PCB life analysis method based on reliability analysis and related equipment

Abstract

The invention relates to the technical field of PCB detection, and provides a PCB life analysis method based on reliability analysis and related equipment. The method comprises the following steps: performing multi-physics field coupling modeling on a to-be-detected circuit board to obtain a multi-field coupling data set, performing space-time alignment processing on a collected multi-source sensor data set to obtain a fusion monitoring data set, and performing feature extraction according to the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix; therefore, a dynamic neural network is used for modeling the failure sensitive feature matrix to obtain corresponding life prediction probability distribution data, and the life prediction probability distribution data is adjusted through actual measurement failure data calibration to obtain an accurate life prediction model. And finally monitoring the circuit board to be detected in real time through the life prediction model to obtain real-time life evaluation data. According to the invention, through fusion of multi-physics field coupling and real-time monitoring data, the service life performance of the target in a complex environment can be evaluated more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of PCB detection, and in particular to a PCB life analysis method based on reliability analysis and related equipment. Background Art

[0002] As the cornerstone of the modern electronics industry, printed circuit boards (PCBs) carry the interconnection and signal transmission functions of electronic components and support the global digitalization process. PCB applications cover key areas such as communications, consumer electronics, and automotive electronics. The irreplaceable nature of PCBs makes the reliability of PCBs directly related to the stable operation of social infrastructure.

[0003] PCB life analysis is extremely important. The safety requirements of PCBs have an absolute priority in the fields of medical equipment and avionics. A failure rate of 0.01% may lead to fatal consequences. In addition, the economic requirements of PCBs prompt companies to optimize maintenance cycles through life prediction.

[0004] Existing PCB life analysis methods include environmental stress testing, electrical performance monitoring, microscopic analysis, and accelerated life experiments based on the Arrhenius model. The standards currently implemented by these methods (for example, IPC-9701) only consider the impact of a single stress factor on PCBs, but the actual use environment of PCBs is complex, and their service life is affected by multiple factors. Therefore, existing PCB life analysis methods lack research on the failure mechanism of multi-physical field coupling. Summary of the invention

[0005] In view of this, the present application provides a PCB life analysis method based on reliability analysis and related equipment to solve the problem of PCB multi-physical field coupling life analysis.

[0006] In a first aspect, the present application provides a PCB life analysis method based on reliability analysis, the method comprising: Establishing an original structural model according to the three-dimensional data set of the circuit board to be inspected, and performing multi-physical field coupling modeling processing on the original structural model to obtain a multi-field coupling data set; Collecting target data of the circuit board to be inspected, and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set; Performing feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix; Modeling the failure sensitive characteristic matrix according to a preset dynamic neural network to obtain life prediction probability distribution data; Performing actual failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model; The circuit board to be inspected is subjected to online integrated monitoring according to the life prediction model to obtain real-time life assessment data.

[0007] In an optional embodiment, the step of establishing an original structural model according to the three-dimensional data set of the circuit board to be inspected, and performing multi-physical field coupling modeling processing on the original structural model to obtain a multi-field coupling data set includes: Constructing a geometric model of the three-dimensional data set to obtain a primary geometric structure model of the circuit board to be inspected; Assigning material parameters according to the geometric structure model of the circuit board to be tested to obtain the original structure model of the circuit board to be tested; Performing thermal field modeling processing on the original structural model to obtain thermal field simulation data; Performing mechanical field modeling processing on the original structural model to obtain mechanical field simulation data; Performing electrochemical field modeling processing on the original structural model to obtain electrochemical field simulation data; The thermal field simulation data, the mechanical field simulation data and the electrochemical field simulation data are coupled to obtain the multi-field coupling data set.

[0008] In an optional embodiment, the collecting of target data of the circuit board to be inspected and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set includes: Collecting target data of the circuit board to be detected to obtain the multi-source sensor data set; Performing time synchronization processing on the multi-source sensor data set to obtain a time series synchronized data set; Performing spatial calibration processing on the time-series synchronization data set to obtain a spatiotemporal calibration data set; De-noising and outlier processing are performed on the spatiotemporal calibration data set to obtain an effective synchronous monitoring data set; The synchronous monitoring data set is subjected to data fusion processing to obtain the fused monitoring data set.

[0009] In an optional implementation, the performing feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix includes: Performing time-frequency domain analysis on the thermal field data in the multi-field coupling data set to obtain frequency-domain characteristic data; Performing spatial gradient calculation processing on the mechanical field data in the multi-field coupling data set to obtain thermal stress distribution characteristic data; Performing ion migration feature extraction processing on the electrochemical field data in the multi-field coupling data set to obtain electrochemical field feature data; Performing wavelet packet transform processing on the vibration data in the fused monitoring data set to obtain vibration mode characteristic data; Performing principal component analysis and dimensionality reduction processing on the frequency domain feature data, the thermal stress distribution feature data, the electrochemical field feature data, and the vibration mode feature data to obtain a low-dimensional data set of main failure features; The feature data in the low-dimensional data set is subjected to weighted synthesis processing to obtain the failure-sensitive feature matrix.

[0010] In an optional embodiment, the modeling process of the failure sensitive feature matrix according to a preset dynamic neural network to obtain life prediction probability distribution data includes: Performing time series analysis on the failure sensitive feature matrix to obtain a time series feature set; Performing network structure design on the time series feature set according to the dynamic neural network to obtain a neural network model; Acquire historical fault data according to the circuit board to be detected, and synthesize the historical fault data and the failure sensitive feature matrix to obtain a training data set; Performing back propagation algorithm processing on the neural network model according to the training data set to obtain a target neural network; The failure sensitive feature matrix is ​​subjected to network inference calculation according to the target neural network to obtain the life prediction probability distribution data.

[0011] In an optional embodiment, the performing of measured failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model includes: Step S51, obtaining measured failure data according to the circuit board to be tested, and performing difference analysis processing according to the life prediction probability distribution data and the measured failure data to obtain error distribution; Step S52, performing back propagation processing on the parameters in the preset initial prediction model according to the error distribution to obtain a calibration loss; Step S53, performing gradient descent processing on the calibration loss to obtain calibration model parameters; Step S54, updating the initial prediction model according to the calibration model parameters to obtain a primary life prediction model; Step S55, verifying the primary life prediction model to obtain an error value; Steps S52 to S55 are repeatedly performed until the error value satisfies a preset error range to obtain the life prediction model.

[0012] In an optional embodiment, the performing online integrated monitoring on the circuit board to be inspected according to the life prediction model to obtain real-time life assessment data includes: Performing real-time acquisition on the circuit board to be inspected to obtain a real-time monitoring data set, and performing time-space synchronization processing and line filtering processing on the real-time monitoring data set to obtain real-time data; Performing integrated calculation processing on the real-time data according to the life prediction model to obtain a real-time life estimation value; The real-time life estimation value is subjected to an abnormality detection process to obtain real-time life assessment data.

[0013] A second aspect of the present application provides a PCB life analysis device based on reliability analysis, the device comprising: A coupling data module, used to establish an original structure model according to the three-dimensional data set of the circuit board to be tested, and perform multi-physical field coupling modeling processing on the original structure model to obtain a multi-field coupling data set; A monitoring data module is used to collect target data of the circuit board to be inspected, and perform spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set; A failure matrix module is used to perform feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix; A prediction data module, used for modeling the failure sensitive feature matrix according to a preset dynamic neural network to obtain life prediction probability distribution data; A prediction model module, used for performing measured failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model; The life evaluation module is used to perform online integrated monitoring on the circuit board to be tested according to the life prediction model to obtain real-time life evaluation data.

[0014] A third aspect of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the PCB life analysis method based on reliability analysis as described above when executing the computer program.

[0015] A fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the PCB life analysis method based on reliability analysis as described above.

[0016] In summary, this application at least includes the following beneficial technical effects: 1. Through multi-physics field coupling modeling, target data collection, feature extraction and dimensionality reduction, combined with a dynamic neural network model, the life of the PCB can be accurately predicted.

[0017] 2. During the life prediction process, the prediction model can be corrected in real time through the actual failure data calibration technology, thus improving the accuracy of the prediction results.

[0018] 3. By integrating and analyzing multiple data sources and combining them with real-time monitoring data, a more reliable and comprehensive lifespan prediction can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is a flow chart of a PCB life analysis method based on reliability analysis provided in an embodiment of the present application; Figure 2 It is a functional module diagram of a PCB life analysis device based on reliability analysis provided in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0022] like Figure 1 , which is a flow chart of a PCB life analysis method based on reliability analysis provided in an embodiment of the present application. The PCB life analysis method based on reliability analysis provided in an embodiment of the present application includes the following steps.

[0023] Step S1, establishing an original structure model according to a three-dimensional data set of a circuit board to be inspected, and performing multi-physical field coupling modeling processing on the original structure model to obtain a multi-field coupling data set.

[0024] The three-dimensional data set of the circuit board to be inspected (hereinafter collectively referred to as the target PCB) acquired by the scanning device is obtained through a wireless communication protocol, so as to generate a primary geometric structure model of the target PCB using an integrated three-dimensional modeling tool (for example, computer-aided design (CAD) software or finite element analysis (FEA) software). The primary geometric structure model includes but is not limited to key structural features such as wiring topology, solder joints, holes, component pins, etc., which are used to accurately reflect the physical appearance characteristics of the target PCB.

[0025] According to the material and composition of the target PCB, appropriate physical parameters are assigned to the primary geometric structure model, where the physical parameters include but are not limited to different material distribution densities, specific heat capacity, elastic modulus, thermal expansion coefficient and other physical properties. For example, the substrate of the circuit board may use epoxy resin composite materials, the soldering metal uses tin-lead alloy or lead-free alloy, and the copper foil layer is a conductive material for electrical connection. It should be understood that the process of assigning material parameters requires not only accurate assignment of the physical properties of each part, but also consideration of the changing behaviors of these substances under working conditions, such as thermal expansion characteristics, stress-strain behavior, and conductive properties.

[0026] The temperature distribution of the target PCB in the working environment is simulated by a preset thermodynamic model. This process involves applying the heat conduction equation to the original structural model, taking into account the Joule heating effect generated by current conduction and the heat exchange effect of the external environment. Thermal field modeling is based on Fourier's law, and thermal field simulation data is obtained by solving the following transient heat conduction equation: in, is the material density, is the specific heat capacity, is the temperature, is the thermal conductivity, Joule heat source (calculated from current density), is the environmental heat exchange item (including but not limited to convection and radiation). By calculating the thermal field simulation data, the temperature distribution of the target PCB under working conditions can be captured, especially the temperature changes around the hot spots (such as BGA solder joints).

[0027] At the same time, mechanical analysis is used to evaluate the stress distribution of the target PCB under stress. In order to analyze the stress response of the target PCB due to thermal expansion, mechanical force or other external loads, a nonlinear elastic-plastic constitutive model is used to perform stress-strain analysis. The mechanical field simulation data is obtained by calculating the following formula: in, is the equivalent stress, is the deviatoric stress tensor, is the coefficient of thermal expansion, is the temperature change, is the elastic modulus. Mechanical analysis can be used to predict the thermal stress concentration area of ​​the circuit board during operation, especially the creep behavior of the solder joints and the fatigue cracks of the copper foil.

[0028] At the same time, the electrochemical field modeling process is used to simulate the ion migration phenomenon of the target PCB during use, especially the growth of conductive anode filaments that may occur under long-term thermal cycles or humidity changes. The electrochemical field simulation data is obtained by calculating the following formula: in, is the ith ion flux, is the diffusion coefficient, is the charge number, is the mobility, is the ion concentration, By simulating the ion migration process based on the Nernst-Planck equation, the electrochemical corrosion process in the target PCB can be simulated, especially in a humid environment, how electrolysis leads to the growth of CAF, thereby affecting the long-term reliability of the circuit board.

[0029] After obtaining the thermal field simulation data, mechanical field simulation data, and electrochemical field simulation data, the thermal, mechanical, and electrochemical effects are combined together by using a coupling algorithm (for example, the multi-field coupling method in finite element analysis (FEA)) to obtain a global physical field model data (i.e., a multi-field coupling data set) that comprehensively considers temperature changes, stress distribution, and ion migration.

[0030] Step S2: collecting target data of the circuit board to be inspected, and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set.

[0031] It should be understood that various types of sensors are deployed in the target PCB to monitor the operating data of the PCB, and are equipped with corresponding acquisition systems. Among them, the sensors include but are not limited to temperature sensors, vibration sensors, humidity sensors, etc., to ensure that the working status of the circuit board can be monitored from different angles. When arranging the sensors, the specific installation location should be selected according to the layout of the circuit board, thermal stress hotspots, and the location of key components (such as solder joints, component pins, and PCB copper foil layers). This ensures that the data of key parts can be captured in a timely manner. Multi-source sensor data sets are obtained from the acquisition system through industrial Ethernet protocols (for example, Profinet), where the multi-source sensor data sets include but are not limited to temperature data, vibration data, humidity data, current or voltage data, etc.

[0032] Since the data collected from multiple sources of sensors often have different timestamps, this difference will affect the accuracy of the data and the subsequent analysis results. Therefore, after obtaining the multi-source sensor data set, the data timestamp error of different sensors is controlled to the microsecond level (±1μs) through network clock synchronization according to the IEEE-1588 precision clock protocol. By setting a reference clock to unify and coordinate the data streams of each sensor, the sampling time of all sensors will be calibrated based on this clock, and the time error will be corrected through the following compensation algorithm, so as to obtain a synchronized data set with a consistent timestamp.

[0033] in, , Collect timestamps for different sensors, The unified timestamp after synchronization. Through time synchronization processing, the timestamps of different sensors are unified to a standard time base to ensure the time consistency of each sensor data. Thus, the multi-source sensor data set is converted into a time-series synchronized data set.

[0034] At the same time, it is necessary to spatially align the timing synchronization data set with the finite element simulation model grid of the target PCB so that the sensor data can be accurately mapped to the spatial nodes of the simulation model. Specifically, the spatial position of the sensor data and the node position of the simulation grid need to be calibrated one-to-one. By accurately measuring the position of the sensor and converting it into the corresponding node in the coordinate system of the finite element grid, the spatial relationship between the sensor and the simulation grid can be determined. It is usually necessary to use transformation methods such as rotation and translation to convert the sensor coordinate system into the simulation grid coordinate system to achieve spatial alignment. The spatial calibration process is specifically implemented through the following formula: in, is the position vector of the sensor in the sensor coordinate system, is the position vector in the finite element mesh, is the coordinate transformation matrix, is the translation vector. The sensor data can be accurately mapped to the finite element mesh nodes through spatial calibration.

[0035] During the sensor data collection and processing process, due to various factors (such as sensor noise, electromagnetic interference, environmental fluctuations, etc.), the collected raw data often contains noise and outliers. Therefore, before further data analysis, it is necessary to perform denoising and outlier detection on the spatiotemporal calibration data set after spatial calibration. Denoising can be performed using algorithms such as Kalman filtering or median filtering. The embodiment of the present application uses the Kalman filtering algorithm to filter the sensor data through prediction and correction steps to reduce the impact of noise. The specific formula is as follows: in, is the estimated value after filtering, is the predicted value, is the Kalman gain matrix, is the observation matrix, is the observed value.

[0036] After filtering, the spatiotemporal calibration data set is subjected to statistical methods (e.g., threshold judgment based on the mean and standard deviation of the data) or anomaly detection algorithms based on machine learning to detect outliers, thereby removing outliers that affect accuracy. By denoising and outlier processing the spatiotemporal calibration data set, the obtained synchronous monitoring data can be ensured to be more accurate and reliable.

[0037] It should be understood that data fusion is the integration of data from multiple sources to obtain a comprehensive and accurate monitoring data set. In the embodiment of the present application, the fusion process is usually performed by methods such as Kalman filtering and particle filtering. At the same time, the data from different sensors are weighted and merged to ensure that the data of each sensor is appropriately valued during the fusion process. Through data fusion processing, a fused monitoring data set (i.e., a fused monitoring data set) can be finally obtained. The fused monitoring data set contains data from multiple source sensors and has temporal and spatial consistency and high precision.

[0038] Step S3: performing feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix.

[0039] Since thermal field data is a multidimensional spatiotemporal data set, in which time and space factors have a profound impact on failure modes, time domain analysis can accurately extract features related to possible failure of PCB boards from thermal field data. The embodiment of the present application uses wavelet transform as a tool for time-frequency analysis. The wavelet packet transform can decompose the signal into different frequency components and analyze the local characteristics of the signal through different scales. For thermal field data, frequency domain information related to different temperature change characteristics, such as transient thermal response, heat flow distribution, etc., especially the change pattern related to the area of ​​thermal stress concentration can be extracted through wavelet packet transform. The specific formula is as follows: in, is the coefficient of wavelet transform, which represents the response of the signal at different scales and displacements. is the original thermal field data, time domain signal. is the wavelet basis function, which represents the local time-frequency characteristics of the signal. is a scale parameter that determines the decomposition scale of the signal. is the translation parameter, which represents the displacement of the signal in the time domain.

[0040] At the same time, the mechanical field data is analyzed through spatial gradient calculation, especially the interaction between temperature and mechanical stress, to determine the thermal stress concentration area. These areas are usually closely related to failure modes such as solder joint cracking and copper foil fatigue. The embodiment of the present application performs a second-order spatial gradient calculation on the temperature distribution in the mechanical field data. The second-order spatial gradient can accurately reflect the severity of temperature changes, that is, it can effectively identify the thermal stress concentration area. By calculating the second-order derivative of the temperature field in different directions, the spatial distribution of thermal stress is obtained. The specific formula is as follows: in, is the second-order gradient of the temperature field, indicating the degree of curvature of the temperature distribution. Represents the second-order derivative of temperature along the three spatial directions of x, y, and z. By calculating the second-order temperature gradient, it is possible to identify areas with drastic temperature changes, which are usually where thermal stress accumulates and may lead to solder joint fatigue or other heat-related failure modes in the PCB board.

[0041] It should be understood that electrochemical processes (e.g., conductive anode filament (CAF) growth) are one of the key factors affecting the life of the target PCB. Therefore, it is necessary to extract ion migration features that may cause CAF growth from the electrochemical field data. The electrochemical field data in the multi-field coupling data set is obtained by calculating the ion migration rate of the electrochemical field data based on the Nernst-Planck equation. The Nernst-Planck equation can describe the migration rate of ions under the action of the electric field, thereby helping to identify the electrochemical reaction area under different electric fields and concentration gradients. The characteristic data related to the CAF growth path can be extracted through the calculation process. Further feature extraction from the electrochemical field data through the preset machine recognition model helps to identify the growth path of the conductive anode filament and help analyze potential electrochemical failure modes.

[0042] The vibration data in the fused monitoring data set can effectively reflect the abnormal conditions caused by thermal stress, mechanical load or electrochemical reaction during the use of the PCB board. The embodiment of the present application also uses wavelet packet transform to analyze the vibration data. The wavelet packet transform can extract the local characteristics of the vibration signal from multiple frequency bands and identify the vibration characteristics related to solder joint cracks or other structural failures. The specific process is similar to the time-frequency analysis of thermal field data. For details, please refer to the calculation process of the time-frequency analysis of thermal field data, which will not be repeated here.

[0043] After obtaining the frequency domain feature data, thermal stress distribution feature data, electrochemical field feature data, and vibration mode feature data, principal component analysis (PCA) is used to reduce the dimensionality of the high-dimensional data set. PCA is a commonly used dimensionality reduction technique that converts data into a new coordinate system to remove redundant features and retain the main features. The reduced-dimensional data set can help better identify the features most relevant to the target PCB failure, and the formula Y=XW is shown.

[0044] Among them, Y is the feature matrix after dimensionality reduction, which contains the main failure characteristics. X is the original feature matrix, which contains multi-dimensional thermal field, mechanical field, vibration and electrochemical field data. W is the feature vector matrix, which contains the coefficients of PCA transformation. The complexity of the data is reduced by dimensionality reduction, while retaining the key information related to the failure, which is convenient for subsequent analysis.

[0045] Finally, the weighted method is used to combine the relative importance of different features to synthesize the reduced features into a final failure-sensitive feature matrix. This process is based on the weighting of the contribution between the features to form a feature matrix (i.e., failure-sensitive feature matrix) for comprehensive evaluation of PCB life.

[0046] Step S4: Modeling the failure sensitive feature matrix according to a preset dynamic neural network to obtain life prediction probability distribution data.

[0047] It should be understood that the failure-sensitive feature matrix usually contains data at multiple different time points, including temperature, stress, vibration and other physical quantities, and these features often have intrinsic correlation in the time dimension. In order to build a neural network model that can process time series data, it is necessary to organize these features into a time series feature set in chronological order. The embodiment of the present application also uses a wavelet transform method to convert time domain signals into frequency domains, and extracts important time features related to failure by performing time-frequency domain analysis on these signals. By analyzing the frequency components of signals such as temperature changes and vibration amplitudes of various components of the target PCB (for example, solder joints, copper foils, etc.), time series features reflecting key failure modes such as fatigue and aging of the target PCB can be extracted. Through time-frequency domain analysis, the changes of each feature at different time scales and frequencies can be obtained, and these time series features will be used as inputs to the neural network model. This process helps to capture the subtle differences between normal operation and potential failures, and enhances the network's ability to identify failure modes.

[0048] The neural network is designed based on the time series feature set and the preset network neural framework (i.e., dynamic neural network). The input layer of the neural network is defined, and the input of this layer is the feature data extracted from the time series feature set. Then, the number of hidden layers and the number of neurons in each layer of the network are designed, and the appropriate activation function (e.g., ReLU, sigmoid, or tanh) is selected. The network design also includes recursive neural network structures, especially long short-term memory (LSTM) or gated recurrent unit (GRU) networks, which are used to capture long-term dependencies in time series. In the specific design, the parameters of the neural network need to be reasonably initialized and the appropriate loss function (e.g., mean square error loss function) needs to be selected to ensure that the network can effectively converge during the training process. Through recursive neural networks (e.g., LSTM), the model can handle the time dependency of the input data and capture the time series patterns related to life prediction. The training process of the network will reasonably distribute the weight of the influence of each time series feature on the final life prediction result.

[0049] At the same time, historical failure data is used as a supervisory signal. Historical failure data contains relevant information when the circuit board to be tested fails in actual use, such as actual life, failure type, etc. These data can be obtained through past tests, laboratory tests, or field data collection. The historical failure data is combined with the failure sensitive feature matrix to form a training data set. Specifically, the historical failure data is used as the target output, and the failure sensitive feature matrix is ​​used as the input feature data. In this way, the neural network can learn the relationship between features and actual failures through historical data, thereby performing effective life prediction.

[0050] After the neural network structure is designed, the neural network is trained using a training data set. The specific training process includes forward propagation and back propagation. Forward propagation passes the input data to the network and generates a predicted output through the calculation of each layer of neurons. Then, based on the error between the predicted result and the actual fault data, the back propagation algorithm is used to adjust the weights and biases in the neural network. During the training process, the gradient descent algorithm or its variants (for example, the Adam optimization algorithm) are used to optimize the network parameters until the loss function is minimized or the set number of iterations is reached. Finally, the trained neural network (i.e., the target neural network) can perform life prediction for new input data.

[0051] After obtaining the trained target neural network, the new failure-sensitive feature matrix is ​​inferred and calculated. At this point, the network will calculate the propagation process of the input features in each layer of neurons through forward propagation, and finally output the life prediction probability distribution data. This data is usually expressed as a probability density function of life, and common forms include Weibull distribution or Log-normal distribution.

[0052] Step S5: calibrate the life prediction probability distribution data with actual failure data and perform prediction modeling processing to obtain a life prediction model.

[0053] Step S51, obtaining failure data of the circuit board to be tested from the actual operating environment. The measured failure data generally comes from instances of failure or damage that occurred during the actual operation process, including but not limited to information such as the actual life of the target circuit board, failure mode, and working environment conditions. The measured failure data is usually obtained through sensor networks, manual inspections, or past historical records. Compare the obtained measured failure data with the life prediction probability distribution data previously generated based on multi-physics field simulation and fusion monitoring data, and perform a difference analysis. Difference analysis is mainly used to evaluate the deviation between the predicted life and the actual life. This deviation reflects the accuracy of the model under specific working environments and actual conditions. Through the formula shown , calculate the error distribution. Among them, is the actual lifespan, To predict life expectancy, is the error distribution. The error distribution is used to calculate the deviation between the quantitative prediction model and the actual data, providing error information for model calibration.

[0054] Step S52, perform back propagation processing of model parameters according to the error distribution data. It should be understood that the preset initial prediction model is obtained by finite element analysis and neural network training in the initial state, and is used to deduce the real-time life estimation value of the target PCB according to the real-time operating condition data of the target PCB. However, due to deviations such as environmental factors and actual working conditions, the initial prediction model may not be completely accurate. Therefore, back propagation is used to adjust the internal parameters of the model to minimize errors and improve the prediction ability of the model. Specifically, the following formula is used: in, is the loss function, where are the parameters of the initial prediction model. is the actual life span of the ith sample. is the predicted life span of the ith sample. N is the total number of samples. Backpropagation is performed by calculating the gradient of the loss function with respect to the model parameters (i.e., calibration loss).

[0055] Step S53: The obtained calibration loss is input into the gradient descent algorithm. Gradient descent is an optimization algorithm used to minimize the loss function. The algorithm gradually reduces the value of the loss function by updating the model parameters. The main steps of gradient descent are to calculate the gradient of the loss function and adjust the parameters according to the gradient direction. The specific update formula of gradient descent is as follows: in, are the values ​​of the model parameters before and after updating, The learning rate is used to control the step size of each parameter update. is the gradient of the loss function with respect to the parameters. Gradient descent is used to move along the gradient direction of the loss function and select an appropriate learning rate (step size) to update the model parameters. The choice of learning rate is crucial. Too large a step size may lead to unstable updates, while too small a step size may make the optimization process too slow.

[0056] Step S54, obtain calibration model parameters by gradient descent. At this time, based on the updated parameters, the initial prediction model needs to be updated to form a preliminarily calibrated primary life prediction model.

[0057] Step S55, after obtaining the primary life prediction model, the primary life prediction model needs to be verified. The accuracy of the primary life prediction model is tested by applying it to new measured failure data. The error value between the predicted result and the actual result is calculated, and the error value is compared with the preset error range to verify the validity of the primary life prediction model. If the error is large, it means that the primary life prediction model still needs further calibration and verification until the accuracy requirements are met. If the error is small, it means that the calibration process has been successful, the primary life prediction model can more accurately reflect the actual failure situation, and the primary life prediction model is determined as the life prediction model.

[0058] The parameters in the preset model are gradually optimized through multiple iterations to ensure that the final life prediction model has high accuracy and robustness.

[0059] Step S6: Perform online integrated monitoring on the circuit board to be tested according to the life prediction model to obtain real-time life assessment data.

[0060] The real-time monitoring data set of the target PCB is acquired in real time through an industrial Ethernet protocol (such as Profinet or Ethernet / IP). The collected real-time monitoring data set includes multi-source and multi-dimensional physical signals. The real-time monitoring data set is subjected to time synchronization, spatial calibration, and linear filtering to obtain accurate real-time data. The time-space synchronization processing and linear filtering processing are similar to the process of step S2. For details, please refer to the relevant processing process of step S2, which will not be repeated here.

[0061] It should be understood that the life prediction model obtained in step S5 is usually constructed based on a probabilistic graphical model, a neural network or other dynamic modeling methods, and the preliminary model training has been completed through the calibration and update of step S5. The input of the life prediction model is the real-time data characteristics closely related to the failure of the target PCB, such as temperature, vibration, and electrochemical stress. Specifically, the input data is passed to each node of the life prediction model, and the connection relationship between the nodes is calculated layer by layer to finally obtain the remaining service life of the target PCB (i.e., the real-time life estimate). This process is based on the life probability distribution data output by the prediction algorithm, including the reliability of life, risk assessment, and life distribution.

[0062] After obtaining the real-time life estimate, anomaly detection is performed to ensure the rationality and accuracy of the prediction results. The goal of anomaly detection is to identify life estimates that deviate significantly from normal usage patterns, so that timely measures can be taken to prevent potential faults or failures. Specifically, through statistical analysis methods, the life distribution and fluctuation range under normal use are obtained based on historical data. This range is used to set a dynamic threshold, and when the real-time life estimate exceeds the threshold, it is considered an anomaly. These anomalies are usually caused by hardware failures, environmental changes, or sensor errors. The formula for anomaly detection is as follows: Among them, RUL i is the remaining life value measured for the i-th time in the historical data, is the mean of historical data, is the standard deviation of historical data, and n is the number of samples of historical data. Perform statistical analysis on historical data and calculate the mean and standard deviation.

[0063] Finally, after anomaly detection processing, the real-time life assessment data obtained can be used for further decision support. If the remaining life is lower than the preset critical value, the system can trigger a maintenance notification or warning to help equipment maintenance personnel intervene in time to avoid losses caused by potential failures.

[0064] The present application is applied to the field of PCB detection technology, by performing multi-physical field coupling modeling on the circuit board to be detected to obtain a multi-field coupling data set, and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set, and then performing feature extraction based on the multi-field coupling data set and the fused monitoring data set to obtain a failure sensitive feature matrix, thereby using a dynamic neural network to model the failure sensitive feature matrix to obtain corresponding life prediction probability distribution data, and then adjusting the life prediction probability distribution data through actual failure data calibration to obtain an accurate life prediction model, and finally using the life prediction model to monitor the circuit board to be detected in real time to obtain real-time life evaluation data. The present application can more accurately evaluate the life performance of a target in a complex environment through the fusion of multi-physical field coupling and real-time monitoring data.

[0065] like Figure 2 As shown, it is a functional module diagram of a PCB life analysis device based on reliability analysis provided in an embodiment of the present application.

[0066] In some embodiments, the PCB life analysis device 2 based on reliability analysis may include a plurality of functional modules composed of computer program segments. The computer program of each program segment in the PCB life analysis device 2 based on reliability analysis may be stored in a memory of a server and executed by at least one processor to execute (see Figure 1 Description) Functions of PCB life analysis method based on reliability analysis.

[0067] In this embodiment, the PCB life analysis device 2 based on reliability analysis can be divided into multiple functional modules according to the functions it performs. The functional modules may include: a coupling data module 21, a monitoring data module 22, a failure matrix module 23, a prediction data module 24, a prediction model module 25 and a life assessment module 26. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0068] The coupling data module 21 is used to establish an original structure model according to the three-dimensional data set of the circuit board to be detected, and perform multi-physical field coupling modeling processing on the original structure model to obtain a multi-field coupling data set.

[0069] In an optional implementation, the coupling data module 21 is specifically used for: Constructing a geometric model of the three-dimensional data set to obtain a primary geometric structure model of the circuit board to be inspected; Assigning material parameters according to the geometric structure model of the circuit board to be tested to obtain the original structure model of the circuit board to be tested; Performing thermal field modeling processing on the original structural model to obtain thermal field simulation data; Performing mechanical field modeling processing on the original structural model to obtain mechanical field simulation data; Performing electrochemical field modeling processing on the original structural model to obtain electrochemical field simulation data; The thermal field simulation data, the mechanical field simulation data and the electrochemical field simulation data are coupled to obtain the multi-field coupling data set.

[0070] The monitoring data module 22 is used to collect target data of the circuit board to be inspected, and perform spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set.

[0071] In an optional implementation, the monitoring data module 22 is specifically used for: Collecting target data of the circuit board to be detected to obtain the multi-source sensor data set; Performing time synchronization processing on the multi-source sensor data set to obtain a time series synchronized data set; Performing spatial calibration processing on the time-series synchronization data set to obtain a spatiotemporal calibration data set; De-noising and outlier processing are performed on the spatiotemporal calibration data set to obtain an effective synchronous monitoring data set; The synchronous monitoring data set is subjected to data fusion processing to obtain the fused monitoring data set.

[0072] The failure matrix module 23 is used to perform feature extraction and dimension reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix.

[0073] In an optional implementation, the failure matrix module 23 is specifically used for: Performing time-frequency domain analysis on the thermal field data in the multi-field coupling data set to obtain frequency-domain characteristic data; Performing spatial gradient calculation processing on the mechanical field data in the multi-field coupling data set to obtain thermal stress distribution characteristic data; Performing ion migration feature extraction processing on the electrochemical field data in the multi-field coupling data set to obtain electrochemical field feature data; Performing wavelet packet transform processing on the vibration data in the fused monitoring data set to obtain vibration mode characteristic data; Performing principal component analysis and dimensionality reduction processing on the frequency domain feature data, the thermal stress distribution feature data, the electrochemical field feature data, and the vibration mode feature data to obtain a low-dimensional data set of main failure features; The feature data in the low-dimensional data set is subjected to weighted synthesis processing to obtain the failure-sensitive feature matrix.

[0074] The prediction data module 24 is used to perform modeling processing on the failure sensitive characteristic matrix according to a preset dynamic neural network to obtain life prediction probability distribution data.

[0075] In an optional implementation, the prediction data module 24 is specifically used to: Performing time series analysis on the failure sensitive feature matrix to obtain a time series feature set; Performing network structure design on the time series feature set according to the dynamic neural network to obtain a neural network model; Acquire historical fault data according to the circuit board to be detected, and synthesize the historical fault data and the failure sensitive feature matrix to obtain a training data set; Performing back propagation algorithm processing on the neural network model according to the training data set to obtain a target neural network; The failure sensitive feature matrix is ​​subjected to network inference calculation according to the target neural network to obtain the life prediction probability distribution data.

[0076] The prediction model module 25 is used to perform measured failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model.

[0077] In an optional implementation, the prediction model module 25 is specifically used for: Step S51, obtaining measured failure data according to the circuit board to be tested, and performing difference analysis processing according to the life prediction probability distribution data and the measured failure data to obtain error distribution; Step S52, performing back propagation processing on the parameters in the preset initial prediction model according to the error distribution to obtain a calibration loss; Step S53, performing gradient descent processing on the calibration loss to obtain calibration model parameters; Step S54, updating the initial prediction model according to the calibration model parameters to obtain a primary life prediction model; Step S55, verifying the primary life prediction model to obtain an error value; Steps S52 to S55 are repeatedly performed until the error value satisfies a preset error range to obtain the life prediction model.

[0078] The life assessment module 26 is used to perform online integrated monitoring on the circuit board to be inspected according to the life prediction model to obtain real-time life assessment data.

[0079] In an optional implementation, the life assessment module 26 is specifically used for: Performing real-time acquisition on the circuit board to be inspected to obtain a real-time monitoring data set, and performing time-space synchronization processing and line filtering processing on the real-time monitoring data set to obtain real-time data; Performing integrated calculation processing on the real-time data according to the life prediction model to obtain a real-time life estimation value; The real-time life estimation value is subjected to an abnormality detection process to obtain real-time life assessment data.

[0080] It should be understood that the various variations and specific embodiments of the methods provided in the above embodiments are also applicable to the PCB life analysis device based on reliability analysis in this embodiment. Through the above detailed description of the PCB life analysis method based on reliability analysis, those skilled in the art can clearly understand the implementation method of the PCB life analysis device based on reliability analysis in this embodiment. For the sake of brevity of the specification, it will not be described in detail here.

[0081] like Figure 3 , which is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0082] In a preferred embodiment of the present invention, the electronic device 3 may include, but is not limited to: a memory 31 , at least one processor 32 and at least one communication bus 33 .

[0083] Those skilled in the art should understand that Figure 3 The structure of the electronic device 3 shown does not constitute a limitation of the embodiment of the present invention, and the electronic device 3 may also include more or less other hardware or software than shown in the figure, or a different arrangement of components.

[0084] In some embodiments, the electronic device 3 is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors and embedded devices.

[0085] It should be noted that the electronic device 3 is only an example, and other existing or future electronic products that are suitable for the present application should also be included in the protection scope of the present application and included here by reference.

[0086] In some embodiments, the memory 31 stores a computer program, and when the computer program is executed by the at least one processor 32, all or part of the steps in the PCB life analysis method based on reliability analysis are implemented. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc memory, magnetic disk memory, magnetic tape memory, or any other computer-readable medium that can be used to carry or store data. Further, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, and the like.

[0087] In some embodiments, the at least one processor 32 is the control core (ControlUnit) of the electronic device 3, and uses various interfaces and lines to connect various components of the entire electronic device 3, and executes various functions and processes data of the electronic device 3 by running or executing programs or modules stored in the memory 31, and calling data stored in the memory 31. For example, when the at least one processor 32 executes the computer program stored in the memory 31, it implements all or part of the steps of the PCB life analysis method based on reliability analysis described in the embodiment of the present application; or implements all or part of the functions of the PCB life analysis device based on reliability analysis. The at least one processor 32 can be composed of an integrated circuit, for example, it can be composed of a single packaged integrated circuit, or it can be composed of multiple integrated circuits with the same function or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips.

[0088] In some embodiments, the at least one communication bus 33 is configured to realize the connection and communication between the memory 31 and the at least one processor 32. Although not shown, the electronic device 3 may also include a power supply (such as a battery) for powering each component. Preferably, the power supply may be logically connected to the at least one processor 32 through a power management device, so as to realize the functions of managing charging, discharging, and power consumption management through the power management device. The power supply may also include any components such as one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 3 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0089] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling an electronic device (which can be a personal computer, electronic device, or network device, etc.) or a processor to execute part of the method described in each embodiment of the present application.

[0090] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0091] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0092] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A PCB life analysis method based on reliability analysis, characterized in that: The method comprises: Establishing an original structural model according to the three-dimensional data set of the circuit board to be inspected, and performing multi-physical field coupling modeling processing on the original structural model to obtain a multi-field coupling data set; Collecting target data of the circuit board to be inspected, and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set; Performing feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix; Modeling the failure sensitive characteristic matrix according to a preset dynamic neural network to obtain life prediction probability distribution data; Performing actual failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model; The circuit board to be inspected is subjected to online integrated monitoring according to the life prediction model to obtain real-time life assessment data.

2. The PCB life analysis method based on reliability analysis according to claim 1 is characterized in that: The method of establishing an original structure model according to the three-dimensional data set of the circuit board to be tested, and performing multi-physical field coupling modeling processing on the original structure model to obtain a multi-field coupling data set includes: Constructing a geometric model of the three-dimensional data set to obtain a primary geometric structure model of the circuit board to be inspected; Assigning material parameters according to the geometric structure model of the circuit board to be tested to obtain the original structure model of the circuit board to be tested; Performing thermal field modeling processing on the original structural model to obtain thermal field simulation data; Performing mechanical field modeling processing on the original structural model to obtain mechanical field simulation data; Performing electrochemical field modeling processing on the original structural model to obtain electrochemical field simulation data; The thermal field simulation data, the mechanical field simulation data and the electrochemical field simulation data are coupled to obtain the multi-field coupling data set.

3. The PCB life analysis method based on reliability analysis according to claim 1 is characterized in that: The step of collecting target data of the circuit board to be inspected and performing spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set includes: Collecting target data of the circuit board to be detected to obtain the multi-source sensor data set; Performing time synchronization processing on the multi-source sensor data set to obtain a time series synchronized data set; Performing spatial calibration processing on the time-series synchronization data set to obtain a spatiotemporal calibration data set; De-noising and outlier processing are performed on the spatiotemporal calibration data set to obtain an effective synchronous monitoring data set; The synchronous monitoring data set is subjected to data fusion processing to obtain the fused monitoring data set.

4. The PCB life analysis method based on reliability analysis according to claim 1 is characterized in that: The performing feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix includes: Performing time-frequency domain analysis on the thermal field data in the multi-field coupling data set to obtain frequency-domain characteristic data; Performing spatial gradient calculation processing on the mechanical field data in the multi-field coupling data set to obtain thermal stress distribution characteristic data; Performing ion migration feature extraction processing on the electrochemical field data in the multi-field coupling data set to obtain electrochemical field feature data; Performing wavelet packet transform processing on the vibration data in the fused monitoring data set to obtain vibration mode characteristic data; Performing principal component analysis and dimensionality reduction processing on the frequency domain feature data, the thermal stress distribution feature data, the electrochemical field feature data, and the vibration mode feature data to obtain a low-dimensional data set of main failure features; The feature data in the low-dimensional data set is subjected to weighted synthesis processing to obtain the failure-sensitive feature matrix.

5. The PCB life analysis method based on reliability analysis according to claim 1, characterized in that: The modeling process of the failure sensitive feature matrix according to a preset dynamic neural network to obtain life prediction probability distribution data includes: Performing time series analysis on the failure sensitive feature matrix to obtain a time series feature set; Performing network structure design on the time series feature set according to the dynamic neural network to obtain a neural network model; Acquire historical fault data according to the circuit board to be detected, and synthesize the historical fault data and the failure sensitive feature matrix to obtain a training data set; Performing back propagation algorithm processing on the neural network model according to the training data set to obtain a target neural network; The failure sensitive feature matrix is ​​subjected to network inference calculation according to the target neural network to obtain the life prediction probability distribution data.

6. The PCB life analysis method based on reliability analysis according to claim 1, characterized in that: The performing of measured failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model comprises: Step S51, obtaining measured failure data according to the circuit board to be tested, and performing difference analysis processing according to the life prediction probability distribution data and the measured failure data to obtain error distribution; Step S52, performing back propagation processing on the parameters in the preset initial prediction model according to the error distribution to obtain a calibration loss; Step S53, performing gradient descent processing on the calibration loss to obtain calibration model parameters; Step S54, updating the initial prediction model according to the calibration model parameters to obtain a primary life prediction model; Step S55, verifying the primary life prediction model to obtain an error value; Steps S52 to S55 are repeatedly performed until the error value satisfies a preset error range to obtain the life prediction model.

7. The PCB life analysis method based on reliability analysis according to claim 1, characterized in that: The performing online integrated monitoring of the circuit board to be inspected according to the life prediction model to obtain real-time life assessment data comprises: Performing real-time acquisition on the circuit board to be inspected to obtain a real-time monitoring data set, and performing time-space synchronization processing and line filtering processing on the real-time monitoring data set to obtain real-time data; Performing integrated calculation processing on the real-time data according to the life prediction model to obtain a real-time life estimation value; The real-time life estimation value is subjected to an abnormality detection process to obtain real-time life assessment data.

8. A PCB life analysis device based on reliability analysis, characterized in that: The device comprises: A coupling data module, used to establish an original structure model according to the three-dimensional data set of the circuit board to be tested, and perform multi-physical field coupling modeling processing on the original structure model to obtain a multi-field coupling data set; A monitoring data module is used to collect target data of the circuit board to be inspected, and perform spatiotemporal alignment processing on the collected multi-source sensor data set to obtain a fused monitoring data set; A failure matrix module is used to perform feature extraction and dimensionality reduction processing on the multi-field coupling data set and the fusion monitoring data set to obtain a failure sensitive feature matrix; A prediction data module, used for modeling the failure sensitive feature matrix according to a preset dynamic neural network to obtain life prediction probability distribution data; A prediction model module, used for performing measured failure data calibration and prediction modeling processing on the life prediction probability distribution data to obtain a life prediction model; The life evaluation module is used to perform online integrated monitoring on the circuit board to be tested according to the life prediction model to obtain real-time life evaluation data.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the PCB life analysis method based on reliability analysis according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the PCB life analysis method based on reliability analysis according to any one of claims 1 to 7 are implemented.

Images