Permanent magnet synchronous motor rack calibration automatic test method and system

Abstract

The invention relates to the technical field of permanent magnet synchronous motors, and discloses a permanent magnet synchronous motor rack calibration automatic test method and system, and the method comprises the steps: collecting the original operation data of a permanent magnet synchronous motor, carrying out the preprocessing of the original operation data, and obtaining the processed motor data; selecting a sample set from the processed motor data, constructing a torque-rotating speed characteristic model of the permanent magnet synchronous motor by adopting a polynomial fitting algorithm based on a least square method, and constructing a motor test model by adopting a multi-layer sensor structure; respectively inputting the processed motor data into a torque-rotating speed characteristic model and a motor test model, and dynamically allocating weights to prediction precision of the torque-rotating speed characteristic model and the motor test model under different working conditions by adopting a weighted average mode to obtain a fused model prediction result; according to the fused model prediction result, generating a permanent magnet synchronous motor rack calibration test report; the testing efficiency is improved, and personal errors are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of permanent magnet synchronous motors, and particularly relates to an automatic test method and system for permanent magnet synchronous motor bench calibration. Background Art

[0002] Permanent magnet synchronous motors are widely used in many fields such as electric vehicles and industrial automation. To ensure the accuracy and reliability of motor performance, motor bench calibration testing is crucial. Traditional calibration testing often relies on manual operation. Testers need to manually set various parameters, start equipment, and record data. The process is cumbersome and error-prone, not only consuming a large amount of human and time costs, but also difficult to guarantee the consistency and accuracy of test results due to the interference of human factors. In addition, in the face of complex test conditions and diverse test requirements, manual testing is difficult to complete efficiently and comprehensively, and cannot meet the high-precision and automated requirements for motor testing in the rapid development of modern industry. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and design an automatic test method and system for permanent magnet synchronous motor bench calibration.

[0004] The first aspect of the present invention provides an automatic test method for permanent magnet synchronous motor bench calibration, and the automatic test method for permanent magnet synchronous motor bench calibration includes the following steps:

[0005] Install the permanent magnet synchronous motor on the bench, collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocess the original operation data to obtain processed motor data;

[0006] Select a sample set from the processed motor data, and use a polynomial fitting algorithm based on the least squares method to construct a torque-speed characteristic model of the permanent magnet synchronous motor, and use a multi-layer perceptron structure to construct a motor test model;

[0007] Input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and use the weighted average method to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the fused model prediction result;

[0008] Generate a permanent magnet synchronous motor bench calibration test report according to the fused model prediction result, where the test report at least includes the basic information of the motor, an overview of the test conditions, a torque-speed characteristic curve, torque change conditions, and key performance index values. The key performance index values at least include the torque constant and back electromotive force coefficient of the motor.

[0009] Optionally, in the first implementation manner of the first aspect of the present invention, installing the permanent magnet synchronous motor on a test bench, collecting the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocessing the original operation data to obtain processed motor data, including:

[0010] Install the permanent magnet synchronous motor on a test bench, and collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor;

[0011] Perform a preliminary screening on the original operation data, perform outlier detection using the 3σ principle, calculate the mean μ and standard deviation σ of the collected torque data and speed data, set the threshold range as (μ - 3σ, μ + 3σ) according to the 3σ principle, and remove the detected outliers from the original operation data to obtain the preliminarily screened data;

[0012] Use the moving average filtering algorithm to smooth the preliminarily screened data, calculate the moving averages of the torque data and speed data, and perform moving average processing according to the moving averages to obtain the smoothed data;

[0013] Use the minimum-maximum normalization method to normalize the smoothed data to obtain the processed motor data.

[0014] Optionally, in the second implementation manner of the first aspect of the present invention, selecting a sample set from the processed motor data, and constructing a torque-speed characteristic model of the permanent magnet synchronous motor using a polynomial fitting algorithm based on the least squares method and constructing a motor test model using a multi-layer perceptron structure, including:

[0015] Divide the speed into low-speed, medium-speed, and high-speed intervals in the processed motor data, and divide the torque into light-load, medium-load, and heavy-load intervals;

[0016] Use the DBSCAN algorithm to cluster the data points in the processed motor data, calculate the number of points within the ∈-neighborhood of each data point. If the number of points within the ∈-neighborhood is greater than or equal to the minimum point threshold, mark the current data point as a core point, add all the points within the ∈-neighborhood of the core point to the same cluster, and repeat the iteration until all core points are processed. For non-core points, if there is a core point within its ∈-neighborhood, add the non-core point to the cluster where the core point is located, otherwise mark it as a noise point, and obtain the data clusters under different working conditions as the clustering result;

[0017] Randomly and with replacement extract samples from the data clusters under different working conditions after clustering, construct m decision trees, calculate the information gain of each sample in the tree, synthesize the results of all decision trees, calculate the average importance score of each sample, and divide the samples with high importance scores to obtain the sample set;

[0018] Based on the sample set, a torque-speed characteristic model of a permanent magnet synchronous motor is constructed using a polynomial fitting algorithm based on the least squares method, and a motor test model is constructed using a multi-layer perceptron structure.

[0019] Optionally, in the third implementation manner of the first aspect of the present invention, the constructing a torque-speed characteristic model of a permanent magnet synchronous motor using a polynomial fitting algorithm based on the least squares method according to the sample set includes:

[0020] Use the least squares method to fit the data in the sample set for each polynomial degree to obtain a fitting model, calculate the sum of squared residuals of the fitting model, and dynamically determine the optimal polynomial degree by combining the Akaike information criterion and the Bayesian information criterion;

[0021] Divide the speed range of the permanent magnet synchronous motor in the sample set into multiple sub-intervals, and assign different weights to the data points according to the distribution of the sample points within the sub-intervals;

[0022] For each speed sub-interval, use the weights determined by the sample points and the optimal polynomial degree to perform locally weighted least squares fitting to obtain a fitting model;

[0023] Calculate the residuals of each data point in the sample set, use the Gaussian process regression algorithm to model the residuals to obtain a residual correction model, and add the prediction result of the fitting model to the output of the residual correction model to obtain the predicted value of the final torque-speed characteristic model, so as to construct the torque-speed characteristic model of the permanent magnet synchronous motor.

[0024] Optionally, in the fourth implementation manner of the first aspect of the present invention, the constructing a motor test model using a multi-layer perceptron structure includes:

[0025] Extract the characteristic data related to the motor performance from the historical data and test data of the operation of the permanent magnet synchronous motor, determine the number of layers of the multi-layer perceptron, and set the number of neurons in the input layer, the number of hidden layers and the number of neurons in each hidden layer, and the number of neurons in the output layer;

[0026] Use the DBO algorithm to optimize the multi-layer perceptron, set the relevant parameters of the DBO algorithm, including the size of the dung beetle population, the number of iterations, and the optimization boundary, initialize the positions of the dung beetle population, each position represents a combination of weights and biases of a multi-layer perceptron, and calculate its fitness value;

[0027] Iteratively update the positions of the dung beetles. When the dung beetles encounter obstacles, they will update their positions by dancing. When the female dung beetles reproduce, a boundary selection strategy is used to determine the reproduction area, continuously search for better combinations of weights and biases, and update the current optimal solution and its fitness value after each iteration until the maximum number of iterations is reached to obtain the motor test model.

[0028] Optionally, in the fifth implementation manner of the first aspect of the present invention, the inputting the processed motor data into the torque-speed characteristic model and the motor test model respectively includes:

[0029] Input the preprocessed data into the torque-speed characteristic model constructed by the polynomial fitting algorithm based on the least squares method, and calculate the corresponding torque prediction value T for each input data N. The expression of the torque-speed characteristic model is T = a0 + a1N + a2N 2 +…+ a p N p , where N is the input data, and a0, a1, a2, …, a p are the polynomial coefficients obtained by least squares fitting;

[0030] Input the preprocessed data into the motor test model, scale the input data through the multi-layer perceptron structure, and input the scaled data into the motor test model optimized by the DBO algorithm. Starting from the input layer, pass through each hidden layer in turn. In each hidden layer, the data is multiplied by the weight matrix of this layer, then added with the bias vector, and then non-linearly transformed through the ReLU function to obtain the output of this hidden layer. After the calculation of the output layer, the prediction result of the motor test model is obtained.

[0031] Optionally, in the sixth implementation manner of the first aspect of the present invention, the method of using weighted average to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the fused model prediction result includes:

[0032] Input the processed motor data into the already constructed torque-speed characteristic model and motor test model respectively for preliminary prediction to obtain preliminary prediction results.

[0033] Let the predicted value of the torque-speed characteristic model be y 1i , the actual value be y i , and the number of samples be n. Then the mean square error of the torque-speed characteristic model is:

[0034]

[0035] Let the predicted value of the motor test model be y 2i . Then the mean square error of the motor test model is:

[0036]

[0037] According to the calculated prediction accuracy evaluation results, dynamically allocate the weights of the torque-speed characteristic model and the motor test model in a weighted average manner;

[0038] According to the assigned weights, the prediction results of the torque-speed characteristic model and the motor test model are weighted and averaged to obtain the prediction results of the fused model.

[0039] The second aspect of the present invention provides a permanent magnet synchronous motor bench calibration automatic test system, which includes a data acquisition module, a model construction module, a model prediction module, and a report generation module. Among them,

[0040] The data acquisition module is used to install the permanent magnet synchronous motor on the bench, collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocess the original operation data to obtain the processed motor data;

[0041] The model construction module is used to select a sample set from the processed motor data, and construct a torque-speed characteristic model of the permanent magnet synchronous motor by using a polynomial fitting algorithm based on the least squares method, and construct a motor test model by using a multi-layer perceptron structure;

[0042] The model prediction module is used to input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and adopt a weighted average method to dynamically assign weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the prediction results of the fused model;

[0043] The report generation module is used to generate a permanent magnet synchronous motor bench calibration test report according to the prediction results of the fused model, where the test report at least includes the basic information of the motor, an overview of the test working conditions, a torque-speed characteristic curve, torque change conditions, and key performance index values. The key performance index values at least include the torque constant and back electromotive force coefficient of the motor.

[0044] Optionally, in the first implementation manner of the second aspect of the present invention, the data acquisition module includes an acquisition sub-module, a rejection sub-module, a smoothing processing sub-module, and a normalization processing sub-module. Among them,

[0045] The acquisition sub-module is used to install the permanent magnet synchronous motor on the bench and collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor;

[0046] The rejection sub-module is used to perform a preliminary screening on the original operation data, detect outliers using the 3σ principle, calculate the mean μ and standard deviation σ of the collected torque data and speed data, set the threshold range as (μ - 3σ, μ + 3σ) according to the 3σ principle, and remove the detected outliers from the original operation data to obtain the preliminarily screened data;

[0047] A smoothing sub-module, which is used to smooth the preliminarily screened data by using a moving average filtering algorithm, calculate the moving averages of torque data and rotational speed data, and perform moving average processing according to the moving averages to obtain smoothed data;

[0048] A normalization sub-module, which is used to normalize the smoothed data by using the min-max normalization method to obtain processed motor data.

[0049] Optionally, in the second implementation manner of the second aspect of the present invention, the model construction module includes a division sub-module, a clustering sub-module, a calculation sub-module, and a construction sub-module, where,

[0050] The division sub-module is used to divide the rotational speed into low-speed, medium-speed, and high-speed intervals and the torque into light-load, medium-load, and heavy-load intervals in the processed motor data;

[0051] The clustering sub-module is used to cluster the data points in the processed motor data by using the DBSCAN algorithm, calculate the number of points in the ∈-neighborhood of each data point. If the number of points in the ∈-neighborhood is greater than or equal to the minimum point threshold, mark the current data point as a core point, add all the points in the ∈-neighborhood of the core point to the same cluster, and repeat the iteration until all core points are processed. For non-core points, if there is a core point in its ∈-neighborhood, add the non-core point to the cluster where the core point is located, otherwise mark it as a noise point, and obtain data clusters under different working conditions as the clustering result;

[0052] The calculation sub-module is used to randomly and with replacement extract samples from the data clusters under different working conditions after clustering, construct m decision trees, calculate the information gain of each sample in the tree, synthesize the results of all decision trees, calculate the average importance score of each sample, and divide the high-importance samples to obtain a sample set;

[0053] The construction sub-module is used to construct a torque-speed characteristic model of a permanent magnet synchronous motor by using a polynomial fitting algorithm based on the least squares method according to the sample set, and construct a motor test model by using a multi-layer perceptron structure.

[0054] In the technical solution provided by the present invention, a permanent magnet synchronous motor is installed on a test bench, and the original operation data of the permanent magnet synchronous motor is collected through a torque sensor and a speed sensor. The original operation data is preprocessed to obtain processed motor data. A sample set is selected from the processed motor data, and a torque-speed characteristic model of the permanent magnet synchronous motor is constructed according to the sample set by using a polynomial fitting algorithm based on the least square method, and a motor test model is constructed by using a multi-layer perceptron structure. The processed motor data is respectively input into the torque-speed characteristic model and the motor test model, and the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions are dynamically weighted by using a weighted average method to obtain a fused model prediction result. According to the fused model prediction result, a calibration test report of the permanent magnet synchronous motor test bench is generated. The automatic test method for calibrating the permanent magnet synchronous motor test bench of the present invention can greatly improve the test efficiency, reduce human errors, provide a strong guarantee for the research and development, production and quality control of the permanent magnet synchronous motor, and promote the development of related industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0056] Figure 1 Schematic diagram of the first embodiment of the automatic test method for calibrating the permanent magnet synchronous motor test bench provided by the embodiment of the present invention;

[0057] Figure 2 Schematic diagram of the second embodiment of the automatic test method for calibrating the permanent magnet synchronous motor test bench provided by the embodiment of the present invention;

[0058] Figure 3 Schematic diagram of the third embodiment of the automatic test method for calibrating the permanent magnet synchronous motor test bench provided by the embodiment of the present invention;

[0059] Figure 4 Schematic diagram of the structure of the automatic test system for calibrating the permanent magnet synchronous motor test bench provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] The terms "first", "second", "third", "fourth", etc. (if any) in the description, claims and the above-mentioned drawings of the present invention are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or equipment.

[0061] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Please refer to Figure 1 The schematic diagram of the first embodiment of the automatic test method for the bench calibration of a permanent magnet synchronous motor provided by the embodiment of the present invention, and this method specifically includes the following steps:

[0062] Step 101: Install the permanent magnet synchronous motor on the bench, collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocess the original operation data to obtain the processed motor data;

[0063] In this embodiment, the permanent magnet synchronous motor is installed on the bench, and the original operation data of the permanent magnet synchronous motor is collected through a torque sensor and a speed sensor; the original operation data is preliminarily screened, and the 3σ principle is used for outlier detection. Calculate the mean μ and standard deviation σ of the collected torque data and speed data. According to the 3σ principle, set the threshold range as (μ - 3σ, μ + 3σ), and remove the detected outliers from the original operation data to obtain the preliminarily screened data; use the moving average filtering algorithm to smooth the preliminarily screened data, calculate the moving average values of the torque data and speed data, and perform moving average processing according to the moving average values to obtain the smoothed data; use the minimum-maximum normalization method to normalize the smoothed data to obtain the processed motor data.

[0064] Step 102: Select a sample set from the processed motor data, and construct a torque-speed characteristic model of the permanent magnet synchronous motor using a polynomial fitting algorithm based on the least squares method, and construct a motor test model using a multi-layer perceptron structure;

[0065] Step 103: Input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and use the weighted average method to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the fused model prediction result;

[0066] In this embodiment, the processed motor data is respectively input into the already constructed torque-speed characteristic model and the motor test model for preliminary prediction, and preliminary prediction results are obtained.

[0067] Let the predicted value of the torque-speed characteristic model be y 1i , and the actual value be y i . Given the number of samples as n, the mean square error of the torque-speed characteristic model is:

[0068]

[0069] Let the predicted value of the motor test model be y 2i , then the mean square error of the motor test model is:

[0070]

[0071] According to the calculated prediction accuracy evaluation results, the weights of the torque-speed characteristic model and the motor test model are dynamically allocated in a weighted average manner;

[0072] According to the allocated weights, the prediction results of the torque-speed characteristic model and the motor test model are weighted and averaged to obtain the fused model prediction result.

[0073] Step 104: Generate a bench calibration test report for the permanent magnet synchronous motor according to the fused model prediction result.

[0074] In this embodiment, the test report at least includes the basic information of the motor, an overview of the test conditions, the torque-speed characteristic curve, the torque change situation, and the numerical values of the key performance indicators. The numerical values of the key performance indicators at least include the torque constant and the back electromotive force coefficient of the motor.

[0075] Please refer to Figure 2 , the schematic diagram of the second embodiment of the automatic bench calibration test method for the permanent magnet synchronous motor provided by the embodiment of the present invention. The method includes:

[0076] Step 201: Divide the rotational speed into low-speed, medium-speed, and high-speed intervals, and divide the torque into light-load, medium-load, and heavy-load intervals in the processed motor data;

[0077] Step 202: Use the DBSCAN algorithm to cluster the data points in the processed motor data. Calculate the number of points within the ε-neighborhood of each data point. If the number of points within the ε-neighborhood is greater than or equal to the minimum point threshold, mark the current data point as a core point, and add all the points within the ε-neighborhood of the core point to the same cluster. Repeat the iteration until all core points are processed. For non-core points, if there is a core point within their ε-neighborhood, add the non-core points to the cluster where the core point is located; otherwise, mark them as noise points. The clustering result is data clusters under different working conditions.

[0078] Step 203: Randomly and with replacement, draw samples from the data clusters under different working conditions after clustering, construct m decision trees, calculate the information gain of each sample in the tree, combine the results of all decision trees, calculate the average importance score of each sample, and divide the high-importance samples to obtain a sample set.

[0079] Step 204: According to the sample set, use the polynomial fitting algorithm based on the least squares method to construct the torque-speed characteristic model of the permanent magnet synchronous motor, and use the multi-layer perceptron structure to construct the motor test model.

[0080] In this embodiment, the data in the sample set is fitted using the least squares method for each polynomial degree to obtain a fitting model. Calculate the sum of squared residuals of the fitting model. Dynamically determine the optimal polynomial degree by combining the Akaike information criterion and the Bayesian information criterion. Divide the speed range of the permanent magnet synchronous motor in the sample set into multiple sub-intervals, and assign different weights to the data points according to the distribution of the sample points within the sub-intervals. For each speed sub-interval, use the weights determined by the sample points and the optimal polynomial degree to perform locally weighted least squares fitting to obtain a fitting model. Calculate the residuals of each data point in the sample set, use the Gaussian process regression algorithm to model the residuals to obtain a residual correction model, and add the prediction result of the fitting model to the output of the residual correction model to obtain the predicted value of the final torque-speed characteristic model, so as to construct the torque-speed characteristic model of the permanent magnet synchronous motor.

[0081] In this embodiment, characteristic data related to the motor performance is extracted from the historical data and test data of the permanent magnet synchronous motor operation, the number of layers of the multi-layer perceptron is determined, the number of neurons in the input layer, the number of hidden layers and the number of neurons in each hidden layer, and the number of neurons in the output layer are set; the multi-layer perceptron is optimized by the DBO algorithm, and the relevant parameters of the DBO algorithm are set, including the size of the dung beetle population, the number of iterations, and the optimization boundary. The position of the dung beetle population is initialized, and each position represents a combination of weights and biases of a multi-layer perceptron, and its fitness value is calculated; the position of the dung beetle is iteratively updated. When the dung beetle encounters an obstacle, it will update its position by dancing. When the female dung beetle reproduces, a boundary selection strategy is used to determine the breeding area, and a better combination of weights and biases is continuously searched. After each iteration, the current optimal solution and its fitness value are updated until the maximum number of iterations is reached, and the motor test model is obtained.

[0082] Please refer to Figure 3 , which is a schematic diagram of the third embodiment of the automatic test method for the permanent magnet synchronous motor bench calibration provided by the embodiment of the present invention. The method includes:

[0083] Step 301: Input the preprocessed data into the torque-speed characteristic model constructed by the polynomial fitting algorithm based on the least squares method, and calculate the corresponding torque prediction value for each input data.

[0084] In this embodiment, for each input data N, the corresponding torque prediction value T is calculated. The expression of the torque-speed characteristic model is T = a0 + a1N + a2N 2 +…+ a p N p , where N is the input data, and a0, a1, a2, …, a p are the polynomial coefficients obtained by least squares fitting;

[0085] Step 302: Input the preprocessed data into the motor test model, perform scaling processing on the input data through the multi-layer perceptron structure, input the scaled data into the motor test model optimized by the DBO algorithm, and start from the input layer, pass through each hidden layer in turn. In each hidden layer, the data is multiplied by the weight matrix of this layer, then added with the bias vector, and then non-linearly transformed through the ReLU function to obtain the output of this hidden layer. After the calculation of the output layer, the prediction result of the motor test model is obtained.

[0086] Please refer to Figure 4 , which is a schematic diagram of the structure of the automatic test system for the permanent magnet synchronous motor bench calibration provided by the embodiment of the present invention. The system includes a data acquisition module, a model construction module, a model prediction module, and a report generation module. Among them,

[0087] The data acquisition module 401 is used to install the permanent magnet synchronous motor on the test bench, collect the original operation data of the permanent magnet synchronous motor through the torque sensor and the speed sensor, and preprocess the original operation data to obtain the processed motor data;

[0088] The model construction module 402 is used to select a sample set from the processed motor data, and construct the torque-speed characteristic model of the permanent magnet synchronous motor by using the polynomial fitting algorithm based on the least square method, and construct the motor test model by using the multi-layer perceptron structure;

[0089] The model prediction module 403 is used to input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and adopt the weighted average method to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the fused model prediction result;

[0090] The report generation module 404 is used to generate the test report for the calibration of the permanent magnet synchronous motor test bench according to the fused model prediction result, where the test report at least includes the basic information of the motor, the overview of the test working conditions, the torque-speed characteristic curve, the torque change situation and the numerical values of the key performance indicators, and the numerical values of the key performance indicators at least include the torque constant and the back electromotive force coefficient of the motor.

[0091] In this embodiment, the data acquisition module includes an acquisition sub-module, a rejection sub-module, a smoothing processing sub-module and a normalization processing sub-module, where,

[0092] The acquisition sub-module is used to install the permanent magnet synchronous motor on the test bench and collect the original operation data of the permanent magnet synchronous motor through the torque sensor and the speed sensor;

[0093] The rejection sub-module is used to perform preliminary screening on the original operation data, detect outliers by using the 3σ principle, calculate the mean μ and the standard deviation σ of the collected torque data and speed data, set the threshold range as (μ - 3σ, μ + 3σ) according to the 3σ principle, and remove the detected outliers from the original operation data to obtain the preliminarily screened data;

[0094] The smoothing processing sub-module is used to perform smoothing processing on the preliminarily screened data by using the moving average filtering algorithm, calculate the moving average values of the torque data and the speed data, and perform moving average processing according to the moving average values to obtain the smoothed data;

[0095] The normalization processing sub-module is used to perform normalization processing on the smoothed data by using the minimum-maximum normalization method to obtain the processed motor data.

[0096] In this embodiment, the model construction module includes a sub-module for division, a clustering sub-module, a calculation sub-module, and a construction sub-module, where

[0097] The sub-module for division is used to divide the rotational speed into low-speed, medium-speed, and high-speed ranges, and the torque into light-load, medium-load, and heavy-load ranges in the processed motor data;

[0098] The clustering sub-module is used to cluster the data points in the processed motor data using the DBSCAN algorithm, calculate the number of points within the ∈-neighborhood of each data point. If the number of points within the ∈-neighborhood is greater than or equal to the minimum point threshold, mark the current data point as a core point, add all the points within the ∈-neighborhood of the core point to the same cluster, and repeat the iteration until all core points are processed. For non-core points, if there is a core point within its ∈-neighborhood, add the non-core point to the cluster where the core point is located, otherwise mark it as a noise point, and obtain data clusters under different working conditions as the clustering result;

[0099] The calculation sub-module is used to randomly sample with replacement from the data clusters under different working conditions after clustering, construct m decision trees, calculate the information gain of each sample in the tree, synthesize the results of all decision trees, calculate the average importance score of each sample, and divide the samples with high importance scores to obtain a sample set;

[0100] The construction sub-module is used to construct the torque-speed characteristic model of the permanent magnet synchronous motor using the polynomial fitting algorithm based on the least squares method according to the sample set, and construct the motor test model using a multi-layer perceptron structure.

[0101] Through the implementation of the above solutions, the original operation data of the permanent magnet synchronous motor is collected by using torque sensors and speed sensors, and the key parameters of the motor during actual operation can be obtained in real time and accurately, providing a reliable basis for subsequent analysis and ensuring that the test results truly reflect the motor operation state; preprocessing the original operation data can remove noise, fill in missing values, etc., improving the data quality and making the subsequent models and analyses based on the processed data more accurate and reliable; constructing a torque-speed characteristic model based on the polynomial fitting algorithm of the least squares method can make full use of the sample set data, accurately fit the relationship between the motor torque and speed, clearly depict this characteristic of the motor, and provide a strong basis for motor performance evaluation and optimization; constructing a motor test model using a multi-layer perceptron structure can comprehensively consider the influence of multiple factors on the motor performance, test and evaluate the motor from multiple dimensions, and comprehensively reflect the operation performance of the motor under different working conditions; inputting the processed data into the two models and using weighted average fusion, dynamically allocating weights according to the model prediction accuracy under different working conditions, can give full play to the advantages of each model, make up for the limitations of a single model, and significantly improve the accuracy of the model in predicting the motor performance; automatically generating a bench calibration test report according to the prediction results of the fused model, realizing the automation of the test process, greatly reducing the manual operation and time costs, improving the test efficiency, and the report is based on the accurate model results, with higher credibility and reference value, providing important support for motor performance evaluation, quality control and subsequent optimization and improvement.

[0102] The above shows and describes 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 by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic test method for the bench calibration of a permanent magnet synchronous motor, characterized in that, The automatic test method for bench calibration of the permanent magnet synchronous motor includes the following steps: Install the permanent magnet synchronous motor on the bench, collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocess the original operation data to obtain processed motor data; Select a sample set from the processed motor data, and use a polynomial fitting algorithm based on the least squares method to construct a torque-speed characteristic model of the permanent magnet synchronous motor, and use a multi-layer perceptron structure to construct a motor test model; Input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and use the weighted average method to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the fused model prediction result; Generate a bench calibration test report for the permanent magnet synchronous motor according to the fused model prediction result, where the test report at least includes the basic information of the motor, an overview of the test working conditions, a torque-speed characteristic curve, torque change conditions, and key performance index values. The key performance index values at least include the torque constant and back electromotive force coefficient of the motor.

2. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 1, characterized in that, The step of installing the permanent magnet synchronous motor on the bench, collecting the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocessing the original operation data to obtain processed motor data includes: Install the permanent magnet synchronous motor on the bench, and collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor; Conduct a preliminary screening of the original operation data, perform outlier detection using the 3σ principle, calculate the mean μ and standard deviation σ of the collected torque data and speed data, set the threshold range as (μ - 3σ, μ + 3σ) according to the 3σ principle, and remove the detected outliers from the original operation data to obtain the preliminarily screened data; Use a moving average filtering algorithm to smooth the preliminarily screened data, calculate the moving average values of the torque data and speed data, and perform moving average processing according to the moving average values to obtain the smoothed data; Use the minimum-maximum normalization method to normalize the smoothed data to obtain the processed motor data.

3. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 1, characterized in that, The step of selecting a sample set from the processed motor data, using a polynomial fitting algorithm based on the least squares method to construct a torque-speed characteristic model of the permanent magnet synchronous motor, and using a multi-layer perceptron structure to construct a motor test model includes: Divide the speed into low-speed, medium-speed, and high-speed intervals, and divide the torque into light-load, medium-load, and heavy-load intervals in the processed motor data; The DBSCAN algorithm is used to cluster the data points in the processed motor data. Calculate the number of points within the ∈-neighborhood of each data point. If the number of points within the ∈-neighborhood is greater than or equal to the minimum point threshold, mark the current data point as a core point, and add all the points within the ∈-neighborhood of the core point to the same cluster. Repeat the iteration until all core points are processed. For non-core points, if there is a core point within its ∈-neighborhood, add the non-core point to the cluster where the core point is located; otherwise, mark it as a noise point. The clustering result is used to obtain data clusters under different working conditions; Randomly sample with replacement from the data clusters under different working conditions after clustering to construct m decision trees. Calculate the information gain of each sample in the tree, and comprehensively consider the results of all decision trees to calculate the average importance score of each sample. Divide the samples with high importance scores to obtain a sample set; According to the sample set, use the polynomial fitting algorithm based on the least squares method to construct the torque-speed characteristic model of the permanent magnet synchronous motor, and use the multi-layer perceptron structure to construct the motor test model.

4. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 3, wherein The method of using the polynomial fitting algorithm based on the least squares method according to the sample set to construct the torque-speed characteristic model of the permanent magnet synchronous motor includes: For each polynomial degree, use the least squares method to fit the data in the sample set to obtain a fitting model. Calculate the sum of squared residuals of the fitting model, and dynamically determine the optimal polynomial degree by combining the Akaike information criterion and the Bayesian information criterion; Divide the speed range of the permanent magnet synchronous motor in the sample set into multiple sub-intervals, and assign different weights to the data points according to the distribution of the sample points within the sub-intervals; For each speed sub-interval, use the weights determined by the sample points and the optimal polynomial degree to perform locally weighted least squares fitting to obtain a fitting model; Calculate the residuals of each data point in the sample set, use the Gaussian process regression algorithm to model the residuals to obtain a residual correction model, and add the prediction result of the fitting model and the output of the residual correction model to obtain the predicted value of the final torque-speed characteristic model, so as to construct the torque-speed characteristic model of the permanent magnet synchronous motor.

5. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 3, characterized in that, The method of using the multi-layer perceptron structure to construct the motor test model includes: Extract the characteristic data related to the motor performance from the historical data and test data of the permanent magnet synchronous motor operation, determine the number of layers of the multi-layer perceptron, and set the number of neurons in the input layer, the number of hidden layers, the number of neurons in each hidden layer, and the number of neurons in the output layer; Use the DBO algorithm to optimize the multi-layer perceptron, set the relevant parameters of the DBO algorithm, including the dung beetle population size, the number of iterations, and the optimization boundary. Initialize the position of the dung beetle population, where each position represents a combination of weights and biases of a multi-layer perceptron, and calculate its fitness value; Iteratively update the position of the dung beetle. When the dung beetle encounters an obstacle, it will update its position by dancing. When the female dung beetle reproduces, use the boundary selection strategy to determine the breeding area, and continuously search for better combinations of weights and biases. Update the current optimal solution and its fitness value after each iteration until the maximum number of iterations is reached to obtain the motor test model.

6. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 1, wherein, The step of inputting the processed motor data into the torque-speed characteristic model and the motor test model respectively includes: Input the preprocessed data into the torque-speed characteristic model constructed by the polynomial fitting algorithm based on the least squares method, and calculate the corresponding torque prediction value T for each input data N. The expression of the torque-speed characteristic model is T = a0 + a1N + a2N 2 +…+ a p N p , where N is the input data, and a0, a1, a2, …, a p are the polynomial coefficients obtained by least squares fitting; Input the preprocessed data into the motor test model. Scale the input data through a multi-layer perceptron structure. Then input the scaled data into the motor test model optimized by the DBO algorithm. Starting from the input layer, pass through each hidden layer in turn. In each hidden layer, the data is multiplied by the weight matrix of this layer, then added with the bias vector, and then undergoes a non-linear transformation through the ReLU function to obtain the output of this hidden layer. After the calculation of the output layer, obtain the prediction result of the motor test model.

7. The automatic test method for bench calibration of a permanent magnet synchronous motor according to claim 1, wherein, The method of using weighted average to dynamically allocate weights to the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the prediction result of the fused model includes: Input the processed motor data into the already constructed torque-speed characteristic model and motor test model respectively for preliminary prediction to obtain preliminary prediction results. Let the predicted value of the torque-speed characteristic model be y 1i , and the actual value be y i . Given the number of samples is n, the mean square error of the torque-speed characteristic model is as follows: Let the predicted value of the motor test model be y 2i , then the mean square error of the motor test model is: According to the calculated prediction accuracy evaluation results, dynamically allocate the weights of the torque-speed characteristic model and the motor test model in a weighted average manner; According to the allocated weights, perform weighted average on the prediction results of the torque-speed characteristic model and the motor test model to obtain the prediction result of the fused model.

8. A permanent magnet synchronous motor bench calibration automatic test system, characterized in that The permanent magnet synchronous motor bench calibration automatic test system includes a data acquisition module, a model construction module, a model prediction module, and a report generation module. Among them, The data acquisition module is used to install the permanent magnet synchronous motor on the bench, collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor, and preprocess the original operation data to obtain the processed motor data; The model construction module is used to select a sample set from the processed motor data, construct the torque-speed characteristic model of the permanent magnet synchronous motor by using the polynomial fitting algorithm based on the least squares method, and construct the motor test model by using a multi-layer perceptron structure; The model prediction module is used to input the processed motor data into the torque-speed characteristic model and the motor test model respectively, and use the method of weighted average to dynamically allocate the weights of the prediction accuracies of the torque-speed characteristic model and the motor test model under different working conditions to obtain the prediction result of the fused model; The report generation module is used to generate a permanent magnet synchronous motor bench calibration test report according to the prediction result of the fused model. The test report at least includes the basic information of the motor, an overview of the test working conditions, the torque-speed characteristic curve, the torque change situation, and the key performance index values. The key performance index values at least include the torque constant and back electromotive force coefficient of the motor.

9. The automatic test system for bench calibration of a permanent magnet synchronous motor according to claim 8, wherein The data acquisition module includes an acquisition sub-module, a rejection sub-module, a smoothing processing sub-module, and a normalization processing sub-module. Among them, The acquisition sub-module is used to install the permanent magnet synchronous motor on the bench and collect the original operation data of the permanent magnet synchronous motor through a torque sensor and a speed sensor; The rejection sub-module is used to perform a preliminary screening on the original operation data. It detects outliers using the 3σ principle, calculates the mean μ and standard deviation σ of the collected torque data and rotational speed data, sets the threshold range as (μ - 3σ, μ + 3σ) according to the 3σ principle, and rejects the detected outliers from the original operation data to obtain the preliminarily screened data; The smoothing processing sub-module is used to perform smoothing processing on the preliminarily screened data by using the moving average filtering algorithm, calculates the moving average values of the torque data and rotational speed data, and performs moving average processing according to the moving average values to obtain the smoothed data; The normalization processing sub-module is used to perform normalization processing on the smoothed data by using the min-max normalization method to obtain the processed motor data.

10. The automatic test system for bench calibration of a permanent magnet synchronous motor according to claim 8, characterized in that, The model construction module includes a division sub-module, a clustering sub-module, a calculation sub-module, and a construction sub-module, where The division sub-module is used to divide the rotational speed into low-speed, medium-speed, and high-speed intervals and the torque into light-load, medium-load, and heavy-load intervals in the processed motor data; The clustering sub-module is used to cluster the data points in the processed motor data by using the DBSCAN algorithm, calculates the number of points within the ∈-neighborhood of each data point. If the number of points within the ∈-neighborhood is greater than or equal to the minimum point threshold, the current data point is marked as a core point, and all points within the ∈-neighborhood of the core point are added to the same cluster. Repeat the iteration until all core points are processed. For non-core points, if there is a core point within its ∈-neighborhood, the non-core point is added to the cluster where the core point is located, otherwise it is marked as a noise point, and the clustering result is obtained to get data clusters under different working conditions; The calculation sub-module is used to randomly and with replacement extract samples from the data clusters under different working conditions after clustering, construct m decision trees, calculate the information gain of each sample in the tree, synthesize the results of all decision trees, calculate the average importance score of each sample, and divide the high-importance samples to obtain a sample set; The construction sub-module is used to construct the torque-speed characteristic model of the permanent magnet synchronous motor by using the polynomial fitting algorithm based on the least squares method according to the sample set, and construct the motor test model by using the multi-layer perceptron structure.

Images