Brushless motor performance test method, apparatus and device, and storage medium

Abstract

The invention relates to the technical field of motor testing, in particular to a brushless motor performance testing method, device and equipment and a storage medium, and the method comprises the steps: S1, collecting a voltage signal, a current signal, a rotating speed signal and a temperature signal during the operation of a brushless motor through a multi-channel sensor, and generating an original performance data set; s2, performing working condition feature extraction and motor type identification processing on the original performance data set based on a machine learning algorithm to generate working condition feature identification data and motor type identification data; by dynamically identifying the motor types and real-time working condition characteristics and combining the intelligent matching strategy library to generate the optimal test scheme, the problem that the prior art cannot adapt to variable working conditions is solved, it is ensured that different types of motors can obtain accurate performance evaluation under different load conditions, and the test adaptability and accuracy are improved.

Description

Technical Field

[0001] This application relates to the field of motor testing technology, and in particular to methods, apparatus, equipment and storage media for testing the performance of brushless motors. Background Technology

[0002] A brushless motor is a typical mechatronic product, consisting of a motor body and a driver. Because brushless DC motors operate in a self-controlled manner, they do not require an additional starting winding on the rotor like synchronous motors that start under heavy loads with frequency conversion speed regulation, nor do they experience oscillations or loss of synchronism during sudden load changes.

[0003] In the field of brushless motor performance testing, existing technologies generally adopt static preset parameter testing schemes, which result in a lack of dynamic adaptability to changing operating conditions during the testing process. When conducting brushless motor performance testing, traditional methods cannot identify changes in motor type and load status in real time, ignore the potential impact of environmental temperature and humidity interference on test data, and the large fluctuations in light, temperature and electromagnetic noise during the testing process may cause test results to be distorted, making it impossible to guarantee the accuracy and reliability of performance evaluation.

[0004] Therefore, this application provides a method, apparatus, equipment, and storage medium for testing the performance of brushless motors. Summary of the Invention

[0005] The purpose of this application is to solve at least one technical problem raised in the background art.

[0006] This application provides a method, apparatus, equipment, and storage medium for testing the performance of brushless motors, including the following steps: S1. Collect voltage, current, speed and temperature signals of the brushless motor during operation using multi-channel sensors to generate raw performance datasets; S2. Based on machine learning algorithms, the original performance dataset is processed to extract operating condition features and identify motor types, generating operating condition feature identification data and motor type identification data. S3. Based on the operating condition feature identification data and motor type identification data, and combined with the preset test strategy library, perform dynamic test strategy matching processing to generate target test strategy type feature data. S4. Based on the target test strategy type feature data, control the servo loading system to simulate the actual working conditions, perform dynamic performance testing on the brushless motor, and generate real-time test waveform data. S5. Real-time monitoring of temperature and humidity parameters is achieved through the environmental compensation module, and environmental interference errors in the real-time test waveform data are corrected to generate calibration test data. S6. Based on statistical analysis models and signal processing algorithms, the calibration test data is analyzed for performance indicators to generate a brushless motor performance evaluation report.

[0007] Preferably, S1 includes: S11. The voltage and current signals of the three-phase windings of the brushless motor are synchronously acquired through a 24-bit high-precision ADC analog-to-digital converter at a sampling frequency of not less than 100kHz to ensure signal fidelity. S12. The rotational speed signal of the motor rotor is acquired through a Hall effect sensor, and the temperature distribution signal of key parts of the motor housing is acquired with a resolution of 0.1℃ through a K-type thermocouple. S13. Align and integrate the voltage signal, current signal, speed signal and temperature signal according to millisecond-level timestamps to construct a multidimensional original performance dataset containing spatiotemporal correlation features.

[0008] Preferably, S2 includes: S21. Input the original performance dataset into the pre-trained ResNet-34 convolutional neural network model. The training process of the model includes: Data samples were extracted from the motor test database and divided into training and validation sets in a 7:3 ratio. The network structure contains 34 convolutional layers, each with a kernel size of 3×3, and uses ReLU as the activation function. The training parameters use the cross-entropy loss function and are iterated through the Adam optimizer for 50 rounds until the validation set loss value converges to below 0.01; S22. Based on the extracted feature vectors, perform cosine similarity matching with a database containing 500+ motor types, and output motor type identification data with confidence scores; S23. Input the preprocessed data into the Markov decision model, and dynamically update the state transition probabilities of the model using the following formula: Where s t Let s be the current working state vector. t+1 Let a be the working state vector at the next moment. t The current test strategy action is N(·), which represents the state transition frequency statistics, λ = 0.35 is the reward correction coefficient, and V(s) is the value of V(s). t+1 ) represents state s t+1 Long-term value assessment.

[0009] Preferably, S3 includes: S31. Establish a test strategy library matrix covering 12 industrial scenarios: Q = (q1, ..., q) k ,…,q 12 ), k = 1, 2, ..., 12; Where q kThis represents the standard operating condition feature combination data corresponding to the k-th test strategy type, specifically including the following typical scenario definitions: q1: Elevator traction machine start-stop operating condition Load mode: 0→150Nm step load; Temperature rise threshold: Temperature rise rate of critical parts ≤ 0.8℃ / s; Vibration limit: Axial vibration energy density ≤ 0.3 J; q2: Electric vehicle driving conditions Load mode: 10-50Hz sinusoidal ripple load; Temperature rise threshold: winding temperature rise ≤ 95℃; Vibration limit: Radial vibration spectrum peak-to-peak value ≤ 0.5 J; Q3: Unloading of UAV motors Load mode: Load is suddenly unloaded from 100% to 10% within 200ms; Temperature rise threshold: shell temperature hysteresis ≥ 5℃ / min; Vibration limit: High-frequency resonant band energy ≤ 0.1 J; Q4: Operating conditions of joint motors in industrial robots Load mode: periodic point-to-point motion load, repeatability accuracy ±0.01Nm; temperature rise threshold: joint unit temperature rise ≤65℃; Vibration limit: End-effector jitter amplitude ≤ 50 μm; q5: Compressor drive motor operating condition Load mode: Constant torque start, load inertia moment 0.12 kg·m 2 ; Temperature rise threshold: Exhaust port temperature rise ≤ 110℃; Vibration limit: Effective value of fundamental frequency vibration velocity ≤ 2.8 mm / s; Q6: Operating conditions of CNC machine tool spindle motor Load mode: High-speed constant power cutting load; Temperature rise threshold: Spindle bearing temperature rise ≤ 70℃; Vibration limit: Dynamic balance grade G1.0; q7: Operating conditions of motors in fan and pump loads Load mode: Square torque characteristic, adjustable start-up time; Temperature rise threshold: Temperature rise of insulated winding ≤ 80℃; Vibration limit: The blade passes through a vibration frequency ≤ 4.0 mm / s; q8: Conveyor motor operating conditions Load mode: constant torque start / stop, frequent forward and reverse rotation; Temperature rise threshold: Gearbox oil temperature rise ≤ 50℃; Vibration limits: Start-stop shock vibration ≤1.0G; Q9: Winding machine tension control motor operating conditions Load mode: constant tension control, linear acceleration and deceleration; Temperature rise threshold: Brake pad temperature rise ≤ 200℃; Vibration limit: Tension fluctuation frequency suppression ≤ ±0.5Hz; q 10 Servo positioning motor operating conditions Load mode: S-curve acceleration and deceleration load, positioning accuracy ±1 pulse; Temperature rise threshold: Encoder temperature rise ≤ 55℃; Vibration limit: No overshoot within the set time; q 11 Operating conditions of brushless motors for home appliances Load mode: Light load, high speed operation, with minimal load fluctuation; Temperature rise threshold: outer casing temperature rise ≤ 40℃; Vibration limits: Sound pressure level ≤ 45 dB; q 12 Heavy-duty crane motor operating conditions; Load mode: Heavy load, low speed start, overload capacity 200%; Temperature rise threshold: winding temperature rise ≤ 125℃; Vibration limits: Avoidance of structural resonant frequencies; S32. Iteratively optimize the matching path using the Q-learning reinforcement learning algorithm, the implementation of which includes: Historical operating condition transfer records are obtained from the Industrial Internet of Things (IIoT) platform. Each record contains a state vector, an action vector, and a reward value. The training set and the validation set are divided in an 8:2 ratio.

[0010] The network structure adopts a fully connected neural network, which includes one input layer, two hidden layers, and one output layer.

[0011] The training parameters used the mean squared error loss function and were iterated through the Adam optimizer for 3000 rounds with a batch size of 64 until the validation set loss value converged to below 0.0001.

[0012] Preferably, S4 includes: S41. Based on the loading mode parameters in the target test strategy type characteristic data, control the magnetic powder brake servo system to generate adjustable step load, ramp load and 0-100Hz sinusoidal ripple load. S42. During dynamic loading, the motor response waveform is acquired synchronously to generate real-time test waveform data including electromagnetic noise spectrum, vibration spectrum characteristics and efficiency-speed curve. The electromagnetic noise power spectral density is calculated using the following formula: Where P noise (f) is the noise power spectral density at frequency f, and its unit is A. 2 / Hz, T is the sampling time, and X(f) is the complex spectrum value of the current signal at frequency f after FFT transformation.

[0013] Preferably, S5 includes: S51. Real-time monitoring of test environment parameters using SHT35 temperature and humidity sensor at 1Hz frequency, generating an environmental compensation coefficient matrix including temperature compensation coefficient and humidity drift coefficient. S52. Using the weighted least squares method, the environmental compensation coefficient matrix is ​​fitted with the real-time test waveform data in a multidimensional error manner, and the calibration test data eliminating environmental interference is output: Where y i For the original test data of the i-th sampling point, T i H represents the temperature measurement value at the corresponding point. i Here are the humidity measurements at the corresponding points, β0, β1, and β2 are the compensation coefficients to be solved, and w i This is a weighting factor based on the signal-to-noise ratio, with a value ranging from 0 to 1.

[0014] Preferably, the device includes: The signal acquisition module is equipped with a 24-bit Δ-Σ architecture ADC and a 16-channel synchronous acquisition card, which is suitable for high-fidelity acquisition of voltage, current, speed and temperature signals; The operating condition identification module integrates a pre-trained machine learning model and a real-time inference engine to extract dynamic operating condition features and identify the motor model. The strategy selection module stores a multi-dimensional test strategy library and dynamically matches the optimal test plan through reinforcement learning algorithms. The test control module connects to a 500W servo loading system and is suitable for performing complex working condition simulations. The environmental compensation module has a built-in industrial-grade temperature and humidity sensor and an adaptive correction algorithm. The data analysis module uses wavelet transform and regression analysis techniques to analyze performance metrics.

[0015] Preferably, the environmental compensation module further includes: A high-frequency vibration sensor with a 100kHz sampling rate collects the spectral characteristics of mechanical vibration under load conditions; Second-order Butterworth electromagnetic interference filter to eliminate high-frequency noise above 50kHz in current signals; The real-time temperature drift correction unit based on PT100 platinum resistance dynamically adjusts the signal gain coefficient according to the temperature change rate of 0.1℃ / s.

[0016] Preferably, the equipment includes: The system comprises a quad-core ARM Cortex-A72 processor, 4GB of LPDDR4 memory, and a computer program stored in eMMC memory. When the computer program is executed by the processor, it implements the brushless motor performance testing method and supports communication with the industrial IoT platform via CAN bus.

[0017] Preferably, the storage medium specifically includes: The data acquisition thread acquires multi-channel sensor signals in real time and generates raw performance datasets with timestamps. Call the pre-built machine learning model library to execute the working condition feature extraction and motor type identification algorithm; Dynamically match test strategies and generate servo loading control instructions based on a reinforcement learning engine; Initiate an environmental compensation thread to perform online correction for temperature and humidity interference; Finally, a standard-compliant PDF test report is generated and uploaded to the cloud-based quality monitoring platform via an Ethernet interface.

[0018] In summary, this application includes at least one of the following beneficial technical effects: 1. The brushless motor performance testing method, apparatus, equipment, and storage medium described in this application solve the problem that existing technologies cannot adapt to changing operating conditions by dynamically identifying motor type and real-time operating condition characteristics and generating the optimal test scheme by combining an intelligent matching strategy library. This ensures that different types of motors can obtain accurate performance evaluation under different load conditions, and improves the adaptability and accuracy of the test.

[0019] 2. The brushless motor performance testing method, apparatus, equipment, and storage medium described in this application solve the problem of distorted results caused by neglecting environmental factors in traditional methods by real-time compensation for environmental temperature and humidity interference and correction of test data errors, eliminating the influence of temperature drift and electromagnetic noise, and ensuring the reliability and consistency of test results in complex industrial environments.

[0020] 3. The brushless motor performance testing method, device, equipment, and storage medium described in this application solve the problem of the one-sidedness of single index evaluation by using multi-dimensional feature fusion analysis of current harmonics, vibration spectrum, temperature gradient, and dynamic learning mechanism. This enables comprehensive analysis of the motor's electromagnetic performance, mechanical stability, and thermal management efficiency, thereby improving fault early warning capabilities and the level of quality monitoring throughout the product lifecycle. Attached Figure Description

[0021] Figure 1 This is a flowchart of the brushless motor performance testing method of the present invention. Detailed Implementation

[0022] The following is in conjunction with the appendix Figure 1 This application will be described in further detail below.

[0023] Example: Please refer to the following carefully. Figure 1 The brushless motor performance testing method, apparatus, equipment, and storage medium include the following steps: S1. Collect voltage, current, speed and temperature signals of the brushless motor during operation using multi-channel sensors to generate raw performance datasets; S2. Based on machine learning algorithms, extract working condition features and identify motor types from the original performance dataset to generate working condition feature identification data and motor type identification data. S3. Based on the operating condition characteristic identification data and motor type identification data, and combined with the preset test strategy library, perform dynamic test strategy matching processing to generate target test strategy type characteristic data. S4. Based on the target test strategy type characteristic data, control the servo loading system to simulate the actual working conditions, perform dynamic performance testing on the brushless motor, and generate real-time test waveform data. S5. Real-time monitoring of temperature and humidity parameters and correction of environmental interference errors in real-time test waveform data through the environmental compensation module to generate calibration test data; S6. Based on statistical analysis models and signal processing algorithms, the calibration test data is analyzed and processed to generate a brushless motor performance evaluation report.

[0024] S1 includes: S11. The voltage and current signals of the three-phase windings of the brushless motor are synchronously acquired through a 24-bit high-precision ADC analog-to-digital converter at a sampling frequency of not less than 100kHz to ensure signal fidelity. S12. The rotational speed signal of the motor rotor is acquired through a Hall effect sensor, and the temperature distribution signal of key parts of the motor housing is acquired with a resolution of 0.1℃ through a K-type thermocouple. S13. Align and integrate the voltage signal, current signal, speed signal and temperature signal according to millisecond-level timestamps to construct a multidimensional original performance dataset containing spatiotemporal correlation features; Signal synchronization calibration formula: The spatiotemporal correlation feature is defined as the dynamic characteristics of mutual correlation and mutual coupling exhibited by sensor signals from different physical quantities of the motor under the precise synchronization relationship of the same, specific, and minute time interval.

[0025] Specifically, this manifests as: the linkage between the sudden change in winding current and the instantaneous response of rotor speed and the temperature change trend at a specific location at the same time point; and the causal or synergistic relationship between the voltage fluctuation pattern caused by load changes and the temperature accumulation effect and vibration spectrum evolution at different time points. This multidimensional signal fusion under precise time alignment captures the dynamic coupling effect of the electro-magnetic-thermal-mechanical multi-physics fields in the actual operation of the motor, providing a high-dimensional data foundation with inherent correlations for subsequent operating condition identification and performance analysis.

[0026] Where τ sync The signal synchronization error is N, where N is the total number of sampling points. For ADC data acquisition timestamps, Here is the Hall sensor timestamp, and SR is the sampling rate correction factor; S2 includes: S21. First, the multi-channel time-series signals in the original performance dataset are converted into a time-frequency graph feature matrix through short-time Fourier transform. Then, the time-frequency graph is input into a pre-trained ResNet-34 convolutional neural network model. The training process of the model includes: Data samples were extracted from the motor test database and divided into training and validation sets in a 7:3 ratio. The network structure contains 34 convolutional layers, each with a kernel size of 3×3, and uses ReLU as the activation function. The training parameters use the cross-entropy loss function and are iterated through the Adam optimizer for 50 rounds until the validation set loss value converges to below 0.01; S22. Based on the extracted feature vectors, perform cosine similarity matching with a database containing 500+ motor types, and output motor type identification data with confidence scores: Where S match For model matching similarity, F k For real-time feature vectors, D k Here, m represents the feature vectors in the database, and m is the number of feature dimensions. S23. Input the preprocessed data into the Markov decision model, and dynamically update the model's state transition probabilities using the following formula: Where s t Let s be the current working state vector. t +1 represents the working condition vector at the next moment, a t The current test strategy action is N(·), which represents the state transition frequency statistics, λ = 0.35 is the reward correction coefficient, and V(s) is the value of V(s). t +1) represents state s tThe long-term value assessment of +1, where s′∈S is the state space.

[0027] S3 includes: S31. Establish a test strategy library matrix covering 12 industrial scenarios: Q = (q1, ..., q) k ,…,q 12 ), k = 1, 2, ..., 12; Where q k This represents the standard operating condition feature combination data corresponding to the kth test strategy type, specifically including the following typical scenario definitions: Q1: Elevator traction machine start-stop operation status Load mode: 0→150Nm step load; Temperature rise threshold: Temperature rise rate of critical parts ≤ 0.8℃ / s; Vibration limit: Axial vibration energy density ≤ 0.3 J; q2: Electric vehicle driving conditions Load mode: 10-50Hz sinusoidal ripple load; Temperature rise threshold: winding temperature rise ≤ 95℃; Vibration limit: Radial vibration spectrum peak-to-peak value ≤ 0.5 J; Q3: Unloading of UAV motors Load mode: Load is suddenly unloaded from 100% to 10% within 200ms; Temperature rise threshold: shell temperature hysteresis ≥ 5℃ / min; Vibration limit: High-frequency resonant band energy ≤ 0.1 J; Q4: Operating conditions of joint motors in industrial robots Load mode: periodic point-to-point motion load, repeatability positioning accuracy ±0.01Nm; Temperature rise threshold: Joint unit temperature rise ≤ 65℃; Vibration limit: End-effector jitter amplitude ≤ 50 μm; q5: Compressor drive motor operating condition Load mode: Constant torque start, load inertia moment 0.12 kg·m 2 ; Temperature rise threshold: Exhaust port temperature rise ≤ 110℃; Vibration limit: Effective value of fundamental frequency vibration velocity ≤ 2.8 mm / s; Q6: Operating conditions of CNC machine tool spindle motor Load mode: High-speed constant power cutting load; Temperature rise threshold: Spindle bearing temperature rise ≤ 70℃; Vibration limit: Dynamic balance grade G1.0; q7: Operating conditions of motors in fan and pump loads Load mode: Square torque characteristic, adjustable start-up time; Temperature rise threshold: Temperature rise of insulated winding ≤ 80℃; Vibration limit: The blade passes through a vibration frequency ≤ 4.0 mm / s; q8: Conveyor motor operating conditions Load mode: constant torque start / stop, frequent forward and reverse rotation; Temperature rise threshold: Gearbox oil temperature rise ≤ 50℃; Vibration limits: Start-stop shock vibration ≤1.0G; Q9: Winding machine tension control motor operating conditions Load mode: constant tension control, linear acceleration and deceleration; Temperature rise threshold: Brake pad temperature rise ≤ 200℃; Vibration limit: Tension fluctuation frequency suppression ≤ ±0.5Hz; q 10 Servo positioning motor operating conditions Load mode: S-curve acceleration and deceleration load, positioning accuracy ±1 pulse; Temperature rise threshold: Encoder temperature rise ≤ 55℃; Vibration limit: No overshoot within the set time; q 11 Operating conditions of brushless motors for home appliances Load mode: Light load, high speed operation, with minimal load fluctuation; Temperature rise threshold: outer casing temperature rise ≤ 40℃; Vibration limits: Sound pressure level ≤ 45 dB; q 12 Heavy-duty crane motor operating conditions; Load mode: Heavy load, low speed start, overload capacity 200%; Temperature rise threshold: winding temperature rise ≤ 125℃; Vibration limits: Avoidance of structural resonant frequencies; S32. Iteratively optimize the matching path using the Q-learning reinforcement learning algorithm. The algorithm implementation includes: Historical operating condition transfer records are obtained from the Industrial Internet of Things (IIoT) platform. Each record contains a state vector, an action vector, and a reward value. The training set and the validation set are divided in an 8:2 ratio.

[0028] The network structure adopts a fully connected neural network, which includes one input layer, two hidden layers, and one output layer.

[0029] The training parameters used the mean squared error loss function and were iterated through the Adam optimizer for 3000 rounds with a batch size of 64 until the validation set loss value converged to below 0.0001.

[0030] Dynamic test strategy matching processing is defined as follows: The real-time identified operating condition feature data is compared with each standard strategy q in the test strategy library matrix Q. k The corresponding combination of working condition features is compared for feature similarity. Based on this, the Q-learning reinforcement learning algorithm is applied to simulate the decision-making process of an agent in a given state to select the optimal action. By iteratively optimizing the matching path, the target test strategy type feature data that best matches the current motor type and real-time operating conditions and includes parameters such as specific loading mode, test frequency, and accuracy requirements is finally output.

[0031] The core of this dynamic matching process lies in intelligently selecting and optimizing the most suitable execution plan for the current test scenario from the pre-set strategy library based on the real-time identified working conditions and motor type. S4 includes: S41. Based on the loading mode parameters in the target test strategy type characteristic data, control the magnetic powder brake servo system to generate adjustable step load, ramp load and 0-100Hz sinusoidal ripple load. S42. During dynamic loading, the motor response waveform is acquired synchronously to generate real-time test waveform data including electromagnetic noise spectrum, vibration spectrum characteristics and efficiency-speed curve. The electromagnetic noise power spectral density is calculated using the following formula: Where P noise (f) is the noise power spectral density at frequency f, and its unit is A. 2 / Hz, T is the sampling duration, ||w|| 2 For the window function energy, wpn[ represents the coefficients of the Hanning window function, n is the sampling point index, x[n] is the discrete current signal, j is the imaginary unit, and e -j2πfn / N Here, is the Fourier transform kernel function, and N is the total number of sampling points; Hanning window definition formula notes: Where wpn] is the window function coefficient of the nth sampling point and its value range is [0,1], and N-1 is the window length adjustment factor.

[0032] S5 includes: S51. Real-time monitoring of test environment parameters is achieved using a multi-channel SHT35 array with a synchronous sampling rate of 100kHz, and effective temperature and humidity feature values ​​are extracted using a sliding window mean filter. Where W = 100 is the width of the sliding window. The filtered temperature value at the i-th sampling point. The filtered humidity value at the i-th sampling point; S52. The environmental compensation coefficient matrix is ​​fitted to the real-time test waveform data using the weighted least squares method to perform multi-dimensional error fitting, and the calibration test data eliminating environmental interference is output: Where y i For the original test data of the i-th sampling point, T i H represents the temperature measurement value at the corresponding point. i Here are the humidity measurements at the corresponding points, β0, β1, and β2 are the compensation coefficients to be solved, and w i This is a weighting factor based on the signal-to-noise ratio, with a value ranging from 0 to 1.

[0033] The device includes: The signal acquisition module is equipped with a 24-bit Δ-Σ architecture ADC and a 16-channel synchronous acquisition card, which is suitable for high-fidelity acquisition of voltage, current, speed and temperature signals; The operating condition identification module integrates a pre-trained machine learning model and a real-time inference engine to extract dynamic operating condition features and identify the motor model. The strategy selection module stores a multi-dimensional test strategy library and dynamically matches the optimal test plan through reinforcement learning algorithms. The test control module connects to a 500W servo loading system and is suitable for performing complex working condition simulations. The environmental compensation module has a built-in industrial-grade temperature and humidity sensor and an adaptive correction algorithm. The data analysis module uses wavelet transform and regression analysis techniques to analyze performance metrics.

[0034] The environmental compensation module further includes: a high-frequency vibration sensor with a 100kHz sampling rate to collect the mechanical vibration spectrum characteristics under load conditions; Second-order Butterworth electromagnetic interference filter to eliminate high-frequency noise above 50kHz in current signals; The real-time temperature drift correction unit based on PT100 platinum resistance dynamically adjusts the signal gain coefficient according to the temperature change rate of 0.1℃ / s.

[0035] The equipment includes: The system consists of an NVIDIA Jetson AGX Orin 64GB module, 64GB of 256-bit LPDDR5 memory, and computer programs stored on a 1TB NVMe SSD. When the computer program is executed by the processor, it implements a method for testing the performance of brushless motors and supports communication with the industrial IoT platform via CAN bus.

[0036] The storage media specifically include: The data acquisition thread acquires multi-channel sensor signals in real time and generates raw performance datasets with timestamps. Call the pre-built machine learning model library to execute the working condition feature extraction and motor type identification algorithm; Dynamically match test strategies and generate servo loading control instructions based on a reinforcement learning engine; Initiate an environmental compensation thread to perform online correction for temperature and humidity interference; Finally, a standard-compliant PDF test report is generated and uploaded to the cloud-based quality monitoring platform via an Ethernet interface.

[0037] The operation steps for brushless motor performance testing are as follows: Step 1: Signal Acquisition and Preprocessing Stage A multi-channel sensor array, including a high-precision current clamp, a Hall effect speed sensor, and a distributed temperature probe, was installed on a high-speed test platform. After starting the brushless motor, three-phase winding current signals, rotor speed signals, and temperature distribution data of key components were simultaneously acquired. The analog signals were converted into digital signal streams using a 24-bit Δ-Σ analog-to-digital converter, and then aligned and integrated according to millisecond-level timestamps to form a spatiotemporally correlated raw performance dataset. This process ensures signal fidelity, providing a high-precision data source for subsequent analysis.

[0038] Step Two: Operating Condition Identification and Strategy Matching Stage First, the multi-channel time-series signals in the original performance dataset are converted into a time-frequency graph feature matrix using a short-time Fourier transform. Then, the time-frequency graph is input into a pre-trained ResNet-34 convolutional neural network model to extract current harmonic distortion features, voltage sag features, and temperature gradient distribution features. Based on the cosine similarity calculation between the feature vectors and the motor type database, a motor model identifier with a confidence score is generated. Combining the frequency of sudden speed changes and load variation trends, a reinforcement learning engine dynamically matches the optimal solution from the test strategy library, including selecting step, ramp, and sinusoidal load modes and precision control parameters, achieving intelligent adaptive testing strategies.

[0039] Step 3: Dynamic Testing and Environmental Compensation Phase The servo loading system generates an adjustable load of 0-200Nm to simulate real-world operating conditions, simultaneously acquiring electromagnetic noise spectrum, vibration spectrum, and efficiency curve data. The environmental compensation module monitors the test environment in real time using temperature and humidity sensors, employing a weighted least squares algorithm to eliminate temperature drift and humidity interference. A high-frequency vibration sensor captures the distribution of mechanical vibration energy, combining this with acceleration signals to analyze the bearing's health status, forming a calibrated multidimensional performance data matrix.

[0040] Step 4: Performance Analysis and Output Stage The transient response characteristics in the calibration data are decomposed using wavelet transform, and core indicators such as efficiency, torque ripple, and thermal stability are calculated using a regression model. The final test report conforming to the standards is generated, including: Electromagnetic compatibility performance level, mechanical vibration health status, dynamic efficiency-speed characteristic curve, and temperature rise limit compliance conclusions.

[0041] The brushless motor performance testing device adopts a hierarchical architecture driven by a six-core processor. Signal acquisition layer: integrates a 16-channel synchronous acquisition card and anti-aliasing filter, supporting synchronous acquisition of ±50A current and 100kHz vibration; Intelligent decision-making layer: Deploy machine learning model library and reinforcement learning engine to achieve closed-loop decision-making of working condition identification, policy matching and load control; Execution control layer: A 500W servo loading system, in conjunction with a magnetic powder brake, accurately reproduces impact loads and continuous speed change conditions; Environmental interference immunity layer: A platinum resistance temperature drift correction unit and a second-order Butterworth filter constitute a dual anti-interference mechanism. Output interaction layer: Supports dual-mode communication of CAN bus and Ethernet, and outputs standardized reports in both PDF and CSV formats.

[0042] The brushless motor performance testing equipment is an industrial-grade testing terminal built based on the ARM Cortex-A72 processor. Real-time processing core: A quad-core processor processes 12 signal streams in parallel at a clock speed of 1.5GHz; Anti-interference storage system: eMMC embedded memory stably stores test programs and databases in an environment of -40℃ to 85℃; Multi-protocol interface: Integrated CAN 2.0B / Modbus TCP interface, supporting connection to production line PLC and cloud quality platform; Dynamic power management: Wide voltage input adapts to workshop voltage fluctuations, ensuring test continuity.

[0043] The storage medium is stored in a two-level high-performance storage architecture. A computer program implements five-thread collaboration. Acquisition thread: Millisecond-level timestamp synchronization of 12 sensor data streams; Recognition thread: Calls the ResNet-34 model library to perform real-time feature extraction and model matching; Control thread: Dynamically adjust the servo loading strategy using the Q-learning algorithm; Compensation thread: Online correction of signal baseline drift caused by temperature and humidity; Reporting thread: Automatically generates interactive test reports containing 3D performance cloud maps.

[0044] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method of testing the performance of a brushless motor, characterised by, It comprises the following steps: S1, collecting voltage signals, current signals, rotating speed signals and temperature signals of the brushless motor during operation through a multi-channel sensor to generate an original performance data set; S2, performing working condition feature extraction and motor type identification processing on the original performance data set based on a machine learning algorithm to generate working condition feature identification data and motor type identification data; S3, performing dynamic test strategy matching processing according to the working condition feature identification data and the motor type identification data in combination with a preset test strategy library to generate target test strategy type feature data; S4, controlling a servo loading system to simulate an actual working condition according to the target test strategy type feature data to perform dynamic performance testing on the brushless motor to generate real-time test waveform data; S5, monitoring temperature and humidity parameters in real time through an environmental compensation module and correcting environmental interference errors in the real-time test waveform data to generate calibrated test data; S6, performing performance index analysis processing on the calibrated test data based on a statistical analysis model and a signal processing algorithm to generate a brushless motor performance evaluation report.

2. The method of claim 1, wherein, The S1 comprises: S11, synchronously collecting voltage signals and current signals of three-phase windings of the brushless motor through a 24-bit high-precision ADC analog-to-digital converter at a sampling frequency of not less than 100 kHz to ensure signal fidelity; S12, collecting rotating speed signals of a motor rotor through a Hall effect sensor and collecting temperature distribution signals of key parts of a motor shell through a K-type thermocouple at a resolution of 0.1℃; S13, aligning and integrating the voltage signals, current signals, rotating speed signals and temperature signals according to millisecond-level time stamps to construct a multi-dimensional original performance data set containing space-time correlation features.

3. The method of claim 1, wherein, The S2 comprises: S21, inputting the original performance data set into a pre-trained ResNet-34 convolutional neural network model to extract current harmonic distortion features, voltage sag features and temperature gradient distribution features; S22, performing cosine similarity matching based on the extracted feature vectors and a motor type database containing 500+ models to output motor type identification data with confidence score; S23, generating feature identification data representing steady-state and transient working conditions according to rotating speed signal mutation frequency and load change trend in combination with a Markov model.

4. The method of claim 1, wherein, The S3 comprises: S31, establishing a test strategy library matrix covering 12 industrial scenarios: Q = (q1,..., q k ,..., q 12 ), k = 1, 2,..., 12; wherein q k represents the standard working condition characteristic combination data corresponding to the kth test strategy type; S32, match the working condition characteristic identification data with q in Q k match, iterate and optimize the matching path through the Q-learning reinforcement learning algorithm, and output the target test strategy type characteristic data containing the loading mode, test frequency and accuracy requirement.

5. The method of claim 1, wherein, The S4 comprises: S41, controlling a magnetic powder brake servo system to generate adjustable step load, ramp load and 0-100 Hz sinusoidal fluctuation load of 0-20 Nm according to load mode parameters in the target test strategy type feature data; S42, synchronously collecting motor response waveforms during dynamic loading to generate real-time test waveform data containing electromagnetic noise spectrum, vibration frequency spectrum features and efficiency-rotating speed curve; The electromagnetic noise power spectrum density is calculated using the formula: where P noise (f) is the noise power spectral density at frequency f and has units of A 2 / Hz, T is the sampling time duration, and X(f) is the complex frequency spectrum value of the current signal after FFT transformation at frequency f.

6. The method of claim 1, wherein, The S5 comprises: S51, monitoring test environment parameters in real time through an SHT35 temperature and humidity sensor at a frequency of 1 Hz to generate an environmental compensation coefficient matrix containing temperature compensation coefficients and humidity drift coefficients; S52, using weighted least squares method to fit the environmental compensation coefficient matrix with real-time test waveform data in multi-dimensional error, outputting calibration test data eliminating environmental interference: where y i is the original test data of the i-th sampling point, T i is the temperature measurement value of the corresponding point, H i is the humidity measurement value of the corresponding point, β0, β1, β2 are compensation coefficients to be solved, w i is a weight factor based on the signal-to-noise ratio, taking a value of 0-1.

7. A brushless motor performance testing device for implementing the brushless motor performance testing method according to any one of claims 1-6, characterized in that, The device comprises: The signal acquisition module is configured with a 24-bit delta-sigma architecture ADC and a 16-channel synchronous acquisition card, and is suitable for high-fidelity acquisition of voltage, current, speed and temperature signals; The working condition recognition module integrates a pre-trained machine learning model and a real-time inference engine, extracts dynamic working condition features and recognizes motor models; The strategy selection module stores a multi-dimensional test strategy library and dynamically matches the optimal test scheme through a reinforcement learning algorithm; The test control module is connected to a 500W servo loading system and is suitable for executing complex working condition simulation; The environmental compensation module has an industrial-grade temperature and humidity sensor and a self-adaptive correction algorithm; The data analysis module analyzes performance indicators based on wavelet transform and regression analysis techniques.

8. The brushless motor performance testing device of claim 7, wherein, The environmental compensation module further comprises: A high-frequency vibration sensor with a sampling rate of 100 kHz acquires mechanical vibration spectrum characteristics under load working conditions; A second-order Butterworth electromagnetic interference filter eliminates high-frequency noise above 50 kHz in the current signal; A real-time temperature drift correction unit based on a PT100 platinum resistor dynamically adjusts the signal gain coefficient according to a temperature change rate of 0.1℃ / s.

9. A brushless motor performance testing apparatus for implementing the brushless motor performance testing method according to any one of claims 1 to 6, characterized by, It comprises: A quad-core ARM Cortex-A72 processor, 4GB LPDDR4 memory and computer programs stored in eMMC memory; When the computer program is executed by the processor, the brushless motor performance test method is realized, and CAN bus and industrial Internet of Things platform communication are supported.

10. A storage medium for testing the performance of a brushless motor, characterized by An industrial-grade SD card stores computer programs, and when the computer programs are executed by the processor, the brushless motor performance test method according to any one of claims 1-6 is realized, specifically comprising: Real-time acquisition of multi-channel sensor signals and generation of time-stamped original performance data sets through data acquisition threads; Call the preset machine learning model library to execute working condition feature extraction and motor type identification algorithm; Based on the reinforcement learning engine, dynamically match the test strategy and generate servo loading control instructions; Start the environmental compensation thread to correct the temperature and humidity interference online; Finally, a standard-compliant PDF format test report is generated and uploaded to the cloud quality monitoring platform through the Ethernet interface.

Images