Analog generation system of high-fidelity ultra-thin earphone structure and training method thereof

Abstract

This invention discloses a high-fidelity ultra-thin headphone structure simulation generation system and its training method, including the construction of a digital twin R&D platform. The platform includes a hardware layer, a software layer, and a data flow. The software layer includes: a geometric modeling module for 3D modeling and mesh generation, wherein the 3D modeling includes various components of the headphone shell and speaker; a material database module for storing material physical parameters and material shape properties; a multiphysics coupling solution module for performing electromagnetic-structural-acoustic joint simulation to simulate the change in tangency between the voice coil and the magnetic field when energized; an assembly stress and tolerance analysis module for simulating the installation process and warning of the risk of ring rubbing; and a process simulation module for simulating glue flow and thermo-electrical-mechanical coupling simulation.

Description

Technical Field

[0001] This invention relates to the field of headphone design technology, and in particular to a simulation generation system and training method for a high-fidelity ultra-thin headphone structure. Background Technology

[0002] With the trend towards thinner and lighter wearable devices such as TWS earphones and smartwatches (especially small products like smart rings), ultra-thin speakers (thickness <3mm) have become a key bottleneck restricting the sound quality experience of end devices.

[0003] The existing technology has the following main drawbacks:

[0004] 1. Traditional R&D relies on blueprints, design models, and the experience of staff to verify the design through repeated mold making, assembly, and testing.

[0005] For ultra-thin speakers, even minor assembly stress or magnetic circuit deviations can cause voice coil rubbing or frequency response distortion, resulting in a single R&D cycle of up to 3 months and an extremely high material loss rate.

[0006] Of course, there are also products that do not have high requirements for sound quality, and their development cycle will be greatly reduced. However, in actual industry design, achieving human voice imitation standards and having good durability are the key design requirements.

[0007] 2. Existing acoustic simulation software (such as COMSOL and ANSYS) often treats the spider as an ideal elastic body when dealing with ultra-thin structures, ignoring the nonlinear stiffness changes caused by glue hardening and interference fit. At the same time, electromagnetic-thermal-mechanical coupling calculations are extremely time-consuming (several hours for a single calculation), which cannot meet the needs of rapid iteration in consumer electronics.

[0008] 3. The magnetic gap of ultra-thin speakers is extremely small (<0.3mm), resulting in severe magnetic leakage and extreme sensitivity to assembly tolerances. Existing design methods cannot accurately predict magnetic flux attenuation and compliance (Cms) drift caused by housing shrinkage or installation pressure during the digital prototype stage, making it difficult to control mass production yield.

[0009] 4. Furthermore, for manufacturers and headphone manufacturers, in order to adapt to different appearances, changes in the shape of the speaker and the change in the spring also affect the sound quality of the speaker. Therefore, constantly designing different shapes also affects the progress of launching new products.

[0010] Therefore, ensuring design consistency and improving design efficiency has always been key to the acoustics industry. Although some companies have good internal analysis systems, they still cannot achieve the trend of design consistency. For example, Chinese patent CN202610214294.0 mainly achieves consistency through assembly, which is still in the later stages of design and R&D, and therefore does not solve the fundamental problem. Summary of the Invention

[0011] The main objective of this invention is to propose a high-fidelity ultra-thin headphone structure simulation generation system. The aim is to design a simulation generation system that can obtain simulated acoustic tests on the products to be generated, thereby reducing the need for multiple prototyping and design iterations, thus shortening the R&D cycle and improving R&D efficiency. At the same time, this design can also be applied to non-ultra-thin headphones and can also be widely used in the field of loudspeakers, improving the innovation of acoustic design.

[0012] To achieve the above objectives, this invention proposes a simulation generation system for a high-fidelity ultra-thin earphone structure, including the construction of a digital twin R&D platform, which includes a hardware layer, a software layer, and a data stream;

[0013] The software layer includes:

[0014] A geometric modeling module, used for 3D modeling and mesh generation, wherein the 3D modeling includes various components of the headphone shell and speaker;

[0015] The material database module is used to store material physical parameters and material shape properties;

[0016] A multiphysics coupling solution module is used to perform electromagnetic-structural-acoustic co-simulation.

[0017] Used to simulate the change in the amount of tangency between the voice coil and the magnetic field when the coil is energized;

[0018] An assembly stress and tolerance analysis module is used to simulate the installation process and provide early warning of the risk of rubbing rings.

[0019] The process simulation module is used to simulate the flow of adhesive and the thermo-electric-mechanical coupling (in actual production, the various components of the loudspeaker, such as magnets (multiple), gaskets, magnetic rings (for confining the magnetic field), voice coils, and voice covers, are generally glued automatically to ensure processing accuracy).

[0020] An acoustic fingerprint recognition and benchmarking module is used to generate a frequency response curve and compare it with the metal detector; (i.e., simulation fidelity).

[0021] A dual-channel collaborative processing module is used to automatically calculate the combination of the production model and manual fine-tuning. The dual-channel collaborative processing module includes a first channel of fully automatic simulation and a second channel of manual collaborative correction.

[0022] A data closure and self-updating module, which is used to record human intervention and drive incremental training of the model;

[0023] The specific steps include:

[0024] Perform fully automated simulation on the first channel:

[0025] The electromagnetic field distribution, vibration modes and acoustic response of the ultrathin loudspeaker are calculated based on the multiphysics solver. The Gaussian process regression (GPR) surrogate model is used to replace part of the finite element calculation, and the predicted frequency response curve is quickly output.

[0026] Perform manual collaborative correction in the second channel:

[0027] When the deviation between the measured data and the simulation data exceeds a preset threshold, a manual intervention loop is triggered, and the simulation parameters are fine-tuned by introducing a leakage magnetic correction factor or a compliant drift model.

[0028] The weights of the material database and surrogate model are updated based on the corrected parameters.

[0029] Specifically, the solution for the electromagnetic field distribution employs an algorithm combining Maxwell's equations with a magnetoresistance model:

[0030] The formula for calculating electromagnetic driving force is: F=∫(J×B)dV;

[0031] in,

[0032] J is the voice coil current density;

[0033] B is the air gap magnetic flux density (solved from Maxwell's equations: ∇×H=J);

[0034] Magnetic flux Φ = ∮SB⋅dS.

[0035] The system optimizes B by changing the shape of the magnetic circuit (T-iron, washer);

[0036] The equivalent circuit method is used to transform it into a second-order differential equation;

[0037] ;

[0038] in:

[0039] Mms: Vibrating mass (voice coil + diaphragm); Rms: Force resistance (affected by air damping and housing sealing);

[0040] Kms: stiffness coefficient (determined by the material of the spring).

[0041] Specifically, the vibration modal analysis employs thin-plate vibration theory and a nonlinear stiffness model, with the following formula:

[0042] Control equation for thin plate vibration:

[0043] ;

[0044] The bonding / snap contact surface between the housing and the speaker is defined by the constraint matrix [C] to simulate the vibration mode shift caused by installation stress (which is the main cause of high losses).

[0045] in,

[0046] D is the bending stiffness.

[0047] w represents the vibration displacement.

[0048] ρ is density.

[0049] h is the thickness;

[0050] nonlinear restoring force model of elastic waves:

[0051] ;

[0052] in,

[0053] Kms is the linear compliance coefficient.

[0054] α is a nonlinear coefficient.

[0055] x is the displacement.

[0056] Specifically, a model is introduced into the first channel for rapid prediction (which can be the large model described below, or an existing trained model can be introduced).

[0057] The mapping relationship is constructed using Gaussian process regression (GPR):

[0058] f(x)∼GP(m(x),k(x,x′));

[0059] The input vector x includes the magnetic ring thickness, magnetic flux, housing wall thickness, and installation preload, while the output vector includes the total harmonic distortion (THD) and sound pressure level (SPL).

[0060] Specifically, the second channel manual collaborative correction includes:

[0061] For micro-control of magnetic flux,

[0062] A leakage flux correction factor η is introduced to compensate for the simulation results:

[0063] ;

[0064] Regarding flexural drift,

[0065] Introducing a glue curing compensation model:

[0066] Kms_real = Kms_sim × (1 + βglue)

[0067] in,

[0068] βglue is the glue curing coefficient.

[0069] It was calculated by engineers based on the measured resonant frequency F0 after actual assembly.

[0070] Specifically, it also includes assembly tolerance analysis methods:

[0071] Using Monte Carlo simulation to randomly generate shell shrinkage rate errors, simulate N virtual assemblies, and calculate the actual air gap:

[0072] ;

[0073] When the minimum gap Gapmin ≤ 0, the system issues a rubbing risk warning.

[0074] A training method for a high-fidelity ultra-thin headphone structure simulation generation system, characterized in that it includes the aforementioned high-fidelity ultra-thin headphone structure simulation generation system, and the training steps include:

[0075] S1: Obtain the three-dimensional structural data and material physical properties of the ultra-thin speaker to be developed, and construct a digital twin model;

[0076] S2: Execute the first channel fully automatic test, call the multiphysics solver to perform electromagnetic field, structural mechanics and acoustic simulation on the digital twin model, and output the first predicted acoustic response;

[0077] S3: Determine whether the first predicted acoustic response meets the preset performance index. If it does not meet the index, mark it as an unqualified solution. If it is in the critical range or the deviation from the measured data exceeds the limit, trigger the second channel manual intervention.

[0078] S4: In response to the engineer's input, the magnetic flux distribution and elasto-compliance parameters in the simulation model are corrected, including the introduction of a leakage magnetic field correction factor η and a compliance drift coefficient β.

[0079] S5: Recalculate the acoustic response based on the corrected parameters and send the corrected data back to the database to update the simulation model.

[0080] Specifically, the system employs a physics-driven loss function, Loss, during training:

[0081] ;

[0082] in,

[0083] λ is the regularization coefficient, which ensures that the model results conform to the physical constraints of Maxwell's equations.

[0084] Advantages of this application:

[0085] 1. Using general 3D modeling software (such as Soliworks, PROE and other mainstream graphic design software) as a foundation, and combining it with databases, material libraries and multiphysics coupling solution modules, preliminary product models and basic parameters can be obtained, thereby reducing the problem of repeated mold making and effectively shortening the product design cycle.

[0086] Furthermore, by using the assembly stress and tolerance analysis module and the process simulation module, the specifications and shape of the speaker can be added, removed, or modified according to actual needs, thereby achieving specific design and the universality of the design, and further improving the production precision of the ultra-thin speaker.

[0087] It is a good design model in the industry and can be widely used in this industry and even large audio systems. However, large audio systems involve complex circuits (such as changes in parameters such as resistance and capacitance, so their precision requirements and design requirements are higher).

[0088] 2. Through a specific training method, namely a dual-channel collaborative processing module, the combination of automated calculation of loudspeaker parameters of the production model and manual fine-tuning is realized. The dual-channel collaborative processing module includes a first channel of fully automatic simulation and a second channel of manual collaborative correction.

[0089] In actual simulation design, due to the influence of tolerances, materials and environment, there is still a deviation between computer parameters and actual design. Therefore, by testing the molded product, human parameters are introduced to achieve fine-tuning of simulation parameters and human parameters, thereby optimizing the product, such as design preferences for low frequency, medium frequency and high frequency, and thus achieving product-specific design.

[0090] 3. In addition, the existing modifications that require experienced engineers will be replaced with model-based adjustments to the production process and automatic predictions, thereby reducing problems in automated production and improving production consistency. Attached Figure Description

[0091] Figure 1 Here is a flowchart of the system training and feedback mechanism;

[0092] Figure 2 This is a comparison diagram of the simulated elastic wave compliance and artificial compensation in this application;

[0093] Figure 3 This is a comparison between the simulated magnetic flux and the manual correction in this application. Detailed Implementation

[0094] The technical solutions in the embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0095] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, top, bottom, inside, outside, vertical, horizontal, longitudinal, counterclockwise, clockwise, circumferential, radial, axial, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0096] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0097] like Figure 1 As shown, a high-fidelity ultra-thin headphone structure simulation generation system includes building a digital twin R&D platform, which includes a hardware layer, a software layer, and a data stream;

[0098] The software layer includes:

[0099] A geometric modeling module, used for 3D modeling and mesh generation, wherein the 3D modeling includes various components of the headphone shell and speaker;

[0100] The material database module is used to store material physical parameters and material shape properties;

[0101] A multiphysics coupling solution module is used to perform electromagnetic-structural-acoustic co-simulation.

[0102] Used to simulate the change in the amount of tangency between the voice coil and the magnetic field when the coil is energized;

[0103] An assembly stress and tolerance analysis module is used to simulate the installation process and provide early warning of the risk of rubbing rings.

[0104] The process simulation module is used to simulate the flow of adhesive and the thermo-electric-mechanical coupling (in actual production, the various components of the loudspeaker, such as magnets (multiple), gaskets, magnetic rings (for confining the magnetic field), voice coils, and voice covers, are generally glued automatically to ensure processing accuracy).

[0105] An acoustic fingerprint recognition and benchmarking module is used to generate a frequency response curve and compare it with the metal detector; (i.e., simulation fidelity).

[0106] A dual-channel collaborative processing module is used to automatically calculate the combination of the production model and manual fine-tuning. The dual-channel collaborative processing module includes a first channel of fully automatic simulation and a second channel of manual collaborative correction.

[0107] A data closure and self-updating module, which is used to record human intervention and drive incremental training of the model;

[0108] The specific steps include:

[0109] Perform fully automated simulation on the first channel:

[0110] The electromagnetic field distribution, vibration modes and acoustic response of the ultrathin loudspeaker are calculated based on the multiphysics solver. The Gaussian process regression (GPR) surrogate model is used to replace part of the finite element calculation, and the predicted frequency response curve is quickly output.

[0111] Perform manual collaborative correction in the second channel:

[0112] When the deviation between the measured data and the simulation data exceeds a preset threshold, a manual intervention loop is triggered, and the simulation parameters are fine-tuned by introducing a leakage magnetic correction factor or a compliant drift model.

[0113] The weights of the material database and surrogate model are updated based on the corrected parameters.

[0114] Step 1: Virtual Inventory

[0115] Export the material library from 3D software (such as SolidWorks) as XML / JSON.

[0116] Key step: Supplementing the "viscoelastic parameter". Standard 3D libraries only provide density and hardness; the loss factor tanδ must be added (which has a significant impact on sound quality).

[0117] Step 2: Assembly Interference Check and Stress Relief

[0118] Operation: Import the shell STL file.

[0119] Simulation: Run a nonlinear statics analysis.

[0120] Judgment criteria: If the maximum stress exceeds 80% of the spring sheet's yield strength, the system will automatically alarm "The assembly stress is too high, and it is recommended to modify the shell chamfer."

[0121] Step 3: Flux Mapping

[0122] Operation: Run the electromagnetic simulation.

[0123] Output: Generates a Flux Density Heatmap.

[0124] Correction: If the central magnetic density is insufficient, the system suggests adjusting the magnet grade (e.g., changing from N48 to N52).

[0125] Step 4: Acoustic Fingerprint Generation

[0126] Operation: Input a standard test signal (such as powder noise).

[0127] Outputs: Predicted frequency response (FR), impedance curve, waterfall plot.

[0128] Comparison: Compare with "Golden Sample" in the database and calculate the similarity index.

[0129] Step 5: Generate a test report (Auto Report)

[0130] The system automatically outputs a "Risk Assessment Report," which indicates:

[0131] Which frequency band might experience distortion due to installation issues?

[0132] Which component has the most sensitive tolerance?

[0133] Training Methodology

[0134] 1. Dataset Construction

[0135] Sample source: Collection of at least 500+ past cases of both successful and unsuccessful ultra-thin speakers.

[0136] Feature engineering:

[0137] Geometric features: magnetic gap height, voice coil winding width.

[0138] Physical characteristics: Bl curve, Cms (mechanical compliance).

[0139] Process characteristics: glue thickness (0.05mm vs 0.1mm), assembly eccentricity.

[0140] 2. Model Training Process

[0141] Pre-training: The Physics-Informed Neural Network (PINN) is trained using data from a general materials library to ensure compliance with the laws of thermodynamics and electromagnetism;

[0142] Fine-tuning: Import the enterprise's private database and fit the "shell resonance peak" for a specific model.

[0143] Generative Adversarial Networks (GANs): Generate data on extreme operating conditions (such as magnetic flux decay under high temperature and humidity) to compensate for insufficient measured data.

[0144] This channel is triggered when the result of the first channel is in the "gray area" (i.e., the simulation is qualified but critical, or the deviation between the measured and simulated results is >3dB). Engineers intervene through the "expert parameter overlay layer".

[0145] 1. Manual Flux Tuning

[0146] Scenario: The simulation shows that the magnetic flux is sufficient, but the actual listening experience is "insufficient" (usually because magnetic leakage is not fully accounted for).

[0147] Intervention tool: Magnetic circuit topology editor

[0148] Manual operation: Engineers manually adjust the "magnetic circuit shunt coefficient" or "washer shape".

[0149] Model correction: Introduce a leakage magnetic field correction factor η into the original formula.

[0150] Bcorrected = η⋅Bsimulated (η is set by the engineer based on experience, such as 0.95);

[0151] Implementation: Instead of rerunning the full model, the system quickly recalculates the Bl product based on the engineer's input η and updates the frequency response curve.

[0152] 2. Manual Spider Tuning (Late-Stage Intervention)

[0153] Scenario: After the ultra-thin speaker is assembled, the glue hardens, causing the spider to stiffen and the f0 drift to occur.

[0154] Intervention tool: compliance compensation slider

[0155] Manual operation: Engineers reverse-engineer the "adhesive hardening coefficient" based on the actual measured f0 after the adhesive is applied.

[0156] Model revision: Introduce the Compliance Shift Model.

[0157] Kms_real = Kms_sim × (1 + βglue)

[0158] βglue: Glue hardening compensation value (manually entered).

[0159] Implementation: Engineers drag the UI slider to adjust the β value, and the system updates the waterfall plot in real time to observe whether the segmentation vibration disappears.

[0160] in Figure 1 , Figure 2 and Figure 3 This is a schematic diagram of a test process for one embodiment.

[0161] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A simulation generation system for a high-fidelity ultra-thin earphone structure, characterized in that, This includes building a digital twin R&D platform, which includes a hardware layer, a software layer, and a data flow. The software layer includes: A geometric modeling module, used for 3D modeling and mesh generation, wherein the 3D modeling includes various components of the headphone shell and speaker; The material database module is used to store material physical parameters and material shape properties; A multiphysics coupling solution module is used to perform electromagnetic-structural-acoustic co-simulation. Used to simulate the change in the amount of tangency between the voice coil and the magnetic field when the coil is energized; An assembly stress and tolerance analysis module is used to simulate the installation process and provide early warning of the risk of rubbing rings. A process simulation module, which is used to simulate the flow of adhesive and the thermo-electric-mechanical coupling. An acoustic fingerprint recognition and benchmarking module is used to generate a frequency response curve and compare it with the metal detector; A dual-channel collaborative processing module is used to automatically calculate the combination of the production model and manual fine-tuning. The dual-channel collaborative processing module includes a first channel of fully automatic simulation and a second channel of manual collaborative correction. A data closure and self-updating module, which is used to record human intervention and drive incremental training of the model; The specific steps include: Perform fully automated simulation on the first channel: The electromagnetic field distribution, vibration modes and acoustic response of the ultrathin loudspeaker are calculated based on the multiphysics solver. The Gaussian process regression surrogate model is used to replace part of the finite element calculation and the predicted frequency response curve is quickly output. Perform manual collaborative correction in the second channel: When the deviation between the measured data and the simulation data exceeds a preset threshold, a manual intervention loop is triggered, and the simulation parameters are fine-tuned by introducing a leakage magnetic correction factor or a compliant drift model. The weights of the material database and surrogate model are updated based on the corrected parameters; A model is introduced into the first channel for rapid prediction. The mapping relationship is constructed using Gaussian process regression (GPR): f(x)∼GP(m(x),k(x,x′)); The input vector x includes the magnetic ring thickness, magnetic flux, shell wall thickness and installation preload, and the output vector includes the total harmonic distortion (THD) and sound pressure level (SPL). The second channel manual collaborative correction specifically includes: For micro-control of magnetic flux, A leakage flux correction factor η is introduced to compensate for the simulation results: ； Regarding flexural drift, Introducing a glue curing compensation model: Kms_real=Kms_sim×(1+βglue); in, βglue is the glue curing coefficient. It was calculated by engineers based on the measured resonant frequency F0 after actual assembly.

2. The simulation generation system for a high-fidelity ultra-thin earphone structure according to claim 1, characterized in that, The electromagnetic field distribution is solved using the following algorithm combining Maxwell's equations and a magnetoresistance model: The formula for calculating electromagnetic driving force is: F=∫(J×B)dV; in, J is the voice coil current density; B is the air gap magnetic flux density; Magnetic flux Φ = ∮SB⋅dS. The system optimizes B by changing the shape of the magnetic circuit; Using the equivalent circuit method, it is transformed into a second-order differential equation: ； in: Mms: Vibrational mass; Rms: Force resistance; Kms: stiffness coefficient.

3. The simulation generation system for a high-fidelity ultra-thin earphone structure according to claim 1, characterized in that, The vibration modal analysis employs thin-plate vibration theory and a nonlinear stiffness model, with the specific formula as follows: Control equation for thin plate vibration: ； in, D is the bending stiffness. w represents the vibration displacement. ρ is density. h is the thickness; nonlinear restoring force model of elastic waves: ； in, Kms is the linear compliance coefficient. α is a nonlinear coefficient. X represents the displacement.

4. The simulation generation system for a high-fidelity ultra-thin earphone structure according to claim 1, characterized in that, It also includes assembly tolerance analysis methods: Using Monte Carlo simulation to randomly generate shell shrinkage rate errors, simulate N virtual assemblies, and calculate the actual air gap: ; When the minimum gap Gapmin ≤ 0, the system issues a rubbing risk warning.

5. A training method for a high-fidelity ultrathin headphone structure simulation generation system, characterized in that, The simulation generation system for the high-fidelity ultra-thin earphone structure according to any one of claims 1-4 includes the following training steps: S1: Obtain the three-dimensional structural data and material physical properties of the ultra-thin speaker to be developed, and construct a digital twin model; S2: Execute the first channel fully automatic test, call the multiphysics solver to perform electromagnetic field, structural mechanics and acoustic simulation on the digital twin model, and output the first predicted acoustic response; S3: Determine whether the first predicted acoustic response meets the preset performance index. If it does not meet the index, mark it as an unqualified solution. If it is in the critical range or the deviation from the measured data exceeds the limit, trigger the second channel manual intervention. S4: In response to the engineer's input, the magnetic flux distribution and elasto-compliance parameters in the simulation model are corrected, including the introduction of a leakage magnetic field correction factor η and a compliance drift coefficient β; S5: Recalculate the acoustic response based on the corrected parameters and send the corrected data back to the database to update the simulation model.

6. The training method for the simulation generation system of the high-fidelity ultra-thin earphone structure according to claim 1, characterized in that, The system employs a physics-driven loss function, Loss, during training. ； in, λ is the regularization coefficient, which ensures that the model results conform to the physical constraints of Maxwell's equations.

Images