Motor manufacturing rotation torque testing method and system

Abstract

The invention discloses a motor manufacturing rotation torque testing method and system, and relates to the technical field of motor testing, and the method comprises the steps: building a preliminary probability model based on normal distribution, obtaining the torque data of a motor under different loads in a testing process, building a likelihood function according to the torque data under different loads, and carrying out the testing of the motor manufacturing rotation torque. And continuously generating new candidate samples through an MH algorithm in combination with a likelihood function and a probability model, checking the convergence of a sample chain by using a GR diagnosis method, constructing posteriori distribution of target parameters by using the reserved candidate samples when a convergence condition is judged to be met, and analyzing and predicting the motor torque based on the posteriori distribution. The test system constructs a likelihood function and posterior distribution for data under different load conditions, can flexibly adapt to complex changes of torque under various load conditions, provides more accurate parameter estimation and distribution characteristic analysis, filters noise interference, and effectively improves the accuracy and stability of test data.

Description

Technical Field

[0001] The present invention relates to the technical field of motor testing, and in particular to a method and system for testing the rotational torque of a motor. Background Art

[0002] Rotational torque testing in motor manufacturing is an important step to ensure motor performance and quality. Rotational torque (or torque) refers to the moment acting on the motor rotor, which directly affects the motor's rotational performance and efficiency. During the motor production and manufacturing process, the rotational torque test can help identify the power capacity and stability that the motor can provide during starting, acceleration, deceleration, normal operation and stopping.

[0003] The prior art has the following defects:

[0004] Traditional torque test methods usually rely on simple statistical analysis and have poor adaptability to nonlinear characteristics and dynamic changes under complex working conditions. Especially under variable load conditions, it is difficult to accurately describe the change law of torque, and it is difficult to maintain test accuracy in high-noise or unstable data environments, especially when the noise is large or the data acquisition is unstable during the motor torque test.

[0005] Based on this, the present invention proposes a motor manufacturing rotation torque testing method and system, which can flexibly adapt to the complex changes of torque under various load conditions by constructing likelihood functions and posterior distributions for data under different load conditions, and provide more accurate parameter estimation and distribution characteristic analysis, and filter noise interference, effectively improving the accuracy and stability of test data. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for testing the rotational torque of a motor in manufacturing, so as to solve the deficiencies in the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for testing the rotation torque of a motor, the testing method comprising the following steps:

[0008] The test system establishes a preliminary probability model based on the normal distribution according to the physical characteristics of the motor torque. During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed based on the torque data under different loads.

[0009] New candidate samples are continuously generated by combining the likelihood function and the probability model through the MH algorithm. After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate samples.

[0010] When entering the burn-in period, the initially unstable candidate samples are screened, and the GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution.

[0011] In a preferred embodiment, the test system establishes a preliminary probability model based on the normal distribution according to the known physical characteristics of the motor torque, including the following steps:

[0012] Obtain the torque fluctuation range of similar motors based on the historical test data of the motor [D min ~D max ], where D min Indicates the minimum torque of the motor, D max It represents the maximum torque of the motor. According to the torque fluctuation range of similar motors and historical test data, the mean torque and standard deviation of the motor are estimated, and then the prior distribution of the mean torque and standard deviation of the torque is defined. The probability model is established by combining the prior distribution, the mean torque and standard deviation of the motor.

[0013] In a preferred embodiment, during the test process, the torque data of the motor under different loads is obtained, and the likelihood function is constructed according to the torque data under different loads, including the following steps:

[0014] Carry out motor torque test under different load conditions and record the torque value of the motor under each load;

[0015] The collected torque data is organized into a structured format, and the torque data under different load conditions is stored as a data set, which includes the torque value and corresponding load information under each load condition;

[0016] Based on the sorted torque data, the corresponding likelihood function is constructed for different load conditions. The function expression is: Where L is the likelihood value, μ is the mean torque, σ is the standard deviation of the torque, and θ i represents the i-th torque value obtained under the test under specific load conditions, n represents the total number of torque data collected under the specific load, is the product symbol, represents the probability density of torque data under a given torque standard deviation σ, Describe each torque value θ i The distance from the mean μ;

[0017] Multiple likelihood values ​​for one load case are multiplied together to obtain the overall likelihood value.

[0018] In a preferred embodiment, the MH algorithm is combined with the likelihood function and the probability model to continuously generate new candidate samples, including the following steps:

[0019] The initial candidate samples are randomly selected. The initial candidate samples are the estimated torque mean and torque standard deviation. The proposed distribution of the mean and standard deviation is defined based on the normal distribution. The proposed distribution of the mean is: is the candidate torque mean, μ (t) is the current torque mean, is the standard deviation of the normal distribution, To control the fluctuation range of the mean torque, the standard deviation is proposed to be distributed as: is the standard deviation of the candidate torque, σ (t) is the current torque standard deviation, is the normal distribution standard deviation, which is used to control the fluctuation range of the torque standard deviation. N(*) represents the normal distribution;

[0020] The candidate torque mean is sampled from the torque mean proposal distribution and the torque standard deviation proposal distribution and the standard deviation of the candidate torque The expression is: ∈ μ is the torque mean disturbance term, from Generated in, ∈ σ is the torque standard deviation disturbance term, from The candidate torque mean is generated and the standard deviation of the candidate torque Combined into new candidate samples.

[0021] In a preferred embodiment, after calculating the acceptance probability of each pair of candidate samples, deciding whether to accept the candidate samples is done by a random process, including the following steps:

[0022] After obtaining all newly generated candidate samples, the acceptance probability of the candidate samples is calculated, and the expression is: In the formula, α is the acceptance probability, is the overall likelihood value of the new candidate sample, is the prior distribution of candidate samples, P(Data|μ (t) ,σ (t) ) is the overall likelihood value of the current candidate sample, P(μ (t) ,σ (t) ) is the prior distribution of the current candidate sample, min(*) means selecting the minimum value;

[0023] Draw a random number u from the uniform distribution N(0,1) and compare the random number u with the acceptance probability α of the candidate sample;

[0024] If u≤α, the candidate sample is accepted, that is

[0025] If u>α, the candidate sample is rejected, that is, (μ (t+1) ,σ (t+1) )=(μ (t) ,σ (t) );

[0026] If the candidate sample is accepted, the sample chain of the torque mean and standard deviation is updated; if the candidate sample is rejected, the current sample chain remains unchanged.

[0027] In a preferred embodiment, screening of initially unstable candidate samples during the burn-in period includes the following steps:

[0028] Obtain several torque means and overall likelihood values ​​of each candidate sample during the iteration process, analyze the fluctuation degree of the candidate sample torque mean to obtain the torque mean fluctuation amplitude, and analyze the overall likelihood value of the candidate sample to calculate the overall likelihood mean;

[0029] The torque mean fluctuation amplitude and the overall likelihood mean are comprehensively calculated to obtain the screening coefficient of the candidate samples. The expression is: sx = α*ML-β*AQ, where α is the screening coefficient, ML is the overall likelihood mean, AQ is the torque mean fluctuation amplitude, α and β are the adjustment coefficients of the overall likelihood mean and the torque mean fluctuation amplitude, respectively, and α and β are both greater than 0;

[0030] After obtaining the screening coefficients of all candidate samples, the smaller the screening coefficient, the more unstable the candidate sample. After sorting all candidate samples from small to large according to the screening coefficient, mark the length of the burn-in period as B, discard the first B candidate samples, and retain the remaining candidate samples.

[0031] In a preferred embodiment, the GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence condition is met, the posterior distribution of the target parameter is constructed using the retained candidate samples, and the motor torque is analyzed and predicted based on the posterior distribution, including the following steps:

[0032] The retained candidate samples are organized into sample chains, and the sample chains are divided into Y independent chains. The standard deviation within each independent chain and the standard deviation between independent chains are calculated respectively. The GR statistic is calculated based on the standard deviation within the independent chain and the standard deviation between independent chains. The expression is:

[0033] In the formula, k is the number of samples in the independent chain, is the GR statistic, C is the standard deviation between independent chains, and W is the standard deviation within independent chains;

[0034] like If the convergence condition is not met, sampling needs to continue. Determine whether the convergence condition is met and organize the retained candidate samples into a data structure;

[0035] The torque mean and torque standard deviation of the posterior distribution are calculated using the retained candidate samples, the torque mean and torque standard deviation of the posterior distribution are visualized through a histogram, and several groups of samples are extracted from the posterior distribution, each group of samples includes the torque mean and torque standard deviation;

[0036] The torque mean of each group of samples is compared with the torque stability range Z min ~Z max For comparison, the samples whose torque mean is within the torque stability range are marked as S w , mark the samples whose torque mean is not within the torque stability range as S x , predict the probability of abnormal occurrence of the motor, the expression is: yc l is the probability of abnormal occurrence;

[0037] The torque standard deviation of each group of samples is compared with the standard deviation threshold, and the samples with torque standard deviation greater than the standard deviation threshold are marked as Q x , mark the samples whose torque standard deviation is less than or equal to the standard deviation threshold as Q w , calculate the motor reliability score, the expression is: kp l Score reliability;

[0038] If the probability of abnormal occurrence is greater than the probability threshold or the reliability score is less than the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a poor set;

[0039] If the probability of abnormal occurrence is greater than the probability threshold, and the reliability score is greater than or equal to the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a medium set;

[0040] If the probability of abnormal occurrence is less than or equal to the probability threshold, and the reliability score is less than the score threshold, it is analyzed that the motor has passed the test this time, and it needs to be monitored more closely, and the motor is classified as a good set;

[0041] If the probability of abnormal occurrence is less than or equal to the occurrence probability threshold, and the reliability score is greater than or equal to the score threshold, it is analyzed that the motor has passed this test and the motor is classified into the excellent set.

[0042] A motor manufacturing rotation torque test system, comprising a construction module, a sample generation module, and an analysis module;

[0043] Building module: According to the physical characteristics of motor torque, a preliminary probability model is established based on normal distribution. During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed based on the torque data under different loads.

[0044] Sample generation module: Continuously generate new candidate samples through the MH algorithm combined with the likelihood function and probability model. After calculating the acceptance probability of each pair of candidate samples, decide whether to accept the candidate samples through a random process;

[0045] Analysis module: When entering the burn-in period, the initially unstable candidate samples are screened, and the GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0047] The present invention establishes a preliminary probability model based on normal distribution, obtains the torque data of the motor under different loads during the test, constructs a likelihood function based on the torque data under different loads, continuously generates new candidate samples by combining the likelihood function and the probability model through the MH algorithm, and uses the GR diagnostic method to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution. By constructing the likelihood function and posterior distribution for the data under different load conditions, the test system can flexibly adapt to the complex changes of torque under various load conditions, and provide more accurate parameter estimation and distribution characteristic analysis, and filter noise interference, effectively improving the accuracy and stability of the test data. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0049] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example 1: Please refer to Figure 1As shown, the present embodiment provides a method for testing the rotation torque of a motor, and the testing method includes the following steps:

[0052] According to the known physical characteristics of the motor torque, the test system establishes a preliminary probability model based on the normal distribution. During the test, the torque data of the motor under different loads are obtained, and the likelihood function is constructed based on the torque data under different loads. The likelihood function is used to describe the possibility of the torque data under different loads, which will be used to calculate the acceptance probability during sampling. New candidate samples are continuously generated by combining the likelihood function and the probability model through the MH (Metropolis-Hastings) algorithm. After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate samples. This sampling process gradually generates new parameter value samples to establish a posterior distribution consistent with the candidate samples. When entering the burn-in period, the initially unstable candidate samples are screened, and the GR (Gelman-Rubin) diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameter, and the motor torque is analyzed and predicted based on the posterior distribution.

[0053] This application establishes a preliminary probability model based on the normal distribution, obtains the torque data of the motor under different loads during the test, constructs a likelihood function based on the torque data under different loads, continuously generates new candidate samples through the MH algorithm combined with the likelihood function and the probability model, and uses the GR diagnostic method to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution. By constructing likelihood functions and posterior distributions for data under different load conditions, the test system can flexibly adapt to the complex changes in torque under various load conditions, provide more accurate parameter estimation and distribution characteristic analysis, and filter noise interference, effectively improving the accuracy and stability of test data.

[0054] Embodiment 2: The test system establishes a preliminary probability model based on the normal distribution according to the known physical characteristics of the motor torque, including the following steps:

[0055] Obtain the torque fluctuation range of similar motors based on the historical test data of the motor [D min ~D max ], where D min Indicates the minimum torque of the motor, D max It represents the maximum torque of the motor. According to the torque fluctuation range of similar motors and historical test data, the mean torque and standard deviation of the motor are estimated, and then the prior distribution of the mean torque and standard deviation of the torque is defined. The probability model is established by combining the prior distribution, the mean torque and standard deviation of the motor.

[0056] In order to better illustrate the above solution, this application example is as follows:

[0057] Suppose we are testing the torque of a motor. According to the test goal, we want to estimate the mean and standard deviation of the motor torque. Since the motor torque is usually distributed around a mean value and the deviation is relatively stable, it can be assumed to follow a normal distribution.

[0058] Through the analysis of the historical test data of the motor, we found that the motor torque usually fluctuates between 40 and 60Nm. Based on these empirical data, we preliminarily estimate that the average value of the motor torque may be close to 50Nm and the standard deviation may be around 5Nm.

[0059] Assume that the prior distribution of the torque mean μ is: μ~N(50,10 2 ). This means that we think the mean is likely to be around 50Nm, but allow it to fluctuate within a wide range.

[0060] Assume that the prior distribution of the standard deviation σ is: 2 ~Inverse-Gamma(2,1), which gives us a rough idea of ​​the standard deviation while allowing smaller values ​​of the standard deviation to be more likely.

[0061] The probability model is established by combining the prior distribution, the torque mean and the torque standard deviation of the motor.

[0062] The initially established probability model defines the probability distribution characteristics of the motor torque and lays the foundation for the subsequent steps of the test system. Subsequently, torque data sampling and likelihood function construction will be carried out based on these parameters.

[0063] During the test, the torque data of the motor under different loads is obtained, and a likelihood function is constructed based on the torque data under different loads. The likelihood function is used to describe the possibility of the torque data under different loads, and will be used to calculate the acceptance probability during sampling, including the following steps:

[0064] Carry out motor torque test under different load conditions and record the torque value of the motor under each load. Ensure the consistency of experimental conditions during the acquisition process to reduce the impact of external factors on the data.

[0065] The collected torque data is organized into a structured format. The torque data under different load conditions can usually be stored as a data set, including the torque value under each load condition and the corresponding load information.

[0066] Based on the sorted torque data, the corresponding likelihood function is constructed for different load conditions. Usually, it is assumed that the torque data obeys the normal distribution under given load conditions, so the likelihood function can be expressed in the form of normal distribution, and the function expression is: Where L is the likelihood value, μ is the mean torque, σ is the standard deviation of the torque, and θ i represents the i-th torque value obtained under the test under specific load conditions, n represents the total number of torque data collected under the specific load, is the product symbol, indicating that for all torque values ​​θ i Since each observation is independent under a given parameter, their likelihood values ​​can be multiplied together to obtain the overall likelihood function, It represents the probability density of torque data under a given torque standard deviation σ. This part is used for standardization so that the probability density function integrates to 1 over its domain. Describe each torque value θ i The distance from the mean μ, the greater the distance, the smaller the probability density value, indicating that the possibility of observing this torque value is lower under the given mean and standard deviation;

[0067] Since the motor is tested multiple times under one load condition to obtain multiple torque values, a likelihood value is calculated corresponding to each torque value, and the multiple likelihood values ​​under one load condition are multiplied to obtain the overall likelihood value.

[0068] The size of the overall likelihood value: This value indicates the likelihood of the observed torque data under specific parameters (mean and standard deviation). A higher likelihood value indicates that the model fits the data well, while a lower likelihood value indicates that the model may not be suitable for the data.

[0069] New candidate samples are continuously generated by combining the likelihood function and the probability model through the MH (Metropolis-Hastings) algorithm, including the following steps:

[0070] The initial candidate samples are randomly selected. The initial candidate samples are the estimated torque mean and torque standard deviation. The proposed distribution of the mean and standard deviation is defined based on the normal distribution. The proposed distribution of the mean is: is the candidate torque mean, μ (t) is the current torque mean, is the standard deviation of the normal distribution, To control the fluctuation range of the mean torque, the standard deviation is proposed to be distributed as: is the standard deviation of the candidate torque, σ (t) is the current torque standard deviation, is the normal distribution standard deviation, which is used to control the fluctuation range of the torque standard deviation. N(*) represents the normal distribution;

[0071] The candidate torque mean is sampled from the torque mean proposal distribution and the torque standard deviation proposal distribution and the standard deviation of the candidate torque The expression is: ∈ μ is the torque mean disturbance term, from Generated in, ∈ σ is the torque standard deviation disturbance term, from The candidate torque mean is generated and the standard deviation of the candidate torque Combined into new candidate samples.

[0072] After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate sample. This sampling process gradually generates new parameter value samples to establish a posterior distribution that matches the candidate sample. It includes the following steps:

[0073] After obtaining all newly generated candidate samples, the acceptance probability of the candidate samples is calculated, and the expression is: In the formula, α is the acceptance probability, is the overall likelihood value of the new candidate sample, is the prior distribution of candidate samples, P(Data|μ (i) ,σ (t) ) is the overall likelihood value of the current candidate sample, P(μ (t) ,σ (t) ) is the prior distribution of the current candidate sample, min(*) means selecting the minimum value;

[0074] Draw a random number u from the uniform distribution N(0,1) and compare the random number u with the acceptance probability α of the candidate sample;

[0075] If u≤α, the candidate sample is accepted, that is

[0076] If u>α, the candidate sample is rejected, that is, (μ (t+1) ,σ (t+1) )=(μ (t) ,σ (t) );

[0077] If the candidate sample is accepted, the sample chain of the torque mean and standard deviation is updated; if the candidate sample is rejected, the current sample chain remains unchanged.

[0078] When entering the burn-in period, the initial unstable candidate samples are screened, including the following steps:

[0079] Obtain several torque means and overall likelihood values ​​during the iteration process of each candidate sample, analyze the fluctuation degree of the candidate sample torque mean to obtain the torque mean fluctuation amplitude, analyze the overall likelihood value of the candidate sample to calculate the overall likelihood mean, the expression is: In the formula, ML is the overall likelihood mean, AQ is the torque mean fluctuation amplitude, n is the total number of candidate samples to obtain the overall likelihood mean, m is the total number of candidate samples to obtain the torque mean, L i is the ith overall likelihood value, avg i is the i-th torque mean, is the average value of the torque.

[0080] The torque mean fluctuation amplitude and the overall likelihood mean are calculated comprehensively to obtain the screening coefficient of the candidate samples. The expression is: sx = α*ML-β*AQ, where is the screening coefficient, ML is the overall likelihood mean, AQ is the torque mean fluctuation amplitude, α and β are the adjustment coefficients of the overall likelihood mean and the torque mean fluctuation amplitude, respectively, and α and β are both greater than 0.

[0081] After obtaining the screening coefficients of all candidate samples, the smaller the screening coefficient, the more unstable the candidate sample. After sorting all candidate samples from small to large according to the screening coefficient, mark the length of the burn-in period as B, discard the first B candidate samples, and retain the remaining candidate samples.

[0082] The GR (Gelman-Rubin) diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence condition is met, the posterior distribution of the target parameter is constructed using the retained candidate samples, and the motor torque is analyzed and predicted based on the posterior distribution, including the following steps:

[0083] The retained candidate samples are organized into sample chains, and the sample chains are divided into Y independent chains. The standard deviation within each independent chain and the standard deviation between independent chains are calculated respectively. The GR statistic is calculated based on the standard deviation within the independent chain and the standard deviation between independent chains. The expression is:

[0084] In the formula, k is the number of samples in the independent chain, is the GR statistic, C is the standard deviation between independent chains, and W is the standard deviation within independent chains;

[0085] like If the convergence condition is not met, sampling needs to continue. Determine whether the convergence condition is met and organize the retained candidate samples into a data structure;

[0086] Calculate the torque mean and torque standard deviation of the posterior distribution using the retained candidate samples;

[0087] In the formula, is the torque mean of the posterior distribution, is the torque standard deviation of the posterior distribution, J is the number of candidate samples retained, θ iis the torque mean of the i-th candidate sample.

[0088] The torque mean and standard deviation of the posterior distribution are visualized through histograms to help intuitively understand the distribution characteristics of the parameters;

[0089] Let the posterior distribution be: θ represents the motor torque, is the torque mean of the posterior distribution, is the torque standard deviation of the posterior distribution, from the posterior distribution Several groups of samples are extracted from the sample set, each of which includes the torque mean and torque standard deviation. The torque mean of each group of samples is compared with the torque stability range Z. min ~Z max For comparison, Z min Represents the minimum torque threshold, Z max represents the maximum torque threshold, and the samples with the mean torque within the torque stability range are marked as S w , mark the samples whose torque mean is not within the torque stability range as S x , predict the probability of abnormal occurrence of the motor, the expression is: yc l The torque standard deviation of each group of samples is compared with the standard deviation threshold, and the samples with torque standard deviation greater than the standard deviation threshold are marked as Q x , mark the samples whose torque standard deviation is less than or equal to the standard deviation threshold as Q w , calculate the motor reliability score, the expression is: kp l Score reliability;

[0090] If the probability of abnormal occurrence is greater than the probability threshold or the reliability score is less than the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a poor set;

[0091] If the probability of abnormal occurrence is greater than the probability threshold, and the reliability score is greater than or equal to the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a medium set;

[0092] If the probability of abnormal occurrence is less than or equal to the probability threshold, and the reliability score is less than the score threshold, it is analyzed that the motor has passed the test this time, and it needs to be monitored more closely, and the motor is classified as a good set;

[0093] If the probability of abnormal occurrence is less than or equal to the probability threshold, and the reliability score is greater than or equal to the score threshold, the motor is analyzed to have passed the test and is classified as an excellent set;

[0094] In the present application, when returning for maintenance, the maintenance priority of the poor set motor is higher than the medium set priority.

[0095] Embodiment 3: A motor manufacturing rotation torque testing system described in this embodiment includes a construction module, a sample generation module, and an analysis module;

[0096] Construction module: According to the known physical characteristics of the motor torque, a preliminary probability model is established based on the normal distribution. During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed based on the torque data under different loads. The likelihood function is used to describe the possibility of the torque data under different loads, which will be used to calculate the acceptance probability during sampling. The probability model and likelihood function are sent to the sample generation module;

[0097] Sample generation module: New candidate samples are continuously generated through the MH (Metropolis-Hastings) algorithm combined with the likelihood function and the probability model. After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate sample. This sampling process gradually generates new parameter value samples to establish a posterior distribution that matches the candidate sample. The accepted candidate sample information is sent to the analysis module;

[0098] Analysis module: When entering the burn-in period, the initially unstable candidate samples are screened, and the GR (Gelman-Rubin) diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution.

[0099] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0100] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0101] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for testing the rotation torque of a motor, characterized in that: The test method comprises the following steps: The test system establishes a preliminary probability model based on the normal distribution according to the physical characteristics of the motor torque. During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed based on the torque data under different loads. New candidate samples are continuously generated by combining the likelihood function and the probability model through the MH algorithm. After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate samples. When entering the burn-in period, the initially unstable candidate samples are screened, and the GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution.

2. A motor manufacturing rotation torque test method according to claim 1, characterized in that: The test system establishes a preliminary probability model based on the normal distribution according to the known physical characteristics of the motor torque, including the following steps: Obtain the torque fluctuation range of similar motors based on the historical test data of the motor [D min ~D max ], where D min Indicates the minimum torque of the motor, D max It represents the maximum torque of the motor. According to the torque fluctuation range of similar motors and historical test data, the mean torque and standard deviation of the motor are estimated, and then the prior distribution of the mean torque and standard deviation of the torque is defined. The probability model is established by combining the prior distribution, the mean torque and standard deviation of the motor.

3. A motor manufacturing rotation torque test method according to claim 2, characterized in that: During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed according to the torque data under different loads, including the following steps: Carry out motor torque test under different load conditions and record the torque value of the motor under each load; The collected torque data is organized into a structured format, and the torque data under different load conditions is stored as a data set, which includes the torque value and corresponding load information under each load condition; Based on the sorted torque data, the corresponding likelihood function is constructed for different load conditions. The function expression is: Where L is the likelihood value, μ is the mean torque, σ is the standard deviation of the torque, and θ i represents the i-th torque value obtained under the test under specific load conditions, n represents the total number of torque data collected under the specific load, is the product symbol, represents the probability density of torque data under a given torque standard deviation σ, Describe each torque value θ i The distance from the mean μ; Multiple likelihood values ​​for one load case are multiplied together to obtain the overall likelihood value.

4. A motor manufacturing rotation torque test method according to claim 3, characterized in that: New candidate samples are continuously generated by combining the MH algorithm with the likelihood function and the probability model, including the following steps: The initial candidate samples are randomly selected. The initial candidate samples are the estimated torque mean and torque standard deviation. The proposed distribution of the mean and standard deviation is defined based on the normal distribution. The proposed distribution of the mean is: is the candidate torque mean, μ(t) is the current torque mean, is the standard deviation of the normal distribution, To control the fluctuation range of the mean torque, the standard deviation is proposed to be distributed as: is the standard deviation of the candidate torque, σ (t) is the current torque standard deviation, is the normal distribution standard deviation, which is used to control the fluctuation range of the torque standard deviation. N(*) represents the normal distribution; The candidate torque mean is sampled from the torque mean proposal distribution and the torque standard deviation proposal distribution and the standard deviation of the candidate torque The expression is: ∈ μ is the torque mean disturbance term, from Generated in, ∈ σ is the torque standard deviation disturbance term, from The candidate torque mean is generated and the standard deviation of the candidate torque Combined into new candidate samples.

5. A motor manufacturing rotation torque test method according to claim 4, characterized in that: After calculating the acceptance probability of each pair of candidate samples, a random process is used to decide whether to accept the candidate samples, including the following steps: After obtaining all newly generated candidate samples, the acceptance probability of the candidate samples is calculated, and the expression is: In the formula, α is the acceptance probability, is the overall likelihood value of the new candidate sample, is the prior distribution of candidate samples, P(Data / μ (t) , σ (t) ) is the overall likelihood value of the current candidate sample, P(μ (t) , σ (t) ) is the prior distribution of the current candidate sample, min(*) means selecting the minimum value; Draw a random number u from the uniform distribution N(0,1), and compare the random number u with the acceptance probability α of the candidate sample; If u≤α, the candidate sample is accepted, that is If u>α, the candidate sample is rejected, that is If the candidate sample is accepted, the sample chain of the torque mean and standard deviation is updated; if the candidate sample is rejected, the current sample chain remains unchanged.

6. A motor manufacturing rotation torque test method according to claim 5, characterized in that: When entering the burn-in period, the initial unstable candidate samples are screened, including the following steps: Obtain several torque means and overall likelihood values ​​of each candidate sample during the iteration process, analyze the fluctuation degree of the candidate sample torque mean to obtain the torque mean fluctuation amplitude, and analyze the overall likelihood value of the candidate sample to calculate the overall likelihood mean; The torque mean fluctuation amplitude and the overall likelihood mean are comprehensively calculated to obtain the screening coefficient of the candidate samples. The expression is: sx = α*ML-β*AQ, where α is the screening coefficient, ML is the overall likelihood mean, AQ is the torque mean fluctuation amplitude, α and β are the adjustment coefficients of the overall likelihood mean and the torque mean fluctuation amplitude, respectively, and α and β are both greater than 0; After obtaining the screening coefficients of all candidate samples, the smaller the screening coefficient, the more unstable the candidate sample. After sorting all candidate samples from small to large according to the screening coefficient, mark the length of the burn-in period as B, discard the first B candidate samples, and retain the remaining candidate samples.

7. A motor manufacturing rotation torque test method according to claim 6, characterized in that: The GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence condition is met, the posterior distribution of the target parameter is constructed using the retained candidate samples. The motor torque is analyzed and predicted based on the posterior distribution, including the following steps: The retained candidate samples are organized into sample chains, and the sample chains are divided into Y independent chains. The standard deviation within each independent chain and the standard deviation between independent chains are calculated respectively. The GR statistic is calculated based on the standard deviation within the independent chain and the standard deviation between independent chains. The expression is: In the formula, k is the number of samples in the independent chain, is the GR statistic, C is the standard deviation between independent chains, and W is the standard deviation within independent chains; like If the convergence condition is not met, sampling needs to be continued. Determine whether the convergence condition is met and organize the retained candidate samples into a data structure; The torque mean and torque standard deviation of the posterior distribution are calculated using the retained candidate samples, the torque mean and torque standard deviation of the posterior distribution are visualized through a histogram, and several groups of samples are extracted from the posterior distribution, each group of samples includes the torque mean and torque standard deviation; The torque mean of each group of samples is compared with the torque stability range Z min ~Z max For comparison, the samples whose torque mean is within the torque stability range are marked as S W , mark the samples whose torque mean is not within the torque stability range as S x , predict the probability of abnormal occurrence of the motor, the expression is: yc l is the probability of abnormal occurrence; The torque standard deviation of each group of samples is compared with the standard deviation threshold, and the samples with torque standard deviation greater than the standard deviation threshold are marked as Q x , mark the samples whose torque standard deviation is less than or equal to the standard deviation threshold as Q w , calculate the motor reliability score, the expression is: kp l Score reliability; If the probability of abnormal occurrence is greater than the probability threshold or the reliability score is less than the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a poor set; If the probability of abnormal occurrence is greater than the probability threshold, and the reliability score is greater than or equal to the score threshold, the motor is analyzed to have failed the test and needs to be returned for repair, and the motor is classified as a medium set; If the probability of abnormal occurrence is less than or equal to the probability threshold, and the reliability score is less than the score threshold, it is analyzed that the motor has passed the test this time, and it needs to be monitored more closely, and the motor is classified as a good set; If the probability of abnormal occurrence is less than or equal to the occurrence probability threshold, and the reliability score is greater than or equal to the score threshold, the motor is analyzed to have passed this test and is classified as an excellent set.

8. A motor manufacturing rotation torque test system, used to implement the test method according to any one of claims 1 to 7, characterized in that: It includes construction module, sample generation module and analysis module; Building module: According to the physical characteristics of motor torque, a preliminary probability model is established based on normal distribution. During the test, the torque data of the motor under different loads is obtained, and the likelihood function is constructed based on the torque data under different loads. Sample generation module: Continuously generate new candidate samples through the MH algorithm combined with the likelihood function and probability model. After calculating the acceptance probability of each pair of candidate samples, decide whether to accept the candidate samples through a random process; Analysis module: When entering the burn-in period, the initially unstable candidate samples are screened, and the GR diagnostic method is used to check the convergence of the sample chain. When it is judged that the convergence conditions are met, the retained candidate samples are used to construct the posterior distribution of the target parameters, and the motor torque is analyzed and predicted based on the posterior distribution.

Images