PID control method for coupling heating of multi-cavity diffusion furnace based on BP neural network

Abstract

The invention relates to a PID (Proportion Integration Differentiation) control method for coupling heating of a multi-cavity diffusion furnace based on a BP (Back Propagation) neural network, which comprises the following steps of: acquiring temperature measurement data of each cavity in the diffusion furnace, and calculating an error between the temperature of each cavity and a preset target temperature; taking the current moment temperature error, the historical moment temperature error, the current moment temperature value, the historical moment temperature error and the control quantity of the previous moment as the input of the trained BP neural network, and predicting to obtain PID control parameters; based on the PID control parameters, through incremental PID control, heating unit power output values of all cavities of the diffusion furnace are calculated, the power of the corresponding heating units is adjusted, and hyper-parameters of the BP neural network are determined by constructing an ARX model of the diffusion furnace. The method can reduce the training cost of equipment operators, improves the stability of the diffusion furnace under different working conditions, and has a wide application prospect in the field of semiconductor manufacturing.

Description

Technical Field

[0001] The present invention relates to the field of multi-cavity heating equipment control, and in particular to a PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network. Background Art

[0002] Diffusion furnaces are core equipment in semiconductor manufacturing processes, and their temperature control accuracy directly impacts the quality of semiconductor devices. Traditional PID (Proportional-Integral-Derivative) control algorithms are widely used in industrial control due to their simple structure and robustness. However, diffusion furnace systems typically exhibit large hysteresis, nonlinearity, and time-varying characteristics, making it difficult to achieve high-precision temperature control with traditional PID control. Key drawbacks include:

[0003] (1) Fixed parameter limitation: Once the PID parameters are set, they cannot be dynamically adjusted during operation and are difficult to adapt to complex temperature changes;

[0004] (2) Large overshoot and steady-state errors: Especially during the system response process, traditional PID control is prone to overshoot, and the steady-state temperature is difficult to strictly maintain at the set value;

[0005] (3) Poor adaptability: When system characteristics change (such as ambient temperature fluctuations or equipment aging), the PID controller cannot automatically adjust parameters, resulting in a decrease in control accuracy.

[0006] Chinese patent application publication number CN112947094A discloses a method for self-adjusting PID parameters for temperature control in a cement rotary kiln. This method uses reinforcement learning and an improved BP neural network to adjust the kiln's PID parameters in real time, resolving the problem of manual adjustment, which prevents rapid and accurate temperature control. This method achieves rapid and precise temperature control in the kiln, improving production efficiency and quality. However, this method does not account for the impact of the interaction between multiple cavities in a diffusion furnace on the furnace terminal temperature, leaving room for improvement in the system's response speed and steady-state control capabilities. Summary of the Invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network, so as to solve or partially solve the problems of mutual effects between the diffusion furnace cavities, the impact on the terminal temperature of the furnace body, and the unsatisfactory system response speed and steady-state control ability.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] One aspect of the present invention provides a PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network, comprising the following steps:

[0010] Obtain temperature measurement data of each cavity in the diffusion furnace and calculate the error between each cavity temperature and the preset target temperature;

[0011] The current temperature error, the historical temperature error, the current temperature value, the historical temperature value and the control quantity at the previous moment are used as the input of the trained BP neural network to predict the PID control parameters;

[0012] Based on the PID control parameters, the power output value of the heating unit of each cavity of the diffusion furnace is calculated through incremental PID control, and the power of the corresponding heating unit is adjusted.

[0013] The hyperparameters of the BP neural network are determined by constructing an ARX model of a diffusion furnace.

[0014] As a preferred technical solution, the heating process of the multi-cavity diffusion furnace includes:

[0015] During the heating phase, the temperature of the cavity with the slowest heating rate is used as the target temperature for all cavities, and PID adjustment is performed;

[0016] In the steady-state stage, the PID control parameters are adjusted independently based on the preset target temperature of each cavity to maintain coordinated temperature control.

[0017] As a preferred technical solution, the construction process of the ARX model includes the following steps:

[0018] Collect the real-time power of the diffusion furnace heating unit and the real-time temperature of each cavity;

[0019] Analyze the temperature coupling relationship of the diffusion furnace, build the ARX model, and realize the modeling of the interaction between various temperature parameters in the furnace;

[0020] The ARX model parameters are optimized based on the physical characteristics of the diffusion furnace and the actual operating conditions.

[0021] As a preferred technical solution, the incremental PID control is implemented using the following formula:

[0022] u(k)=u(k-1)+K p [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)

[0023] +e(k-2)]

[0024] Among them, u(k) is the power control quantity of the heating unit at time k, that is, the power of the heating wire; e(k) is the difference between the actual temperature of the cavity at time k and the set temperature, K p ,K i ,Kd They are proportional, integral and differential coefficients respectively, which serve as PID control parameters.

[0025] As a preferred technical solution, the heating unit is a heating wire.

[0026] As a preferred technical solution, the loss function of the BP neural network is optimized through mean square error.

[0027] As a preferred technical solution, the activation function of the BP neural network is Sigmoid or ReLU.

[0028] As a preferred technical solution, the BP neural network includes an input layer, a hidden layer and an output layer connected in sequence, wherein the input layer receives temperature error information and historical control quantity information, the hidden layer learns system characteristics through nonlinear mapping, and the output layer is used to generate PID control parameters.

[0029] Another aspect of the present invention provides an electronic device comprising: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the aforementioned PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network.

[0030] Another aspect of the present invention provides a computer-readable storage medium comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the aforementioned PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network.

[0031] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0032] (1) Improving the system response speed and steady-state control capability of the diffusion furnace: This invention establishes an ARX model of the diffusion furnace to perform system identification modeling and effectively characterize the temperature coupling relationship between the various cavities. Combined with the adaptive capability of the BP network, it achieves dynamic optimization and adjustment of the PID parameters, thereby quickly responding to temperature changes during the heating process and maintaining high-precision temperature control in the steady-state phase.

[0033] (2) Implementing staged control in the heating phase and the steady-state phase to improve the pertinence of the control strategy: The present invention divides the temperature control process into the heating phase and the steady-state phase. In the heating phase, the lowest speed cavity is used to perform unified heating scheduling to ensure the coordination of temperature changes in each cavity and reduce local overheating. In the steady-state phase, independent adjustments are made based on each target temperature, so that the system can achieve lower energy consumption and higher temperature control accuracy while ensuring temperature consistency. This strategy fully takes into account the balance between system response and energy consumption management.

[0034] (3) Reduce the workload of PID parameter tuning and reduce debugging and maintenance costs: Compared with the traditional PID system that requires a lot of experiments and manual tuning, the present invention can continuously and adaptively optimize the control parameters after the initial setting through the BP neural network self-learning function, significantly reducing the trial and error cost and maintenance difficulty during the engineering deployment process.

[0035] (4) Enhance the robustness and adaptability of the system to operating disturbances: In actual industrial scenarios such as fluctuations in temperature sensor data, interference from external heat sources, or changes in cavity heat capacity, the present invention can use neural networks to perform nonlinear modeling of system behavior and update parameters in real time, thereby maintaining the stability and reliability of the temperature control system and effectively suppressing the impact of external disturbances on the system.

[0036] (5) The multi-cavity collaborative control capability is strong and can adapt to the diffusion furnace with complex structure: This method can be extended to the diffusion furnace equipment with multiple temperature control zones. It is not only suitable for symmetrical structures, but also for asymmetric multi-cavity systems with complex thermal coupling and large differences in thermal inertia. It has good versatility and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of a PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network in an embodiment;

[0038] Figure 2 Schematic diagram of the BP neural network structure in the embodiment;

[0039] Figure 3 Schematic diagram of the diffusion furnace system identification result based on the ARX model in the embodiment;

[0040] Figure 4 The comparison results of the traditional PID, fuzzy PID and BPPID controllers of the present invention are shown in the examples;

[0041] Figure 5 The temperature control effect under different power upper and lower limit conditions in the embodiment;

[0042] Figure 6 Schematic diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts should fall within the scope of protection of the present invention.

[0044] Example 1

[0045] To address the aforementioned issues with the existing technology, this embodiment provides a PID control method for coupled heating in a multi-cavity diffusion furnace based on a BP neural network. This method aims to improve the temperature control accuracy of the diffusion furnace, reduce overshoot and steady-state error, and achieve coordinated multi-cavity temperature control. This method adaptively adjusts PID parameters through a BP neural network, enabling the controller to dynamically adjust the heating wire power based on temperature error, improving the system's response speed and steady-state control capabilities.

[0046] See also Figure 1 , this method mainly includes the following steps:

[0047] (1) Collect temperature sensor data from multiple cavities in the diffusion furnace and calculate the error between the temperature of each cavity and the target set temperature.

[0048] (2) The BP neural network is used to adaptively adjust the PID control parameters, where the input of the BP neural network includes: temperature error at the current moment, temperature error at the previous moment, temperature errors at the previous two moments, temperature value at the current moment, temperature value at the previous moment, and control value at the previous moment.

[0049] Among them, see Figure 2 The BP neural network adopts a three-layer structure, consisting of an input layer, a hidden layer, and an output layer. The input layer receives temperature error information and historical control data; the hidden layer learns system characteristics through nonlinear mapping; and the output layer generates optimal PID control parameters. The neural network's activation function uses Sigmoid or ReLU to enhance nonlinear fitting capabilities, and the network loss function is optimized using mean squared error (MSE).

[0050] The BP neural network hyperparameters are determined by constructing a simulation model based on autoregressive (ARX) modeling of a multi-cavity coupled diffusion furnace. The model specifically includes the following steps:

[0051] High-precision sensors are used to collect real-time data on the temperature of the diffusion furnace heating wire and each cavity, as well as the power input and surface temperature data of the corresponding heating wire, to form a complete time series input and output data set.

[0052] Based on the collected historical data, we conduct an in-depth analysis of the heat conduction between multiple cavities and their mutual influence mechanisms. Using the ARX modeling method, we build a system model of a multiple-input single-output (MISO) structure:

[0053]

[0054] Where T i (t) is the current temperature of the i-th cavity; P j (t) is the heating power of the jth cavity; a ij (z-1 ) is the regression coefficient polynomial representing the temperature effect, reflecting the coupling between temperatures; b ij (z -1 ) is the polynomial of the control input response to temperature; e i (t) is the modeling error; n is the total number of cavities. If i≠j, then a ij ,b ij ≠0 indicates the presence of inter-cavity coupling.

[0055] During the modeling process, the cross-influence terms between adjacent cavities are explicitly retained, thereby accurately reflecting the multi-cavity temperature coupling characteristics inside the diffusion furnace.

[0056] Combined with the physical partition characteristics of the diffusion furnace structure, the distribution of the heating wires and the actual process conditions, the appropriate model order is selected, and the least squares method is used to calculate the ARX model parameters. Perform identification and construct the regression matrix Φ. The target vector Y is the temperature measurement value. Solve:

[0057]

[0058] Subsequently, the model is iteratively optimized based on the control accuracy requirements and experimental data feedback to improve the simulation model's ability to predict the dynamic behavior of the system.

[0059] See also Figure 3 Schematic diagram of the accuracy of the ARX model of the diffusion furnace. Figure 3 The fitting results of the ARX model in a multi-cavity diffusion furnace system based on four sets of experimental data (data1 to data4) are shown. Figure 3 (a) shows the training data used for modeling. In each sub-graph, the blue solid line represents the actual temperature measurement, the black dashed line represents the model fitting output, and the fitting accuracy (goodness of fit percentage) is marked in the upper right corner. The figure clearly shows the performance of the model under different dynamic and steady-state conditions:

[0060] Figure 3 (b) Data 1 is a standard heating process. The model has a goodness of fit of more than 97% in all cavity temperature control fittings, with the highest reaching 99.41%. The fitting curves are almost consistent with the actual measurements. Figure 3 (c) Data 2 shows the system under steady-state perturbation vibration. This data shows the small fluctuation response of the system in steady state. Although the goodness of fit is slightly lower (e.g., 57-77%), the curve trend is still good, indicating that it still has certain predictive ability under noise and perturbation. Figure 3 (d) Data 4 is a long-running process with a long model test time span (approximately 14,000 seconds), simulating multi-cycle control and disturbance changes. The goodness of fit of the temperature of each cavity remains above 95%, reflecting the model's stable modeling capability under long-term control.

[0061] Figure 3 It is verified that the ARX model constructed in the present invention can accurately fit the actual temperature response under various working conditions (heating, steady state, disturbance, and long period), and has the following capabilities: the ability to model the temperature of multiple cavities and the ability to respond to the effects of power input and coupling with adjacent cavities; and it provides a high-quality fitting data basis for the BP neural network intelligent controller.

[0062] In this embodiment, the temperature coupling relationship of the diffusion furnace refers to the mutual influence of the temperatures between the various cavities, that is, the power adjustment of the heating wire not only affects the target cavity temperature, but also affects the temperature of the adjacent cavity; the temperature of different heating wire distribution ends of the diffusion furnace is affected by the internal temperature of the cavity and the adjacent heating wires; by modeling the coupling relationship between the diffusion furnace cavity temperature and the heating wire temperature and power, the control strategy is optimized to achieve precise control of temperature regulation.

[0063] (3) The optimal PID parameters are calculated through forward propagation, and the network weights are adjusted using the backpropagation gradient descent algorithm to minimize the control error.

[0064] (4) An incremental PID control algorithm is used to calculate the heating wire power output value based on the optimized parameters output by the BP neural network.

[0065] The PID controller uses the incremental PID calculation formula:

[0066] u(k)=u(k-1)+K p [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)

[0067] +e(k-2)]

[0068] Among them, K p ,K i ,K d They are proportional, integral and differential coefficients respectively, which are adjusted in real time by BP neural network.

[0069] (5) According to the calculated heating power, the heating wire power of each cavity of the diffusion furnace is adjusted to stabilize the temperature of each cavity at the set value. In the heating stage, the target temperature of the cavity is set differently to achieve coordinated heating between the cavities. Specifically, the reference temperature is set for the cavity with the slowest heating rate, and the other cavities use the cavity with the slowest heating rate as the reference temperature.

[0070] Adopt temperature coupling strategy to keep temperature consistency in different cavities during heating process

[0071] (6) Continuously collect temperature data and dynamically adjust PID parameters according to the BP neural network output to achieve real-time temperature control of the diffusion furnace.

[0072] In order to verify the effectiveness of the solution of this embodiment, a Simulink simulation test environment was built to conduct comparative experiments on traditional PID, fuzzy PID and BPPID. Figure 4 The figure shows the simulation experiment results, which show that the BPPID controller constructed based on the ARX model in this embodiment is superior to the traditional method in terms of response speed, steady-state error, overshoot, etc.

[0073] In summary, this method is based on the fusion algorithm of BP neural network and PID control, combined with Simulink simulation modeling and PCL interface actual equipment control, to achieve the optimization of the heating rate of the diffusion furnace and the precise regulation of the steady-state temperature. This method adaptively adjusts the PID control parameters through the BP neural network, effectively improving the adaptive ability of temperature control and reducing the complexity of manual parameter setting. Compared with traditional PID control, this method is suitable for the coordinated temperature control of multi-cavity diffusion furnaces, which can ensure that the temperatures of different cavities rise synchronously and maintain the steady-state temperature in the cavity fluctuating within ±0.7°C. This method can reduce the training cost of equipment operators, improve the stability of the diffusion furnace under different working conditions, and reduce the experimental testing cost of PID parameter adjustment. It has broad application prospects in the field of semiconductor manufacturing.

[0074] In practical applications, such as Figure 5 As shown, this method can achieve the temperature of the diffusion furnace to rise at a specified rate.

[0075] Example 2

[0076] Based on Example 1, see Figure 6 This embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network as described in Example 1.

[0077] like Figure 6 As mentioned above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 Of course, in addition to software implementation, the present invention does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0078] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0079] Example 3

[0080] This embodiment provides a computer-readable storage medium, including one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network as described in Example 1.

[0081] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0082] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network, characterized in that: The steps include: Obtain temperature measurement data of each cavity in the diffusion furnace and calculate the error between each cavity temperature and the preset target temperature; The current temperature error, the historical temperature error, the current temperature value, the historical temperature value and the control quantity at the previous moment are used as the input of the trained BP neural network to predict the PID control parameters; Based on the PID control parameters, the power output value of the heating unit of each cavity of the diffusion furnace is calculated through incremental PID control, and the power of the corresponding heating unit is adjusted. The hyperparameters of the BP neural network are determined by constructing an ARX model of a diffusion furnace.

2. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The heating process of the multi-cavity diffusion furnace includes: During the heating phase, the temperature of the cavity with the slowest heating rate is used as the target temperature for all cavities, and PID adjustment is performed; In the steady-state stage, the PID control parameters are adjusted independently based on the preset target temperature of each cavity to maintain coordinated temperature control.

3. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The construction process of the ARX model includes the following steps: Collect the real-time power of the diffusion furnace heating unit and the real-time temperature of each cavity; Analyze the temperature coupling relationship of the diffusion furnace, build the ARX model, and realize the modeling of the interaction between various temperature parameters in the furnace; The ARX model parameters are optimized based on the physical characteristics of the diffusion furnace and the actual operating conditions.

4. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The incremental PID control is implemented using the following formula: u(k)=u(k-1)+K p [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)+e(k-2)] Among them, u(k) is the power control quantity of the heating unit at time k, that is, the power of the heating wire; e(k) is the difference between the actual temperature of the cavity at time k and the set temperature, K p ,K i ,K d They are proportional, integral and differential coefficients respectively, which serve as PID control parameters.

5. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The heating unit is a heating wire.

6. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The loss function of the BP neural network is optimized through mean square error.

7. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The activation function of the BP neural network is Sigmoid or ReLU.

8. The PID control method for coupled heating of a multi-cavity diffusion furnace based on BP neural network according to claim 1 is characterized in that: The BP neural network includes an input layer, a hidden layer and an output layer connected in sequence, wherein the input layer receives temperature error information and historical control quantity information, the hidden layer learns system characteristics through nonlinear mapping, and the output layer is used to generate PID control parameters.

9. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory stores one or more programs, wherein the one or more programs include instructions for executing the PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that It includes one or more programs for execution by one or more processors of an electronic device, and the one or more programs include instructions for executing the PID control method for coupled heating of a multi-cavity diffusion furnace based on a BP neural network as described in any one of claims 1-8.

Images