Automatic testing method and system for vehicle-mounted PCB mainboard

Abstract

The invention relates to the technical field of mainboard testing, and discloses an automatic testing method and system for a vehicle-mounted PCB mainboard, and the method comprises the steps: collecting a multi-dimensional environment stress original data set of the PCB mainboard; performing feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set; performing fault feature analysis on a target node of the PCB mainboard based on the five-dimensional state vector set to obtain a PCB fault feature vector; performing adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix; pCB intermittent fault state estimation is carried out based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix, a fault state estimation value and fault position information are obtained, the pertinence and accuracy of fault detection are improved, the false alarm rate is effectively reduced, and component-level fault accurate positioning is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mainboard testing, and particularly relates to a vehicle-mounted PCB mainboard automatic testing method and system. BACKGROUND

[0002] As the core carrier of the automotive electronic system, the vehicle-mounted PCB mainboard carries key functional modules such as engine control, braking system, infotainment, etc. However, the vehicle-mounted environment has characteristics such as drastic temperature change, frequent vibration and impact, obvious humidity fluctuation, unstable power supply, and complex electromagnetic interference, and the coupling effect of these multi-dimensional environmental stresses can easily lead to intermittent faults such as solder joint cracks, micro-fractures, and false soldering of the PCB mainboard, which seriously threatens the safety and reliability of vehicle operation.

[0003] The traditional PCB fault detection method mainly relies on single parameter monitoring or offline detection technology, and cannot effectively capture the intermittent fault characteristics under multi-dimensional environmental stress cycle excitation. The existing detection technology generally has problems such as low detection accuracy, high false alarm rate, and inability to accurately locate the fault position. SUMMARY

[0004] The present application provides a vehicle-mounted PCB mainboard automatic testing method and system, which improves the pertinence and accuracy of fault detection, effectively reduces the false alarm rate, and realizes accurate fault positioning at the component level.

[0005] In a first aspect, the present application provides a vehicle-mounted PCB mainboard automatic testing method, which comprises: Collecting a multi-dimensional environmental stress original data set of the PCB mainboard; Performing feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set; Performing fault feature analysis on a target node of the PCB mainboard based on the five-dimensional state vector set to obtain a PCB fault feature vector; Performing adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix; Performing PCB intermittent fault state estimation based on the PCB fault feature vector, the five-dimensional state vector set, and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault position information.

[0006] In combination with the first aspect, in a first implementation manner of the first aspect of the present application, the collecting of the multi-dimensional environmental stress original data set of the PCB mainboard comprises: The temperature cycle parameter collection is performed on the surface of the PCB mainboard to obtain temperature original data; the vibration acceleration parameter collection is performed on the PCB mainboard to obtain vibration original data; the environmental humidity gradient parameter collection is performed on the PCB mainboard to obtain humidity original data; the power supply fluctuation parameter collection is performed on the vehicle-mounted power supply system to which the PCB mainboard belongs to obtain power supply original data; and the environmental electromagnetic interference parameter collection is performed on the PCB mainboard to obtain electromagnetic interference original data. The temperature original data, the vibration original data, the humidity original data, the power supply original data and the electromagnetic interference original data are subjected to data aggregation and time synchronization processing to obtain a multi-dimensional environmental stress original data set.

[0007] In combination with the first aspect, in a second implementation manner of the first aspect of the application, the feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set comprises: The temperature original data in the multi-dimensional environmental stress original data set is subjected to absolute temperature value extraction, temperature change rate calculation and temperature change rate feature calculation to obtain a temperature feature vector; The vibration original data in the multi-dimensional environmental stress original data set is subjected to fast Fourier transform processing and power spectral density calculation to obtain a vibration feature vector; The humidity original data, the power supply original data and the electromagnetic interference original data in the multi-dimensional environmental stress original data set are respectively subjected to statistical feature extraction and spectrum analysis processing to obtain a humidity feature vector, a power supply feature vector and an electromagnetic interference feature vector; The temperature feature vector, the vibration feature vector, the humidity feature vector, the power supply feature vector and the electromagnetic interference feature vector are combined to obtain the five-dimensional state vector set.

[0008] In combination with the first aspect, in a third implementation manner of the first aspect of the application, the target node of the PCB mainboard is subjected to fault feature analysis based on the five-dimensional state vector set to obtain a PCB fault feature vector, which comprises: A sinusoidal signal of a preset frequency range is injected into the target node of the PCB mainboard by a carrier signal injection method, and the injection intensity and frequency parameters of the sinusoidal signal are adjusted according to the temperature feature vector, the vibration feature vector and the humidity feature vector in the five-dimensional state vector set to obtain an equivalent impedance parameter; An eye diagram analysis is performed on the digital signal line of the target node, and the eye height parameter, the eye width parameter and the jitter parameter are monitored and calculated in real time according to the electromagnetic interference feature vector in the five-dimensional state vector set to obtain a signal integrity parameter; perform noise measurement and transient response characteristic analysis on a power plane of the target node, and adjust a measurement threshold and a response time window according to a power feature vector in the five-dimensional state vector set to obtain a power integrity parameter; perform vector mapping on the equivalent impedance parameter, the signal integrity parameter and the power integrity parameter to obtain a PCB fault feature vector.

[0009] In a fourth implementation form of the first aspect, the adaptive fading factor calculation based on the five-dimensional state vector set comprises: perform two-norm calculation on the five-dimensional state vector set to obtain an environmental stress intensity value; determine a material characteristic parameter combination according to a material type of the PCB mainboard, wherein a first material parameter of the FR4 material is 10.5 and a second material parameter is 0.35, a first material parameter of the high Tg material is 15.8 and a second material parameter is 0.28, and a first material parameter of the high-frequency material is 12.3 and a second material parameter is 0.31; calculate an adaptive fading factor of the environmental stress intensity value and the material characteristic parameter combination; perform inverse scaling and forward scaling adjustment on a reference system noise covariance matrix and a reference measurement noise covariance matrix based on the adaptive fading factor to obtain a dynamically adjusted covariance matrix.

[0010] In a fifth implementation form of the first aspect, the calculation of the adaptive fading factor of the environmental stress intensity value and the material characteristic parameter combination comprises: perform square operation on the environmental stress intensity value to obtain a square result, and perform quotient calculation on the square result divided by a first material parameter in the material characteristic parameter combination to obtain a base parameter of exponential operation; perform negative sign processing on the base parameter of exponential operation, and perform exponential operation through an exponential function to obtain an exponential operation result; perform product operation on the exponential operation result and a second material parameter in the material characteristic parameter combination to obtain a fading adjustment amount, and perform subtraction operation on the fading adjustment amount and the value 1 to obtain the adaptive fading factor.

[0011] In a sixth implementation form of the first aspect, the PCB intermittent fault state estimation based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix comprises: The state dimension of the PCB fault feature vector is determined around a state distribution to obtain a sigma point set through deterministic sigma point selection. State prediction and observation prediction are performed on the sigma point set and the five-dimensional state vector set through a nonlinear state mapping function and a nonlinear observation mapping function respectively to obtain state prediction values and observation prediction values. A prediction covariance matrix and a cross covariance matrix are calculated based on the state prediction values, the observation prediction values and the dynamically adjusted covariance matrix, and a matrix inversion operation is performed based on the prediction covariance matrix and the cross covariance matrix to obtain a filter gain parameter. Residuals are calculated from the PCB fault feature vector and the observation prediction values, and state updates are performed on the state prediction values in combination with the filter gain parameter to obtain fault state estimation values and fault location information.

[0012] In a seventh implementation manner of the first aspect, the state dimension of the PCB fault feature vector is determined around a state distribution to obtain a sigma point set through deterministic sigma point selection, and the sigma point set is obtained through the following steps: A state dimension is determined according to the PCB fault feature vector, and a scaling parameter is calculated by subtracting the state dimension from the value 3 to obtain a scaling parameter of unscented transformation; The dynamically adjusted covariance matrix is subjected to Cholesky decomposition to obtain a decomposition result, and the decomposition result is subjected to product operation with the state dimension and the scaling parameter to obtain a covariance square root matrix; A current state estimation value is taken as a center sigma point, and the current state estimation value is subjected to addition operation and subtraction operation with each column vector of the covariance square root matrix to obtain a positive sigma point set and a negative sigma point set; The center sigma point, the positive sigma point set and the negative sigma point set are merged to obtain a sigma point set.

[0013] In an eighth implementation manner of the first aspect, the residuals are calculated from the PCB fault feature vector and the observation prediction values, and the state updates are performed on the state prediction values in combination with the filter gain parameter to obtain the fault state estimation values and the fault location information, and the steps include: The PCB fault feature vector and the observation prediction values are subjected to subtraction operation to obtain an observation residual vector; Absolute value calculation is performed on each vector element of the observation residual vector to obtain a plurality of vector element absolute values, and the plurality of vector element absolute values are compared with a preset threshold value respectively to obtain a residual absolute value vector and a threshold comparison result; Based on the threshold comparison result, the residual absolute value vector is calculated by a Huber function to obtain a robust weight adjustment factor; Element-wise product operation is performed on the observation residual vector and the robust weight adjustment factor to obtain a product operation result, and the state prediction value is subjected to an addition operation in combination with the filter gain parameter and the product operation result to obtain a fault state estimation value; Based on the abnormal position of the electrical parameter in the fault state estimation value, component coordinate mapping calculation is performed to obtain fault location information.

[0014] In a second aspect, the present application provides a vehicle-mounted PCB mainboard automatic test system, which comprises: A collection module is configured to collect a multi-dimensional environmental stress original data set of a PCB mainboard. A feature extraction module is configured to perform feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set. A fault feature analysis module is configured to perform fault feature analysis on a target node of the PCB mainboard based on the five-dimensional state vector set to obtain a PCB fault feature vector. A calculation module is configured to perform adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix. A state estimation module is configured to perform intermittent fault state estimation of the PCB based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information.

[0015] The technical scheme provided by the application comprises the following steps: collecting temperature cycles, vibration acceleration, humidity gradients, power fluctuations and electromagnetic interference in five dimensions of environmental parameters, constructing a complete vehicle-mounted environmental stress model, and comprehensively reflecting the real working environment of the PCB mainboard to provide comprehensive environmental excitation information for fault detection. An adaptive fading factor calculation method based on the characteristics of PCB materials is adopted, which can dynamically adjust the detection parameters according to different material types (FR4, high Tg, high-frequency material), significantly improve the pertinence and accuracy of fault detection, and effectively reduce the false alarm rate. The improved unscented Kalman filter is used for nonlinear state estimation, and the sigma point selection and weight calculation can accurately capture the dynamic change process of the intermittent fault of the PCB, realize accurate tracking and prediction of the fault state. Through the carrier signal injection method, impedance analysis, eye diagram analysis technology for signal integrity detection, noise measurement for power integrity analysis, a multi-dimensional fault feature vector is constructed, which can comprehensively describe the change characteristics of the electrical performance of the PCB. The Huber function is used for robust weight adjustment, which can effectively suppress the influence of abnormal measurement values on fault state estimation, and enhance the stability and reliability of the detection system in complex vehicle-mounted environments. Through the component coordinate mapping calculation of the abnormal position of the electrical parameters, the abstract electrical parameter abnormality can be converted into specific physical position information, and the fault is accurately positioned at the component level.

[0016] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and achieved by the structure particularly pointed out in the description, claims and drawings.

[0017] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 An embodiment schematic diagram of the vehicle-mounted PCB mainboard automatic test method in the embodiment of the present application is shown in the figure. Figure 2 An embodiment schematic diagram of the vehicle-mounted PCB mainboard automatic test system in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0019] In order to make the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the present application will be described in detail below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] The terms "comprising" and "having" and any variations thereof used in referring to elements of a process, method, system, product or apparatus, are intended to cover the process, method, system, product or apparatus "consisting of" the listed elements, but not excluding additional elements. For example, a process, method, system, product or apparatus that includes a list of steps or elements is not limited to only those steps or elements but can include other steps or elements not expressly listed or inherent to such process, method, system, product or apparatus.

[0021] To facilitate the understanding of the present embodiment, first of all, a kind of vehicle-mounted PCB mainboard automatic test method disclosed in the present embodiment is introduced in detail. As shown in Figure Figure 1 The method comprises the following steps: 101, collect the multi-dimensional environmental stress original data set of PCB mainboard; It can be understood that the execution subject of the present application can be a vehicle-mounted PCB mainboard automatic test system, and can also be a terminal or a server, and the specific place is not limited. The present embodiment takes the server as the execution subject for example.

[0022] Specifically, a multi-dimensional physical environment monitoring system suitable for complex working conditions of vehicles is constructed, which covers key environmental excitation dimensions affecting the performance of PCBs, including temperature, vibration, humidity, power fluctuation and electromagnetic interference. In terms of temperature parameter collection, multiple high-precision PT100 platinum resistance sensors are uniformly arranged on the surface of the PCB mainboard, with 5 to 8 temperature measurement points arranged to ensure coverage of key thermal sensitive component areas and capture of spatial distribution characteristics of thermal stress. The sampling frequency is set to 10 Hz, the temperature measurement range is from -50℃ to 150℃, the measurement accuracy reaches ±0.1℃, and is used to record the temperature change trajectory during thermal cycling; in the vibration acceleration collection process, a high-sensitivity piezoelectric three-axis acceleration sensor is selected and fixedly installed at the edge and center of the PCB mainboard to obtain comprehensive structural response data. The sensor sensitivity is set to 100 mV / g, the frequency response covers 5 Hz to 3000 Hz, and the sampling frequency reaches 5000 Hz, which can effectively capture high-frequency vibration impacts caused by engine, chassis system or road condition changes during vehicle operation; for humidity gradient monitoring, a capacitive humidity sensor is used, with 3 to 5 measurement points arranged in the peripheral and internal cavity areas of the mainboard to form a humidity gradient field. The sampling frequency is 1 Hz, the measurement range covers 0 to 100% relative humidity, and the measurement accuracy reaches ±2% RH, which can reflect the influence of spatial and temporal changes of environmental humidity on the insulation state and solder joints of the PCB; power fluctuation parameter collection is performed by a high-precision voltage sensor to continuously monitor the voltage input of the vehicle power supply system at a millisecond level. The sampling frequency is 1000 Hz, the measurement accuracy reaches ±0.01V, and is used to record voltage spikes, instantaneous drops and low-frequency fluctuations; the collection of electromagnetic interference relies on a near-field probe array, which covers a frequency range of 150 kHz to 3 GHz, and can realize real-time sensing of the electromagnetic field strength changes around the PCB mainboard, especially in typical scenarios such as ignition start, CAN bus communication and wireless signal interference. After independent collection of various environmental data, each type of signal is transmitted to the central processing unit through a distributed data collection architecture and a hardware anti-interference processing module. The module includes an analog signal conditioning circuit, an analog-to-digital conversion module and a filtering structure. The filtering part uses multi-stage band-pass filtering and adaptive wavelet denoising technology to process high-frequency interference and baseline drift in the signal; the data processing system uses a CAN-FD high-speed communication bus for data aggregation, with a transmission rate of 5 Mbps to ensure efficient transmission of multi-channel parallel data; after data collection and aggregation, a unified time synchronization processing is performed, a time synchronization mechanism based on clock alignment and hardware timestamp calibration is adopted, so that the sampling results of all collection channels are strictly time-domain corresponding at a time resolution of 1 ms, and finally five types of data, including temperature, vibration, humidity, power and electromagnetic interference, are fused into a unified time series matrix, thereby forming a multi-dimensional environmental stress original data set.

[0023] 102. Feature extraction and normalization processing are performed on the multi-dimensional environmental stress raw data set to obtain a five-dimensional state vector set; Specifically, the five types of original data of temperature, vibration, humidity, power supply and electromagnetic interference are respectively structured to be embedded in a high-dimensional state expression space with unified dimension, unified scale and coupling relationship. For temperature original data, the absolute temperature value sequence is extracted to describe the static thermal field distribution, and the first derivative of temperature, i.e. the rate of change of temperature with time, is calculated. The fluctuation trend of the rate of change, i.e. the second derivative, is further extracted to construct a temperature feature vector, which contains the steady-state temperature level, thermal stress change speed and acceleration information, thereby reflecting the dynamic behavior pattern in the thermal shock process. For vibration original data, the power spectrum density distribution is calculated by converting the vibration original data from time domain to frequency domain through fast Fourier transform, and the energy distribution is partitioned and counted according to the preset frequency band, for example, the frequency spectrum is divided into three regions of 10-100Hz, 100-500Hz and 500-2000Hz, and the spectral energy of each region and the root mean square acceleration value of the whole are extracted to form a vibration feature vector describing the vibration intensity, frequency concentration and resonance characteristics. When processing humidity, power supply fluctuation and electromagnetic interference original data, the fusion of statistical features and frequency domain structure is adopted, and the time domain statistical features such as mean, standard deviation, maximum / minimum value and slope change trend are extracted for each type of signal to characterize its stability and fluctuation degree. At the same time, the frequency structure information is extracted through spectrum analysis technology. The relative change vector can be formed by combining the change rate and environmental gradient for the humidity part, the voltage peak-peak value and frequency distribution density are considered for the power supply part, and the electromagnetic field strength density in different frequency bands is extracted for the electromagnetic interference part, which form the humidity feature vector, the power supply feature vector and the electromagnetic interference feature vector respectively. In order to eliminate the inconsistency of dimension and scale between different physical quantities, a unified normalization processing mechanism is introduced, and the Min-Max method is used to map all features to the [0, 1] interval. The temperature feature vector, vibration feature vector, humidity feature vector, power supply feature vector and electromagnetic interference feature vector are concatenated and combined to form a five-dimensional state vector set.

[0024] 103. Based on the five-dimensional state vector set, the target node of the PCB mainboard is analyzed to obtain a PCB fault feature vector; Specifically, the carrier signal injection method is used to actively stimulate the target node in the PCB mainboard, and a micro-amplitude sinusoidal signal of a specific frequency range is injected on the node, which covers 20MHz to 500MHz to ensure sufficient detection of high-frequency response changes caused by multiple types of structural abnormalities such as solder cracking, virtual welding or trace degradation. The amplitude and frequency of the injected signal are dynamically adjusted according to the current five-dimensional state vector set and the environmental state associated with the node position, especially in combination with the local thermal shock reflected in the temperature feature vector, the structural stress excitation intensity indicated in the vibration feature vector, and the insulation condition change embodied in the humidity feature vector. The injection strength and frequency of the sinusoidal signal are adjusted in real time to enhance the sensitivity of the injected signal to abnormal responses and improve the accuracy and stability of the equivalent impedance extraction. In this process, the equivalent impedance parameter set of the target node is extracted through amplitude-frequency response modeling and phase delay analysis, which is used to reflect the physical connection state and the integrity of the local conduction path. At the same time, in order to evaluate the quality state of the target node in the digital signal transmission process, eye diagram analysis is performed on its main signal path, and the eye height, eye width and jitter amount indicators under continuous code type are obtained through oscilloscope, and the electromagnetic interference feature vector in the five-dimensional state vector set is combined to dynamically compensate and constrain the key parameters in the eye diagram analysis process. Especially when the electromagnetic interference intensity is enhanced, the analysis threshold is reduced to improve the sensitivity, and the synchronization window is adjusted to avoid false identification caused by interference fluctuations, so as to accurately capture the changes in signal integrity and extract signal integrity parameters containing dynamic jitter influence and synchronization mismatch information. In order to identify the intermittent response instability caused by unstable power supply or power structure resonance, the noise and transient response characteristics of the power plane region where the target node is located are detected, and the voltage ripple amplitude, recovery time and load disturbance response are used to characterize the power integrity. The selection of detection parameters also depends on the power feature vector in the five-dimensional state vector set as a reference benchmark. When the voltage fluctuation is severe or the power supply frequency is close to the critical working point, the system automatically adjusts the upper and lower limits of the noise measurement threshold and the observation time window of the transient response to avoid boundary distortion and response signal loss, ensuring that the obtained power stability indicators have sufficient resolution and dynamic range. The equivalent impedance parameters, signal integrity parameters and power integrity parameters obtained in the above three dimensions are uniformly mapped into a vector to construct a PCB fault feature vector with clear multi-dimensional association of structural response, power generation behavior and transmission characteristics.

[0025] 104、based on the five-dimensional state vector set, an adaptive fading factor is calculated to obtain a dynamically adjusted covariance matrix; Specifically, the five-dimensional state vector set obtained in the current time window is subjected to two norm calculation, and the solution process of the norm is equivalent to embedding the temperature feature vector, the vibration feature vector, the humidity feature vector, the power supply feature vector and the electromagnetic interference feature vector into a Euclidean space together, and obtaining a scalar value as the strength index of the current environmental stress through vector module length calculation, which can reflect the overall effect of the current environment on the PCB mainboard. On this basis, according to the material type used by the PCB mainboard to be tested, the corresponding first material parameter and second material parameter are extracted from the preset material characteristic database as the model adjustment basis. When the tested mainboard is FR4 standard material, the first material parameter is 10.5 and the second material parameter is 0.35; if it is a high Tg material, the corresponding parameters are 15.8 and 0.28; if it is a high frequency material, 12.3 and 0.31 are used as the parameter combination. This set of material characteristic parameters reflects the sensitivity and degradation rate of different materials to thermal, mechanical and electrical coupling stress response, and thus serves as an adjustment factor for the environmental stress response threshold in the filtering model. Combined with the environmental stress intensity value calculated above and the selected material parameter combination, an adaptive decay factor is calculated by using an exponential decay structure to construct a nonlinear response relationship. The essence of the factor is to simulate the rapid decay response of the system to the noise model under high-intensity environmental stress, that is, when the environmental disturbance is severe, the system increases the state update speed to enhance the tracking ability of the sudden state, and in a stable environment, it maintains strong smoothness to suppress excessive response. Based on the adaptive decay factor, the reference system noise covariance matrix and the reference measurement noise covariance matrix in the filter are dynamically adjusted, wherein the system noise covariance matrix is enhanced by inverse scaling to increase the tolerance to process noise during state evolution, thereby improving the response speed to rapid state changes; at the same time, the measurement noise covariance matrix is weighted and contracted by forward scaling to suppress the disturbance of sudden observation anomalies to the state estimation result, thereby maintaining the stability and convergence of the observation channel. The covariance matrix adjustment strategy of one increase and one decrease not only constitutes a joint dynamic constraint mechanism for state and observation noise in theory, but also exhibits strong adaptive robustness when the environmental stress mutates or high-frequency interference exists, so that the filter can not only maintain the real-time performance of the estimation result, but also avoid false alarms or convergence instability caused by excessive response. Through dynamic updating of the system and observation covariance matrices, the system can be adjusted according to the environmental stress change and material sensitivity at each time, and a dynamic covariance structure with working condition perception ability is constructed.

[0026] The environmental stress intensity value is squared to amplify the response amplitude of high intensity stress and enhance the resolution capability of subsequent operation. The square result is taken as the dividend, and the first material parameter in the material characteristic parameter combination corresponding to the current test PCB mainboard is calculated by the quotient value. The first material parameter is a value positively correlated with the sensitivity of the material to external disturbance response, which is used as a reference factor to calculate the amplitude of the adjustment index, thereby obtaining the base parameter in the exponential operation. The base parameter needs to be processed with a negative sign to reflect the nonlinear decay trend that the stronger the stress, the more intense the system response adjustment. By executing the exponential function operation with the negative value as input, an attenuation factor original response value between zero and one is obtained. This value has good convergence and adjustment continuity due to the use of exponential function mapping, and is suitable for responding to the sudden changes of stress excitation in the vehicle environment. The result of the exponential operation is multiplied by the second material parameter in the material characteristic parameter combination, and the second parameter represents the expansion factor of the material state change in the high stress environment. The result of multiplying the exponential response value represents the required covariance weight adjustment amplitude of the material under the current stress intensity, that is, the decay adjustment amount. The value 1 is subtracted from the decay adjustment amount to obtain the final adaptive decay factor. The closer the value of the decay factor to 1, the smaller the environmental stress, and the system tends to maintain the original filter stability. When the factor value decreases, it indicates that the environmental stress intensity increases, and the system needs to appropriately accelerate the response ability to state changes.

[0027] 105. Perform PCB intermittent fault state estimation based on the PCB fault feature vector, the five-dimensional state vector set, and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information.

[0028] Specifically, taking the state dimension of the PCB fault feature vector as the benchmark, a deterministic sigma point selection operation is performed around the state distribution center at the current time. This selection process simulates the overall state distribution characteristics by arranging multiple discrete representative points around the state mean in a specific structure. In order to ensure that these sigma points accurately reflect the variance and covariance structure between state variables, a square root decomposition operation is performed based on the dynamically adjusted covariance matrix, and an offset is constructed using the square root matrix to achieve symmetric distribution and structural integrity of each sigma point in high-dimensional space, forming a sigma point set as the input basis for subsequent nonlinear propagation. The sigma point set is processed through nonlinear state mapping functions and nonlinear observation mapping functions, and combined with the five-dimensional state vector set corresponding to the current time, the state propagation and observation response simulation of each sigma point is completed, i.e., the evolution behavior of the sigma point in the state space at the next time is estimated through the state mapping function, and the corresponding sensor response or measurement output in the observation space is simulated through the observation mapping function, so as to obtain the state prediction value and observation prediction value of each sigma point, respectively, and the state prediction mean and observation prediction mean are summarized in a weighted form. The offset difference between the state prediction result of all sigma points and its mean is multiplied, and the prediction covariance matrix is constructed by stacking, which is used to describe the overall uncertainty of the system in the state propagation process. At the same time, the joint offset relationship between the state variables and the observation variables is calculated, and the cross-covariance matrix is constructed through the same weighted operation, which reflects the sensitivity and conduction effect of state disturbance on observation output. Based on the matrix structure relationship between the prediction covariance matrix and the cross-covariance matrix, the inverse operation is performed to obtain the filter gain parameter, which will be used as a residual adjustment factor in the update link. The size of the parameter is directly related to the sensitivity of the system to observation errors, and controls the state correction step and response amplitude. The residual error between the PCB fault feature vector and the observation prediction value generated by the sigma point is calculated to obtain the deviation between the measurement value at the current time and the predicted observation, and the residual error is fused with the aforementioned filter gain parameter to correct the original state prediction value, thereby obtaining the updated fault state estimation value. This estimation value not only contains quantitative evaluation of fault strength, persistence and evolution trend, but also completes the inference of fault location information by combining with the circuit topology structure and node correspondence, thereby realizing the two-dimensional positioning of intermittent faults in the time axis and physical structure.

[0029] The dimension of the current state is determined according to the dimension information of the PCB fault feature vector, and the dimension corresponds to the number of high-dimensional electrical fault indicators such as the extracted equivalent impedance parameters, signal integrity parameters and power integrity parameters. Based on this dimension, the scaling parameter for controlling the breadth of sigma point distribution in the unscented transformation is obtained by subtracting 3 from the calculation of the current state dimension, which directly affects the distribution radius of the subsequent sigma point in the state space and its representativeness, and is an important adjustment factor between the number of dimensions and the state coverage capability. To ensure the spatial structure of the current state uncertainty in the sigma point selection process, the dynamic adjustment covariance matrix for modeling the noise structure is subjected to Cholesky decomposition. Cholesky decomposition is a way to decompose a positive definite symmetric matrix into a lower triangular matrix multiplied by its transpose, which has numerical stability and can guarantee that the output matrix has strict orthogonal properties, which is suitable for positive and negative direction expansion of state disturbance. After obtaining the decomposition result, multiply it by the state dimension and the scaling parameter respectively to construct the covariance square root matrix. Each column of the matrix is a principal axis offset vector in a certain disturbance direction in the current state space, representing the standard offset step of the state distribution in that direction. The state estimate at the current time is taken as the center point of the sigma point set, and the forward and negative disturbance expansion operations are performed. The center point is added to each column vector of the covariance square root matrix to obtain the positive sigma point set, and the center point is subtracted from the same column vector to obtain the negative sigma point set. This symmetry processing ensures that the overall sigma point set forms a symmetric expansion structure with equal probability and equal distance around the state distribution, so as to cover the main uncertainty area in the state space and have high fidelity reflection capability for the nonlinear function mapping characteristics. The center sigma point, the positive sigma point set and the negative sigma point set are combined into a complete sigma point set to form a high-dimensional representative point array with a total number of 2n+1 state points, where n is the state dimension.

[0030] The PCB fault feature vector is subtracted from the observed prediction value to form a set of observation residual vectors representing the difference between the actual observation and the theoretical prediction. This vector reflects the dynamic deviation of the current system in a specific electrical feature dimension and provides directional correction basis for subsequent state correction. The absolute value operation is performed on all elements in the observation residual vector to eliminate the directional influence and only retain the amplitude characteristics of the deviation, forming a set of absolute values of the vector elements. The set of absolute values is compared with the preset multi-dimensional threshold set element by element to construct a set of threshold comparison results reflecting whether each deviation is out of bounds. Based on the threshold comparison results, the absolute values of the observation residual vector are calculated by the Huber function. The Huber function is a robust loss function that takes into account both least squares estimation and least absolute value estimation. It remains linear in response when the error is small, and becomes a slowly varying response when the error is too large, thereby preserving the contribution of small residuals to the correction while weakening the destructive effect of large residuals on state update. The output obtained after processing by the function is the robust weight adjustment factor, reflecting the correction confidence of the system in each fault feature dimension under the current environmental stress and observation disturbance. The original observation residual vector is multiplied by the corresponding robust weight adjustment factor element by element to obtain a set of weighted and adjusted correction input vectors, which are then multiplied by the filter gain parameters obtained in the previous stage, and finally added to the state prediction value to realize the correction and update of the fault state, obtaining the fault state estimation vector containing the estimated values of each electrical feature at the current time. The components corresponding to the maximum abnormal deviation or continuous abnormality in the fault state estimation value are extracted, and combined with the topological structure and component layout information of the PCB mainboard, the component coordinate mapping calculation is performed to map the abnormal indicators in the state vector dimension to the specific physical component number and coordinate position, forming the final fault location information.

[0031] In the embodiment of the application, by simultaneously collecting environmental parameters in five dimensions of temperature cycle, vibration acceleration, humidity gradient, power fluctuation and electromagnetic interference, a complete vehicle-mounted environmental stress model is constructed, which can comprehensively reflect the real working environment of the PCB mainboard and provide comprehensive environmental excitation information for fault detection. An adaptive fading factor calculation method based on the characteristics of PCB materials is adopted, which can dynamically adjust the detection parameters according to different material types (FR4, high Tg, high-frequency material), significantly improve the pertinence and accuracy of fault detection, and effectively reduce the false alarm rate. An improved unscented Kalman filter is used for nonlinear state estimation, and through sigma point selection and weight calculation, the dynamic change process of intermittent faults of the PCB can be accurately captured, and accurate tracking and prediction of the fault state can be realized. Through carrier signal injection method for impedance analysis, eye diagram analysis technology for signal integrity detection, and noise measurement for power integrity analysis, a multi-dimensional fault feature vector is constructed, which can comprehensively describe the change characteristics of the electrical performance of the PCB. Robust weight adjustment is performed using Huber function, which can effectively suppress the influence of abnormal measurement values on fault state estimation, and enhance the stability and reliability of the detection system in complex vehicle-mounted environments. Through component coordinate mapping calculation of the abnormal position of the electrical parameters, the abstract electrical parameter abnormality can be converted into specific physical location information, and accurate fault positioning at the component level can be realized.

[0032] In a specific embodiment, the process of performing step 101 can specifically include the following steps: The surface of the PCB mainboard is subjected to temperature cycle parameter collection to obtain temperature raw data; the PCB mainboard is subjected to vibration acceleration parameter collection to obtain vibration raw data; the PCB mainboard is subjected to environmental humidity gradient parameter collection to obtain humidity raw data; the vehicle-mounted power supply system to which the PCB mainboard belongs is subjected to power fluctuation parameter collection to obtain power raw data; and the PCB mainboard is subjected to environmental electromagnetic interference parameter collection to obtain electromagnetic interference raw data. The temperature raw data, vibration raw data, humidity raw data, power raw data and electromagnetic interference raw data are subjected to data aggregation and time synchronization processing to obtain a multi-dimensional environmental stress raw data set.

[0033] Specifically, in terms of temperature cycle parameter collection, to capture the periodic thermal stress experienced by the PCB mainboard surface under different vehicle operating conditions such as start-stop, idle, high-speed running, etc., a temperature monitoring unit with multiple high-precision PT100 type platinum resistors is used. This sensor has excellent linear response characteristics and good long-term stability. The sensor points are evenly distributed on the high-heat-density area of the mainboard, around the core processing module, and adjacent to the edge signal interface, with 5 to 8 independent measurement points deployed, and temperature data is continuously acquired at a sampling frequency of 10 times per second. This temperature monitoring unit supports a working range of -50℃ to 150℃, with an accuracy of ±0.1℃, and can accurately track the thermal stress trend caused by environmental changes, component heating or heat diffusion, and export temperature change rate and temperature difference gradient parameters as needed for subsequent analysis. In the vibration acceleration parameter collection, considering the frequent and wide frequency spectrum of vibration sources in the vehicle environment, such as engine excitation, road unevenness, chassis resonance, etc., which will be transmitted to the PCB mainboard through structural coupling, a piezoelectric three-axis acceleration sensor is configured and fixed to the key structural nodes of the mainboard, including the central processor directly above, the connection part of the heat dissipation metal component, and the suspended PCB edge area. The sensor has a sensitivity of 100 mV / g, a response frequency range of 5 Hz to 3000 Hz, and a sampling frequency of 5000 Hz, ensuring that all effective vibration components from low-frequency fluctuations to high-frequency impacts can be captured, and early identification of energy aggregation in abnormal frequency bands can be performed. After analog conditioning and band-pass filtering, the output signal can suppress power frequency interference and high-frequency false peaks to the greatest extent. For environmental humidity gradient parameter collection, the interaction between indoor humidity fluctuation and PCB mainboard surface moisture absorption characteristics is considered, and the response speed and spatial resolution of the deployed capacitive humidity sensor need to be considered. The sensor is arranged in the upper, middle, and lower layer distribution structure inside the mainboard packaging shell, with 3 to 5 measurement points recording the spatio-temporal changes of humidity in the microenvironment, a sampling frequency of 1 Hz, a measurement range of 0 to 100% RH, and an accuracy of ±2% RH. The system monitors the humidity gradient, i.e. the relative humidity difference between the mainboard surface and the surrounding space, and analyzes the humidity change rate simultaneously, providing basic data support for analyzing typical humidity-induced problems such as insulation performance changes, solder point moisture erosion, and surface dew trends. For power fluctuation parameter collection, due to the dynamic events such as voltage spikes, load fluctuations, and sudden discontinuities faced by the vehicle power supply system, a voltage sensor module with high time resolution is deployed directly connected to the mainboard power input. The module has a sampling frequency of 1000 Hz and an accuracy of ±0.01V, supports real-time recording of voltage ripple, transient drop, voltage climb rate, etc. key characteristics, reflects the power stability through voltage standard deviation, spectral energy density, and peak-to-peak value indicators, and is suitable for monitoring vehicle voltage stabilization system performance degradation, EMC coupling interference, and power discontinuity phenomena caused by thermal shutdown.The environmental electromagnetic interference parameter collection realizes real-time monitoring of high-frequency spatial electric field intensity by using a near-field probe array. The working frequency range covers 150 kHz to 3 GHz, which can capture interference components generated by multiple sources such as communication systems, high-frequency oscillators, and wireless charging systems. The array is deployed 5-10 mm above the surface of the mainboard to form a scanning surface array. The collection system can output electromagnetic intensity spectrum in real time and support weighted filtering processing for different frequency bands. For example, for the 150 kHz to 30 MHz segment, common-mode interference is focused on; for the 30 MHz to 300 MHz segment, radiation interference is focused on; and for the 300 MHz to 3 GHz segment, high-frequency modulated signal leakage is monitored, thereby constructing a multi-scale expression model of electromagnetic interference in the frequency spectrum dimension. After independent collection of five types of environmental raw data, all signals are uploaded to the central processing unit through the distributed data bus. To achieve structured fusion, the data from each channel is preprocessed through hardware filtering, analog-to-digital conversion, abnormality rejection, and dynamic smoothing. Time synchronization is performed after all data sampling. The synchronization mechanism uses a timestamped trigger system combined with a 1 ms level clock alignment strategy to ensure that the data from different channels correspond accurately on the time axis, avoiding data mismatch caused by sampling delay or clock drift. The five types of data after aggregation, alignment, and cleaning form a complete multi-dimensional environmental stress raw data set, which retains the dynamic evolution trajectories of temperature, vibration, humidity, power, and electromagnetic field in space and time.

[0034] In a specific embodiment, the process of performing step 102 can specifically include the following steps: The temperature raw data in the multi-dimensional environmental stress raw data set is subjected to absolute temperature value extraction, temperature change rate calculation, and temperature change rate feature calculation to obtain a temperature feature vector; The vibration raw data in the multi-dimensional environmental stress raw data set is subjected to fast Fourier transform processing and power spectral density calculation to obtain a vibration feature vector; The humidity raw data, power raw data, and electromagnetic interference raw data in the multi-dimensional environmental stress raw data set are respectively subjected to statistical feature extraction and spectral analysis processing to obtain a humidity feature vector, a power feature vector, and an electromagnetic interference feature vector; The temperature feature vector, the vibration feature vector, the humidity feature vector, the power feature vector, and the electromagnetic interference feature vector are combined to obtain a five-dimensional state vector set.

[0035] Specifically, in the process of temperature raw data processing, based on the temperature time series data provided by multiple point sensors on the surface of the PCB mainboard, an absolute temperature extraction operation is performed to directly obtain the average temperature value of multiple measuring points at the current time to reflect the steady-state level of the overall thermal environment of the mainboard. At the same time, in order to accurately capture the dynamic change trend caused by thermal shock, power fluctuation or abnormal structure heat dissipation, first-order difference calculation is performed on the temperature time series curve to obtain the speed information of temperature change with time, i.e. temperature change rate, and then the change amplitude of the change rate is calculated by continuous difference or sliding window method to form a high-order characteristic quantity of temperature change rate, which is used to reflect the nonlinear aggravation or frequent fluctuation behavior of thermal disturbance. By combining the absolute temperature value, temperature change rate and change rate, a three-dimensional temperature feature vector is constructed. For the processing of vibration raw data, the collected three-axis acceleration signals are projected into the frequency domain space through fast Fourier transform to obtain the distribution of signal energy in different frequency bands. In order to enhance the usability of frequency domain features, power spectral density calculation is performed on the transformation result, which reflects the energy intensity contained in the vibration signal under unit frequency bandwidth and reveals the dominant frequency band and resonance peak position in the frequency spectrum structure. According to the typical vibration frequency spectrum distribution of vehicle, the frequency range is divided into three representative intervals, such as 10Hz to 100Hz, 100Hz to 500Hz and 500Hz to 2000Hz, and the integral energy density is calculated in each interval to form a multi-dimensional frequency band energy vector, which is combined with the overall root mean square value of the vibration signal to form a four-dimensional vibration feature vector. This vector can effectively depict the energy distribution characteristics under the dominance of different vibration sources, and reflect the relative dominant relationship between low and high frequency components through the frequency band energy proportion, which provides direct support for identifying vibration sensitive faults caused by resonance or structural unevenness. In the processing of humidity, power supply and electromagnetic interference data, similar statistical and spectral dual analysis methods are used to complete feature extraction. For humidity data, the current humidity mean value, humidity change rate and humidity gradient between the mainboard surface and the environment are extracted, which together reflect the static level and dynamic evolution of the humid conditions suffered by the mainboard. Through sliding time window, the periodic disturbance amplitude and frequency of the humidity signal are identified, and spectral analysis is performed to reveal the potential condensation rhythm or evaporation period, thereby constructing a humidity feature vector that includes both statistical quantities and frequency characteristics. For power supply raw data, the steady-state offset and instantaneous fluctuation of the voltage time series signal are focused on, and the voltage mean value, fluctuation standard deviation and peak-to-peak value are calculated to describe its time stability, and the frequency energy density in the range of 0 to 1kHz is analyzed to identify whether there is a voltage stabilization out of control, switching disturbance or power supply coupling problem, thereby constructing a power supply feature vector that integrates statistical and spectral attributes.For electromagnetic interference signals, the energy response values of three frequency bands of 150 kHz to 30 MHz, 30 MHz to 300 MHz and 300 MHz to 3 GHz are extracted by using a band-pass filter in sections, which respectively represent low-frequency common-mode interference, medium-frequency radio frequency coupling and high-frequency radiation leakage phenomena, and the maximum amplitude, mean square value and high-frequency energy ratio are counted to generate an electromagnetic interference feature vector with spatial breadth and frequency depth. In order to construct the above five types of feature vectors into a unified representation dimension, eliminate dimensional differences and improve the computational efficiency of subsequent state estimation, the Min-Max normalization mechanism is used to compress and map each dimension of the feature quantity, so that its value range is limited in the unified [0, 1] interval, so that the feature quantities of different physical properties have the basic conditions for comparison and fusion. The normalized temperature feature vector, vibration feature vector, humidity feature vector, power supply feature vector and electromagnetic interference feature vector are concatenated and combined in a fixed order to generate a five-dimensional state vector set with consistent structure, time alignment and complete information.

[0036] In a specific embodiment, the process of performing step 103 can specifically include the following steps: The target node of the PCB mainboard is injected with a sinusoidal signal of a predetermined frequency range by a carrier signal injection method, and the injection strength and frequency parameters of the sinusoidal signal are adjusted according to the temperature feature vector, the vibration feature vector and the humidity feature vector in the five-dimensional state vector set to obtain equivalent impedance parameters; Eye diagram analysis is performed based on the digital signal line of the target node, and the eye height parameter, eye width parameter and jitter parameter are monitored and calculated in real time according to the electromagnetic interference feature vector in the five-dimensional state vector set to obtain signal integrity parameters; Noise measurement and transient response characteristic analysis are performed on the power plane of the target node, and the measurement threshold and response time window are adjusted according to the power supply feature vector in the five-dimensional state vector set to obtain power integrity parameters; The equivalent impedance parameters, signal integrity parameters and power integrity parameters are mapped into a vector to obtain a PCB fault feature vector.

[0037] Specifically, a carrier signal injection method is used as the basic testing technique. A low-amplitude sinusoidal signal within a preset frequency range is injected into known critical functional nodes or fault-sensitive areas on the PCB motherboard. The injection frequency is set between 20MHz and 500MHz to ensure that it covers the characteristic frequency band of typical digital circuits while being sufficient to excite high-frequency impedance mutations caused by microstructural defects such as solder joint cracking, interlayer solder joint defects, and via degradation. The frequency and amplitude are dynamically adjusted based on the temperature, vibration, and humidity eigenvectors in the real-time five-dimensional state vector set to enhance the stimulus signal's responsiveness to environmental stress coupling effects. For example, when the temperature eigenvector indicates rapid heating or strong thermal cycling, the system increases the frequency density of the injected signal to stimulate microfracture reflections caused by thermal expansion and contraction. When the vibration eigenvector indicates strong mechanical disturbances, the injection amplitude is appropriately increased to ensure that the impedance ambiguity caused by vibration-coupled mechanical loops is penetrated. When large humidity gradients or rapid changes are detected in the humidity signature, the signal frequency is reduced and a calibration cycle is added to mitigate parasitic conduction errors caused by surface microcondensation. Through an injection mechanism linked to environmental conditions, equivalent impedance parameters, including resistance, capacitance, and inductance coupling characteristics, are calculated based on the signal's amplitude attenuation, phase delay, and echo morphology changes at both ends of the node. These parameters reflect the complex impedance response of the target node under non-ideal structures or in the early stages of failure. Focusing on typical high-speed digital signal lines on motherboards, such as main control MCU outputs, communication module transmit lines, or memory data buses, eye diagram analysis is performed using a high-bandwidth real-time oscilloscope system. By superimposing the eye diagram structure formed by continuously acquired signal waveforms, three core metrics for the current line are determined: eye height, eye width, and jitter. Eye height represents the amplitude margin between signal logic levels, eye width reflects the sampling stability range within the bit period, and jitter reflects signal edge drift caused by clock instability, electromagnetic interference, or sudden load changes. To ensure the analysis results are sensitive to environmental changes, the eye diagram calculation window and calibration reference level are adjusted in real time based on the electromagnetic interference characteristic vectors in the five-dimensional state vector set. When the electromagnetic interference characteristics reflect the presence of energy concentration trends in the medium or high frequency bands, the eye diagram sampling clock is fine-tuned to avoid periodic obstructions caused by modulation interference. At the same time, the analysis tolerance is adjusted according to the interference intensity to improve the recognition accuracy of slight edge disturbances. When the interference spectrum span is large or the intensity changes significantly, the analysis concentration of the target node signal is enhanced through weighted processing, taking into account the spatial characteristics of the interference source, to ensure that the extracted signal integrity parameters can fully cover distortion behaviors such as waveform distortion, timing misalignment, and pattern overlap caused by changes in external electromagnetic fields in the transmission path. In power integrity analysis, the system observes the power supply plane where the target node is located and collects transient response and steady-state noise signals around its operating voltage rail to evaluate its stability performance in response to events such as load disturbances, current spikes, and EMI coupling.In the process, the actual working waveform of the target power supply channel is obtained by a high-speed voltage sampling module, and characteristic parameters such as power ripple amplitude, noise frequency spectrum, recovery time and drop duration are extracted; the measurement threshold and response time window are dynamically adjusted according to the power characteristic vector in the five-dimensional state vector set. When the standard deviation of voltage fluctuation in the power characteristic increases or the peak-to-peak value changes dramatically, the system automatically reduces the noise threshold to capture more detailed fluctuations; when the spectrum analysis result shows that the energy of a specific frequency band increases rapidly, the response window will be shortened to track dynamic interference events in real time, thereby improving the discrimination ability of short-period power failure or voltage distortion, and obtaining power integrity parameters. The equivalent impedance parameters, signal integrity parameters and power integrity parameters are mapped into vectors, and the PCB fault characteristic vector is generated by vector combination.

[0038] In a specific embodiment, the process of performing step 104 can specifically include the following steps: The five-dimensional state vector set is subjected to two-norm calculation to obtain an environmental stress intensity value; According to the material type of the PCB mainboard, the material characteristic parameter combination is determined, wherein the first material parameter is 10.5 and the second material parameter is 0.35 for FR4 material, the first material parameter is 15.8 and the second material parameter is 0.28 for high Tg material, and the first material parameter is 12.3 and the second material parameter is 0.31 for high-frequency material; An adaptive fading factor of the environmental stress intensity value and the material characteristic parameter combination is calculated; Based on the adaptive fading factor, the reciprocal scaling and forward scaling adjustment are performed on the reference system noise covariance matrix and the reference measurement noise covariance matrix to obtain a dynamically adjusted covariance matrix.

[0039] Specifically, the five-dimensional state vector set is subjected to two-norm calculation to obtain a scalar value as the environmental stress intensity value at the current time. This value is essentially the overall synthesis result of the five types of environmental stress levels, reflecting not only the absolute amplitude of the current external disturbance, but also retaining the relative coupling information between different stress sources in the vector structure, thus having good environmental driving perception ability. According to the material type of the measured PCB mainboard, a parameter combination corresponding to its physical properties is extracted from the pre-defined material parameter database. This parameter combination includes two key values: the first material parameter and the second material parameter. The first parameter is used to adjust the exponential sensitivity of environmental stress to the fading factor, and the second parameter is used to adjust the linear weight of the final covariance adjustment amplitude. According to the engineering standard setting, the first material parameter of FR4 material is 10.5, and the second material parameter is 0.35, the high Tg material is 15.8 and 0.28 respectively, and the high frequency material uses 12.3 and 0.31 as the matching combination. The parameter setting is based on a large number of thermal-electric-mechanical physical field joint simulation and accelerated aging test results, and its physical meaning is that the response lag and degradation threshold of different material structures to environmental excitation are different, so introducing material attribute parameters into the covariance matrix updating strategy helps to improve the modeling accuracy and engineering adaptability. The environmental stress intensity value calculated by the foregoing and the selected material characteristic parameter combination are substituted into the adaptive fading mechanism. This mechanism uses a combination of exponential suppression and linear scaling mapping logic, aiming to maintain the convergence of the system covariance structure when the environmental stress is small, and quickly amplify the system response sensitivity when the stress is enhanced, thereby enhancing the recognition ability of the mutant state. In the calculation process, the environmental stress intensity value is squared to enhance the difference in its high value interval, then the square result is mapped with the first material parameter to construct an exponential input parameter, which is adjusted in sign and solved by an exponential function to obtain a suppression value between zero and one, then multiplied by the second material parameter to obtain a fading adjustment value, and finally the adaptive fading factor is calculated by subtracting the adjustment value, whose final value can gradually decrease with the increase of environmental disturbance, thus forming a dynamic response function with physical self-consistency. The fading factor is used as a weight adjustment factor in the updating process of the system noise covariance matrix and the measurement noise covariance matrix. To maintain the sensitivity of state propagation to external disturbances, the inverse scaling of the reference system noise covariance matrix is performed, that is, when the fading factor decreases, the system noise covariance is amplified, the state propagation step is increased, and the response to state mutation is enhanced; while for the measurement noise covariance matrix, the forward scaling mechanism is adopted, that is, when the fading factor decreases, the measurement noise covariance decreases accordingly, so as to weaken the tolerance to observation disturbance and enhance the correction effect of observation on the estimation process. The dynamic adjusted covariance matrix is obtained.

[0040] In a specific embodiment, the process of performing the step of calculating the adaptive fading factor of the combination of the environmental stress intensity value and the material characteristic parameter can specifically include the following steps: Square the environmental stress intensity value to obtain a square result, and divide the square result by the first material parameter in the combination of material characteristic parameters to obtain a quotient value, thereby obtaining a base parameter of the exponential operation; Negatively process the base parameter of the exponential operation, and perform an exponential operation through an exponential function to obtain an exponential operation result; Multiply the exponential operation result by the second material parameter in the combination of material characteristic parameters to obtain a fading adjustment amount, and subtract the value 1 from the fading adjustment amount to obtain the adaptive fading factor.

[0041] Specifically, the environmental stress intensity value is squared to enhance its discrimination in the high stress interval, so that the stronger the environmental disturbance, the faster the square value grows, thereby improving the response sensitivity of the fading mechanism in the high-risk state. After obtaining the square result, the first material parameter corresponding to the current measured PCB mainboard material is extracted according to the preset material characteristic parameter combination, which is set according to the thermal-mechanical-electrical comprehensive resistance characteristics of the material under the condition of multi-physical field coupling, and is used to adjust the amplitude weight of the current environmental stress in the calculation. The squared environmental stress intensity value is used as the dividend, and the first material parameter is used as the divisor to calculate the quotient, obtaining a proportionally adjusted value, which is the base parameter of the exponential operation used to construct the nonlinear fading function. Its physical meaning is that when the environmental stress is small, even if the amplitude is not high after squaring, the exponential base value will still be small after scaling by a large first material parameter, and the subsequent exponential output will tend to 1, while when the environmental stress is severe or the material resistance value is low, the base will quickly increase, constructing a strong inhibition gradient for the exponential function. The base parameter of the exponential operation is processed with a negative sign to convert it into a negatively increasing variable, thereby incorporating it into the fading model of the exponential function. The introduction of the negative sign ensures that the larger the base, the smaller the result of the exponential operation, achieving a mechanism that the system responds faster and adjusts the weight more greatly in the state of higher environmental intensity. The system performs exponential function operation, takes the negative value as input and calculates its exponential result, obtaining a nonlinear value that is always less than or equal to 1, which has the mathematical properties of smooth transition and rapid convergence. Its result presents a monotonically decreasing trend with the change of environmental intensity, especially under high stress conditions, it has the ability to quickly approach zero, thereby effectively promoting the estimation system to improve the sensitivity to sudden states. According to the second material parameter in the material characteristic parameter combination, the above exponential operation result is multiplied by the second material parameter to obtain the fading adjustment amount. The second material parameter represents the response scaling factor of the material when facing disturbances. The larger the value, the more likely the material is to change its structure or electrical characteristics under actual stress, so the state adjustment amplitude required should also be larger; on the contrary, if the parameter is small, it indicates that the material itself has strong environmental stability, and the state estimation can maintain high stability under external disturbances. By multiplying the exponential operation result with the second parameter, a dynamic adjustment factor is obtained that combines the current stress level and material sensitivity. In order to form the final adaptive fading factor, the value 1 is subtracted from the above fading adjustment amount to obtain a fading coefficient ranging from 0 to 1. The smaller the value, the higher the environmental disturbance state or the stronger the material response, the state prediction covariance needs to be improved and the state estimation response speed needs to be enhanced; when the value tends to 1, it indicates that the system is in a relatively stable environment or the material has good resistance, and the estimation system can maintain a low dynamic adjustment amplitude, keeping smoothness and filtering stability.The fading factor is then introduced into the noise covariance adjustment mechanism of the state estimation algorithm, performing inverse scaling on the system noise covariance matrix and forward scaling on the measurement noise covariance matrix, thereby forming a dynamic balance between state propagation and observation update.

[0042] In a specific embodiment, the process of performing step 105 can specifically include the following steps: The state dimension based on the PCB fault feature vector is determined around the state distribution by deterministic sigma point selection, and the sigma point set is calculated by square root decomposition of the dynamically adjusted covariance matrix; The sigma point set and the five-dimensional state vector set are respectively state predicted and observation predicted by the nonlinear state mapping function and the nonlinear observation mapping function, to obtain the state prediction value and the observation prediction value; The predicted covariance matrix and the cross covariance matrix are calculated based on the state prediction value, the observation prediction value and the dynamically adjusted covariance matrix, and the matrix inversion operation is performed based on the predicted covariance matrix and the cross covariance matrix, to obtain the filtering gain parameter; The residual is calculated by the PCB fault feature vector and the observation prediction value, and the state prediction value is updated based on the filtering gain parameter to obtain the fault state estimation value and the fault location information.

[0043] Specifically, based on the PCB fault feature vector, the dimension size of the current state space is determined by analyzing the parameter dimension it contains. The fault feature vector includes multiple electrical characteristic indicators with physical orientation, such as equivalent impedance parameters, signal integrity parameters, and power integrity parameters. The dimension directly determines the structural dimension of the subsequent state distribution. In order to select the deterministic sigma point required for the unscented Kalman filter, 2n+1 sigma points are constructed around the current estimated state around the state space, where n is the state dimension. The selection process is based on the dynamically adjusted covariance matrix at the current time. The square root matrix of the covariance matrix is obtained by square root decomposition, and the column vectors of the matrix are added and subtracted to the current state estimate to generate positive and negative offset points, and together with the center state estimate, they form a complete sigma point set. The set maintains a symmetrical distribution in high-dimensional space, which can maximize the cooperative information between state variables while ensuring computational stability. The generated sigma point set is input into the nonlinear state mapping function, which describes the change rule of state variables in the evolution process over time, and combines with the current five-dimensional environmental state vector set to simulate the propagation path of each sigma point in the state space at the next time. Thus, a set of state prediction values of all sigma points is obtained. In this process, the state mapping function uses a hybrid function based on physical modeling to model the influence mechanism of environmental variables such as temperature change, humidity gradient, and vibration spectrum intensity on electrical parameters, making the state propagation process physically constrained and engineering interpretable. At the same time, each predicted sigma point is also input into the nonlinear observation mapping function, which describes the mapping relationship of state variables in the sensor observation domain, and finally outputs a set of observation prediction values, representing the signal behavior characteristics that the sensor should observe under the current state assumption. The offset between the state prediction values of all predicted sigma points and their mean value is weighted and accumulated to construct a prediction covariance matrix, which reflects the uncertainty range of state variables in the prediction process. At the same time, the joint offset between the state prediction value and the observation prediction value is weighted and superimposed to construct the cross-covariance matrix between state and observation, which reflects the influence degree of state change on observation change. After obtaining the two covariance structures, perform matrix inversion operation, combine the combination relationship of the prediction observation covariance matrix and the cross-covariance matrix, and calculate the filter gain parameter at the current time. The gain value is essentially a weight coefficient between state prediction and observation feedback, which determines whether the system tends to keep the prediction result or follow the actual observation data when updating the state.The residual between the observed PCB fault feature vector and the observed predicted value is calculated to obtain the error vector between the actual sensor response at the current moment and the model prediction. This residual vector, after being adjusted and weighted by the filter gain, is used as a state correction and added to the state prediction value to complete the update process of the current state estimate. Based on the spatial correspondence between the various electrical parameters in the state estimate, a topological mapping function is used to map the parameter anomaly location to the specific motherboard component coordinate space. Combined with the circuit design and physical layout information, the component area most likely to have structural defects or intermittent failures is located, forming complete fault location information including state value, confidence level, coordinate location, and parameter components.

[0044] In a specific embodiment, the execution step of selecting a deterministic sigma point around the state distribution based on the state dimension of the PCB fault feature vector to obtain a set of sigma points may specifically include the following steps: Determine the state dimension according to the PCB fault feature vector, and calculate the scaling parameter by subtracting the state dimension from the value 3 to obtain the scaling parameter of the untraceable transformation; Perform Cholesky decomposition on the dynamically adjusted covariance matrix to obtain a decomposition result, and perform a product operation on the decomposition result with the state dimension and the scaling parameter to obtain a covariance square root matrix; The current state estimate is used as the central sigma point, and the current state estimate is added and subtracted from each column vector of the covariance square root matrix to obtain a positive sigma point set and a negative sigma point set; The central sigma point, the positive sigma point set and the negative sigma point set are combined to obtain the sigma point set.

[0045] Specifically, the state dimension is determined according to the PCB fault feature vector at the current time, which contains a series of highly structured physical quantities such as equivalent impedance parameters, signal integrity parameters, and power integrity parameters. The dimension represents the parameter freedom of the current system state in the modeling process, directly affecting the spatial dimension of the sigma point and the covariance propagation structure. After obtaining the state dimension, the system performs a basic scaling parameter calculation based on the unscented transformation principle. The scaling parameter is obtained by subtracting the state dimension from the constant value 3, which is used to adjust the weight distribution and spatial expansion of the subsequent sigma point generation process. The smaller the scaling parameter value, the larger the state dimension, and the system will tend to be more concentrated in the sigma point distribution, reducing the risk of computational instability in high-dimensional state space. After setting the scaling parameter, the Cholesky decomposition is performed on the dynamic adjustment covariance matrix used at the current time. The covariance matrix is derived from the fusion of the five-dimensional environmental state vector, the adaptive fading factor adjustment, and the estimated covariance structure of the historical state evolution. Its essence is a structured representation of the joint uncertainty between system state variables. Cholesky decomposition, as a square root operation of a symmetric positive definite matrix, can stably decompose the covariance matrix into a lower triangular matrix and its transpose, ensuring the controllability of the output structure in numerical stability and spatial directionality. The decomposition result actually constitutes the uncertainty principal axis matrix in the current state estimation space, and each column represents the standard deviation of a certain independent disturbance direction. The product operation of the lower triangular matrix after Cholesky decomposition, the state dimension, and the scaling parameter calculated earlier gives the covariance square root matrix. The significance of this operation is to quantify the theoretical structure of the covariance matrix into a vector scale that can be added or subtracted, and adjust its influence range through the scaling factor, so that the finally generated sigma points not only satisfy the theoretical variance structure but also maintain sufficient discrete representation to support the nonlinear propagation process. Each column vector in the covariance square root matrix represents the standard deviation in a certain principal direction relative to the current state estimate. Combined with the center state estimate, the positive and negative directions are offset to construct the complete sigma point set. The state estimate at the current time is selected as the center point of the sigma point generation process, which is the position of the first sigma point. Its position in the state space represents the maximum a posteriori probability point at the current time. To complete the construction of the sigma point set, each column vector of the covariance square root matrix is added to the state estimate based on the center point to form a set of positive sigma points. At the same time, the same column vector is subtracted from the state estimate with the opposite sign to generate a corresponding set of negative sigma points.The sigma point construction method centered on the state estimation value and symmetrically expanded to each main shaft direction ensures that all sigma points cover the current state distribution in a statistical sense, and makes them have strong representativeness in the nonlinear transformation process, so that the accuracy of the transformed expected value and covariance structure is maintained to the greatest extent in the subsequent state propagation and observation prediction links. The central sigma point, the positive sigma point set and the negative sigma point set are combined to form a complete sigma point set. The set contains 2n+1 points in total, where n is the state dimension, has the ability to construct an approximate Gaussian distribution in a high-dimensional state space, and can realize the integral approximation of a complex nonlinear function at a low computational cost. It is the basic structure of the core computing framework of the unscented Kalman filter.

[0046] In a specific embodiment, the process of performing residual calculation on the PCB fault feature vector and the observation prediction value, and combining the filter gain parameter to update the state prediction value to obtain the fault state estimation value and the fault location information can specifically include the following steps: Subtracting the observation residual vector from the PCB fault feature vector and the observation prediction value; Calculating the absolute value of each vector element of the observation residual vector to obtain a plurality of vector element absolute values, and comparing the plurality of vector element absolute values with a preset threshold value respectively to obtain a residual absolute value vector and a threshold comparison result; Based on the threshold comparison result, the residual absolute value vector is calculated by the Huber function to obtain a robust weight adjustment factor; Element-wise multiplication operation is performed on the observation residual vector and the robust weight adjustment factor to obtain a multiplication operation result, and the multiplication operation result is combined with the filter gain parameter to perform addition operation on the state prediction value to obtain the fault state estimation value; Based on the abnormal position of the electrical parameter in the fault state estimation value, component coordinate mapping calculation is performed to obtain the fault location information.

[0047] Specifically, subtract the PCB fault feature vector from the observed predicted value, and perform one-to-one difference calculation on the resistance, capacitance, signal integrity or power integrity and other feature components contained in the actual observation vector and the corresponding observation value output by the theoretical prediction model to generate an observation residual vector. The residual vector is not only a direct expression of model error, but also can amplify the difference signal when containing mutation items or observation abnormalities, becoming the input basis of the robustness adjustment mechanism. Perform absolute value calculation operation on each element of the observation residual vector to construct a residual absolute value vector that represents error amplitude without direction. Compare each of the above obtained absolute value elements with the preset residual threshold set one by one, and each vector dimension corresponds to an error tolerance upper limit. The threshold is preset according to system experience data, noise model statistical characteristics or engineering configuration parameters, and reflects the maximum deviation amplitude allowed by the system for various physical indicators. When the absolute residual of a certain dimension exceeds the corresponding threshold, it is considered as an abnormal observation behavior, and the part below the threshold is considered as a deviation within the acceptable range. Through the comparison operation, two intermediate results are obtained, one is the absolute value set of the original residual, and the other is the "whether the threshold is exceeded" judgment result of each dimension, which is used as the decision input for subsequent segmented weight adjustment. Based on the threshold comparison result, Huber function is introduced as the core robustness weight construction mechanism, and segmented function calculation is performed on the residual absolute value of each dimension. Huber function maintains linear response to error when the error is small, maintaining the properties of least squares estimation; while in the interval where the error exceeds the threshold, the function form is compressed to the error amplitude, forming an approximate least absolute value estimation behavior, thereby effectively suppressing the severe influence of large observation deviation on state estimation. According to the judgment result of whether each dimension exceeds the threshold, the linear segment or square root segment of Huber function is applied to the residual value of the dimension, and a set of robustness weight adjustment factors with the same dimension as the residual vector is output. This set of factors represents the correction proportion of each dimension in state update in the form of floating point coefficients, where the closer to 1 indicates that the observation of the dimension is more reliable and should be fully corrected, and the closer to 0 indicates that the observation dimension may be affected by strong interference or nonlinear mutation and should reduce its correction weight to preserve the system's prediction trend. Perform element-by-element multiplication operation on the observation residual vector and the set of robustness weight factors to generate a set of weighted and adjusted correction vectors, each component of which is an observation deviation within the control range and can be considered as the reliable part of the observation update in estimation. Linearly fuse the weighted residual with the filtering gain parameter obtained by the unscented propagation to obtain a set of comprehensive correction vectors, and finally add them to the original state prediction value to realize the estimation update of the fault state at the current time.The update not only considers the stability of the system prediction model, but also dynamically adjusts the observation intervention degree according to the quality of the actual observation signal, so that the state estimation result still maintains high convergence precision and error suppression ability under non-ideal conditions such as sudden noise, electrical interference and signal distortion. The fault state estimation value is updated to extract the abnormal amplitude of the electrical parameter of the state vector in the dimension with the largest abnormal amplitude, and the part with the abnormal amplitude exceeding the experience set threshold is used as the potential failure index. Combined with the PCB main board structure topology, the component layout file and the electrical connection relationship diagram, the circuit path of the abnormal parameter is mapped, and finally the corresponding physical component or local area of the PCB is located. The mapping process considers the mapping matrix between the state variable and the physical space node, and the structure is derived from the device attribute file and the electrical connection netlist in the design stage, which can convert the abnormal structure index in the estimation vector into specific component number, packaging position, layout coordinate and even logical signal line identifier connected thereto. The final output fault location information includes the failure parameter dimension, abnormal degree and confidence, and also provides spatial reference required for maintenance and review in the form of physical coordinates.

[0048] The above describes the vehicle-mounted PCB main board automatic test method in the embodiment of the application. The vehicle-mounted PCB main board automatic test system in the embodiment of the application is described below. Please refer to Figure 2 An embodiment of the vehicle-mounted PCB main board automatic test system in the embodiment of the application includes: The acquisition module 201 is configured to acquire a multi-dimensional environmental stress original data set of the PCB main board. The feature extraction module 202 is configured to perform feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set. The fault feature analysis module 203 is configured to perform fault feature analysis on the target node of the PCB main board based on the five-dimensional state vector set to obtain a PCB fault feature vector. The calculation module 204 is configured to perform adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix. The state estimation module 205 is configured to perform intermittent fault state estimation of the PCB based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information.

[0049] Through the cooperation of the above-mentioned components, by simultaneously collecting temperature cycles, vibration acceleration, humidity gradients, power fluctuations and electromagnetic interference of five-dimensional environmental parameters, a complete vehicle-mounted environmental stress model is constructed, which can fully reflect the real working environment of the PCB mainboard and provide comprehensive environmental excitation information for fault detection. Using the adaptive fading factor calculation method based on the characteristics of PCB materials, the detection parameters can be dynamically adjusted according to different material types (FR4, high Tg, high-frequency materials), which can significantly improve the pertinence and accuracy of fault detection and effectively reduce the false alarm rate. The improved unscented Kalman filter is used for nonlinear state estimation, and through sigma point selection and weight calculation, the dynamic change process of intermittent faults of PCB can be accurately captured, and the accurate tracking and prediction of fault state can be realized. Through the carrier signal injection method for impedance analysis, eye diagram analysis technology for signal integrity detection, noise measurement for power integrity analysis, a multi-dimensional fault feature vector is constructed, which can fully describe the change characteristics of the electrical performance of PCB. The Huber function is used for robust weight adjustment, which can effectively suppress the influence of abnormal measurement values on fault state estimation, and enhance the stability and reliability of the detection system in complex vehicle-mounted environment. Through the component coordinate mapping calculation of the abnormal position of the electrical parameters, the abstract electrical parameter abnormality can be converted into specific physical location information, and the fault is accurately positioned at the component level.

[0050] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, system and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0051] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (read-only memory, ROM), a random access memory (random access memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0052] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalent replacements; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for automated testing of a vehicle-mounted PCB motherboard, characterized in that: include: Collect the original data set of multi-dimensional environmental stress of PCB motherboard; Performing feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set; Performing fault feature analysis on a target node of the PCB mainboard based on the five-dimensional state vector set to obtain a PCB fault feature vector; performing adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix; The intermittent fault state of the PCB is estimated based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information.

2. The vehicle-mounted PCB motherboard automated testing method according to claim 1, characterized in that: The acquisition of the original multi-dimensional environmental stress data set of the PCB mainboard includes: Temperature cycle parameters are collected on the surface of the PCB mainboard to obtain temperature raw data; vibration acceleration parameters are collected on the PCB mainboard to obtain vibration raw data; environmental humidity gradient parameters are collected on the PCB mainboard to obtain humidity raw data; power fluctuation parameters are collected on the vehicle power system to which the PCB mainboard belongs to obtain power raw data; environmental electromagnetic interference parameters are collected on the PCB mainboard to obtain electromagnetic interference raw data; The temperature raw data, the vibration raw data, the humidity raw data, the power raw data and the electromagnetic interference raw data are aggregated and time-synchronized to obtain a multi-dimensional environmental stress raw data set.

3. The vehicle-mounted PCB motherboard automated testing method according to claim 1, characterized in that: The feature extraction and normalization processing are performed on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set, including: Extracting absolute temperature values, calculating temperature change rates, and calculating temperature change rate characteristics of the original temperature data in the multidimensional environmental stress original data set to obtain a temperature characteristic vector; Performing fast Fourier transform processing and power spectrum density calculation on the original vibration data in the multi-dimensional environmental stress original data set to obtain a vibration eigenvector; Performing statistical feature extraction and spectrum analysis on the humidity raw data, power raw data, and electromagnetic interference raw data in the multidimensional environmental stress raw data set to obtain a humidity feature vector, a power feature vector, and an electromagnetic interference feature vector; The temperature eigenvector, the vibration eigenvector, the humidity eigenvector, the power eigenvector, and the electromagnetic interference eigenvector are vector-combined to obtain a five-dimensional state vector set.

4. The vehicle-mounted PCB motherboard automated testing method according to claim 1, characterized in that: The performing fault feature analysis on the target node of the PCB mainboard based on the five-dimensional state vector set to obtain the PCB fault feature vector includes: A sinusoidal signal within a preset frequency range is injected into a target node of the PCB mainboard by a carrier signal injection method, and the injection intensity and frequency parameters of the sinusoidal signal are adjusted according to the temperature eigenvector, the vibration eigenvector, and the humidity eigenvector in the five-dimensional state vector set to obtain an equivalent impedance parameter; performing an eye diagram analysis based on a digital signal line of the target node, and performing real-time monitoring and calculation of an eye height parameter, an eye width parameter, and a jitter parameter based on an electromagnetic interference characteristic vector in the five-dimensional state vector set to obtain a signal integrity parameter; Performing noise measurement and transient response characteristic analysis on the power plane of the target node, and adjusting a measurement threshold and a response time window according to a power characteristic vector in the five-dimensional state vector set to obtain a power integrity parameter; Vector mapping is performed on the equivalent impedance parameter, the signal integrity parameter, and the power integrity parameter to obtain a PCB fault feature vector.

5. The vehicle-mounted PCB mainboard automated testing method according to claim 1, characterized in that: The adaptive fading factor calculation based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix includes: Performing a two-norm calculation on the five-dimensional state vector set to obtain an environmental stress intensity value; Determine a combination of material characteristic parameters according to the material type of the PCB mainboard, wherein the first material parameter corresponding to the FR4 material is 10.5 and the second material parameter is 0.35, the first material parameter corresponding to the high Tg material is 15.8 and the second material parameter is 0.28, and the first material parameter corresponding to the high frequency material is 12.3 and the second material parameter is 0.31; Calculating an adaptive attenuation factor of a combination of the environmental stress intensity value and the material characteristic parameter; Based on the adaptive fading factor, reciprocal scaling and forward scaling are performed on the reference system noise covariance matrix and the reference measurement noise covariance matrix to obtain a dynamically adjusted covariance matrix.

6. The method for automated testing of a vehicle-mounted PCB motherboard according to claim 5, wherein: The calculating of the adaptive attenuation factor of the combination of the environmental stress intensity value and the material characteristic parameter includes: Performing a square operation on the environmental stress intensity value to obtain a square result, and dividing the square result by the first material parameter in the material characteristic parameter combination to calculate a quotient to obtain a base parameter of an exponential operation; The base parameter of the exponential operation is minus-signed, and the exponential operation is performed using an exponential function to obtain an exponential operation result; A product operation is performed on the exponential operation result and the second material parameter in the material characteristic parameter combination to obtain a fading adjustment amount, and a value 1 is subtracted from the fading adjustment amount to obtain an adaptive fading factor.

7. The method for automated testing of a vehicle-mounted PCB motherboard according to claim 1, wherein: The estimating the intermittent fault state of the PCB based on the PCB fault feature vector, the five-dimensional state vector set, and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information includes: Deterministic sigma point selection is performed around the state distribution based on the state dimension of the PCB fault feature vector, and a sigma point set is obtained by square root decomposition calculation of the dynamically adjusted covariance matrix; Performing state prediction and observation prediction on the sigma point set and the five-dimensional state vector set respectively by a nonlinear state mapping function and a nonlinear observation mapping function to obtain a state prediction value and an observation prediction value; Calculating a prediction covariance matrix and a cross covariance matrix based on the state prediction value, the observation prediction value, and the dynamically adjusted covariance matrix, and performing a matrix inversion operation based on the prediction covariance matrix and the cross covariance matrix to obtain a filter gain parameter; The residual calculation is performed on the PCB fault feature vector and the observed prediction value, and the state prediction value is updated in combination with the filter gain parameter to obtain the fault state estimation value and fault location information.

8. The method for automated testing of a vehicle-mounted PCB motherboard according to claim 7, wherein: The state dimension based on the PCB fault feature vector is used to select deterministic sigma points around the state distribution to obtain a sigma point set, including: Determine a state dimension according to the PCB fault feature vector, and calculate a scaling parameter by subtracting the state dimension from a value of 3 to obtain a scaling parameter of an untraceable transformation; Performing Cholesky decomposition on the dynamically adjusted covariance matrix to obtain a decomposition result, and performing a product operation on the decomposition result, the state dimension, and the scaling parameter to obtain a covariance square root matrix; Taking the current state estimate as the central sigma point, and performing addition and subtraction operations on the current state estimate and each column vector of the covariance square root matrix, respectively, to obtain a positive sigma point set and a negative sigma point set; The central sigma point, the positive sigma point set and the negative sigma point set are combined to obtain a sigma point set.

9. The method for automated testing of a vehicle-mounted PCB motherboard according to claim 7, wherein: The residual calculation is performed on the PCB fault feature vector and the observed prediction value, and the state prediction value is updated in combination with the filter gain parameter to obtain the fault state estimation value and the fault location information, including: performing a subtraction operation on the PCB fault feature vector and the observed prediction value to obtain an observation residual vector; Performing absolute value calculation on each vector element of the observed residual vector to obtain a plurality of vector element absolute values, and comparing the plurality of vector element absolute values ​​with preset thresholds to obtain a residual absolute value vector and a threshold comparison result; Based on the threshold comparison result, the residual absolute value vector is segmented calculated by Huber function to obtain a robustness weight adjustment factor; Performing an element-by-element product operation on the observation residual vector and the robustness weight adjustment factor to obtain a product operation result, and performing an addition operation on the state prediction value in combination with the filter gain parameter and the product operation result to obtain a fault state estimation value; Component coordinate mapping calculation is performed based on the abnormal position of the electrical parameter in the fault state estimation value to obtain fault location information.

10. An automated testing system for a vehicle-mounted PCB motherboard, characterized in that: Used to perform the vehicle-mounted PCB mainboard automated testing method according to any one of claims 1 to 9, the vehicle-mounted PCB mainboard automated testing system comprises: Acquisition module, used to collect the original data set of multi-dimensional environmental stress of PCB mainboard; A feature extraction module is used to perform feature extraction and normalization processing on the multi-dimensional environmental stress original data set to obtain a five-dimensional state vector set; A fault feature analysis module is used to perform fault feature analysis on a target node of the PCB mainboard based on the five-dimensional state vector set to obtain a PCB fault feature vector; A calculation module, configured to calculate an adaptive fading factor based on the five-dimensional state vector set to obtain a dynamically adjusted covariance matrix; A state estimation module is used to estimate the intermittent fault state of the PCB based on the PCB fault feature vector, the five-dimensional state vector set and the dynamically adjusted covariance matrix to obtain a fault state estimation value and fault location information.

Images