Real-time monitoring method and system for deformation of strain sensor embedded in loudspeaker diaphragm

Abstract

The application discloses a strain sensor deformation real-time monitoring method and system embedded in a loudspeaker diaphragm, and relates to the technical field of intelligent control of electroacoustic transducers; a micrometer flexible sensor network is conformally sprayed on the surface of the diaphragm to collect a micro-strain sequence of the diaphragm in real time without destroying zero acoustic parameters; a one-dimensional sequence is reconstructed into a continuous three-dimensional micro-deformation topology by using an orthogonal basis matrix, and a spatial nonlinear acoustic error is extracted in combination with a rigid body reference model to perform Volterra active acoustic compensation; the fatigue degree of the diaphragm is calculated in real time based on rain flow counting and cumulative damage theory, a frequency band compression limit is automatically triggered when the fatigue is critical, and the fatigue degree is used as a safety constraint factor to dynamically limit the step length of acoustic compensation; the application breaks the offline limitation of traditional laser vibration measurement, effectively offsets the acoustic distortion caused by segmented vibration, constructs a dynamic restraint closed loop of sound quality compensation and mechanical anti-tearing, and realizes the balance between high-fidelity driving of the loudspeaker and the limit physical life.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for electroacoustic transducers, and more specifically, relates to a method and system for real-time monitoring of deformation of strain sensors embedded in loudspeaker diaphragms. Background Technology

[0002] As the core electroacoustic transducer that converts electrical signals into sound signals, loudspeakers are widely used in professional monitoring, consumer electronics, and car audio systems. With the development of modern digital audio technology, loudspeakers are evolving towards higher power, smaller size, and deeper low frequencies. This requires the loudspeaker's core sound-producing component—the diaphragm—to withstand enormous electromagnetic driving forces while undergoing high-frequency, long-stroke reciprocating motion, frequently pushing the physical stress of the diaphragm material to its limits.

[0003] Ideally, a speaker diaphragm should move synchronously as a rigid body, like a piston. However, in real-world applications (especially in mid-to-high frequency ranges and high-dynamic-range music), the diaphragm inevitably experiences "breakup modes," meaning that different areas of the diaphragm deform asynchronously (e.g., standing waves, swaying), leading to severe nonlinear acoustic distortion (e.g., intermodulation distortion and higher harmonic distortion). Currently, the industry primarily relies on laser Doppler vibrometers (such as the Klippel system) in laboratory environments to observe diaphragm deformation. This optical approach is bulky and extremely expensive, suitable only for offline testing during the R&D phase, and difficult to monitor in real-time during actual music playback at the user end. In the field of active control, existing dynamic coil feedback (MFB) technology typically involves attaching an accelerometer to the voice coil frame, but this approach only acquires one-dimensional motion data of the voice coil along a single axis, making it difficult to perceive the microscopic deformation topology of the entire diaphragm surface. In addition, existing DSP predistortion compensation algorithms are mostly based on static mathematical models (such as the traditional Volterra series), lacking real-time physical feedback from the diaphragm itself, and belong to the "open-loop blind guessing" type of compensation.

[0004] This reveals the following problems with existing technologies: 1. There is a lack of means to acquire multi-dimensional deformation data of the diaphragm surface in real time and without loss under actual working conditions; 2. Existing acoustic compensation algorithms lack closed-loop guidance from physical deformation data, making it difficult to handle spatial nonlinear distortion caused by segmented vibration; 3. Traditional acoustic compensation focuses solely on reducing distortion (increasing driving force), neglecting the mechanical fatigue limit of the diaphragm material, which can easily accelerate the mechanical tearing of the diaphragm under high power, and lacks a mechanism for balancing sound quality and lifespan. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the problems in related technologies, this invention provides a method and system for real-time monitoring of deformation of a strain sensor embedded in a loudspeaker diaphragm, thereby overcoming the aforementioned technical problems existing in the prior art.

[0007] (II) Technical Solution

[0008] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0009] S1. Construct a network of miniature strain sensors made of flexible piezoresistive material and establish a low-mass additional signal extraction path along the non-sound-producing physical structure of the loudspeaker;

[0010] S2. When the loudspeaker is driven to produce sound by an audio electrical signal, the analog resistance change signal output by each node in the micro strain sensor network is collected based on the low-mass-added signal extraction path, and a one-dimensional digital strain time series is obtained by analog-to-digital conversion.

[0011] S3. Spatial mapping calculation is performed on the one-dimensional digital strain time series using the orthogonal basis matrix of deformation modes to obtain a continuous three-dimensional micro-deformation topology matrix.

[0012] S4. Extract the extreme values ​​of the one-dimensional digital strain time series and calculate the current cumulative fatigue damage degree; when the cumulative fatigue damage degree reaches the damage degree threshold, locate the fatigue occurrence area and generate a dynamic pressure limit attenuation command.

[0013] S5. The continuous three-dimensional micro-deformation topology matrix and the ideal rigid piston motion reference model of the loudspeaker diaphragm are compared by difference and weighted summation by area integral to obtain the spatial nonlinear error scalar signal.

[0014] The original audio signal is pre-filtered using a dynamic compression and attenuation command to obtain a safe audio signal.

[0015] S6. Using the accumulated fatigue damage as a boundary constraint, based on the spatial nonlinear error scalar signal and the safety audio electrical signal, an inverse acoustic compensation signal is dynamically generated through an adaptive nonlinear filter and superimposed on the safety audio electrical signal to drive the speaker.

[0016] Preferably, step S1 includes the following steps:

[0017] S11. Extract the geometric feature parameters of the loudspeaker diaphragm; the geometric feature parameters include the outer diameter of the surround, the inner diameter of the surround, the radius of the middle section of the cone, and the radius of the joint between the voice coil and the skeleton;

[0018] S12. Based on the geometric feature parameters, define the spatial topology of the micro-strain sensor network; the spatial topology includes a first ring array arranged at the connection between the folded ring and the diaphragm, and a second radial array arranged in the middle section of the diaphragm.

[0019] S13. Using nano-graphene conductive ink, the micro-strain sensor network is directly printed on the back of the speaker diaphragm according to the spatial topology layout through aerosol conformal inkjet printing process.

[0020] Using nano-silver wire ink to print ultra-fine conductive traces, the pins of all sensor nodes are brought together to the voice coil skeleton and led out through multi-strand nylon wires woven inside the speaker centering support to generate a low-quality additional signal output path.

[0021] Preferably, step S2 includes the following steps:

[0022] S21. Convert the resistance change rate caused by the deformation of the graphene sensor into a millivolt-level analog voltage signal, and amplify it with a fixed gain to obtain the amplified voltage signal.

[0023] S22. Perform digital-to-analog conversion on the amplified voltage signal to obtain the converted voltage signal;

[0024] S23. Perform digital filtering and sequence reconstruction on the converted voltage signal to obtain a one-dimensional digital strain time series.

[0025] Preferably, step S3 includes the following steps:

[0026] S31. Define the surface polar coordinate mesh system of the loudspeaker diaphragm; divide the diaphragm surface into M discrete mesh nodes, where the value of M is greater than the number of sensor nodes N contained in the micro-strain sensor network in S1;

[0027] S32. Obtain the real-time deformation mode orthogonal basis matrix; the dimension of the deformation mode orthogonal basis matrix is ​​M×N; the internal elements of the deformation mode orthogonal basis matrix represent the fixed physical mapping weights between the local strain of N sensor nodes and the three-dimensional spatial displacement of M mesh nodes;

[0028] S33. At the current sampling time, reorganize the one-dimensional digital strain time series generated in S2 into a column vector with dimension N×1;

[0029] S34. Based on the orthogonal basis matrix and column vector of the deformation mode, the column vector of dimension M×1 is calculated in real time using the matrix multiplication formula to obtain the continuous three-dimensional micro-deformation topology matrix.

[0030] Preferably, the real-time deformation mode orthogonal basis matrix in S32 includes the following steps:

[0031] S321. Input a logarithmic sweep frequency test signal to the speaker equipped with the intelligent diaphragm hardware carrier in S1;

[0032] S322. Simultaneously record the digital strain calibration sequence output by N sensors, and simultaneously scan and record the real physical Z-axis displacement sequence of M grid nodes using a laser Doppler vibrometer.

[0033] S323. Construct the least squares objective function with L2 regularization term;

[0034] S324. Solve the deformation mode orthogonal basis matrix that minimizes the objective function of the least squares method by using the singular value decomposition algorithm, and obtain the real-time deformation mode orthogonal basis matrix.

[0035] Preferably, step S4 includes the following steps:

[0036] S41. Based on the one-dimensional digital strain time series in S2, the rainflow counting algorithm is used to count the strain cycle amplitude and the corresponding real-time cycle occurrence number of each sensor node within a preset sliding time window.

[0037] S42, Call the speaker diaphragm material pre-written into memory The fatigue life curve equation calculates the limit number of cycles at which the material will undergo irreversible fracture under the strain cycle amplitude.

[0038] S43. The limit number of cycles for irreversible fracture of the material based on the real-time cycle occurrence number and strain cycle amplitude; calculate the damage increment within the current sliding time window using the Palmgren-Miner linear cumulative damage theory;

[0039] The damage increment within the current sliding time window is accumulated to obtain the updated real-time cumulative fatigue damage of the current node;

[0040] S44. When the cumulative fatigue damage degree corresponding to any sensor node is greater than or equal to the safety threshold, it is determined that the area is about to undergo physical tearing, the dynamic compression attenuation command is immediately triggered and generated, and output to step S5 to limit the audio input.

[0041] Preferably, obtaining the dynamic compression attenuation command in S44 includes the following steps:

[0042] S441. When the real-time cumulative fatigue damage of a certain node reaches the safety threshold, the system backtracks the rain flow counting records statistically obtained by the rain flow counting algorithm in S41 and extracts the center excitation frequency that causes the maximum strain cycle amplitude.

[0043] S442. Control the digital frequency divider to generate a notch attenuation band with a bandwidth of 1 octave at the center excitation frequency, and obtain the dynamic compression attenuation command.

[0044] Preferably, step S5 includes the following steps:

[0045] S51: Receive the original audio electrical signal. When the signal passes through the digital frequency divider, it is controlled by the dynamic compression and attenuation command generated by S4 to perform pre-notch processing, reduce the energy of dangerous frequency bands, and obtain a safe audio electrical signal.

[0046] S52. Construct an ideal rigid body piston motion reference model; collect the input current generated by the speaker voice coil based on the safe audio electrical signal in real time, and calculate the ideal axial displacement scalar of the voice coil at the current moment by combining the force coefficient; assign the ideal axial displacement scalar to the M grid nodes defined in S3 with equal amplitude to generate an ideal displacement matrix of dimension M×1.

[0047] S53. Subtract the continuous three-dimensional micro-deformation topology matrix in S3 from the ideal displacement matrix element by element to obtain a spatial residual matrix with dimension M×1.

[0048] S54. Perform area integral weighted summation based on acoustic radiation efficiency on the spatial residual matrix to extract the spatial nonlinear error scalar signal in single scalar form;

[0049] Preferably, step S6 includes the following steps:

[0050] S61. Construct a second-order Volterra adaptive filter in a digital signal processor; after filtering the safety audio electrical signal generated in step S51 through the secondary acoustic path transfer function of the loudspeaker, obtain the reference signal;

[0051] S62. Based on the reference signal and the spatial nonlinear error scalar signal, and introducing the real-time cumulative fatigue damage degree generated in S4 as a safety penalty variable, dynamically calculate the adaptive step size factor containing fatigue constraints.

[0052] S63. Based on the filter-x minimum mean square error algorithm, and introducing the cumulative fatigue damage degree generated in S4 as a safety penalty variable, the optimal filter weight vector at the current sampling time is obtained through the restricted weight update iterative formula of the filter.

[0053] S64. Perform Volterra convolution operation on the current safety audio electrical signal using the optimal filter weight vector to calculate the reverse acoustic compensation signal; after phase reversal of the reverse acoustic compensation signal, superimpose it onto the safety audio electrical signal and output it to the power amplifier to drive the speaker.

[0054] Preferably, a real-time deformation monitoring system for a strain sensor embedded in a loudspeaker diaphragm is used to implement the real-time deformation monitoring method for the strain sensor embedded in the loudspeaker diaphragm. The system includes an intelligent diaphragm physical module, a high-frequency signal acquisition module, a DSP reconstruction module, a fatigue calculation and protection generation module, and a constrained acoustic compensation execution module, wherein:

[0055] The aforementioned intelligent diaphragm physical module is used to perform step S1 and provides a hardware carrier with a micro-strain sensor network and a low-mass additional signal extraction path.

[0056] The aforementioned high-frequency signal acquisition module is used to execute step S2, connect the signal output path, and convert the analog weak electrical signal into a one-dimensional digital strain time series.

[0057] The aforementioned DSP reconstruction module is used to execute step S3, receiving a one-dimensional digital strain time series to reconstruct a continuous three-dimensional microscopic deformation topology matrix.

[0058] The fatigue calculation and protection generation module described above is used to execute step S4, calculate the cumulative fatigue damage, generate dynamic pressure limit attenuation instructions, and upload the offline structural optimization dataset to the cloud.

[0059] The aforementioned restricted acoustic compensation execution module is used to execute steps S5 and S6, receive the dynamic compression attenuation command issued by S4 to generate a safe audio electrical signal, and simultaneously receive the cumulative fatigue damage degree issued by S4 as the calculation boundary. Based on the three-dimensional topology error, it runs the restricted Volterra filter and outputs a safe reverse acoustic compensation signal.

[0060] (III) Beneficial Effects

[0061] The present invention has the following beneficial effects:

[0062] This invention employs aerosol conformal printing technology to construct a micron-scale flexible piezoresistive sensor network within the diaphragm surface of a loudspeaker. This reduces the proportion of total added mass in the equivalent vibrational mass and, without altering the original acoustic physical characteristics of the loudspeaker (such as resonant frequency and quality factor), overcomes the limitations of traditional laser vibration meters used offline, transforming the loudspeaker into intelligent hardware with real-time, in-situ body sensing capabilities.

[0063] This invention introduces a deformation mode orthogonal basis matrix, which instantly reconstructs one-dimensional discrete strain data into a high-resolution three-dimensional micro-deformation topology; further, by integrating the acoustic radiation efficiency area, the complex physical space residual is reduced to a single acoustic nonlinear error scalar; this provides an accurate objective function for the Volterra filter, realizes reverse acoustic cancellation for segmented vibrations, and reduces intermodulation distortion and harmonic distortion in the mid-to-high frequency band.

[0064] This invention is the first to introduce rainflow counting and Miner's cumulative damage theory into the electroacoustic control layer; it calculates the mechanical fatigue damage of the diaphragm in real time and embeds it as the highest priority safety penalty factor into the filter's step size update formula; when the diaphragm ages or faces the risk of tearing, the system can automatically reduce the acoustic compensation intensity and perform compression attenuation of specific frequency bands; it solves the defect of traditional compensation algorithms that blindly pursue sound quality and accelerate device damage, achieves an adaptive balance between acoustic fidelity and mechanical life, and provides digital twin data for the development of next-generation products.

[0065] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0066] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0067] Figure 1 This is a flowchart illustrating the real-time deformation monitoring method of the strain sensor embedded in the loudspeaker diaphragm according to the present invention.

[0068] Figure 2 A schematic diagram of the module of the strain sensor deformation real-time monitoring system embedded in the speaker diaphragm of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0070] To address the technical problems raised in the background section, please refer to [link / reference]. Figure 1 This invention provides a method and system for real-time monitoring of strain sensor deformation embedded in a loudspeaker diaphragm, comprising:

[0071] S1. Construct a network of miniature strain sensors made of flexible piezoresistive material and establish a low-mass additional signal extraction path along the non-sound-producing physical structure of the loudspeaker;

[0072] S2. When the loudspeaker is driven to produce sound by an audio electrical signal, the analog resistance change signal output by each node in the micro strain sensor network is collected based on the low-mass-added signal extraction path, and a one-dimensional digital strain time series is obtained by analog-to-digital conversion.

[0073] S3. Spatial mapping calculation is performed on the one-dimensional digital strain time series using the orthogonal basis matrix of deformation modes to obtain a continuous three-dimensional micro-deformation topology matrix.

[0074] S4. Extract the extreme values ​​of the one-dimensional digital strain time series and calculate the current cumulative fatigue damage degree; when the cumulative fatigue damage degree reaches the damage degree threshold, locate the fatigue occurrence area and generate a dynamic pressure limit attenuation command.

[0075] S5. The continuous three-dimensional micro-deformation topology matrix and the ideal rigid piston motion reference model of the loudspeaker diaphragm are compared by difference and weighted summation by area integral to obtain the spatial nonlinear error scalar signal.

[0076] The original audio signal is pre-filtered using a dynamic compression and attenuation command to obtain a safe audio signal.

[0077] S6. Using the accumulated fatigue damage as a boundary constraint, based on the spatial nonlinear error scalar signal and the safety audio electrical signal, an inverse acoustic compensation signal is dynamically generated through an adaptive nonlinear filter and superimposed on the safety audio electrical signal to drive the speaker.

[0078] This invention integrates a flexible sensing network non-destructively onto the surface of a loudspeaker diaphragm, transforming physical deformation into a one-dimensional digital sequence without altering the original acoustic parameters. This sequence is then instantaneously reconstructed into a continuous three-dimensional microscopic deformation topology using an orthogonal basis matrix. Based on the three-dimensional deformation, nonlinear acoustic errors are extracted for active acoustic compensation. The mechanical fatigue damage of the diaphragm is measured in real time and used as a safety constraint factor to dynamically limit the step size of acoustic compensation and perform frequency band limiting. This solves the distortion problem caused by loudspeaker split vibration under high-power drive, while avoiding mechanical tearing of the diaphragm due to overcompensation, achieving a balance between high-fidelity sound quality and physical lifespan.

[0079] The above embodiment S1 includes the following steps:

[0080] S11. Extract the geometric feature parameters of the loudspeaker diaphragm; the geometric feature parameters include the outer diameter of the surround, the inner diameter of the surround, the radius of the middle section of the cone, and the radius of the joint between the voice coil and the skeleton;

[0081] In specific implementation, the above embodiment S11 is as follows: Taking a professional-grade 6.5-inch Kevlar composite fiber mid-bass speaker as an example, by reading the three-dimensional CAD engineering drawing data of the speaker, the key stress ring parameters are extracted using computer-aided geometric analysis: the outer diameter R of the surround. out =75.0mm, inner diameter of the surround (i.e., the physical connection boundary between the surround and the cone material) R surround =65.0mm, equivalent radiation radius R of the middle section of the cone mid=40.0mm, radius R at the voice coil frame joint coil =18.0mm; these parameters define the physical coordinate system for the subsequent sensor placement;

[0082] S12. Based on the geometric feature parameters, define the spatial topology of the micro-strain sensor network; the spatial topology includes a first ring array arranged at the connection between the folded ring and the diaphragm, and a second radial array arranged in the middle section of the diaphragm.

[0083] In specific implementation, the above embodiment S12 specifically involves: in order to capture the swing segmentation mode and concentric circle standing wave mode that are prone to occur in the high-frequency band, in R surround On a circumference of 65.0 mm, eight sensor nodes (numbered F1 to F8) with dimensions of 2 mm × 5 mm are evenly arranged at 30-degree central angles, with their major axes arranged radially, specifically for monitoring nonlinear tensile and reverse deformation at the boundary; in R mid Four sensor nodes of the same size (numbered F9 to F12) are evenly arranged at 90-degree intervals on a circumference of 40.0 mm to monitor the longitudinal wave transmission strain of the cone body.

[0084] S13. Using nano-graphene conductive ink, the micro-strain sensor network is directly printed on the back of the speaker diaphragm according to the spatial topology layout through aerosol conformal inkjet printing process.

[0085] Using nano-silver wire ink to print ultra-fine conductive traces, the pins of all sensor nodes are converged to the voice coil skeleton and led out through multi-strand nylon wires woven inside the speaker centering support, generating a low-mass additional signal lead-out path; the total additional mass of this path is strictly controlled to within 0.5% of the speaker's equivalent vibrating mass.

[0086] In specific implementation, the above embodiment S13 is as follows: the nozzle moving speed is set to 10 mm / s, the sheath gas flow rate is 25 sccm, and graphene-based flexible piezoresistive material with a thickness strictly controlled at 5.0 ± 0.2 micrometers is directly sprayed onto the back of the Kevlar diaphragm to form 12 strain resistance grids; ultra-fine conductive traces with a linewidth of 15 micrometers and a thickness of 2 micrometers are printed using nano-silver wire ink; 24 leads converge inward to the center R. coil At 18.0mm, it is seamlessly connected and led out by 24-core special ultra-fine brocade wire (0.05mm diameter per wire) woven inside the centering support through a high-temperature hot pressing process; according to precision balance measurement, the total added mass of this network and lead-out path is only 0.035 grams, accounting for 0.233% of the speaker's own 15.0-gram equivalent vibrating mass, which is lower than the industry's 5% acoustic parameter drift threshold;

[0087] Furthermore, to address the technical challenges of sensor interface delamination and fatigue fracture under long-term, high-amplitude reciprocating motion of the loudspeaker, as an optional embodiment, after the micro-strain sensor network and conductive traces are printed, a 2-micron-thickness parylene-C conformal coating is deposited on the back of the diaphragm using chemical vapor deposition (CVD). This coating has elastic modulus and adhesion, anchoring the sensor network to the Kevlar substrate at the molecular level, ensuring that delamination does not occur under hundreds of millions of alternating stresses.

[0088] Meanwhile, to address the issue that multiple signal lines might alter the mechanical properties of the spider, the 24-core specially made ultra-fine nylon yarn employs a centrally symmetrical, modulus-matched weaving process. Specifically, carbon nanotube (CNT) conductive yarn with an equivalent Young's modulus similar to that of the original spider substrate (such as Conex bulletproof fiber) is used instead of traditional copper wire. These 24 conductive yarns are divided into eight uniform groups and woven in the same direction along the eight centrally symmetrical radial corrugation valleys of the spider. This symmetrical distribution and modulus-matched design ensures that the radial forces of the spider's mechanical compliance (Cms) and mechanical drag (Rms) in the Z-axis direction remain absolutely balanced, thus physically ensuring that the speaker's Thiele-Small parameters do not drift significantly.

[0089] This embodiment successfully constructed a micron-scale strain sensor network and lead-out path on the back of the speaker diaphragm through precise geometric feature extraction and aerosol conformal printing process; it achieved the integration of sensor and diaphragm, and endowed the pure mechanical diaphragm with in-situ, real-time proprioceptive capability without changing the original acoustic physical characteristics of the speaker such as resonant frequency and quality factor, thus overcoming the problem of traditional external sensors destroying sound quality.

[0090] The above embodiment S2 includes the following steps:

[0091] S21. Convert the resistance change rate caused by the deformation of the graphene sensor into a millivolt-level analog voltage signal, and amplify it with a fixed gain to obtain the amplified voltage signal.

[0092] In specific implementation, the above embodiment S21 specifically involves: constructing a multi-channel bridge conversion and amplification circuit; connecting the 24 brocade wires led out in S1 to an external high-frequency signal acquisition front end, the front end containing 12 independent 1 / 4 bridge Wheatstone bridge circuits, converting the minute resistance change rate (ΔR / R, ranging from 0.1% to 2%) generated by the graphene sensor due to deformation into a millivolt-level analog voltage signal, and amplifying it with a fixed gain (e.g., 100 times) through a low-noise instrumentation amplifier (INA);

[0093] S22. Perform digital-to-analog conversion on the amplified voltage signal to obtain the converted voltage signal;

[0094] In specific implementation, the above embodiment S22 specifically involves: inputting the amplified 12-channel analog voltage signals into three parallel 4-channel 24-bit high-precision audio-grade analog-to-digital converters (ADCs); in order to satisfy the Nyquist sampling law and accurately capture high-frequency harmonics of segmented vibration up to 5kHz, the sampling clock frequency of the ADC is strictly set and locked to 192kHz.

[0095] S23. Perform digital filtering and sequence reconstruction on the converted voltage signal to obtain a one-dimensional digital strain time series.

[0096] In specific implementation, the above embodiment S22 is as follows: the digital signal output by the ADC enters the field programmable gate array (FPGA) through the I2S digital audio bus. The FPGA runs a digital high-pass filter with a cutoff frequency of 5Hz to remove the DC offset caused by ambient temperature drift. Finally, in each sampling clock cycle t, the 12 channels of clean dynamic strain data are packaged into a 12×1 column vector, that is, a one-dimensional digital strain time series S(t).

[0097] This embodiment combines a Wheatstone bridge with a 192kHz high-frequency ADC to achieve high-speed, high-fidelity acquisition of diaphragm micro-deformation; it overcomes the shortcomings of conventional industrial strain gauges, which have low sampling rates and are difficult to capture acoustic vibrations, and accurately converts weak physical deformations into lossless digital time series, providing a high-quality real-time data source for subsequent DSP algorithms.

[0098] The above embodiment S3 includes the following steps:

[0099] S31. Define the surface polar coordinate mesh system of the loudspeaker diaphragm; divide the diaphragm surface into M discrete mesh nodes, where the value of M is greater than the number of sensor nodes N contained in the micro-strain sensor network in S1;

[0100] In specific implementation, the above embodiment S31 specifically involves: establishing a polar coordinate mesh model of the 6.5-inch diaphragm inside the DSP; dividing it into 10 rings along the radial direction and 36 angles along the circumference (one node every 10 degrees), generating a total of M=360 mesh nodes; these 360 ​​nodes are sufficient to depict the spatial undulation wavelength of the diaphragm at a maximum frequency of 5kHz with high resolution.

[0101] S32. Obtain the real-time deformation mode orthogonal basis matrix Φ; the dimension of the deformation mode orthogonal basis matrix Φ is M×N; the internal elements of the deformation mode orthogonal basis matrix Φ represent the fixed physical mapping weights between the local strain of N sensor nodes and the three-dimensional spatial displacement of M mesh nodes;

[0102] The acquisition of the real-time deformation mode orthogonal basis matrix Φ in embodiment S32 above includes the following steps:

[0103] S321. Input a logarithmic sweep test signal containing frequency components from 20Hz to 20kHz to the speaker equipped with the intelligent diaphragm hardware carrier in S1.

[0104] S322, Simultaneously record the digital strain calibration sequence S from the outputs of N sensors. cal (t), and simultaneously scanned and recorded the real physical Z-axis displacement sequence D of M grid nodes using a laser Doppler vibrometer. cal (t);

[0105] S323. Construct the least squares objective function with L2 regularization term;

[0106] In specific implementation, the above embodiment S323 is as follows: the least squares objective function J(Φ) = ||D cal (t)-Φ*S cal (t)|| 2 +λ||Φ|| 2 , where λ is the regularization penalty coefficient to prevent overfitting caused by inverting the ill-conditioned matrix;

[0107] It should be noted that, considering the highly nonlinear and spatially complex physical characteristics of loudspeaker split vibration, the construction of the orthogonal basis matrix Φ of the deformation modes in this invention is not a simple point-to-point underdetermined mapping, but rather based on the modal superposition principle in acoustic structure dynamics. As an optional embodiment, during the offline calibration stage, the 360-dimensional real physical displacement sequence D obtained by the laser vibrometer is used... cal (t) Perform principal component analysis (PCA) or intrinsic orthogonal decomposition (POD) to extract the first K (e.g., K=8) dominant orthogonal physical modes of the loudspeaker's segmented vibration (including the first-order axial piston mode, the second-order rocking mode, the third-order concentric circle bending mode, etc.).

[0108] The layout of the 12 micro-strain sensor nodes is specifically optimized for the stress peak / trough positions of the K-order principal modes. Therefore, matrix Φ is essentially a composite transformation matrix that includes strain-modal participation coefficient mapping and modal-3D spatial expansion. It first maps the 12-dimensional discrete strain vector into K-dimensional orthogonal modal participation coefficients, and then generates a 360-dimensional continuous spatial displacement by linearly combining these K orthogonal modal bases. By restricting the solution space to a K-dimensional principal mode subspace that conforms to physical laws, the mathematically underdetermined ill-conditioned dimensionality reduction problem is solved, giving the model the ability to generalize and reconstruct complex combined modes under the excitation of unknown music signals.

[0109] Furthermore, the L-curve criterion is used to optimize the regularization parameter; the residual norm ||D cal (t)-Φ*S cal Using (t)|| as the x-axis and the solution norm ||Φ|| as the y-axis, a double logarithmic curve is plotted. By calculating the maximum curvature point (i.e., the inflection point) of the curve, the optimal balance point is found that can ensure mapping accuracy while avoiding a surge in matrix element values ​​(overfitting). After L-curve analysis of 50 sets of samples with different frequency sweeps, the λ values ​​corresponding to the inflection points are all stably distributed between 0.009 and 0.012. Therefore, in the engineering implementation, the average value of 0.01 is taken as the fixed parameter.

[0110] S324. Solve the deformation mode orthogonal basis matrix Φ that minimizes the least squares objective function J(Φ) by using the singular value decomposition (SVD) algorithm to obtain the real-time deformation mode orthogonal basis matrix;

[0111] S33. At the current sampling time t, reorganize the one-dimensional digital strain time series generated in S2 into a column vector S(t) with dimension N×1;

[0112] S34. Based on the orthogonal basis matrix and column vector of the deformation modes, the column vector D(t) of dimension M×1 is calculated in real time using the matrix multiplication formula D(t)=Φ×S(t) to obtain the continuous three-dimensional micro-deformation topology matrix; the column vector D(t) represents the vertical physical displacement of M grid nodes on the diaphragm relative to the stationary plane at the current time t, and the column vector D(t) constitutes the continuous three-dimensional micro-deformation topology matrix.

[0113] In specific implementation, the above embodiment S34 is as follows: at a certain moment t when the user plays music, the DSP reads the data S(t) from 12 sensors; for example, S1 on the left shows a strain of +150 microstrain (tension), and S5 on the right shows a strain of -120 microstrain (compression); the DSP performs hardware multiplication and addition operations of D(t)=Φ×S(t); the calculated 360-dimensional vector D(t) clearly shows that the left half of the diaphragm has shifted upward by 0.5mm and the right half has shifted downward by -0.4mm; the system successfully restores the limited local sensor data into a three-dimensional panoramic topology map of the diaphragm "swaying left and right";

[0114] The above embodiments establish the mathematical mapping relationship between local strain and global displacement through rigorous offline laser calibration and SVD matrix solution; in real-time operation, only a single matrix multiplication with very low computing power is required to expand 12 discrete data points into a high-definition three-dimensional deformation topology of 360 points; eliminating the real-time dependence on expensive optical instruments, enabling consumer-grade audio equipment to have laboratory-level observation capabilities.

[0115] The above embodiment S4 includes the following steps:

[0116] S41. Based on the one-dimensional digital strain time series in S2, the strain cycle amplitude experienced by each sensor node within a preset sliding time window is statistically analyzed using a rainflow counting algorithm. and the corresponding number of real-time loop occurrences n k ;

[0117] S42, Call the speaker diaphragm material pre-written into memory (Strain-Life) Fatigue Curve Equation Calculate the amplitude during strain cycles The number of limiting cycles at which the material undergoes irreversible fracture; where N k For the amplitude of strain cycles The limiting cycle number at which the material undergoes irreversible fracture; a and b are material constants;

[0118] Furthermore, before the loudspeakers leave the factory, standard tensile and compressive fatigue tests were conducted on the Kevlar composite fiber material used in this batch, according to the fatigue performance testing standards for polymer-based composite materials, on a material testing machine to obtain different strain cycle amplitudes. The number of fracture cycles N k The test data will be plotted in a logarithmic coordinate system (log -log N k Linear regression fitting is performed in the sample; the intercept of the fitted line is obtained by least squares method, which is parameter a (measured as 25.5 for this batch of materials), and the slope of the line is parameter b (measured as 6.2 for this batch of materials).

[0119] S43, Based on the real-time cycle occurrence count and strain cycle amplitude S k The limiting number of cycles at which the material undergoes irreversible fracture is determined; using the Palmgren-Miner linear cumulative damage theory, the damage increment ΔD=∑(n) within the current sliding time window is calculated. k / N k );

[0120] The damage increment within the current sliding time window is summed to obtain the updated real-time cumulative fatigue damage D of the current node. total new =D total old +ΔD; where D total old The cumulative fatigue damage before the update; D total The range of its value is [0, 1];

[0121] In specific implementation, the above embodiment S43 specifically involves: the DSP continuously running a rainflow counter in the background; after the user listens continuously for 1 hour (sliding window), the amplitude experienced by the loop S1 node is statistically analyzed. The large cycle of k=1000 micro-strain n k =5000 times; Substitute The strain-life fatigue curve equation is used to calculate the ultimate life N at that strain. k = =10 (25.5−18.6) =10 6.9 ≈7,943,282 times; the damage increment ΔD was calculated using the Palmgren-Miner formula: 5000 / 7943282 = 0.00063; this increment was added to the historical damage values ​​recorded in memory to obtain the updated D. total new ;

[0122] S44, the real-time cumulative fatigue damage degree D total new The real-time output to step S5 serves as a safety constraint factor for the adaptive filter; simultaneously, when the cumulative fatigue damage degree corresponding to any sensor node is ≥ the safety threshold of 0.85, it is determined that the region is about to undergo physical tearing, and the dynamic compression attenuation command is immediately triggered and generated, and output to step S5 to limit the audio input.

[0123] The acquisition of the dynamic compression attenuation command in S44 includes the following steps:

[0124] S441, When the real-time cumulative fatigue damage degree D of a certain node total new When the safety threshold of 0.85 is reached, the system backtracks the rainflow counting records collected by the rainflow counting algorithm in S41 and extracts the central excitation frequency F that caused the maximum strain cycle amplitude. c ;

[0125] S442, Control the digital frequency divider to activate at the center frequency F c A notch attenuation band with a bandwidth of one octave is generated at the location; the attenuation depth G of the notch attenuation band ain The quantitative calculation formula is G ain (dB) = −20 × (D) total new -0.8) / 0.2;

[0126] Synchronously extract the continuous three-dimensional micro-deformation topology matrix output at trigger time S3, extract the three-dimensional coordinate system of the grid node where the deformation extreme value occurs, and correlate it with the central excitation frequency F. cThe accumulated fatigue damage was packaged to generate an offline structural optimization dataset. The structural optimization dataset visually shows that under 60Hz drive, severe asymmetric inward wrinkling and collapse occurred on the left side of the diaphragm ring (S1 area). The dataset was uploaded to the cloud server through the Internet of Things interface to complete the digital twin data closed loop. Based on this clear quantitative data, the R&D engineers specifically thickened the angle area corresponding to the S1 ring by 0.2mm or changed the damping adhesive formula in the mold design of the next generation product, thereby realizing the physical structure closed-loop optimization based on real user big data.

[0127] This embodiment transforms the abstract speaker aging process into precise quantitative indicators through rainflow counting and Miner's cumulative damage theory, and generates dynamic compression and limiting instructions accordingly. It changes the traditional passive waiting situation of audio equipment to be damaged, and realizes predictive health management. By generating notch attenuation to implement protection in specific fatigue frequency bands, it effectively curbs the propagation of fatigue cracks while preserving the overall sound quality to the greatest extent, extends the physical service life of the speaker, and provides optimized data for the next generation of products.

[0128] The above embodiment S5 includes the following steps:

[0129] S51: Receive the original audio electrical signal. When this signal passes through the digital frequency divider, it undergoes pre-notch filtering under the control of the dynamic compression and attenuation command generated by S4 to reduce the energy of dangerous frequency bands, resulting in a safe audio electrical signal x. safe (t);

[0130] S52. Construct an ideal rigid body piston motion reference model; collect the input current generated by the speaker voice coil based on the aforementioned safe audio electrical signal in real time, and combine it with the force coefficient B. l Calculate the ideal axial displacement scalar Z of the voice coil at the current moment. ideal (t); Assign the ideal axial displacement scalar to the M mesh nodes defined in S3 with equal magnitude to generate an ideal displacement matrix D of dimension M×1. ideal (t);

[0131] In specific implementation, the above embodiment S52 is as follows: multiply the real-time input current by the force coefficient B1 to obtain the electromagnetic driving force, and then input the driving force into the second-order linear mechanical impedance transfer function composed of the equivalent vibrating mass of the loudspeaker, mechanical damping and support compliance to obtain the ideal axial displacement scalar.

[0132] Furthermore, the ideal rigid piston motion reference model does not refer to a speaker maintaining absolute planar rigid translation across all frequency bands; the reference model calculates the ideal axial displacement scalar D. idealFollowing (t), a spatial low-pass roll-off filter function (such as the Bessel roll-off model) matching the speaker diameter will be further connected in series; this roll-off model simulates the natural, symmetrical mechanical decoupling phenomenon of the cone at high frequencies; therefore, the generated ideal displacement matrix Z ideal (t) represents an ideal acoustic radiation state that is smooth, symmetrical, and free of parasitic resonances; Z ideal The spatial residual matrix E is obtained by subtracting the actual deformation D(t) from the spatial residual matrix E(t). space (t), eliminating natural radial high-frequency attenuation, and extracting asymmetric rocking modes, local standing waves, and high-order harmonic distortions that cause severe hearing degradation; through the above steps, the accuracy of the error scalar signal e(t) in physical acoustics is ensured;

[0133] The second-order linear mechanical impedance transfer function: Where m represents the equivalent vibrating mass of the loudspeaker; R m Indicates mechanical damping; C m Indicates compliance; i(t) represents the real-time acquired input current of the speaker voice coil; d represents differentiation;

[0134] S53. Combine the continuous three-dimensional microscopic deformation topology matrix D(t) in S3 with the ideal displacement matrix D ideal (t) is subtracted element by element to obtain a spatial residual matrix E of dimension M×1. space (t)=D(t)−D ideal (t);

[0135] S54, regarding the spatial residual matrix E space (t) Perform area integral weighted summation based on acoustic radiation efficiency to extract the spatial nonlinear error scalar signal e(t) in single scalar form;

[0136] In specific implementation, the above embodiment S54 is as follows: the specific quantization calculation formula for the spatial nonlinear error scalar signal e(t) in S54 is: Among them, E space (i,t) represents the displacement residual value of the i-th grid node in the spatial residual matrix at time t; r i η is the physical radial distance from the i-th grid node to the center of the diaphragm; Δr is the radial width of the grid; η is the acoustic radiation impedance coupling coefficient, used to compensate for the directivity gain of high-frequency sound waves;

[0137] Furthermore, a finite element model of the 6.5-inch loudspeaker placed on a standard baffle was established. Within the 2kHz to 5kHz frequency band (segmenting the main vibration frequency range), the coupling efficiency between the actual normal vibration velocity at each node on the diaphragm surface and the far-field sound pressure was calculated. Simulation results show that, due to the cone-shaped geometric diffusion effect of the cone, its high-frequency sound radiation impedance real part (radiation resistance) has an average gain coefficient of approximately 1.2 times compared to the piston radiation impedance of an ideal infinite plane. Therefore, η was set to 1.2 to ensure that the error integral accurately reflects the true sound power distortion.

[0138] This embodiment reduces the complex three-dimensional topological matrix into a single error scalar by differential comparison of the rigid body reference model and area integral based on acoustic radiation efficiency; it transforms the spatial physical displacement residual of the pure mechanical dimension into the nonlinear sound pressure error of the acoustic dimension. Without adding external acoustic sensors such as microphones, it quantifies the actual contribution of segmented vibration to auditory distortion, providing a target error function with clear direction and low computational cost for subsequent adaptive acoustic compensation algorithms.

[0139] S6. Using the accumulated fatigue damage as a boundary constraint, based on the spatial nonlinear error scalar signal and the safety audio electrical signal, an inverse acoustic compensation signal is dynamically generated through an adaptive nonlinear filter and superimposed on the safety audio electrical signal to drive the speaker.

[0140] S61. Construct a second-order Volterra adaptive filter in a digital signal processor; convert the security audio electrical signal x generated in step S51 into a digital signal processor. safe (t) After filtering by the secondary acoustic path transfer function of the loudspeaker, the reference signal X is obtained. 1 safe (n);

[0141] S62. Based on the reference signal and the spatial nonlinear error scalar signal e(n), and introducing the real-time cumulative fatigue damage degree Dtotal generated in S4. new As a safety penalty variable, the adaptive step size factor μ, which includes fatigue constraints, is dynamically calculated;

[0142] In specific implementation, the above embodiment S62 specifically involves: setting the quantization formula μ=[α / (||X 1 safe (n)|| 2 +β)]×(1-D total ) 2 ; where α is set to 0.005 and β is set to 1e-4; by introducing (1-D total ) 2 This feature allows the filter to automatically reduce the compensation step size when the diaphragm fatigue increases, thus avoiding the risk of mechanical breakage caused by forced compensation.

[0143] S63. Based on the filter-x minimum mean square error (FxLMS) algorithm, and introducing the cumulative fatigue damage degree D generated in S4. total As a safety penalty variable, the optimal filter weight vector W at the current sampling time is obtained through the restricted weight update iterative formula of the filter. n+1 ;

[0144] In specific implementation, the above embodiment S63 is specifically as follows: W n+1 =W n +μ×e(n)×X 1 safe (n); where W n Let X be the filter weight vector at the current time; e(n) is the value in step S54; 1 safe (n) is the reference signal after the safety audio electrical signal is filtered by the speaker secondary acoustic path transfer function;

[0145] S64. Using the optimal filter weight vector W n+1 For the current secure audio electrical signal x safe (t) is subjected to Volterra convolution operation to calculate the reverse acoustic compensation signal y(t); the reverse acoustic compensation signal y(t) is phase-reversed and superimposed on the safety audio electrical signal x. safe In (t), the output is sent to the power amplifier to drive the speaker;

[0146] In specific implementation, the above embodiment S6 specifically refers to: when the speaker is in a brand new state (D total When =0), (1-0) 2 =1, the step size factor μ remains at its maximum value, the filter weights converge rapidly, and the generated inverse acoustic compensation signal y(t) has the largest amplitude. At this time, the driving current containing y(t) flows through the voice coil. This inverse acoustic compensation signal y(t) is essentially a broadband predistortion energy sequence. Since complex spatial segmentation vibrations are often excited by standing waves formed by the reflection of a fundamental wave of a specific frequency and its intermodulation products at the diaphragm boundary, the predistortion electromagnetic force generated by the voice coil does not directly pinch the diaphragm in space, but injects canceling energy with the same amplitude and opposite phase as the distortion energy at the excitation source in the time domain. By changing the boundary conditions of the excitation source, the energy excitation mechanism of a specific spatial mode (such as the rocking mode) is destroyed, thereby suppressing the occurrence of segmentation vibration at the physical source. The intermodulation distortion (IMD) at 3kHz high power is reduced from 8.5% to 0.9%.

[0147] When S4 determines that the speaker is severely aged (e.g., D), total When (=0.85), (1-0.85) 2=0.0225, the restricted step size factor μ is automatically reduced to the original 2.25%; at this time, the weight update is extremely slow, the amplitude of the calculated compensation signal y(t) is forcibly suppressed to a low level, the system no longer pulls the aging diaphragm to exchange for sound quality, and avoids the diaphragm tearing on the spot;

[0148] This embodiment constructs a constrained acoustic compensation mechanism by embedding mechanical fatigue into the step size update formula of the Volterra filter. It solves the defect of traditional active distortion compensation technology that blindly pursues sound quality and accelerates device damage. The system can establish a mechanism of fault-oriented safety and smooth degradation processing between acoustic fidelity and mechanical physical life based on the real-time health status of the diaphragm. When the speaker is in a healthy state, the system fully releases computing resources to pursue the ultimate in extremely low distortion. When mechanical fatigue approaches the physical limit, the system actively and smoothly reduces the compensation gain, sacrificing some sound quality performance under extreme conditions in exchange for the device to avoid catastrophic physical tearing. This dynamic trade-off strategy solves the problem of traditional audio equipment being prone to distortion or burnout under extreme conditions, and realizes intelligent safety drive.

[0149] Please see Figure 2 A real-time deformation monitoring system for a strain sensor embedded in a loudspeaker diaphragm is provided, which implements a method for real-time monitoring of the deformation of the strain sensor embedded in the loudspeaker diaphragm. The system includes an intelligent diaphragm physical module, a high-frequency signal acquisition module, a DSP reconstruction module, a fatigue calculation and protection generation module, and a constrained acoustic compensation execution module, wherein:

[0150] The aforementioned intelligent diaphragm physical module is used to perform step S1 and provides a hardware carrier with a micro-strain sensor network and a low-mass additional signal extraction path.

[0151] The aforementioned high-frequency signal acquisition module is used to execute step S2, connect the signal output path, and convert the analog weak electrical signal into a one-dimensional digital strain time series.

[0152] The aforementioned DSP reconstruction module is used to execute step S3, receiving a one-dimensional digital strain time series to reconstruct a continuous three-dimensional microscopic deformation topology matrix.

[0153] The fatigue calculation and protection generation module described above is used to execute step S4, calculate the cumulative fatigue damage, generate dynamic pressure limit attenuation instructions, and upload the offline structural optimization dataset to the cloud.

[0154] The aforementioned restricted acoustic compensation execution module is used to execute steps S5 and S6, receive the dynamic compression attenuation command issued by S4 to generate a safe audio electrical signal, and simultaneously receive the cumulative fatigue damage degree issued by S4 as the calculation boundary. Based on the three-dimensional topology error, it runs the restricted Volterra filter and outputs a safe reverse acoustic compensation signal.

[0155] This invention upgrades traditional blind-driven loudspeakers into intelligent transducer systems with self-awareness and self-protection capabilities. By integrating a flexible sensing network into the diaphragm without damage, it overcomes the offline limitations of traditional optical vibration instruments, converting invisible physical segmentation vibrations into three-dimensional digital topology in real time. It extracts spatial nonlinear errors, enabling the adaptive acoustic compensation algorithm to eliminate high-frequency intermodulation distortion and achieve acoustic fidelity. It constructs a two-way constraint closed loop between sound quality and lifespan, embedding the real-time calculated mechanical fatigue damage degree as a bottom-level constraint factor into the control algorithm. When the diaphragm ages or faces the risk of tearing, the system automatically limits the compensation intensity and performs frequency band compression, solving the engineering problem of prioritizing sound quality versus lifespan under high-power drive, and achieving an adaptive balance between high-fidelity sound quality and extreme physical lifespan.

[0156] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0157] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for real-time monitoring of deformation of a strain sensor embedded in a loudspeaker diaphragm, characterized in that, Includes the following steps: S1. Construct a network of miniature strain sensors made of flexible piezoresistive material and establish a low-mass-added signal extraction path along the non-sound-producing physical structure of the loudspeaker; S2. When the loudspeaker is driven to produce sound by an audio electrical signal, the analog resistance change signal output by each node in the micro strain sensor network is collected based on the low-mass-added signal extraction path, and a one-dimensional digital strain time series is obtained by analog-to-digital conversion. S3. Spatial mapping calculation is performed on the one-dimensional digital strain time series using the orthogonal basis matrix of deformation modes to obtain a continuous three-dimensional micro-deformation topology matrix. S4. Extract the extreme values ​​of the one-dimensional digital strain time series and calculate the current cumulative fatigue damage degree; when the cumulative fatigue damage degree reaches the damage degree threshold, locate the fatigue occurrence area and generate a dynamic pressure limit attenuation command. S5. The continuous three-dimensional micro-deformation topology matrix and the ideal rigid piston motion reference model of the loudspeaker diaphragm are compared by difference and weighted by area integral to obtain a spatial nonlinear error scalar signal; the original audio electrical signal is pre-filtered by a dynamic compression attenuation command to obtain a safe audio electrical signal. S6. Using the accumulated fatigue damage as a boundary constraint, based on the spatial nonlinear error scalar signal and the safety audio electrical signal, an inverse acoustic compensation signal is dynamically generated through an adaptive nonlinear filter and superimposed on the safety audio electrical signal to drive the speaker.

2. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S1 includes the following steps: S11. Extract the geometric feature parameters of the loudspeaker diaphragm; the geometric feature parameters include the outer diameter of the surround, the inner diameter of the surround, the radius of the middle section of the cone, and the radius of the joint between the voice coil and the skeleton; S12. Based on the geometric feature parameters, define the spatial topology of the micro-strain sensor network; the spatial topology includes a first ring array arranged at the connection between the folded ring and the diaphragm, and a second radial array arranged in the middle section of the diaphragm. S13. Using nano-graphene conductive ink, the micro-strain sensor network is directly printed on the back of the speaker diaphragm according to the spatial topology layout through aerosol conformal inkjet printing process. Using nano-silver wire ink to print ultra-fine conductive traces, the pins of all sensor nodes are converged to the voice coil skeleton and led out through multi-strand nylon wires woven inside the speaker centering support, generating a low-quality additional signal output path.

3. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S2 includes the following steps: S21. Convert the resistance change rate caused by the deformation of the graphene sensor into a millivolt-level analog voltage signal, and amplify it with a fixed gain to obtain the amplified voltage signal. S22. Perform digital-to-analog conversion on the amplified voltage signal to obtain the converted voltage signal; S23. Perform digital filtering and sequence reconstruction on the converted voltage signal to obtain a one-dimensional digital strain time series.

4. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S3 includes the following steps: S31. Define the polar coordinate grid system of the loudspeaker diaphragm surface; divide the diaphragm surface into M discrete grid nodes, where the value of M is greater than the number of sensor nodes N contained in the micro-strain sensor network in S1; S32. Obtain the real-time deformation mode orthogonal basis matrix; the dimension of the deformation mode orthogonal basis matrix is ​​M×N; the internal elements of the deformation mode orthogonal basis matrix represent the fixed physical mapping weights between the local strain of N sensor nodes and the three-dimensional spatial displacement of M mesh nodes; S33. At the current sampling time, reorganize the one-dimensional digital strain time series generated in S2 into a column vector with dimension N×1; S34. Based on the orthogonal basis matrix and column vector of the deformation mode, the column vector of dimension M×1 is calculated in real time using the matrix multiplication formula to obtain the continuous three-dimensional micro-deformation topology matrix.

5. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 4, characterized in that, The real-time deformation mode orthogonal basis matrix in S32 includes the following steps: S321. Input a logarithmic sweep frequency test signal to the speaker equipped with the intelligent diaphragm hardware carrier in S1; S322. Simultaneously record the digital strain calibration sequence output by N sensors, and simultaneously scan and record the real physical Z-axis displacement sequence of M grid nodes using a laser Doppler vibrometer. S323. Construct the least squares objective function with L2 regularization term; S324. Solve the deformation mode orthogonal basis matrix that minimizes the objective function of the least squares method by using the singular value decomposition algorithm, and obtain the real-time deformation mode orthogonal basis matrix.

6. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S4 includes the following steps: S41. Based on the one-dimensional digital strain time series in S2, the rainflow counting algorithm is used to count the strain cycle amplitude and the corresponding real-time cycle occurrence number of each sensor node within a preset sliding time window. S42, Call the speaker diaphragm material pre-written into memory The fatigue life curve equation calculates the limit number of cycles at which the material will undergo irreversible fracture under the strain cycle amplitude. S43. The limit number of cycles for irreversible fracture of the material based on the real-time cycle occurrence number and strain cycle amplitude; calculate the damage increment within the current sliding time window using the Palmgren-Miner linear cumulative damage theory; The damage increment within the current sliding time window is accumulated to obtain the updated real-time cumulative fatigue damage of the current node; S44. When the cumulative fatigue damage degree corresponding to any sensor node is greater than or equal to the safety threshold, it is determined that the sensor node area will experience physical tearing, and the dynamic pressure limit attenuation command is immediately triggered and generated.

7. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 6, characterized in that, The acquisition of the dynamic compression attenuation command in S44 includes the following steps: S441. When the real-time cumulative fatigue damage of a certain node reaches the safety threshold, the system backtracks the rain flow counting records statistically obtained by the rain flow counting algorithm in S41 and extracts the center excitation frequency that causes the maximum strain cycle amplitude. S442 controls the digital frequency divider to generate a notch attenuation band with a bandwidth of 1 octave at the center excitation frequency, thus obtaining a dynamic compression attenuation command.

8. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S5 includes the following steps: S51: Receive the original audio electrical signal. When the signal passes through the digital frequency divider, it is controlled by the dynamic compression and attenuation command generated by S4 to perform pre-notch processing, reduce the energy of dangerous frequency bands, and obtain a safe audio electrical signal. S52. Construct an ideal rigid body piston motion reference model; collect the input current generated by the speaker voice coil based on the safe audio electrical signal in real time, and calculate the ideal axial displacement scalar of the voice coil at the current moment by combining the force coefficient; assign the ideal axial displacement scalar to the M grid nodes defined in S3 with equal amplitude to generate an ideal displacement matrix of dimension M×1. S53. Subtract the continuous three-dimensional micro-deformation topology matrix in S3 from the ideal displacement matrix element by element to obtain a spatial residual matrix with dimension M×1. S54. Perform area integral weighted summation based on acoustic radiation efficiency on the spatial residual matrix to extract the spatial nonlinear error scalar signal in single scalar form.

9. The method for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm according to claim 1, characterized in that, S6 includes the following steps: S61. Construct a second-order Volterra adaptive filter in a digital signal processor; after filtering the safety audio electrical signal generated in step S51 through the secondary acoustic path transfer function of the loudspeaker, obtain the reference signal; S62. Based on the reference signal and the spatial nonlinear error scalar signal, and introducing the real-time cumulative fatigue damage degree generated in S4 as a safety penalty variable, dynamically calculate the adaptive step size factor containing fatigue constraints. S63. Based on the filter-x minimum mean square error algorithm, and introducing the cumulative fatigue damage degree generated in S4 as a safety penalty variable, the optimal filter weight vector at the current sampling time is obtained through the restricted weight update iterative formula of the filter. S64. Perform Volterra convolution operation on the current safety audio electrical signal using the optimal filter weight vector to calculate the reverse acoustic compensation signal; after phase reversal of the reverse acoustic compensation signal, superimpose it onto the safety audio electrical signal and output it to the power amplifier to drive the speaker.

10. A real-time deformation monitoring system for a strain sensor embedded in a loudspeaker diaphragm, characterized in that, The system for real-time deformation monitoring of a strain sensor embedded in a loudspeaker diaphragm as described in any one of claims 1-9 comprises: Intelligent diaphragm physics module: used to perform the S1 step, providing a hardware carrier with a micro strain sensor network and a low-mass additional signal extraction path; High-frequency signal acquisition module: used to execute step S2, connect the signal output path, and convert the analog weak electrical signal into a one-dimensional digital strain time series; DSP Reconstruction Module: Used to execute the S3 step, receiving a one-dimensional digital strain time series to reconstruct a continuous three-dimensional microscopic deformation topology matrix; Fatigue Calculation and Protection Generation Module: This module is used to execute step S4, calculate the cumulative fatigue damage, generate dynamic compression attenuation instructions, and upload the offline structural optimization dataset to the cloud. The constrained acoustic compensation execution module is used to execute steps S5 and S6, receive the dynamic compression attenuation command issued by S4 to generate a safe audio electrical signal, and simultaneously receive the cumulative fatigue damage degree issued by S4 as the calculation boundary. Based on the three-dimensional topology error, it runs the constrained Volterra filter and outputs a safe reverse acoustic compensation signal.

Images