A comprehensive testing system and method for electric motors used in electric bicycles

Abstract

This invention discloses a comprehensive testing system and method for electric bicycle motors. By acquiring real-time time-series data of the electric bicycle motor under operating and testing conditions, multi-dimensional feature vectors characterizing the performance status of the electric bicycle motor are extracted. A first predictive evaluation model is constructed using these multi-dimensional feature vectors as input, resulting in a real-time health status quantitative score and fault risk classification for the electric bicycle motor. This provides a reference for the optimal design scheme of the electric bicycle motor. Simultaneously, a second predictive evaluation model is constructed based on the multi-dimensional feature vectors and historical state sequences to predict the remaining service life of the electric bicycle motor, providing a basis for designing the optimal service life of the electric bicycle motor. The combined output of the two models achieves comprehensive performance prediction of the electric bicycle motor, enabling manufacturers to produce electric bicycle motors that meet testing standards while reducing production costs, achieving the best cost-effectiveness.

Description

Technical Field

[0001] This invention relates to the field of electric motor testing technology, and in particular to a comprehensive testing system and method for electric motors used in electric bicycles. Background Technology

[0002] Electric bicycles, with their convenience, environmental friendliness, and affordability, have become an important means of personal short-distance transportation. As the core power component of an electric bicycle, the performance and reliability of the electric motor directly determine the overall riding experience, range, and safety. Therefore, effective and accurate testing of electric bicycle motors is a crucial step in ensuring product quality, protecting user safety, and optimizing after-sales service.

[0003] Currently, the testing of electric bicycle motors generally employs dynamometer benches for factory performance testing. This testing is typically conducted under specific loads, such as rated points, measuring key parameters like motor speed, torque, current, voltage, and efficiency, and strictly comparing them against design specifications to determine product qualification. However, this testing method is a sampling or end-point inspection, unable to quickly assess the full-condition performance of every electric bicycle motor. Secondly, the test conditions are singular and the time is short, making it difficult to expose potential inconsistencies or early latent faults (such as minor inter-turn short circuits or slight magnet unevenness). Finally, the test results are a set of discrete "pass / fail" data, unable to form a digital health record that can be used to track the long-term performance evolution of electric bicycle motors.

[0004] In addition, the new standard has added new technical requirements for the performance of electric motors used in electric bicycles, imposing strict requirements from the front end of the production and design of electric motors. However, in daily testing, the failure rate of electric motor samples submitted by companies for testing is high.

[0005] To address the aforementioned issues, this solution establishes a first predictive evaluation model that provides a reference for the optimal design of electric bicycle motors. Simultaneously, a second predictive evaluation model provides a basis for the optimal service life of electric bicycle motors. The combined output of the two models enables comprehensive performance prediction of electric bicycle motors, providing reliable data support for enterprises during the design phase of electric bicycle motor products. This allows enterprises to reduce production costs and achieve optimal cost-effectiveness while meeting testing standards. Summary of the Invention

[0006] The main objective of this invention is to provide a comprehensive testing system and method for electric motors used in electric bicycles. This system comprehensively utilizes multi-source data and integrates advanced algorithms to achieve a comprehensive intelligent testing system that leaps from single performance testing to full lifecycle health management, effectively solving the problems in the background technology.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A comprehensive testing method for electric motors used in electric bicycles includes the following steps:

[0009] Acquire real-time timing data of electric bicycle motors during operation and testing;

[0010] The real-time time series data is processed to extract a multi-dimensional feature vector characterizing the performance state of the motor;

[0011] The multi-dimensional feature vector is input into the first prediction and evaluation model to obtain a real-time health status quantitative score and fault risk classification of the electric motor for electric bicycles.

[0012] A second prediction and evaluation model is constructed based on the multi-dimensional feature vector and historical state sequence to estimate the remaining service life of the electric motor for electric bicycles.

[0013] Based on the combined results of the first and second prediction and evaluation models, a comprehensive performance prediction for the electric motor used in electric bicycles is obtained.

[0014] The system receives maintenance result feedback data from the maintenance terminal, uses the fault feature vector and the confirmed fault category as new labeled samples, and performs incremental learning and optimization on the supervised learning-based classification model.

[0015] Cluster analysis is performed on the factory inspection feature vectors of multiple motors of the same production batch or model to generate a quality distribution report and identify abnormal individuals that deviate from the standard group;

[0016] By aggregating the characteristic data of the entire life cycle of motors of the same model, fitting the common degradation trajectory of their key features, and establishing a standard healthy degradation baseline for motors of this model.

[0017] The real-time time series data includes the raw data of electrical operating parameters, thermodynamic state parameters, and external operating conditions and environmental parameters required to establish the first and second prediction and evaluation models.

[0018] The multi-dimensional feature vector includes data features extracted from the original data;

[0019] Furthermore, the first prediction and evaluation model is constructed using the following method:

[0020] First, a normal operating condition benchmark is constructed by using the historical feature vectors of the electric motor of an electric bicycle under known health conditions through a health benchmark modeling model.

[0021] Subsequently, the supervised learning-based classification model is trained using a multi-layer nonlinear transformation model based on the historical feature vectors and their corresponding state labels, resulting in a classifier that can map the input feature vectors to specific health states or fault types.

[0022] Furthermore, the second prediction and evaluation model includes a health indicator trajectory matching life prediction model and a degradation state filtering life prediction model.

[0023] The health indicator trajectory matching life prediction model specifically includes: a health indicator extraction unit, used to generate a one-dimensional health indicator based on multi-dimensional feature vectors; a trajectory matching unit, used to calculate the similarity distance between the real-time sequence of the one-dimensional health indicator and the curves in the historical degradation curve library; and a life prediction unit, used to estimate the useful life of the remaining electric motor for electric bicycles by weighted calculation based on the similarity matching results.

[0024] The degradation state filtering lifetime prediction model specifically includes: a state space model construction unit, used to establish a degradation probability model of the electric motor for electric bicycles containing state equations and observation equations; an online state estimation unit, used to recursively estimate the posterior distribution of the health state based on Kalman filtering or particle filtering algorithms; and a lifetime prediction unit, used to simulate future degradation paths and output the probability distribution of the remaining lifetime of the electric motor for electric bicycles based on the state equations and failure thresholds.

[0025] A comprehensive testing system for electric motors used in electric bicycles, comprising:

[0026] The data acquisition module collects and uploads real-time timing data of the electric motor of the electric bicycle during operation or testing.

[0027] The feature vector extraction module, connected to the data acquisition module, is used to process the real-time time series data to extract multi-dimensional feature vectors characterizing the performance status of the electric motor of the electric bicycle.

[0028] The first prediction and evaluation model construction module is connected to the feature vector extraction module and is used to obtain a real-time health status quantitative score and fault risk classification of the electric motor of the electric bicycle by taking the multi-dimensional feature vector as input.

[0029] The second prediction and evaluation model construction module is connected to the feature vector extraction module and is used to construct a second prediction and evaluation model based on the multi-dimensional feature vector and historical state sequence to estimate the remaining service life of the motor.

[0030] The detection result output module is connected to the first prediction and evaluation model construction module and the second prediction and evaluation model construction module to output the comprehensive performance prediction of the electric motor for electric bicycles.

[0031] The system also includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor, when executing the program, can implement a comprehensive testing method for electric motors used in electric bicycles.

[0032] The present invention has the following beneficial effects:

[0033] Compared with existing technologies, this solution can construct a dynamic performance profile of the electric motor of electric bicycles throughout its entire life cycle by continuously collecting real-time time-series data, reflecting its continuous state changes under real and variable operating conditions, thereby overcoming the technical problem that traditional detection can only obtain isolated data at a specific point in time and under a specific load.

[0034] Compared with existing technologies, this solution trains the first prediction and evaluation model based on big data technology, which can identify complex and coupled fault modes. At the same time, the extracted feature vectors provide a physically interpretable basis for the model evaluation results, thereby significantly improving the accuracy and interpretability of fault diagnosis in the detection process.

[0035] Compared with existing technologies, this solution, by constructing a second predictive evaluation model, can predict the failure risk of motors in the early stages of performance degradation, realizing the transformation from post-fault diagnosis to pre-health prediction, which is conducive to formulating corresponding maintenance strategies in advance based on the prediction results. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a comprehensive testing method for an electric motor used in electric bicycles according to the present invention.

[0037] Figure 2 This is a schematic diagram of the structure of a comprehensive testing system for electric motors used in electric bicycles according to the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0039] Example 1

[0040] In this embodiment, a health indicator trajectory matching lifespan prediction model is used to construct the second prediction and evaluation model. (See [link]). Figure 1 The flowchart shown is a comprehensive testing method for electric motors used in electric bicycles, including the following steps:

[0041] By communicating with the electric bicycle motor controller, real-time timing data of the electric bicycle motor under operating and testing conditions can be obtained;

[0042] Process real-time time-series data to extract multi-dimensional feature vectors that characterize the performance status of electric motors used in electric bicycles;

[0043] By inputting multi-dimensional feature vectors into the preset first prediction and evaluation model, the real-time health status quantitative score and fault risk classification of the motor are obtained.

[0044] A second predictive evaluation model is constructed based on multi-dimensional feature vectors and historical state sequences to estimate the remaining service life of the motor.

[0045] Based on the combined results of the first and second prediction and evaluation models, a comprehensive performance prediction for the electric motor used in electric bicycles is obtained.

[0046] The system receives maintenance result feedback data from the maintenance terminal, uses the fault feature vector and the confirmed fault category as new labeled samples, and performs incremental learning and optimization on the supervised learning-based classification model.

[0047] Cluster analysis is performed on the factory inspection feature vectors of multiple electric bicycle motors from the same production batch or model to generate a quality distribution report and identify abnormal individuals that deviate from the standard group;

[0048] By aggregating the characteristic data of the entire life cycle of the electric motors for the same model of electric bicycle, fitting the common degradation trajectory of their key characteristics, and establishing a standard healthy degradation baseline for the electric motors of this model of electric bicycle.

[0049] The following section provides a further explanation of this solution, detailing the specific implementation process, which includes the following steps:

[0050] Step S1: Standardization of the data acquisition system

[0051] S11: Define data specifications: Specify the required raw signal, sampling frequency, and data packet format.

[0052] S12: Deploy data acquisition terminals:

[0053] Production line: A high-precision data acquisition card is installed on the dynamometer stand for electric bicycle motors to synchronously record data from the electric bicycle motor controller and the torque and speed sensor data of the stand.

[0054] Prototype vehicles / fleet: Deploy intelligent controllers or on-board data recorders with edge computing and data uploading capabilities.

[0055] Service outlets: Equipped with standardized portable intelligent diagnostic tools that can connect to the vehicle controller and perform standard testing procedures.

[0056] Step S2: Building a full lifecycle data warehouse

[0057] S21: Multi-source data aggregation: Establish data pipelines to receive data streams from production line testing stations, road test fleets, and after-sales service outlets.

[0058] S22: Data Association and Labeling

[0059] Create a unique electronic file for each electric bicycle motor, linking it to its production batch, model, and serial number.

[0060] Key annotations for maintenance data: Associate maintenance work orders (such as "replace the right bearing of the electric motor of an electric bicycle") with operating data for a period of time before and after the failure to form "fault-data" sample pairs.

[0061] The final status of the motors used in scrapped electric bicycles is marked, and their total operating mileage / time and final failure mode are recorded.

[0062] Step S3: Construction and Validation of Health Indicators

[0063] S31: Feature candidate pool generation: Based on domain knowledge, a large number of potential degradation-related features (such as efficiency, torque ripple, specific subharmonic amplitude, normalized current, etc.) are calculated from the original data.

[0064] S32: Extraction of Key Health Indicators

[0065] Principal component analysis or autoencoder is used to reduce the dimensionality of high-dimensional features, and the first principal component is extracted as the comprehensive health index HI(t).

[0066] Alternatively, monotonicity and correlation indicators can be used for screening, and the single physical characteristic (such as "efficiency value under rated current") with the strongest monotonicity with operating time / mileage and the most relevant to known faults can be selected as HI(t).

[0067] S33: Indicator Verification: Plot the curve of HI(t) as a function of mileage / time to verify that it generally exhibits a monotonically decreasing trend and is aligned with known major maintenance events at specific time points.

[0068] Step S4: State-space model construction and parameter initialization

[0069] S41: Model Selection and Definition

[0070] State equations: used to describe hidden degenerate states. In one possible implementation, the evolution of the state hides the degenerate state. Represented as: ,in This represents the cumulative mileage (or running time) during the period from the (k-1)th to the kth observation.

[0071] Observational Equations (Measurement Models): Establishing Health Indicators With hidden state In one possible implementation, the relationship between health indicators With hidden state The relationship can be represented as: Typically, it is assumed that A=1, meaning the observed value equals the state value plus noise.

[0072] Initial parameter estimation:

[0073] Using a complete HI sequence of electric motors of the same model from new to scrap in historical data, the model hyperparameters are initially estimated using maximum likelihood estimation or Bayesian estimation: drift coefficient λ and process noise variance. Measurement noise variance and initial state distribution .

[0074] Set a failure threshold based on engineering experience or statistical analysis. .when If this occurs, the motor is deemed to have failed.

[0075] Step S5: Online Data Preprocessing and Feature Calculation

[0076] Real-time reception of timing data streams from the motor of a single electric bicycle.

[0077] Aggregate data by fixed time windows (e.g., daily) or mileage windows (e.g., every 100 kilometers).

[0078] Within each window, the same feature calculation process as in the offline phase is executed to obtain the health index z for the current window. k .

[0079] Step S6: Real-time state estimation (filtering)

[0080] Initialization: For the electric motor of the new electric bicycle, use offline estimation. Beliefs as the initial state.

[0081] Recursive Bayesian filtering (taking Kalman filtering as an example):

[0082] Prediction step: Based on the state equation, predict the prior state distribution for the next time step.

[0083] Update step: Obtain new observations Then, calculate the Kalman gain. Update the posterior state estimate.

[0084] Output: The optimal estimate H{k|k} of the hidden degenerate state at the current moment and its uncertainty P{k|k}.

[0085] Step S7: Probability Prediction of Remaining Useful Life

[0086] RUL distribution calculation: based on the currently estimated state Based on the state equations, the state first exceeds the threshold after calculating the remaining mileage l. The probability of.

[0087] RUL Point Estimation and Intervals: Output the Expected Value of RUL As a prediction of lifespan, a confidence interval (such as the upper and lower bounds of a 90% confidence level) is also given.

[0088] Health Score: Maps the predicted RUL to a health score of 0-100: ,in This represents the typical design life of the electric motor used in this model of electric bicycle.

[0089] Step S8: Early Warning and Decision Support

[0090] Tiered early warning:

[0091] Attention level: When If the predicted lower bound of RUL is lower than a certain threshold, a message will be displayed indicating "performance degradation, attention recommended".

[0092] Warning level: When or When the value is below the short-term threshold (as of the next maintenance cycle), a "recommended planned inspection" warning will be issued.

[0093] Alert level: When Or, if a sudden change in status is detected, an alarm will be issued stating "High risk, immediate repair recommended".

[0094] Maintenance suggestion generation: Based on the output of the fault classification model (such as "the current feature is 70% similar to the bearing fault mode"), the warning information includes preliminary fault probability and maintenance suggestions.

[0095] Step S9: Online Model Learning and Adaptation

[0096] Online parameter updates: Once enough new observation data has been accumulated for the electric motor of the electric bicycle, the model parameters of that individual motor are dynamically updated using sliding window maximum likelihood estimation or Bayesian online learning methods. This enables personalized predictions.

[0097] Threshold dynamic calibration: Based on the actual failure data of the same batch of electric bicycle motors, statistically analyze the distribution of group failure thresholds and dynamically calibrate. This makes it more in line with reality.

[0098] Step S10: Closed-loop verification and knowledge base update

[0099] Prediction-Result Comparison: When the electric motor of an electric bicycle is finally repaired or scrapped, the actual failure time / mileage is compared with the historical prediction curve to calculate the prediction error.

[0100] Model performance evaluation: Periodically analyze the prediction error indicators (such as root mean square error and mean absolute percentage error) of all electric motors used in electric bicycles to evaluate the overall performance of the model.

[0101] Knowledge base enhancement: The complete "run-prediction-failure" data chain is stored as a new high-quality labeled sample in the historical degradation library. This data is used to retrain the health indicator extraction model and optimize the initial model parameters, forming a closed loop of continuous improvement.

[0102] Example 2:

[0103] In this embodiment, a degradation state filtering lifetime prediction model is used to construct the second prediction and evaluation model. (See [link]). Figure 1 The flowchart shown is a comprehensive testing method for electric motors used in electric bicycles, including the following steps:

[0104] By communicating with the electric bicycle motor controller, real-time timing data of the electric bicycle motor under operating and testing conditions can be obtained;

[0105] Process real-time time-series data to extract multi-dimensional feature vectors that characterize the performance status of electric motors used in electric bicycles;

[0106] By inputting multi-dimensional feature vectors into the preset first prediction and evaluation model, a real-time health status quantitative score and fault risk classification of the electric motor for electric bicycles are obtained.

[0107] A second prediction and evaluation model is constructed based on multi-dimensional feature vectors and historical state sequences to estimate the remaining service life of electric motors for electric bicycles.

[0108] Based on the combined results of the first and second prediction and evaluation models, a comprehensive performance prediction for the electric motor used in electric bicycles is obtained.

[0109] The system receives maintenance result feedback data from the maintenance terminal, uses the fault feature vector and the confirmed fault category as new labeled samples, and performs incremental learning and optimization on the classification model.

[0110] Cluster analysis is performed on the factory inspection feature vectors of multiple electric bicycle motors from the same production batch or model to generate a quality distribution report and identify abnormal individuals that deviate from the standard group;

[0111] By aggregating the characteristic data of the entire life cycle of the electric motors for the same model of electric bicycle, fitting the common degradation trajectory of their key characteristics, and establishing a standard healthy degradation baseline for the electric motors of this model of electric bicycle.

[0112] The following section provides a further explanation of this solution, detailing the specific implementation process, which includes the following steps:

[0113] Step S1: Standardization of the data acquisition system

[0114] S11: Define data specifications: Clearly define the required raw signal, sampling frequency, and data packet format.

[0115] S12: Deploy data acquisition terminals:

[0116] Production line: A high-precision data acquisition card is installed on the dynamometer stand for electric bicycle motors to synchronously record data from the electric bicycle motor controller and the torque and speed sensors on the stand.

[0117] Prototype vehicles / fleet: Deploy intelligent controllers or on-board data recorders with edge computing and data uploading capabilities.

[0118] Service outlets: Equipped with standardized portable intelligent diagnostic tools that can connect to the vehicle controller and perform standard testing procedures.

[0119] Step S2: Building a full lifecycle data warehouse

[0120] S21: Multi-source data aggregation: Establish data pipelines to receive data streams from production line testing stations, road test fleets, and after-sales service outlets.

[0121] S22: Data Association and Labeling

[0122] Create a unique electronic file for each electric bicycle motor, linking it to its production batch, model, and serial number.

[0123] Key annotations for maintenance data: Associate maintenance work orders (such as "replace the right bearing") with operating data for a period of time before and after the failure to form "failure-data" sample pairs.

[0124] The final status of the motors used in scrapped electric bicycles is marked, and their total operating mileage / time and final failure mode are recorded.

[0125] Step S3: Construction and Validation of Health Indicators

[0126] S31: Feature candidate pool generation: Based on domain knowledge, a large number of potential degradation-related features (such as efficiency, torque ripple, specific subharmonic amplitude, normalized current, etc.) are calculated from the original data.

[0127] S32: Extraction of Key Health Indicators

[0128] Principal component analysis or autoencoder is used to reduce the dimensionality of high-dimensional features, and the first principal component is extracted as the comprehensive health index HI(t).

[0129] Alternatively, monotonicity and correlation indicators can be used for screening, and the single physical characteristic (such as "efficiency value under rated current") with the strongest monotonicity with operating time / mileage and the most relevant to known faults can be selected as HI(t).

[0130] S33: Indicator Verification: Plot the curve of HI(t) as a function of mileage / time to verify that it generally exhibits a monotonically decreasing trend and is aligned with known major maintenance events at specific time points.

[0131] Step S4: Build a historical degradation library

[0132] Collect a large number of complete HI(t) curves of electric motors used in similar electric bicycles from brand new to failure, denoted as , where τ is the relative lifetime (from 0 to 1);

[0133] The currently observed health indicator sequence Perform similarity matching with a pre-stored library of complete historical degradation curves, and calculate the dynamic time-normalized distance or Euclidean distance;

[0134] Step S5: RUL Estimation

[0135] S51: Based on the failure times of several most similar historical curves, estimate the remaining useful life of the current electric bicycle motor using a weighted average. Specifically:

[0136] S52: Suppose we find the k most similar historical curves, whose total lifetimes are respectively... The current running time is The predicted RUL can then be estimated using the following formula: ,in Let be the weighting coefficient of the j-th historical curve, and .

[0137] S53: Health Score: Maps the predicted RUL to a health score of 0-100. ,in This represents the typical design life of the electric motor used in this model of electric bicycle.

[0138] Step S6: Early Warning and Decision Support

[0139] Tiered early warning:

[0140] Attention level: When If the predicted lower bound of RUL is lower than a certain threshold, a message will be displayed indicating "performance degradation, attention recommended".

[0141] Warning level: When or When the value is below the short-term threshold (as of the next maintenance cycle), a "recommended planned inspection" warning will be issued.

[0142] Alert level: When Or, if a sudden change in status is detected, an alarm will be issued stating "High risk, immediate repair recommended".

[0143] Maintenance suggestion generation: Based on the output of the fault classification model (such as "the current feature is 70% similar to the bearing fault mode"), the warning information includes preliminary fault probability and maintenance suggestions.

[0144] Step S7: Online Model Learning and Adaptation

[0145] Online parameter update: Once enough new observation data has been accumulated for the electric motor of the electric bicycle, the model parameters of the individual electric motor are dynamically updated using sliding window maximum likelihood estimation or Bayesian online learning methods. This enables personalized predictions.

[0146] Threshold dynamic calibration: Based on the actual failure data of the same batch of electric bicycle motors, statistically analyze the distribution of group failure thresholds and dynamically calibrate. This makes it more in line with reality.

[0147] Step S8: Closed-loop verification and knowledge base update

[0148] Prediction-Result Comparison: When the electric motor of an electric bicycle is finally repaired or scrapped, the actual failure time / mileage is compared with the historical prediction curve to calculate the prediction error.

[0149] Model performance evaluation: Periodically analyze the prediction error indicators (such as root mean square error and mean absolute percentage error) of all electric motors used in electric bicycles to evaluate the overall performance of the model.

[0150] Knowledge base enhancement: The complete "run-prediction-failure" data chain is stored as a new high-quality labeled sample in the historical degradation library. This data is used to retrain the health indicator extraction model and optimize the initial model parameters, forming a closed loop of continuous improvement.

[0151] Example 3:

[0152] This invention also provides a comprehensive testing system for electric motors used in electric bicycles, see [link to relevant documentation]. Figure 2 The system architecture diagram shown includes:

[0153] The data acquisition module collects and uploads real-time timing data of the electric motor of the electric bicycle during operation or testing.

[0154] The feature vector extraction module, connected to the data acquisition module, is used to process real-time time series data to extract multi-dimensional feature vectors that characterize the performance status of the electric motor in electric bicycles.

[0155] The first prediction and evaluation model construction module is connected to the feature vector extraction module. It is used to obtain the real-time health status quantitative score and fault risk classification of the electric motor of electric bicycle by taking multi-dimensional feature vectors as input.

[0156] The second prediction and evaluation model construction module, connected to the feature vector extraction module, is used to construct the second prediction and evaluation model based on multi-dimensional feature vectors and historical state sequences to estimate the remaining service life of the electric motor for electric bicycles.

[0157] The detection result output module is connected to the first prediction and evaluation model construction module and the second prediction and evaluation model construction module to output the comprehensive performance prediction of the electric motor for electric bicycles.

[0158] The system also includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor, when executing the program, is able to implement a comprehensive testing method for electric motors used in electric bicycles.

[0159] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A comprehensive testing method for electric motors used in electric bicycles, characterized in that, The method includes: Real-time time-series data of the electric motor for electric bicycles under operating and testing conditions are acquired; the real-time time-series data is processed to extract multi-dimensional feature vectors characterizing the performance status of the electric motor for electric bicycles; the multi-dimensional feature vectors are input into a first prediction and evaluation model to obtain a quantitative score of the real-time health status and a fault risk classification of the electric motor for electric bicycles; a second prediction and evaluation model is constructed based on the multi-dimensional feature vectors and historical state sequences to estimate the remaining service life of the electric motor for electric bicycles; based on the combined results of the first and second prediction and evaluation models, a comprehensive performance prediction of the electric motor for electric bicycles is obtained. The real-time time series data includes the raw data of electrical operating parameters, thermodynamic state parameters, and external operating conditions and environmental parameters required to establish the first and second predictive evaluation models; The multi-dimensional feature vector includes data features extracted from the original data; The first predictive assessment model includes a health benchmark modeling model and a classification model based on supervised learning; The second prediction and assessment model includes a health indicator trajectory matching lifespan prediction model and a degradation state filtering prediction model; The first prediction and evaluation model is constructed using the following method: First, a normal operating condition benchmark is constructed by using the historical feature vectors of the electric motor of an electric bicycle under known health conditions through a health benchmark modeling model. Subsequently, the supervised learning-based classification model is trained using a multi-layer nonlinear transformation model based on the feature vector and its corresponding state label to obtain a classifier that can map the input feature vector to a specific health state or fault type. The second prediction and evaluation model includes a health indicator trajectory matching life prediction model and a degradation state filtering life prediction model; The health indicator trajectory matching life prediction model includes: a health indicator extraction unit for generating a one-dimensional health indicator based on multi-dimensional feature vectors; a trajectory matching unit for calculating the similarity distance between the real-time sequence of the one-dimensional health indicator and the curves in the historical degradation curve library; and a life prediction unit for estimating the remaining useful life of the electric motor for electric bicycles based on the similarity matching results through weighted calculation. The degradation state filtering lifetime prediction model includes: a state space model construction unit, used to establish a degradation probability model of the electric motor for electric bicycles containing state equations and observation equations; an online state estimation unit, used to recursively estimate the posterior distribution of the health state based on Kalman filtering or particle filtering algorithms; and a lifetime prediction unit, used to simulate future degradation paths and output the probability distribution of the remaining lifetime of the electric motor for electric bicycles based on state equations and failure thresholds. The method further includes: The system receives maintenance result feedback data from the maintenance terminal, uses the fault feature vector and the confirmed fault category as new labeled samples, and performs incremental learning and optimization on the supervised learning-based classification model. The method further includes: Cluster analysis is performed on the factory inspection feature vectors of multiple electric bicycle motors from the same production batch or model to generate a quality distribution report and identify abnormal individuals that deviate from the standard group; By aggregating the characteristic data of the entire life cycle of the electric motors for the same model of electric bicycle, fitting the common degradation trajectory of their key characteristics, and establishing a standard healthy degradation baseline for the electric motors of this model of electric bicycle.

2. A comprehensive testing system for electric motors used in electric bicycles, used to implement the method as described in claim 1, characterized in that, include: The data acquisition module collects and uploads real-time timing data of the electric motor of the electric bicycle during operation or testing. The feature vector extraction module, connected to the data acquisition module, is used to process the real-time time series data to extract multi-dimensional feature vectors characterizing the performance status of the electric motor of the electric bicycle. The first prediction and evaluation model construction module is connected to the feature vector extraction module and is used to obtain a real-time health status quantitative score and fault risk classification of the electric motor of the electric bicycle by taking the multi-dimensional feature vector as input. The second prediction and evaluation model construction module is connected to the feature vector extraction module and is used to construct a second prediction and evaluation model based on the multi-dimensional feature vector and historical state sequence to estimate the remaining service life of the electric motor for electric bicycles. The detection result output module is connected to the first prediction and evaluation model construction module and the second prediction and evaluation model construction module to output the comprehensive performance prediction of the electric motor for electric bicycles.

3. The comprehensive testing system for electric motors of electric bicycles according to claim 2, characterized in that, The system further includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, is able to implement the comprehensive testing method for an electric motor of an electric bicycle as described in claim 1.

Images