Neodymium-iron-boron magnet processing control method and system

Abstract

The present application belongs to the technical field of magnet processing control, and particularly relates to a neodymium-iron-boron magnet processing control method and system. The method comprises the following steps: calculating a phase change region thermal disturbance factor according to the relative deviation of the dynamic heat flux integral of the time period composed of the start time and the end time of the magnet finished product compared with the qualified finished product, and correcting the information entropy of the temperature change amount of all time windows of the thermal history temperature curve of the magnet finished product to obtain the thermal history structure entropy of the magnet finished product, and combining the peak temperature and the total processing time to form a dynamic thermal history feature vector of the magnet finished product, which is used to train a Gaussian process regression model capable of predicting magnetic properties, and obtain the magnetic property prediction mean and the magnetic property prediction variance of the just processed magnet finished product, which is used to determine whether the just processed magnet finished product is taken as a product to be inspected and is subjected to re-inspection. The present application improves the objectivity and comprehensiveness of feature extraction and the prediction accuracy of magnetic properties.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of magnet processing control. More particularly, the present application relates to a neodymium iron boron magnet processing control method and system. BACKGROUND

[0002] Neodymium iron boron (NdFeB) permanent magnet material, as the third generation of rare earth permanent magnet material, is widely used in consumer electronics, new energy vehicles, wind power and other key fields due to its excellent magnetic properties.

[0003] The final magnetic properties of neodymium iron boron magnets, especially their coercivity and remanence, show a very high sensitivity to the thermal history they experience during the production process; temperature fluctuations at any stage, from sintering, aging treatment to subsequent machining cooling, can lead to significant differences and inconsistencies in the magnetic properties of the final product.

[0004] In order to achieve accurate control of magnet performance, the industry has begun to introduce advanced machine learning algorithms; among them, Gaussian Process Regression (GPR) is considered a very promising control technology because it can quantify the uncertainty of the prediction result while giving the predicted value; by establishing a Gaussian Process Regression model between the thermal history and the final magnetic properties, the magnet performance can be predicted and the process adjustment can be guided.

[0005] However, there is a core technical problem in the existing control method based on Gaussian Process Regression: the Gaussian Process Regression model itself cannot directly process raw, continuous time series data; therefore, in the scenario of predicting magnet performance, it cannot directly process raw, continuous thermal history temperature series; the application premise is that a set of discrete, fixed features must be manually extracted from the continuous thermal history temperature series, such as peak temperature, specific temperature interval residence time, etc.; this feature extraction method is highly dependent on the prior knowledge and subjective experience of engineers, limiting the prediction accuracy of the Gaussian Process Regression model and restricting its ability to achieve higher consistency and customized control in high-end magnet manufacturing. SUMMARY

[0006] To solve the technical problem that the Gaussian Process Regression is highly dependent on discrete features extracted according to the prior knowledge and subjective experience of engineers, limiting the prediction accuracy of the Gaussian Process Regression model, the present application provides solutions in the following aspects.

[0007] In a first aspect, the present application provides a neodymium iron boron magnet processing control method, comprising:

[0008] Obtain the start and end times of the thermal history temperature curve of the finished magnet entering and leaving the critical phase change temperature interval; calculate the dynamic heat flux integral of the finished magnet in the time period consisting of the start and end times based on the thermal history temperature curve; calculate the phase change zone thermal disturbance factor of the finished magnet based on the relative deviation of the dynamic heat flux integral of the finished magnet compared to the qualified finished magnet in the time period consisting of the start and end times;

[0009] The thermal history temperature curve of the finished magnet is divided into multiple time windows of equal length, and the temperature change in each time window is calculated; according to the thermal disturbance factor of the phase change zone of the finished magnet, the information entropy of the temperature change in all time windows is corrected to obtain the thermal history structural entropy of the finished magnet;

[0010] The thermal history structural entropy, peak temperature, and total processing time of the finished magnet product are combined to form a dynamic thermal history feature vector of the finished magnet product. A Gaussian process regression model is trained based on the dynamic thermal history feature vector and magnetic properties of the finished magnet product in historical production data.

[0011] The trained Gaussian process regression model is used to obtain the predicted mean and variance of the magnetic properties of the newly processed finished magnets, which are used to determine whether the newly processed finished magnets should be treated as products to be inspected and re-inspected.

[0012] The present invention realizes the automatic and objective extraction of high-quality features from raw data by constructing a series of interrelated dynamic indicators such as dynamic heat flux integral, phase change zone thermal disturbance factor and thermal history structure entropy, avoiding the subjectivity and one-sidedness of manual feature selection, and can capture deep dynamic information that is easily overlooked by traditional methods, thereby improving the objectivity and comprehensiveness of feature extraction.

[0013] Furthermore, a dynamic eigenvector with greater physical significance and information completeness is used to construct a Gaussian process regression model capable of predicting magnetic properties. This enables the Gaussian process regression model constructed by the present invention to more deeply understand the complex nonlinear relationship between thermal history and final magnetic properties, thereby significantly improving the accuracy of magnetic property prediction and providing a solid foundation for achieving high consistency in magnetic properties.

[0014] Preferably, the thermal history temperature curve is composed of temperature data continuously collected at each key stage of the entire process of producing the finished magnet; and the key phase change temperature range is determined according to the NdFeB material corresponding to the finished magnet.

[0015] Preferably, the step of obtaining the start and end time of the heat history temperature curve of the finished magnet entering and leaving the key phase change temperature range comprises: for the temperature of the finished magnet at each time in the heat history temperature curve, if the finished magnet is at time The temperature is greater than or equal to and the temperature at time is less than , the time is taken as the starting time when the thermal history temperature curve enters the critical phase transition temperature interval; if the temperature of the magnet product at time is greater than and the temperature at time is equal to or less than , the time is taken as the ending time when the thermal history temperature curve leaves the critical phase transition temperature interval. , are the lower limit and the upper limit of the temperature of the critical phase transition temperature interval, respectively.

[0016] Preferably, based on the thermal history temperature curve, the dynamic heat flux integral of the magnet product in the time period composed of the starting time and the ending time is calculated, including:

[0017] ; in the formula, is the dynamic heat flux integral of the magnet product in the time period ; the time period is composed of the starting time and the ending time when the thermal history temperature curve of the magnet product enters and leaves the critical phase transition temperature interval; is a weight function, is the temperature value at time ; is the first derivative of temperature with respect to time.

[0018] The present application can better distinguish the two processes of short-time severe cooling and long-time mild cooling which have completely different effects on the internal stress of the magnet by constructing the dynamic heat flux integral as an index to quantify the severity of the thermal shock suffered by the magnet product in a specific time period.

[0019] Preferably, the calculation formula of the weight function is:

[0020] ; in the formula, is a natural exponential function; represents the temperature standard deviation, and , , are the lower limit and the upper limit of the temperature of the critical phase transition temperature interval, respectively; is the Curie temperature of the processed Nd-Fe-B material.

[0021] Preferably, the calculation of the phase change zone thermal disturbance factor of the finished magnet includes: the qualified finished product refers to the finished magnet corresponding to the qualified NdFeB magnet; the phase change zone thermal disturbance factor of the finished magnet is The calculation formula is:

[0022] Where, For the finished magnet in the time period [ ] dynamic heat flux integral; The time period for qualified finished products[ ] dynamic heat flux integral, The time period for all qualified finished products[ ] is the mean value of the dynamic heat flux integral; [ ] is the starting time when the thermal history temperature curve of the qualified finished product enters and leaves the key phase change temperature range and end time The composition time period.

[0023] The present invention obtains the thermal disturbance factor of the phase change zone of the finished magnet by analyzing the dynamic heat flux integral during the cooling process. This can reflect the degree of thermal disturbance of the actual processing process compared with the ideal process, and further refine the description of the heat treatment stability of the finished magnet in the key phase change temperature range.

[0024] Preferably, the information entropy of the temperature variation in all time windows is corrected according to the thermal disturbance factor of the phase change zone of the finished magnet to obtain the thermal history structural entropy of the finished magnet, including:

[0025] Where, is the thermal history structural entropy of the finished magnet; is the preset quantity; For the The proportion of temperature change within a time window, is the information entropy of the temperature change in all time windows; is the thermal disturbance factor of the phase change zone of the finished magnet; Indicates taking the absolute value.

[0026] The present invention combines the thermal disturbance of the local key area, namely the thermal disturbance factor of the phase change zone, with the morphological change of the global curve, namely the information entropy, to form a characteristic, namely the thermal history structural entropy, which can comprehensively evaluate the global complexity and disorder of the entire thermal history curve, and is used to characterize the quality of the thermal history of the finished magnet.

[0027] Preferably, the temperature variation within the time window is equal to the difference between the maximum and minimum values ​​of the temperature within the time window; The percentage of temperature change within a time window wherein, , are the temperature variation amounts in the first , the second time windows, respectively.

[0028] Preferably, the judging whether the just-processed magnet product is the product to be inspected and performing re-inspection includes: judging whether a magnetic property predicted mean value of the just-processed magnet product is within a target range of the magnetic property and whether a magnetic property predicted variance of the just-processed magnet product exceeds a preset threshold value; if the magnetic property predicted mean value of the just-processed magnet product is within the target range of the magnetic property and the magnetic property predicted variance is less than the preset threshold value, it is indicated that the just-processed magnet product meets the requirements; otherwise, it is indicated that the just-processed magnet product is likely to not meet the process requirements, the just-processed magnet product is marked as the product to be inspected, and the magnetic property of the product to be inspected is re-inspected.

[0029] The present application has stronger self-adaptive ability for different neodymium iron boron materials, and through quantitative evaluation of the quality of the thermal history and combined uncertainty output of the Gaussian process regression model, the control decision is more reliable, the intelligentization and robustness of the control method are enhanced, and the waste rate caused by process fluctuation is effectively reduced.

[0030] In the second aspect, the present application provides a neodymium iron boron magnet processing control system, comprising a processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the above-mentioned neodymium iron boron magnet processing control method is realized.

[0031] By adopting the above technical scheme, the above-mentioned neodymium iron boron magnet processing control method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal equipment is manufactured according to the memory and the processor, and the use is convenient.

[0032] The present application has the advantages of:

[0033] The present application realizes automatic and objective extraction from original data to high-quality features by constructing a series of interrelated dynamic indexes such as dynamic heat flux integral, phase change region thermal disturbance factor and thermal history structure entropy, avoids subjectivity and one-sidedness of manual feature selection, can capture deep dynamic information easily ignored by traditional methods, and improves objectivity and comprehensiveness of feature extraction.

[0034] Further, a dynamic feature vector with more physical meaning and information completeness is adopted to construct a Gaussian process regression model capable of predicting the magnetic performance, so that the Gaussian process regression model constructed by the application can more deeply understand the complex nonlinear relationship between the thermal history and the final magnetic performance, thereby significantly improving the prediction accuracy of the magnetic performance and providing a solid foundation for realizing high consistency of the magnetic performance. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 is a flowchart schematically showing a neodymium-iron-boron magnet processing control method in the application;

[0036] Figure 2 is a process flow of a neodymium-iron-boron magnet.

[0037] Figure 3 is a flowchart schematically showing step S2. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0039] The specific embodiments of the application will be described in detail below with reference to the drawings.

[0040] The embodiments of the application disclose a neodymium-iron-boron magnet processing control method, referring to Figure 1 , comprising steps S1-S4:

[0041] S1, continuously collecting temperature data of the magnet finished product in each key stage of the whole process to form a thermal history temperature curve of the magnet finished product; and determining a key phase transition temperature interval of the magnet finished product according to the corresponding neodymium-iron-boron material of the magnet finished product.

[0042] It should be noted that the manufacturing of the neodymium-iron-boron adopts a powder metallurgy process, that is, raw materials containing a certain ratio, such as neodymium, dysprosium, iron, cobalt, niobium, praseodymium, aluminum, boron iron, etc., are smelted into alloy steel ingots through a medium-frequency induction smelting furnace, then crushed into 3-5 μm (millimeter) powder, and pressed into shape in a magnetic field. The green body after shaping is sintered to be dense in a vacuum sintering furnace and tempered and aged, so that a magnet blank with certain magnetic performance is obtained. After the magnet blank is processed through grinding, drilling, slicing and other processes, and then surface treated, the neodymium-iron-boron magnet required by the user is obtained.

[0043] Exemplarily, the process flow of the neodymium-iron-boron magnet is as shown in Figure 2 .

[0044] It should be further explained that the final magnetic properties of NdFeB magnets, especially their coercive force and remanence, are extremely sensitive to the thermal history they experience during the production process. Therefore, it is necessary to track and manage the thermal history of a single finished magnet from sintering to machining.

[0045] Specifically, on the production line, a unique identification is established for each finished magnet; an array of non-contact infrared temperature sensors is deployed at key workstations such as the entrance and exit of the sintering furnace, the entrance and exit of the tempering and aging furnace, the cooling channel, and machining (including the grinding area, drilling area, and slicing area). The temperature data of each finished magnet at each key stage of the entire process is continuously collected at a fixed sampling frequency (1Hz), forming a time series data from high temperature to room temperature with time as the axis for each finished magnet, which serves as the thermal history temperature curve of each finished magnet.

[0046] Furthermore, the critical phase transition temperature range of NdFeB material is recorded as [ ] This interval is the most sensitive area for magnet microstructure and magnetic domain formation.

[0047] in, 、 represent the left and right boundaries of the critical phase transition temperature range, It is the lower limit of the critical phase transition temperature range, indicating the lowest temperature at which the material begins to undergo a phase transition or a significant change in physical properties. It is the upper limit of the critical phase transition temperature range, indicating the highest temperature at which the phase transition of the material or the significant change of physical properties ends.

[0048] It should be noted that the method for obtaining the critical phase transition temperature range of NdFeB material is well known to those skilled in the art and will not be described in detail here.

[0049] Finally, the critical phase transition temperature range of the processed NdFeB material is used as the critical phase transition temperature range of the finished magnet product.

[0050] S2. Construct the thermal history structure entropy of the finished magnet product based on the thermal history temperature curves and key phase transition temperature ranges of the finished magnet product and qualified finished products.

[0051] It should be noted that in the scenario of predicting the performance of the magnet, the Gaussian process regression model itself cannot directly process the original continuous thermal history temperature sequence, and its application premise is that a set of discrete and fixed features, such as peak temperature and temperature interval residence time, must be artificially extracted from the continuous temperature curve. This feature extraction method is highly dependent on the prior knowledge and subjective experience of engineers, and has two limitations: first, it is easy to overlook non-intuitive but crucial dynamic thermal patterns, such as small oscillations or nonlinear changes in speed during the cooling process; second, the selection of features is subjective and limited, resulting in a feature vector that cannot fully and objectively represent the complete thermal history information. These limitations limit the prediction accuracy of the Gaussian process regression model and restrict its ability to achieve higher consistency and customized control in high-end magnet manufacturing.

[0052] Further, to solve the problem of limited prediction accuracy caused by the Gaussian process regression model relying on artificial feature extraction, the embodiment constructs a dynamic feature that can represent the internal physical process and structural complexity of the thermal history, i.e., the thermal history structural entropy of the magnet product, to replace traditional artificial experience features, thereby improving the prediction accuracy of the Gaussian process regression model and the robustness of the control system.

[0053] The flowchart of step S2 refers to Figure 3 , including steps S201 to S204, specifically:

[0054] S201, according to the thermal history temperature curve of the magnet product and the key phase transition temperature interval, and the starting time and ending time of the thermal history temperature curve of the magnet product and the qualified product entering and leaving the key phase transition temperature interval are obtained respectively.

[0055] According to the thermal history temperature curve of the magnet product and the key phase transition temperature interval, the starting time and ending time of the thermal history temperature curve of the magnet product entering and leaving the key phase transition temperature interval are obtained.

[0056] Specifically, for the temperature of the magnet product at each time in the thermal history temperature curve:

[0057] (1) If the temperature of the magnet product at time is greater than or equal to and the temperature of the magnet product at time is less than , the time is taken as the starting time of the thermal history temperature curve entering the key phase transition temperature interval.

[0058] (2) If the temperature of the magnet product at time is greater than and the temperature of the magnet product at time the temperature of the magnet product is equal to or less than the start time is the end time when the thermal history temperature curve of the magnet product leaves the critical phase transition temperature interval.

[0059] wherein, , are the lower and upper limits of the temperature of the critical phase transition temperature interval, respectively.

[0060] Further, according to the start time when the thermal history temperature curve of the magnet product enters the critical phase transition temperature interval and the end time when the thermal history temperature curve of the magnet product leaves the critical phase transition temperature interval, a time period is formed.

[0061] It should be noted that the time period obtained is the specific analysis of the cooling path through the critical phase transition temperature interval of the magnet product in the process of processing the magnet product.

[0062] Further, the magnet products corresponding to the plurality of qualified neodymium-iron-boron magnets are taken as a plurality of qualified products.

[0063] Further, according to the thermal history temperature curve of each qualified product and the critical phase transition temperature interval, the start time and the end time when the thermal history temperature curve of each qualified product enters and leaves the critical phase transition temperature interval are obtained.

[0064] Further, according to the start time when the thermal history temperature curve of the qualified product enters the critical phase transition temperature interval and the end time when the thermal history temperature curve of the qualified product leaves the critical phase transition temperature interval, a time period is formed.

[0065] It should be noted that by analyzing the thermal history temperature curve of the magnet product corresponding to the qualified neodymium-iron-boron magnet and the critical phase transition temperature interval, the time period obtained is the specific analysis of the cooling path through the critical phase transition temperature interval of the qualified product in the process of processing the qualified product.

[0066] S202, based on the thermal history temperature curve, the dynamic heat flux integral of the time period formed by the start time and the end time of the magnet product and the qualified product is calculated, respectively.

[0067] It should be noted that the embodiment quantifies the degree of thermal shock that the magnet product withstands in a specific time period by constructing the dynamic heat flux integral as an index, and the dynamic heat flux integral can reflect the cumulative effect and dynamic characteristics of the temperature change of the magnet product in the processing process.

[0068] Specifically, based on the thermal history temperature curves of the finished magnet product and the qualified finished product, the dynamic heat flux integrals of the finished magnet product and the qualified finished product in the time period consisting of the start time and the end time are calculated respectively.

[0069] Then the finished magnet is produced in the time period [ The calculation formula of the dynamic heat flux integral is:

[0070] ;

[0071] In the formula, [ ] is the starting time when the thermal history temperature curve of the finished magnet enters and leaves the critical phase change temperature range and end time the time period of the composition; For the finished magnet in the time period [ ] dynamic heat flux integral; For the moment The temperature value, is a weight function and is temperature-dependent; It is the first derivative of temperature with respect to time, indicating the instantaneous rate of temperature change.

[0072] The calculation formula of the weight function is:

[0073] ;

[0074] Where, For the moment Temperature value; is the Curie temperature of the processed NdFeB material; is the natural exponential function; represents the temperature standard deviation, and , 、 are the lower and upper temperature limits of the critical phase transition temperature range, respectively.

[0075] Among them, the Curie temperature is the critical point at which a magnetic material changes from ferromagnetism or ferrimagnetism to paramagnetism. Below this temperature, the material behaves as a ferromagnet and the magnetization intensity is stable. Above this temperature, thermal motion destroys the orderly arrangement of the magnetic moments, the material becomes a paramagnet, and the magnetic susceptibility is significantly reduced. The Curie temperature of commercial NdFeB materials is usually between 310°C and 400°C. The method for obtaining the Curie temperature of NdFeB materials is well known to those skilled in the art and will not be repeated here.

[0076] It should be noted that the weight function constructed based on the Curie temperature can highlight the importance of this critical temperature zone.

[0077] It is to be noted that the obtained dynamic heat flux integral is proportional to the intensity and duration of temperature change in the time period , which can better distinguish between short-term severe cooling and long-term mild cooling, two processes that have a completely different impact on the internal stress of the magnet, compared to the average cooling rate, a conventional feature for building Gaussian process regression models. A smooth heat treatment process will correspond to a smaller dynamic heat flux integral .

[0078] Secondly, the calculation formula of the dynamic heat flux integral of the qualified product in the time period is as follows:

[0079] ;

[0080] In the formula, the time period composed of the start time and the end time when the thermal history temperature curve of the qualified product enters and leaves the critical phase transition temperature interval; is the dynamic heat flux integral of the qualified product in the time period ; is the temperature value at time , is a weight function and is related to temperature; is the first derivative of temperature with respect to time, representing the instantaneous temperature change rate.

[0081] S203, according to the relative deviation of the dynamic heat flux integral of the magnet product compared to the qualified product in the time period composed of the start time and the end time, calculate the phase transition zone thermal disturbance factor of the magnet product.

[0082] It is to be noted that the cooling rate has an important influence on the microstructure and magnetic properties of neodymium-iron-boron magnets. Too fast or too slow cooling may lead to uneven grain size, increased residual stress or decreased magnetic properties. By analyzing the dynamic heat flux integral during the cooling process, the heat treatment stability of the magnet product in the critical phase transition temperature interval can be further refined.

[0083] It is further to be noted that the phase transition zone thermal disturbance factor of the magnet product is constructed by the relative deviation of the actual dynamic heat flux from the reference dynamic heat flux, which is used to measure the irregularity of the thermal history temperature curve of the magnet product when passing through the critical phase transition temperature interval.

[0084] Specifically, according to the dynamic heat flux integral of the magnet product in the time period compared to the time period ] is used to calculate the thermal disturbance factor in the phase change zone of the finished magnet.

[0085] The calculation formula for the thermal disturbance factor of the phase change zone of the finished magnet is:

[0086] ;

[0087] Where, is the thermal disturbance factor of the phase change zone of the finished magnet; For the finished magnet in the time period [ ] dynamic heat flux integral; The time period for qualified finished products[ ] dynamic heat flux integral, The time period for all qualified finished products[ ] is the mean value of the dynamic heat flux integral; [ ] is the starting time when the thermal history temperature curve of the finished magnet enters and leaves the critical phase change temperature range and end time The time period of the composition; ] is the starting time when the thermal history temperature curve of the qualified finished product enters and leaves the key phase change temperature range and end time The composition time period.

[0088] Among them, the thermal disturbance factor of the phase change zone of the obtained magnet product is It is a specific analysis from macroscopic dynamic heat flux to microstructural sensitive area, which can reflect the degree of thermal disturbance in actual processing compared with the ideal process. Close to 0, indicating that the actual processing process is very stable; if the thermal disturbance factor of the phase change zone A large positive value indicates that there are unexpectedly drastic temperature fluctuations in the key phase change region during the actual processing, which is often the root cause of microcracks or magnetic domain structure defects.

[0089] S204. Divide the thermal history temperature curve of the finished magnet into multiple time windows of equal length, and calculate the temperature change in each time window; based on the thermal disturbance factor of the phase change zone of the finished magnet, correct the information entropy of the temperature change in all time windows to obtain the thermal history structural entropy of the finished magnet.

[0090] It should be noted that in order to comprehensively evaluate the global complexity and disorder of the entire thermal history curve, this embodiment uses the thermal disturbance factor of the phase change region to Starting from this indicator and combining it with the information entropy theory, the thermal history structure entropy indicator was constructed. By combining the thermal disturbance in the local key area with the morphological change of the global curve, a final and highly condensed feature was formed.

[0091] Specifically, the thermal history temperature curve of the magnet product is divided into equal-length time windows, a preset number; the temperature change amount in each time window is calculated, which is equal to the difference between the maximum and minimum values of the temperature in the time window; the temperature change amount proportion in each time window is calculated, which is the proportion of the temperature change amount in each time window to the total temperature change amount.

[0092] wherein, for the first time window, the temperature change amount proportion , , are the temperature change amounts in the first time window and the second time window, respectively. The probability distribution of the heat change distribution on the thermal history temperature curve can be described.

[0093] wherein, the preset number The specific value of the preset number can be set according to the actual application scene and requirements, and the value range of the preset number is [10, 35], and the preset number is set to 20 in the present application.

[0094] Further, according to the phase change region thermal disturbance factor of the magnet product, the information entropy of the temperature change amount in all time windows is corrected to obtain the thermal history structural entropy of the magnet product; and the calculation formula of the thermal history structural entropy of the magnet product is:

[0095] ;

[0096] In the formula, is the thermal history structural entropy of the magnet product; is the preset number; is the phase change region thermal disturbance factor of the magnet product; represents taking an absolute value; is the temperature change amount proportion in the first time window, which can describe the probability distribution of the heat change distribution on the thermal history temperature curve, is a logarithmic function with 2 as the base.

[0097] wherein, is a penalty term, which is derived from the phase change region thermal disturbance factor , which can amplify the thermal disturbance degree of the key phase change temperature interval and couple it into the calculation of the thermal history structural entropy.

[0098] wherein, is the information entropy of the temperature change amount in all time windows, used to measure the uniformity of the entire cooling process, the more uneven the process, the higher the information entropy value, for example, the temperature changes very quickly in some stages of the process, and other stages are stagnant.

[0099] It should be noted that the thermal history structure entropy of the obtained magnet product not only measures the overall unevenness of the thermal history temperature curve of the magnet product, but also particularly amplifies the negative impact of instability in the key phase transition temperature interval through the penalty term; when the thermal history temperature curve of a magnet product is not only uneven, that is, the information entropy of the thermal history temperature curve is large, but also has a severe disturbance in the key phase transition temperature interval, that is, the phase transition region thermal disturbance factor is large, the thermal history structure entropy of the magnet product will increase significantly; and the thermal history structure entropy of the magnet product can represent the quality of the thermal history of the magnet product.

[0100] S3, combine the thermal history structure entropy, peak temperature and total processing time of the magnet product to form a dynamic thermal history feature vector of the magnet product; based on the dynamic thermal history feature vector and the magnetic properties of the magnet product in the historical production data, a Gaussian process regression model is trained.

[0101] Since Gaussian process regression does not directly process the original time series data, but needs to convert the continuous thermal history temperature curve of each magnet product into a feature vector composed of a group of discrete features; therefore, the thermal history structure entropy , peak temperature and total processing time of the magnet product are combined to form a dynamic thermal history feature vector [ ] of the magnet product, which replaces the feature set obtained by relying on artificial experience in traditional methods to describe the features of the thermal history.

[0102] Among them, the peak temperature and the total processing time belong to the global physical quantity of the thermal history of the magnet product, and the peak temperature and the total processing time are obtained by analyzing the thermal history temperature curve of the magnet product.

[0103] Specifically, the peak temperature refers to the highest temperature experienced by the magnet product during the processing process, i.e., the maximum value of the temperature value in the thermal history temperature curve of the magnet product; the thermal history temperature curve of the magnet product reflects the change of its temperature with time during the processing process, therefore, the total processing time refers to the time experienced during the entire process from heating to cooling, i.e., the time difference between the end time and the start time in the thermal history temperature curve of the magnet product.

[0104] Further, using historical production data, the dynamic thermal history feature vector of the magnet product in the historical production data is used as input, and the magnetic properties of the magnet product in the historical production data are used as output, to train a Gaussian process regression model, which constructs a probability distribution model about the "input-output" relationship by learning the distribution therein.

[0105] The magnetic properties include remanence, magnetic induction coercivity, and intrinsic coercivity.

[0106] It should be noted that by constructing a series of interrelated dynamic indicators such as dynamic heat flux integral, phase change region thermal disturbance factor, and thermal history structural entropy, the objectivity and comprehensiveness of feature extraction are improved, thereby significantly improving the prediction accuracy of magnetic properties and providing a solid foundation for achieving high consistency of magnetic properties.

[0107] S4, through the trained Gaussian process regression model, the magnetic property prediction mean and the magnetic property prediction variance of the just-processed magnet product are obtained, which are used to determine whether the just-processed magnet product is to be inspected and re-inspected.

[0108] Specifically, in actual production, for the just-processed magnet product, its dynamic thermal history feature vector is calculated in real time and input into the trained Gaussian process regression model, and the model will output two key information, which are the magnetic property prediction mean and the magnetic property prediction variance of the just-processed magnet product.

[0109] The magnetic property prediction mean is the core output of the model, which is the most likely prediction result of the final magnetic property of the magnet product, and the magnetic property prediction variance is the confidence of the prediction result, which can be used to quantify the uncertainty of the model.

[0110] Further, it is determined whether the magnetic property prediction mean of the just-processed magnet product is within the target range of the magnetic properties, and whether the magnetic property prediction variance of the just-processed magnet product exceeds the preset threshold:

[0111] (1) If the magnetic property prediction mean of the just-processed magnet product is within the target range of the magnetic properties, and the magnetic property prediction variance is less than the preset threshold, it means that the just-processed magnet product meets the requirements.

[0112] (2) Otherwise, it is indicated that the just processed magnet product is likely to not meet the process requirements, the just processed magnet product is marked as a to-be-inspected product, and the magnetic properties of the to-be-inspected product are re-inspected.

[0113] In the table 1, the target ranges of the magnetic properties of various neodymium-iron-boron materials are shown, and the target ranges of the remanence, the magnetic induction coercive force and the intrinsic coercive force of different neodymium-iron-boron materials are included.

[0114] Table 1

[0115]

[0116] It should be noted that the biggest advantage of the Gaussian process regression is that the uncertainty of the prediction result can be quantified, and the control system can make decisions based on the confidence provided by the Gaussian process regression model, that is, the magnetic property prediction variance, and only the magnets meeting the performance requirements in the high confidence interval are released, and the magnets with low confidence are re-inspected, thereby improving the robustness and reliability of the quality control.

[0117] The embodiment of the present application also discloses a neodymium-iron-boron magnet processing control system, comprising a processor and a memory, and the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the neodymium-iron-boron magnet processing control method according to the present application is realized.

[0118] The above system also includes a communication bus and a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

Claims

1. A method of controlling the processing of neodymium-iron-boron magnets, characterized in that The method comprises the following steps: obtaining the start time and the end time of the thermal history temperature curve of the magnet product entering and leaving the key phase change temperature interval; based on the thermal history temperature curve, calculating the dynamic heat flux integral of the time period composed of the start time and the end time of the magnet product; calculating the phase change region thermal disturbance factor of the magnet product according to the relative deviation of the dynamic heat flux integral of the time period composed of the start time and the end time of the magnet product compared with the qualified product; dividing the thermal history temperature curve of the magnet product into multiple equal-length time windows, calculating the temperature change in each time window; according to the phase change region thermal disturbance factor of the magnet product, correcting the information entropy of the temperature change in all time windows to obtain the thermal history structure entropy of the magnet product; combining the thermal history structure entropy, the peak temperature and the total processing time of the magnet product to form the dynamic thermal history feature vector of the magnet product; based on the dynamic thermal history feature vector and the magnetic property of the magnet product in the historical production data, training a Gaussian process regression model; obtaining the magnetic property prediction mean and the magnetic property prediction variance of the just-processed magnet product by the trained Gaussian process regression model, for judging whether the just-processed magnet product is a product to be inspected and performing re-inspection.

2. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. The thermal history temperature curve is composed of temperature data continuously collected in each key stage of the whole process of producing the magnet product; the key phase change temperature interval is determined according to the corresponding neodymium iron boron material of the magnet product.

3. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. The start time and the end time of the thermal history temperature curve of the magnet product entering and leaving the key phase change temperature interval are obtained, comprising: For the temperature of the magnet product at each time in the thermal history temperature curve, if the temperature of the magnet product at time is greater than or equal to and the temperature at time is less than , time is taken as the start time of the critical phase transition temperature interval of the thermal history temperature curve; If the temperature of the magnet product at time is greater than and the temperature at time is equal to or less than , the time is taken as the end time of the critical phase transition temperature range of the thermal history temperature curve; , are the lower and upper temperature limits of the critical phase transition temperature interval, respectively.

4. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. The dynamic heat flux integral of the time period composed of the start time and the end time of the magnet product is calculated based on the thermal history temperature curve, comprising: ； Where, For the finished magnet in the time period [ ]Dynamic heat flux integral;[ ] is the starting time when the thermal history temperature curve of the finished magnet enters and leaves the critical phase change temperature range and end time the time period of the composition; is the weight function, For the moment Temperature value; is the first derivative of temperature with respect to time.

5. The method of claim 4, wherein the step of applying a coating of a material to the surface of the sintered magnet comprises applying a coating of a material to the surface of the sintered magnet to form a coating of a material on the surface of the sintered magnet. The calculation formula of the weight function is: ； wherein is the natural exponential function; denotes the temperature standard deviation, and , , are the lower and upper temperature limits of the critical phase transition temperature interval, respectively; is the Curie temperature of the processed NdFeB material.

6. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. The phase change region thermal disturbance factor of the magnet product is calculated, comprising: The qualified finished product refers to a finished magnet product corresponding to a qualified neodymium-iron-boron magnet; and a thermal disturbance factor of a phase transition zone of the finished magnet product is calculated by the following formula: ； wherein is the dynamic heat flux integral of the magnet finished product over the time period ] is the dynamic heat flux integral of the finished product over the time period is the dynamic heat flux integral of all finished products over the time period is the average of the dynamic heat flux integral of all finished products over the time period is the time period consisting of the start time and the end time at which the thermal history temperature profile of the finished product enters and exits the critical phase transition temperature interval, respectively.​​ 7. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. According to the phase change region thermal disturbance factor of the magnet product, the information entropy of the temperature change in all time windows is corrected to obtain the thermal history structure entropy of the magnet product, comprising: ； In the formula, is the structure entropy of the thermal history of the magnet product; is the preset number; is the first is the proportion of the temperature change amount in the first is the information entropy of the temperature change amount in all time windows; is the phase transition region thermal disturbance factor of the magnet product; represents taking the absolute value.

8. The method of claim 7, wherein the step of applying a coating of a material to the surface of the sintered magnet comprises applying a coating of a material to the surface of the sintered magnet. The temperature change amount in the time window is equal to the difference between the maximum value and the minimum value of the temperature in the time window; the temperature change amount in the first time window accounts for , , The temperature change amount in the first time window and the temperature change amount in the second time window, respectively.

9. The method of claim 1, wherein the step of applying a coating to the surface of the sintered magnet comprises applying a coating to the surface of the sintered magnet to form a coating having a thickness of 0.1 to 1.0 μm. Whether the just-processed magnet product is a product to be inspected and performing re-inspection, comprising: determining whether the magnetic property prediction mean of the just-processed magnet product is within the target range of the magnetic property, and whether the magnetic property prediction variance of the just-processed magnet product exceeds the preset threshold value; if the magnetic property prediction mean of the just-processed magnet product is within the target range of the magnetic property, and the magnetic property prediction variance is less than the preset threshold value, it means that the just-processed magnet product meets the requirements; otherwise, it means that the just-processed magnet product is likely to not meet the process requirements, and the just-processed magnet product is marked as a product to be inspected, and the magnetic property of the product to be inspected is re-inspected.

10. A neodymium-iron-boron magnet processing control system, characterized by, The method comprises the following steps: a processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, a neodymium iron boron magnet processing control method according to any one of claims 1-9 is realized.

Images