A method and system for intelligent detection of defects in BMS boards

Abstract

This invention discloses an intelligent defect detection method and system for BMS boards. The method aims to construct a circuit state spectrum library by collecting multi-domain abnormal signals and performing frequency domain fusion, identifying abnormal impedance modes and generating defect excitation points; performing gradient analysis on thermal infrared distribution signals to generate a thermal flux topological field, extracting the location of abnormal heat sources and establishing thermoelectric coupling coordinates; injecting a frequency-sweeping signal into the defect excitation point to generate defect response characteristics, and generating an enhanced detection domain through resonant amplification; performing spectral decomposition on the enhanced detection domain to identify defect resonant frequencies, and constructing a phase detection network using multi-frequency superposition excitation and phase weaving; generating a defect location matrix by harmonic crosstalk mining based on the phase detection network and the defect impedance distribution domain; reconstructing the defect electric field topology by performing impedance gradient analysis on the defect location matrix to achieve precise defect coordinate location; and finally, completing defect type identification and severity level assessment through hierarchical labeling and tiered damage recursive chain analysis.

Description

Technical Field

[0001] This invention relates to the field of electronic equipment testing technology, and in particular to an intelligent detection method and system for BMS board defects. Background Technology

[0002] As the core control unit of the battery management system, the BMS board undertakes critical functions such as battery status monitoring, charge and discharge control, and safety protection. Its reliability directly affects the safety and lifespan of the entire battery system. However, under complex operating environments, BMS boards are prone to various defects such as solder joint cracking, wire breakage, insulation degradation, and component aging. These defects are often highly concealed and develop rapidly, posing a serious threat to the safe operation of the battery system.

[0003] Existing BMS board inspection technologies primarily rely on single-signal source analysis methods, such as resistance measurement, temperature monitoring, or impedance analysis. These methods suffer from limitations such as insufficient detection accuracy, inaccurate defect localization, and inability to identify early, minute defects. Furthermore, traditional inspection methods lack in-depth analysis of the correlations between various physical phenomena, making it difficult to establish precise mappings between defect characteristics and physical locations. This hinders intelligent defect classification and quantitative assessment of severity, severely restricting the technological development of BMS board fault prediction and preventative maintenance. Summary of the Invention

[0004] This invention provides a method and system for intelligent detection of defects in BMS boards. It aims to fully integrate multi-domain abnormal signals such as impedance spectrum data, thermal infrared distribution signals, and circuit trace information to accurately identify and locate key detection elements such as defect excitation points, abnormal heat source locations, and defect resonant frequencies. Furthermore, it reveals the spatial distribution and damage propagation mechanism of defects through techniques such as frequency sweep signal injection, multi-frequency superposition excitation, and phase detection networks, ultimately forming a multi-level, high-precision defect location matrix and severity level assessment system.

[0005] The first aspect of this invention proposes an intelligent detection method for defects in BMS boards, comprising the following steps:

[0006] Collect multi-domain abnormal signals and circuit routing information during the operation of the BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. Perform frequency domain fusion on the multi-domain abnormal signals to construct a circuit state spectrum library.

[0007] Abnormal impedance modes are identified by the circuit state spectrum library, and reverse excitation analysis is performed on the abnormal impedance modes to generate defect excitation points. A frequency sweep signal is injected into the defect excitation points to generate defect response features, and the defect response features are resonantly amplified to generate an enhanced detection domain.

[0008] Gradient analysis is performed on the thermal infrared distribution signal to generate a thermal flow topology field. Abnormal heat source locations are extracted from the thermal flow topology field. The abnormal heat source locations are geometrically correlated with the circuit trace information to generate thermoelectric coupling coordinates. Impedance mapping is performed based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain.

[0009] The enhanced detection domain is subjected to spectral decomposition to identify the defect resonant frequency point. At the defect resonant frequency point, multi-frequency superposition excitation is performed to generate a superimposed excitation field. The superimposed excitation field is phase-woven to construct a phase detection network. Based on the phase detection network and the defect impedance distribution domain, harmonic crosstalk is mined to generate a defect location matrix.

[0010] The defect location matrix is ​​analyzed by impedance gradient to generate a gradient tracking field. The field strength distribution around the defect is reconstructed based on the gradient tracking field to generate the defect electric field topology. The defect electric field topology is used to locate the field strength peak to generate the precise coordinates of the defect.

[0011] Based on the precise coordinates of the defects, layered annotation is performed to extract the defect type and severity level, thus completing the intelligent detection of defects on the BMS board.

[0012] A second aspect of this invention provides an intelligent detection system for BMS board defects, comprising:

[0013] The signal acquisition module is used to acquire multi-domain abnormal signals and circuit routing information during the operation of the BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. The multi-domain abnormal signals are fused in the frequency domain to construct a circuit state spectrum library.

[0014] The excitation detection module is used to identify abnormal impedance modes through the circuit state spectrum library, perform reverse excitation analysis on the abnormal impedance modes to generate defect excitation points, inject a frequency sweep signal into the defect excitation points to generate defect response features, and perform resonant amplification on the defect response features to generate an enhanced detection domain.

[0015] The thermoelectric analysis module is used to perform gradient analysis on the thermal infrared distribution signal to generate a thermal flow topology field, extract the location of abnormal heat source from the thermal flow topology field, geometrically correlate the location of abnormal heat source with the circuit routing information to generate thermoelectric coupling coordinates, and perform impedance mapping based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain.

[0016] The spectrum localization module is used to perform spectrum decomposition on the enhanced detection domain to identify the defect resonant frequency point, perform multi-frequency superposition excitation at the defect resonant frequency point to generate a superimposed excitation field, perform phase weaving on the superimposed excitation field to construct a phase detection network, and perform harmonic crosstalk mining based on the phase detection network and the defect impedance distribution domain to generate a defect localization matrix.

[0017] The precise positioning module is used to perform impedance gradient analysis on the defect positioning matrix to generate a gradient tracking field, reconstruct the field strength distribution around the defect based on the gradient tracking field to generate the defect electric field topology, and use the defect electric field topology to perform field strength peak positioning to generate the precise coordinates of the defect.

[0018] The results output module is used to extract the defect type and severity level by performing layered annotation based on the precise coordinates of the defect, and to complete the intelligent detection of defects on the BMS board.

[0019] The beneficial effects of this invention are reflected in the following points: First, by using frequency domain fusion technology of multi-domain abnormal signals, impedance spectrum data and thermal infrared distribution signals are used to construct a circuit state spectrum library, enabling accurate identification of abnormal impedance modes. Furthermore, defect excitation points are generated through reverse excitation analysis, and combined with frequency sweep signal injection and resonant amplification processing, weak defect signals are converted into strong detectable signals, improving the sensitivity and reliability of defect detection. Second, by generating a thermal flux topological field through gradient analysis of thermal infrared distribution signals, a geometric relationship is established between the location of abnormal heat sources and circuit trace information, forming a thermoelectric coupling coordinate system. Impedance mapping technology is used to generate a defect impedance distribution domain, achieving a precise spatial correspondence between thermal phenomena and circuit defects, thus improving the accuracy of defect location. Finally, a phase detection network is constructed using multi-frequency superposition excitation and phase weaving technology. Combined with harmonic crosstalk mining methods, a defect location matrix is ​​generated. Precise defect coordinates are determined through impedance gradient analysis and electric field topology reconstruction. Finally, defect type identification and severity assessment are completed based on tiered damage recursive chain analysis, forming a complete technical chain from signal acquisition to defect diagnosis.

[0020] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0021] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0022] Unless otherwise specified or defined, the same reference numerals in different figures represent the same or similar technical features, and different reference numerals may be used to represent the same or similar technical features.

[0023] Figure 1 This is a flowchart illustrating an intelligent detection method for BMS board defects according to the present invention.

[0024] Figure 2 This is a structural block diagram of an intelligent detection system for BMS board defects according to the present invention. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0027] The technical solutions of the embodiments of this application are described below.

[0028] like Figure 1 As shown, this embodiment of the invention provides an intelligent detection method for BMS board defects, including the following steps S110-S160:

[0029] Step S110: Collect multi-domain abnormal signals and circuit trace information during the operation of the BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. Perform frequency domain fusion on the multi-domain abnormal signals to construct a circuit state spectrum library.

[0030] Specifically, multi-domain abnormal signals and circuit trace information are collected during the operation of the BMS board. Impedance analyzers and thermal infrared imaging devices are deployed at key nodes of the BMS board to simultaneously collect multi-domain abnormal signals, including impedance spectrum data and thermal infrared distribution signals. The impedance analyzer employs a four-wire measurement method, with a frequency scanning range covering 1Hz to 1MHz, performing impedance spectrum measurements at the input and output ports of key components such as battery management chips, power MOSFETs, and current sensing resistors. A small signal excitation is applied during the measurement process to avoid interfering with the normal operation of the BMS. The thermal infrared imaging device uses non-contact temperature measurement technology, fixedly installed above the BMS board, to monitor the temperature distribution changes on the board surface in real time. The thermal infrared sensor uses a microbolometer array, achieving a temperature resolution of 0.1℃ and a spatial resolution covering every square millimeter of the board surface. Circuit trace information is simultaneously collected, obtaining detailed information such as the geometry of the traces, conductor width, interlayer spacing, and via distribution through PCB design files and 3D scanning technology. A digital model of the circuit traces is established, including the start and end coordinates, path length, cross-sectional area, and material properties of each trace. All acquired data is recorded synchronously via a high-speed data acquisition card to ensure time alignment and spatial matching between multi-domain abnormal signals and circuit routing information.

[0031] In some embodiments, the step of constructing a circuit state spectrum library by frequency domain fusion of the multi-domain abnormal signals includes: identifying impedance abnormal frequency bands in the impedance spectrum data; generating thermoelectric correlation points based on temperature correlation between the impedance abnormal frequency bands and the thermal infrared distribution signal; using the thermoelectric correlation points to excite changes in circuit trace current distribution to generate a current distribution domain; and constructing a circuit state spectrum library based on the current distribution domain.

[0032] Identify impedance anomaly bands in impedance spectrum data. The collected impedance spectrum data is segmented by frequency to establish a baseline impedance spectrum model under normal operating conditions. This baseline model is established through statistical analysis of a large amount of impedance spectrum data under normal operating conditions, including the mean and standard deviation of impedance values ​​at each frequency point. The deviation between the real-time impedance spectrum and the baseline model is calculated as D(f) = |Z(f) - Z_ref(f)| / σ(f), where Z(f) is the measured impedance, Z_ref(f) is the mean baseline impedance, and σ(f) is the standard deviation of the baseline impedance. When the deviation exceeds a set threshold, the corresponding frequency segment is marked as an impedance anomaly band. A sliding window technique is used for continuous detection of the anomaly bands; the window size is determined based on the BMS control cycle. Spectral features are extracted from the identified impedance anomaly bands, and parameters such as the center frequency, bandwidth, peak amplitude, and quality factor are calculated. A classification system for impedance anomaly bands is established, classifying them into three categories based on their anomaly characteristics: capacitive anomalies, inductive anomalies, and resistive anomalies. Record the timestamp, spatial location, and anomaly type for each impedance anomaly frequency band to provide accurate frequency domain characteristic information for subsequent temperature correlation analysis.

[0033] Thermoelectric correlation points are generated based on temperature correlation between impedance anomaly frequency bands and thermal infrared distribution signals. The identified impedance anomaly frequency band information is spatially and temporally matched with the simultaneous thermal infrared distribution signal. A spatial coordinate system for the BMS board is established, unifying impedance measurement points and thermal infrared pixels under the same coordinate system. Temperature gradient analysis is performed on the thermal infrared distribution signal, calculating the amplitude and direction of the temperature gradient for each pixel. Temperature anomaly regions are identified, defined as areas where the temperature value exceeds the normal operating temperature range or the temperature gradient exceeds a threshold. The spatial distance and temperature correlation between the measurement points corresponding to the impedance anomaly frequency bands and the temperature anomaly regions are calculated. The Pearson correlation coefficient is used to evaluate the correlation strength between the degree of impedance anomaly and the degree of temperature anomaly: r = Σ(Z_i - Z_mean)(T_i - T_mean) / √[Σ(Z_i - Z_mean)]. 2 Σ(T_i-T_mean) 2 [Thermoelectric correlation point], where Z_i is the impedance anomaly, T_i is the temperature anomaly, and Z_mean and T_mean are the mean values ​​of impedance and temperature, respectively. When the correlation coefficient exceeds the critical value and the spatial distance is within a reasonable range, the location is marked as a thermoelectric correlation point. The coordinates, impedance anomaly characteristics, temperature anomaly characteristics, and correlation strength of the thermoelectric correlation points are recorded to establish a thermoelectric correlation point database.

[0034] For example, the step of generating a current distribution domain by stimulating the change in circuit trace current distribution using the thermoelectric correlation point includes: analyzing the circuit trace current state based on the thermoelectric correlation point to generate a redistribution scheme; monitoring the current density of the redistribution scheme to form a density change curve; identifying the current path through the density change curve to generate a dominant current path; and spatially mapping the dominant current path to generate a current distribution domain.

[0035] A current redistribution scheme is generated based on the analysis of circuit trace current states at thermoelectric correlation points. For identified thermoelectric correlation points, their impact mechanism on the current state of surrounding circuit traces is analyzed in detail. Anomalies at thermoelectric correlation points can lead to increased local resistance or poor contact, forcing current to seek new conduction paths. An equivalent circuit model of the circuit traces is established, modeling each trace as a combination of resistance, inductance, and capacitance. An additional impedance component is introduced at the thermoelectric correlation point location to simulate the impact of anomalies on circuit characteristics. Kirchhoff's laws are used to analyze the current distribution in each trace, calculating the current redistribution under abnormal conditions. An optimization objective function for current redistribution is established, aiming to minimize power consumption and maximize current distribution uniformity. A genetic algorithm is used to solve for the optimal current distribution scheme. In the algorithm, individuals represent different current distribution strategies, and the fitness function comprehensively considers factors such as power consumption, temperature rise, and electromagnetic compatibility. Through multiple generations of evolution and selection, the optimal current redistribution scheme under abnormal conditions of thermoelectric correlation points is obtained. The redistribution scheme includes the target current value, current distribution ratio, and optimized circuit parameter settings for each trace. Record the implementation conditions, optimization objectives, and expected effects of the redistribution scheme to provide a reference standard for subsequent current density monitoring.

[0036] The current density of the redistribution scheme is monitored to generate density change curves. Circuit parameters are adjusted according to the generated redistribution scheme to implement the current redistribution. High-precision current sensors are deployed on key circuit traces to monitor the current density changes in real time. These current sensors utilize the Hall effect principle, featuring high bandwidth and low noise characteristics, enabling them to capture rapid changes in current density. The monitoring process covers the entire time cycle of the redistribution scheme implementation, recording data from the initial state to steady state. Time-series data of current density is established, including the current density amplitude, rate of change, and trend at each monitoring point. Digital filtering techniques are used to preprocess the monitoring data to eliminate the influence of noise and interference signals. Statistical characteristics of the current density are calculated, including parameters such as maximum, minimum, average, and standard deviation. A curve of current density versus time is plotted, with the horizontal axis representing time and the vertical axis representing the current density amplitude. The density change curve clearly shows the evolution trajectory of current density during the redistribution process, including transient response and steady-state characteristics. Segmented analysis of the density change curve is performed to identify the change characteristics and transition points at different stages.

[0037] Dominant current paths are identified through density variation curves. Based on these curves, the distribution and priority of current along different paths are analyzed. The peak values ​​of the density variation curves correspond to the paths with the highest current density, which carry the main current conduction tasks. A threshold determination method is used to identify dominant current paths, defining paths whose current density exceeds a certain proportion of the total current as dominant paths. The current carrying capacity and conduction efficiency of each path are calculated, taking into account factors such as path resistance, geometry, and heat dissipation. A path dominance evaluation index, P=I, is established. 2 R / A, where I is the path current, R is the path resistance, and A is the path cross-sectional area. A path with high dominance indicates that it plays a dominant role in current conduction. All identified current paths are ranked by dominance to determine the primary dominant current paths and secondary auxiliary paths. The stability and reliability of the dominant current paths are analyzed, and their current-carrying capacity under different operating conditions is evaluated. The geometric parameters, electrical characteristics, and thermal characteristics of the dominant current paths are recorded to establish a path characteristic database.

[0038] A current distribution domain is generated by spatially mapping the dominant current paths. The identified dominant current paths are precisely located and visualized in 3D space. A 3D geometric model of the BMS board is established, including complete spatial information of circuit traces, components, and heat dissipation structures. The 2D topology of the dominant current paths is mapped to the 3D geometric model, determining the spatial coordinates and direction of each path. Spline interpolation is used to smooth the paths, generating continuous 3D curves. A local coordinate system is established on each dominant current path, defining the tangential, normal, and subnormal directions of the path. The current density distribution around the path is calculated, considering the skin effect and proximity effect. A mesh model of the current distribution domain is established, dividing the 3D space into regular mesh cells. Current density vectors are interpolated and calculated in each mesh cell, containing information on the magnitude and direction of the current. The current distribution domain is visualized using color coding and vector arrows; color intensity represents current density magnitude, and arrow direction represents current direction. The current distribution domain provides a complete 3D image of the current distribution inside the BMS board, clearly showing the current flow patterns and concentrated areas. A mathematical description of the current distribution domain is established, and the current density at each point in space is represented by a vector field function.

[0039] A circuit state spectral library is constructed based on the current distribution domain. The generated current distribution domain data is categorized and organized according to different anomaly types, anomaly degrees, and spatial locations. Feature vectors for circuit states are established, encompassing statistical, spatial, and time-varying characteristics of the current distribution domain. Statistical characteristics include parameters such as the mean, variance, peak value, and RMS value of the current density. Spatial distribution characteristics are extracted using principal component analysis to describe the spatial patterns of the current distribution. Time-varying characteristics are extracted using time-frequency analysis to identify the changing patterns of the current distribution over time, recognizing periodic and non-periodic variation patterns. Clustering algorithms are used to classify the circuit state feature vectors under different operating conditions, establishing a classification system for circuit states. Each circuit state corresponds to a specific current distribution pattern and anomaly feature combination. The data structure of the circuit state spectral library is constructed, including state identifiers, feature vectors, current distribution domain data, and associated impedance spectrum and temperature distribution information. An indexing mechanism for the spectral library is established to support rapid retrieval of corresponding circuit states based on anomaly features. The spectral library adopts a dynamically expanding design, capable of continuously receiving and classifying new circuit state data.

[0040] Step S120: Identify abnormal impedance modes through the circuit state spectrum library, perform reverse excitation analysis on the abnormal impedance modes to generate defect excitation points, inject a frequency sweep signal into the defect excitation points to generate defect response features, and perform resonant amplification on the defect response features to generate an enhanced detection domain.

[0041] Specifically, abnormal impedance patterns are identified using a circuit state spectrum library. Real-time acquired impedance spectrum data is compared and analyzed with standard patterns stored in the library, and pattern matching techniques are used to identify abnormal impedance characteristics in the current circuit state. Feature vectors for impedance patterns are established, including key parameters such as peak position, peak amplitude, half-peak width, quality factor, and phase characteristics of the frequency response curve. Euclidean distance and cosine similarity are used to calculate the matching degree between the measured impedance spectrum and each pattern in the circuit state spectrum library. When the matching degree exceeds a set threshold, the corresponding pattern is marked as an abnormal impedance pattern. A support vector machine classifier is used to classify the identified abnormal impedance patterns into different types, such as device aging, loose connection, solder joint cracking, and insulation degradation. Each abnormal impedance pattern corresponds to a specific physical defect mechanism and fault development trend. A severity assessment system for abnormal impedance patterns is established, determining the abnormality level based on the degree to which the pattern deviates from the normal state. Accurate identification of abnormal impedance patterns successfully distinguishes different types of circuit defect characteristics.

[0042] Inverse excitation analysis is performed on anomalous impedance modes to generate defect excitation points. Reverse engineering methods are used to analyze the physical defect locations and mechanisms that generate anomalous impedance modes. A causal mapping between impedance anomalies and physical defects is established, and the influence of different types of defects on the impedance spectrum is analyzed using finite element method. The gradient descent method is used to solve the inverse problem, deducing the most probable defect locations and defect parameters from the observed anomalous impedance modes. An objective function for defect parameter optimization is established: F = Σ|Z_measured(f) - Z_calculated(f,p)| 2 Where Z_measured represents the measured impedance, Z_calculated represents the calculated impedance, p represents the defect parameter, and f represents the frequency. The optimal combination of defect parameters, including defect location coordinates, defect size, defect type, and defect severity, is determined by minimizing the objective function. Multiple random initializations are performed using the Monte Carlo method to improve the reliability and accuracy of the back-analysis. Uncertainty analysis is performed on the solution results to evaluate the confidence intervals of the defect location and parameters. The defect location determined by the back-analysis is marked as the defect excitation point. Precise localization of the defect excitation point achieves an effective conversion from frequency domain characteristics to spatial location.

[0043] In some embodiments, the step of injecting a swept-frequency signal at the defect excitation point to generate defect response characteristics includes: acquiring the electromagnetic resonance characteristics of the defect excitation point; generating a resonance excitation point by inducing defect resonance based on the electromagnetic resonance characteristics; generating an echo excitation wave by echo amplifying the swept-frequency signal through the resonance excitation point; and generating defect response characteristics based on the echo excitation wave.

[0044] The electromagnetic resonance characteristics of the defect excitation point were obtained. At the determined location of the defect excitation point, the electromagnetic resonance characteristics of the point were measured in detail using a network analyzer and a near-field scanning probe. The near-field scanning probe, employing miniature electric and magnetic field probes, is capable of measuring the spatial distribution of the electromagnetic field at a sub-millimeter scale. A three-dimensional electromagnetic field description around the defect excitation point was established, including the spatial distribution of electric field strength, magnetic field strength, and Poynting vector. The input impedance of the defect excitation point as a function of frequency was measured, and resonant and anti-resonant points on the impedance curve were identified. The quality factor Q = ωL / R at the resonant frequency was calculated, where ω represents the angular frequency, L represents the equivalent inductance, and R represents the equivalent resistance. The electromagnetic resonance modes of the defect excitation point were analyzed, including different field distribution modes such as TM mode, TE mode, and hybrid mode. The impedance discontinuity of the defect excitation point was analyzed using time-domain reflectometry, identifying the reflection and scattering characteristics caused by the defect. An equivalent circuit description of the defect excitation point was established, including lumped parameters such as resistance, inductance, capacitance, and mutual inductance. The parameter values ​​of the equivalent circuit were determined from the measurement data using parameter extraction methods. The complete measurement of the electromagnetic resonance characteristics revealed the inherent frequency characteristics of the defect excitation point.

[0045] This study utilizes electromagnetic resonance characteristics to induce resonance in defects, generating resonant excitation points. By leveraging the obtained electromagnetic resonance characteristics of the defect excitation point, a specialized resonance induction scheme is designed to enhance the defect detection signal. The dominant resonant modes in the electromagnetic resonance characteristics are analyzed to determine the most sensitive resonant frequency and corresponding field distribution mode. A resonant coupling structure is designed at the dominant resonant frequency, using structures such as microstrip lines, helical antennas, or resonant rings to form a strong coupling with the defect excitation point. The coupling coefficient k = M / √(L1L2) is calculated, where M represents mutual inductance, and L1 and L2 represent the self-inductance of the two resonant structures. The geometric parameters and spatial position of the coupling structure are optimized to maximize coupling efficiency and resonance enhancement. A three-dimensional full-wave analysis is established using electromagnetic field numerical calculations to study the electromagnetic field distribution and energy transfer mechanism during the resonance induction process. Selective excitation of the defect resonance mode is achieved by adjusting the amplitude, phase, and frequency of the excitation signal. A control strategy for resonance induction is established, dynamically adjusting the excitation parameters based on real-time measured resonant responses. Under optimal resonance induction conditions, the electromagnetic field intensity in the defect region is significantly enhanced, forming a local high-field-intensity region. These high field strength regions constitute resonant excitation points, significantly enhancing the concentration of electromagnetic field energy at the defect location.

[0046] Echo excitation waves are generated by echo amplifying the swept frequency signal through the resonant excitation point. At the identified resonant excitation point, its enhanced electromagnetic field characteristics are used to amplify the injected swept frequency signal. The high Q-value of the resonant excitation point significantly amplifies the signal near the resonant frequency, while suppressing signals deviating from the resonant frequency. An adaptive swept frequency strategy is designed to adjust the frequency step and residence time of the swept frequency signal according to the resonant characteristics of the resonant excitation point. A fine frequency step and a longer residence time are used near the resonant frequency to ensure sufficient excitation of the resonant response. Multiple reflection paths are formed around the resonant excitation point using delay lines and reflection structures to extend the signal's residence time in the resonant region. A mathematical description of echo amplification is established, considering factors such as multiple reflections, resonance enhancement, and loss attenuation. The echo amplification coefficient A = |H(f)| is calculated. 2 Here, H(f) represents the transfer function containing multiple reflections. Autocorrelation and cross-correlation analysis techniques are used to extract useful defect information from complex echo signals. A time-delay spectrum analysis of the echo signal is established to identify the time delay and amplitude characteristics corresponding to different reflection paths.

[0047] Defect response characteristics are generated based on echo excitation waves. The echo-amplified excitation wave signal is used to acquire corresponding response signals at various key nodes of the BMS circuit, establishing complete defect response characteristic data. A high-sampling-rate digital oscilloscope and a multi-channel data acquisition system are used to acquire the response signals, ensuring the capture of subtle changes in the response signals. The transfer relationship between the echo excitation wave and the response signal is established, and the parameters of the transfer function are determined using a system identification method. The amplitude, phase, frequency, and time delay characteristics of the response signal are analyzed to extract characteristic parameters related to the defect. The harmonic distortion (THD) of the response signal is calculated as √(V²). 2 +V3 2 +...+V n 2 ) / V1, where V1 represents the fundamental amplitude, and V2, V3, etc. represent the harmonic amplitudes, V n This represents the amplitude of the nth harmonic. Spectral analysis is used to identify characteristic frequency components in the response signal, establishing a correspondence between defect characteristic frequencies and physical defects. A classification system for defect response characteristics is established, categorizing defects into multiple classes based on different characteristics of the response signal. Machine learning techniques are used to train a defect identification program, establishing a mapping from response characteristics to defect type and severity. A standardized representation of defect response characteristics is constructed, incorporating multi-dimensional information such as amplitude spectrum, phase spectrum, time-domain waveform, and statistical characteristics. The generation of defect response characteristics completes the conversion process from echo excitation waves to identifiable characteristic signals.

[0048] The defect response characteristics are resonantly amplified to generate an enhanced detection domain. The resonant frequencies in the defect response characteristics are analyzed, corresponding to the inherent vibration modes of the defect structure. A resonant amplifier circuit is designed, using a high-Q LC resonant circuit or a quartz crystal resonator, to tune the resonant frequency of the amplifier circuit to the characteristic frequency of the defect. Lock-in amplification technology is employed, using a reference signal and the defect response signal for coherent detection, effectively suppressing noise and amplifying the useful signal. An adaptive filter is established, dynamically adjusting the filter parameters according to the spectral distribution of the defect response characteristics to maximize the signal-to-noise ratio. Digital signal processing techniques are used to further enhance the amplified signal, including frequency domain filtering, time domain averaging, and correlation detection. The spatial distribution of the enhanced detection domain is calculated, mapping the amplified defect response signal into the three-dimensional space of the BMS board. The enhanced detection domain displays the intensity distribution and influence range of the defect signal, clearly showing the signal enhancement effect at the defect location. Quantitative indicators for the enhanced detection domain are established, including parameters such as signal enhancement factor, detection sensitivity, and spatial resolution. Resonant amplification converts the weak defect signal into a strong, detectable signal.

[0049] Step S130: Gradient analysis is performed on the thermal infrared distribution signal to generate a thermal flow topology field. Abnormal heat source locations are extracted from the thermal flow topology field. The abnormal heat source locations are geometrically correlated with the circuit trace information to generate thermoelectric coupling coordinates. Impedance mapping is performed based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain.

[0050] Specifically, a heat flux topology field is generated by gradient analysis of the thermal infrared distribution signal. The thermal infrared distribution signal acquired by the BMS board is rearranged according to pixel coordinates to establish a two-dimensional temperature distribution matrix T(x,y). The temperature gradient field ∇T=(∂T / ∂x,∂T / ∂y) is calculated, where ∇T represents the temperature gradient vector, and ∂T / ∂x and ∂T / ∂y represent the rates of temperature change in the x and y directions, respectively. The central difference scheme is used for gradient calculation to ensure the accuracy and stability of the gradient calculation. The magnitude and direction of the gradient vector are analyzed. The gradient magnitude reflects the drasticness of temperature change, and the gradient direction indicates the direction of heat flow. The heat flux density equation q=-k∇T is established, where q represents the heat flux density vector and k represents the thermal conductivity. This equation describes the relationship between heat flow and temperature gradient. The heat flux topology field is constructed through the spatial distribution of the heat flux density vector. The heat flux topology field visually displays the heat transfer paths and heat flow concentration areas inside the BMS board.

[0051] Extracting anomalous heat source locations from the heat flux topology. Identifying regions of abnormally concentrated heat flux density within the generated heat flux topology, these regions correspond to heat sources and heat dissipation bottlenecks in the circuit. Calculating the heat flux divergence ∇·q, where positive values ​​indicate heat source locations and negative values ​​indicate heat sink locations. A threshold-based method is used to identify anomalous heat sources; when the heat flux divergence exceeds the statistical threshold under normal operating conditions, the location is marked as an anomalous heat source. Analyzing the spatial distribution characteristics of anomalous heat sources, including their geometry, area, and thermal intensity. Calculating the thermal power of an anomalous heat source P = ∫∫(∇·q)dA, where P represents thermal power and A represents the area of ​​the heat source region; the total thermal power of the heat source is calculated by integration. Establishing a classification system for anomalous heat sources, categorizing them into point heat sources, line heat sources, and area heat sources based on their power level and spatial size. Using connected component analysis, adjacent anomalous heat sources are clustered to identify heat source clusters and independent heat sources. The precise extraction of abnormal heat source locations enables spatial localization from the heat flow field to the specific heat source.

[0052] Thermoelectric coupling coordinates are generated by geometrically associating the location of abnormal heat sources with circuit trace information. A unified coordinate system for the BMS board is established, mapping the locations of abnormal heat sources and circuit trace information to the same spatial reference frame. The spatial distance between the abnormal heat source location and the circuit trace is calculated, and the nearest neighbor relationship is determined using the Euclidean distance formula. When the distance between the abnormal heat source location and the circuit trace is less than a set threshold, a geometric association relationship is established. The geometric characteristics of the circuit trace are analyzed, including trace width, trace length, via location, and pad distribution. A heat source-trace association matrix is ​​established, where matrix elements represent the association strength between the abnormal heat source and a specific circuit trace. The calculation of association strength comprehensively considers factors such as spatial distance, thermal power, and trace current carrying capacity. Thermoelectric coupling coordinates are calculated using a weighted average method, and the coordinate values ​​reflect the coupling position between the heat source and the circuit trace. The generation of thermoelectric coupling coordinates establishes a precise spatial correspondence between thermal phenomena and circuit structure.

[0053] In some embodiments, the step of generating a defect impedance distribution domain based on the thermoelectric coupling coordinates by impedance mapping includes: identifying high-resistance regions by measuring circuit trace impedance based on the thermoelectric coupling coordinates; evaluating the solder joint contact state using the high-resistance regions to form a solder joint quality distribution; calibrating the solder joint quality distribution by resistance values ​​to generate a calibration resistance matrix; and spatially mapping the calibration resistance matrix to generate a defect impedance distribution domain.

[0054] This method uses thermoelectric coupling coordinates to measure circuit trace impedance and identify high-resistance regions. Miniature impedance measurement probes with nanoscale tips are positioned at designated locations on the thermoelectric coupling coordinates, enabling precise resistance measurements at the micrometer scale. A constant current source injects a microamp-level test current while simultaneously measuring the corresponding voltage drop, calculating the local resistance value R=V / I. Dense sampling is performed at different locations along the circuit trace, with the sampling interval set to 1 / 10 of the trace width to ensure the capture of minute impedance changes. A spatial distribution map of the circuit trace impedance is established, with the horizontal axis representing the location along the trace and the vertical axis representing the impedance value. Abnormal peak values ​​in the impedance distribution map are analyzed; when the local impedance exceeds twice the normal impedance value, the area is marked as a high-resistance region. The geometric parameters of the high-resistance region are calculated, including the length, width, and peak impedance of the high-resistance segment. Statistical analysis methods are used to establish the distribution patterns of the high-resistance region, identifying its clustering and dispersion characteristics. The identification of high-resistance regions reveals locations of abnormal impedance concentrations in the circuit trace, which typically correspond to poor contact or material defects.

[0055] Solder joint contact status is evaluated using high-resistance regions to form a solder joint quality distribution. The identified high-resistance regions are spatially matched with the solder joint distribution map of the BMS board to analyze the correspondence between high-resistance regions and solder joint locations. An evaluation index system for solder joint contact status is established, including three dimensions: contact resistance, contact area, and mechanical strength. Ultrasonic testing technology is used to analyze the internal structure of solder joints within high-resistance regions, identifying defects such as voids, cracks, and delamination. The contact resistance of solder joints is measured; the contact resistance of normal solder joints is typically in the milliohm range, while the contact resistance of defective solder joints can reach the ohm range. A solder joint quality rating standard is established, classifying solder joints into four levels: excellent, good, average, and poor. The solder joint quality index Q = (R_normal / R_actual) × (A_actual / A_design) is calculated, where R_normal represents normal contact resistance, R_actual represents actual contact resistance, A_actual represents actual contact area, and A_design represents designed contact area. A color-coding technique is used to create a solder joint quality distribution map, with different colors representing different quality levels.

[0056] A calibration resistance matrix is ​​generated by calibrating the resistance values ​​of solder joint quality distribution. Based on the evaluation results of the solder joint quality distribution, a quantitative relationship between solder joint quality grade and resistance value is established. Standard resistors are used to calibrate the resistance of solder joints of different quality grades, establishing quality-resistance calibration curves. The calibration resistance value for excellent-grade solder joints is set to the design resistance value, for good-grade to 1.2 times the design value, for average-grade to 1.5 times the design value, and for poor-grade to more than twice the design value. A calibration resistance matrix R_matrix is ​​established, where the row index represents the x-coordinate of the solder joint, the column index represents the y-coordinate, and the matrix elements represent the calibration resistance value at the corresponding position. Data filling is performed on the calibration resistance matrix, filling the regions between solder joints with resistance values ​​using linear interpolation. Statistical characteristics of the calibration resistance matrix are calculated, including parameters such as the mean resistance, standard deviation, and extreme value distribution. A data structure for the calibration resistance matrix is ​​established, employing a sparse matrix storage format to reduce memory usage.

[0057] A defect impedance distribution domain is generated by spatially mapping the calibration resistor matrix. Discrete resistance values ​​in the calibration resistor matrix are mapped to the continuous three-dimensional space of the BMS board, establishing a spatial distribution function of impedance. Cubic spline interpolation is used to spatially expand the calibration resistor matrix, generating a smooth impedance distribution surface. The gradient field of the impedance distribution is calculated; the gradient magnitude reflects the drastic change in impedance, and the gradient direction indicates the direction of impedance increase. Impedance isosurfaces are established, connecting spatial points with the same impedance value, clearly showing the hierarchical structure of the impedance distribution. Volume rendering technology is used to generate a three-dimensional visualization image of the defect impedance distribution domain, with different colors and transparency representing different impedance levels. The volume distribution of the defect impedance distribution domain is calculated, and the spatial volume proportion within different impedance ranges is statistically analyzed. A mathematical description of the defect impedance distribution domain is established, and a polynomial function is used to fit the spatial variation law of the impedance distribution. The generation of the defect impedance distribution domain completes the transformation from two-dimensional solder joint information to a three-dimensional spatial impedance field, clearly showing the abnormal impedance distribution pattern inside the BMS board.

[0058] Step S140: Spectral decomposition is performed on the enhanced detection domain to identify the defect resonant frequency point. Multi-frequency superposition excitation is performed at the defect resonant frequency point to generate a superimposed excitation field. Phase weaving is performed on the superimposed excitation field to construct a phase detection network. Harmonic crosstalk is mined based on the phase detection network and the defect impedance distribution domain to generate a defect location matrix.

[0059] Specifically, the enhanced detection domain is decomposed to identify the resonant frequency points of the defect. The enhanced detection domain signal generated in the previous steps is input into the spectrum analysis system, and the signal is transformed into the frequency domain using Fast Fourier Transform (FFT). The enhanced detection domain signal contains rich frequency components, where the energy concentration at specific frequencies corresponds to the resonant characteristics of the defect. The power spectral density PSD = |X(f)| is calculated. 2 Here, X(f) represents the frequency domain representation of the signal, and the power spectral density reflects the energy distribution of each frequency component. Peak detection technology is used to identify significant peaks in the power spectrum, which correspond to the characteristic resonant frequencies of the defect. A screening criterion for resonant frequencies is established: when the peak amplitude exceeds 10 times the background noise and the peak bandwidth is less than 5% of the center frequency, the frequency is marked as a defect resonant frequency. The distribution pattern of defect resonant frequencies is analyzed to identify the relationship between the fundamental frequency and each harmonic frequency. The quality factor Q = f0 / Δf of the resonant frequency is calculated, where f0 represents the resonant frequency and Δf represents the half-power bandwidth; a high Q value indicates that the defect has strong resonant characteristics. A frequency library of defect resonant frequencies is established, containing the center frequency, amplitude, bandwidth, and phase information of each resonant frequency.

[0060] A multi-frequency superimposed excitation field is generated at the defect resonant frequency. Based on the identified defect resonant frequency, a multi-frequency synchronous excitation scheme is designed, injecting excitation signals simultaneously at each resonant frequency. A multi-channel signal generator is used to generate sinusoidal signals of different frequencies, with each signal's frequency precisely set to the center frequency of the defect resonant frequency. The superimposed waveform of the multi-frequency signals is calculated as s(t) = ΣᵢAᵢsin(2πfᵢt + φᵢ), where Aᵢ represents the amplitude of the i-th frequency component, fᵢ represents the frequency, and φᵢ represents the initial phase. The amplitude and phase relationship of each frequency component is optimized to maximize the excitation effect at the defect location. Phase-locked loop technology is used to ensure phase synchronization between multiple frequency components, avoiding excitation effect attenuation caused by phase drift. Superimposed excitation signals are simultaneously injected at multiple locations on the BMS board, forming a spatially distributed excitation field. The amplitude and phase response of the superimposed excitation field at different locations are monitored to establish a spatial distribution map of the excitation field. The intensity of the superimposed excitation field reaches its peak at the defect location and remains at a low level in the normal region, thus achieving selective excitation of the defect.

[0061] A phase detection network is constructed by phase weaving the superimposed excitation field. Phase detection probes are arranged in a grid pattern on the BMS board surface, with the probe spacing set to 1 / 4 of the signal wavelength to ensure spatial resolution of the phase detection. Each probe measures the phase information of the superimposed excitation field at that location, and a lock-in amplifier is used to achieve high-precision phase measurement. The topology of the phase detection network is established, connecting the phase relationships between adjacent probes with directed edges to form a complete network graph. The phase difference Δφ = φⱼ - φᵢ between adjacent probes is calculated; the phase difference reflects the phase change of the signal during spatial propagation. Phase unwrapping techniques are used to handle phase jump problems, ensuring the continuity of phase measurement. A mathematical description of the phase detection network is established, using an adjacency matrix to represent the network connections and a weight matrix to represent the phase difference information. The propagation characteristics of the phase detection network are analyzed, identifying the main propagation paths and propagation delays.

[0062] In some embodiments, the step of generating a defect location matrix by mining harmonic crosstalk based on the phase detection network and the defect impedance distribution domain includes: performing phase difference analysis on the phase detection network to identify phase jump points; determining resistance abrupt change regions in the defect impedance distribution domain based on the phase jump points; performing current path tracing on the resistance abrupt change regions to form crosstalk conduction paths; and performing coordinate calibration on the crosstalk conduction paths to construct a defect location matrix.

[0063] Phase difference analysis is performed on the phase detection network to identify phase transition points. Within the constructed phase detection network, the phase difference variation between adjacent probes is analyzed one by one. The first derivative dΔφ / dx and the second derivative d... 2 Δφ / dx 2The first derivative reflects the rate of phase change, and the second derivative reflects the acceleration of phase change. A criterion for phase jumps is established: when the first derivative of the phase difference exceeds π / 4 rad / mm and the second derivative shows a sign change, the location is marked as a phase jump point. A sliding window technique is used to smooth the phase difference sequence, with the window size set to three times the probe spacing to eliminate the influence of measurement noise. The spatial distribution characteristics of phase jump points are analyzed to identify isolated jump points and continuous jump regions. The amplitude and direction of the phase jump are calculated; the jump amplitude reflects the severity of the defect, and the jump direction indicates an abnormal signal propagation pattern. A classification system for phase jump points is established, dividing jump points into three categories based on jump amplitude: strong jumps, medium jumps, and weak jumps. Strong jump points usually correspond to severe circuit defects, medium jump points to minor defects, and weak jump points may be measurement errors or environmental interference. Successful identification of phase jump points locates abnormal nodes in the phase detection network.

[0064] For example, determining the resistance abrupt change region in the defect impedance distribution domain based on the phase jump point includes: using the phase jump point as the search center to detect resistance values ​​in the surrounding area to obtain resistance distribution data; analyzing the impedance change rate based on the resistance distribution data to generate a change rate distribution; continuously narrowing the search range using the change rate distribution to improve positioning accuracy; and determining the search area as the resistance abrupt change region when the positioning accuracy meets the requirements.

[0065] Resistance distribution data is acquired by detecting resistance values ​​in the surrounding area using the phase transition point as the search center. A polar coordinate search grid is established with the identified phase transition point coordinates as the search origin, with a radial search step size of 0.1 mm and an angular step size of 15 degrees. Precision resistance measurements are performed at each grid point using a four-probe microresistance measuring device, with the measurement current set to the microampere level to avoid affecting the circuit. A data acquisition protocol for resistance distribution is established, including the measurement sequence, measurement parameters, and data recording format. The resistance distribution data includes the coordinates, resistance value, measurement time, and ambient temperature of each measurement point. Interpolation methods are used to fill in the resistance values ​​at locations not directly measured, generating a continuous resistance distribution function. The statistical characteristics of the resistance distribution data are analyzed, calculating the mean, variance, maximum, and minimum resistance values. A contour map of the resistance distribution is created, connecting points with the same resistance value to visually display the spatial variation pattern of the resistance.

[0066] The impedance change rate is analyzed based on resistance distribution data to generate a rate of change distribution. The spatial gradient of the resistance distribution data is calculated using numerical differentiation methods, with a central difference scheme employed to ensure accuracy. The radial impedance change rate is calculated as dR / dr = (R(r+Δr)-R(r-Δr)) / (2Δr), where R(r) represents the resistance value at a distance r from the origin, and Δr represents the radial step size. The angular impedance change rate is calculated as dR / dθ = (R(θ+Δθ)-R(θ-Δθ)) / (2Δθ), where θ represents the angular coordinate, and Δθ represents the angular step size. A vector representation of the impedance change rate is established as ∇R = (dR / dr, dR / dθ), where the magnitude of the vector represents the intensity of the change rate, and the direction of the vector represents the direction of maximum change. The distribution characteristics of the impedance change rate are analyzed to identify extreme points and zero points. Extreme points correspond to the boundary positions of abrupt changes in resistance, while zero points correspond to the stable regions of the resistance distribution. High-rate-of-change regions are identified using a threshold segmentation method. Regions with a rate of change exceeding the mean plus twice the standard deviation are marked as high-rate-of-change regions. A visual representation of the rate of change distribution is established, using a vector field plot to display the magnitude and direction of the rate of change.

[0067] The search range is continuously narrowed using the rate of change distribution to improve positioning accuracy. Based on the analysis of the rate of change distribution, the region with the largest rate of change is identified as the new search center. An adaptive search strategy is adopted, dynamically adjusting the search range and sampling density according to the current positioning accuracy. When the rate of change distribution shows a clear directionality, the search range is narrowed along the direction of the largest rate of change. A shrinkage criterion for the search range is established, reducing the search range to 70% of the previous size each time, while increasing the sampling density by 50%. A multi-scale analysis method is used to analyze the rate of change distribution at different spatial scales, from the overall distribution at a coarse scale to the local features at a fine scale. A positioning accuracy evaluation index is established: positioning accuracy η = σ_target / σ_current, where σ_target represents the target accuracy and σ_current represents the standard deviation of the current measurement. When the positioning accuracy is less than a set threshold, the process of narrowing the search range stops. An iterative optimization method is used to continuously improve the search strategy, with each iteration improving the positioning accuracy.

[0068] When the positioning accuracy meets the requirements, the search area is identified as a resistance abrupt change region. A criterion for judging positioning accuracy is established: if the change in positioning accuracy after three consecutive searches is less than 5% and the absolute positioning error is less than 0.05 mm, the positioning accuracy is considered to meet the requirements. Statistical tests are used to evaluate the reliability of the search results, calculating the confidence interval and significance level. The geometric characteristics of the final search area are analyzed, including parameters such as area, perimeter, major axis direction, and shape factor. A boundary description of the resistance abrupt change region is established, and B-spline curves are used to fit the region boundary to ensure its smoothness and continuity. Resistance distribution statistics within the resistance abrupt change region are calculated, including maximum resistance value, minimum resistance value, peak resistance gradient, and abrupt change amplitude. A quality evaluation system for the resistance abrupt change region is established, comprehensively considering factors such as abrupt change intensity, region size, and boundary clarity. The identified resistance abrupt change region information is recorded in a database, including region coordinates, geometric parameters, resistance characteristics, and determination time.

[0069] Crosstalk conduction paths are formed by tracing current paths through the resistance abrupt change region. Within the defined resistance abrupt change region, the actual current conduction path and mechanism are analyzed. Circuit network equations are established using Ohm's law and Kirchhoff's laws to calculate the current distribution along different paths. The high impedance characteristics of the resistance abrupt change region force the current to choose alternative conduction paths, forming an abnormal current distribution pattern. An optimization criterion for the current path is established: the current always chooses the path with the least impedance for conduction. Dijkstra's shortest path search is used to find the optimal current conduction path, with path weights set to the resistance values ​​of the corresponding line segments. The bifurcation and merging phenomena of the current paths are analyzed to identify the primary and secondary conduction paths. The current distribution ratio of each conduction path is calculated, with the primary path carrying the majority of the current and the secondary paths carrying smaller currents. A mathematical model of the crosstalk conduction path is established, and the path topology is represented using graph theory. The propagation characteristics of the crosstalk signal along the conduction path are analyzed, including parameters such as propagation delay, signal attenuation, and frequency response.

[0070] A defect location matrix is ​​constructed by performing coordinate calibration on the crosstalk propagation path. The tracked crosstalk propagation path is precisely calibrated in the coordinate system of the BMS board, establishing a spatial coordinate description of the path. GPS coordinate calibration technology is used to determine the precise location of key nodes on the path, achieving micrometer-level accuracy. A data structure for the path coordinates is established, including the coordinate information of the path's start point, end point, intermediate nodes, and turning points. Geometric parameters of the path are calculated, including total path length, radius of curvature, and turning angles. The intersection points of multiple crosstalk propagation paths are analyzed; these intersection points typically correspond to the locations of critical defects in the circuit. A mathematical framework for the defect location matrix is ​​established, with the matrix dimension set to the pixel resolution of the BMS board. The defect probability value P(i,j)=Σ for each matrix element is calculated. k w k δk (i,j), where w k δ represents the weight of the k-th path. k (i,j) represents the indicator function of the path passing through position (i,j). Normalization is used to map the defect probability value to the 0-1 interval, facilitating the quantitative assessment of defect severity. A visual representation of the defect location matrix is ​​established, using a heatmap to display the spatial distribution of defect probabilities. The construction of the defect location matrix completes the quantitative mapping from crosstalk paths to precise defect locations.

[0071] Step S150: The defect location matrix is ​​analyzed by impedance gradient to generate a gradient tracking field. The field strength distribution around the defect is reconstructed based on the gradient tracking field to generate the defect electric field topology. The defect electric field topology is used to locate the peak field strength to generate the precise coordinates of the defect.

[0072] Specifically, the defect location matrix is ​​analyzed using impedance gradient analysis to generate a gradient tracking field. The defect location matrix constructed in the previous steps is numerically differentiated to calculate the spatial gradient of each element. The Sobel operator is used to calculate the x-direction gradient ∇_xP and y-direction gradient ∇_yP of the defect location matrix, where P represents the probability value of the defect location matrix. A gradient vector field G = (∇_xP, ∇_yP) is established, where the magnitude of the vector represents the drastic change in defect probability, and the direction of the vector indicates the direction of the fastest increase in defect probability. The gradient magnitude |G| = √(∇_xP) is calculated. 2 +(∇_yP) 2 The gradient direction angle θ = arctan(∇_yP / ∇_xP) is used. Gaussian filtering is applied to smooth the gradient field, eliminating noise and discretization errors. The streamline equations for the gradient tracking field are established, with streamlines extending along the gradient direction and connecting regions with similar gradient magnitudes. The divergence ∇·G and curl ∇×G of the gradient tracking field are calculated; the divergence reflects the convergence and divergence characteristics of the gradient, and the curl reflects the rotational characteristics of the gradient.

[0073] In some embodiments, the step of reconstructing the field strength distribution around the defect based on the gradient tracking field to generate the defect electric field topology includes: identifying the field strength polarity based on the gradient tracking field to generate a positive field strength region and a negative field strength region; using the positive field strength region to perform field strength balancing processing on the negative field strength region to form a uniform field strength region; performing topology reconstruction on the uniform field strength region to generate topological structural units; and performing geometric optimization on the topological structural units to construct the defect electric field topology.

[0074] Field strength polarity identification based on gradient tracking field generates positive and negative field strength regions. The divergence characteristics of the gradient vector in the gradient tracking field are analyzed, and the divergence value ∇·G at each spatial point is calculated. A positive divergence value indicates that the region is the source point of the gradient, corresponding to a positive field strength region. A negative divergence value indicates that the region is the sink point of the gradient, corresponding to a negative field strength region. A threshold for determining field strength polarity is established; when the absolute value of the divergence value exceeds the mean plus one standard deviation, polarity is marked. Connectivity analysis is used to cluster adjacent regions of the same polarity, forming continuous positive and negative field strength regions. Geometric parameters of each polarity region are calculated, including region area, perimeter, centroid coordinates, and orientation angle. The spatial distribution patterns of positive and negative field strength regions are analyzed to identify the symmetry and periodicity characteristics of the polarity regions. Intensity levels of the polarity regions are established, classifying regions into three levels—strong polarity, medium polarity, and weak polarity—based on the magnitude of the divergence value. Strong polarity regions typically correspond to the core of a defect, medium polarity regions correspond to the extent of a defect's influence, and weak polarity regions correspond to the edge of a defect.

[0075] A uniform field strength region is formed by balancing the negative field strength region using a positive field strength region. The spatial distribution relationship between the positive and negative field strength regions is analyzed to identify matching positive and negative polarity pairs. The field strength balance condition between the positive and negative field strength regions is calculated, and the balance equation Σq_positive + Σq_negative = 0 is established using the law of charge conservation. A field strength balance adjustment strategy is designed to achieve overall field strength balance by adjusting the intensity of the polarity regions. An optimization objective function for field strength balance is established: F = Σ(E_i - E_target). 2 Here, E_i represents the electric field strength value of the i-th region, and E_target represents the target equilibrium electric field strength. The optimal solution for electric field balance is obtained using the Lagrange multiplier method to determine the adjustment magnitude and direction for each region. During the balancing process, the physical rationality of the electric field distribution is maintained, avoiding non-physical abrupt changes in electric field strength. An evaluation index for a uniform electric field region is established, calculating the uniformity of the electric field distribution U = 1 - σ / μ, where σ represents the standard deviation of the electric field strength, and μ represents the mean of the electric field strength. When the uniformity exceeds 0.8, a satisfactory electric field balance effect is considered achieved. The formation of a uniform electric field region eliminates the interference of polarity imbalance on the electric field topology construction.

[0076] Topological reconstruction of uniform field intensity regions generates topological structural units. Topological analysis is performed on the resulting uniform field intensity regions to identify internal connectivity and structural features. The uniform field intensity regions are discretized into triangular meshes using the Delaunay triangulation method, with each triangle serving as a basic topological unit. Adjacency relationships between topological structural units are established, constructing a connection graph between units. The types of topological structural units are analyzed, categorized into isolated, connected, and hub types based on their connectivity. Isolated units have low connectivity and are typically located at the edges of the field intensity distribution. Connected units have moderate connectivity and constitute the main body of the topology. Hub units have high connectivity and are key nodes in the topological network. The geometric characteristics of the topological structural units are calculated, including unit area, shape factor, and orientation angle. A weighting mechanism for the topological structural units is established, with weights determined based on the internal field intensity distribution and connectivity importance. Topology simplification techniques are used to merge similar topological structural units, reducing topological complexity while preserving key structural features.

[0077] Geometric optimization of topological structural elements is used to construct the defect electric field topology. Geometric optimization methods are employed to improve the shape and distribution of the topological structural elements, enhancing the accuracy and stability of the topological description. An objective function for geometric optimization is established, comprehensively considering the regularity of element shape, uniformity of distribution, and rationality of connections. Mesh optimization techniques are used to adjust the vertex positions of the topological structural elements, improving their geometric quality. Geometric quality indices of the elements are calculated, including parameters such as aspect ratio, interior angle deviation, and area ratio. Mesh smoothing techniques are used to eliminate jagged boundaries and irregular protrusions in the topology. Optimization criteria for topological connections are established, adjusting the connection relationships between elements to improve the overall structure of the topology. Topology refinement techniques are used to increase the density of topological elements in regions with large field strength gradients, improving the accuracy of local descriptions. A quality evaluation system for the defect electric field topology is established, including evaluation indicators such as topological integrity, continuity, and accuracy. Visualization techniques are used to display the optimized defect electric field topology, employing color coding and contour lines to show the spatial distribution of field strength. The construction of the defect electric field topology completes the geometric reconstruction from discrete elements to a continuous field distribution, accurately describing the electric field structure characteristics around the defects.

[0078] This method utilizes the defect electric field topology to generate precise defect coordinates by locating peak electric field strength. Global maxima of the electric field strength are searched within the constructed defect electric field topology; these points correspond to the most probable defect locations. A gradient ascent method is used to generate precise defect coordinates starting from multiple initial points, ensuring that a global optimum is found rather than a local extremum. The magnitude of the electric field strength and spatial coordinates of each maxima are calculated, creating a list of candidate defect locations. The stability of the maxima is analyzed by calculating the eigenvalues ​​of the Hessian matrix of the electric field strength; positive eigenvalues ​​indicate stable maxima. A weighted average method is used to fuse the coordinate information of multiple maxima, with the weights set to the corresponding electric field strengths. Confidence intervals for the precise defect coordinates are calculated to assess the uncertainty in coordinate determination. A coordinate accuracy evaluation standard is established, and the reliability of the coordinates is evaluated using cross-validation. Sub-pixel interpolation is used to improve the accuracy of coordinate localization, achieving an interpolation accuracy of 1 / 10 of the original grid spacing.

[0079] Step S160: Based on the precise coordinates of the defects, perform layered annotation to extract the defect type and severity level, and complete the intelligent detection of defects on the BMS board.

[0080] In some embodiments, the step of extracting the defect type and severity level by hierarchical annotation based on the precise coordinates of the defect includes: constructing a core damage region based on the precise coordinates of the defect; performing chain reaction analysis on the core damage region to obtain the sweep coefficient; obtaining a tiered damage recursive chain based on the sweep coefficient; and generating the defect type and severity level based on the tiered damage recursive chain.

[0081] A core damage region is constructed based on the precise coordinates of the defect. The precise coordinates of the defect are used as the center point for damage analysis, and a spatial description of the damage region is established around this point. A radial expansion method is used to progressively expand the search radius outward from the precise coordinates of the defect, identifying the surrounding area affected by the defect. The initial search radius is set to 0.5 mm, and the expansion range is dynamically adjusted according to the influence intensity of the defect. Changes in various electrical parameters are monitored during the expansion process; when a parameter deviates from its normal value by more than 10%, the affected area is included in the damage range. A quantitative description of the damage degree is established, calculating the damage intensity D(r) = D0 × exp(-r / λ), where D0 represents the damage intensity at the defect center, r represents the radial distance from the defect center, and λ represents the damage attenuation length constant. The geometry of the damage region is analyzed; most defect damage regions are elliptical, with the major axis usually aligned with the current conduction direction. The boundary contour of the core damage region is calculated, and spatial points with the same damage intensity are connected using equal damage lines. A hierarchical system for the damage region is established, dividing the region into three levels: a severely damaged core region, a moderately damaged transition region, and a lightly damaged peripheral region. The construction of the core damage region completes the precise spatial definition of the defect's influence range.

[0082] Chain reaction analysis was performed on the core damage region to obtain the sweep efficiency. Around the constructed core damage region, a ring scan method was used to gradually expand the detection range outwards, measuring the actual affected area of ​​the chain reaction. The ring scan started from the boundary of the core damage region and expanded outwards by 0.2 mm each time, with multiple detection points arranged on each scan ring to measure electrical parameters. Detection parameters included key indicators such as local resistance, voltage distribution, and current density. When the change in detection parameters was less than 5%, the chain reaction effect was considered to have ended, and this scan ring was defined as the outer boundary of the affected area. The area of ​​the affected region, A_affected, was calculated using numerical integration methods for irregular affected areas. A grid system for area calculation was established, with a grid size of 0.1 mm × 0.1 mm. The number of affected pixels within the grid was counted and converted into the actual area. The area of ​​the core damage region, A_core, was measured. The core damage region was defined as the area where the electrical parameters deviated from normal values ​​by more than 20%. Boundary tracing technology was used to accurately determine the contour of the core damage region, and the actual area enclosed by the contour was calculated. The sweepability coefficient W = A_affected / A_core is calculated, which quantifies the ability of defect damage to spread outward from the core region. The sweepability coefficient typically ranges from 1.5 to 8.0, with higher values ​​indicating stronger diffusion. A grading standard for the sweepability coefficient is established: W < 2.0 indicates limited impact, 2.0 ≤ W < 4.0 indicates moderate diffusion impact, and W ≥ 4.0 indicates strong diffusion impact. Obtaining the sweepability coefficient provides a quantitative description of the defect's impact range.

[0083] A tiered damage recursive chain is obtained based on the sweep efficiency W. Using the obtained sweep efficiency W, a mathematical description and propagation path of the damage propagating step-by-step in space are established. The damage propagation process is discretized into multiple propagation levels, each corresponding to a different spatial distance and damage intensity. A recursive equation for tiered damage is established: D_{n+1}=W×D_n×α_n, where D_n represents the damage intensity of the nth level, D_{n+1} represents the damage intensity of the next level, and α_n represents the propagation attenuation coefficient of the nth level. The propagation attenuation coefficient α_n is determined based on the material properties and geometry of the propagation path, typically ranging from 0.6 to 0.9. The spatial distribution of damage intensity at each level is calculated; the damage intensity of the first level is equal to the average damage intensity of the core damage area, and the damage intensity of subsequent levels decreases sequentially. A termination condition for damage propagation is established: the propagation chain is considered terminated when the damage intensity of a certain level falls below a detection threshold or the number of propagation levels exceeds a preset upper limit. The spatial topology of the tiered damage recursive chain is analyzed, and the direction and intensity distribution maps of damage propagation are plotted. Identify key propagation paths in the propagation chain; these paths typically extend along the direction of current conduction with the lowest resistance. Calculate characteristic parameters of the recursive chain, including chain length L, total propagation distance R, and average propagation intensity I. Establish a data structure for the recursive chain, recording the spatial coordinates, damage intensity, and propagation direction information for each stage. Obtaining the tiered damage recursive chain reveals the complete propagation process and impact path of defect damage.

[0084] Defect types and severity levels are generated based on the tiered damage propagation chain. An automatic defect type identification mechanism is established by analyzing the characteristic patterns of the tiered damage propagation chain. Different types of defects have unique damage propagation patterns and propagation chain characteristics. Solder joint cracking defects exhibit a short-range strong impact, with few propagation stages but high damage intensity at each stage. Conductor breakage defects exhibit a long-range weak impact, with many propagation stages but low damage intensity at each stage. Insulation degradation defects exhibit a diffuse impact, with a radial propagation direction. Device aging defects exhibit a progressive impact, with damage intensity increasing slowly over time. A feature vector F=[L,I,D,S] is established for each defect type, where L represents the propagation chain length, I represents the average damage intensity, D represents the propagation directionality, and S represents the propagation speed. A support vector machine classifier is used to automatically identify defect types based on the feature vector. A severity level assessment system is established, comprehensively considering factors such as the area of ​​the core damage region, the magnitude of the sweep coefficient, and the length of the propagation chain. The Comprehensive Severity Index (CSI) is calculated as (A_core × W × L) / (A_total × W_max × L_max), where A_total, W_max, and L_max represent the maximum possible values ​​of the corresponding parameters. The CSI is mapped to four levels: CSI < 0.25 for minor, 0.25 ≤ CSI < 0.5 for moderate, 0.5 ≤ CSI < 0.75 for severe, and CSI ≥ 0.75 for extremely severe. The generation of defect types and severity levels completes the entire process of intelligent defect detection for BMS boards, realizing a complete technical chain from signal acquisition to defect diagnosis.

[0085] To implement the intelligent detection method for BMS board defects corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See [link to documentation]. Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a BMS board defect intelligent detection system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The BMS board defect intelligent detection system 200 provided in this embodiment includes:

[0086] Signal acquisition module 201 is used to acquire multi-domain abnormal signals and circuit routing information during the operation of BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. Frequency domain fusion is performed on the multi-domain abnormal signals to construct a circuit state spectrum library.

[0087] Excitation detection module 202 is used to identify abnormal impedance modes through the circuit state spectrum library, perform reverse excitation analysis on the abnormal impedance modes to generate defect excitation points, inject a frequency sweep signal into the defect excitation points to generate defect response features, and perform resonant amplification on the defect response features to generate an enhanced detection domain.

[0088] Thermoelectric analysis module 203 is used to perform gradient analysis on the thermal infrared distribution signal to generate a thermal flow topology field, extract the abnormal heat source location from the thermal flow topology field, geometrically correlate the abnormal heat source location with the circuit trace information to generate thermoelectric coupling coordinates, and perform impedance mapping based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain.

[0089] The spectrum localization module 204 is used to perform spectrum decomposition on the enhanced detection domain to identify the defect resonant frequency point, perform multi-frequency superposition excitation at the defect resonant frequency point to generate a superposition excitation field, perform phase weaving on the superposition excitation field to construct a phase detection network, and perform harmonic crosstalk mining based on the phase detection network and the defect impedance distribution domain to generate a defect localization matrix.

[0090] The precise positioning module 205 is used to perform impedance gradient analysis on the defect positioning matrix to generate a gradient tracking field, reconstruct the field strength distribution around the defect based on the gradient tracking field to generate the defect electric field topology, and use the defect electric field topology to perform field strength peak positioning to generate the precise coordinates of the defect.

[0091] The result output module 206 is used to extract the defect type and severity level by performing layered annotation based on the precise coordinates of the defect, and to complete the intelligent detection of defects on the BMS board.

[0092] The BMS board defect intelligent detection system 200 described above can implement the BMS board defect intelligent detection method of the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0093] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0094] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for intelligent detection of defects in BMS boards, characterized in that, include: Collect multi-domain abnormal signals and circuit routing information during the operation of the BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. Perform frequency domain fusion on the multi-domain abnormal signals to construct a circuit state spectrum library. Abnormal impedance modes are identified by the circuit state spectrum library, and reverse excitation analysis is performed on the abnormal impedance modes to generate defect excitation points. A frequency sweep signal is injected into the defect excitation points to generate defect response features, and the defect response features are resonantly amplified to generate an enhanced detection domain. Gradient analysis is performed on the thermal infrared distribution signal to generate a thermal flow topology field. Abnormal heat source locations are extracted from the thermal flow topology field. The abnormal heat source locations are geometrically correlated with the circuit trace information to generate thermoelectric coupling coordinates. Impedance mapping is performed based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain. The enhanced detection domain is subjected to spectral decomposition to identify the defect resonant frequency point. At the defect resonant frequency point, multi-frequency superposition excitation is performed to generate a superimposed excitation field. The superimposed excitation field is phase-woven to construct a phase detection network. Based on the phase detection network and the defect impedance distribution domain, harmonic crosstalk is mined to generate a defect location matrix. The defect location matrix is ​​analyzed by impedance gradient to generate a gradient tracking field. The field strength distribution around the defect is reconstructed based on the gradient tracking field to generate the defect electric field topology. The defect electric field topology is used to locate the field strength peak to generate the precise coordinates of the defect. Based on the precise coordinates of the defects, layered annotation is performed to extract the defect type and severity level, thus completing the intelligent detection of defects on the BMS board.

2. The method according to claim 1, characterized in that, The step of constructing a circuit state spectrum library by frequency domain fusion of the multi-domain anomalous signals includes: Identify impedance anomaly bands in the impedance spectrum data; Thermoelectric correlation points are generated by correlating the impedance anomaly frequency band with the thermal infrared distribution signal at temperature. The current distribution domain is generated by stimulating the change in the current distribution of the circuit traces at the thermoelectric correlation point. A circuit state spectrum library is constructed based on the current distribution domain.

3. The method according to claim 1, characterized in that, The step of injecting a swept-frequency signal at the defect excitation point to generate defect response features includes: Obtain the electromagnetic resonance characteristics of the defect excitation point; Based on the aforementioned electromagnetic resonance characteristics, defect resonance is induced to generate resonance excitation points. The frequency sweep signal is amplified by the resonant excitation point to generate an echo excitation wave. Defect response characteristics are generated based on the echo excitation wave.

4. The method according to claim 1, characterized in that, The process of generating a defect impedance distribution domain based on the thermoelectric coupling coordinates includes: Based on the thermoelectric coupling coordinates, circuit trace impedance measurement is performed to identify high-resistance regions. The high-resistance region is used to evaluate the contact state of the solder joints and form a solder joint quality distribution. The resistance value of the solder joint quality distribution is calibrated to generate a calibration resistance matrix. The calibration resistor matrix is ​​spatially mapped to generate a defect impedance distribution domain.

5. The method according to claim 1, characterized in that, The step of generating a defect location matrix by harmonic crosstalk mining based on the phase detection network and the defect impedance distribution domain includes: Phase difference analysis is performed on the phase detection network to identify phase transition points; Based on the phase jump point, a resistance abrupt change region is determined in the defect impedance distribution domain; The current path tracing of the resistance abrupt region forms a crosstalk conduction path; A defect location matrix is ​​constructed by performing coordinate calibration on the crosstalk propagation path.

6. The method according to claim 1, characterized in that, The process of reconstructing the electric field distribution around the defect based on the gradient tracking field to generate the defect electric field topology includes: Based on the gradient tracking field, field strength polarity is identified to generate positive and negative field strength regions. The negative field strength region is balanced using the positive field strength region to form a uniform field strength region. The uniform field strength region is topologically reconstructed to generate topological structural units; Geometric optimization is performed on the aforementioned topological structural units to construct a defect electric field topology.

7. The method according to claim 1, characterized in that, The step of extracting defect type and severity level by hierarchical annotation based on the precise coordinates of the defect includes: Construct the core damage region based on the precise coordinates of the defect; Chain reaction analysis was performed on the core damage area to obtain the sweep efficiency. The tiered damage recursion chain is obtained based on the aforementioned sweep coefficient; Defect types and severity levels are generated based on the aforementioned tiered damage recursion chain.

8. The method according to claim 2, characterized in that, The method of generating a current distribution domain by utilizing the thermoelectric correlation point to excite the change in circuit trace current distribution includes: A redistribution scheme is generated based on the circuit trace current state analysis of the thermoelectric correlation point. Current density monitoring is performed on the redistribution scheme to generate density change curves; The dominant current path is generated by identifying the current path through the density change curve. The dominant current path is spatially mapped to generate a current distribution domain.

9. The method according to claim 5, characterized in that, Determining the resistance abrupt change region in the defect impedance distribution domain based on the phase jump point includes: The phase transition point is used as the search center to detect the resistance value of the surrounding area and obtain resistance distribution data. Based on the resistance distribution data, the rate of change of impedance is analyzed to generate a rate of change distribution. The search range is continuously narrowed using the rate of change distribution to improve positioning accuracy; When the positioning accuracy meets the requirements, the search area is determined to be the resistance change region.

10. A BMS board defect intelligent detection system, characterized in that, include: The signal acquisition module is used to acquire multi-domain abnormal signals and circuit routing information during the operation of the BMS board. The multi-domain abnormal signals include impedance spectrum data and thermal infrared distribution signals. The multi-domain abnormal signals are fused in the frequency domain to construct a circuit state spectrum library. The excitation detection module is used to identify abnormal impedance modes through the circuit state spectrum library, perform reverse excitation analysis on the abnormal impedance modes to generate defect excitation points, inject a frequency sweep signal into the defect excitation points to generate defect response features, and perform resonant amplification on the defect response features to generate an enhanced detection domain. The thermoelectric analysis module is used to perform gradient analysis on the thermal infrared distribution signal to generate a thermal flow topology field, extract the location of abnormal heat source from the thermal flow topology field, geometrically correlate the location of abnormal heat source with the circuit routing information to generate thermoelectric coupling coordinates, and perform impedance mapping based on the thermoelectric coupling coordinates to generate a defect impedance distribution domain. The spectrum localization module is used to perform spectrum decomposition on the enhanced detection domain to identify the defect resonant frequency point, perform multi-frequency superposition excitation at the defect resonant frequency point to generate a superimposed excitation field, perform phase weaving on the superimposed excitation field to construct a phase detection network, and perform harmonic crosstalk mining based on the phase detection network and the defect impedance distribution domain to generate a defect localization matrix. The precise positioning module is used to perform impedance gradient analysis on the defect positioning matrix to generate a gradient tracking field, reconstruct the field strength distribution around the defect based on the gradient tracking field to generate the defect electric field topology, and use the defect electric field topology to perform field strength peak positioning to generate the precise coordinates of the defect. The results output module is used to extract the defect type and severity level by performing layered annotation based on the precise coordinates of the defect, and to complete the intelligent detection of defects on the BMS board.

Images