Silver coating adhesive force detection method based on multi-frequency impedance imaging

Abstract

The invention relates to the technical field of material detection, in particular to a silver coating adhesive force detection method based on multi-frequency impedance imaging, which comprises the following steps: applying a dual-mode multi-frequency excitation signal to the surface of a silver coating through an attached microelectrode array, synchronously acquiring voltage-current response data, and generating a dual-mode multi-frequency impedance original data set; performing time-varying impedance phase-locked separation on the dual-mode multi-frequency impedance original data set, extracting a dynamic impedance component of a coating-matrix interface region, and reconstructing a three-dimensional sub-interface impedance distribution diagram; and outputting an adhesive force quantized value through an oscillation frequency-adhesive force conversion model according to the spatial oscillation frequency characteristics of the impedance phase in the three-dimensional sub-interface impedance distribution diagram. According to the invention, a double-phase-locked loop system is adopted to realize the time-varying component demodulation of the impedance response of the silver coating-matrix interface, the low-frequency dynamic impedance signal modulated by silver ion migration is effectively extracted, and the weak change of the interface bonding state is sensitively reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of material detection, and in particular to a method for detecting the adhesion of a silver plating layer based on multi-frequency impedance imaging. Background Art

[0002] As a functional metal coating with excellent conductivity, oxidation resistance and welding properties, silver plating is widely used in key fields such as microelectronic devices, precision connectors, high-frequency communication components, aviation cables, and medical conductive interfaces. In actual service, it not only needs to meet electrical performance requirements, but also places extremely high demands on the adhesion between the coating and the substrate. Once the adhesion is poor, it is easy to lead to increased contact resistance, local heat accumulation, coating peeling and even electrical failure, posing a serious threat to system safety and long-term reliability.

[0003] Current methods for testing the adhesion of silver coatings in the industry primarily include mechanical stripping, scanning ultrasound, eddy current imaging, or salt spray testing. It is worth noting that complex microscopic failure mechanisms often exist in the silver coating during the interfacial bonding process, such as silver ion migration, local stress concentration, dendrite growth, or stress peeling of the coating. These problems do not initially manifest as obvious macroscopic delamination, but can cause non-uniform fluctuations in interfacial impedance at the micron or submicron scale, ultimately affecting adhesion stability. Therefore, the ability to dynamically identify and spatially locate these early microscopic interface disturbances will significantly improve the foresight and accuracy of adhesion quality control. Summary of the Invention

[0004] The present invention provides a silver-plated layer adhesion detection method based on multi-frequency impedance imaging, and a high-resolution adhesion detection method based on multi-frequency excitation, spatial imaging and model conversion, so as to realize the accurate identification, quantitative evaluation and real-time monitoring of the adhesion state of the silver-plated layer, thereby fundamentally improving the reliability assessment level of the material interface.

[0005] A method for detecting the adhesion of a silver coating based on multi-frequency impedance imaging comprises the following steps: S1, applies dual-mode multi-frequency excitation signals to the silver-plated surface through an attached microelectrode array, synchronously collects voltage-current response data, and generates a dual-mode multi-frequency impedance raw data set; S2, performing time-varying impedance phase-locked separation on the dual-mode multi-frequency impedance raw data set, extracting the dynamic impedance component of the coating-substrate interface region, and reconstructing a three-dimensional sub-interface impedance distribution map; S3, outputting a quantized adhesion value through an oscillation frequency-adhesion conversion model according to the spatial oscillation frequency characteristics of the impedance phase in the three-dimensional sub-interface impedance distribution diagram.

[0006] Optionally, the dual mode includes a current mode and a voltage mode, and the application of the dual mode multi-frequency excitation signal includes applying a current mode excitation signal through the annular electrode group of the attached microelectrode array, and simultaneously applying a voltage mode excitation signal through the dot matrix electrode group.

[0007] Optionally, the current mode excitation signal and the voltage mode excitation signal are injected in orthogonal timing, the current mode excitation is started at the zero crossing point of the voltage mode excitation, and the duration is one quarter of the voltage cycle; During current mode stimulation, voltage response data of the array electrode group were collected synchronously; During voltage-mode excitation, current response data of the ring electrode set are acquired synchronously.

[0008] Optionally, S1 also includes associating the voltage response data in the current mode with the corresponding frequency points to generate a first impedance subset; associating the current response data in the voltage mode with the corresponding frequency points to generate a second impedance subset; and merging the first impedance subset and the second impedance subset to form a dual-mode multi-frequency impedance original data set.

[0009] Optionally, the time-varying impedance phase-locked separation includes a dual phase-locked loop configuration, including a first phase-locked loop and a second phase-locked loop, wherein; Setting a first phase-locked loop to lock the phase of a fundamental frequency excitation signal, wherein the fundamental frequency excitation signal is the swept frequency excitation signal in S1; The second phase-locked loop is set to lock the phase of a low characteristic frequency, where the low characteristic frequency is related to the migration rate of silver ions at the coating-substrate interface.

[0010] Optionally, the extraction of the dynamic impedance component of the coating-substrate interface region includes inputting the dual-mode multi-frequency impedance raw data set into a dual phase-locked loop to extract the dynamic impedance component modulated by the characteristic frequency. ,right Perform time-frequency decoupling, including: When the frequency point f is fixed, the time domain oscillation component is obtained ; At a fixed time point t, obtain the frequency domain response component .

[0011] Optionally, reconstructing the three-dimensional sub-interface impedance distribution map includes: Based on the frequency domain response components , the compressed sensing algorithm is used to reconstruct the xy plane impedance distribution; Based on the time domain oscillation component , calculate the impedance gradient in the depth direction by phase delay inversion; The plane distribution and depth gradient are integrated to generate a three-dimensional sub-interface impedance distribution map.

[0012] Optionally, the extraction of the spatial oscillation frequency characteristics includes selecting a preset depth layer in the three-dimensional sub-interface impedance distribution map, extracting the impedance phase distribution of the preset depth layer parallel to the coating surface, and generating a two-dimensional phase matrix; performing a two-dimensional fast Fourier transform on the two-dimensional phase matrix to obtain its spectrum energy map; and extracting the main oscillation frequency component with the largest energy in the spectrum energy map as the spatial characteristic frequency feature.

[0013] Optionally, S3 includes calculating an energy share coefficient based on the ratio of the spectral energy of the main oscillation frequency component to the total spectral energy, and inputting the main oscillation frequency component and its corresponding energy share coefficient into a preset oscillation frequency-adhesion conversion model to output an adhesion quantization value.

[0014] Optionally, S3 also includes classifying adhesion levels based on the adhesion quantization value, comparing the adhesion quantization value with a preset level judgment threshold, and when the adhesion quantization value is greater than or equal to a first threshold, determining it as level A adhesion; when the adhesion quantization value is between the first threshold and a second threshold, determining it as level B adhesion; when the adhesion quantization value is less than the second threshold, determining it as level C adhesion.

[0015] Beneficial effects of the present invention: 1. The present invention constructs a dual-mode multi-frequency excitation mechanism and adopts a dual phase-locked loop system to demodulate the time-varying component of the silver-plated layer-substrate interface impedance response, effectively extracting the low-frequency dynamic impedance signal modulated by silver ion migration. Compared with traditional static resistance methods or low-frequency electrochemical analysis, this method can more sensitively reflect subtle changes in the interface bonding state and is suitable for identifying hidden adhesion defects such as early cracks, fine dendrites, and micro-delamination. The present invention reduces detection blind spots when identifying abnormal micro-area adhesion.

[0016] 2. Based on the reconstruction of three-dimensional sub-interface impedance, the present invention proposes a spatial oscillation feature extraction method with the main oscillation frequency and spectral energy ratio as dual factors, and constructs an adhesion quantification model based on an exponential-logarithmic combination to achieve a mapping from physical signal characteristics to structural adhesion performance. This model fully considers the impact of interface roughness on the increase in main frequency and the contribution of damage range to the decrease in adhesion strength, and exhibits stable discrimination capabilities on a variety of substrates.

[0017] 3. The present invention performs two-dimensional Fourier transform and main frequency identification on the impedance phase distribution of the selected depth layer. Only local phase data is required to quickly complete the micro-area adhesion quantification and classification. Combined with the set multi-level threshold judgment rules, multi-level adhesion level information can be output. This is suitable for the rapid screening and process closed-loop control of batch silver coating products. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 Schematic diagram of the detection method flow in an embodiment of the present invention; Figure 2 Schematic diagram of the dual phase-locked loop working mechanism of an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art may also implement some known technologies in other alternative ways. The accompanying drawings are only for describing the embodiments in more detail and are not intended to limit the present invention in any specific way.

[0021] like Figure 1-Figure 2 As shown, a method for detecting the adhesion of a silver coating based on multi-frequency impedance imaging comprises the following steps: S1, a dual-mode multi-frequency excitation signal is applied to the silver-plated surface through an attached microelectrode array, and the voltage-current response data are simultaneously collected to generate a dual-mode multi-frequency impedance raw data set.

[0022] S11, excitation application stage: applying current mode excitation signals and voltage mode excitation signals respectively through the ring electrode group and the dot matrix electrode group in the attached microelectrode array, specifically including: S111, the current mode excitation signal is a swept frequency AC current signal with a frequency range of 0.1–1 kHz, and its current density is controlled at 0.05–0.1 mA / mm²; Low-frequency current excitation signals have a greater electromagnetic penetration depth (based on the current density distribution model in the conductor-electrolyte system), can penetrate the surface of the silver plating layer, and act on the bonding interface between silver and the substrate (such as copper and aluminum). This type of low-frequency excitation can more sensitively reflect the polarization process of charge at the interface trap and the response of microstructural voids (such as the capacitance-resistance change in the adhesion weakening area), and is suitable for detecting problems such as bond strength degradation.

[0023] S112, the voltage mode excitation signal is a swept frequency AC voltage signal with a frequency range of 10–100 kHz and a voltage amplitude controlled at 0.3–0.5 V; High-frequency AC voltage mode excitation has a skin effect: the electrical signal tends to concentrate and propagate on the surface of the conductor, with a penetration depth of less than 5μm. It is used to capture interference features such as the fine structure of the silver layer surface and electrochemical reaction noise (such as bubbles and oxide films), while also suppressing their contamination of low-frequency data acquisition. High-frequency voltage excitation can also quickly and stably input energy, avoiding errors caused by uneven current distribution.

[0024] S113 , the two excitation signals are injected and controlled in quadrature sequence: the current mode excitation is triggered at the zero-crossing point of the voltage mode excitation, and the duration is 1 / 4 of the voltage mode excitation period.

[0025] In the dual-mode incentive structure: Dot-matrix electrodes are used for high-frequency voltage excitation: The plum blossom-shaped dot-matrix electrodes are evenly distributed and densely arranged, providing fine voltage injection control in space and are suitable for applying high-frequency signals with fine amplitude control. Dot-shaped excitation forms a local electric field on the surface of the silver layer, which can produce a response distribution with high spatial resolution.

[0026] Ring electrodes are used for low-frequency current excitation: the circular ring structure naturally forms an enclosing uniform current injection path, which can stimulate stable current density over a large range. Its structure is conducive to achieving uniform energy distribution on a large-area contact interface and improving the effectiveness of low-frequency excitation penetration.

[0027] S12, data acquisition: synchronous response data acquisition in two excitation modes: During current mode stimulation, voltage response data of the array electrode group were collected synchronously; During voltage-mode excitation, current response data of the ring electrode set are acquired synchronously.

[0028] In the present invention, dual-mode excitation refers to the use of two different electrical signal injection methods to detect the silver-plated layer under test: Current mode excitation: injecting a known AC current into the silver layer , measuring the voltage response generated , and thus calculate the impedance ; Voltage mode excitation: Apply a known AC voltage to the silver layer , measuring the current response generated , and thus calculate the impedance ; These two modes are dual to each other in electrical principles, that is, they extract the system impedance characteristics from the perspectives of excitation current / measurement voltage and excitation voltage / measurement current respectively.

[0029] During the data collection phase: a. Current mode excitation + dot matrix electrode voltage acquisition: During current mode excitation, the excitation signal is injected from the ring electrode group into the silver layer. The current forms a flowing electric field in the coating, which causes a potential difference at different spatial points. The dot matrix electrode group is densely distributed (9-point plum blossom array), so it can collect local voltage responses at multiple spatial locations. This produces a spatially labeled voltage response distribution map, reflecting the differences in the response of different regions to the same current stimulus. Surface response mapping can be achieved, revealing impedance variations at the coating-substrate interface in different regions, particularly sensitive to areas of weak adhesion.

[0030] b. Voltage mode excitation + ring electrode current acquisition: During voltage mode excitation, the excitation signal is applied from the dot matrix electrode group to the surface, forming a local high-frequency electric field. The voltage excitation guides the electron movement on the surface of the silver layer, inducing an induced current at the ring electrode. By measuring the response current of the ring electrode , combined with a known excitation voltage , the equivalent impedance can be inferred in reverse. The ring structure has a strong integration area and coupling capability, can effectively integrate the response signal under the voltage mode, can realize the overall phenomenon extraction of the high-frequency electric field induced response, effectively suppress point errors, and is suitable for surface conductivity change detection.

[0031] S13, dataset generation stage: After completing data acquisition under dual-mode excitation, the collected voltage and current response data need to be used for impedance calculation according to the excitation frequency and injection method, and a complete impedance data set needs to be constructed.

[0032] Current mode data processing: During current excitation, the voltage response data collected by the dot matrix electrode group is calculated correspondingly with the known current injection value to obtain the impedance value at each frequency point, forming a first-class impedance data set. This set records the changes in the voltage response of the coating structure when a fixed current is injected at different frequencies.

[0033] Voltage mode data processing: During voltage excitation, the current response data collected by the annular electrode group is calculated against the known voltage injection value to obtain the second type of impedance data set, which reflects the material's current conductivity performance under fixed voltage excitation conditions at different frequencies.

[0034] The two types of impedance data sets are merged using frequency as the index to form a complete impedance dataset containing both current mode and voltage mode components, namely the dual-mode multi-frequency impedance raw dataset. Each data entry carries the corresponding frequency, excitation method, measurement result, and spatial location information.

[0035] The core meaning is: Retain the response information of different layers and structures of the material in the two modes; Provide basic data support for subsequent spatial distribution imaging and quantitative evaluation of the adhesion of the silver coating; Ensure that the frequency layers of the original data are clear and the spatial labels are complete.

[0036] The specific data set is generated as follows: The collected response data is associated with the injection mode according to the frequency dimension to generate a dual-mode multi-frequency impedance raw data set, including: For current mode (current excitation, voltage response), the first impedance subset is generated according to the following equation: ; For voltage mode (voltage excitation, current response), the second impedance subset is generated according to the following formula: ; Merge the two impedance subsets above, aligning them on the same frequency grid and unified time axis, to form the complete dual-mode multi-frequency impedance raw data set: ; in, Indicates the frequency The voltage response generated by current mode excitation is Indicates the injection current value of the current mode, Indicates the injection voltage value of the voltage mode, Indicates the frequency The current response generated by voltage mode excitation is Indicates the excitation frequency, the range is 0.1-1kHz for current mode and 10-100kHz for voltage mode. 、 They are the impedance values ​​calculated in current / voltage mode respectively. Represents the dual-mode multi-frequency impedance raw data set.

[0037] Orthogonal timing injection mechanism: Taking the voltage mode frequency of 1kHz as an example, the excitation waveform and injection timing are as follows: Voltage cycle ms, current mode injection at the voltage signal zero crossing point, continuous ; Effectively avoid co-frequency coupling and excitation interference, the measured crosstalk rate is lower than ; The orthogonal timing injection mechanism avoids signal crosstalk. When current and voltage signals are injected in parallel, there is significant electromagnetic crosstalk, such as cross-induction and nonlinear mixing, which distorts the response signal. Embedding the current excitation signal into the zero-crossing point of the voltage excitation signal (that is, when the voltage waveform crosses the zero point and the electric field strength is minimum) can minimize the interference of the voltage waveform on the current response channel; limiting the current excitation duration to 1 / 4 of the voltage excitation cycle can ensure that the signal injection window is short and the impact is controllable, while not interfering with the frequency sampling accuracy. Local electrolytic reactions often occur during the silver plating layer inspection process, leading to bubble generation, which in turn changes the contact impedance. The orthogonal timing strategy avoids the simultaneous loading of two excitation signals, reduces the superposition of interface polarization and the concentrated outbreak of electrolytic bubbles, and enhances signal stability.

[0038] Table 1 Microelectrode array structure

[0039] S2, performing time-varying impedance phase-locked separation on the dual-mode multi-frequency impedance raw data set, extracting the dynamic impedance component of the coating-substrate interface region, and reconstructing a three-dimensional sub-interface impedance distribution map.

[0040] S21, dual phase-locked loop configuration stage: setting the first phase-locked loop and the second phase-locked loop.

[0041] The first phase-locked loop ensures that when processing the impedance signal, it always maintains strict phase synchronization with the swept frequency excitation signal applied in S1, thereby accurately separating the response component corresponding to the fundamental frequency and avoiding measurement errors caused by phase drift at the excitation end.

[0042] The second phase-locked loop is that the migration of silver ions on the interface will introduce an extremely low-frequency modulation component in the impedance signal. This characteristic frequency is directly related to the interface bonding state. By locking the phase of this characteristic frequency, the slow-changing dynamic process of the interface layer can be tracked, and the interface adhesion change can be extracted separately from the overall response.

[0043] The purpose of using dual phase-locked loops is to simultaneously maintain phase lock on the high-frequency excitation signal and the extremely low-frequency interface characteristic signal, achieve high signal-to-noise ratio extraction of the dynamic impedance component of the coating-substrate interface, and improve the stability of interface detection.

[0044] In summary, The original impedance spectrum is a frequency × time graph. The dual phase-locked loop (DPLL) acts as an intelligent filtering mechanism, identifying and extracting the portions of this graph most relevant to interface changes, outputting a clearer dynamic impedance graph. By splitting this graph into rows or columns (i.e., fixed frequency or fixed time), the temporal variation patterns and frequency response structures can be observed separately, achieving time-frequency decoupling.

[0045] Specifically include: Configure the first phase-locked loop to lock the fundamental frequency phase of the sweep frequency excitation signal in S1, which corresponds to the frequency point of the applied dual-mode excitation signal. ; A second phase-locked loop is configured to lock the characteristic phase of the characteristic frequency of 0.1–2 Hz. This frequency is related to the migration behavior of silver ions at the coating-substrate interface and is calculated as follows: ;in, represents the silver ion migration rate, represents the rate factor (prefactor), represents the activation energy of silver ions, represents the Boltzmann constant, is the temperature, and the migration behavior estimated by the above formula corresponds to an electrical modulation frequency of 0.1-2 Hz. The above formula calculates the migration rate of silver ions at the coating-substrate interface, which reflects the speed at which ions cross the energy barrier at different temperatures based on the exponential relationship of the thermal activation process. By calculating this migration rate, the corresponding electrical modulation characteristic frequency (0.1-2 Hz) can be deduced and used as the locking target of the second phase-locked loop, so as to extract only the dynamic impedance component related to the migration of silver ions at the interface and avoid interference from other irrelevant frequency components. The locked 0.1-2 Hz characteristic frequency is essentially the characteristic modulation signal derived from the silver ion migration process at the coating-substrate interface, so the frequency extracted from this frequency The dynamic impedance component is actually the response signal of the interface area.

[0046] S22, impedance component extraction stage: the dual-mode multi-frequency impedance raw data set is input into the dual phase-locked loop to extract the dynamic impedance component under the characteristic frequency modulation ; Perform time-frequency decoupling on the dynamic impedance component: Fixed frequency point When , extract the corresponding time domain oscillation component: ; Fixed time point When , extract the corresponding frequency domain response component: .

[0047] is the dynamic impedance component, which indicates the frequency and time The impedance response value under is used to reflect the dynamic electrical behavior of the structure under the modulation of the characteristic frequency of silver ion migration; is the time domain oscillation component, indicating that at a fixed frequency The fluctuation of impedance over time reflects the ion migration dynamics at the coating-substrate interface; is the frequency domain response component, indicating that at a fixed time The impedance distribution under different frequency excitations reflects the frequency response characteristics of the interface material, such as capacitance and conductivity distribution.

[0048] Specifically, the goal of S2 is to extract the weak signal caused by the dynamic changes of the coating-substrate interface from the impedance data obtained by S1, and to expand it from the two perspectives of time change and frequency response to prepare for three-dimensional reconstruction.

[0049] The impedance data obtained in the current mode and voltage mode in S1 constitute a complete impedance data set. , which is a complex impedance value organized by different excitation frequencies. In actual acquisition, these impedance values ​​are continuously updated in time series, that is, for each frequency point , all collected the impedance change trajectory over a period of time. Therefore, this input data is actually a two-dimensional structure: The horizontal axis is the frequency point, such as 0.1kHz, 0.5kHz, 1kHz; The vertical axis is the time point, which is sampled once per second and lasts for a few seconds; Each point is a complex impedance value, carrying magnitude and phase.

[0050] This two-dimensional impedance data is sent to a dual phase-locked loop, where each frequency point is processed independently.

[0051] The first phase-locked loop zeroes the signal at each frequency point, that is, aligns its phase with the excitation signal, eliminates errors caused by system delays and signal drift, and obtains a net impedance change signal with interface changes as the main modulation source (excluding the influence of the excitation end).

[0052] The second phase-locked loop extracts the characteristic frequency component in the impedance signal caused by the slowly changing behavior of the interface (such as silver ion migration). This characteristic frequency component is usually between 0.1–2 Hz and is a very low-frequency, very weak modulation. After phase-locked processing, it can be captured from the background to form a signal representing the dynamic response of the interface.

[0053] After processing by two phase-locked loops, a new data structure is output: at each frequency point, a dynamic impedance sequence that fluctuates slowly over time is extracted. This is the dynamic impedance component. This data structure is still two-dimensional (but the content has changed): The horizontal axis is the frequency point; The vertical axis is the time point; However, each point represents a small impedance fluctuation at the coating-substrate interface due to ion migration at that frequency and at that moment.

[0054] Next, this dynamic impedance data can be analyzed from two perspectives: A. Fix the frequency and observe the time changes to obtain the time domain component: Select a certain frequency point and observe the curve of the dynamic impedance changing with time at this frequency, which reflects the activity level and change law of the interface at this depth (corresponding to the frequency penetration layer).

[0055] B. Fix the time, observe different frequencies, and obtain the frequency domain components: Select a certain time point and view the response at different frequencies at that moment, which reflects the difference in electrical distribution of the interface in the entire spatial range (different frequencies correspond to different depths).

[0056] These two perspectives can be used for subsequent processing respectively: The results in the time direction are used to estimate the depth variation trend of the attachment interface; The frequency-direction results are used to construct a transverse impedance profile.

[0057] S23, 3D reconstruction stage: S231, Planar direction (xy plane) reconstruction: based on frequency domain response components ,Using the compressed sensing algorithm, the impedance plane diagram under the electrode spatial distribution is solved: ; means: make the signal as sparse as possible in a certain transform domain while satisfying the measurement error limit.

[0058] Objective function The impedance image after discrete cosine transform or Wavelet transform is expected ,Under this transformation, it is as sparse as possible (i.e., there are only a few non-zero coefficients), which is the core idea of ​​compressed sensing, used to restore high-resolution images.

[0059] Constraints Request reconstructed impedance image , after measuring the matrix After mapping, the actual collected data The error between (calculated in Euclidean norm) cannot exceed the tolerance range .

[0060] in, represents the discrete cosine transform (DCT) sparse basis, represents the measurement matrix consisting of electrode distribution, represents the impedance image vector to be reconstructed, represents the measured frequency domain component data, represents the error tolerance term (control robustness), Indicates constraints.

[0061] The role of the compressed sensing algorithm is to obtain a complete signal through sparse characteristics and optimized reconstruction when the data dimension is insufficient, thereby solving the problem of insufficient impedance image acquisition data due to the limited number of electrodes, and thus inferring the complete impedance distribution.

[0062] The steps of the compressed sensing algorithm are as follows: Construct a measurement matrix: Construct an electrode response matrix A based on the actual electrode arrangement to reflect the influence of electrodes at different positions on each pixel.

[0063] Select sparse transform: Select a sparse basis such as discrete cosine transform (DCT) to map the image Z into a sparse coefficient domain.

[0064] Set tolerance limits: Set the upper limit of the error based on the measurement error tolerance, allowing a certain degree of numerical deviation.

[0065] Solve the optimal image: solve through iterative algorithm, and finally restore the complete impedance map with the strongest sparsity and minimum error.

[0066] Image inverse transformation: The reconstruction result is converted from the sparse domain back to the original image domain to obtain a complete and continuous impedance distribution map in the x–y plane.

[0067] S232, depth direction (z axis) reconstruction: based on the time domain oscillation component The impedance gradient in the sub-interface direction is calculated by the phase change rate: ; in, represents the impedance phase angle, Indicates the current frequency point. represents the vacuum permeability, Indicates the electrical conductivity of the coating material, The rate of change of phase with time is expressed by Calculated, It is a complex impedance sequence collected over a period of time at a fixed frequency, where each time point corresponds to a complex number, whose amplitude represents the impedance strength, and the phase angle represents the phase relationship between current and voltage. By extracting the phase angle of this complex impedance sequence at each time point (that is, calculating the amplitude of the complex number), a curve of phase change over time can be obtained. This curve reflects the tiny phase drift caused by processes such as silver ion migration at the coating-substrate interface at this frequency. The derivative of this phase curve along the time axis is calculated to calculate its rate of change, that is, the rate of change of phase over time. This rate of change represents the degree of delay or response lag encountered by the signal when passing through the interface layer, and can be further used to estimate the impedance gradient under the interface and reflect the structural distribution in the depth direction.

[0068] The purpose of S232 is to reconstruct the impedance variation in the depth direction (z-axis) of the coating-substrate interface, that is, to infer the structural state of the interface layer from the surface downward. The impedance phase trajectory that changes with time at a certain frequency point has been extracted, that is, the time domain oscillation component. This oscillation component contains the slow change of the impedance phase caused by microscopic processes such as the migration of silver ions in the interface layer. By calculating the rate of change of the phase angle in the time dimension, the propagation delay of the current in the interface layer can be inferred. Since electrical signals of different frequencies have different penetration depths, combined with parameters such as frequency, conductivity and electromagnetic constants, the gradient of the impedance in the depth direction can be calculated based on the phase change rate. This gradient reflects the vertical change trend of the electrical distribution in the sub-interface area, and is then used to determine the distribution characteristics of the interface adhesion strength.

[0069] S233, the x-y plane reconstruction results are fused with the z-direction gradient to generate a complete three-dimensional sub-interface impedance distribution map: ; Its depth resolution can reach 0.2μm, which is better than traditional electrochemical impedance spectroscopy technology.

[0070] First, the reconstruction result of the x-y plane is a transverse impedance distribution map reconstructed at each time point or frequency point through algorithms such as compressed sensing, reflecting the local impedance level of the interface at different spatial locations. Simultaneously, by analyzing the time-domain impedance phase change at each frequency, the corresponding depth-direction impedance gradient is calculated. This gradient information describes the downward trend of the impedance value from the surface at each transverse coordinate point. The key to gradient fusion is to start from each pixel point on the x-y plane and, based on its corresponding impedance gradient information, layer by layer superimpose or integrate the impedance changes at different depths along the z-axis to construct a three-dimensional impedance structure composed of multiple superimposed cross sections. Finally, through the joint interpolation and continuous processing of all plane maps and depth gradients, a continuous three-dimensional impedance distribution map is formed, which not only retains the transverse spatial resolution but also introduces the vertical depth layer information, realizing three-dimensional imaging of the interface adhesion layer structure.

[0071] Simply put, gradient fusion is based on the lateral impedance map. By superimposing the depth gradient information, the impedance changes of each layer are deduced point by point, and finally a complete three-dimensional sub-interface distribution map is constructed.

[0072] S3, outputting a quantized adhesion value through an oscillation frequency-adhesion conversion model according to the spatial oscillation frequency characteristics of the impedance phase in the three-dimensional sub-interface impedance distribution diagram.

[0073] S31, spatial oscillation frequency extraction stage: In the three-dimensional sub-interface impedance distribution map, select the target depth layer ( m), extract the phase distribution along the direction parallel to the coating surface to form a two-dimensional phase matrix: ; Specifically, first select a depth layer, extract the impedance phase information of the corresponding position in the depth layer, ignore the impedance amplitude, and only retain the phase angle data to form a phase distribution diagram on a two-dimensional plane. Then, in this plane, scan each grid point at equal intervals along the x and y directions to collect its phase value, thereby forming a complete two-dimensional phase matrix. Each element of this two-dimensional phase matrix represents the local phase state of a point in the direction parallel to the coating surface, reflecting whether there are microscopic disturbances or defects on the interface. This matrix serves as the basic input for subsequent spatial frequency analysis (such as Fourier transform) to identify possible periodic changes or local damage areas in the interface structure; select The depth layer serves as the frequency analysis layer. This depth layer is usually located in the actual contact transition zone between the silver layer and the substrate material, that is, near the adhesion interface. This depth layer is affected by the microstructure of the silver layer and directly reflects the feedback of the substrate to the carrier behavior. Therefore, its electrical response can best reflect the real changes in the interface bonding state. In the three-dimensional sub-interface impedance diagram, the rate of change of the impedance phase is not consistent at different depths. Experiments show that the phase change near the surface is relatively gentle. At about 0.2μm, the phase perturbation caused by interface micro-defects, micro-bubbles or silver ion migration is most concentrated, showing a clear spatial oscillation feature. Layers that are too shallow are easily interfered by surface oxide layers, detection noise, etc., affecting the accuracy of the spectrum; while layers that are too deep have weak signal penetration and large energy attenuation, and the extracted phase changes are not obvious.

[0074] A two-dimensional fast Fourier transform is performed on the two-dimensional phase matrix, which includes taking the two-dimensional phase matrix extracted from the selected depth layer as input data and performing a fast Fourier transform (FFT) on the two-dimensional phase matrix simultaneously in the x and y directions, that is, converting the original spatial domain data into the frequency domain. This transformation process will identify whether there are spatial features such as periodic changes, edge mutations, or local high-frequency disturbances in the phase distribution, and represent them as a spectrum energy map. The final spectrum energy map is a new two-dimensional matrix, in which each point corresponds to a specific spatial frequency combination, and the numerical value represents the energy intensity of the frequency component in the original phase map. By analyzing this spectrum map, the dominant oscillation frequency and its spatial energy ratio can be found, which can be used to judge the uniformity of the interface structure, the degree of defects, and the scope of their influence, expressed as: ; Extract the main oscillation frequency component, which is defined as the frequency modulus corresponding to the position with the maximum spectral energy halo: ;in, represents the plane directional phase distribution diagram, represents the two-dimensional Fourier transform operator, represents a two-dimensional spatial frequency domain image, Indicates the main oscillation frequency.

[0075] The most important spatial oscillation frequency is extracted from the phase distribution of the interface to reflect the degree of microscopic disturbance of the interface structure between the silver-plated layer and the substrate. It can identify whether there are obvious periodic changes or local damage areas on the interface, and quantify this change with a representative frequency value, providing key characteristic indicators for subsequent adhesion evaluation.

[0076] S32, damage feature association stage: Calculate the energy ratio of the main oscillation frequency component in the spectrum: ;in, Represents the energy value of the main oscillation frequency position in the spectrum, Represents the total energy of the entire spectrum. This ratio is used to measure the intensity weight of the main oscillation frequency in the entire spectrum.

[0077] S33, adhesion quantization conversion stage: the main oscillation frequency Its energy ratio At the same time, input the frequency-adhesion conversion model and output the corresponding adhesion value : ; in, represents the model coefficient related to the base material of the silver plating layer, (for copper substrate), is the quantitative value of adhesion. Adhesion grades are classified according to the standards in Table 1 below.

[0078] Model factors: Item 1 : Used to describe the microscopic oscillation characteristics of the interface phase. The higher the oscillation frequency, the rougher the interface or the more severe the damage, and the lower the adhesion.

[0079] Item 2 : Used to reflect the proportion of the main frequency oscillation area in the overall spectrum. The larger the damage range, the higher the proportion and the worse the adhesion.

[0080] In the frequency-adhesion conversion model: The spatial phase oscillation characteristic (i.e., the main oscillation frequency) and its energy fraction in the spectrum are converted into a physically interpretable and quantifiable adhesion index. This model, centered on the influence of the actual interface microstate on the phase distribution, establishes a quantitative relationship between these two parameters, frequency and energy fraction, and adhesion.

[0081] Main oscillation frequency : The higher the frequency, the more dramatic changes there are in the phase distribution, reflecting the presence of finer or denser structural disturbances at the interface, such as microcracks, interface roughness, and dendrite growth. The adhesion is worse, so an exponential decay function is introduced to describe the negative correlation between increasing frequency and adhesion.

[0082] Spectrum energy ratio The higher the percentage, the greater the energy occupied by the main oscillation frequency in the entire spectrum, indicating the presence of large-scale structural defects or periodic perturbations in the interface. Using a logarithmic function to characterize this effect avoids linear amplification of local changes and enhances the stability and adaptability of the model.

[0083] The overall model is a two-factor weighted structure, in which the exponential function controls the main tone and represents frequency sensitivity; the logarithmic term is a correction term that reflects the proportion of the damaged area.

[0084] The model building process is as follows: Preliminary experimental sampling: Under various typical adhesion conditions, including intact, micro-cracked, and delaminated, the corresponding main oscillation frequency and spectrum share values ​​are obtained through actual measurements. Standard adhesion tests, including peel tests or ASTM adhesion grade scores, are then performed on each group of samples.

[0085] Parameter relationship fitting: nonlinear regression method is used to use the measured 、 With known adhesion value A training set was constructed and multiple function models (linear, polynomial, logarithmic, exponential, and exponential-logarithmic mixed) were compared. Finally, it was determined that the combination of exponential + logarithmic best fits the actual change trend and has the smallest fitting residual.

[0086] Parameter determination: The optimal parameter group under the copper matrix is ​​obtained by least square fitting: : represents the theoretical maximum adhesion force when the interface is intact; : Indicates the attenuation rate of frequency to adhesion; : The weighted measure of the control energy proportion.

[0087] Different matrices (such as steel and nickel) are fitted separately through the same process.

[0088] Table 2 Adhesion grade determination table

[0089] Table 2 above is based on the output of the oscillation frequency-adhesion force conversion model, which was fitted through extensive sample experiments in conjunction with standard adhesion testing methods. Specifically, the researchers measured the main oscillation frequency and energy fraction on samples with different adhesion states. The quantitative values ​​output by the model were then compared with the actual peeling grade to ultimately determine the corresponding relationship between the AF value and the grading threshold. The purpose of Table 2 is to convert continuous adhesion strength quantification values ​​into grading criteria that can be directly applied in engineering. This allows test results of grades A, B, and C to intuitively correspond to good, medium, and poor adhesion states. This not only facilitates rapid assessment of batch product quality but also ensures that the output of this method is consistent with international testing standards. This method can replace destructive testing methods in scenarios such as electronic connectors and communication components, enabling non-destructive online grading.

[0090] The adhesion grade classification refers to the grade scoring method in existing domestic and international standards (ASTM D3359), combined with the statistical results of actual test samples, the adhesion quantification values ​​output by the conversion model are divided into three levels: Grade A (AF ≥ 80): The interface structure is complete, the impedance phase changes smoothly, the frequency is low, and the proportion is low; the actual measurement corresponds to 5B–4B in ASTM, representing an ideal adhesion state.

[0091] Grade B (40≤AF<80): Localized microcracks or slight voids exist at the interface, and moderately high-frequency components appear in the phase; corresponds to 3B–2B in ASTM, indicating moderate adhesion strength, which may affect long-term reliability.

[0092] Grade C (AF<40): Large-area delamination, desorption, or dendrite penetration, with high-frequency energy concentrated in the spectrum and accounting for a large proportion; corresponds to 1B–0B in ASTM, indicating severe adhesion failure.

[0093] Table 3 Adhesion grade mapping and material parameters

[0094] The α, β, and γ parameters in Table 3 are derived from extensive experimental data fitting of samples from various substrate materials (e.g., copper, nickel, and steel). By comparing the relationship between frequency characteristics, energy fraction, and actual adhesion strength, a nonlinear regression method was used to determine the optimal parameter set, reflecting the physical properties of different materials. Table 3 provides the model's unique parameter configurations for different material systems, making the conversion model universal across materials. For example, in copper-based electronic connectors, the copper parameters in the table can be directly used for adhesion assessment. Similarly, accurate determinations can be made on steel-based structural components or nickel-based electronic devices using the corresponding parameters.

[0095] Among them, α determines the upper limit of the overall adhesion and reflects the maximum adhesion strength under the ideal interface state; β controls the decay rate of the main oscillation frequency to the adhesion force, reflecting the sensitivity of the material interface to microscopic perturbations; γ controls the correction amplitude of the spectrum ratio on adhesion, reflecting the material's tolerance to local damage area.

[0096] 1. Copper (Cu): Copper and silver easily form a good metallic bond, with medium to high interfacial bonding strength. Its surface is relatively soft, prone to slight oscillations due to microcracks or ion migration, and has moderate frequency sensitivity. Therefore, the setting is α = 85: medium to high maximum adhesion; β = 0.12: moderate frequency influence; γ = 7.5: moderate sensitivity to damaged areas.

[0097] 2. Nickel (Ni): Nickel has a stronger tendency to oxidize on the surface, making the silver-plated layer more susceptible to interfacial isolation. Its overall adhesion is weaker and it is more sensitive to frequency oscillations. Therefore, the following settings are set: α = 78: This lowers the upper limit of adhesion; β = 0.15: This more strongly reflects the tendency of adhesion to attenuate with increasing frequency; and γ = 6.8: This has a slightly smaller impact on the percentage of damaged area, making it suitable for a more uniform failure mechanism.

[0098] 3. Steel (Fe): Steel has a high surface hardness, making uniform adhesion of the silver layer difficult and resulting in high interface roughness. However, once bonded, it is highly stable and insensitive to local defects. Therefore, the following settings are made: α = 92: adhesion is highest under ideal bonding conditions; β = 0.10: the oscillation frequency has a weaker effect on adhesion attenuation; and γ = 8.2: the damage fraction has a greater impact, reflecting its high sensitivity to macroscopic delamination or uneven stress.

[0099] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0100] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for detecting adhesion of a silver coating based on multi-frequency impedance imaging, characterized in that: The following steps are involved: S1, applies dual-mode multi-frequency excitation signals to the silver-plated surface through the microelectrode array, synchronously collects voltage-current response data, and generates a dual-mode multi-frequency impedance raw data set; S2, performing time-varying impedance phase-locked separation on the dual-mode multi-frequency impedance raw data set, extracting the dynamic impedance component of the coating-substrate interface region, and reconstructing a three-dimensional sub-interface impedance distribution map; S3, outputting a quantized adhesion value through an oscillation frequency-adhesion conversion model according to the spatial oscillation frequency characteristics of the impedance phase in the three-dimensional sub-interface impedance distribution diagram.

2. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 1, wherein: The dual-mode multi-frequency excitation signal includes a current mode and a voltage mode. The application of the dual-mode multi-frequency excitation signal includes applying a current mode excitation signal through the annular electrode group of the attached microelectrode array and applying a voltage mode excitation signal through the dot matrix electrode group.

3. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 2, wherein: The current mode excitation signal and the voltage mode excitation signal are injected in orthogonal timing, the current mode excitation is started at the zero crossing point of the voltage mode excitation, and the duration is one quarter of the voltage cycle; During current mode stimulation, voltage response data of the array electrode group were collected synchronously; During voltage-mode excitation, current response data of the ring electrode set are acquired synchronously.

4. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 3, wherein: The S1 also includes associating the voltage response data in the current mode with the corresponding frequency points to generate a first impedance subset; associating the current response data in the voltage mode with the corresponding frequency points to generate a second impedance subset; and merging the first impedance subset and the second impedance subset to form a dual-mode multi-frequency impedance original data set.

5. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 1, wherein: The time-varying impedance phase-locked separation includes a dual phase-locked loop configuration, including a first phase-locked loop and a second phase-locked loop, wherein; Setting a first phase-locked loop to lock the phase of a fundamental frequency excitation signal, wherein the fundamental frequency excitation signal is the swept frequency excitation signal in S1; The second phase-locked loop is set to lock the phase of a low characteristic frequency, where the low characteristic frequency is related to the migration rate of silver ions at the coating-substrate interface.

6. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 5, wherein: The method of extracting the dynamic impedance component of the coating-substrate interface region includes inputting the dual-mode multi-frequency impedance raw data set into a dual phase-locked loop, and extracting the dynamic impedance component modulated by the characteristic frequency. ,right Perform time-frequency decoupling, including: When the frequency point f is fixed, the time domain oscillation component is obtained ; At a fixed time point t, obtain the frequency domain response component .

7. The method for detecting adhesion of a silver coating based on multi-frequency impedance tomography according to claim 6, wherein: The reconstructing of the three-dimensional sub-interface impedance distribution map comprises: Based on the frequency domain response components , the compressed sensing algorithm is used to reconstruct the xy plane impedance distribution; Based on the time domain oscillation component , calculate the impedance gradient in the depth direction by phase delay inversion; The plane distribution and depth gradient are integrated to generate a three-dimensional sub-interface impedance distribution map.

8. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 1, wherein: The extraction of the spatial oscillation frequency characteristics includes selecting a preset depth layer in the three-dimensional sub-interface impedance distribution map, extracting the impedance phase distribution of the preset depth layer parallel to the coating surface, and generating a two-dimensional phase matrix; performing a two-dimensional fast Fourier transform on the two-dimensional phase matrix to obtain its spectrum energy map; and extracting the main oscillation frequency component with the largest energy in the spectrum energy map as the spatial characteristic frequency feature.

9. The method for detecting adhesion of a silver coating based on multi-frequency impedance imaging according to claim 8, wherein: The S3 includes calculating an energy share coefficient based on the ratio of the spectrum energy of the main oscillation frequency component to the total spectrum energy, and inputting the main oscillation frequency component and its corresponding energy share coefficient into a preset oscillation frequency-adhesion conversion model to output an adhesion quantization value.

10. The method for detecting adhesion of a silver coating based on multi-frequency impedance tomography according to claim 9, wherein: S3 further includes classifying the adhesion level according to the adhesion quantization value, comparing the adhesion quantization value with a preset level determination threshold, and determining the adhesion level as Class A when the adhesion quantization value is greater than or equal to a first threshold; determining the adhesion level as Class B when the adhesion quantization value is between the first and second thresholds; and determining the adhesion level as Class C when the adhesion quantization value is less than the second threshold.

Images