PCB analysis and prediction method and device based on deep learning, equipment and medium

Abstract

The invention provides a PCB analysis and prediction method and device based on deep learning, equipment and a medium, and the method comprises the steps: obtaining a PCB data set through collecting historical PCB design parameters of different dimensions in design, testing and manufacturing; performing principal component analysis on the PCB data set to obtain data after dimension reduction; constructing a training sample set and a test sample set based on the data after dimension reduction; training the constructed deep learning model by using the training sample set and the test sample set to obtain a prediction model; the current PCB design parameters are analyzed and predicted based on the prediction model, a corresponding prediction result is obtained, the prediction accuracy, comprehensiveness and efficiency are greatly improved, and meanwhile the time cost and resource consumption in the PCB design process are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the field of PCB technology, and in particular to a PCB analysis and prediction method, device, equipment, and medium based on deep learning. Background Art

[0002] Printed Circuit Board (PCB) is one of the most basic components in the electronics industry and is widely used in various electronic devices.

[0003] PCB design is the process of designing and optimizing the layout, routing, and manufacturing of a PCB based on a circuit schematic using electronic design automation (EDA) software. The quality and efficiency of PCB design directly impact the performance, reliability, and cost of electronic products.

[0004] Currently, most prediction methods focus solely on cost and time estimation, lacking comprehensive forecasting capabilities for key parameters such as electrical performance and thermal characteristics, making them incapable of comprehensive analysis and forecasting of PCBs. Furthermore, traditional methods often rely on simplified empirical formulas or statistical models. As PCB design complexity continues to increase, these methods have become less accurate and less adaptable to evolving technical conditions and market demands.

[0005] Therefore, there is an urgent need for a PCB analysis and prediction method, device, equipment and medium based on deep learning to improve the above problems. Summary of the Invention

[0006] The purpose of the present invention is to provide a PCB analysis and prediction method, device, equipment and medium based on deep learning, which can improve the accuracy and comprehensiveness of PCB analysis and prediction.

[0007] In a first aspect, the present invention provides a PCB analysis and prediction method based on deep learning, comprising:

[0008] Collect historical PCB design parameters from different dimensions during design, testing, and manufacturing to obtain a PCB dataset;

[0009] Performing principal component analysis on the PCB data set to obtain dimension-reduced data;

[0010] Constructing a training sample set and a test sample set based on the data after dimensionality reduction;

[0011] Use the training sample set and the test sample set to train the deep learning model and obtain the prediction model;

[0012] The current PCB design parameters are analyzed and predicted based on the prediction model to obtain corresponding prediction results.

[0013] The method of the present invention has the following beneficial effects: a PCB data set is obtained by collecting historical PCB design parameters of different dimensions in design, testing and manufacturing; principal component analysis is performed on the PCB data set to obtain dimensionality-reduced data; a training sample set and a test sample set are constructed based on the dimensionality-reduced data; a deep learning model is trained using the training sample set and the test sample set to obtain a prediction model; current PCB design parameters are analyzed and predicted based on the prediction model to obtain corresponding prediction results, which not only greatly improves the accuracy, comprehensiveness and efficiency of the prediction, but also greatly reduces the time cost and resource consumption in the PCB design process.

[0014] Optionally, the PCB design parameters include design parameters related to circuit layout and logic in the design phase, test parameters related to performance in the test phase, and manufacturing parameters related to physical manufacturing process and cost in the manufacturing phase;

[0015] Wherein, the design parameters include layout parameters, circuit design, component selection, electrical connection and power distribution;

[0016] Said test parameters include material selection, manufacturing process, surface treatment, quality control and tolerance design;

[0017] The manufacturing parameters include electrical performance, thermal performance, mechanical stability, reliability, cost and sales volume.

[0018] Optionally, the layout parameters include board size, shape, number of layers, and layout and alignment of components;

[0019] The circuit design includes the width, spacing, routing of circuit lines, and the size and distribution of pads;

[0020] The component selection includes component type, packaging specifications and pressure and heat resistance performance;

[0021] The electrical connection includes the design of the connection point, the connectivity of the circuit and the integrity of the signal;

[0022] The power distribution includes the layout of power lines and ground lines, and the design of power layers;

[0023] The material selection includes substrate material, conductive layer material and solder resist ink;

[0024] The manufacturing process includes lamination, drilling, metallization, welding and assembly;

[0025] The surface treatment includes gold plating, tin plating or chemical copper plating;

[0026] Said quality control includes inspection and testing standards during the manufacturing process;

[0027] The tolerance design includes tolerance and accuracy requirements for each manufacturing step;

[0028] Said electrical performance includes signal integrity, noise level and electromagnetic compatibility;

[0029] The thermal properties include heat distribution, heat dissipation capacity, thermal stress and thermal expansion;

[0030] The mechanical stability includes bending resistance and vibration resistance;

[0031] The reliability includes service life, failure rate and resistance to environmental factors.

[0032] Optionally, the deep learning model is a feedforward neural network;

[0033] The feedforward neural network consists of two hidden layers, the first hidden layer contains 32 neurons, the second hidden layer contains 16 neurons, and the activation function of each neuron is a ReLU activation function.

[0034] Optionally, the output layer of the feedforward neural network is a single neuron layer, and the mean square error is used as the loss function to quantify the gap between the predicted value and the actual value, and the Adam optimizer is used to adjust the weight.

[0035] Optionally, also include:

[0036] Preprocessing the collected PCB design parameters to obtain preprocessed data;

[0037] Performing correlation analysis on the preprocessed data to obtain a PCB data set;

[0038] The preprocessing includes outlier processing and missing value processing.

[0039] Optionally, principal component analysis is performed on the PCB dataset to obtain dimensionally reduced data including:

[0040] Performing standardization processing on the PCB data set to obtain standardized data;

[0041] Calculating a covariance matrix of the standardized data;

[0042] Performing eigenvalue decomposition on the covariance matrix to obtain the eigenvalues of the principal components;

[0043] Select the first k principal components with the largest eigenvalues according to the size of the eigenvalues;

[0044] Construct a feature vector based on the selected k principal components;

[0045] Projecting the PCB dataset onto the principal components to obtain dimensionally reduced data;

[0046] Wherein, k is a positive integer.

[0047] Optionally, the prediction results include cost prediction results and performance prediction results, and the performance prediction results include electrical and thermal characteristics and physical strength.

[0048] Optionally, also include:

[0049] Collect historical PCB design parameters of N target objects in different dimensions during design, testing, and manufacturing to obtain a PCB dataset for each target object, where N is a positive integer;

[0050] Perform principal component analysis on the PCB dataset of each target object to obtain the data after dimensionality reduction;

[0051] Constructing a training sample set and a test sample set based on the data after dimensionality reduction;

[0052] Use the training sample set and the test sample set to train the deep learning model and obtain the prediction model for each target object;

[0053] The prediction model of each target object is regularly updated and optimized to obtain an optimized model.

[0054] In a second aspect, the present invention provides a deep learning-based PCB analysis and prediction device, comprising modules / units for executing any of the possible design methods described in the first aspect. These modules / units can be implemented in hardware, or by executing corresponding software implementations in hardware.

[0055] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a program that can be run on the processor, and when the program is executed by the processor, the electronic device implements a method for executing any possible design of any of the above aspects.

[0056] In a fourth aspect, the present invention provides a readable storage medium, wherein the readable storage medium stores a program, and when the program is executed, it implements any possible design method of any of the above aspects.

[0057] For the beneficial effects of the second to fourth aspects, please refer to the description of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A schematic diagram of a process flow of a PCB analysis and prediction method based on deep learning provided by an embodiment of the present invention;

[0059] Figure 2 A schematic diagram of the principle of correlation analysis provided by an embodiment of the present invention;

[0060] Figure 3 A schematic diagram of the structure of a PCB analysis and prediction device based on deep learning provided by an embodiment of the present invention;

[0061] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein should be the common meanings understood by people with ordinary skills in the field to which the present invention belongs. The words "including" and similar words used in this article mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0063] like Figure 1 As shown, the present invention provides a PCB analysis and prediction method based on deep learning, comprising:

[0064] S101, collect historical PCB design parameters in different dimensions during design, testing, and manufacturing to obtain a PCB dataset.

[0065] In some embodiments, the PCB design parameters include circuit layout and logic-related design parameters in the design stage, performance-related test parameters in the test stage, and manufacturing parameters related to physical manufacturing processes and costs in the manufacturing stage; wherein, the design parameters include layout parameters, circuit design, component selection, electrical connection and power distribution; the test parameters include material selection, manufacturing process, surface treatment, quality control and tolerance design; the manufacturing parameters include electrical performance, thermal performance, mechanical stability, reliability, cost and sales volume.

[0066] In some specific embodiments, the layout parameters include the size, shape, number of layers of the board, and the layout and alignment of components; the circuit design includes the width, spacing, routing path of the circuit lines, and the size and distribution of the pads; the component selection includes the component type, packaging specifications and voltage and heat resistance; the electrical connection includes the design of the connection points, the connectivity of the circuit and the integrity of the signal; the power distribution includes the layout of the power line and the ground line, and the design of the power layer; the material selection includes the substrate material, the conductive layer material and the solder mask ink; the manufacturing process includes lamination process, drilling, metallization, welding and assembly; the surface treatment includes gold plating, tin plating or chemical copper plating; the quality control includes inspection and testing standards during the manufacturing process; the tolerance design includes the tolerance and precision requirements for each manufacturing step; the electrical performance includes signal integrity, noise level and electromagnetic compatibility; the thermal performance includes heat distribution, heat dissipation capacity, thermal stress and thermal expansion; the mechanical stability includes bending resistance and vibration resistance; the reliability includes service life, failure rate and resistance to environmental factors.

[0067] In other embodiments, the method further includes: preprocessing the collected PCB design parameters to obtain preprocessed data; performing correlation analysis on the preprocessed data to obtain a PCB data set to exclude low-correlation features and reduce redundancy and unnecessary computational burden; the preprocessing includes outlier processing and missing value processing to ensure the quality of data used for analysis.

[0068] S102, performing principal component analysis on the PCB dataset to obtain dimensionally reduced data.

[0069] In some embodiments, performing principal component analysis on the PCB dataset to obtain dimensionally reduced data includes: normalizing the PCB dataset to obtain standardized data; calculating a covariance matrix of the standardized data; performing eigenvalue decomposition on the covariance matrix to obtain the sizes of eigenvalues of the principal components; selecting the top k principal components with large eigenvalues based on the sizes of the eigenvalues; constructing eigenvectors based on the selected k principal components; projecting the PCB dataset onto the principal components to obtain dimensionally reduced data; wherein k is a positive integer that reduces the dimension of the data, eliminates multicollinearity between features, extracts the most influential features, and selects the main components that can maximize the variability of the data through principal component analysis, thereby providing a more refined and effective feature set for model training.

[0070] S103: constructing a training sample set and a test sample set based on the data after dimensionality reduction.

[0071] S104: Use the training sample set and the test sample set to train the constructed deep learning model to obtain a prediction model.

[0072] In some embodiments, the deep learning model is a feedforward neural network; the feedforward neural network consists of two hidden layers, and the first hidden layer contains 32 neurons, and the second hidden layer contains 16 neurons. The activation function of each neuron is a ReLU activation function, which increases the model's ability to handle nonlinear problems while also avoiding the problem of gradient disappearance.

[0073] In some specific embodiments, the output layer of the feedforward neural network is a single neuron layer, and the mean square error is used as the loss function to quantify the gap between the predicted value and the actual value. The Adam optimizer is used to adjust the weights, and the advantages of Momentum and RMSprop of stochastic gradient descent are combined to achieve fast and stable convergence.

[0074] S105: Analyze and predict the current PCB design parameters based on the prediction model to obtain corresponding prediction results.

[0075] In some embodiments, the prediction results include cost prediction results and performance prediction results, and the performance prediction results include electrical and thermal characteristics and physical strength.

[0076] The embodiment of the present invention comprehensively considers the characteristics of the design stage, the testing stage, and the manufacturing stage. From the different levels of design concept to physical implementation, the parameters of different stages are integrated in the prediction model to achieve a more comprehensive prediction of the overall performance and cost of the PCB. In addition, the parameters of the design and manufacturing stages can be adjusted by the designer, and the parameters of the testing stage are used to verify the performance, cost, etc. of the PCB. The prediction results obtained by the prediction model can help designers determine whether the current design and manufacturing plan meets the actual required product performance. If not, the design can be readjusted. The method of the present invention combines traditional data processing technology with advanced deep learning algorithms to provide a comprehensive solution for PCB design parameter analysis and prediction. This method can significantly improve the accuracy and efficiency of the prediction, while greatly reducing the time cost and resource consumption in the design process.

[0077] In other embodiments, customized data collection can be performed for different customers and manufacturers to ensure coverage of a wide range of application scenarios and technical requirements, meet multi-source data and personalized data needs, collect historical PCB design parameters of N target objects (e.g., different customers and manufacturers) in different dimensions of design, testing, and manufacturing, and obtain a PCB dataset for each target object, where N is a positive integer. The data is classified according to the data source (different customers or manufacturers) to implement targeted processing and analysis. Specific preprocessing methods: Specific preprocessing steps are implemented for data from different sources, such as standardization, outlier processing, missing value filling, etc., to ensure data consistency and quality.

[0078] A principal component analysis is performed on the PCB dataset of each target object to obtain the data after dimensionality reduction. A principal component analysis is performed independently on the dataset of each customer or manufacturer to capture its unique data characteristics and variability. Based on the personalized principal component analysis results, the main components that best represent the data characteristics of a specific customer or manufacturer are selected.

[0079] A training sample set and a test sample set are constructed based on the reduced-dimensional data; the training sample set and the test sample set are used to train the constructed deep learning model to obtain a prediction model for each target object, and an independent deep learning model is trained for each customer or manufacturer's data set to adapt to its unique data characteristics and needs, and the performance of each customized model is optimized to ensure the best prediction effect on its specific data set.

[0080] Based on the prediction model of each target object, a mechanism is established to regularly update and optimize the prediction model to obtain an optimized model to adapt to changes in the data and needs of each target object.

[0081] For ease of understanding, this embodiment further illustrates the specific implementation method of the above method in combination with a specific application scenario system, such as Figure 2 As shown, the specific steps include:

[0082] Step a: Data collection and preprocessing

[0083] Data collection: Collect PCB data sets from design, testing, and manufacturing to ensure coverage of multi-dimensional information at different stages. For example, design parameters in the design stage prioritize layout parameters, circuit design, component selection, electrical connections, and power distribution; test parameters in the testing stage prioritize material selection, manufacturing process, surface treatment, quality control, and tolerance design; and manufacturing parameters in the manufacturing node prioritize electrical performance, thermal performance, mechanical stability, reliability, cost, and sales volume.

[0084] Data preprocessing: This includes removing outliers, processing missing values, and eliminating noise to obtain preprocessed data and ensure the quality of data used for analysis. For outlier processing, statistical metrics (such as mean and standard deviation) can be used to define outliers. For example, data points outside the range of ±3 times the standard deviation of the mean are considered outliers. Outliers are then replaced with the median, mean, or values estimated using a predictive model (such as a regression model). For missing value processing, methods such as linear interpolation, polynomial interpolation, spline interpolation, or multiple interpolation can be applied to estimate missing values based on adjacent points.

[0085] Correlation analysis: Figure 2 As shown in Figure 1, the Pearson correlation coefficient is used to perform correlation analysis on the preprocessed data to obtain the PCB data set, excluding low-correlation features and reducing redundancy and unnecessary computational burden.

[0086] Step b, principal component analysis

[0087] This step primarily involves performing principal component analysis (PCA) on the PCB dataset to reduce the data dimension, eliminate multicollinearity between features, and extract the most influential features. PCA selects the main components that best reflect data variability, providing a more refined and effective feature set for model training.

[0088] Normalization: Before performing principal component analysis, the PCB dataset is first normalized to obtain normalized data to ensure that each feature has the same weight in the analysis. Normalization usually involves subtracting the mean of the feature and dividing by its standard deviation.

[0089] Covariance Matrix Calculation: Calculates the covariance matrix of the standardized data. The covariance matrix reflects the linear relationship between data features and is a key component of principal component analysis.

[0090] Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to determine the eigenvalues of the principal components. Eigenvalue decomposition can reveal the main direction of change in the data set and sort by eigenvalue to determine the importance of each principal component.

[0091] Principal component selection: The top k most important principal components are selected based on their eigenvalues. The selection principle is usually based on the cumulative contribution rate, that is, the cumulative contribution rate of the top k principal components reaches a preset threshold (such as 95%), which means that they can explain 95% of the variation in the data.

[0092] Construct eigenvectors: Construct eigenvectors based on the selected principal components. Eigenvectors are obtained by eigenvalue decomposition of the covariance matrix. They are orthogonal and form a new feature space.

[0093] Data Projection: Project the PCB dataset onto the selected principal components to obtain the reduced dimensionality data. This is done by matrix multiplication of the original data matrix and the eigenvector matrix.

[0094] Step c, deep learning model training

[0095] Model Architecture: A fully connected feedforward neural network (FNN) was constructed. The network's input layer was designed to accept feature vectors processed with PCA and normalized. The network consists of two hidden layers, the first containing 32 neurons and the second containing 16 neurons. Each neuron uses the ReLU (Rectified Linear Unit) activation function, which improves the model's ability to handle nonlinear problems while also preventing the vanishing gradient problem.

[0096] Loss Function and Optimizer: The model's output layer is a single-neuron layer that predicts continuous cost values. Mean Squared Error (MSE) is used as the loss function to quantify the difference between the predicted and actual values. The Adam optimizer is used to adjust weights. It combines the advantages of momentum and RMSprop from stochastic gradient descent to achieve fast and stable convergence.

[0097] Overfitting prevention and control: Dropout layers are inserted between dense layers to randomly discard some connections to enhance the generalization ability of the model.

[0098] Training and Validation: During training, the reduced data is divided into a training set and a validation set. The model is learned on the training set, and its performance is evaluated on the validation set. This helps monitor and avoid overfitting while also adjusting the network parameters.

[0099] Batch size and number of iterations: When training the model, an appropriate batch size (such as 32 or 64) is selected to balance computational efficiency and memory limitations. At the same time, a sufficiently large number of epochs is set to ensure that the model has enough time to learn, but early stopping is used to avoid unnecessary computation.

[0100] Performance evaluation: At the end of each epoch, the performance of the model is evaluated on the validation set, and the MSE and other possible evaluation indicators (such as MAE, mean squared error) are recorded.

[0101] Model saving and restoration: After training is completed, save the model weights and architecture so that the model can be quickly loaded in subsequent applications to make new predictions or continue training.

[0102] Step d, model prediction and optimization

[0103] Predictive Applications: Trained and validated deep learning models can be deployed in production environments to predict new PCB design parameters in real time. These models can predict a range of key metrics, such as cost estimates, estimated manufacturing time, signal integrity, thermal characteristics, and more, providing comprehensive design evaluations for production.

[0104] Optimization Recommendations: Based on the model output, targeted optimization suggestions can be made. For example, if the predicted cost exceeds the budget, a material change or design modification to reduce the number of layers can be recommended. These suggestions are automatically generated by analyzing the difference between the predicted results and the target parameters and incorporating empirical rules of PCB manufacturing.

[0105] Continuous Learning and Optimization: Supports online learning, enabling continuous updating and adjustment of its prediction logic based on the latest data. In this way, the model can adapt to new design trends and changes in manufacturing technology, maintaining its prediction accuracy.

[0106] Step e, user interaction interface

[0107] Interface design: An intuitive graphical user interface (GUI) can be configured through which users can input PCB design parameters such as size, number of layers, material type, etc., select the desired performance prediction results, and intuitively view the results through charts and numerical indicators.

[0108] Interactive Exploration: Users can explore different design options within the interface and instantly view the predicted results for each option. This feature allows users to adjust parameters and compare the impact of different designs on cost and performance, helping designers make more informed decisions.

[0109] Feedback Mechanism: This invention integrates a feedback loop where users can provide feedback on the actual design output based on the predicted results. For example, if the actual cost is much lower than the predicted cost, the user can input this data and the prediction model will be adjusted and improved with this new information.

[0110] The advantages of this embodiment are as follows:

[0111] 1. Improve prediction accuracy: Ability to provide high-precision predictions of PCB design parameters such as cost, time, and various performance indicators, helping to reduce design iterations and speed up product launch.

[0112] 2. Comprehensive performance evaluation: Not limited to cost prediction, the present invention can also evaluate and predict multiple performance parameters of PCB designs, including electrical, thermal characteristics, and physical strength, providing designers with comprehensive performance analysis.

[0113] 3. Reduce design costs and improve design efficiency: By predicting costs and performance during the design phase, this invention can help avoid expensive later revisions, thereby reducing overall design and manufacturing costs. With fast and accurate parameter predictions, designers can evaluate more design options in a shorter time, significantly improving design efficiency.

[0114] 4. Real-time feedback and continuous learning: Combining continuous learning and user feedback mechanisms, the system can continuously optimize the prediction algorithm based on the latest design results and market dynamics, ensuring that the prediction results remain accurate and relevant over time.

[0115] 5. Adaptability and scalability: The present invention takes into account different PCB design requirements and diverse user scenarios, has good adaptability and scalability, and can be easily extended to new prediction tasks and application areas.

[0116] 6. Data-driven decision support: By providing data-driven insights, the present invention supports a more scientific and precise decision-making process, reducing the risk of decisions based on intuition, experience or insufficient information.

[0117] In summary, the present invention not only excels in design parameter prediction, but more importantly, it provides comprehensive support and advantages at multiple levels, helping designers achieve a more efficient, accurate, and sustainable PCB design process.

[0118] like Figure 3 As shown, based on the above-mentioned PCB analysis and prediction method based on deep learning, the present invention provides a PCB analysis and prediction device based on deep learning, including: an acquisition unit 301, used to acquire historical PCB design parameters of different dimensions in design, testing and manufacturing to obtain a PCB data set; an analysis unit 302, used to perform principal component analysis on the PCB data set to obtain data after dimensionality reduction; a construction unit 303, used to construct a training sample set and a test sample set based on the data after dimensionality reduction; a training unit 304, used to train a deep learning model using the training sample set and the test sample set to obtain a prediction model; a prediction unit 305, used to analyze and predict current PCB design parameters based on the prediction model to obtain corresponding prediction results.

[0119] It should be understood that all relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module, and will not be repeated here.

[0120] In other embodiments of the present invention, an electronic device 400 is disclosed. Figure 4 As shown, the system may include: one or more processors 401; a memory 402; a display 403; one or more applications (not shown); and one or more computer programs 404. The above components may be connected via one or more communication buses 405. The one or more computer programs 404 are stored in the memory 402 and configured to be executed by the one or more processors 401. The one or more computer programs 404 include instructions, which may be used to execute the following instructions: Figure 1 The various steps in the corresponding embodiments.

[0121] While the embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations of these embodiments are possible. However, it should be understood that such modifications and variations are within the scope and spirit of the present invention as set forth in the claims. Furthermore, the invention described herein is susceptible to other embodiments and may be practiced or implemented in a variety of ways.

Claims

1. A PCB analysis and prediction method based on deep learning, characterized in that: include: Collect historical PCB design parameters from different dimensions during design, testing, and manufacturing to obtain a PCB dataset; Performing principal component analysis on the PCB data set to obtain dimension-reduced data; Constructing a training sample set and a test sample set based on the data after dimensionality reduction; Use the training sample set and the test sample set to train the deep learning model and obtain the prediction model; The current PCB design parameters are analyzed and predicted based on the prediction model to obtain corresponding prediction results.

2. The method according to claim 1, characterized in that The PCB design parameters include design parameters related to circuit layout and logic in the design phase, test parameters related to performance in the test phase, and manufacturing parameters related to physical manufacturing process and cost in the manufacturing phase; Wherein, the design parameters include layout parameters, circuit design, component selection, electrical connection and power distribution; Said test parameters include material selection, manufacturing process, surface treatment, quality control and tolerance design; The manufacturing parameters include electrical performance, thermal performance, mechanical stability, reliability, cost and sales volume.

3. The method according to claim 2, characterized in that The layout parameters include board size, shape, number of layers, and component placement and alignment; The circuit design includes the width, spacing, routing of circuit lines, and the size and distribution of pads; The component selection includes component type, packaging specifications and pressure and heat resistance performance; The electrical connection includes the design of the connection point, the connectivity of the circuit and the integrity of the signal; The power distribution includes the layout of power lines and ground lines, and the design of power layers; The material selection includes substrate material, conductive layer material and solder resist ink; The manufacturing process includes lamination, drilling, metallization, welding and assembly; The surface treatment includes gold plating, tin plating or chemical copper plating; Said quality control includes inspection and testing standards during the manufacturing process; The tolerance design includes tolerance and accuracy requirements for each manufacturing step; Said electrical performance includes signal integrity, noise level and electromagnetic compatibility; The thermal properties include heat distribution, heat dissipation capacity, thermal stress and thermal expansion; The mechanical stability includes bending resistance and vibration resistance; The reliability includes service life, failure rate and resistance to environmental factors.

4. The method according to claim 1, wherein The deep learning model is a feedforward neural network; The feedforward neural network consists of two hidden layers, the first hidden layer contains 32 neurons, the second hidden layer contains 16 neurons, and the activation function of each neuron is a ReLU activation function.

5. The method according to claim 4, characterized in that The output layer of the feedforward neural network is a single neuron layer. The mean square error is used as the loss function to quantify the gap between the predicted value and the actual value, and the Adam optimizer is used to adjust the weight.

6. The method according to claim 1, characterized in that Also includes: Preprocessing the collected PCB design parameters to obtain preprocessed data; Performing correlation analysis on the preprocessed data to obtain a PCB data set; The preprocessing includes outlier processing and missing value processing.

7. The method according to claim 1, characterized in that The principal component analysis of the PCB dataset is performed to obtain the following data after dimensionality reduction: Performing standardization processing on the PCB data set to obtain standardized data; Calculating a covariance matrix of the standardized data; Performing eigenvalue decomposition on the covariance matrix to obtain the eigenvalues of the principal components; Select the first k principal components with the largest eigenvalues according to the size of the eigenvalues; Construct a feature vector based on the selected k principal components; Projecting the PCB dataset onto the principal components to obtain dimensionally reduced data; Wherein, k is a positive integer.

8. The method according to claim 1, characterized in that The prediction results include cost prediction results and performance prediction results, and the performance prediction results include electrical and thermal characteristics and physical strength.

9. The method according to claim 1, characterized in that Also includes: Collect historical PCB design parameters of N target objects in different dimensions during design, testing, and manufacturing to obtain a PCB dataset for each target object, where N is a positive integer; Perform principal component analysis on the PCB dataset of each target object to obtain the data after dimensionality reduction; Constructing a training sample set and a test sample set based on the data after dimensionality reduction; Use the training sample set and the test sample set to train the deep learning model and obtain the prediction model for each target object; The prediction model of each target object is regularly updated and optimized to obtain an optimized model.

10. A PCB analysis and prediction device based on deep learning, used in the method according to any one of claims 1 to 9, characterized in that: include: The acquisition unit is used to collect historical PCB design parameters of different dimensions in design, testing and manufacturing to obtain a PCB data set; An analysis unit, configured to perform principal component analysis on the PCB data set to obtain dimension-reduced data; A construction unit, configured to construct a training sample set and a test sample set based on the dimensionality-reduced data; The training unit is used to train the deep learning model using the training sample set and the test sample set to obtain a prediction model; The prediction unit is used to analyze and predict the current PCB design parameters based on the prediction model to obtain corresponding prediction results.

11. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a program that can be run on the processor, and when the program is executed by the processor, the electronic device implements the method according to any one of claims 1 to 9.

12. A readable storage medium having a program stored therein, characterized in that: When the program is executed, the method according to any one of claims 1 to 9 is implemented.

Images