A frequency modulation based adaptive inductance design method

Abstract

This application relates to the field of inductor design technology, and particularly to an adaptive inductor design method based on frequency modulation. It includes the following steps: S1, collecting the basic physical parameters of the inductor and calculating the reference resistivity ρ_ref; S2, sampling and obtaining corresponding resistivity samples within the frequency range; S3, training a neural network model to fit the frequency-resistivity relationship; S4, searching for the optimal frequency f_opt, such that the actual resistivity is ≤ ρ_ref and the inductance value is close to the target value; S5, verifying and correcting the inductor parameters. This invention achieves precise design of high-frequency inductors through dynamic modeling and adaptive optimization of frequency-resistivity, effectively solving the performance degradation problem caused by the skin effect, and can be widely applied to high-frequency circuit systems in fields such as communication and energy.

Description

Technical Field

[0001] This application relates to the field of inductor design technology, and in particular to an adaptive inductor design method based on frequency modulation. Background Technology

[0002] Traditional inductor design relies on static parameter calculations at a fixed frequency, using empirical formulas (such as the Wheeler formula) or finite element simulations to determine inductance values ​​and core parameters. However, at high frequencies (>1MHz), the skin effect causes current to concentrate at the surface, reducing the effective cross-sectional area and increasing resistivity. This leads to inductance values ​​deviating from design targets, increased temperature rise, and decreased efficiency. While existing technologies utilize variable inductors (such as core bias adjustment) or frequency compensation circuits (such as LC tuning networks) to achieve parameter adjustment, they rely on preset models or closed-loop feedback, lacking the ability to adaptively model the frequency-resistivity nonlinear relationship, making it difficult to optimize performance in real time across a wide frequency range. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides an adaptive inductor design method based on frequency modulation, comprising the following steps:

[0004] S1. Collect the basic physical parameters of the inductor and calculate the reference resistivity ρ_ref;

[0005] S2. Sample within the frequency range and obtain the corresponding resistivity sample;

[0006] S3. Train a neural network model to fit the frequency-resistivity relationship;

[0007] S4. Search for the optimal frequency f_opt, so that the actual resistivity is ≤ ρ_ref and the inductance value is close to the target value;

[0008] S5. Verify and correct the inductor parameters.

[0009] Preferably, in step S1, collecting the basic physical parameters of the inductor and calculating the reference resistivity ρ_ref includes the following steps:

[0010] S101. Collect the physical parameters of the target inductor: conductor material conductivity σ, permeability μ0, and core relative permeability μ0. r The number of coil turns N, the wire diameter d, and the cross-sectional area of ​​the magnetic core A;

[0011] S102. Set the initial operating frequency f0 and the skin effect threshold K_skin (characterizing the acceptable resistivity deviation rate, taken as 5%-10%).

[0012] S103. Calculate the reference resistivity ρ_ref.

[0013] ,

[0014] Where ρ0 is the DC conductivity (σ -1 ), δ0 is the standard skin depth .

[0015] Preferably, in step S2, sampling and obtaining the corresponding resistivity sample within the frequency range includes the following steps:

[0016] S201, Frequency modulation range: Set the frequency modulation interval [f_min, f_max];

[0017] S202, Sample generation: Sample N frequency points {f1, f2, ..., f_N} with a step size Δf (e.g., 1kHz); for each frequency point f_i, measure its corresponding actual resistivity ρ_i through electromagnetic simulation (e.g., HFSS) or experiment.

[0018] Preferably, in step S201, the frequency modulation interval [f_min, f_max] is [f0﹣50%f0, ​​f0+50%f0].

[0019] Preferably, in step S3, training the neural network model to fit the frequency-resistivity relationship includes the following steps:

[0020] S301. Construct a BP neural network model: The input layer is the frequency f_i, the hidden layer contains 2 layers (each layer has 32 and 16 neurons respectively), and the output layer is the resistivity prediction value ρ_pred;

[0021] S302. Training data: Using {f_i, ρ_i} as samples, the mean squared error (MSE) is used as the loss function, and the optimizer is Adam (learning rate 0.001).

[0022] S303. Model Validation: The model converges when the prediction error (|ρ_pred-ρ_i| / ρ_i) < 3%.

[0023] Preferably, in step S4, searching for the optimal frequency includes the following steps:

[0024] S401. Based on the trained model, traverse the frequency interval [f_min, f_max] and calculate ρ_pred for each f_i.

[0025] S401. Filter the frequency points that satisfy ρ_pred ≤ ρ_ref, and select the ones that make the inductance value L ( The frequency f_opt that is closest to the target value L_target (where l is the magnetic circuit length) is taken as the optimal operating frequency.

[0026] Preferably, in step S5, the inductor parameter correction and verification includes the following steps:

[0027] S501, Recalculate the coil impedance Z_opt based on f_opt ( (where R_ac is the AC resistance).

[0028] S502. Test the Q value (quality factor) and temperature rise of the inductor under f_opt through experiments. If the Q value is ≥30 and the temperature rise is ≤40℃, the design is complete; otherwise, return to step S3, adjust the hidden layer parameters of the neural network and retrain.

[0029] As can be seen from the above, the following beneficial effects can be obtained by applying the method provided in this application:

[0030] 1. Adaptive optimization: No need to preset frequency model, the optimal operating frequency is dynamically matched through data-driven methods, effectively suppressing resistivity deviation caused by skin effect;

[0031] 2. Wide frequency band applicability: Real-time frequency modulation can be achieved in the range of 1MHz to 100MHz, which is suitable for high-frequency communication, radar and other scenarios;

[0032] 3. Improved efficiency: Compared with traditional fixed frequency design, this method can increase the Q value of the inductor by 15%~20% and reduce the temperature rise by 8~12℃. Detailed Implementation

[0033] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0034] Example

[0035] To address the aforementioned technical problems, this embodiment provides an adaptive inductor design method based on frequency modulation.

[0036] Includes the following steps:

[0037] S1. Collect the basic physical parameters of the inductor and calculate the reference resistivity ρ_ref;

[0038] In step S1, the acquisition of the basic physical parameters of the inductor and the calculation of the reference resistivity ρ_ref include the following steps:

[0039] S101. Collect the physical parameters of the target inductor: conductor material conductivity σ, permeability μ0, and core relative permeability μ0. r The number of coil turns N, the wire diameter d, and the cross-sectional area of ​​the magnetic core A;

[0040] S102. Set the initial operating frequency f0 and the skin effect threshold K_skin (characterizing the acceptable resistivity deviation rate, taken as 5%-10%).

[0041] S103. Calculate the reference resistivity ρ_ref.

[0042] ,

[0043] Where ρ0 is the DC conductivity (σ -1 ), δ0 is the standard skin depth .

[0044] S2. Sample within the frequency range and obtain the corresponding resistivity sample;

[0045] In step S2, sampling and obtaining the corresponding resistivity sample within the frequency range includes the following steps:

[0046] S201, Frequency modulation range: Set the frequency modulation interval [f_min, f_max]; where the frequency modulation interval [f_min, f_max] is [f0﹣50%f0, ​​f0+50%f0].

[0047] S202, Sample generation: Sample N frequency points {f1, f2, ..., f_N} with a step size Δf (e.g., 1kHz); for each frequency point f_i, measure its corresponding actual resistivity ρ_i through electromagnetic simulation (e.g., HFSS) or experiment.

[0048] S3. Train a neural network model to fit the frequency-resistivity relationship;

[0049] In step S3, training the neural network model to fit the frequency-resistivity relationship includes the following steps:

[0050] S301. Construct a BP neural network model: The input layer is the frequency f_i, the hidden layer contains 2 layers (each layer has 32 and 16 neurons respectively), and the output layer is the resistivity prediction value ρ_pred;

[0051] S302. Training data: Using {f_i, ρ_i} as samples, the mean squared error (MSE) is used as the loss function, and the optimizer is Adam (learning rate 0.001).

[0052] S303. Model Validation: The model converges when the prediction error (|ρ_pred-ρ_i| / ρ_i) < 3%.

[0053] S4. Search for the optimal frequency f_opt, so that the actual resistivity is ≤ ρ_ref and the inductance value is close to the target value;

[0054] In step S4, searching for the optimal frequency includes the following steps:

[0055] S401. Based on the trained model, traverse the frequency interval [f_min, f_max] and calculate ρ_pred for each f_i.

[0056] S401. Filter the frequency points that satisfy ρ_pred ≤ ρ_ref, and select the ones that make the inductance value L ( The frequency f_opt that is closest to the target value L_target (where l is the magnetic circuit length) is taken as the optimal operating frequency.

[0057] S5. Verify and correct the inductor parameters.

[0058] In step S5, the inductor parameter correction and verification includes the following steps:

[0059] S501, Recalculate the coil impedance Z_opt based on f_opt ( (where R_ac is the AC resistance).

[0060] S502. Test the Q value (quality factor) and temperature rise of the inductor under f_opt through experiments. If the Q value is ≥30 and the temperature rise is ≤40℃, the design is complete; otherwise, return to step S3, adjust the hidden layer parameters of the neural network and retrain.

[0061] This embodiment provides a specific implementation case: designing a high-frequency inductor with target parameters: L=10μH, f0=20MHz, Q≥40.

[0062] Step S1: Use enameled copper wire (σ=5.8×10) 7 S / m, d=0.5mm), the magnetic core is ferrite (μ r =2000, A=1cm², N=50 turns, K_skin=8%, calculated ρ_ref=1.72×10 -8 Ω·m;

[0063] Step S2: In the frequency range [10MHz, 30MHz], sample N=200 points and obtain ρ_i through simulation;

[0064] Step S3: Train the BP neural network; after 500 iterations, the MSE converges to 1.2 × 10⁻⁶. -17 ;

[0065] Step S4: The search yields f_opt = 22.5MHz, at which point ρ_pred = 1.68 × 10 -8 Ω·m (<ρ_ref), L=9.8μH (error 2%);

[0066] Step S5: The measured Q value is 45, and the temperature rise is 32℃, which meets the design requirements. Compared with the traditional fixed frequency design, this method can increase the inductor Q value by 15%~20% and reduce the temperature rise by 8~12℃.

[0067] This invention achieves precise design of high-frequency inductors through frequency-resistivity dynamic modeling and adaptive optimization, effectively solving the performance degradation problem caused by the skin effect. It can be widely used in high-frequency circuit systems in fields such as communications and energy.

[0068] The embodiments described above do not constitute a limitation on the scope of protection of this technical solution. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the above embodiments should be included within the scope of protection of this technical solution.

Claims

1. An adaptive inductor design method based on frequency modulation, characterized in that: Includes the following steps: S1. Collect the basic physical parameters of the inductor and calculate the reference resistivity ρ_ref; S2. Sample within the frequency range and obtain the corresponding resistivity sample; S3. Train a neural network model to fit the frequency-resistivity relationship; S4. Search for the optimal frequency f_opt, so that the actual resistivity is ≤ ρ_ref and the inductance value is close to the target value; S5. Verify and correct the inductor parameters.

2. The adaptive inductor design method based on frequency modulation according to claim 1, characterized in that: In step S1, the acquisition of the basic physical parameters of the inductor and the calculation of the reference resistivity ρ_ref include the following steps: S101. Collect the physical parameters of the target inductor: conductor material conductivity σ, permeability μ0, and core relative permeability μ0. r The number of coil turns N, the wire diameter d, and the cross-sectional area of ​​the magnetic core A; S102, Set the initial operating frequency f0 and the skin effect threshold K_skin; S103. Calculate the reference resistivity ρ_ref. ， Where ρ0 is the DC conductivity (σ -1 ), δ0 is the standard skin depth .

3. The adaptive inductor design method based on frequency modulation according to claim 1, characterized in that: In step S2, sampling and obtaining the corresponding resistivity sample within the frequency range includes the following steps: S201, Frequency modulation range: Set the frequency modulation interval [f_min, f_max]; S202, Sample generation: Sample N frequency points {f1, f2, ..., f_N} with a step size Δf; for each frequency point f_i, measure its corresponding actual resistivity ρ_i through electromagnetic simulation (such as HFSS) or experiment.

4. The adaptive inductor design method based on frequency modulation according to claim 3, characterized in that: In step S201, the frequency modulation interval [f_min, f_max] is [f0﹣50%f0, ​​f0+50%f0].

5. The adaptive inductor design method based on frequency modulation according to claim 1, characterized in that: In step S3, training the neural network model to fit the frequency-resistivity relationship includes the following steps: S301. Construct a BP neural network model: The input layer is the frequency f_i, the hidden layer contains 2 layers (each layer has 32 and 16 neurons respectively), and the output layer is the resistivity prediction value ρ_pred; S302. Training data: Using {f_i, ρ_i} as samples, mean squared error (MSE) is used as the loss function; S303. Model Validation: The model converges when the prediction error (|ρ_pred-ρ_i| / ρ_i) < 3%.

6. The adaptive inductor design method based on frequency modulation according to claim 1, characterized in that: In step S4, searching for the optimal frequency includes the following steps: S401. Based on the trained model, traverse the frequency interval [f_min, f_max] and calculate ρ_pred for each f_i. S401. Filter the frequency points that satisfy ρ_pred ≤ ρ_ref, and select the ones that make the inductance value L ( The frequency f_opt that is closest to the target value L_target (where l is the magnetic circuit length) is taken as the optimal operating frequency.

7. The adaptive inductor design method based on frequency modulation according to claim 1, characterized in that: In step S5, the inductor parameter correction and verification includes the following steps: S501, Recalculate the coil impedance Z_opt based on f_opt ( (where R_ac is the AC resistance). S502. Test the Q value (quality factor) and temperature rise of the inductor under f_opt through experiments. If the Q value is ≥30 and the temperature rise is ≤40℃, the design is complete; otherwise, return to step S3, adjust the hidden layer parameters of the neural network and retrain.

Images