A three-pin waveguide impedance matching method based on gradient descent

Abstract

The application provides a three-pin waveguide impedance matching method based on gradient descent, and relates to the technical field of microwave, which comprises the following steps: obtaining the reflection coefficient of the current hardware system, and presetting a target reflection coefficient threshold; based on the reflection coefficient and the gradient descent algorithm, iteratively calculating the insertion depth of the three pins to obtain the actual displacement amount of the pins, and adjusting the insertion depth of the three pins based on the actual displacement amount and the initial position; based on the current position of the pins, obtaining the change amplitude of the reflection coefficient, obtaining the optimization state according to the change amplitude and executing the corresponding escape strategy to obtain the updated reflection coefficient; and judging whether the target reflection coefficient threshold is reached based on the updated reflection coefficient. The application can ensure that the amplitude of the reflection coefficient is within the preset threshold, so that the utilization rate of microwave energy by the load is maintained above 99% during microwave heating.

Description

Technical Field

[0001] This invention relates to the field of microwave technology, and more specifically, to a three-pin waveguide impedance matching method based on gradient descent. Background Technology

[0002] Microwave heating technology, with its advantages of cleanliness, high efficiency, and strong controllability, has been widely used in industrial and scientific research fields such as materials synthesis, food processing, and medical sterilization. This technology utilizes microwave electromagnetic energy to directly act on dielectric materials, achieving bulk heating through molecular polarization and relaxation effects. However, in practical applications, the impedance mismatch between the microwave transmission path and the load to be heated severely restricts the improvement of system energy efficiency and performance. Therefore, three-pin impedance matching devices have been widely introduced to solve this problem.

[0003] Traditional three-pin impedance matching schemes employ matching algorithms based on equivalent circuit models. This involves establishing a mapping relationship between pin insertion depth and equivalent circuit parameters, combined with an enumeration search strategy to achieve impedance matching. The scheme first establishes an equivalent circuit model of the three-pin waveguide, derives port scattering parameters and reverse equivalent impedance, constructs a pin configuration database, and then determines the optimal pin position through a lookup table. However, because this scheme relies on the accuracy of the equivalent circuit model, and the computational cost of the enumeration search increases exponentially with search accuracy, it suffers from low computational efficiency, sensitivity to model errors, and limited real-time tracking capability for dynamic loads when handling high-precision matching requirements. Summary of the Invention

[0004] This invention aims to address the problems of low computational efficiency, sensitivity to model errors, limited real-time dynamic load tracking capability, and the learning rate selection dilemma and susceptibility to local optima in traditional three-pin impedance matching schemes, as well as the inherent limitations of gradient descent algorithms. It provides a three-pin waveguide impedance matching method with high matching efficiency, strong stability, and adaptability to dynamic load changes. To achieve the above objectives, the technical solution adopted by this invention is as follows: Step S100: Obtain the reflection coefficient of the current hardware system, preset the target reflection coefficient threshold, and place the three pins in the initial position. The preset target reflection coefficient threshold is -20dB. The reflection coefficient should be less than or equal to the target reflection coefficient threshold.

[0005] The hardware system of the present invention includes a main control module, a drive module, a reflection coefficient detection module, and a three-pin module. The main control module is connected to the driver (i.e., the stepper motor) and electrically connected to the reflection coefficient detection module. The drive module is connected one-to-one with the stepper motor in the three-pin module. The three-pin module is set in a three-pin rectangular waveguide. The probe of the reflection coefficient detection module is deployed in the waveguide to couple the microwave power signal, forming a complete closed-loop control hardware architecture.

[0006] Step S200: Iteratively calculate the insertion depth of the three pins based on the reflection coefficient and gradient descent algorithm to obtain the actual displacement of the pins, and adjust the insertion depth of the three pins based on the actual displacement and the initial position. Step S300: Obtain the change range of the reflection coefficient based on the current position of the pin, obtain the optimization state according to the change range and execute the corresponding escape strategy to obtain the updated reflection coefficient; Step S400: Determine whether the target reflection coefficient threshold has been reached based on the updated reflection coefficient. If it has been reached, impedance matching is completed. Otherwise, repeat steps S200 to S300 based on the updated reflection coefficient.

[0007] As a preferred embodiment, step S200 includes: Step S210: Obtain the change in reflection coefficient of each pin when it moves in the positive and negative directions respectively, and obtain the gradient direction of the pin based on the change. Step S220: Dynamically select the proportional-integral parameter set according to the current reflection coefficient, and obtain the effective learning rate of the current iteration based on the selected proportional-integral parameter set; Step S230: Obtain the displacement of the pin based on the gradient direction and the effective learning rate, and introduce an impulse term to correct the displacement to obtain the actual displacement of the pin.

[0008] As a preferred embodiment, obtaining the gradient direction of the pin based on the change includes: If both positive and negative movement reduce the reflection coefficient, then the direction that reduces the reflection coefficient more significantly is chosen as the gradient direction. If both positive and negative movement increase the reflection coefficient, then the gradient direction is 0, and the current position of the pin remains unchanged. If positive and negative movements have opposite effects on the reflection coefficient, then the direction that reduces the reflection coefficient should be chosen as the gradient direction.

[0009] As a preferred embodiment, the step of dynamically selecting the proportional-integral parameter set based on the current reflection coefficient includes: When the reflection coefficient is greater than the first threshold, the first set of proportional-integral parameters is used; When the first threshold ≥ the reflection coefficient > the second threshold, the second set of proportional-integral parameters is used, wherein the proportional parameter in the second set of proportional-integral parameters is less than the proportional parameter in the first set of proportional-integral parameters, and the integral parameter in the second set of proportional-integral parameters is less than the integral parameter in the first set of proportional-integral parameters. When the reflection coefficient is less than or equal to the second threshold, a third set of proportional-integral parameters is used, wherein the proportional parameter in the third set of proportional-integral parameters is less than the proportional parameter in the second set of proportional-integral parameters, and the integral parameter in the third set of proportional-integral parameters is less than the integral parameter in the second set of proportional-integral parameters; the first threshold is 0.3, and the second threshold is 0.1.

[0010] As a preferred embodiment, the correction of the displacement by introducing an impulse term includes: The current impulse value is obtained by weighted summing the displacement and the impulse value of the previous iteration period. The actual displacement of the pin is obtained based on the current impulse value.

[0011] As a preferred embodiment, step S300 specifically includes: Step S310: Obtain the change range of the reflection coefficient based on the current position of the pin; Step S320: Obtain the current optimization state based on the change magnitude; Step S330: If the current optimization state is trapped in a local optimum, then execute the escape strategy of impulse enhancement and position perturbation to escape the local optimum and obtain the updated reflection coefficient; the condition for being trapped in a local optimum is that the decrease in the magnitude of the reflection coefficient in consecutive iterations is less than a preset improvement threshold. If the current optimization state is a normal improvement state, then the updated reflection coefficient is obtained directly.

[0012] As a preferred embodiment, the escape strategy of performing impulse enhancement and positional perturbation to escape the local optimum and obtain the updated reflection coefficient includes: S3201, Increase the weighting coefficient for the impulse term; S3202. If the decrease in reflection coefficient after increasing the impulse is less than the preset improvement threshold, the pin position is restored to the historical best position, and a random displacement is applied based on that position; the historical best position is the insertion depth of the pin when the reflection coefficient is the smallest in multiple iterations. S3203, Return to step S200.

[0013] The beneficial effects of this invention are as follows: This invention enhances noise immunity and improves the reliability of gradient estimation by employing a bidirectional gradient measurement method. It resolves the contradiction between fast response and stable convergence in traditional fixed proportional-integral parameters by setting a three-order proportional-integral parameter set, thus avoiding oscillation overshoot in the small reflection region. Furthermore, it effectively solves the local optimum problem by using the impulse method and a positional perturbation escape strategy, forming a closed-loop tuning system with autonomous decision-making capabilities. This not only effectively addresses the inherent problems of traditional methods in terms of convergence speed, local optima, and dynamic adaptability, but also ensures that the system can stably converge to the global optimum. Secondly, this invention does not require the establishment of an equivalent circuit model or a pre-built matching database. It directly controls the optimization process through hardware and combines an adaptive learning rate gradient descent algorithm to automatically increase the step size in steep regions of the objective function to accelerate convergence and decrease the step size in flat regions for fine-tuning, which significantly improves the matching speed.

[0014] Furthermore, by measuring the reflection coefficient in real time and dynamically adjusting the pin position, this invention effectively overcomes the model dependence and response delay problems of traditional methods, making it suitable for impedance matching requirements under different load conditions.

[0015] Finally, by ensuring that the amplitude of the reflection coefficient is within a preset threshold, the present invention can guarantee that the load's utilization rate of microwave energy remains above 99% during microwave heating, thus significantly improving the energy efficiency and performance of the microwave system.

[0016] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the three-pin waveguide impedance matching method based on gradient descent in an embodiment of the present invention.

[0019] Figure 2 This is a framework diagram of the hardware system for three-pin waveguide impedance matching in an embodiment of the present invention.

[0020] Figure 3This is a simulation diagram illustrating the optimization of matching results in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] This embodiment provides a three-pin waveguide impedance matching method based on gradient descent, as shown in Figure 1. The method includes the following steps: Step S100: Obtain the reflection coefficient of the current hardware system, preset the target reflection coefficient threshold, and place the three pins in the initial position; See Figure 2 In this invention, the rectangular waveguide is defined as a BJ26 rectangular waveguide with a length of 86.36 mm, a width of 43.18 mm, and an operating frequency of f = 2.45 GHz. The initial positions of the three pins are zeroed, and the reflection coefficient S11 of the current hardware system is obtained through a reflection coefficient detection module. The preset target reflection coefficient threshold is -20 dB, and the amplitude of the reflection coefficient S11 is ≤ -20 dB. The reflection coefficient detection module includes detector one, detector two, detector three and probes. The amplitude value of the current reflection coefficient S11 is obtained by reading the power value coupled by the three probes through detector one, detector two and detector three respectively.

[0024] Step S200: Based on the reflection coefficient S11 and the gradient descent algorithm, iteratively calculate the insertion depth of the three pins to obtain the actual displacement of the pins, and adjust the insertion depth of the three pins based on the actual displacement and the initial position. Step S200 includes: Step S210: Obtain the change in reflection coefficient S11 of each pin when it moves in the positive and negative directions respectively, and obtain the gradient direction of the pin based on the change. The specific process of step S210 is as follows: Step S2101: Obtain the movement step size of each pin during each debugging session, and perform gradient calculations for both positive and negative movements for each pin. The specific formulas are as follows: ; ; in, It is a positive gradient. The gradient is negative. The objective function is related to the reflection coefficient. The current position of the pin. This is the step size for the pin to move.

[0025] Step S2102: Determine the gradient direction of the pin based on the calculation results of the positive and negative gradients: If both positive and negative shifts reduce the reflection coefficient S11, meaning both the positive and negative gradients are negative, it indicates that both directions can improve the matching effect. Therefore, the direction that more significantly reduces the reflection coefficient S11 is chosen as the gradient direction. ; If both positive and negative movements increase the reflection coefficient S11, then the gradient direction is 0, and the current position of the pin remains unchanged; that is, if both the positive and negative gradients are positive, it means that both directions will degrade the matching effect, so the gradient direction is 0, and the current position of the pin remains unchanged. ; If positive and negative movements have opposite effects on the reflection coefficient S11, i.e., one positive and one negative, the direction that reduces the reflection coefficient S11 is chosen as the gradient direction.

[0026] This bidirectional gradient measurement method can comprehensively evaluate the direction of improvement, avoid the one-sidedness of unilateral measurement, effectively detect local extreme points, and adaptively adjust the test step size to adapt to different reflection levels, which significantly enhances the noise resistance and improves the reliability of gradient estimation.

[0027] Step S220: Dynamically select the proportional-integral parameter set according to the current reflection coefficient S11, and obtain the effective learning rate of the current iteration based on the selected proportional-integral parameter set; The proportional-integral parameter group is dynamically selected based on the current amplitude of the reflection coefficient S11. The specific parameter settings are as follows: ; ; The first threshold is 0.3, the second threshold is 0.1, and these three thresholds can be adaptively adjusted according to the operating frequency or waveguide size in the actual application scenario.

[0028] The effective learning rate after adding the corresponding proportional-integral parameter to the learning rate used in the original gradient descent algorithm. The expression is: ; ; in, The base learning rate can be preset according to the actual situation; To control the output, For discrete-time indexing, The sampling time interval, Error signal = Set value - Measured value; :Proportion; :integral; Therefore, the final optimized output expression after adding the proportional-integral parameter is: ,in, This represents the gradient value.

[0029] Step S230: Obtain the displacement of the pin based on the gradient direction and the effective learning rate, and introduce an impulse term to correct the displacement to obtain the actual displacement of the pin.

[0030] In this step, it should be noted that the gradient direction, i.e., the gradient vector, is first determined. and effective learning rate The initial displacement is calculated using the following expression: ; Secondly, regarding the displacement amount The current impulse value is obtained by weighted summation with the impulse value of the previous iteration cycle. To solve the problem that the gradient descent algorithm is prone to getting trapped in local optima, an impulse term is introduced. By introducing the weighted cumulative effect of historical update, even if the gradient is very small, the historically accumulated impulse can still provide enough momentum to keep the parameter update moving in the original direction and achieve the goal of escaping the local optimum. The specific definition of the impulse term is: ; in, This is the current impulse value, initially set to 0. This is the impulse value from the previous impulse. It is the displacement; In this embodiment, the momentum influencing factor is... =0.7; when When the value is 0.7, the effective memory length of the impulse is L≈3.33. That is, before the third iteration, the influence of the impulse will be halved, ensuring that it can effectively escape the local optimum and will not cause system instability due to excessive memory length.

[0031] Based on the current impulse value Obtain the actual displacement of the pin Ultimately, the actual displacement of the pin for: ; Based on the calculated actual displacement The initial position of the pins is determined, and the insertion depth of the three pins is adjusted to complete one iteration of position adjustment.

[0032] In step S200, a stepped proportional-integral calculation method is used. The tuning process is divided into three reflection regions—large, medium, and small—based on the amplitude of the reflection coefficient S11, and different methods are applied to each region. , By adjusting parameters and output amplitude, adaptive control is achieved, effectively resolving the contradiction between fast response and stable convergence in traditional fixed PID (proportional-integral-derivative) parameters. It can quickly reduce large reflections and avoid oscillation overshoot in the small reflection zone. At the same time, the dynamic output limiting mechanism prevents excessive movement, significantly improving the efficiency, stability and robustness of the dynamic adjustment process.

[0033] Step S300: Based on the current position of the pin, obtain the change range of the reflection coefficient S11, obtain the optimization state according to the change range and execute the corresponding escape strategy to obtain the updated reflection coefficient S11; Step S300 includes: Step S310: Obtain the change range of the reflection coefficient S11 based on the current position of the pin; Step S320: Obtain the current optimization state based on the change magnitude; Step S330: If the current optimization state is trapped in a local optimum, that is, the decrease in the reflection coefficient S11 in N consecutive iterations is less than the preset improvement threshold. (In this embodiment, N=6, =0.01 (can be adjusted according to the accuracy requirements of the actual hardware system), indicating that the current gradient is approaching 0, the gradient descent algorithm and PID control are stalled. At this time, the escape strategy of impulse enhancement and position disturbance is executed to escape the local optimum and obtain the updated reflection coefficient S11. Step S340: If the current optimization state is a normal improvement state, then directly obtain the updated reflection coefficient S11.

[0034] In step S330, the specific process of executing the escape strategy of impulse enhancement and position perturbation to escape the local optimum and obtaining the updated reflection coefficient S11 includes: Step S3301: Add impulse item The weighting coefficients will affect the momentum factor. Adjust it to 1.5 times the original value, activate the 1.5 times impulse enhancement, and superimpose the amplified historical movement trend onto the PID control output to attempt to break through the local optimum along the original improvement direction; Step S3302: Based on the adjusted momentum influence factor Recalculate the actual displacement of the pin. The pin position is adjusted, the adjusted reflection coefficient S11 is obtained, and its decrease is calculated. If the decrease in reflection coefficient S11 after increasing the impulse is still less than the preset improvement threshold, then... If the cumulative number of iterations without improvement reaches M (M=12 in this embodiment), then the pin position is restored to the historical best position, which is the insertion depth of the pin when the reflection coefficient S11 is the minimum in multiple iterations; and a random displacement is applied independently to each pin based on this position.

[0035] Step S3303: Initialize the hardware system and return to step S200.

[0036] The escape strategy of this invention adopts a three-level progressive response strategy based on a strict counting threshold. It utilizes impulse memory to avoid the blindness of completely random escape. Through the coordinated work of impulse decay and random exploration mechanism, it effectively obtains global search capability while maintaining local rapid convergence.

[0037] Step S400: Determine whether the target reflection coefficient threshold has been reached based on the updated reflection coefficient S11. If it has been reached, impedance matching is completed. Otherwise, repeat steps S200 to S300 based on the updated reflection coefficient.

[0038] See Figure 3 , Figure 3 To optimize the matching results, when the reflection coefficient S11 can be kept within a preset threshold (S11≤-20dB), the corresponding energy utilization efficiency reaches 99% or higher. This invention obtains the gradient direction by integrating bidirectional gradient calculation and combines it with proportional-integral algorithm to optimize the effective learning rate and impulse optimization, achieving high-precision matching and forming a closed-loop tuning system with autonomous decision-making capabilities.

[0039] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0041] 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 variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A three-pin waveguide impedance matching method based on gradient descent, characterized in that, include: Step S100: Obtain the reflection coefficient of the current hardware system, preset the target reflection coefficient threshold, and place the three pins in the initial position; Step S200: Iteratively calculate the insertion depth of the three pins based on the reflection coefficient and gradient descent algorithm to obtain the actual displacement of the pins, and adjust the insertion depth of the three pins based on the actual displacement and the initial position. Step S300: Obtain the change range of the reflection coefficient based on the current position of the pin, obtain the optimization state according to the change range and execute the corresponding escape strategy to obtain the updated reflection coefficient; Step S400: Determine whether the target reflection coefficient threshold has been reached based on the updated reflection coefficient. If it has been reached, impedance matching is completed. Otherwise, repeat steps S200 to S300 based on the updated reflection coefficient.

2. The three-pin waveguide impedance matching method based on gradient descent according to claim 1, characterized in that, Step S200 includes: The change in reflection coefficient of each pin is obtained when it moves in the positive and negative directions, and the gradient direction of the pin is obtained based on the change. The proportional-integral parameter set is dynamically selected based on the current reflection coefficient, and the effective learning rate for the current iteration is obtained based on the selected proportional-integral parameter set. The displacement of the pin is obtained based on the gradient direction and the effective learning rate, and an impulse term is introduced to correct the displacement to obtain the actual displacement of the pin.

3. The three-pin waveguide impedance matching method based on gradient descent according to claim 2, characterized in that, The step of obtaining the gradient direction of the pin based on the change includes: If both positive and negative movement reduce the reflection coefficient, then the direction that reduces the reflection coefficient more significantly is chosen as the gradient direction. If both positive and negative movement increase the reflection coefficient, then the gradient direction is 0, and the current position of the pin remains unchanged. If positive and negative movements have opposite effects on the reflection coefficient, then the direction that reduces the reflection coefficient should be chosen as the gradient direction.

4. The three-pin waveguide impedance matching method based on gradient descent according to claim 3, characterized in that, The step of dynamically selecting the proportional-integral parameter set based on the current reflection coefficient includes: When the reflection coefficient is greater than the first threshold, the first set of proportional-integral parameters is used; When the first threshold ≥ the reflection coefficient > the second threshold, the second set of proportional-integral parameters is used, wherein the proportional parameter in the second set of proportional-integral parameters is less than the proportional parameter in the first set of proportional-integral parameters, and the integral parameter in the second set of proportional-integral parameters is less than the integral parameter in the first set of proportional-integral parameters. When the reflection coefficient is less than or equal to the second threshold, a third set of proportional-integral parameters is used, wherein the proportional parameter in the third set of proportional-integral parameters is less than the proportional parameter in the second set of proportional-integral parameters, and the integral parameter in the third set of proportional-integral parameters is less than the integral parameter in the second set of proportional-integral parameters.

5. The three-pin waveguide impedance matching method based on gradient descent according to claim 4, characterized in that, The first threshold is 0.3, and the second threshold is 0.

1.

6. The three-pin waveguide impedance matching method based on gradient descent according to claim 2, characterized in that, The introduction of an impulse term to correct the displacement includes: The current impulse value is obtained by weighted summing the displacement and the impulse value of the previous iteration period. The actual displacement of the pin is obtained based on the current impulse value.

7. The three-pin waveguide impedance matching method based on gradient descent according to claim 1, characterized in that, Step S300 includes: The change in the reflection coefficient is obtained based on the current position of the pin; The current optimization state is obtained based on the magnitude of the change; If the current optimization state gets stuck in a local optimum, then the escape strategy of impulse enhancement and position perturbation is executed to escape the local optimum and obtain the updated reflection coefficient; If the current optimization state is a normal improvement state, then the updated reflection coefficient is obtained directly.

8. The three-pin waveguide impedance matching method based on gradient descent according to claim 7, characterized in that, The condition for falling into a local optimum is that the decrease in the magnitude of the reflection coefficient is less than a preset improvement threshold in multiple consecutive iterations.

9. The three-pin waveguide impedance matching method based on gradient descent according to claim 8, characterized in that, An escape strategy involving impulse enhancement and positional perturbation is employed to escape the local optimum and obtain updated reflection coefficients, including: Increase the weighting factor for the impulse term; If the decrease in reflection coefficient after increasing the impulse is less than the preset improvement threshold, the pin position is restored to the historical optimal position, and a random displacement is applied based on that position; the historical optimal position is the insertion depth of the pin when the reflection coefficient is the minimum in multiple iterations. Return to step S200.

10. The three-pin waveguide impedance matching method based on gradient descent according to claim 1, characterized in that, The target reflectance threshold is -20dB.

Images