PCBA reliability prediction method and device, storage medium and program product

Abstract

The invention discloses a PCBA reliability prediction method and device, a storage medium and a program product, and relates to the technical field of printed circuit boards, according to the PCBA reliability prediction method and device, simulation sub-models of a temperature field, a mechanical field and an electric field are constructed through actually measured temperature data of a PCBA sample plate, coupling processing is carried out, and multi-physical field distribution of the PCBA in an actual working environment can be simulated more accurately. On the basis, simulation analysis can be carried out on multiple fault modes, and multi-physical field distribution data in different fault modes can be obtained. By analyzing the multi-physical field distribution data, the reliability of the PCBA can be predicted, and the prediction capability can help technicians to recognize possible failure risks in advance, so that preventive measures are taken, and the service life of a product is prolonged. And fault prediction and reliability analysis are carried out through the simulation model, so that the number of times and range of physical testing can be reduced, and the development cost and time are saved.

Description

Technical Field

[0001] The present application relates to the technical field of printed circuit boards, and in particular, to a method, device, storage medium and program product for predicting the reliability of a PCBA. Background Art

[0002] In the production and manufacturing process of a Printed Circuit Board Assembly (PCBA), temperature detection is particularly important.

[0003] In the related art, a single sensor, a thermometer and a static threshold are used to judge whether the temperature of the PCBA is abnormal. It is difficult to capture abnormal problems under complex working conditions, resulting in low accuracy of temperature detection. Summary of the Invention

[0004] The present application provides a method, device, storage medium and program product for predicting the reliability of a PCBA, so as to at least solve the problem of low accuracy of temperature detection in the related art.

[0005] The present application provides a method for predicting the reliability of a PCBA, including:

[0006] Collecting the temperature of a printed circuit board assembly sample to obtain measured temperature data;

[0007] According to the measured temperature data, respectively constructing a temperature field simulation sub-model, a mechanical field simulation sub-model and an electric field simulation sub-model corresponding to the sample;

[0008] Performing coupling processing on the temperature field simulation sub-model, the mechanical field simulation sub-model and the electric field simulation sub-model to obtain a multi-physical field simulation model;

[0009] Simulating various failure modes through the multi-physical field simulation model to obtain multi-physical field distribution data corresponding to each failure mode;

[0010] According to the multi-physical field distribution data corresponding to each failure mode, predicting the reliability of the sample to obtain reliability-related information.

[0011] The present application also provides a device for predicting the reliability of a PCBA, including:

[0012] An acquisition module, configured to collect the temperature of a printed circuit board assembly sample to obtain measured temperature data;

[0013] A construction module, configured to respectively construct a temperature field simulation sub-model, a mechanical field simulation sub-model and an electric field simulation sub-model corresponding to the sample according to the measured temperature data;

[0014] A coupling module for coupling a temperature field simulation sub-model, a mechanical field simulation sub-model, and an electric field simulation sub-model to obtain a multi-physical field simulation model;

[0015] A simulation module for simulating various fault modes through the multi-physical field simulation model to obtain multi-physical field distribution data corresponding to each fault mode;

[0016] A prediction module for predicting the reliability of a sample based on the multi-physical field distribution data corresponding to each fault mode to obtain reliability-related information.

[0017] This application also provides an electronic device, including: a memory for storing a computer program; a processor for implementing the steps of any of the above PCBA reliability prediction methods when executing the computer program.

[0018] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program implements the steps of any of the above PCBA reliability prediction methods when executed by a processor.

[0019] This application also provides a computer program product including a computer program, which implements the steps of any of the above PCBA reliability prediction methods when executed by a processor.

[0020] This application constructs simulation sub-models of the temperature field, mechanical field, and electric field through the measured temperature data of the PCBA sample and performs coupling processing, which can more accurately simulate the multi-physical field distribution of the PCBA in the actual working environment. This accurate simulation helps to identify potential fault modes. Based on this, simulation analysis can be carried out for various fault modes to obtain multi-physical field distribution data under different fault modes. By analyzing the multi-physical field distribution data, the reliability of the PCBA can be predicted. This prediction ability can help technicians identify possible failure risks in advance, thereby taking preventive measures and extending the service life of the product. And through the simulation model for fault prediction and reliability analysis, the number and scope of physical tests can be reduced, thus saving development costs and time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the embodiments of this application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a schematic diagram of a system architecture provided by an embodiment of this application;

[0023] Figure 2 It is a schematic flowchart of a PCBA reliability prediction method provided by an embodiment of the present application;

[0024] Figure 3 It is a schematic flowchart of another PCBA reliability prediction method provided by an embodiment of the present application;

[0025] Figure 4 It is a schematic flowchart of an anomaly detection process provided by an embodiment of the present application;

[0026] Figure 5 It is a schematic structural diagram of a PCBA reliability prediction device provided by an embodiment of the present application;

[0027] Figure 6 It is a schematic structural diagram of an electronic device provided by the present application. Detailed implementation manners

[0028] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0029] It should be noted that in the description of the present application, the terms "include", "comprise" or any other variant thereof are intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. In the description of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0030] First, some terms related to the present application are explained:

[0031] OPC-UA protocol: OPC Unified Architecture (UA) protocol. OPC originated from "OLE for Process Control", which means OLE for process control. OLE (Object Linking and Embedding) is a software technology used to share data and functions between different applications.

[0032] LSTM-GRU Hybrid Neural Network: A neural network constructed by combining two different types of recurrent neural network (RNN) structures, namely the Long Short-Term Memory network (LSTM) and the Gated Recurrent Unit (GRU).

[0033] PINN Network: Physics-Informed Neural Networks.

[0034] To enable those skilled in the art of the present technology to better understand the solution of this application, the following provides a further detailed description of this application in conjunction with the accompanying drawings and specific implementation manners.

[0035] In combination with the specific application environment architecture or specific hardware architecture on which the execution of the PCBA reliability prediction method depends, the specific application environment architecture or specific hardware architecture is described herein.

[0036] Figure 1 It is a schematic diagram of a system architecture provided by an embodiment of this application. Refer to Figure 1 , the system architecture includes a PCBA sample 10, a temperature sensor array 20, a data acquisition unit 30, and an electronic device 40.

[0037] The PCBA sample 10 is an additional sample, not the PCBA in the actual production environment. The PCBA sample 10 is used to simulate the actual production process of the PCBA. To detect the temperature of the PCBA sample 10, a temperature sensor array 20 is provided for the PCBA sample 10.

[0038] Optionally, the temperature sensor array 20 is set in the key thermally sensitive areas of the PCBA sample 10. The key thermally sensitive areas include at least one of the Ball Grid Array (BGA) area and the device surface. For example, the BGA area is the location where the Central Processing Unit (CPU) or Graphics Processing Unit (GPU) is located. The devices include at least one of power devices and connectors. For example, the power device can be at least one of a Metal Oxide Semiconductor (MOS) transistor, an inductor, or other possible devices.

[0039] The temperature sensor array 20 is a multimodal sensor array, that is, it includes multiple different types of temperature sensors. Optionally, the temperature sensor array 20 includes at least one of a thermocouple array and a distributed optical fiber temperature measurement unit.

[0040] Optionally, a thermocouple array is arranged in the BGA area. Among them, the thermocouple array includes multiple thermocouple monitoring points. The node density and sampling frequency of the thermocouple array can be set according to actual needs, and this embodiment does not limit this, for example, the node density is 4 monitoring points per square centimeter, and the sampling frequency is 200 Hertz (Hz). For example, the installation method of the thermocouple array can be: using a flexible film thermocouple with a thickness ≤ 0.1 millimeter (mm), pasted on the surface of the component through thermal conductive silicone, avoiding welding or drilling.

[0041] Optionally, a distributed optical fiber temperature measurement unit is surface-mounted on the device. Among them, the spatial resolution of the distributed optical fiber temperature measurement unit can be set according to actual needs, and this embodiment does not limit this, for example, the spatial resolution can be 0.5 mm. The spatial resolution refers to the minimum distance interval in the distributed optical fiber temperature measurement unit that can distinguish two adjacent temperature anomaly points (or temperature change points). For example, the installation method of the distributed optical fiber temperature measurement unit can be: embedding the optical fiber ribbon into the edge slot of the PCBA template 10 or fixing it with a high-temperature resistant tape, and monitoring the temperature using the principle of light scattering without damaging the structure of the PCBA template 10.

[0042] Optionally, the temperature sensor array 20 can also be set at specific positions around the PCBA template 10 to perform a global temperature scan on the PCBA template 10. Correspondingly, the temperature sensor array 20 also includes an infrared thermal imaging unit. Exemplarily, the infrared thermal imaging unit can be set on a fixed bracket perpendicular to the upper part of the PCBA template 10, and a suitable distance needs to be maintained from the PCBA template 10 to ensure that the lens field of view covers the entire template surface. Among them, the wavelength range and frame rate of the infrared thermal imaging unit can be set according to actual needs, and this embodiment does not limit this, for example, the wavelength range can be 8 - 14 micrometers (μm), and the frame rate is not less than 30 frames per second (fps) to ensure real-time performance.

[0043] This application realizes the real-time acquisition of the temperature field of the PCBA through a multi-modal temperature sensor array 20 and a non-destructive installation method, and has a three-dimensional monitoring ability: the infrared thermal imaging unit (area coverage), the thermocouple array (single-point accuracy), and the distributed optical fiber (continuous path) form a "plane-point-line" collaborative monitoring network. Optionally, the temperature sensor array 20 adopts a star topology and is connected to the data acquisition unit 30 through a Controller Area Network (CAN) bus to ensure the real-time nature of data transmission. Among them, the star topology includes a central node, and other nodes are all connected to this central node. For example, each thermocouple monitoring point is integrated through the array and accessed as a node to the central node. The entire optical fiber of the distributed optical fiber temperature measurement unit is used as an overall node and accessed to the central node, and the infrared thermal imaging unit is accessed to the central node as a node.

[0044] Optionally, the data acquisition unit 30 uses the Precision Time Protocol (PTP) to achieve time synchronization of the temperature sensor array 20. For example, the clock deviation is controlled within ±1 microsecond (μs) to achieve microsecond-level time synchronization.

[0045] The data acquisition unit 30 is used to denoise the temperature signals output by the temperature sensor array 20 to obtain the denoised temperature signals. The data acquisition unit 30 is also used to send the denoised temperature signals to the electronic device 40.

[0046] The data acquisition unit 30 integrates an adaptive filtering algorithm. Optionally, the adaptive filtering algorithm is a wavelet transform and Kalman filter fusion algorithm. Correspondingly, the data acquisition unit 30 can be used to denoise the high-frequency noise (such as electromagnetic interference) in the temperature signals collected by the thermocouple array through wavelet transform, and can also be used to perform dynamic error compensation on the low-frequency drift (such as line impedance error) in the temperature signals collected by the distributed optical fiber temperature measurement unit through Kalman filter. It can also be used to comprehensively process the noise (such as pixel offset) in the temperature signals collected by the infrared thermal imaging unit through the wavelet transform and Kalman filter fusion algorithm. Optionally, the db4 wavelet basis is selected for wavelet transform.

[0047] Optionally, the data acquisition unit 30 can be any hardware module with data processing and transmission functions, and this application does not limit this. For example, the data acquisition unit 30 can be a CPU, or it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0048] This application uses a multimodal sensor array to achieve an organic combination of macroscopic global scanning and microscopic local monitoring. Through a collaborative detection mechanism, the temperature detection accuracy is improved, ensuring the accurate capture of small thermal anomalies. Paired with a wavelet transform-Kalman filter fusion algorithm, it effectively suppresses environmental electromagnetic interference and significantly enhances the detection reliability of weak thermal signals.

[0049] The data acquisition unit 30 is also in communication with the electronic device 40 . The data acquisition unit 30 is also used to convert the processed temperature signal into temperature data and send the temperature data to the electronic device 40 .

[0050] The electronic device 40 is used to execute the PCBA reliability prediction method provided herein, modeling the PCBA template 10 to obtain a multi-physics simulation model. The multi-physics simulation model can be considered a digital twin of the PCBA template 10. It can be seen that the data acquisition unit 30 establishes a channel that enables real-time data transmission between the PCBA template 10 and its digital twin.

[0051] Alternatively, the electronic device 40 is intended to be various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers.

[0052] The PCBA reliability prediction method provided in the present application is performed by a PCBA reliability prediction device, which is disposed in an electronic device, such as the electronic device 40 described above.

[0053] Figure 2 A flow chart of a PCBA reliability prediction method provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the PCBA reliability prediction method includes:

[0054] S201: collecting temperature of the PCBA sample to obtain measured temperature data.

[0055] The PCBA sample is an additional sample and is not the PCBA in the actual production environment.

[0056] This embodiment simulates the actual production process of PCBA through a PCBA sample, and detects the temperature of the PCBA sample in real time to obtain measured temperature data.

[0057] The measured temperature data includes the measured temperatures of multiple monitoring points on the PCBA sample.

[0058] S202: Based on the measured temperature data, a temperature field simulation sub-model, a mechanical field simulation sub-model, and an electric field simulation sub-model corresponding to the PCBA sample are respectively constructed.

[0059] The temperature field simulation sub-model is used to simulate the temperature change of the PCBA sample board.

[0060] The mechanical field simulation sub-model is used to simulate the stress change of the PCBA sample board.

[0061] The electric field simulation sub-model is used to simulate the change of current density of the PCBA sample board.

[0062] The temperature field simulation sub-model, mechanical field simulation sub-model and electric field simulation sub-model corresponding to the PCBA sample board can be constructed using simulation software. The simulation software can be selected according to actual needs, and this embodiment does not limit this.

[0063] S203: Couple the temperature field simulation sub-model, mechanical field simulation sub-model and electric field simulation sub-model to obtain a multi-physical field simulation model.

[0064] Among them, the multi-physical field simulation model can be regarded as the digital twin model of the PCBA sample board. The temperature field of the PCBA will change during the actual production process, and the mechanical field and electric field will change due to the change of the temperature field. Therefore, the temperature field simulation sub-model, mechanical field simulation sub-model and electric field simulation sub-model of the PCBA sample board can be coupled to realize the coupling of multi-physical fields, so as to comprehensively analyze the PCBA sample board according to the multi-physical field simulation model.

[0065] S204: Simulate various fault modes through the multi-physical field simulation model to obtain the multi-physical field distribution data corresponding to each fault mode.

[0066] The fault mode is used to describe the possible faults of the PCBA in the actual production scenario. The specific content of the fault mode and the number of various fault modes can be set according to actual needs, and this application does not limit this.

[0067] The multi-physical field distribution data includes temperature field distribution data, stress distribution data and current density distribution data.

[0068] In this embodiment, a multi-physical field simulation model is constructed through S203, and then fault simulation is applied based on the constructed model through S204.

[0069] S205: Predict the reliability of the PCBA sample board according to the multi-physical field distribution data corresponding to each fault mode to obtain reliability-related information.

[0070] Among them, the reliability-related information is used to describe the reliability of the PCBA sample board.

[0071] This application constructs simulation submodels of the temperature field, mechanical field, and electric field through the measured temperature data of the PCBA prototype, and performs coupling processing, which can more accurately simulate the multi-physical field distribution of the PCBA in the actual working environment. This accurate simulation helps to identify potential failure modes. Based on this, simulation analysis can be carried out for various failure modes to obtain the multi-physical field distribution data under different failure modes. By analyzing the multi-physical field distribution data, the reliability of the PCBA can be predicted. This prediction ability can help technicians identify possible failure risks in advance, so as to take preventive measures and extend the service life of the product. And through the simulation model for failure simulation and reliability prediction, the number and scope of physical tests can be reduced, thus saving development costs and time.

[0072] Figure 3 FIG. is a schematic flowchart of another PCBA reliability prediction method provided by an embodiment of this application. As Figure 3 shown, the PCBA reliability prediction method includes:

[0073] S301: Collect the temperature of the PCBA prototype through a multi-modal sensor array to obtain the original temperature data.

[0074] Among them, the multi-modal sensor array includes a thermocouple array, a distributed optical fiber temperature measurement unit, and an infrared thermal imaging unit. The specific implementation process of collecting the temperature of the PCBA prototype through the multi-modal sensor array can refer to the relevant description of the temperature sensor array 20 in the above system architecture, which will not be elaborated here.

[0075] S302: Denoise the original temperature data to obtain the measured temperature data.

[0076] In some embodiments, the electronic device can be directly communicatively connected to the multi-modal sensor array, so as to obtain the original temperature data and denoise the original temperature data to obtain the measured temperature data. The specific implementation process of the denoising process is the same as that of the data acquisition unit 30 in the above system architecture, which will not be elaborated here.

[0077] S301-S302 is an implementation method for collecting the temperature of the PCBA prototype to obtain the measured temperature data. The original temperature data collected through the multi-modal sensor array is more comprehensive, and by denoising the original temperature data, more accurate measured temperature data can be obtained.

[0078] In some other embodiments, the electronic device may also be communicatively connected to a data acquisition unit. The data acquisition unit obtains the original temperature data collected by the multi-modal sensor array, performs denoising processing on the original temperature data to obtain the measured temperature data, and sends the measured temperature data to the electronic device. Accordingly, steps S301 - S302 may be replaced with: obtaining the measured temperature data obtained by performing temperature acquisition and denoising processing on the PCBA sample board sent by the data acquisition unit. For the specific implementation process, refer to the relevant description of the data acquisition unit 30 in the above system architecture, which will not be elaborated here.

[0079] S303: Construct a temperature field simulation sub-model corresponding to the PCBA sample board according to the measured temperature data.

[0080] In an alternative implementation, the specific implementation process of S303 includes S3031 - S3032:

[0081] S3031: For the PCBA sample board, construct a heat conduction sub-model, a convective heat dissipation sub-model, and a radiative heat transfer sub-model respectively;

[0082] S3032: Perform coupling processing on the heat conduction sub-model, the convective heat dissipation sub-model, and the radiative heat transfer sub-model to obtain the temperature field simulation sub-model.

[0083] Optionally, S3031 is specifically implemented as:

[0084] Using simulation software, create an original structure model of the PCBA. The original structure model includes multiple components; and set initial material parameters for each component. The original structure model is a three-dimensional model, and the material parameters of the components are used to represent material properties. For example, the multiple components include components such as a PCB substrate, a BGA, capacitors, resistors, connectors, etc. For example, the material parameters include the material thermal conductivity (also known as the heat conduction rate), density, specific heat capacity, convective heat transfer coefficient, emissivity, and other possible parameters.

[0085] Construct a heat conduction sub-model for the original structure model. Among them, the heat conduction sub-model is used to simulate the heat conduction between the PCBA sample board and the components. Specifically, the heat conduction sub-model is represented in the form of a heat conduction equation, and the heat conduction equation can be established based on Fourier's law. The material parameters required for modeling include the material thermal conductivity, density, and specific heat capacity. For example, the thermal conductivity includes: the thermal conductivity of the copper foil is set to 401 W / (m·K), and the thermal conductivity of the FR-4 substrate is set to 0.3 W / (m·K).

[0086] Exemplarily, the heat conduction equation is , where ρ is the material density, c is the specific heat capacity, T is the temperature field distribution, t is the time, k is the thermal conductivity, and Q is the internal heat source.

[0087] A convective heat dissipation sub-model is constructed for the original structure model. Among them, the convective heat dissipation sub-model is used to simulate the heat exchange between air and the surface of the PCBA sample board, that is, forced convection. Specifically, the k-ε turbulence model is used for modeling. The material parameters required for modeling include the convective heat transfer coefficient. In addition, the fan wind speed parameters required for modeling are dynamically updated through measured data (such as the real-time wind speed during equipment operation).

[0088] A radiative heat transfer sub-model is constructed for the original structure model. Among them, the radiative heat transfer sub-model is used to simulate the radiative heat exchange between components and the shell. Specifically, it is modeled based on the Stefan-Boltzmann law. The parameters required for modeling include the emissivity, and the emissivity is set according to the surface treatment process of the component (such as plating, oxide layer), and can be obtained through material manuals or experimental measurements.

[0089] When the PCBA sample board is working, the internal components transfer heat to the shell through heat conduction (heat conduction sub-model), the shell dissipates heat through air convection (convective heat dissipation sub-model), and at the same time radiates heat to the surrounding environment (radiative heat transfer sub-model). Separately modeling will cause errors, and the collaborative description of the three types of heat transfer processes can be achieved through coupling.

[0090] After coupling the three types of heat transfer processes, the formed heat balance equation is strongly non-linear. That is to say, the temperature field simulation sub-model can be expressed in the form of a non-linear heat balance equation. The non-linear heat balance equation f(T)=0 is used to describe the physical relationship between the temperature field distribution T and the material parameters.

[0091] Optionally, in order to improve the simulation accuracy of the temperature field simulation sub-model, the temperature field simulation sub-model can be reversely optimized. Correspondingly, after S3032, the specific implementation process of S303 also includes S3033 - S3037:

[0092] S3033: Use the Newton-Raphson algorithm to solve the non-linear heat balance equation to obtain the temperature field distribution data under the current material parameters.

[0093] Among them, the initial material parameters of the non-linear heat balance equation are known, and the equation is a non-linear equation about the temperature field distribution T, which can be directly solved by the Newton-Raphson algorithm.

[0094] The temperature field distribution data includes the simulation temperatures of each node in the simulation grid. The simulation grid is obtained by discretizing the three-dimensional model of the PCBA into a finite number of tiny units (such as hexahedrons, tetrahedrons).

[0095] S3034: Determine the residual matrix according to the measured temperature data of the PCBA sample board and the temperature field distribution data.

[0096] Among them, the measured temperature data is collected by a multimodal sensor array. For the convenience of calculation, the temperature data collected by a thermocouple array and a distributed optical fiber temperature measurement unit can be used.

[0097] The measured temperature data includes the measured temperatures of multiple monitoring points, and the temperature field distribution data includes the simulated temperatures of each node in the simulation grid. Therefore, the multiple monitoring points and the multiple nodes in the simulation grid can be spatially aligned first, and then the residuals can be calculated. The alignment method can refer to the alignment method in the related technology, which will not be elaborated in this embodiment.

[0098] Among them, the residual matrix includes the residuals of each aligned point.

[0099] S3035: Determine the root mean square error (RMSE) according to the residual matrix.

[0100] S3036: If the root mean square error of the residuals is greater than the preset error value, perform reverse optimization on the material parameters in the temperature field simulation sub-model to obtain an optimized temperature field simulation sub-model.

[0101] Among them, the preset error value can be set according to actual needs, which is not limited in this embodiment. For example, if the temperature unit is °C, the preset error value can be 2, 2.5, 3 or other values.

[0102] If the root mean square error of the residuals is less than or equal to the preset error value, it indicates that the error between the measured temperature and the simulated temperature is small, and there is no need to perform reverse optimization on the temperature field simulation sub-model.

[0103] If the root mean square error of the residuals is greater than the preset error value, it indicates that the error between the measured temperature and the simulated temperature is large, and the accuracy of the temperature field simulation sub-model is not high. Therefore, the temperature field simulation sub-model can be reversely optimized, and by updating the material parameters, the model output can be made closer to the measured temperature.

[0104] Optionally, the way to update the material parameters can be implemented by an optimization algorithm, such as gradient descent, Gauss-Newton method, which is not limited in this embodiment.

[0105] S3037: Repeat the above operations until the root mean square error of the residuals is less than or equal to the preset error value.

[0106] The core of the reverse optimization is to minimize the error between the measured temperature and the simulated temperature. By repeating S3033 - S3036 until the root mean square error of the residuals converges, the material parameters at this time are the optimal solution.

[0107] For example, the iteration tolerance of the Newton-Raphson algorithm can be set to 1×10 -6 Or other suitable values.

[0108] In this embodiment, each update of the material parameters requires re-solving the temperature field distribution data. Due to its quadratic convergence characteristic, the Newton-Raphson algorithm can quickly obtain a high-precision temperature solution, providing a reliable gradient calculation basis for parameter optimization.

[0109] It should be noted that the temperature field simulation sub-model can be periodically reversely optimized to ensure the model accuracy. For example, the period can be 20 minutes, 30 minutes, 40 minutes, etc., and this embodiment does not make any limitation in this regard.

[0110] By comparing the measured temperature data with the simulated temperature data, the material parameters are updated to form a dynamic correction mechanism of "simulation-measurement-optimization", ensuring that the model always closely approximates the actual working conditions. Thus, for the PCBA in the actual production environment, if the process changes, the actual temperature field distribution also changes, and the temperature field simulation sub-model can be automatically updated with the dynamic change of the process, and the long-term simulation accuracy fluctuation remains within a small range.

[0111] In an optional implementation manner, the simulation grid corresponding to the PCBA sample can also be dynamically grid-subdivided. Correspondingly, the method provided by this application further includes:

[0112] Determine the temperature gradient between adjacent nodes in the simulation grid corresponding to the PCBA sample according to the temperature field distribution data; if the temperature gradient is greater than the preset gradient value, grid encryption is performed on the grid area where the adjacent nodes are located to obtain a new simulation grid.

[0113] Among them, the temperature field distribution data is obtained by solving S3033. The temperature gradient G is the product of the temperature difference between adjacent nodes and the distance between nodes, and the calculation formula is: G = ΔxΔT, where ΔT is the temperature difference between adjacent nodes and Δx is the distance between nodes.

[0114] If the temperature gradient is greater than the preset gradient value, it indicates that the temperature difference between adjacent nodes is large. Therefore, grid encryption can be performed to subdivide the grid. Among them, the preset gradient value can be 5 °C / mm or other suitable values, and this embodiment does not make any limitation in this regard.

[0115] For example, the initial grid size is set to 1mm × 1mm × 0.2mm. When the temperature gradient > 5 °C / mm, the grid is automatically encrypted. For example, the original grid side length is refined from 1mm to 0.2mm, and the new grid size is 0.2mm × 0.2mm × 0.05mm.

[0116] In this embodiment, high-temperature gradient regions (such as solder joints and the edges of power devices) require denser grids to capture subtle temperature changes and reduce the phenomenon of calculation errors caused by sparse grids, thereby improving the local calculation accuracy through dynamic grid subdivision.

[0117] S304: Construct a mechanical field simulation sub-model corresponding to the PCBA prototype board.

[0118] Construct a mechanical field simulation sub-model for the original structure model. Temperature changes cause changes in thermal strain, which in turn cause stress changes. Correspondingly, the mechanical field simulation sub-model is expressed in the form of mechanical equations. The mechanical equation is , where is the stress divergence, is the body force, i is the free index, and j is the dummy index. Through the mechanical equation, simulate the stress distribution of the PCBA prototype board caused by temperature changes.

[0119] S305: Construct an electric field simulation sub-model corresponding to the PCBA prototype board.

[0120] Construct an electric field simulation sub-model for the original structure model. Temperature changes cause changes in conductivity, which in turn cause changes in current density. Correspondingly, the electric field simulation sub-model is expressed in the form of electric field equations. The electric field equation is , where, is the current density, and , is the temperature function of conductivity, is the electric field, , is the electric potential. Through the electric field equation, simulate the change in current density of the PCBA prototype board caused by temperature changes.

[0121] S306: Perform coupling processing on the temperature field simulation sub-model, the mechanical field simulation sub-model, and the electric field simulation sub-model to obtain a multi-physical field simulation model.

[0122] Specific coupling method: First, perform thermal simulation through the temperature field simulation sub-model to obtain temperature field distribution data; map the temperature field distribution data to the mechanical field simulation sub-model and the electric field simulation sub-model, that is, use the temperature field distribution data as the "temperature field input" for thermal-mechanical coupling and thermal-electric coupling respectively; perform simulations on the mechanical field simulation sub-model and the electric field simulation sub-model respectively to achieve multi-physical field solution under the basic conditions.

[0123] Most simulation software supports the "field coupling" function and can automatically complete temperature mapping.

[0124] S307: Use the multi-physical field simulation model to simulate various fault modes respectively to obtain multi-physical field distribution data corresponding to each fault mode.

[0125] In an optional implementation, the specific implementation process of S307 includes: for each fault mode, perform the following operations respectively, and the following operations include:

[0126] Inject the fault mode into the temperature field simulation sub - model in a parameterized form;

[0127] Perform numerical solution on the temperature field simulation sub - model to obtain temperature field distribution data;

[0128] Use the temperature field distribution data as input to drive the numerical solution of the mechanical field simulation sub - model to obtain stress distribution data;

[0129] Use the temperature field distribution data as input to drive the numerical solution of the electric field simulation sub - model to obtain current density distribution data.

[0130] Among them, the current density distribution data includes the current density of each node in the simulation grid. The stress distribution data includes the stress of each node in the simulation grid.

[0131] Optionally, preset multiple fault modes, and convert each fault mode into boundary conditions of the temperature field or changes in material parameters to obtain fault parameters. When simulating the fault mode, inject the corresponding fault parameters into the temperature field simulation sub - model and solve the temperature field distribution data.

[0132] Among them, adjustable fault parameters are defined for each fault mode. For example, when the heat sink falls off, the convective coefficient decreases by 50%, which is achieved by modifying the boundary conditions; another example is that component parameter drift is achieved by modifying the temperature coefficient of resistors / capacitors.

[0133] Optionally, the fault mode can be randomly selected from multiple fault modes through a preset probability distribution (such as uniform distribution or Poisson distribution), or injected according to the test scenario sequence (such as simulating thermal faults first and then mechanical faults) to verify the robustness of the PCBA prototype under different fault combinations.

[0134] Exemplarily, multiple fault modes at least include the following items:

[0135] 1. Heat sink falls off: resulting in a decrease in the convective coefficient.

[0136] 2. Component overheats and burns out: Simulate that some key components (such as power devices, CPUs, etc.) have excessive temperature due to excessive power consumption, poor heat dissipation, etc., and finally burn out, thereby affecting the normal operation of the entire PCBA.

[0137] 3. Component parameter drift: For example, the parameters of components such as resistors and capacitors drift due to temperature changes, resulting in changes in circuit performance, such as signal anomalies and function failures.

[0138] 4. Solder joint is poorly soldered: During the welding process, due to improper welding process or inaccurate temperature control, there is a phenomenon of poor soldering at the solder joint. When the temperature changes, the resistance of the poorly soldered joint will change, which may cause the circuit to be intermittently connected, affecting the stability of the PCBA.

[0139] 5. Solder joint cracking: Under the action of long-term temperature cycling or mechanical stress, the solder joints may crack. This will lead to a deterioration of the electrical connection between the components and the PCB substrate, or even a complete disconnection, resulting in failures.

[0140] 6. Pin short circuit: Short circuits occur between component pins due to improper soldering, foreign object contamination, etc., causing abnormal current flow, resulting in local overheating, and may also affect the normal operation of related circuits, etc.

[0141] In this embodiment, through the simulation of failure modes, the re - solution of multi - physical fields under failure conditions is realized.

[0142] S308: According to the multi - physical field distribution data corresponding to each failure mode, perform reliability prediction on the PCBA sample board to obtain reliability - related information.

[0143] Among them, the reliability - related information is used to describe the reliability of the PCBA sample board. The reliability - related information includes at least one of signal attenuation information, regional fracture information, and electromigration information. The signal attenuation information includes the possibility of signal attenuation in the first area on the PCBA sample board. The regional fracture information is used to indicate the possibility of fracture in the second area on the PCBA sample board. The electromigration information is used to indicate the possibility of electromigration in the third area on the PCBA sample board.

[0144] In an optional implementation manner, the specific implementation process of S308 includes: for each failure mode, perform the following operations, and the following operations include S3081 - S3083:

[0145] S3081: According to the temperature field distribution data after the failure mode injection and the temperature field distribution data before the failure mode injection, determine the wire resistance change data.

[0146] The wire resistance change data includes the resistance change amount of each node in the simulation grid.

[0147] Temperature change will directly cause a change in resistance. For example, when the temperature increases, the resistance increases. Correspondingly, for each node in the simulation grid, determine the resistance change amount of the node according to the temperature change amount of the node. Among them, the temperature change amount of the node is the difference between the simulation temperature before the failure mode injection and the simulation temperature after the failure mode injection.

[0148] The resistance change amount of the node is calculated by the formula where ΔR represents the resistance change amount, is the initial resistance, that is, the resistance value when the temperature is the reference temperature (usually 20 °C or 0 °C), ΔT is the temperature change amount, α is the temperature coefficient, and α = 0.0039 / °C.

[0149] S3082: Determine signal attenuation information based on the wire resistance change data.

[0150] The attenuation of a signal during transmission in a wire is positively correlated with the resistance. Therefore, an increase in resistance will cause signal attenuation. Optionally, S3082 is specifically implemented as follows: If the resistance change amount of any node in the simulation grid indicates an increase in the resistance of that node, then the area where that node is located is taken as the first area, and signal attenuation information is generated to indicate the possible existence of signal attenuation in the first area.

[0151] S3083: Determine area fracture information based on the stress distribution data after fault mode injection and the stress distribution data before fault mode injection.

[0152] An increase in temperature forms a hot spot area, which affects the mechanical properties of the material, such as an increase in stress, exacerbating the fracture risk. Optionally, S3083 is specifically implemented as follows: If the stress change amount of any node in the simulation grid indicates an increase in the stress of that node, then the area where that node is located is taken as the second area, and area fracture information is generated to indicate the possible existence of fracture in the second area. Among them, the stress change amount of the node is the difference between the stress data before fault mode injection and the stress data after fault mode injection.

[0153] S3084: Determine electromigration information based on the current density distribution data after fault mode injection and the current density distribution data before fault mode injection.

[0154] Temperature changes will cause changes in conductivity. To satisfy the electric field equation, the current density will be adjusted adaptively. For example, if the conductivity of a certain area decreases due to an increase in temperature, then the current density in that area will decrease. To maintain current continuity, the current density in the adjacent area needs to increase accordingly. Areas with concentrated current density (such as solder joint corners and wire cross-section changes) are high-risk areas for electromigration and belong to high-risk solder joints, which may cause copper foil melting or substrate carbonization.

[0155] Optionally, S3084 is specifically implemented as follows: If the current density change amount of any node in the simulation grid indicates an increase in the current density of that node, then the area where that node is located is taken as the third area, and electromigration information is generated to indicate the possible existence of electromigration in the third area. Among them, the current density change amount of the node is the difference between the current density data before fault mode injection and the current density data after fault mode injection.

[0156] Through thermo-electro-mechanical coupling analysis, the impact of faults on signal integrity (such as signal attenuation caused by resistance changes), mechanical reliability (such as solder joint cracking), and electromigration phenomena is predicted, realizing the prediction of reliability and providing an improvement basis for the subsequent production and manufacturing of PCBA.

[0157] In Figure 2 orFigure 3 Based on the illustrated embodiments, anomaly detection can also be performed on the temperature field simulation sub-model. Correspondingly, Figure 4 FIG. is a schematic flow chart of an anomaly detection process provided by an embodiment of the present application. As Figure 4 shown, the anomaly detection process includes:

[0158] S401: Obtain the temperature field distribution data output by the temperature field simulation sub-model.

[0159] Optionally, the implementation manner of this step is the same as that of S3033 and will not be elaborated here.

[0160] S402: Determine whether there is an anomaly in the temperature field distribution data through a preset detection algorithm.

[0161] Among them, the preset detection algorithm can be any one of a thermal behavior benchmark model, an isolation forest-autoencoder hybrid algorithm, or a temperature evolution prediction model.

[0162] S403: If there is an anomaly, generate an anomaly detection result.

[0163] The anomaly detection result is used to indicate that there is an anomaly in the temperature field distribution data. Specifically, the anomaly detection result includes the position information of the node and the temperature anomaly value. The node is the node in the simulation grid where the temperature is abnormal. The position information is the node coordinates in the simulation grid. The temperature anomaly value is the difference between the simulation temperature of the node and the reference temperature, and the reference temperature is determined by the preset detection algorithm.

[0164] S404: Match the anomaly detection result with the fault types recorded in the database to obtain the target fault type.

[0165] A large number of PCBA-related fault types are stored in the database, and each fault type may cause temperature anomalies in the PCBA.

[0166] Optionally, the database is connected to a rule engine, and multiple temperature anomaly association rules are set in the rule engine. Then, through the rule engine, the anomaly detection result can be matched with the fault types recorded in the database according to the temperature anomaly association rules to achieve fault traceability.

[0167] Specifically, parse the anomaly feature parameters from the anomaly detection result. The anomaly feature parameters include the position information and the temperature anomaly value. Screen the candidate rules from multiple temperature anomaly association rules according to the position information, and then determine whether the temperature anomaly value meets the trigger condition of the candidate rules. If it meets, use the fault type corresponding to the candidate rule as the target fault type.

[0168] For example, the trigger condition for a temperature anomaly correlation rule is that the node is located in the solder joint area and the temperature anomaly value is greater than 8°C, and the matching fault type is poor solder joint. Another example, the trigger condition for another temperature anomaly correlation rule is that the node is located in the core area of the device and the temperature anomaly value is greater than 15°C, and the matching fault type is poor heat dissipation of the device. Another example, the trigger condition for yet another temperature anomaly correlation rule is multi-point temperature anomaly (≥3 nodes) and the average value of the temperature anomaly values is greater than 6°C, and the matching fault type is aging of power supply components.

[0169] Optionally, the database also stores data related to product manufacturability design, including product design specifications, manufacturing process parameters, material characteristics, cost data, and previous design cases, etc. For example, the database can be at least one of a research and development design library, a manufacturing process problem library, or a Design for Manufacturing (DFM) database.

[0170] S405: Feed the anomaly detection result and the target fault type back to the manufacturing execution system.

[0171] Optionally, feed the anomaly detection result and the target fault type back to the manufacturing execution system through a data communication protocol in the field of industrial automation (such as the OPC-UA protocol).

[0172] By using the anomaly detection result and the matched target fault type as the feedback information of the manufacturing execution system, so that technicians can optimize the manufacturing process of the PCBA, realizing the closed-loop feedback optimization of the PCBA production and manufacturing link.

[0173] In some embodiments, the temperature field simulation sub-model can be periodically subjected to anomaly detection, so as to feed back to the manufacturing execution system in a timely manner. The period of anomaly detection can be set according to actual needs, and this embodiment does not limit it.

[0174] S403 - S405 provide a processing method for the situation where there are anomalies. If it is determined through a preset detection algorithm that the temperature field distribution data has no anomalies, no processing is performed, and wait for the next anomaly detection.

[0175] In some embodiments, the PCBA design version can be bound to the anomaly detection result based on the blockchain traceability mechanism, which is convenient for subsequent tracing of quality problems.

[0176] In Figure 4 Based on the illustrated embodiment, S402 includes at least one of the following implementation manners:

[0177] The first implementation method: Use the trained thermal behavior benchmark model to determine whether there is an abnormality in the temperature field distribution data. Among them, the thermal behavior benchmark model is used to determine the predicted temperature range according to the current working condition data of the PCBA, so as to predict the possible temperature range that the PCBA may be in currently.

[0178] Optionally, the specific implementation process of using the trained thermal behavior benchmark model to determine whether there is an abnormality in the temperature distribution data includes:

[0179] Obtain the current working condition data of the PCBA sample; input the working condition data as input data into the thermal behavior benchmark model, and output the predicted temperature range through the thermal behavior benchmark model; determine whether there is an abnormality in the temperature field distribution data according to the predicted temperature range.

[0180] Among them, the working condition data includes at least one of electrical parameters, physical environment parameters, and operating state parameters. The electrical parameters include at least one of the voltage, current, power consumption, and impedance of each component. The physical environment parameters include at least one of the temperature field distribution (temperature of each node), humidity, and vibration frequency / amplitude. The operating state parameters include at least one of signal transmission delay, clock frequency stability, and data throughput.

[0181] The predicted temperature range includes the temperature range of each node in the simulation grid, and this temperature range contains the interval in which the temperature of the node may fluctuate under the current working condition. Correspondingly, for each node, if the simulated temperature of the node in the temperature field distribution data is within the temperature range of the node, it means that the temperature of the node is normal. On the contrary, if the simulated temperature of the node in the temperature field distribution data is outside the temperature range of the node, it means that the temperature of the node is abnormal. When determining the temperature anomaly value of the node, the temperature median in the temperature range of the node can be regarded as the reference temperature.

[0182] Optionally, the training process of the thermal behavior benchmark model includes:

[0183] Obtain sample data and label data; the sample data includes the historical working condition data of the PCBA, and the label data includes the measured temperature range under the working condition represented by the sample working condition data;

[0184] Using the label data as supervised data, iteratively train the initial thermal behavior benchmark model according to the sample data to obtain the trained thermal behavior benchmark model.

[0185] To improve the training accuracy, a large amount of sample data and the corresponding label data of the sample data can be obtained. For example, the sample data can be the normal working condition data of the PCBA in the actual production environment within 1000 hours (such as collecting a set of working condition data every 1 minute or 1 hour), and the label data is the measured temperature range collected by the sensor under the corresponding working condition.

[0186] By training a thermal behavior benchmark model, it is possible to predict an accurate temperature range based on current operating condition data, and then, using this temperature range as a threshold range, determine whether there are abnormalities in the temperature field distribution data, with relatively high accuracy.

[0187] Optionally, the model architecture of the thermal behavior benchmark model can be an LSTM-GRU hybrid neural network. Table 1 shows the network architecture of an LSTM-GRU hybrid neural network.

[0188] Table 1

[0189]

[0190] It should be noted that Table 1 is only an example. In actual applications, the model architecture of the thermal behavior benchmark model can also be in other forms, and this application does not make any limitations in this regard.

[0191] The second implementation method: Use an Isolation Forest-Autoencoder hybrid algorithm to determine whether there are abnormalities in the temperature field distribution data.

[0192] Isolation Forest quickly identifies outliers based on a tree structure and is suitable for sparse anomalies in high-dimensional data. Autoencoders capture data distribution features through reconstruction error and are more sensitive to local and cluster anomalies. The Isolation Forest-Autoencoder hybrid algorithm combines the fast screening ability of Isolation Forest and the reconstruction accuracy advantage of Autoencoders.

[0193] Specifically, use Isolation Forest to perform preliminary anomaly detection on the temperature field distribution data to quickly identify samples with obvious anomalies; extract the samples considered normal by Isolation Forest for training the autoencoder, and construct a multi-layer perceptron autoencoder to learn the feature representation of normal temperature distribution; calculate the reconstruction error of each sample as the anomaly score of the autoencoder; perform weighted fusion on the scores of Isolation Forest and the autoencoder to obtain the final anomaly score; only the samples considered abnormal by both models are marked as abnormal. Among them, a sample refers to the simulated temperature of a node. Since the autoencoder will output the reconstructed temperature value for each node, which reflects the fitting result of the autoencoder to "normal temperature", when determining the temperature anomaly value of a node, the reconstructed temperature value of this node can be regarded as the reference temperature.

[0194] Among them, the autoencoder takes minimizing the reconstruction error as the training objective. In the temperature field distribution data, only the nodes that simultaneously meet the Isolation Forest threshold (such as score > 0.7) and the autoencoder threshold (such as reconstruction error > 3σ) are marked as abnormal, and reconstruction error > 3σ is based on the 3σ principle of the normal distribution.

[0195] Among them, the contamination rate of the isolation forest algorithm can be set to 0.1% or other possible values, which is not limited in this application. Setting the contamination rate to 0.1% means that during the detection process, it is expected that only 0.1% of the data points in the data will be determined as outliers.

[0196] The isolation forest-autoencoder hybrid algorithm adopted in this implementation method takes into account both local outlier recognition and global pattern learning.

[0197] The third implementation method: Use the trained temperature evolution prediction model to determine whether there are abnormalities in the temperature field distribution data. Among them, the temperature evolution prediction model is used to roll predict the temperature field change information of the PCBA in the next period of time.

[0198] Optionally, the specific implementation process of using the trained temperature evolution prediction model to determine whether there are abnormalities in the temperature field distribution data includes:

[0199] Take the temperature field distribution data and the corresponding load parameters as input data, input them into the trained temperature evolution prediction model, and use the temperature evolution prediction model to roll predict the temperature field change information at each time point within the preset duration in the future; determine whether there are abnormalities according to the temperature field change information.

[0200] This implementation method takes the temperature field distribution data currently simulated by the temperature field simulation sub-model and the current load parameters of the PCBA template as the input data of the temperature evolution prediction model, and outputs the temperature field change information at each time point within the preset duration in the future.

[0201] Among them, the load parameters refer to various external input conditions or working state parameters that affect the temperature field distribution of the PCBA template. These parameters will directly or indirectly cause changes in the temperatures of each node of the PCBA and are important input variables for the temperature evolution prediction model to perform rolling prediction. Exemplarily, the load parameters include at least one of electrical load parameters, environmental load parameters, and structural load parameters. The electrical load parameters include at least one of the working current, voltage, power consumption of each component, and the voltage of the power input. The environmental load parameters include at least one of the environmental temperature, humidity, heat dissipation conditions (such as air flow rate, flow rate of forced air cooling / water cooling), and environmental air pressure. The structural load parameters include at least one of the material properties of the PCBA template (such as material thermal conductivity, copper foil thickness), component layout density, and packaging method.

[0202] Among them, the preset duration can be set according to actual needs, such as 10 minutes, 20 minutes, etc., which is not limited in this application. The time step can also be set according to actual needs, such as 1 second, 2 seconds, etc., which is not limited in this application.

[0203] Regarding the temperature field change information at each time point within a preset future duration, the temperature field change information includes the predicted temperature values of each node in the simulation grid at this time point. Correspondingly, the operation of determining whether there is an anomaly based on the temperature field change information is specifically implemented as follows: for each node, if the difference between the predicted temperature values of the node at two adjacent time points is greater than the preset difference, it is determined that the node has an anomaly, and directly this difference is determined as the temperature anomaly value of the node; conversely, if the difference between the predicted temperature values of the node at two adjacent time points is less than or equal to the preset difference, it is determined that the node has no anomaly. Among them, the preset difference can be set according to actual needs, and this embodiment does not limit it, for example, 2°C, 3°C, etc.

[0204] Through the temperature evolution prediction model, the temperature field change information within a future period of time can be predicted, so as to provide a basis for preventive maintenance.

[0205] Optionally, the training process of the temperature evolution prediction model includes: iteratively training the initial temperature evolution prediction model by minimizing the loss function to obtain the trained temperature evolution prediction model. Among them, the loss function is determined based on the heat conduction equation and the data fitting error; the data fitting error is the mean square error between the predicted temperature and the sample temperature.

[0206] Among them, the sample data required for training is the historical temperature data and load parameters of the PCBA at each time point in the past period of time; the historical temperature data includes the measured temperatures of each node. The PCBA can be the PCBA in the actual production environment.

[0207] The loss function consists of two parts: the physical constraint term and the data-driven loss. The physical constraint term is the residual of the heat conduction equation, and the data-driven loss is the data fitting error, that is, the mean square error between the predicted temperature and the sample temperature. The sample temperature is the measured temperature.

[0208] Exemplarily, the loss function is shown as the following formula:

[0209] L = L data + λL physics

[0210] Among them, L is the model loss, L data is the data-driven loss, L physics is the physical constraint term, λ is the weight coefficient, and λ can be set according to experience, and this application does not limit it.

[0211] By determining the loss function of the model based on the heat conduction equation and the data fitting error, the trained model can predict more accurately.

[0212] This embodiment provides multiple implementation methods for determining whether the temperature field distribution data is abnormal. In practical applications, at least one implementation method can be flexibly selected, with strong flexibility.

[0213] Based on the above embodiments, in some embodiments, different images and information can be superimposed and displayed through a mixed reality device (such as a HoloLens2 device). This includes real-time infrared images, and the transparency of the images can be set to 50%; it also includes simulated temperature field isotherms, and the interval of the isotherms can be set to 5°C; it also includes a difference cloud map, where red in the difference cloud map indicates that the measured value is greater than the simulated value, and blue indicates that the measured value is less than the simulated value. Among them, the real-time infrared image is generated based on the temperature data output by the infrared thermal imaging unit. The simulated temperature field isotherms and the difference cloud map are generated based on the temperature field distribution data output by the temperature field simulation sub-model. With the help of visualization technology, through the superimposed display of images and information, it is convenient for technicians to view, reduces the fault location time, and thus improves the production operation and maintenance efficiency.

[0214] Based on the above embodiments, in some embodiments, an edge-cloud collaborative computing architecture can also be deployed. For example, data preprocessing and real-time anomaly detection are performed at the edge, and model training is performed in the cloud. The edge-cloud collaborative computing architecture has a small delay in real-time anomaly detection at the edge and a fast model training speed in the cloud, and supports the expansion of a thousand-node scale.

[0215] Through the description of the above implementation manners, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation manner.

[0216] Figure 5 It is a schematic structural diagram of a PCBA reliability prediction device provided by an embodiment of the present application. As Figure 5 shown, the PCBA reliability prediction device 50 includes:

[0217] An acquisition module 501, configured to collect the temperature of the PCBA sample to obtain measured temperature data;

[0218] A construction module 502, configured to respectively construct a temperature field simulation sub-model, a mechanical field simulation sub-model, and an electric field simulation sub-model corresponding to the PCBA sample according to the measured temperature data;

[0219] A coupling module 503, configured to perform coupling processing on the temperature field simulation sub-model, the mechanical field simulation sub-model, and the electric field simulation sub-model to obtain a multi-physical field simulation model;

[0220] The simulation module 504 is used to perform simulations for multiple fault modes respectively through a multi-physics simulation model to obtain multi-physics distribution data corresponding to each fault mode;

[0221] The prediction module 505 is used to perform reliability prediction on the PCBA sample board according to the multi-physics distribution data corresponding to each fault mode to obtain reliability-related information.

[0222] In a possible implementation manner, the acquisition module 501 is used to:

[0223] Collect the temperature of the PCBA sample board through a multi-modal sensor array to obtain the original temperature data; the multi-modal sensor array includes a thermocouple array, a distributed optical fiber temperature measurement unit, and an infrared thermal imaging unit;

[0224] Perform denoising processing on the original temperature data to obtain the measured temperature data.

[0225] In a possible implementation manner, the construction module 502 is used to:

[0226] For the PCBA sample board, respectively construct a heat conduction sub-model, a convective heat dissipation sub-model, and a radiative heat transfer sub-model;

[0227] Perform coupling processing on the heat conduction sub-model, the convective heat dissipation sub-model, and the radiative heat transfer sub-model to obtain a temperature field simulation sub-model; the temperature field simulation sub-model is represented in the form of a non-linear heat balance equation;

[0228] Use the Newton-Raphson algorithm to solve the non-linear heat balance equation to obtain the temperature field distribution data under the current material parameters;

[0229] Determine the residual matrix according to the measured temperature data and the temperature field distribution data of the PCBA sample board;

[0230] Determine the root mean square error of the residuals according to the residual matrix;

[0231] Judge whether the root mean square error of the residuals is greater than the preset error value;

[0232] If the root mean square error of the residuals is greater than the preset error value, update the material parameters of the temperature field simulation sub-model to obtain an optimized temperature field simulation sub-model;

[0233] Repeat the above iterative operation until the root mean square error of the residuals is less than or equal to the preset error value.

[0234] In a possible implementation manner, the construction module 502 is further used to:

[0235] Determine the temperature gradient between adjacent nodes in the simulation grid corresponding to the PCBA sample board according to the temperature field distribution data;

[0236] If the temperature gradient is greater than a preset gradient value, grid encryption is performed on the grid region where adjacent nodes are located to obtain a new simulation grid.

[0237] In a possible implementation, the simulation module 504 is configured to:

[0238] For each fault mode, the following operations are respectively performed, and the following operations include:

[0239] Inject the fault mode into the temperature field simulation sub-model in a parameterized form;

[0240] Perform numerical solution on the temperature field simulation sub-model to obtain temperature field distribution data;

[0241] Using the temperature field distribution data as input, drive the mechanical field simulation sub-model to perform numerical solution to obtain stress distribution data;

[0242] Using the temperature field distribution data as input, drive the electric field simulation sub-model to perform numerical solution to obtain current density distribution data.

[0243] In a possible implementation, the prediction module 505 is configured to:

[0244] For each fault mode respectively, the following operations are performed, and the following operations include:

[0245] Determine the wire resistance change data according to the temperature field distribution data after the injection of the fault mode and the temperature field distribution data before the injection of the fault mode;

[0246] Determine the signal attenuation information according to the wire resistance change data;

[0247] Determine the regional fracture information according to the stress distribution data after the injection of the fault mode and the stress distribution data before the injection of the fault mode;

[0248] Determine the electromigration information according to the current density distribution data after the injection of the fault mode and the current density distribution data before the injection of the fault mode.

[0249] In a possible implementation, it further includes a detection module, which is configured to:

[0250] Obtain the temperature field distribution data output by the temperature field simulation sub-model;

[0251] Determine whether there is an abnormality in the temperature field distribution data through a preset detection algorithm;

[0252] If there is an abnormality, generate an abnormality detection result;

[0253] Match the abnormality detection result with the fault types recorded in the database to obtain the target fault type;

[0254] Feedback the anomaly detection result and the target fault type to the manufacturing execution system.

[0255] In a possible implementation, the detection module is configured to perform at least one of the following:

[0256] Use the trained thermal behavior benchmark model to determine whether there is an anomaly in the temperature field distribution data; or,

[0257] Use the isolation forest-autoencoder hybrid algorithm to determine whether there is an anomaly in the temperature field distribution data; or,

[0258] Use the trained temperature evolution prediction model to determine whether there is an anomaly in the temperature field distribution data.

[0259] In a possible implementation, the detection module is configured to:

[0260] Obtain the current working condition data of the PCBA sample;

[0261] Input the working condition data as input data into the thermal behavior benchmark model, and output the predicted temperature range through the thermal behavior benchmark model;

[0262] Determine whether there is an anomaly in the temperature field distribution data according to the predicted temperature range;

[0263] In a possible implementation, it further includes a training module, which is configured to:

[0264] Obtain sample data and label data; the sample data includes the historical working condition data of the PCBA, and the label data includes the measured temperature range under the working condition represented by the sample working condition data;

[0265] Using the label data as supervised data, iteratively train the initial thermal behavior benchmark model according to the sample data to obtain the trained thermal behavior benchmark model.

[0266] In a possible implementation, the detection module is configured to:

[0267] Input the temperature field distribution data and the corresponding load parameters as input data into the temperature evolution prediction model, and roll-predict the temperature field change information at each time point within a preset future duration through the temperature evolution prediction model;

[0268] Determine whether there is a potential anomaly according to the temperature field change information.

[0269] In a possible implementation, it further includes a training module, which is configured to:

[0270] Iteratively train the initial temperature evolution prediction model by minimizing the loss function to obtain a trained temperature evolution prediction model; wherein the loss function is determined based on the heat conduction equation and the data fitting error; the data fitting error is the mean square error between the predicted temperature and the sample temperature.

[0271] For the description of the features in the corresponding embodiment of the PCBA reliability prediction device 50, reference can be made to the relevant description in the corresponding embodiment of the PCBA reliability prediction method, which will not be elaborated here one by one.

[0272] Figure 6 The structure diagram of an electronic device provided by this application. This electronic device is also Figure 1 the electronic device 40 shown. As Figure 6 shown, the electronic device 40 provided in this embodiment includes: a processor 401 and a memory 402. Optionally, the electronic device 40 further includes a communication component 403. Among them, the processor 401, the memory 402, and the communication component 403 are connected through a bus.

[0273] In the specific implementation process, the memory 402 is used to store computer programs. The processor 401 executes the computer programs stored in the memory 402, so that the processor 401 executes the above-mentioned embodiment of the PCBA reliability prediction method.

[0274] For the specific implementation process of the processor 401, reference can be made to the above method embodiment, and its implementation principle and technical effects are similar, which will not be elaborated here in this embodiment.

[0275] In the above embodiment, it should be understood that the processor can be a central processing unit (Central Processing Unit, CPU), and can also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in combination with the application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0276] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0277] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0278] Embodiments of the present application also provide a computer-readable storage medium, in which a computer program is stored. Wherein, the computer program is configured to execute the steps in any of the above embodiments of the PCBA reliability prediction method when running.

[0279] In an exemplary embodiment, the above computer-readable storage medium may include, but is not limited to: various media such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disc that can store a computer program.

[0280] Embodiments of the present application also provide a computer program product. The above computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above embodiments of the PCBA reliability prediction method.

[0281] Embodiments of the present application also provide another computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above embodiments of the PCBA reliability prediction method.

[0282] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0283] The above has introduced in detail a PCBA reliability prediction solution provided by the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A PCBA reliability prediction method, characterized in that, Including: Collecting the temperature of a printed circuit board assembly sample to obtain measured temperature data; Respectively constructing a temperature field simulation sub-model, a mechanical field simulation sub-model, and an electric field simulation sub-model corresponding to the sample according to the measured temperature data; Performing coupling processing on the temperature field simulation sub-model, the mechanical field simulation sub-model, and the electric field simulation sub-model to obtain a multi-physical field simulation model; Simulating various fault modes through the multi-physical field simulation model to obtain multi-physical field distribution data corresponding to each of the fault modes; Predicting the reliability of the sample according to the multi-physical field distribution data corresponding to each of the fault modes to obtain reliability-related information.

2. The method according to claim 1, characterized in that The collecting the temperature of a printed circuit board assembly sample to obtain measured temperature data includes: Collecting the temperature of the sample through a multi-modal sensor array to obtain raw temperature data; the multi-modal sensor array includes a thermocouple array, a distributed optical fiber temperature measurement unit, and an infrared thermal imaging unit; 3. The method according to claim 1, characterized in that Performing denoising processing on the raw temperature data to obtain the measured temperature data. According to the measured temperature data, constructing the temperature field simulation sub-model corresponding to the sample includes: Respectively constructing a heat conduction sub-model, a convective heat dissipation sub-model, and a radiative heat transfer sub-model for the sample; Performing coupling processing on the heat conduction sub-model, the convective heat dissipation sub-model, and the radiative heat transfer sub-model to obtain the temperature field simulation sub-model; the temperature field simulation sub-model is represented in the form of a non-linear heat balance equation; Using the Newton-Raphson algorithm to solve the non-linear heat balance equation to obtain temperature field distribution data under the current material parameters; Determining a residual matrix according to the measured temperature data of the sample and the temperature field distribution data; Determining the root mean square error of the residuals according to the residual matrix; Judging whether the root mean square error of the residuals is greater than a preset error value; If the root mean square error of the residuals is greater than the preset error value, updating the material parameters of the temperature field simulation sub-model to obtain an optimized temperature field simulation sub-model; 4. The method according to claim 3, wherein Repeating the above iterative operation until the root mean square error of the residuals is less than or equal to the preset error value. It also includes: Determining the temperature gradient between adjacent nodes in the simulation grid corresponding to the sample according to the temperature field distribution data; 5. The method according to claim 1, wherein If the temperature gradient is greater than a preset gradient value, performing grid encryption on the grid area where the adjacent nodes are located to obtain a new simulation grid. The simulating various fault modes through the multi-physical field simulation model to obtain multi-physical field distribution data corresponding to each of the fault modes includes: For each of the fault modes, respectively performing the following operations, and the following operations include: Injecting the fault mode into the temperature field simulation sub-model in a parameterized form; Performing numerical solution on the temperature field simulation sub-model to obtain temperature field distribution data; Using the temperature field distribution data as input to drive the mechanical field simulation sub-model to perform numerical solution to obtain stress distribution data; Taking the temperature field distribution data as input, driving the electric field simulation sub-model to perform numerical solution to obtain current density distribution data.

6. The method according to claim 5, wherein Predicting the reliability of the sample according to the multi-physical field distribution data respectively corresponding to each of the failure modes to obtain reliability-related information, including: For each of the failure modes, perform the following operations, where the following operations include: Determine the wire resistance change data according to the temperature field distribution data after failure mode injection and the temperature field distribution data before failure mode injection; Determine the signal attenuation information according to the wire resistance change data; Determine the regional fracture information according to the stress distribution data after failure mode injection and the stress distribution data before failure mode injection; Determine the electromigration information according to the current density distribution data after failure mode injection and the current density distribution data before failure mode injection.

7. The method according to claim 1, wherein Further includes: Obtain the temperature field distribution data output by the temperature field simulation sub-model; Determine whether there is an abnormality in the temperature field distribution data through a preset detection algorithm; If there is an abnormality, generate an abnormality detection result; Match the abnormality detection result with the failure types recorded in the database to obtain the target failure type; Feed back the abnormality detection result and the target failure type to the manufacturing execution system.

8. The method according to claim 7, wherein Determining whether there is an abnormality in the temperature field distribution data through the preset detection algorithm includes at least one of the following: Using the trained thermal behavior benchmark model to determine whether there is an abnormality in the temperature field distribution data; or, Using the isolation forest-autoencoder hybrid algorithm to determine whether there is an abnormality in the temperature field distribution data; or, Using the trained temperature evolution prediction model to determine whether there is an abnormality in the temperature field distribution data.

9. The method according to claim 8, wherein Using the trained thermal behavior benchmark model to determine whether there is an abnormality in the temperature distribution data includes: Obtain the current working condition data of the sample; Take the working condition data as input data and input it into the thermal behavior benchmark model, and output the predicted temperature range through the thermal behavior benchmark model; Determine whether there is an abnormality in the temperature field distribution data according to the predicted temperature range.

10. The method according to claim 9, wherein The training process of the thermal behavior benchmark model includes: Obtain sample data and label data; the sample data includes the historical working condition data of the printed circuit board assembly, and the label data includes the measured temperature range under the working condition represented by the sample working condition data; Using the label data as supervised data, iteratively train the initial thermal behavior benchmark model according to the sample data to obtain the trained thermal behavior benchmark model.

11. The method according to claim 8, characterized in that Using the trained temperature evolution prediction model to determine whether there is an abnormality in the temperature field distribution data includes: Take the temperature field distribution data and the corresponding load parameters as input data and input them into the temperature evolution prediction model, and roll-predict the temperature field change information at each time point within a preset future duration through the temperature evolution prediction model; Determine whether there is a potential abnormality according to the temperature field change information.

12. The method according to claim 11, wherein The training process of the temperature evolution prediction model includes: The initial temperature evolution prediction model is iteratively trained by minimizing the loss function to obtain the trained temperature evolution prediction model; wherein the loss function is determined based on the heat conduction equation and the data fitting error; the data fitting error is the mean square error between the predicted temperature and the sample temperature.

13. An electronic device, characterized in that, It includes: A memory for storing computer programs; A processor for implementing the steps of the PCBA reliability prediction method according to any one of claims 1 to 12 when executing the computer program.

14. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, wherein the computer program implements the steps of the PCBA reliability prediction method according to any one of claims 1 to 12 when executed by a processor.

15. A computer program product, comprising a computer program, characterized in that, The computer program implements the steps of the PCBA reliability prediction method according to any one of claims 1 to 12 when executed by a processor.

Images