

[0072] Analysis Report Generation: Automatically generate analysis reports, including fault detection history, system performance evaluation, trend prediction, etc., to provide detailed fault analysis materials for maintenance personnel.

[0073] Remote Fault Diagnosis: Through the cloud platform, remote diagnosis and maintenance can be achieved. Maintenance personnel can remotely access the system, obtain real-time data, analyze the cause of the fault, and put forward repair suggestions.

[0074] Maintenance Record Management: Record each maintenance and repair operation, including maintenance time, maintenance personnel, fault type, and processing process, etc., for reference in long-term maintenance.

[0075] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in any respect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Accordingly, all changes that fall within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. An intelligent motorcycle monitoring method based on visual recognition, characterized in that: The intelligent motorcycle monitoring method is specifically as follows: Deploy sensors, collect sensor data and pre-process them through the on-board computing unit; Extracting feature data from preprocessed sensor data; Based on the extracted feature data, an operation status evaluation model is trained through data analysis and machine learning; Real-time monitoring of potential faults is performed and the operation status evaluation model is used to evaluate the faults. When a fault occurs, maintenance personnel can view the fault information, equipment status and historical records through the background system to quickly locate and repair the fault. The fault location and repair plan are displayed visually.

2. The intelligent motorcycle monitoring method based on visual recognition according to claim 1 is characterized in that: The deployment of sensors and collection of sensor data are specifically as follows: Deploy sensors; The sensors include vibration sensors, temperature sensors, pressure sensors and sound sensors; Collect sensor data in real time; The sensor data includes vibration data, temperature data, pressure data and sound data.

3. The intelligent motorcycle monitoring method based on visual recognition according to claim 2 is characterized in that: The vehicle-mounted computing unit performs preprocessing, specifically: The preprocessing includes data cleaning, denoising and data normalization; Set timestamp; Align different sensor data by timestamp.

4. The intelligent motorcycle monitoring method based on visual recognition according to claim 3 is characterized in that: Extracting feature data from the preprocessed sensor data is specifically as follows: S4-1, preliminary screening by LASSO regression; S4-2. For the preliminary screening results, the dimension is reduced by principal component analysis to obtain the characteristic data [R1, R2, ..., R n ]; R i ∈[R1, R2, …, R n ]; where n represents the total number of feature data, and n is a positive integer; R i Represents the i-th feature data, i represents the quantity label of the feature data, i is a positive integer, 1≤i≤n.

5. The intelligent motorcycle monitoring method based on visual recognition according to claim 4 is characterized in that: Based on the extracted feature data, the operation status evaluation model is trained through data analysis and machine learning, specifically: The specific operating status evaluation formula is: ； Where S represents the current operating status evaluation value; w i represents the adjustable weight corresponding to the i-th feature data, the sum of which is 1; φ i (R i ) represents the nonlinear health score function of the i-th feature data; η i (R i ) represents the sensitivity function of the i-th feature data; β represents the feature interaction factor; δ(R) represents the degree of coordination between different feature data; λ represents the time decay coefficient; S(t) represents the time state fitting function of the operating state evaluation value in the history; T represents the operating time; According to the operating status evaluation formula, training is performed through machine learning.

6. The intelligent motorcycle monitoring method based on visual recognition according to claim 5 is characterized in that: Specifically: The nonlinear health scoring function is specifically: ； Among them, R i 1 and R i 2 is a preset first scoring threshold and a second scoring threshold; The sensitivity function is specifically: ； Among them, k i Represents the covariance coefficient between the i-th feature data and the running status evaluation value; The degree of coordination is specifically: ； Among them, ξ ij represents the correlation coefficient between the feature data with quantity label i and the feature data with quantity label j; j represents the quantity label of the feature data, j is a positive integer, 1≤j≤n; R j Represents the jth feature data.

7. The intelligent motorcycle monitoring method based on visual recognition according to claim 6 is characterized in that: Potential faults are monitored in real time and evaluated through the operation status evaluation model. When a fault occurs, maintenance personnel view fault information, equipment status and historical records through the background system to locate and repair the fault, specifically: Setting a first operating state threshold according to nameplate information of different components; Setting up early warning mechanisms; Real-time information is collected and evaluated through the operation status evaluation model. If the current operation status evaluation value is less than the first operation status threshold, the relevant component is determined to be faulty and an early warning reminder is issued. Maintenance personnel view fault information, equipment status and historical records through the background system to locate and repair the fault.

8. The intelligent motorcycle monitoring method based on visual recognition according to claim 7 is characterized in that: Visually display the fault location and maintenance plan, specifically: Build a digital twin diagram and mark the location of the faulty component; Use color coding to highlight the status of faulty parts; Allow maintenance experts to view fault conditions through the remote system and guide on-site maintenance.

9. An intelligent motorcycle monitoring system based on visual recognition, using the intelligent motorcycle monitoring method based on visual recognition described in any one of claims 1 to 8, characterized in that: The intelligent motorcycle monitoring system includes a data acquisition module, a preprocessing module, a feature extraction module, an operation status evaluation model module, a fault monitoring and alarm module, a visual display module and a remote maintenance support module; The data acquisition module is used to deploy sensors and collect sensor data in real time to provide raw data; The preprocessing module is used to perform data preprocessing, including data cleaning, denoising and data normalization; Provide timestamp alignment to ensure the consistency of different sensor data in the time dimension; The feature extraction module is used to extract key features from the preprocessed sensor data; It includes two steps: preliminary screening and dimensionality reduction; The operating status evaluation model module is used to evaluate the real-time operating status and provide a basis for fault monitoring; The fault monitoring and alarm module is used to detect potential faults in real time, provide early warnings and trigger alarms for maintenance personnel to perform repairs; The visualization display module is used to intuitively present fault location and maintenance solutions through digital twin and data visualization technology; The remote maintenance support module is used to store historical data, generate analysis reports, and provide remote support for maintenance personnel.

Images