A temperature field adaptive control method and system of a neodymium iron boron vacuum sintering furnace

Abstract

The present application belongs to the technical field of temperature control, and particularly relates to a temperature field adaptive control method and system for a Nd-Fe-B vacuum sintering furnace, comprising the following steps: collecting the actual measured temperatures of each thermocouple in the upper and lower zones of the furnace according to a control period, and obtaining the average temperature of the upper zone and the average temperature of the lower zone by averaging the actual measured temperatures of each thermocouple, respectively; calculating the difference between the maximum and minimum of the actual measured temperatures of each thermocouple to obtain the instantaneous temperature difference; taking the maximum of each instantaneous temperature difference in a preset time window as the temperature difference extreme value; multiplying the temperature difference extreme value by a preset expansion coefficient to obtain the upper limit of the domain, and scaling the fuzzy set boundary in proportion to the upper limit of the domain; and counting the activation frequency of each fuzzy rule in the preset window, calculating the information entropy according to each activation frequency, and obtaining the rule activation entropy. The present application effectively overcomes the control inaccuracy problem caused by the dramatic change in the temperature difference magnitude in each sintering stage, and significantly improves the remanence consistency and yield of the same batch of products.

Description

Technical Field

[0001] This invention belongs to the field of temperature control technology, specifically relating to an adaptive temperature field control method and system for a NdFeB vacuum sintering furnace. Background Technology

[0002] Neodymium iron boron (NdFeB) magnets, as an indispensable core functional material in modern industry, rely heavily on precise temperature control during the sintering process for their superior magnetic properties. In standard NdFeB vacuum sintering furnace production, the entire sintering cycle typically lasts up to sixteen hours, involving a degassing section at 400-600 degrees Celsius and a main sintering section at 1050-1100 degrees Celsius. Due to the significant temperature differences between these two stages and the uneven distribution of thermal inertia across the furnace, maintaining real-time and precise uniformity of the furnace's thermal field throughout the entire sintering cycle, and ensuring smooth temperature tracking of each zone to the target process curve, becomes a crucial requirement for guaranteeing consistent remanent magnetic properties across batches of products.

[0003] To meet the temperature control requirements of the aforementioned sintering process, the industry currently widely adopts traditional proportional-integral-derivative (PID) control technology based on independent closed-loop control for multiple temperature zones, or introduces a fuzzy controller with fixed parameters on this basis. This type of existing technology typically involves arranging multiple thermocouples in different physical zones within the furnace for independent temperature measurement, and calculating the absolute deviation and rate of change of the deviation between the real-time temperature and the set target. Subsequently, the control system directly inputs these deviation data into the fuzzy controller, which has pre-defined domain boundaries and control rules, for logical reasoning, and then outputs fixed compensation parameters. Finally, it attempts to smooth out the temperature differences between zones by adjusting the output power of the heating elements in each zone.

[0004] However, the aforementioned existing technologies have inherent control logic flaws when faced with the huge temperature jumps between the degassing section and the main sintering section. Because their fuzzy set universe of discourse and rule table are set to static dead zones throughout the entire sintering cycle, when the actual temperature difference inside the furnace far exceeds the preset universe of discourse, a large amount of deviation data will be truncated, causing the controller to completely lose its analytical resolution regarding temperature magnitude. Simultaneously, the rigid rule table cannot adaptively match the actual thermodynamic state of the current stage, easily leading to slow system response or severe power overshoot. This blind spot in the adaptation of fixed parameters over a wide temperature range ultimately results in persistently high remanence dispersion in the same batch of NdFeB products, failing to meet the stringent quality control requirements of high-end magnetic materials. Summary of the Invention

[0005] This invention provides a temperature field adaptive control method and system for a NdFeB vacuum sintering furnace, to solve the technical problem that the existing fixed-parameter fuzzy controller has an adaptation blind zone when facing a wide temperature difference range, resulting in high residual magnetism dispersion of NdFeB products in the same batch.

[0006] In a first aspect, the present invention provides a method for adaptive temperature field control of a NdFeB vacuum sintering furnace, comprising the following steps: The measured temperatures of each thermocouple in the upper and lower zones of the furnace are collected according to the control cycle, and the average values ​​are calculated to obtain the average temperature of the upper and lower zones. The difference between the maximum and minimum measured temperatures of each thermocouple is calculated to obtain the instantaneous temperature difference. The maximum instantaneous temperature difference within a preset time window is taken as the extreme value of the temperature difference; the extreme value of the temperature difference is multiplied by a preset expansion coefficient to obtain the upper limit of the universe of discourse, and the boundary of the fuzzy set is scaled proportionally according to the upper limit of the universe of discourse. The activation frequency of each fuzzy rule within a preset window is counted, and the information entropy is calculated based on each activation frequency to obtain the rule activation entropy. When the rule activation entropy is lower than the preset entropy threshold, the fuzzy set is re-divided according to the upper limit of the domain of discourse, and the output values ​​of each rule are recalibrated to obtain an updated rule table. Calculate the upper zone normalization deviation and lower zone normalization deviation relative to the process target temperature, respectively. Use the change rates of the upper zone normalization deviation and lower zone normalization deviation as inputs, perform fuzzy inference through the update rule table, and obtain the upper zone correction increment. Use the change rates of the lower zone normalization deviation and lower zone normalization deviation as dual inputs, perform fuzzy inference through the update rule table, and obtain the lower zone correction increment. Output the upper zone adjustment amount and lower zone adjustment amount according to the upper zone correction increment and lower zone correction increment, respectively.

[0007] Its effect is as follows: by calculating the instantaneous temperature difference extreme value, a dynamic domain of discourse is established, and the information entropy is calculated based on the activation frequency as the trigger benchmark for rule reconstruction. The spatial range is closely linked with the rule utilization state, ensuring that the deviation data falls within the high resolution range, eliminating the adaptation blind zone of the fixed fuzzer under wide temperature difference conditions, and greatly improving the convergence accuracy of partition control.

[0008] Furthermore, multiplying the extreme temperature difference by a preset expansion coefficient to obtain the upper limit of the universe of discourse includes: multiplying the extreme temperature difference by a preset expansion coefficient to obtain the upper limit of the candidate universe of discourse; within a continuous preset verification window after the upper limit of the candidate universe of discourse takes effect, counting the number of sampling points in each control cycle whose instantaneous temperature difference does not exceed the upper limit of the candidate universe of discourse to obtain the number of qualified points; dividing the number of qualified points by the total number of sampling points in the verification window to obtain the universe of discourse coverage rate; when the universe of discourse coverage rate is not lower than the preset coverage rate threshold, the upper limit of the candidate universe of discourse is confirmed as the upper limit of the universe of discourse.

[0009] Its effect is as follows: by counting the number of qualified points whose instantaneous temperature difference does not exceed the upper limit of the candidate within the verification window, and calculating the domain coverage as the basis for the validity judgment, a rigorous post-verification mechanism is constructed, which effectively prevents the domain setting from being too conservative and causing the real-time deviation data to be truncated, thus solidly ensuring the global validity of the inference output.

[0010] Furthermore, when the domain coverage is lower than the preset coverage threshold, the method further includes: increasing the preset expansion coefficient to a preset upward adjustment value, multiplying the extreme temperature difference by the increased preset expansion coefficient to obtain a new upper limit of the candidate domain; recalculating the domain coverage, and repeating the above process until the domain coverage is not lower than the preset coverage threshold.

[0011] Furthermore, after using the maximum instantaneous temperature difference within a preset time window as the extreme value of the temperature difference, the method further includes: subtracting the extreme value of the temperature difference in the current statistical window from the extreme value of the temperature difference in the previous statistical window and dividing by the length of the preset time window to obtain the rate of change of the temperature difference; when the rate of change of the temperature difference is higher than a first rate threshold, shortening the update interval of the upper limit of the domain of discourse to the first interval; when the rate of change of the temperature difference is lower than a second rate threshold, extending the update interval of the upper limit of the domain of discourse to the second interval; when the rate of change of the temperature difference is not higher than the first rate threshold and not lower than the second rate threshold, maintaining the preset default interval; wherein, the first interval is less than the preset default interval, and the second interval is greater than the preset default interval.

[0012] Its effect is that by calculating the rate of change of the extreme values ​​of temperature difference between adjacent statistical windows, the update interval of the upper limit of the domain of discourse is adaptively shortened or extended, enabling the control system to quickly follow during the thermal field jump period and avoid introducing invalid disturbances during the stable heat preservation period, thus achieving an efficient and smooth physical match between the update frequency and the actual process dynamics.

[0013] Furthermore, the preset entropy threshold is determined in the following way: after the first complete sintering cycle is completed, the activation entropy of each rule calculated throughout the sintering cycle is obtained; the activation entropy of each rule is sorted from smallest to largest, and the activation entropy value of the rule located at the preset quantile after sorting is taken as the preset entropy threshold.

[0014] Furthermore, the output of the upper and lower zone adjustment quantities based on the upper and lower zone correction increments includes: obtaining the basic proportional, integral, and derivative parameters of the PID controller for each zone; limiting the correction components of each parameter in the upper and lower zone correction increments to the preset proportional range of the corresponding basic proportional, integral, and derivative parameters to obtain the limited correction increment; summing the limited correction increment with the corresponding basic parameters to obtain the final PID parameters for each zone; and calculating the upper and lower zone adjustment quantities based on the final PID parameters for each zone.

[0015] Its effect is that by limiting the correction increment of each zone to the preset ratio range of the basic parameters, and then summing it with the basic parameters to output the final adjustment amount, it effectively prevents the parameter correction from becoming excessively divergent under extreme temperature difference changes, eliminates the risk of output reversal or integral saturation from the bottom layer, and greatly improves the stability of the response of the bottom actuator.

[0016] Furthermore, re-dividing the fuzzy sets according to the upper limit of the universe of discourse includes: dividing the interval from the upper limit of the negative universe of discourse to the upper limit of the positive universe of discourse into a predetermined number of fuzzy sets at equal intervals; dividing the interval length by the predetermined number to obtain the fuzzy set interval; and determining the boundary position of each fuzzy set sequentially from the upper limit of the negative universe of discourse, using the fuzzy set interval as the step size.

[0017] Furthermore, before collecting the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the method also includes setting the length of the preset time window to be an integer multiple of the control cycle.

[0018] Furthermore, after collecting the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the process also includes: detecting outliers in the measured temperatures of each thermocouple, replacing the sampled values ​​that exceed the preset range of the corresponding thermocouple with the sampled values ​​of the thermocouple in the previous control cycle; performing mean filtering on the replaced measured temperatures of each thermocouple to obtain the filtered measured temperatures of the thermocouple; and using the filtered measured temperatures of the thermocouples to perform subsequent steps.

[0019] Its effect is as follows: by replacing the over-range sampled values ​​with historical normal values ​​and performing mean filtering on the replaced data, it effectively intercepts sudden electromagnetic interference spikes in the industrial field, completely eliminates the hidden danger of injecting inferior and sudden data into the control link, and provides a highly reliable data source for subsequent advanced adaptive inference.

[0020] Secondly, the present invention provides a temperature field adaptive control system for a NdFeB vacuum sintering furnace, comprising a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned temperature field adaptive control method for the NdFeB vacuum sintering furnace is implemented.

[0021] The beneficial effects are: by extracting the extreme values ​​of temperature difference at multiple points and dynamically scaling the upper limit of the domain of discourse, introducing rules to activate entropy and synchronously reconstruct the rule table, and combining normalized deviation to execute partition control, the control logic can accurately match the magnitude of temperature difference at each stage, overcome the overshoot and sluggishness problems that are easily caused by fixed parameters, and significantly improve the uniformity of the thermal field in the furnace and the yield of the final product. Attached Figure Description

[0022] Figure 1 This is a flowchart of the temperature field adaptive control method for the NdFeB vacuum sintering furnace of the present invention.

[0023] Figure 2 This is a comparison curve of the temperature rise following effect of the present invention.

[0024] Figure 3 This is a graph showing the dynamic change of the upper limit of the domain of discussion of this invention with the extreme value of temperature difference. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] An embodiment of the temperature field adaptive control method for the NdFeB vacuum sintering furnace provided by the present invention: like Figure 1 As shown, the adaptive temperature field control method for a NdFeB vacuum sintering furnace includes the following steps: S1: Collect the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, and calculate the average values ​​to obtain the average temperature of the upper and lower zones; calculate the difference between the maximum and minimum measured temperatures of each thermocouple to obtain the instantaneous temperature difference.

[0027] In one embodiment, to continuously provide a stable time base and data analysis scale, time parameters need to be set before data acquisition occurs. Before acquiring the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the control cycle is set to an integer value ranging from 5 to 30 seconds; the length of the preset time window is set to an integer multiple of the control cycle, and the preset time window length ranges from 45 to 90 seconds. For example, considering the physical response thermal inertia of this type of vacuum sintering furnace, the control cycle is set to 10 seconds, and the preset time window length is set to 60 seconds, meaning that one preset time window precisely contains 6 control cycles.

[0028] After obtaining the time reference, the measured temperatures of the eight thermocouples inside the furnace are synchronously collected according to the control cycle. To ensure the reliability of the basic data and filter out electromagnetic interference from the industrial site, after collecting the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the process includes outlier detection for each thermocouple's measured temperature. Sample values ​​exceeding the preset range of the corresponding thermocouple are replaced with the sample values ​​from the previous control cycle. The replaced measured temperatures are then averaged to obtain the filtered measured temperatures. Subsequent steps are performed using these filtered measured temperatures. In essence, by replacing out-of-range spikes with historical normal values ​​from the previous control cycle and combining this with moving average filtering to smooth short-term fluctuations, the risk of injecting substandard data into the subsequent calculation process and causing control oscillations is effectively prevented.

[0029] Based on the filtered thermocouple measured temperatures obtained from the above collection and cleaning, in order to quantify the current overall thermal field level of each region, it is necessary to reduce the dimensionality of the high-dimensional single-point temperature data and aggregate it into regional indicators. The average temperature of the upper region and the average temperature of the lower region satisfy the following relationship:

[0030]

[0031] In the formula, The average temperature of the upper region is expressed in degrees Celsius. The average temperature of the lower region is expressed in degrees Celsius. For the first The measured temperature of the thermocouple after filtering in the current control cycle, in degrees Celsius; Take 1 to 4 to represent the upper zone thermocouple. We take 5 to 8 to represent the thermocouples in the lower zone. Understandably, by using a linear mapping method with an arithmetic mean, the discrete temperature measurement data is collapsed into representative variables that reflect the overall thermal potential of the upper and lower zones, providing an intuitive and reliable benchmark for subsequent judgment of regional temperature deviations.

[0032] Meanwhile, in order to monitor the global temperature non-uniformity within the furnace in real time, it is necessary to extract the range characteristics of the global temperature, and the instantaneous temperature difference satisfies the following relationship:

[0033] In the formula, The instantaneous temperature difference during the current control cycle is expressed in degrees Celsius. For the current control cycle number The filtered temperature of the thermocouple is measured in degrees Celsius. This is the maximum value among the actual measured temperatures of the eight filtered thermocouples in the entire furnace during the current control cycle, in degrees Celsius. This represents the minimum measured temperature of all eight filtered thermocouples in the furnace during the current control cycle, expressed in degrees Celsius. Understandably, this formula precisely quantifies the most severe non-uniformity of the furnace's thermal field at the current instant by finding the extreme temperature boundaries in spatial dimensions. This provides the primary physical basis for subsequently determining whether the fuzzy set's domain of discourse needs to expand.

[0034] By setting up pre-time parameters, detecting and filtering out hardware anomalies, and extracting temperature uniformity collapse and range features in the spatial dimension, not only are random interference signals in the industrial field effectively isolated, but also the complex eight-channel discrete single-point data are accurately condensed into upper zone temperature uniformity, lower zone temperature uniformity and instantaneous temperature difference with clear physical orientation, thus building a solid and closed-loop data source foundation for subsequent advanced adaptive control.

[0035] S2, take the maximum value of each instantaneous temperature difference within the preset time window as the extreme value of temperature difference; multiply the extreme value of temperature difference by the preset expansion coefficient to obtain the upper limit of the universe of discourse, and scale the boundary of the fuzzy set proportionally according to the upper limit of the universe of discourse.

[0036] In one embodiment, to quantify the upper limit of the actual magnitude of the non-uniformity of the thermal field inside the furnace during the current sintering stage, the maximum value of each instantaneous temperature difference within a preset time window is taken as the extreme value of the temperature difference; the extreme value of the temperature difference satisfies the following relationship:

[0037] In the formula, This represents the extreme value of the temperature difference, with the dimension of degrees Celsius. Preset time window; For each sampling moment within a preset time window; The instantaneous temperature difference at each sampling time is expressed in degrees Celsius. Understandably, by finding the temporal peak characteristics of the spatial range within a complete sliding window, the true temperature difference jump top line of the furnace thermal field at different stages such as the degassing section and the main sintering section can be effectively and stably captured, thus serving as reliable data support for the subsequent adaptive expansion of the spatial benchmark by the fuzzy converter.

[0038] After obtaining the extreme temperature difference values, in order to dynamically adapt to the rapid changes in the sintering furnace thermal field caused by the switching of process stages, after taking the maximum value of the instantaneous temperature difference within a preset time window as the extreme temperature difference value, the method further includes subtracting the extreme temperature difference value of the current statistical window from the extreme temperature difference value of the previous statistical window and dividing by the length of the preset time window to obtain the temperature difference change rate. When the temperature difference change rate is higher than a first rate threshold, the update interval of the upper limit of the universe of discourse is shortened to the first interval; when the temperature difference change rate is lower than a second rate threshold, the update interval of the upper limit of the universe of discourse is extended to the second interval; when the temperature difference change rate is neither higher than the first rate threshold nor lower than the second rate threshold, the preset default interval is maintained. The first interval is less than the preset default interval, and the second interval is greater than the preset default interval. The temperature difference change rate satisfies the following relationship:

[0039] In the formula, The rate of change of temperature difference is expressed in degrees Celsius per second. This represents the extreme temperature difference within the current statistical window, measured in degrees Celsius. This represents the extreme temperature difference of the previous statistical window, measured in degrees Celsius. The length of the preset time window is measured in seconds. For example, the preferred first rate threshold is 0.083 degrees Celsius per second, and the preferred second rate threshold is 0.017 degrees Celsius per second; correspondingly, the preferred first interval is 10 seconds, the preferred second interval is 180 seconds, and the preferred default interval is 60 seconds. Understandably, when the rate of temperature change is large, it indicates that the thermal field is in a rapid change period, such as when the degassing section enters the main sintering section. In this case, the update interval is significantly shortened to quickly follow the state change; conversely, when the thermal field tends to be in a stable holding period, the update interval is extended to avoid introducing invalid system disturbances, thereby achieving intelligent adaptive matching of the domain update frequency with process dynamics.

[0040] After determining the update interval and extreme temperature difference, to ensure that the newly generated fuzzy set universe of discourse has sufficient margin and truly reflects the current temperature difference distribution, the extreme temperature difference is multiplied by a preset expansion coefficient to obtain the upper limit of the universe of discourse, including multiplying the extreme temperature difference by the preset expansion coefficient to obtain the upper limit of the candidate universe of discourse. Within a continuous preset verification window after the upper limit of the candidate universe of discourse takes effect, the number of sampling points whose instantaneous temperature difference in each control cycle does not exceed the upper limit of the candidate universe of discourse is counted to obtain the number of qualified points. The number of qualified points is divided by the total number of sampling points in the verification window to obtain the universe of discourse coverage. When the universe of discourse coverage is not lower than the preset coverage threshold, the upper limit of the candidate universe of discourse is confirmed as the upper limit of the universe of discourse. The upper limit of the candidate universe of discourse satisfies the following relationship:

[0041] In the formula, is the upper bound of the candidate universe of discourse, with the dimension in degrees Celsius; The expansion coefficient is a preset value, which is dimensionless, and its initial preferred value is 1.2. This represents the extreme value of the temperature difference, with the dimension in degrees Celsius. Simultaneously, the universe coverage used for verification satisfies the following relationship:

[0042] In the formula, The domain coverage is dimensionless. The number of qualifying points within a continuously preset verification window is expressed in units of "points". The number of sampling points within the continuously preset verification window is expressed in units of "points". For example, a preferred preset verification window is 60 seconds, and the preset coverage threshold is set to 0.85. Understandably, the closer the universe coverage is to 1, the more fully the actual measured temperature difference data falls within the newly generated candidate universe interval. Through rigorous posterior statistical coverage review, it prevents a large amount of real-time deviation data from being truncated due to an overly conservative preset expansion coefficient, thus solidly ensuring the effective resolution of subsequent fuzzy inference.

[0043] In another embodiment, to address the situation where the initially calculated upper limit of the candidate domain cannot encompass the extreme boundary of the actual temperature difference due to abrupt changes in the thermal field, when the domain coverage is lower than a preset coverage threshold, the method further includes increasing the preset expansion coefficient to a preset upward adjustment value. The increased preset expansion coefficient is then used to recalculate the extreme temperature difference by the increased preset expansion coefficient to obtain a new upper limit of the candidate domain. The domain coverage is then recalculated, and the above process is repeated until the domain coverage is not lower than the preset coverage threshold. It should be noted that each time a cyclic adjustment is triggered, the preset upward adjustment value is preferably set to increase by 0.1 based on the current preset expansion coefficient. Understandably, this cyclic recalculation mechanism ensures that once the candidate domain is found to be insufficiently spacious, the expansion coefficient is automatically and progressively increased to expand the domain boundary a second or even multiple times until the instantaneous temperature difference data during the verification period fully falls within the safe control range, thus constructing a robust domain boundary fallback search closed loop.

[0044] Based on the instantaneous temperature difference extracted from the front end and its rate of change monitored, the timing and pace of the domain update are intelligently allocated. At the same time, the upper limit of the candidate domain is generated by combining the preset expansion coefficient and the closed-loop standard verification of the domain coverage and the adaptive cyclic adjustment mechanism are introduced. This not only avoids the data saturation and truncation distortion caused by the fixed domain from the source, but also completely overcomes the problem of fuzzy inaccuracy caused by the large difference in thermal field characteristics at each sintering stage from the perspective of engineering practice and fault tolerance. This builds a highly flexible and reliable decision boundary for issuing high-precision fuzzy inference conclusions.

[0045] S3: Calculate the activation frequency of each fuzzy rule within the preset window, and calculate the information entropy based on each activation frequency to obtain the rule activation entropy; when the rule activation entropy is lower than the preset entropy threshold, re-divide the fuzzy set according to the upper limit of the domain and recalibrate the output value of each rule to obtain the updated rule table.

[0046] In one embodiment, to verify the matching degree between the original fuzzy rule table and the new universe of discourse generated in the previous steps, and to promptly trigger the synchronous replacement of the rule table when a serious mismatch is detected, it is necessary to first statistically analyze the activation status of each fuzzy rule. Specifically, the activation frequency of each fuzzy rule within a preset window is statistically analyzed. To ensure the validity of the statistical data and avoid contamination of old data due to sudden changes in the spatial benchmark, when the upper limit of the universe of discourse is updated, the statistical window for rule activation entropy is reset, and the calculation of the activation frequency of each fuzzy rule is restarted from the effective time of the upper limit update, so as to eliminate the interference of the activation frequency data before the upper limit update on the calculation result of the rule activation entropy after the update. Understandably, since the change of the upper limit of the universe of discourse will cause a sudden change in the mapping position of the deviation data, forcibly resetting the statistical window can ensure that the subsequently calculated evaluation index fully and purely reflects the actual rule utilization under the new universe of discourse benchmark. Based on this, the information entropy is calculated according to each activation frequency to obtain the rule activation entropy, which satisfies the following relationship:

[0047] In the formula, Entropy is activated by the rule and is dimensionless; For the first The activation frequency of a fuzzy rule within a statistical window, dimensionless; The value is the fuzzy rule number, ranging from 1 to 49. Understandably, this formula borrows the principle of Shannon's information entropy to construct a quantitative index for measuring the uniformity of fuzzy rule distribution. The smaller the rule activation entropy, the more concentrated the fuzzy inference is on a few rules, and the more uneven the utilization of the rule table, thus providing a precise mathematical characterization for judging whether a mismatch has occurred in the rule table. For example, the length of the statistical window is preferably set to 300 seconds.

[0048] After obtaining the rule activation entropy, it needs to be compared with a preset benchmark to determine whether to trigger the update mechanism. When the rule activation entropy is lower than the preset entropy threshold, the rule table is updated. To ensure that the benchmark can adaptively fit the inherent thermal characteristics of the specific vacuum sintering furnace, the preset entropy threshold is determined as follows: After the first complete sintering cycle, the activation entropy of each rule calculated throughout the sintering cycle is obtained; the activation entropy of each rule is sorted from smallest to largest, and the activation entropy value of the rule at the preset quantile after sorting is taken as the preset entropy threshold. It should be noted that by extracting the entropy value corresponding to the most uneven period in the historical full-cycle record as the trigger benchmark, this method can automatically calibrate the preset entropy threshold within one to two sintering cycles, completely eliminating the tedious and error-prone manual parameter tuning process. For example, the preset quantile is preferably set to 10%, and the preset entropy threshold reference value is 2, which physically corresponds to the extreme mismatch state where fuzzy inference is severely concentrated within about four rules.

[0049] When the rule activation entropy is lower than a preset entropy threshold, it is determined that there is a serious mismatch between the current rule table and the latest effective upper limit of the universe of discourse. At this time, the fuzzy sets are re-divided according to the upper limit of the universe of discourse, and the output values ​​of each rule are recalibrated to obtain an updated rule table. The re-division of fuzzy sets according to the upper limit of the universe of discourse includes: dividing the interval from the negative upper limit of the universe of discourse to the positive upper limit of the universe of discourse into a preset number of fuzzy sets at equal intervals; dividing the interval length by the preset number to obtain the fuzzy set interval; and determining the boundary position of each fuzzy set sequentially from the negative upper limit of the universe of discourse using the fuzzy set interval as the step size. Among them, the upper limit of the universe of discourse, which has been strictly verified in the previous steps, is used as a new absolute spatial benchmark. The boundaries of each fuzzy set, usually preferably seven, are re-anchored by equal interval division, which completely eliminates the local inference congestion caused by the stretching or compression of the spatial range. After the boundary division is completed, the output values ​​of each rule corresponding to the new boundary are simultaneously recalibrated based on the experience logic of the specific sintering process of heating and holding, and finally an updated rule table that perfectly matches the new universe of discourse range is output.

[0050] By calculating the activation entropy of the rules, the uniformity of rule table utilization is monitored in real time and quantitatively. When the entropy value is too low, the fuzzy set boundary is automatically re-divided into equal intervals and the rule output value is synchronously reconstructed. This achieves precise data linkage between spatial domain update and internal rule table update, fundamentally solving the core pain point of long-term mismatch between fuzzy set and rules caused by the lack of synchronization of rule tables in dynamic domain scenarios. This ensures that the fuzzy controller maintains extremely high inference resolution and control accuracy throughout the entire cycle of drastic temperature jumps in the degassing section and the main sintering section.

[0051] S4 calculates the upper zone normalization deviation and lower zone normalization deviation relative to the process target temperature, respectively. The change rates of the upper zone normalization deviation and lower zone normalization deviation are used as inputs, and fuzzy inference is performed through the update rule table to obtain the upper zone correction increment. The change rates of the lower zone normalization deviation and lower zone normalization deviation are used as dual inputs, and fuzzy inference is performed through the update rule table to obtain the lower zone correction increment. The upper zone adjustment amount and lower zone adjustment amount are output according to the upper zone correction increment and lower zone correction increment, respectively.

[0052] In one embodiment, to convert the adaptively updated universe of discourse reference and the matched update rule table from the preceding steps into actual control input signals, the upper region normalization deviation and lower region normalization deviation of the upper and lower region average temperatures relative to the process target temperature are calculated, respectively. Before calculating the normalization deviation, the actual upper region temperature difference deviation and lower region temperature difference deviation are calculated, which satisfy the following relationship:

[0053]

[0054] In the formula, This refers to the temperature difference deviation in the upper zone, measured in degrees Celsius. This refers to the temperature difference deviation in the lower zone, measured in degrees Celsius. The average temperature of the upper region is expressed in degrees Celsius. The average temperature of the lower region is expressed in degrees Celsius. The target temperature is measured in degrees Celsius. Understandably, by calculating the difference between the average temperature of a region and the target temperature, the absolute physical degree to which each region deviates from the sintering target is directly reflected.

[0055] Subsequently, based on the aforementioned absolute deviation and the upper bound of the universe of discourse finally confirmed in the preceding steps, the upper region normalization deviation and the lower region normalization deviation satisfy the following relationship:

[0056]

[0057] In the formula, The normalization deviation in the upper region is dimensionless. The normalization deviation for the lower region is dimensionless. This refers to the temperature difference deviation in the upper zone, measured in degrees Celsius. This refers to the temperature difference deviation in the lower zone, measured in degrees Celsius. Let be the upper bound of the universe of discourse, with the dimension in degrees Celsius. Understandably, dividing the absolute temperature difference deviation by the dynamically updated upper bound of the universe of discourse can strictly map the deviation and converge it to the standard normalized interval of -1 to 1, ensuring that the input quantity strictly corresponds to the boundary of the fuzzy set re-divided by the previous step, and completely avoiding the problem of decreased inference resolution caused by drastic changes in the magnitude of temperature difference.

[0058] Meanwhile, to capture the dynamic convergence or divergence trend of temperature difference deviating from the process target, the change rates of the upper region normalized deviation and the lower region normalized deviation are used as inputs. Fuzzy inference is performed through the update rule table to obtain the upper region correction increment. Similarly, the change rates of the lower region normalized deviation and the lower region normalized deviation are used as dual inputs. Fuzzy inference is performed through the update rule table to obtain the lower region correction increment. The change rates of the normalized deviations in each region satisfy the following relationship:

[0059]

[0060] In the formula, The rate of change of the normalized deviation in the upper region is expressed in seconds. The rate of change of the lower region normalization deviation is expressed in seconds. The upper region normalization bias at the current sampling time is dimensionless; The upper zone normalization deviation of the previous control cycle is dimensionless. The lower region normalization bias at the current sampling time is dimensionless; The lower zone normalization deviation of the previous control cycle is dimensionless. To control the cycle, the unit of measurement is seconds; This is the current sampling time; This refers to the sampling time of the previous control cycle. Understandably, by introducing the rate of change of the deviation as the second input dimension of the fuzzy controller, the control logic gains the ability to predict temperature change trends in advance. By feeding this dual input into the update rule table of the previous step, which is based on rule-activated entropy synchronous calibration, and performing cross-fuzzy inference, the upper and lower correction increments can be accurately calculated separately.

[0061] After obtaining the correction increments for each zone, to prevent the control parameters from diverging or exceeding limits due to direct superposition of increments under extreme operating conditions, the upper and lower zone adjustment quantities are output based on the upper and lower zone correction increments, respectively. This includes: obtaining the basic proportional, integral, and derivative parameters of the PID controller for each zone; limiting the correction components of each parameter in the upper and lower zone correction increments to the preset proportional range of the corresponding basic proportional, integral, and derivative parameters to obtain the limited correction increment; summing the limited correction increment with the corresponding basic parameters to obtain the final PID parameters for each zone; and calculating the upper and lower zone adjustment quantities based on the final PID parameters for each zone. The upper and lower zone correction increments are each composed of three dimensions: a proportional correction component, an integral correction component, and a derivative correction component. The preset proportional range is preferably limited to the range of -50% to +50%. Understandably, the basic parameters calibrated during the process debugging phase represent the baseline of the equipment's basic thermal response. By strictly limiting and pruning the correction components derived from fuzzy inference, it is ensured that the final superimposed output of each PID parameter always remains positive and does not exceed 1.5 times the basic parameter. This eliminates the risk of output reversal, integral saturation, or even runaway from the underlying logic when extreme temperature differences occur.

[0062] It establishes a complete data and logic flow loop, from multi-point temperature data acquisition and dimensionality reduction, adaptive scaling of the domain space, and internal logic reconstruction of the rule table, to the final partitioned fuzzy inference and amplitude limiting protection. The upper and lower zones are solved independently without interference. Faced with the complex coupled conditions of over-temperature in the upper zone and under-temperature in the lower zone, the corresponding controllers can accurately adjust the power regulation of the corresponding heating elements, guiding the temperature of each zone to converge smoothly, quickly, and accurately towards the process target temperature.

[0063] like Figure 2 As shown in the figure, this figure is a comparison of the temperature following effect of the method of the present invention and the traditional method. When the temperature difference changes drastically from the degassing section to the main sintering section, the traditional fixed parameter controller has an adaptation blind zone, resulting in obvious temperature overshoot and slow response. However, the method of the present invention achieves smooth and accurate following of the temperature of each zone to the process target curve throughout the entire cycle by dynamically scaling the domain of discourse and reconstructing the rule table.

[0064] like Figure 3 As shown in the figure, the upper limit of the universe of discourse in this embodiment of the invention dynamically and adaptively changes with the extreme value of instantaneous temperature difference. When the system detects a sudden change in the extreme value of temperature difference caused by the switching of the sintering stage, the controller can amplify the upper limit of the universe of discourse in real time, effectively avoiding the problem of poor control accuracy caused by data truncation at the top in traditional control, and ensuring the global effectiveness of fuzzy logic.

[0065] An embodiment of the temperature field adaptive control system for the NdFeB vacuum sintering furnace provided by the present invention: The temperature field adaptive control system of the NdFeB vacuum sintering furnace includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned temperature field adaptive control method of the NdFeB vacuum sintering furnace is implemented.

[0066] The temperature field adaptive control system of the NdFeB vacuum sintering furnace also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0067] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0068] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for adaptive temperature field control in a NdFeB vacuum sintering furnace, characterized in that, Includes the following steps: The measured temperatures of each thermocouple in the upper and lower zones of the furnace are collected according to the control cycle, and the average values ​​are calculated to obtain the average temperature of the upper and lower zones. The difference between the maximum and minimum measured temperatures of each thermocouple is calculated to obtain the instantaneous temperature difference. The maximum instantaneous temperature difference within a preset time window is taken as the extreme value of the temperature difference; the extreme value of the temperature difference is multiplied by a preset expansion coefficient to obtain the upper limit of the universe of discourse, and the boundary of the fuzzy set is scaled proportionally according to the upper limit of the universe of discourse. The activation frequency of each fuzzy rule within a preset window is counted, and the information entropy is calculated based on each activation frequency to obtain the rule activation entropy. When the rule activation entropy is lower than the preset entropy threshold, the fuzzy set is re-divided according to the upper limit of the domain of discourse, and the output values ​​of each rule are recalibrated to obtain an updated rule table. Calculate the upper zone normalization deviation and lower zone normalization deviation relative to the process target temperature, respectively. Use the change rates of the upper zone normalization deviation and lower zone normalization deviation as inputs, perform fuzzy inference through the update rule table, and obtain the upper zone correction increment. Use the change rates of the lower zone normalization deviation and lower zone normalization deviation as dual inputs, perform fuzzy inference through the update rule table, and obtain the lower zone correction increment. Output the upper zone adjustment amount and lower zone adjustment amount according to the upper zone correction increment and lower zone correction increment, respectively.

2. The adaptive temperature field control method for the NdFeB vacuum sintering furnace according to claim 1, characterized in that, The process of multiplying the extreme temperature difference by a preset expansion coefficient to obtain the upper limit of the universe of discourse includes: multiplying the extreme temperature difference by a preset expansion coefficient to obtain the upper limit of the candidate universe of discourse; within a continuous preset verification window after the upper limit of the candidate universe of discourse takes effect, counting the number of sampling points in each control cycle whose instantaneous temperature difference does not exceed the upper limit of the candidate universe of discourse to obtain the number of qualified points; dividing the number of qualified points by the total number of sampling points in the verification window to obtain the universe of discourse coverage rate; when the universe of discourse coverage rate is not lower than the preset coverage rate threshold, the upper limit of the candidate universe of discourse is confirmed as the upper limit of the universe of discourse.

3. The adaptive temperature field control method for the NdFeB vacuum sintering furnace according to claim 2, characterized in that, When the domain coverage is lower than the preset coverage threshold, the process further includes: increasing the preset expansion coefficient to a preset upward adjustment value, multiplying the extreme temperature difference by the increased preset expansion coefficient to obtain a new upper limit of the candidate domain; recalculating the domain coverage, and repeating the above process until the domain coverage is not lower than the preset coverage threshold.

4. The adaptive temperature field control method for the NdFeB vacuum sintering furnace according to claim 1, characterized in that, After taking the maximum instantaneous temperature difference within a preset time window as the extreme value of the temperature difference, the method further includes: subtracting the extreme value of the temperature difference in the current statistical window from the extreme value of the temperature difference in the previous statistical window and dividing by the length of the preset time window to obtain the rate of change of the temperature difference; when the rate of change of the temperature difference is higher than the first rate threshold, shortening the update interval of the upper limit of the domain of discourse to the first interval; when the rate of change of the temperature difference is lower than the second rate threshold, extending the update interval of the upper limit of the domain of discourse to the second interval; when the rate of change of the temperature difference is not higher than the first rate threshold and not lower than the second rate threshold, maintaining the preset default interval; wherein, the first interval is less than the preset default interval, and the second interval is greater than the preset default interval.

5. The adaptive temperature field control method for the NdFeB vacuum sintering furnace according to claim 1, characterized in that, The preset entropy threshold is determined as follows: after the first complete sintering cycle, the activation entropy of each rule calculated throughout the sintering cycle is obtained; the activation entropy of each rule is sorted from smallest to largest, and the activation entropy value of the rule located at the preset quantile after sorting is taken as the preset entropy threshold.

6. The temperature field adaptive control method for the NdFeB vacuum sintering furnace according to claim 1, characterized in that, The output of upper and lower zone adjustment quantities based on the upper and lower zone correction increments includes: obtaining the basic proportional, integral, and derivative parameters of the PID controller for each zone; limiting the correction components of each parameter in the upper and lower zone correction increments to the preset proportional range of the corresponding basic proportional, integral, and derivative parameters to obtain the limited correction increment; summing the limited correction increment with the corresponding basic parameters to obtain the final PID parameters for each zone; and calculating the upper and lower zone adjustment quantities based on the final PID parameters for each zone.

7. The adaptive temperature field control method for a NdFeB vacuum sintering furnace according to claim 1, characterized in that, Re-dividing fuzzy sets according to the upper limit of the universe of discourse includes: dividing the interval from the upper limit of the negative universe of discourse to the upper limit of the positive universe of discourse into a preset number of fuzzy sets at equal intervals; dividing the interval length by the preset number to obtain the fuzzy set interval; and determining the boundary position of each fuzzy set sequentially from the upper limit of the negative universe of discourse, using the fuzzy set interval as the step size.

8. The adaptive temperature field control method for a NdFeB vacuum sintering furnace according to claim 1, characterized in that, Before collecting the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the process also includes setting the length of the preset time window to be an integer multiple of the control cycle.

9. The adaptive temperature field control method for a NdFeB vacuum sintering furnace according to claim 1, characterized in that, After collecting the measured temperatures of each thermocouple in the upper and lower zones of the furnace according to the control cycle, the process also includes: detecting outliers in the measured temperatures of each thermocouple, replacing the sampled values ​​that exceed the preset range of the corresponding thermocouple with the sampled values ​​of the thermocouple in the previous control cycle; performing mean filtering on the replaced measured temperatures of each thermocouple to obtain the filtered measured temperatures of the thermocouple; and using the filtered measured temperatures of the thermocouples to perform subsequent steps.

10. A temperature field adaptive control system for a NdFeB vacuum sintering furnace, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the temperature field adaptive control method of the NdFeB vacuum sintering furnace according to any one of claims 1-9 is implemented.

Images