Multi-stage temperature accurate control method based on neural network algorithm

Abstract

The invention discloses a multi-stage temperature accurate control method based on a neural network algorithm, relates to the field of oven temperature control, and particularly comprises the steps of setting a final target temperature # imgabs0 #, and dividing a heating process of heating to the final target temperature # imgabs1 # into a rapid temperature rise stage, a temperature rise control stage and an accurate temperature control stage, an intermediate target temperature # imgabs2 # is set in each stage so as to gradually approach the final target temperature # imgabs3 #; according to the established oven system model, the influence of different intermediate target temperatures # imgabs4 # on temperature response is analyzed, and a proper intermediate target temperature # imgabs5 # combination is found; the heating temperature # imgabs6 # of the heating wire and the temperature # imgabs7 # in the oven are detected in real time through a temperature sensor, and then segmented control is conducted; through segmented control, the temperature rise rate and control precision of each stage are optimized, overshoot is reduced, the response speed is increased, the target temperature is gradually approached, and temperature overshoot caused by one-time large-amplitude temperature rise is avoided.

Description

Technical Field

[0001] The invention relates to the technical field of oven temperature control, and in particular to a multi-stage temperature precision control method based on a neural network algorithm. Background Art

[0002] In the temperature control of the oven, it is required to quickly heat up to improve efficiency, and to minimize temperature overshoot to ensure cooking quality and safety. However, due to the nonlinearity, time-varying and hysteresis characteristics of the oven system, traditional single PID control is difficult to meet the above requirements. Summary of the invention

[0003] In order to solve the defects of the prior art, the present invention provides a multi-stage temperature precision control method based on a neural network algorithm.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0005] The present invention provides a multi-stage temperature precision control method based on a neural network algorithm, comprising the following steps:

[0006] Step 1: Set the final target temperature and blower speed value , will be heated to the final target temperature The heating process is divided into a rapid heating stage, a controlled heating stage and a precise temperature control stage, and an intermediate target temperature is set in each stage. , where i=1, 2, 3, to gradually approach the final target temperature ;

[0007] Step 2: Establish an oven system model and analyze different intermediate target temperatures based on the established oven system model The impact on temperature response, find the intermediate target temperature that minimizes the total rise time and overshoot combination;

[0008] Step 3: Set up a temperature sensor to use the temperature sensor to measure the heating temperature of the heating wire during the heating process. The temperature inside the oven Conduct real-time detection;

[0009] Step 4: Start heating. First, in the rapid heating stage, the heating wire is heated at full power to make the heating temperature Rapidly rise to the intermediate target temperature Then switch the control mode of the heating wire to PID control or fuzzy self-tuning PID control, and gradually adjust the power output of the heating wire to make the temperature inside the oven Gradually increase to the next intermediate target temperature , the neural network algorithm is trained on the previous PID parameters to obtain the parameter output model of the heating wire. , , The best predicted PID parameters are obtained through the parameter output model, and the heating wire is PID controlled according to the best predicted PID parameters to make the temperature inside the oven Rise to the last intermediate target temperature , and continue to heat up to the set final target temperature Get closer and finally maintain the temperature inside the oven At the set final target temperature The corresponding preset small range fluctuations.

[0010] As a preferred technical solution of the present invention, the method of establishing the oven system model in step 2 is to first establish the energy balance equation of the heating wire according to the principle of heat transfer:

[0011]

[0012] The above formula is recorded as formula 1, where: Is the quality of the heating wire, is the specific heat capacity of the heating wire, is the function of the heating temperature of the heating wire changing with time, is the real-time input electrical power function, is the function of convective heat transfer rate changing with time, is the function of radiation heat transfer rate changing with time;

[0013] The convective heat transfer formula is:

[0014] The above formula is recorded as formula 2, where: : Convective heat transfer rate as a function of time; : Convective heat transfer coefficient as a function of time, is the surface temperature of the heating wire, The value of The value of is the air temperature in the heating chamber.

[0015] According to the thermal emissivity formula,

[0016] The above is recorded as formula 3, where: is the function of the radiation heat transfer rate changing with time, is the surface emissivity of the heating wire, is the Stefan–Boltzmann constant, is the function of the heating temperature of the heating wire changing with time, is the absolute temperature of the surrounding environment. The influence of the surrounding temperature in the same heating process is negligible and is regarded as a constant here. is the surface area of ​​the heating wire;

[0017] Establish the energy balance equation for the oven air fluid:

[0018]

[0019] The above formula is recorded as formula 4, where is the air quality inside the oven; is the specific heat capacity of air, is a function of heat loss changing with time; according to the above formulas 1, 2, 3, and 4, we can get a system model of the oven including heat transfer and fluid mechanics:

[0020]

[0021] Analyze different intermediate target temperatures based on the oven system model The impact on the oven temperature response, find multiple sets of intermediate target temperatures that make the total rise time short and the overshoot small , and set the intermediate target temperature Optimize to get the best combination of intermediate target temperature , the equation at this time also serves as the solution condition for solving the performance index function.

[0022] As a preferred technical solution of the present invention, the intermediate target temperature The optimization method is to establish a function that defines performance indicators :

[0023]

[0024] In the above formula is the control end time, is a trade-off factor used to balance temperature error and power change rate, is the time-varying function of input power, It indicates the temperature inside the oven. It represents the final target temperature; then for multiple groups of intermediate target temperatures Combined simulation, take the corresponding temperature response and control performance, the temperature response and control performance measurement indicators include: rise time , Overshoot , Adjust time ; Select the intermediate target temperature that minimizes both the temperature response and the control performance measure combination.

[0025] As a preferred technical solution of the present invention, when training the model, a series of optimal PID parameter values ​​are obtained by using the data training set and the gradient descent method. The PID principle formula is as follows:

[0026]

[0027] Perform parameter tuning and take the critical proportional gain , is the minimum value for the system to produce sustained oscillation , critical period is the corresponding oscillation period, then the following PID parameters are obtained:

[0028]

[0029] The calculated value here is used as the starting value, and the adjustment process uses the gradient descent method to find the optimal value for training. , , Values, thereby obtaining parameter values ​​for data training;

[0030] Among them is is the proportional adjustment coefficient in PID control, is the integral adjustment coefficient in PID control, It is the differential adjustment coefficient in PID control. Using the data training set, a parameter output model of the heating wire based on the neural network algorithm is constructed;

[0031] The nth output value is: , , , through temperature The curve is judged and selected, and the rise time is selected , Overshoot , Adjust time The smallest curve, and then select the best , , The values ​​are taken as elements in the training set.

[0032] As a preferred technical solution of the present invention, the method of performing fuzzy self-tuning PID control on the heating wire in step 4 is to input power to the heating wire The control error is:

[0033]

[0034] Using fuzzy logic, the input control error changes over time function and control error change rate Right now , adjust PID parameters, and the indicators to be measured include rise time :Temperature from 10% to 90% Time required to set value; overshoot : Maximum percentage exceeding the set value; adjustment time : The time required for the system to enter and remain within the range of ±2% of the set value; the smaller Tr, Mp, and Ts are, the better.

[0035] As a preferred technical solution of the present invention, the method of performing neural network algorithm training on the previous PID parameters in step 4 is to record the input, output and optimal PID parameters of the oven system model under different working conditions as a data training set;

[0036] Use the data training set to train the neural network algorithm; , , For input, (t) is the automatic input read by the sensor, V is the wind speed value input by the user, expressed in K p , K i , K d As output, the optimal PID parameters are used as labels.

[0037] The beneficial effects of the present invention are:

[0038] 1. This multi-stage temperature precision control method based on neural network algorithm sets multiple intermediate target temperatures , using a phased control strategy. In the initial stage, the heating wire runs at full power to make its heating temperature Rapidly rise to the first intermediate target temperature This strategy utilizes the thermal inertia of the heating wire and can quickly increase the temperature inside the oven in the shortest time, greatly shortening the preheating time and improving the working efficiency of the equipment. After reaching the intermediate target temperature, it switches to fine PID control or fuzzy self-tuning PID control to gradually adjust the power output of the heating wire. and the temperature inside the oven Real-time monitoring and control of the temperature inside the oven Gradually increase to the next intermediate target temperature , the neural network algorithm is trained for the PID parameters in the temperature control stage, and the parameter output model of the heating wire is obtained. After training, the target temperature is given. , , The best predicted PID parameters are obtained through the parameter output model, and the heating wire is PID controlled according to the best predicted PID parameters to make the temperature inside the oven Rise to the last intermediate target temperature , and continue to heat up to the set final target temperature Get closer and finally maintain the temperature inside the oven At the set final target temperature In summary, the present invention optimizes the temperature rise rate and control accuracy of each stage through segmented control, reduces overshoot, improves response speed, and gradually approaches the target temperature to avoid temperature overshoot caused by a one-time large-scale temperature increase. In the initial stage, a larger heating power can be allowed to speed up the temperature rise speed. When approaching the target temperature, the control is refined to improve temperature stability.

[0039] 2. This multi-stage temperature precision control method based on neural network algorithm establishes the state space model of the oven according to heat transfer and fluid mechanics, which can accurately estimate the time constant and lag time of the system, thereby effectively compensating for thermal inertia and lag in the control strategy and improving the accuracy and response speed of temperature control.

[0040] 3. The present invention reduces the ineffective power output of the heating wire through precise temperature control, avoiding overheating and frequent power adjustment. The staged control strategy enables the heating wire to operate at full power when needed, and reduces the power when approaching the target temperature, thereby optimizing the energy utilization efficiency and reducing the operating cost of the equipment. In addition, the present invention can ensure the uniformity and stability of the oven temperature, avoiding the influence of temperature fluctuations on the food baking process, which not only improves the baking quality of the food, but also reduces the food defect rate caused by unstable temperature, thereby improving the overall quality of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 It is a flow chart of a multi-stage temperature precision control method based on a neural network algorithm of the present invention;

[0043] Figure 2 This is a temperature adjustment effect diagram of a multi-stage temperature precision control method based on a neural network algorithm of the present invention;

[0044] Figure 3 This is a comparison chart of the effects of a multi-stage temperature precision control method based on a neural network algorithm of the present invention and the effects of not using the method of the present invention. DETAILED DESCRIPTION

[0045] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0046] Example: Figure 1 As shown, the present invention provides a multi-stage temperature precision control method based on a neural network algorithm, comprising the following steps:

[0047] Set the final target temperature and blower speed value , will be heated to the final target temperature The heating process is divided into a rapid heating stage, a controlled heating stage and a precise temperature control stage, and an intermediate target temperature is set in each stage. , where i=1, 2, 3, to gradually approach the final target temperature ;

[0048] Step 2: Establish an oven system model and analyze different intermediate target temperatures based on the established oven system model The impact on temperature response, find the intermediate target temperature that minimizes the total rise time and overshoot combination;

[0049] Step 3: Set up a temperature sensor to use the temperature sensor to measure the heating temperature of the heating wire during the heating process. The temperature inside the oven Conduct real-time detection;

[0050] Step 4: Start heating. First, in the rapid heating stage, the heating wire is heated at full power to make the heating temperature Rapidly rise to the intermediate target temperature Then switch the control mode of the heating wire to PID control or fuzzy self-tuning PID control, and gradually adjust the power output of the heating wire to make the temperature inside the oven Gradually increase to the next intermediate target temperature , the neural network algorithm is trained on the previous PID parameters to obtain the parameter output model of the heating wire. , , The best predicted PID parameters are obtained through the parameter output model, and the heating wire is PID controlled according to the best predicted PID parameters to make the temperature inside the oven Rise to the last intermediate target temperature , and continue to heat up to the set final target temperature Get closer and finally maintain the temperature inside the oven At the set final target temperature The corresponding preset small range fluctuations. By setting multiple intermediate target temperatures , making it heat up to temperature Rapidly rise to the first intermediate target temperature This strategy uses the thermal inertia of the heating wire to quickly increase the temperature inside the oven in the shortest possible time. , greatly shortening the preheating time and improving the working efficiency of the equipment; when reaching the intermediate target temperature Then switch the control mode of the heating wire to PID control or fuzzy self-tuning PID control, and gradually adjust the power output of the heating wire to make the temperature inside the oven Gradually increase to the next intermediate target temperature , the neural network algorithm is trained for the PID parameters in the temperature control stage, and the parameter output model of the heating wire is obtained. After training, the target temperature is given. , , The best predicted PID parameters are obtained through the parameter output model, and the heating wire is PID controlled according to the best predicted PID parameters to make the temperature inside the oven Rise to the last intermediate target temperature , and continue to heat up to the set final target temperature Get closer and finally maintain the temperature inside the oven At the set final target temperature Small fluctuation range can effectively suppress temperature overshoot and ensure that the temperature stably reaches and maintains the final target temperature about.

[0051] The present invention optimizes the temperature rise rate and control accuracy of each stage through segmented control, reduces overshoot, improves response speed, and gradually approaches the target temperature to avoid temperature overshoot caused by a one-time large-scale temperature increase. In the initial stage, a larger heating power can be allowed to accelerate the heating speed. When approaching the target temperature, the control is refined to improve temperature stability.

[0052] Since the oven may be affected by load changes (such as the type and amount of food in the oven) and ambient temperature fluctuations in actual use, fuzzy self-tuning PID control and neural network PID parameter tuning methods are introduced. By adjusting the parameters of the PID controller in real time, the algorithm can adapt to changes in system characteristics and maintain stable control performance. This adaptive capability improves the robustness of the system, allowing the oven to maintain the best temperature control effect under various working conditions.

[0053] The method for establishing the oven system model in step 2 is to first establish the energy balance equation of the heating wire according to the principle of heat transfer:

[0054]

[0055] The above formula is recorded as formula 1, where: Is the quality of the heating wire, is the specific heat capacity of the heating wire, is the function of the heating temperature of the heating wire changing with time, is the real-time input electrical power function, is the function of convective heat transfer rate changing with time, is the function of radiation heat transfer rate changing with time;

[0056] The convective heat transfer formula is:

[0057] The above formula is recorded as formula 2, where: : Convective heat transfer rate as a function of time; : Convective heat transfer coefficient as a function of time, is the surface temperature of the heating wire. Since the distance between the sensor and the heating wire is relatively close (considering that the actual installation cannot be directly placed on the surface of the heating wire), and the actual measurement results show that The value of The value of is the air temperature in the heating chamber;

[0058] According to the thermal emissivity formula,

[0059] The above is recorded as formula 3, where: is the function of the radiation heat transfer rate changing with time, is the surface emissivity of the heating wire, is the Stefan–Boltzmann constant, is the function of the heating temperature of the heating wire changing with time, is the absolute temperature of the surrounding environment. The influence of the surrounding temperature in the same heating process is negligible and is regarded as a constant here. is the surface area of ​​the heating wire;

[0060] Establish the energy balance equation for the oven air fluid:

[0061]

[0062] The above formula is recorded as formula 4, where is the quality of air inside the oven; is the specific heat capacity of air, is a function of heat loss changing with time; according to the above formulas 1, 2, 3, and 4, we can get a system model of the oven including heat transfer and fluid mechanics:

[0063]

[0064] Analyze different intermediate target temperatures based on the oven system model The impact on the oven temperature response, find multiple sets of intermediate target temperatures that make the total rise time short and the overshoot small , and set the intermediate target temperature Optimize to get the best combination of intermediate target temperature , the equation at this time also serves as the solution condition for solving the performance index function. The present invention accurately describes the dynamic behavior of the system through precise physical modeling, combined with the principles of heat transfer and fluid mechanics. Using experimental data and parameter identification methods, the algorithm can accurately estimate the time constant and lag time of the system, thereby effectively compensating for thermal inertia and lag in the control strategy, and improving the accuracy and response speed of temperature control.

[0065] Wherein, the intermediate target temperature The optimization method is to establish a function that defines performance indicators :

[0066]

[0067] In the above formula is the control end time, is a trade-off factor used to balance temperature error and power change rate, is the time-varying function of input power, It indicates the temperature inside the oven. It represents the final target temperature; then for multiple groups of intermediate target temperatures Combine MATLAB / Simulink simulation to obtain the corresponding temperature response and control performance. The indicators for measuring temperature response and control performance include: rise time , Overshoot , Adjust time ; Select the intermediate target temperature that minimizes both the temperature response and the control performance measure This effectively achieves rapid temperature ramping while minimizing temperature overshoot.

[0068] Among them, when training the model, a series of optimal PID parameter values ​​are obtained through the data training set using the gradient descent method. The PID principle formula is as follows:

[0069]

[0070] Perform parameter tuning and take the critical proportional gain , is the minimum value for the system to produce sustained oscillation , critical period is the corresponding oscillation period, then the following PID parameters are obtained:

[0071]

[0072] The calculated value here is used as the starting value, and the adjustment process uses the gradient descent method to find the optimal value for training. , , Values, thereby obtaining parameter values ​​for data training;

[0073] Among them is is the proportional adjustment coefficient in PID control, is the integral adjustment coefficient in PID control, It is the differential adjustment coefficient in PID control. Using the data training set, a parameter output model of the heating wire based on the neural network algorithm is constructed;

[0074] The nth output value is: , , , through temperature The curve is judged and selected, and the rise time is selected , Overshoot , Adjust time The smallest curve, and then select the best , , The present invention successfully solves the key problem in the temperature control of commercial ovens. It takes into account the requirements of rapid temperature rise and minimization of temperature overshoot, improves the accuracy and stability of temperature control, optimizes energy efficiency, and enhances the system's adaptability and robustness.

[0075] The method for performing fuzzy self-tuning PID control on the heating wire in step 4 is to input power to the heating wire. The control error is:

[0076]

[0077] Using fuzzy logic, the input control error changes over time function and control error change rate Right now , adjust PID parameters, and the indicators to be measured include rise time :Temperature from 10% to 90% Time required to set value; overshoot : Maximum percentage exceeding the set value; adjustment time : The time required for the system to enter and remain within the range of ±2% of the set value; the smaller Tr, Mp, and Ts are, the better.

[0078] Wherein, the method of performing neural network algorithm training on the previous PID parameters in step 4 is to record the input, output and optimal PID parameters of the oven system model under different working conditions as a data training set;

[0079] Use the data training set to train the neural network algorithm; , , For input, (t) is the automatic input read by the sensor, V is the wind speed value input by the user, expressed in K p , K i , K d As output, the optimal PID parameters are used as labels, where is the proportional adjustment coefficient in PID control, is the integral adjustment coefficient in PID control, It is the differential adjustment coefficient in PID control. Using the data training set, a parameter output model of the heating wire based on the neural network algorithm is constructed. Using a large amount of experimental data, the neural network is trained using the back propagation algorithm to predict the optimal PID parameters. By adjusting the parameters of the PID controller in real time, the algorithm can adapt to changes in system characteristics and maintain stable control performance. This adaptive capability improves the robustness of the system, allowing the oven to maintain the best temperature control effect under various working conditions. After constructing a system neural network algorithm model, in the subsequent temperature setting process of the oven, different ambient temperatures, different wind speeds set by the user, and different target temperatures can all obtain the optimal PID parameters without manually adjusting the PID parameters.

[0080] The method provided by the present invention is used for testing. Figure 2 As shown, the curve is the temperature in the cavity measured separately. The temperature starts to rise before 16:00:00, the preset temperature is 170°C, and then the temperature is reduced. At 17:05:00, the preset temperature is 305°C, and the temperature starts to rise from 60°C, an overshoot occurs, and it stabilizes at about 305°C. Then the temperature is reduced to 17:35:00, and then the method of the present invention is used to perform three staged temperature increase tests. The method of the present invention is used to increase the temperature in the following three increases.

[0081] like Figure 3As shown, the method of the present invention is used for temperature control before 22:25:00, and the method of the present invention is turned off for temperature control after 22:25:00. The light-colored curve is the temperature of the cavity, and the dark-colored curve is the temperature of the heating wire. It can be clearly seen that due to the large inertia of the system, the temperature in the cavity seriously lags behind the real-time temperature of the heating tube.

[0082] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-stage temperature precision control method based on a neural network algorithm, characterized in that: The following steps are involved: Step 1: Set the final target temperature and blower speed value , will be heated to the final target temperature The heating process is divided into a rapid heating stage, a controlled heating stage and a precise temperature control stage, and an intermediate target temperature is set in each stage. , where i=1, 2, 3, to gradually approach the final target temperature ; Step 2: Establish an oven system model and analyze different intermediate target temperatures based on the established oven system model The impact on temperature response, find the intermediate target temperature that minimizes the total rise time and overshoot combination; Step 3: Set up a temperature sensor to use the temperature sensor to measure the heating temperature of the heating wire during the heating process. The temperature inside the oven Conduct real-time detection; Step 4: Start heating. First, in the rapid heating stage, the heating wire is heated at full power to make the heating temperature Rapidly rise to the intermediate target temperature Then switch the control mode of the heating wire to PID control or fuzzy self-tuning PID control, and gradually adjust the power output of the heating wire to make the temperature inside the oven Gradually increase to the next intermediate target temperature , the neural network algorithm is trained on the previous PID parameters to obtain the parameter output model of the heating wire. , , The best predicted PID parameters are obtained through the parameter output model, and the heating wire is PID controlled according to the best predicted PID parameters to make the temperature inside the oven Rise to the last intermediate target temperature , and continue to heat up to the set final target temperature Get closer and finally maintain the temperature inside the oven At the set final target temperature The corresponding preset small range fluctuations.

2. The multi-stage temperature precision control method based on a neural network algorithm according to claim 1 is characterized in that: The method of establishing the oven system model in step 2 is to first establish the energy balance equation of the heating wire according to the principle of heat transfer: ， The above formula is recorded as formula 1, where: Is the quality of the heating wire, is the specific heat capacity of the heating wire, is the function of the heating temperature of the heating wire changing with time, is the real-time input electrical power function, is the function of convective heat transfer rate changing with time, is the function of radiation heat transfer rate changing with time; The convective heat transfer formula is: , The above formula is recorded as formula 2, where: : Convective heat transfer rate as a function of time; : Convective heat transfer coefficient as a function of time, is the surface temperature of the heating wire, The value of The value of is the air temperature in the heating chamber; According to the thermal emissivity formula, , The above is recorded as formula 3, where: is the function of the radiation heat transfer rate changing with time, is the surface emissivity of the heating wire, is the Stefan–Boltzmann constant, is the function of the heating temperature of the heating wire changing with time, is the absolute temperature of the surrounding environment, is the surface area of ​​the heating wire; Establish the energy balance equation for the oven air fluid: ， The above formula is recorded as formula 4, where is the quality of air inside the oven; is the specific heat capacity of air, is a function of heat loss changing with time; according to the above formulas 1, 2, 3, and 4, we can get a system model of the oven including heat transfer and fluid mechanics: ， Analyze different intermediate target temperatures based on the oven system model The impact on the oven temperature response, find multiple sets of intermediate target temperatures that make the total rise time short and the overshoot small , and set the intermediate target temperature Optimize to get the best combination of intermediate target temperature , the equation at this time also serves as the solution condition for solving the performance index function.

3. The multi-stage temperature precision control method based on a neural network algorithm according to claim 2 is characterized in that: The intermediate target temperature The optimization method is to establish a function that defines performance indicators : ， In the above formula is the control end time, is a trade-off factor used to balance temperature error and power change rate, is the time-varying function of input power, It indicates the temperature inside the oven. It represents the final target temperature; then for multiple groups of intermediate target temperatures Combined simulation, take the corresponding temperature response and control performance, the temperature response and control performance measurement indicators include: rise time , Overshoot , Adjust time ; Select the intermediate target temperature that minimizes both the temperature response and the control performance measure combination.

4. The multi-stage temperature precision control method based on a neural network algorithm according to claim 1 is characterized in that: When training the model, a series of optimal PID parameter values ​​are obtained through the data training set using the gradient descent method. The PID principle formula is as follows: ， Perform parameter tuning and take the critical proportional gain , is the minimum value for the system to produce sustained oscillation , critical period is the corresponding oscillation period, then the following PID parameters are obtained: ， The calculated value here is used as the starting value, and the adjustment process uses the gradient descent method to find the optimal value for training. , , Values, thereby obtaining parameter values ​​for data training; Among them is is the proportional adjustment coefficient in PID control, is the integral adjustment coefficient in PID control, It is the differential adjustment coefficient in PID control. Using the data training set, a parameter output model of the heating wire based on the neural network algorithm is constructed; The nth output value is: , , , through temperature The curve is judged and selected, and the rise time is selected , Overshoot , Adjust time The smallest curve, and then select the best , , The values ​​are taken as elements in the training set.

5. The multi-stage temperature precision control method based on a neural network algorithm according to claim 1 is characterized in that: The method of performing fuzzy self-tuning PID control on the heating wire in step 4 is to input power to the heating wire The control error is: ， Using fuzzy logic, the input control error changes over time function and control error change rate Right now , adjust PID parameters, and the indicators to be measured include rise time :Temperature from 10% to 90% Time required to set value; overshoot : Maximum percentage exceeding the set value; adjustment time : The time required for the system to enter and remain within the range of ±2% of the set value; the smaller Tr, Mp, and Ts are, the better.

6. A multi-stage temperature precision control method based on a neural network algorithm according to any one of claims 1 to 5, characterized in that: The method for performing neural network algorithm training on the previous PID parameters in step 4 is to record the input, output and optimal PID parameters of the oven system model under different working conditions as a data training set; use the data training set to perform neural network algorithm training; wherein the recorded , , For input, (t) is the automatic input read by the sensor, V is the wind speed value input by the user, expressed in K p , K i , K d As output, the optimal PID parameters are used as labels.

Images