Variable frequency microwave heating control method and system of vacuum furnace

Abstract

The invention relates to the technical field of microwave heating control, and discloses a variable-frequency microwave heating control method and system for a vacuum furnace, and the method comprises the following steps: S101, collecting first feature data of M blocks of a material, the first feature data comprising temperature, humidity, pressure, vacuum degree, microwave frequency and heating time; s102, preprocessing the first feature data to obtain a second feature sequence; s103, analyzing the moisture content of the material based on a preset moisture content prediction model, and outputting the moisture content of the M blocks; and S104, the control parameters of the blocks are optimized according to an improved particle swarm algorithm, optimal control parameters are obtained, and the control parameters comprise the microwave frequency and the heating time. According to the method, the microwave frequency and the heating time are optimized in a zoning mode, the moisture content prediction model is combined, and the drying uniformity and adaptability are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of microwave heating control, and in particular relates to a variable frequency microwave heating control method and system for a vacuum furnace. Background Art

[0002] The vacuum furnace microwave heating technology combines microwave energy and vacuum environment. Due to its rapid heating, high efficiency and energy saving, and suitability for heat-sensitive materials, it is widely used in material drying, sintering and heat treatment. However, this technology still has some shortcomings that need to be solved in practical applications.

[0003] First, heating unevenness is the core issue. In a vacuum environment, the distribution of microwave electromagnetic fields is affected by the shape of the material, the dielectric properties and the cavity design, resulting in inconsistent energy absorption, local overheating or insufficient heating, which affects the drying effect and material quality. Secondly, the dynamic control capability is insufficient. Existing technologies mostly use fixed frequency and power settings, lack real-time monitoring and adjustment, and are difficult to adapt to the dynamic changes in material properties and drying requirements, resulting in low drying efficiency or excessive energy consumption. In addition, energy utilization is not optimized enough. Different materials require differentiated heating parameters during the drying stage, but existing equipment lacks the ability to coordinate multi-parameter optimization of microwave frequency and heating time, which can easily lead to reduced efficiency. Summary of the invention

[0004] The invention provides a variable frequency microwave heating control method and system for a vacuum furnace, which solve the technical problems of uneven drying and insufficient dynamic control capability in the related art.

[0005] The present invention provides a variable frequency microwave heating control method for a vacuum furnace, comprising the following steps:

[0006] S101, evenly dividing the material in the vacuum furnace into M blocks, collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter;

[0007] S102, preprocessing the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index;

[0008] S103, analyzing the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second feature sequence, and the output is the moisture content of M blocks within a second preset time period;

[0009] S104, optimizing the control parameters of the block according to the improved particle swarm algorithm, and constructing the objective function through the moisture content prediction model, thereby obtaining the optimal control parameters, wherein the control parameters include: microwave frequency and heating time.

[0010] Further, the first feature data is preprocessed to obtain a second feature sequence, and the specific steps include:

[0011] S201, calculating the difference between the feature value of each block and the average feature value of its adjacent blocks, and if the difference is greater than a first preset threshold, treating the feature value of the block as an abnormal value and replacing it with the average feature value of its adjacent blocks;

[0012] S202, extracting a heating effect index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the heating effect index feature is: ,in, Represents the heating effect index characteristics of the block at the nth time point, represents the comprehensive weight coefficient, Indicates the microwave frequency of the block at the nth time point, N indicates the number of time points, and n indicates the nth time point. and represent the temperature of the block at the nth and n-1th time points respectively, t represents the first preset time interval, and Indicates the pressure of the block at the nth and n-1th time points, Indicates the initial pressure of the block, Indicates the humidity of the block at the nth time point, Indicates the initial humidity of the block, Indicates the vacuum degree of the block at the nth time point, Indicates the initial vacuum degree of the block, represents the pressure balance coefficient;

[0013] S203, extracting a water removal rate feature according to the processed first feature data, and splicing it after the first feature data. The calculation formula of the water removal rate feature is: ,in, represents the water removal rate characteristics of the block at the nth time point, represents the coupling coefficient that balances humidity, pressure, and microwave frequency, Indicates the humidity of the block at the n-1th time point, represents the proportionality coefficient between equilibrium pressure and humidity, Indicates the initial microwave frequency of the block;

[0014] S204, extracting a thermomechanical coupling index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the thermomechanical coupling index feature is: ,in, Represents the thermal coupling index characteristics of the block at the nth time point, represents the normalization coefficient;

[0015] S205, using a z-score normalization method to normalize the spliced ​​first feature data, and combining the processed first feature data at N time points to obtain a second feature sequence.

[0016] Furthermore, the gated recurrent unit is improved according to the short board effect and the redundancy effect, and a moisture content prediction model is constructed based on the improved gated recurrent unit. The structure of the improved gated recurrent unit includes: an input layer, a gating mechanism, a candidate hidden state layer, a redundant information compensation layer, a hidden state combination layer and an output layer;

[0017] The input layer is used to input the second feature sequence;

[0018] The gating mechanisms include: update gate, reset gate, and redundancy gate;

[0019] The candidate hidden state layer is used to generate candidate hidden states for the current time step;

[0020] The redundant information compensation layer is used to compensate the candidate hidden states according to the redundant features extracted by the redundant gate;

[0021] The hidden state combination layer is used to generate the hidden state of the current time step;

[0022] The output layer is used to output the hidden state of the last time step as the first updated feature and input it into the first classifier. The classification space of the first classifier represents the moisture content of the M blocks in the second preset time period.

[0023] Furthermore, the calculation formula of the gating mechanism includes:

[0024] Update Gate: ;

[0025] Reset the gate: ;

[0026] Redundant Doors: ;

[0027] in, represents the nth sequence unit of the second characteristic sequence, represents the hidden state of the n-1th sequence unit of the second feature sequence, , and represents the update gate, reset gate and redundant gate of the nth sequence unit of the second feature sequence, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the update gate, represents the first adjustment coefficient, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the reset gate, represents the time step weight coefficient, t represents the first preset time interval, , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the redundant gate, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

[0028] Furthermore, the calculation formulas for the candidate hidden state layer, the redundant information compensation layer, and the hidden state combination layer are:

[0029] Candidate hidden state layer: ;

[0030] Redundant information compensation layer: ;

[0031] Hidden state combination layer: ;

[0032] in, , and They represent the candidate hidden state, redundant feature and hidden state of the nth sequence unit of the second feature sequence respectively, , and A first weight coefficient, a second weight coefficient and a bias parameter representing the candidate hidden state, represents the redundant weight coefficient, Represents element-wise multiplication.

[0033] Furthermore, the control parameters of the block are optimized according to the improved particle swarm algorithm. The specific steps include:

[0034] S301, generating a particle swarm including K particles that meets the constraint conditions based on a two-dimensional Laplace distribution, where K is a custom parameter;

[0035] Each particle includes a position code and a velocity code, the position code is represented by the first particle vector, and the velocity code is represented by the second particle vector;

[0036] S302, constructing an objective function according to the water content prediction model;

[0037] S303, calculating the fitness value of each particle according to the objective function and the current position of each particle;

[0038] S304, updating the first particle vector and the second particle vector of the particles in the next generation particle swarm according to the first update strategy that introduces the Logistic map;

[0039] S305, when the fitness value of a particle in the particle swarm with the current iteration number j is greater than the maximum fitness value of the particle in history, the maximum fitness value of the particle in history is updated to the current fitness value; the particle with the largest fitness value in the particle swarm with the current iteration number j is taken as the global optimal solution;

[0040] S306, the current number of iterations is increased by 1. When the current number of iterations reaches the maximum number of iterations, the global optimal solution is output as the optimal control parameter, otherwise, the process returns to step S303, where the maximum number of iterations is a custom parameter.

[0041] Furthermore, the specific steps of S301 include:

[0042] S401, determining the parameter range of microwave frequency and heating time according to the constraint conditions;

[0043] S402, calculating the mean of the microwave frequency and the mean of the heating time according to the second characteristic sequence, and then obtaining the center and scale parameter of the two-dimensional Laplace distribution;

[0044] S403, initializing the first particle vector according to the center and scale parameters of the two-dimensional Laplace distribution, and then obtaining the second particle vector, and combining them to obtain a particle swarm containing K particles.

[0045] Furthermore, the calculation formula of the objective function is:

[0046] ;

[0047] in, represents the fitness value of the particle, Indicates the number of blocks, i and u both indicate block indexes, and They respectively represent the moisture contents of the i-th block and the u-th block obtained according to the moisture content prediction model.

[0048] The present invention also provides a variable frequency microwave heating control system for a vacuum furnace, comprising:

[0049] A data acquisition module, used for evenly dividing the material in the vacuum furnace into M blocks, and collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter;

[0050] A data preprocessing module preprocesses the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index;

[0051] A moisture content prediction module analyzes the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second characteristic sequence, and the output is the moisture content of M blocks within a second preset time period;

[0052] The control parameter optimization module optimizes the control parameters of the block according to the improved particle swarm algorithm, and constructs the objective function through the moisture content prediction model to obtain the optimal control parameters, where the control parameters include: microwave frequency and heating time.

[0053] The beneficial effects of the present invention are as follows: the present invention solves the problem of uneven drying caused by uneven energy distribution in traditional drying technology by dividing the material in the vacuum furnace into multiple blocks and independently optimizing the microwave frequency and heating time of each block; adopts an optimization method based on a moisture content prediction model to minimize the moisture content difference of each block, thereby significantly improving the consistency of drying quality;

[0054] The present invention uses an improved gated circulation unit to construct a moisture content prediction model, and introduces a combination mechanism of short-board effect and redundancy effect, so that the model can still maintain high-precision prediction capabilities when facing noise, missing data or outliers, providing reliable reference data for optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 The present invention is a flow chart of a variable frequency microwave heating control method for a vacuum furnace. DETAILED DESCRIPTION

[0056] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0057] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0058] like Figure 1 As shown, a variable frequency microwave heating control method for a vacuum furnace comprises the following steps:

[0059] S101, evenly dividing the material in the vacuum furnace into M blocks, collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter;

[0060] S102, preprocessing the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index;

[0061] S103, analyzing the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second feature sequence, and the output is the moisture content of M blocks within a second preset time period;

[0062] S104, optimizing the control parameters of the block according to the improved particle swarm algorithm, and constructing the objective function through the moisture content prediction model, thereby obtaining the optimal control parameters, wherein the control parameters include: microwave frequency and heating time.

[0063] In one embodiment of the present invention, a temperature sensor is arranged in the vacuum furnace to collect the temperature inside the furnace cavity, a humidity sensor is arranged to collect the humidity inside the furnace cavity, and a vacuum pressure sensor is arranged to collect the pressure and vacuum degree inside the furnace cavity. By equipping each block with an independent microwave generator, the microwave frequency of each block is monitored, and an independent heating control module is used to record the heating time of each block.

[0064] In one embodiment of the present invention, the first feature data is preprocessed to obtain a second feature sequence, and the specific steps include:

[0065] S201, calculating the difference between the feature value of each block and the average feature value of its adjacent blocks, and if the difference is greater than a first preset threshold, treating the feature value of the block as an abnormal value and replacing it with the average feature value of its adjacent blocks;

[0066] S202, extracting a heating effect index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the heating effect index feature is: ,in, Represents the heating effect index characteristics of the block at the nth time point, represents the comprehensive weight coefficient, Indicates the microwave frequency of the block at the nth time point, N indicates the number of time points, and n indicates the nth time point. and represent the temperature of the block at the nth and n-1th time points respectively, t represents the first preset time interval, and Indicates the pressure of the block at the nth and n-1th time points, Indicates the initial pressure of the block, Indicates the humidity of the block at the nth time point, Indicates the initial humidity of the block, Indicates the vacuum degree of the block at the nth time point, Indicates the initial vacuum degree of the block, represents the pressure balance coefficient;

[0067] S203, extracting a water removal rate feature according to the processed first feature data, and splicing it after the first feature data. The calculation formula of the water removal rate feature is: ,in, represents the water removal rate characteristics of the block at the nth time point, represents the coupling coefficient that balances humidity, pressure, and microwave frequency, Indicates the humidity of the block at the n-1th time point, represents the proportionality coefficient between equilibrium pressure and humidity, Indicates the initial microwave frequency of the block;

[0068] S204, extracting a thermomechanical coupling index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the thermomechanical coupling index feature is: ,in, Represents the thermal coupling index characteristics of the block at the nth time point, represents the normalization coefficient;

[0069] S205, using a z-score normalization method to normalize the spliced ​​first feature data, and combining the processed first feature data at N time points to obtain a second feature sequence.

[0070] In one embodiment of the present invention, a gated recurrent unit is improved according to the short board effect and the redundancy effect, and a moisture content prediction model is constructed based on the improved gated recurrent unit. The structure of the improved gated recurrent unit includes: an input layer, a gating mechanism, a candidate hidden state layer, a redundant information compensation layer, a hidden state combination layer and an output layer;

[0071] The input layer is used to input the second feature sequence;

[0072] The gating mechanism includes: update gate, reset gate and redundancy gate; the update gate determines the proportion of information in the hidden state that needs to be updated; the reset gate controls which parts of the previous hidden state need to be cleared to calculate the candidate hidden state of the current time step; the redundancy gate dynamically controls the extraction and utilization of redundant information to compensate for the shortcomings in the hidden state;

[0073] The candidate hidden state layer is used to generate candidate hidden states for the current time step;

[0074] The redundant information compensation layer is used to compensate the candidate hidden states according to the redundant features extracted by the redundant gate;

[0075] The hidden state combination layer is used to generate the hidden state of the current time step;

[0076] The output layer is used to output the hidden state of the last time step as the first updated feature and input it into the first classifier. The classification space of the first classifier represents the moisture content of the M blocks in the second preset time period.

[0077] The weakest board effect refers to the fact that certain key factors or features in the system become bottlenecks in the overall performance due to their low weights, insufficient importance or limited capabilities. Specifically, in time series modeling, some neurons in the hidden state may be critical to the task, but their dynamic weights are low or are affected by noise, which makes the model unable to fully utilize this information, thereby limiting the overall predictive ability. The redundancy effect refers to the existence of redundant or less important information in the system. Although this information contributes little under normal circumstances, it can compensate for the failure of key information or interference, thereby improving the stability and robustness of the system. In time series modeling, the redundancy effect saves secondary information in the hidden state to ensure that even if the main features fail, the model can still use this redundant information for effective prediction.

[0078] When capturing important features, traditional gated recurrent units may limit the expressive power of the model due to the low dynamic weights of some candidate hidden states (similar to the short board effect). The redundant gate dynamically manages redundant information to make up for the shortcomings of the short board candidate hidden states and ensure that key information is not lost, thereby balancing the short board limitation and redundant utilization. When faced with noise, missing data or failure of key features, the redundant gate can activate the secondary information in the hidden state for compensation, avoiding the degradation of model performance due to the failure of a single feature. For example, in the moisture content prediction of the present invention, this mechanism can enhance the adaptability to uncertainty and outliers.

[0079] In one embodiment of the present invention, the calculation formula of the gating mechanism includes:

[0080] Update Gate: ;

[0081] Reset the gate: ;

[0082] Redundant Doors: ;

[0083] in, represents the nth sequence unit of the second characteristic sequence, represents the hidden state of the n-1th sequence unit of the second feature sequence, , and represents the update gate, reset gate and redundant gate of the nth sequence unit of the second feature sequence, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the update gate, represents the first adjustment coefficient, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the reset gate, represents the time step weight coefficient, t represents the first preset time interval, , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the redundant gate, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

[0084] In one embodiment of the present invention, the calculation formulas for the candidate hidden state layer, the redundant information compensation layer and the hidden state combination layer are:

[0085] Candidate hidden state layer: ;

[0086] Redundant information compensation layer: ;

[0087] Hidden state combination layer: ;

[0088] in, , and They represent the candidate hidden state, redundant feature and hidden state of the nth sequence unit of the second feature sequence respectively, , and A first weight coefficient, a second weight coefficient and a bias parameter representing the candidate hidden state, represents the redundant weight coefficient, Represents element-by-element multiplication; through and Regulate hidden state generation and extract compensation information through redundant gates , improve stability.

[0089] In one embodiment of the present invention, a mean square error function is used as a loss function of a moisture content prediction model, and the predicted value of the loss function is obtained based on the output of the moisture content prediction model. The true value of the loss function represents the actual measured value, wherein the moisture content can be inferred by measuring the dielectric constant of the material block.

[0090] In one embodiment of the present invention, the control parameters of the block are optimized according to the improved particle swarm algorithm, and the specific steps include:

[0091] S301, generating a particle swarm including K particles that meets the constraint conditions based on a two-dimensional Laplace distribution, where K is a custom parameter;

[0092] Each particle includes a position code and a velocity code. The position code is represented by the first particle vector, which is expressed as: , the velocity encoding is represented by the second particle vector, which is expressed as: ,in, represents the microwave frequency and heating time from the first block to the Mth block, and They represent the microwave frequency and heating time of the Mth block respectively, and the value of each dimension of the second particle vector is 10% of the value of the corresponding dimension of the first particle vector;

[0093] The constraints include: microwave frequency and heating time are within the practical range that meets the vacuum furnace operating conditions;

[0094] S302, constructing an objective function according to the water content prediction model;

[0095] S303, calculating the fitness value of each particle according to the objective function and the current position of each particle;

[0096] S304, updating the first particle vector and the second particle vector of the particles in the next generation particle swarm according to the first update strategy that introduces the Logistic map, the update formula of the first particle vector is: , the update formula of the second particle vector is: ,in, It represents the second particle vector of the particle in the particle swarm when the iteration number is j+1. represents the inertia weight coefficient, represents the second particle vector of the particle in the particle swarm when the number of iterations is j, and denote the first learning coefficient and the second learning coefficient respectively, and is a random number ranging from 0 to 1. It represents the optimal second particle vector in the history of the particle when the number of iterations is j. The optimal second particle vector represents the speed of the particle with the highest fitness value. represents the optimal second particle vector of the entire particle swarm when the number of iterations is j, represents the first particle vector of the particle in the particle swarm when the number of iterations is j, Indicates the disturbance intensity coefficient, ranging from 0.01 to 0.1. represents the chaos control parameter, preferably, Take 3.9, It represents the perturbation value when the number of iterations is j. It is worth noting that the initial value of the perturbation value is The value range is from 0 to 1. represents the first particle vector of the particle in the particle swarm when the number of iterations is j+1, Represents the chaotic sequence generated by Logistic mapping;

[0097] S305, when the fitness value of a particle in the particle swarm with the current iteration number j is greater than the maximum fitness value of the particle in history, the maximum fitness value of the particle in history is updated to the current fitness value; the particle with the largest fitness value in the particle swarm with the current iteration number j is taken as the global optimal solution;

[0098] S306, the current number of iterations is increased by 1. When the current number of iterations reaches the maximum number of iterations, the global optimal solution is output as the optimal control parameter, otherwise the process returns to step S303, wherein the maximum number of iterations is a custom parameter. Preferably, the maximum number of iterations is set to 200.

[0099] In one embodiment of the present invention, by initializing the particle swarm based on the two-dimensional Laplace distribution, the improved algorithm can intelligently generate the initial solution according to the parameter constraints and the historical mean. This initialization method not only ensures the diversity of the solution, but also can flexibly adjust the particle distribution according to the range of microwave frequency and heating time. At the same time, the introduction of the disturbance term enhances the adaptability to complex constraint problems in the particle update process, effectively improves the algorithm's ability to handle nonlinear and multi-objective problems, and makes it more suitable for practical scenarios such as drying uniformity optimization.

[0100] In one embodiment of the present invention, the specific steps of S301 include:

[0101] S401, determining the parameter range of microwave frequency and heating time according to the constraint conditions;

[0102] S402, calculating the mean of the microwave frequency and the mean of the heating time according to the second characteristic sequence, and then obtaining the center and scale parameters of the two-dimensional Laplace distribution. Specifically, the probability density function of the two-dimensional Laplace distribution is: ,in, represents a two-dimensional Laplace distribution, and represent the microwave frequency and heating time of the mth block respectively, represents the center of the two-dimensional Laplace distribution, i.e., a vector consisting of the mean of the microwave frequency and the mean of the heating time, b represents the scale parameter of the two-dimensional Laplace distribution, and , express and The Manhattan distance between

[0103] S403, initializing the first particle vector according to the center and scale parameters of the two-dimensional Laplace distribution, and then obtaining the second particle vector, and combining them to obtain a particle swarm containing K particles.

[0104] In one embodiment of the present invention, the calculation formula of the objective function is:

[0105] ;

[0106] in, represents the fitness value of the particle, Indicates the number of blocks, i and u both indicate block indexes, and They respectively represent the moisture contents of the i-th block and the u-th block obtained according to the moisture content prediction model.

[0107] In one embodiment of the present invention, a variable frequency microwave heating control system for a vacuum furnace is also provided, comprising:

[0108] A data acquisition module, used for evenly dividing the material in the vacuum furnace into M blocks, and collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter;

[0109] A data preprocessing module preprocesses the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index;

[0110] A moisture content prediction module analyzes the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second characteristic sequence, and the output is the moisture content of M blocks within a second preset time period;

[0111] The control parameter optimization module optimizes the control parameters of the block according to the improved particle swarm algorithm, and constructs the objective function through the moisture content prediction model to obtain the optimal control parameters, where the control parameters include: microwave frequency and heating time.

[0112] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A variable frequency microwave heating control method for a vacuum furnace, characterized in that: The following steps are involved: S101, evenly dividing the material in the vacuum furnace into M blocks, collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter; S102, preprocessing the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index; S103, analyzing the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second feature sequence, and the output is the moisture content of M blocks within a second preset time period; S104, optimizing the control parameters of the block according to the improved particle swarm algorithm, and constructing the objective function through the moisture content prediction model, thereby obtaining the optimal control parameters, wherein the control parameters include: microwave frequency and heating time.

2. The variable frequency microwave heating control method of a vacuum furnace according to claim 1, characterized in that: Preprocessing the first feature data to obtain a second feature sequence includes: S201, calculating the difference between the feature value of each block and the average feature value of its adjacent blocks, and if the difference is greater than a first preset threshold, treating the feature value of the block as an abnormal value and replacing it with the average feature value of its adjacent blocks; S202, extracting a heating effect index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the heating effect index feature is: ,in, represents the heating effect index characteristic of the block at the nth time point, represents the comprehensive weight coefficient, Indicates the microwave frequency of the block at the nth time point, N indicates the number of time points, and n indicates the nth time point. and represent the temperature of the block at the nth and n-1th time points respectively, t represents the first preset time interval, and Indicates the pressure of the block at the nth and n-1th time points, Indicates the initial pressure of the block, Indicates the humidity of the block at the nth time point, Indicates the initial humidity of the block, Indicates the vacuum degree of the block at the nth time point, Indicates the initial vacuum degree of the block, represents the pressure balance coefficient; S203, extracting a water removal rate feature according to the processed first feature data, and splicing it after the first feature data. The calculation formula of the water removal rate feature is: ,in, represents the water removal rate characteristics of the block at the nth time point, represents the coupling coefficient that balances humidity, pressure, and microwave frequency, Indicates the humidity of the block at the n-1th time point, represents the proportionality coefficient between equilibrium pressure and humidity, Indicates the initial microwave frequency of the block; S204, extracting a thermomechanical coupling index feature according to the processed first feature data, and concatenating the feature after the first feature data. The calculation formula of the thermomechanical coupling index feature is: ,in, Represents the thermal coupling index characteristics of the block at the nth time point, represents the normalization coefficient; S205, using a z-score normalization method to normalize the spliced ​​first feature data, and combining the processed first feature data at N time points to obtain a second feature sequence.

3. The variable frequency microwave heating control method of a vacuum furnace according to claim 1, characterized in that: The gated recurrent unit is improved according to the short board effect and the redundant effect, and a moisture content prediction model is constructed based on the improved gated recurrent unit. The structure of the improved gated recurrent unit includes: an input layer, a gating mechanism, a candidate hidden state layer, a redundant information compensation layer, a hidden state combination layer and an output layer; The input layer is used to input the second feature sequence; The gating mechanisms include: update gate, reset gate, and redundancy gate; The candidate hidden state layer is used to generate candidate hidden states for the current time step; The redundant information compensation layer is used to compensate the candidate hidden states according to the redundant features extracted by the redundant gate; The hidden state combination layer is used to generate the hidden state of the current time step; The output layer is used to output the hidden state of the last time step as the first updated feature and input it into the first classifier. The classification space of the first classifier represents the moisture content of the M blocks in the second preset time period.

4. The variable frequency microwave heating control method of a vacuum furnace according to claim 3, characterized in that: The calculation formula of the gating mechanism includes: Update Gate: ; Reset the gate: ; Redundant Doors: ; in, represents the nth sequence unit of the second characteristic sequence, represents the hidden state of the n-1th sequence unit of the second feature sequence, , and represents the update gate, reset gate and redundant gate of the nth sequence unit of the second feature sequence, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the update gate, represents the first adjustment coefficient, , and Respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the reset gate, represents the time step weight coefficient, t represents the first preset time interval, , and They respectively represent the first weight coefficient, the second weight coefficient and the bias parameter of the redundant gate, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

5. The variable frequency microwave heating control method for a vacuum furnace according to claim 4, characterized in that: The calculation formulas for the candidate hidden state layer, redundant information compensation layer, and hidden state combination layer are: Candidate hidden state layer: ; Redundant information compensation layer: ; Hidden state combination layer: ; in, , and They represent the candidate hidden state, redundant feature and hidden state of the nth sequence unit of the second feature sequence respectively, , and A first weight coefficient, a second weight coefficient and a bias parameter representing the candidate hidden state, represents the redundant weight coefficient, Represents element-wise multiplication.

6. The variable frequency microwave heating control method for a vacuum furnace according to claim 1, characterized in that: The control parameters of the block are optimized according to the improved particle swarm algorithm. The specific steps include: S301, generating a particle swarm including K particles that meets the constraint conditions based on a two-dimensional Laplace distribution, where K is a custom parameter; Each particle includes a position code and a velocity code, the position code is represented by the first particle vector, and the velocity code is represented by the second particle vector; S302, constructing an objective function according to the water content prediction model; S303, calculating the fitness value of each particle according to the objective function and the current position of each particle; S304, updating the first particle vector and the second particle vector of the particles in the next generation particle swarm according to the first update strategy that introduces the Logistic map; S305, when the fitness value of a particle in the particle swarm with the current iteration number j is greater than the maximum fitness value of the particle in history, the maximum fitness value of the particle in history is updated to the current fitness value; the particle with the largest fitness value in the particle swarm with the current iteration number j is taken as the global optimal solution; S306, the current number of iterations is increased by 1. When the current number of iterations reaches the maximum number of iterations, the global optimal solution is output as the optimal control parameter, otherwise, the process returns to step S303, where the maximum number of iterations is a custom parameter.

7. The variable frequency microwave heating control method for a vacuum furnace according to claim 6, characterized in that: The specific steps of S301 include: S401, determining the parameter range of microwave frequency and heating time according to the constraint conditions; S402, calculating the mean of the microwave frequency and the mean of the heating time according to the second characteristic sequence, and then obtaining the center and scale parameter of the two-dimensional Laplace distribution; S403, initializing the first particle vector according to the center and scale parameters of the two-dimensional Laplace distribution, thereby obtaining the second particle vector, and combining them to obtain a particle swarm including K particles.

8. The variable frequency microwave heating control method for a vacuum furnace according to claim 6, characterized in that: The objective function is calculated as follows: ； in, represents the fitness value of the particle, Indicates the number of blocks, i and u both indicate block indexes, and They respectively represent the moisture contents of the i-th block and the u-th block obtained according to the moisture content prediction model.

9. A variable frequency microwave heating control system for a vacuum furnace, characterized in that: A variable frequency microwave heating control method for implementing a vacuum furnace according to any one of claims 1 to 8, comprising: A data acquisition module, used for evenly dividing the material in the vacuum furnace into M blocks, and collecting first characteristic data of the M blocks within a first preset time period at a first preset time interval, wherein the first characteristic data includes: temperature, humidity, pressure, vacuum degree, microwave frequency and heating time, and M is a custom parameter; A data preprocessing module preprocesses the first feature data to obtain a second feature sequence, wherein the nth sequence unit of the second feature sequence represents the preprocessed first feature data at the nth time point, and n represents a time point index; A moisture content prediction module analyzes the moisture content of the material based on a preset moisture content prediction model, the input of the moisture content model is the second characteristic sequence, and the output is the moisture content of M blocks within a second preset time period; The control parameter optimization module optimizes the control parameters of the block according to the improved particle swarm algorithm, and constructs the objective function through the moisture content prediction model to obtain the optimal control parameters, where the control parameters include: microwave frequency and heating time.

Images