Electronic component classification prediction method and apparatus

Abstract

This invention provides a method and apparatus for classifying and predicting electronic components. The method involves obtaining a simplified 3D circuit board model of a printed circuit board (PCB) to be used for electronic heat dissipation simulation. The simplified 3D circuit board model includes physical models of electronic components. The simplified 3D circuit board model is input into a trained deep learning model incorporating a graph neural network to obtain classification labels for the physical models of electronic components within the 3D circuit board model. The deep learning model extracts geometric features and arrangement position information from the boundary representation data of the physical models of electronic components in the simplified 3D circuit board model, and predicts the classification labels of the physical models of electronic components based on these geometric features and arrangement position information. This invention can accurately classify electronic components.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an electronic component classification and prediction method, an electronic component classification and prediction device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] When performing electronic heat dissipation simulation on a PCB (Printed Circuit Board), each electronic component on the PCB has its own specific thermal characteristics, including but not limited to parameters such as heat generation, thermal resistance, and thermal capacity. These parameters directly affect the heat distribution and heat dissipation effect of the entire PCB. Only by accurately identifying the electronic components and inputting the corresponding physical property parameters can the accuracy and reliability of the electronic heat dissipation simulation be ensured. Therefore, it is crucial to accurately identify the electronic components on the PCB. Summary of the Invention

[0003] In view of the above problems, a method and apparatus for classifying and predicting electronic components are proposed to overcome or at least partially solve these problems. The specific technical solution is as follows:

[0004] In a first aspect of the present invention, a method for classifying and predicting electronic components is provided, the method comprising:

[0005] Obtain a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein the simplified 3D circuit board model includes solid models of electronic components; the simplified 3D circuit board model has fewer geometric features than the unsimplified 3D circuit board model.

[0006] The simplified 3D circuit board model is input into the trained deep learning model to obtain the classification label of the electronic component entity model in the 3D circuit board model output by the deep learning model; the deep learning model extracts geometric features and layout position information from the boundary representation data features of the electronic component entity model in the simplified 3D circuit board model, so as to predict the classification label of the electronic component entity model in the simplified 3D circuit board model based on the geometric features and the layout position information; the deep learning model is a deep learning model combined with graph neural network.

[0007] Optionally, the deep learning model includes an encoder and a neural network model. The simplified 3D circuit board model is input into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, including:

[0008] The encoder is used to extract the geometric features of the electronic component entity model in the simplified 3D circuit board model; wherein, the geometric features include geometric information and topological structure;

[0009] The simplified 3D circuit board model and the geometric features are input into the neural network model to obtain the classification labels of the electronic component entity models in the simplified 3D circuit board model output by the neural network model.

[0010] Optionally, before inputting the simplified 3D circuit board model into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, the method further includes:

[0011] Obtain a sample electronic component entity model; wherein, the sample electronic component entity model has corresponding actual geometric features, the actual geometric features including actual geometric information and actual topological structure;

[0012] The encoder to be trained is self-supervised using the physical model of the sample electronic components.

[0013] When the encoder reaches the preset training conditions after self-supervised training, the encoder that has completed training is obtained.

[0014] Optionally, before inputting the simplified 3D circuit board model into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, the method further includes:

[0015] Obtain a sample 3D circuit board model of a sample printed circuit board; wherein, the sample 3D circuit board model includes sample electronic component entity models, and the sample electronic component entity models have corresponding actual classification labels;

[0016] Calculate the center-to-center distance between the center points of the sample electronic component solid model;

[0017] Calculate the affinity matrix between the sample electronic component entity models based on the distance between the center points;

[0018] The connection relationships between the sample electronic components are determined based on the affinity matrix between the sample electronic component entity models and a preset threshold.

[0019] The physical model of the sample electronic components is converted into a coding unit;

[0020] For each coding unit of the sample printed circuit board, in reverse order of the distance between the center points of the coding units, the geometric feature vector of the coding unit is combined with the geometric feature vectors of other coding units that are connected to it to construct sub-graph data, and all the sub-graph data of the sample printed circuit board are combined into an input sequence.

[0021] The input sequence is used to pre-train the neural network model to be trained.

[0022] Optionally, determining the connection relationships between the sample electronic components based on the affinity matrix between the sample electronic component entity models and a preset threshold includes:

[0023] When the affinity matrix between the sample electronic component entity models is greater than the preset threshold, it is determined that there is a connection relationship between the sample electronic component entity models.

[0024] Optionally, pre-training the neural network model to be trained using the input sequence includes:

[0025] The encoding units in the input sequence are randomly masked to obtain the masked input sequence;

[0026] The masked input sequence is input into the neural network model to be trained, so that the neural network model can predict the predicted classification label of the masked coding unit based on the unmasked coding unit in the input sequence.

[0027] Calculate the loss value based on the predicted classification label and the actual classification label corresponding to the encoded unit after masking;

[0028] When the loss value reaches the preset convergence condition, the pre-trained neural network model is obtained.

[0029] Optionally, the neural network model includes a multilayer perceptron, which is used to classify the model based on the feature vector corresponding to the first encoding unit in the input sequence output by the neural network model, and to fine-tune the pre-trained neural network model by outputting the predicted classification label corresponding to the first encoding unit, so as to obtain a trained deep learning model.

[0030] In a first aspect of the present invention, an electronic component classification and prediction device is also provided, the device comprising:

[0031] A 3D circuit board model acquisition module is used to acquire a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein, the simplified 3D circuit board model includes solid models of electronic components; the simplified 3D circuit board model has fewer geometric features than the unsimplified 3D circuit board model.

[0032] The classification label acquisition module is used to input the simplified 3D circuit board model into a trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model; the deep learning model extracts geometric features and arrangement position information from the boundary representation data features of the electronic component entity models in the simplified 3D circuit board model, so as to predict the classification labels of the electronic component entity models in the simplified 3D circuit board model based on the geometric features and the arrangement position information; the deep learning model is a deep learning model combined with a graph neural network.

[0033] In another aspect of the present invention, a computer-readable storage medium is also provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform any of the above-described electronic component classification and prediction methods.

[0034] In another aspect of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the above-described electronic component classification and prediction methods.

[0035] Compared with related technologies, the embodiments of the present invention have at least the following advantages:

[0036] In this embodiment of the invention, a simplified three-dimensional circuit board model of the printed circuit board to be simulated for electronic heat dissipation is obtained. The simplified three-dimensional circuit board model may include physical models of electronic components. Furthermore, the simplified three-dimensional circuit board model has fewer geometric features than the unsimplified three-dimensional circuit board model. The simplified three-dimensional circuit board model is input into a trained deep learning model combined with a graph neural network. The deep learning model can output the classification labels corresponding to the physical models of electronic components in the simplified three-dimensional circuit board model. The deep learning model can extract geometric features and arrangement position information from the boundary representation data features of the physical models of electronic components in the simplified three-dimensional circuit board model. Thus, the deep learning model can predict the classification labels of the physical models of electronic components in the simplified three-dimensional circuit board model based on the geometric features and arrangement position information. In the embodiment of this invention, when performing electronic heat dissipation simulation on a printed circuit board, the deep learning model can accurately predict the geometric features and arrangement information of the electronic component entity model in the simplified three-dimensional circuit board model. Then, based on the geometric features and arrangement information, the classification labels of the electronic components in the printed circuit board are determined, which improves the identification accuracy of the electronic components. Furthermore, based on the identified classification labels, the physical property parameters required for the simulation of the electronic components can be accurately determined, thereby ensuring the accuracy and reliability of the electronic heat dissipation simulation. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0038] Figure 1 This is a flowchart illustrating the steps of an electronic component classification and prediction method provided in an embodiment of the present invention.

[0039] Figure 2 This is a structural block diagram of an electronic component classification and prediction device provided in an embodiment of the present invention;

[0040] Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0042] Currently, the identification of electronic components on printed circuit boards (PCBs) mainly relies on manual inspection, which is inefficient and prone to errors, especially when there are a large number of electronic components on the PCB.

[0043] With the development of artificial intelligence technology, it has become possible to automatically identify electronic components on printed circuit boards using deep learning models. However, most deep learning models currently classify and identify electronic components by extracting their geometric features. In electronic heat dissipation simulation, electronic components are usually simplified to reduce the differences between their geometric features. Since there are many types of electronic components and some have similar geometric shapes, it is difficult to achieve the required accuracy for electronic heat dissipation simulation by simply relying on the geometric features of electronic components.

[0044] In practical implementation, the arrangement of electronic components on printed circuit boards (PCBs) follows certain regularities (e.g., the relative positions of electronic components, the layout patterns of electronic components with different classification labels on the PCB, etc.). Identifying electronic components using their positional information on the PCB helps improve identification accuracy. Therefore, the AI ​​(Artificial Intelligence) model, i.e., the deep learning model in this embodiment of the invention, extracts the geometric features of electronic components while combining their positional information on the PCB, thus comprehensively identifying electronic components on the PCB based on multiple factors. This embodiment of the invention extracts both the geometric features and the positional information of electronic components on the PCB. By combining these geometric features and positional information, predictive identification of electronic components can be performed, enabling accurate and rapid electronic heat dissipation simulation based on the identification results. It should be noted that the geometric features are 3D geometric features (three-dimensional geometric features). Among them, geometric features can include at least the area of ​​the face of the geometric object, the length of the edge, the radius, etc., as well as the UV mesh features of the face. The topological structure mainly includes the connection relationship between faces through common edges, forming the topological relationship of the graph structure.

[0045] Reference Figure 1 The above is a flowchart of the steps of an electronic component classification and prediction method provided in an embodiment of the present invention, as follows: Figure 1 As shown, the method may specifically include the following steps:

[0046] Step 101: Obtain a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein the simplified 3D circuit board model includes solid models of electronic components; the simplified 3D circuit board model has fewer geometric features than the unsimplified 3D circuit board model.

[0047] In practice, electronic heat dissipation simulation is a method that uses computer simulation technology to predict and optimize the heat dissipation effect of printed circuit boards. In this way, the heat dissipation performance of printed circuit boards can be estimated in advance during the design stage, and potential problems of printed circuit boards can be identified, such as excessively high temperature in local areas of printed circuit boards or excessively high temperature in areas where key electronic components are located. This allows for early design optimization of printed circuit boards, ensuring the reliability and stability of products using printed circuit boards.

[0048] In practical implementation, the three-dimensional circuit board model (3D model) of the printed circuit board can include the physical models of electronic components (physical models / 3D models), such as physical models of onboard devices, heat sinks, axial fans, planar fans, blowers, and centrifugal fans.

[0049] Based on an actual simulation scenario, this invention obtains a simplified three-dimensional circuit board model of the printed circuit board to be simulated for electronic heat dissipation. The simplified three-dimensional circuit board model includes physical models of electronic components, and the simplified three-dimensional circuit board model has fewer geometric features than the unsimplified three-dimensional circuit board model. This can improve simulation efficiency, especially when there are a large number of electronic components on the printed circuit board.

[0050] In one optional embodiment of the present invention, the simplification of the three-dimensional circuit board model can specifically involve simplifying the complex geometry of the electronic component entity model in the three-dimensional circuit board model to obtain an electronic component entity model with a relatively simple geometry, thereby reducing the processing complexity of the computer on the three-dimensional circuit board model and improving processing efficiency.

[0051] Step 102: Input the simplified 3D circuit board model into the trained deep learning model to obtain the classification label of the electronic component entity model in the 3D circuit board model output by the deep learning model; the deep learning model extracts geometric features and layout position information from the boundary representation data features of the electronic component entity model in the simplified 3D circuit board model, so as to predict the classification label of the electronic component entity model in the simplified 3D circuit board model based on the geometric features and the layout position information; the deep learning model is a deep learning model combined with a graph neural network.

[0052] In this embodiment of the invention, a pre-trained deep learning model (AI model) combining a graph neural network can be pre-deployed in the computer. The deep learning model is a Graph-Transformer, which is a deep learning method that combines the advantages of graph neural networks (GNN) and Transformers (a neural network architecture based on attention mechanisms). Compared with traditional models such as CNNs (convolutional neural networks), it has significant advantages in processing non-Euclidean data (such as graph structures, temporal relationships, etc.). In other words, based on the deep learning model, the required data can be accurately extracted from the simplified three-dimensional circuit board model.

[0053] It should be noted that the simplified 3D circuit board model loses a large number of geometric features. Therefore, it is much more difficult to predict the category of electronic components based on the simplified 3D circuit board model. For this reason, in this embodiment of the invention, a deep learning model combining graph neural networks is used to predict the entity model of the simplified electronic components on the 3D circuit board by combining geometric features and layout position information. In this way, even though the simplified 3D circuit board model loses a large number of geometric features, the category of electronic components can still be accurately predicted by combining layout position information with a deep learning model combining graph neural networks.

[0054] Specifically, the simplified 3D circuit board model can be input into a deep learning model. The deep learning model can extract geometric features and arrangement information from the B-Rrep data features (boundary representation data features) of each electronic component entity model in the simplified 3D circuit board model. Then, based on the geometric features and arrangement information, it can predict the classification labels of each electronic component entity model in the simplified 3D circuit board model, such as onboard devices, heat sinks, axial fans, and planar fans.

[0055] In practical implementation, each electronic component in a printed circuit board has its own specific thermal characteristics, including but not limited to parameters such as heat generation, thermal resistance, and thermal capacity. These parameters directly affect the heat distribution and heat dissipation effect of the entire printed circuit board. Therefore, in this embodiment of the invention, by determining the classification label corresponding to each electronic component entity model, the physical property parameters required for the simulation of each electronic component entity model can be accurately determined based on the classification label. Then, electronic heat dissipation simulation of the printed circuit board is performed based on the physical property parameters of the electronic component entity model, ensuring the accuracy and reliability of the electronic heat dissipation simulation.

[0056] In the above-mentioned electronic component classification prediction method, a simplified three-dimensional circuit board model of the printed circuit board to be simulated for electronic heat dissipation is obtained. The simplified three-dimensional circuit board model may include physical models of electronic components. Furthermore, the simplified three-dimensional circuit board model has fewer geometric features than the unsimplified three-dimensional circuit board model. The simplified three-dimensional circuit board model is input into a trained deep learning model combined with a graph neural network. The deep learning model can output the classification labels corresponding to the physical models of electronic components in the simplified three-dimensional circuit board model. The deep learning model can extract geometric features and arrangement position information from the boundary representation data features of the physical models of electronic components in the simplified three-dimensional circuit board model. Thus, the deep learning model can predict the classification labels of the physical models of electronic components in the simplified three-dimensional circuit board model based on the geometric features and arrangement position information. In the embodiment of this invention, when performing electronic heat dissipation simulation on a printed circuit board, the deep learning model can accurately predict the geometric features and arrangement information of the electronic component entity model in the simplified three-dimensional circuit board model. Then, based on the geometric features and arrangement information, the classification labels of the electronic components in the printed circuit board are determined, which improves the identification accuracy of the electronic components. Furthermore, based on the identified classification labels, the physical property parameters required for the simulation of the electronic components can be accurately determined, thereby ensuring the accuracy and reliability of the electronic heat dissipation simulation.

[0057] In one embodiment of the present invention, the deep learning model may include an encoder and a neural network model. Step 102 involves inputting the simplified three-dimensional circuit board model into the trained deep learning model to obtain the classification labels of the electronic component entity models in the three-dimensional circuit board model output by the deep learning model, including:

[0058] The encoder is used to extract the geometric features of the electronic component entity model in the simplified 3D circuit board model; wherein, the geometric features include geometric information and topological structure;

[0059] The simplified 3D circuit board model and the geometric features are input into the neural network model to obtain the classification labels of the electronic component entity models in the simplified 3D circuit board model output by the neural network model.

[0060] The geometric features of an electronic component physical model can include geometric information and topology. Specifically, geometric information refers to the structural information of the electronic component physical model, such as its shape, size, and positional relationships. Topology refers to the interrelationships and connection methods of the electronic component physical model in the printed circuit board. The topology describes the interrelationships between geometric elements such as points, lines, and surfaces of the electronic component physical model, as well as their relative storage addresses in memory space. These interrelationships determine how the electronic component physical model is constructed, modified, and rendered in CAD (Computer-Aided Design).

[0061] In this embodiment of the invention, the deep learning model may include a pre-trained generator encoder (GNN (Graph Neural Network) - AutoEncoder) and a neural network model (Transformer). The encoder can predict the geometric features (including geometric information and topological structure) of the electronic component entity model, and the neural network model can predict the classification label of the electronic component entity model based on the geometric features and the simplified 3D circuit board model. During prediction, the neural network model can predict the proximity matrix between the electronic component entity models, and the density matrix can determine whether there is a connection relationship between the electronic component entity models.

[0062] In one embodiment of the present invention, before step 102, in which the simplified three-dimensional circuit board model is input into the trained deep learning model to obtain the classification labels of the electronic component entity models in the three-dimensional circuit board model output by the deep learning model, the method may further include:

[0063] Obtain a sample electronic component entity model; wherein, the sample electronic component entity model has corresponding actual geometric features, the actual geometric features including actual geometric information and actual topological structure;

[0064] The encoder to be trained is self-supervised using the physical model of the sample electronic components.

[0065] When the encoder reaches the preset training conditions after self-supervised training, the encoder that has completed training is obtained.

[0066] In this embodiment of the invention, geometric features can be extracted from the physical model of electronic components on a printed circuit board using an encoder. The encoder capable of extracting geometric features can be trained. Specifically, a suitable pre-trained generative model (i.e., the encoder to be trained, such as a diffusion-based architecture, specifically a GNN-AutoEncoder model structure) can be selected. The encoder of the pre-trained generative model is capable of extracting geometric features from the physical model of electronic components.

[0067] In one example of this invention, self-supervised pre-training is applied to acquire B-Rrep data of individual electronic component 3D models (sample electronic component solid models) from a large number of open-source CAD files. The B-Rrep data defines the boundaries between internal and external points by representing the solid model as a set of connected surface elements. The B-Rrep data includes actual geometric features. Then, the generated encoder to be trained is applied to perform self-supervised pre-training based on the sample electronic component solid models. During training, the encoder learns the geometric information and topological structure of the electronic component solid models, thereby improving the effectiveness of the encoder's pre-extraction of geometric features. This enables the encoder to accurately extract key features of the electronic component solid models and obtain feature vectors, which are then used as the geometric features of the corresponding electronic components. The encoder is considered trained when it meets preset training conditions, such as reaching a preset number of training iterations or the error between the encoder's output geometric features and the actual geometric features being within a preset error range.

[0068] This invention obtains the CAD geometric features of different electronic component entity models in CAD files as input to the encoder, and directly trains and learns the geometric features of each electronic component entity model, making full use of the geometric features of electronic components and effectively improving the recognition accuracy of geometric features of electronic components.

[0069] In one embodiment of the present invention, before step 102, in which the simplified three-dimensional circuit board model is input into the trained deep learning model to obtain the classification labels of the electronic component entity models in the three-dimensional circuit board model output by the deep learning model, the method may further include:

[0070] Obtain a sample 3D circuit board model of a sample printed circuit board; wherein, the sample 3D circuit board model includes sample electronic component entity models, and the sample electronic component entity models have corresponding actual classification labels;

[0071] Calculate the center-to-center distance between the center points of the sample electronic component solid model;

[0072] Calculate the affinity matrix between the sample electronic component entity models based on the distance between the center points;

[0073] The connection relationships between the sample electronic components are determined based on the affinity matrix between the sample electronic component entity models and a preset threshold.

[0074] The physical model of the sample electronic components is converted into a coding unit;

[0075] For each coding unit of the sample printed circuit board, sub-graph data is constructed by combining the geometric feature vectors of the coding unit and other coding units that are connected in reverse order of the distance between the center points of the coding units, and all the sub-graph data of the sample printed circuit board are combined into an input sequence.

[0076] The input sequence is used to pre-train the neural network model to be trained.

[0077] In this embodiment of the invention, an input sequence is constructed based on the distribution of electronic component entity models on the printed circuit board. Specifically, based on the distribution location information of the electronic component entity models on the printed circuit board, the distance between the center points of each electronic component entity model is calculated. A proximity matrix between the electronic component entity models is obtained through the center point distance calculation. Based on the proximity matrix and a preset threshold, it is determined whether there is a connection relationship between the electronic component entity models. If there is a connection relationship, it indicates that there are edges between the electronic component entity models, and a subgraph data is constructed. Then, based on the subgraph data, each electronic component entity model is converted into an encoding unit (token) for pre-training of a Transformer (neural network model). Specifically, each electronic component entity model is used as the starting token, and the tokens are used according to the relationship between the electronic component entity model and other electronic components. If a connection exists between the entity models, and if so, then other electronic component entity models are used as tokens to construct subgraph data in reverse order of center point distance. The subgraph data of all electronic component entity models then form the input sequence of the Transformer. This input sequence consists of several geometric feature vectors, where each node represents a geometric feature of an electronic component. The geometric feature is represented by a geometric feature vector. The input sequence reflects the arrangement and position information of the electronic component entity models. Then, the neural network model to be trained using the input sequence can be trained to obtain a model capable of predicting the center point distance between electronic component entity models. Based on the center point distance, a proximity matrix is ​​predicted, and then based on the proximity matrix and a preset threshold, the existence of a connection between the electronic component entity models is predicted.

[0078] In one embodiment of the present invention, determining the connection relationship between the sample electronic components based on the proximity matrix between the sample electronic component entity models and a preset threshold may include:

[0079] When the affinity matrix between the sample electronic component entity models is greater than the preset threshold, it is determined that there is a connection relationship between the sample electronic component entity models.

[0080] In this embodiment of the invention, if the affinity matrix between sample electronic component entity models is greater than a preset threshold, it can be determined that there is a connection relationship between the sample electronic component entity models; conversely, if the affinity matrix between sample electronic component entity models is less than or equal to the preset threshold, it can be determined that there is no connection relationship between the sample electronic component entity models.

[0081] In practical implementation, center point distance and proximity are inversely related. The smaller the center point distance between electronic components, the greater the proximity in the proximity matrix; conversely, the larger the center point distance, the smaller the proximity in the proximity matrix. For example, the proximity values ​​in the proximity matrix are compared with a preset threshold. If the value is less than the preset threshold, it is set to 0; if it is greater than or equal to the preset threshold, it is set to 1. This results in a proximity matrix containing only 0s and 1s, which serves as the adjacency matrix. 1 indicates a connection between sample electronic component entity models, and 0 indicates no connection between sample electronic component entity models. Thus, connection relationships can be described based on the adjacency matrix.

[0082] In some embodiments of the present invention, the calculation process of the affinity matrix may include:

[0083]

[0084] in, These are the coordinates of the center point of node i. These are the coordinates of the center point of node j. It is the distance between the center points of node i and node j.

[0085] The intimacy value (numerical value) in the intimacy matrix can be calculated in the following way:

[0086]

[0087] in, This represents the affinity value in the affinity matrix. Generally, the greater the distance between the center points, the lower the probability of a connection between the nodes (electronic components). The calculated affinity values ​​can then be combined to form an affinity matrix.

[0088] In practice, the connection relationships of electronic components in a 3D circuit board model are usually determined based on electrical connections (such as wires and solder joints). However, in real-world applications, there may be non-electrical connections between electronic components (such as thermal coupling or physical proximity). These connections are not explicitly shown in the 3D circuit board model, especially in simplified 3D circuit board models, where some electrical connections may be lost.

[0089] In this embodiment of the invention, different types of connection relationships are required for different simulation tasks. For example, in thermal simulation tasks, additional thermal coupling information of electronic components is needed, and this information may be lost in the simplified 3D circuit board model. Therefore, this embodiment of the invention proposes to determine whether there is a connection relationship between electronic components based on the center point distance. Especially when the simplified 3D circuit board model fails to show non-electrical connection relationships (such as thermal coupling, physical proximity), the center point distance can serve as a supplementary means to ensure that important connection relationships in the 3D circuit board model are not ignored. For example, if two components are not directly connected electrically, but their center points are close, they are likely to exhibit a strong thermal coupling relationship in electronic heat dissipation simulation. In addition, this embodiment of the invention can also meet the simulation needs of multiple simulation tasks by adjusting preset thresholds according to different simulation task requirements (such as different simulation task types, simulation accuracy requirements, simulation speed, types of electronic components in the 3D circuit board model, working environment, etc.).

[0090] For example, in high-precision thermal simulation, the preset threshold is lowered to capture more thermal coupling relationships; in fast electrical simulation, the preset threshold is increased to reduce the number of connections and improve simulation speed; for high-power components or high-temperature environments, the preset threshold is further lowered to enhance the accuracy of thermal management. Thus, by flexibly adjusting the threshold, this embodiment of the invention can adapt to diverse simulation needs and ensure the accuracy and efficiency of simulation results.

[0091] In one embodiment of the present invention, pre-training the neural network model to be trained using the input sequence may include:

[0092] The encoding units in the input sequence are randomly masked to obtain the masked input sequence;

[0093] The masked input sequence is input into the neural network model to be trained, so that the neural network model can predict the predicted classification label of the masked coding unit based on the unmasked coding unit in the input sequence.

[0094] Calculate the loss value based on the predicted classification label and the actual classification label corresponding to the encoded unit after masking;

[0095] When the loss value reaches the preset convergence condition, the pre-trained neural network model is obtained.

[0096] In this embodiment of the invention, masking prediction technology is used for model training during the neural network model training process. Specifically, by randomly masking a portion of the input sequence's encoding units (tokens), the neural network model needs to predict the predicted classification label corresponding to the masked data based on the unmasked portion (unmasked portion), i.e., the predicted classification label of the masked encoding unit. Then, the loss value is calculated based on the predicted classification label and the actual classification label corresponding to the masked encoding unit. If the loss value reaches a preset convergence condition, for example, when the loss value is within a preset loss value range, the trained neural network model can be obtained. In addition, the adjacency matrix (connection relationship) is also incorporated during neural network model training, enabling the neural network model to learn more representative and universal feature representations.

[0097] In one embodiment of the present invention, the neural network model includes a multilayer perceptron, which is used to classify according to the feature vector corresponding to the first coding unit in the input sequence output by the neural network model, so as to output a predicted classification label corresponding to the first coding unit.

[0098] Among them, Transformer (neural network model) can also include MLP (Multilayer Perceptron) network. MLP network is a feedforward neural network composed of multiple layers of neurons, which can learn complex nonlinear mapping relationships.

[0099] In this embodiment of the invention, after the input sequence is processed by a pre-trained transformer model, the feature vector of the first token can be obtained. The MLP is then applied to fine-tune the classification of the feature vector to obtain the final classification and recognition result of the electronic component entity model, i.e., the classification label.

[0100] B-Rrep data, or Boundary Representation Data, is a geometric shape representation method widely used in solid modeling and computer-aided design (CAD). B-Rrep data defines the boundaries between internal and external points through connected surface elements and records in detail the geometric information of all geometric elements constituting the shape and their interconnected topological structure. Therefore, B-Rrep data is suitable for representing objects with complex geometric shapes and topological structures, such as electronic components on a PCB board. Based on the geometric information and topological structure recorded in the B-Rrep data, the performance and behavior of electronic components on an actual PCB board can be accurately simulated. This invention extracts the B-Rrep data of the solid models of various electronic components on a PCB board, constructs subgraph data based on the spatial distribution (layout position information) of the components in the solid models, and predicts the classification labels of the electronic components based on an AI model (deep learning model) using GNN and Transformer architectures, thereby achieving the classification of electronic components.

[0101] To enable those skilled in the art to better understand the embodiments of the present invention, a specific example is provided below for illustration. The process of predicting electronic components in a printed circuit board based on a deep learning model according to the embodiments of the present invention may include the following steps:

[0102] Step 1: Pre-extract features using a pre-trained generative model encoder.

[0103] Choosing a suitable pre-trained generative model (e.g., a diffusion-based architecture, such as the GNN-AutoEncoder model structure) enables its encoder to extract geometric features from PCB 3D models (solid models of electronic components / 3D models of electronic components). In practical applications, a large amount of open-source, independent B-Rrep data for 3D electronic component models is collected and input into the pre-trained generative model for self-supervised pre-training. During self-supervised pre-training, the B-Rrep reconstruction task enhances the encoder's ability to extract geometric information and topological structure of electronic components, thereby improving the effectiveness of the encoder's pre-extracted geometric features. Specifically, Input B-Rrep: B-Rrep data of the 3D model of electronic components; GNN tokenizer: converts the B-Rrep data of each component on the PCB board into graph structure data according to the connection relationship, treats each component as a tokenizer of the transformer, and then arranges them in reverse order according to the distance between the components to form a sequence as the input of the transformer; Surface_Conv and Curve_Conv: sample the faces and edges of the B-Rrep data using UV coordinate grids to obtain point coordinates, and perform convolution operations on the point coordinates; Node Update: updates the node information in the B-Rrep data according to the message passing mechanism of the graph neural network; Edge Update: updates the edge information in the B-Rrep data according to the message passing mechanism of the graph neural network; GNN: Graph Neural Network in deep learning; Reconstruct: reconstructs or reconstructs the B-Rrep data using the learned node or graph representations.

[0104] In a preferred embodiment of the present invention, the pre-trained generative model is a Graph-Transformer deep learning model. Graph-Transformer is a deep learning method that combines the advantages of graph neural networks and Transformers. Specifically, it aggregates neighbor information through message passing mechanisms (such as GNNs), preserves the topological structure of electronic components, and captures the dependency relationship between any two electronic components in the graph through the self-attention of Transformers. Therefore, it can obtain geometric information and topological structure more accurately than models such as CNNs in complex relationship prediction tasks, especially in scenarios where the data itself has an explicit relational structure, such as simplified three-dimensional circuit board models with a large number of electronic components.

[0105] Step 2: Constructing sub-graph data for electronic components on the PCB board:

[0106] Taking each electronic component on the PCB board as the center, the distance between the center points of other electronic components is calculated. The proximity matrix between electronic components is obtained by calculating the center point distance. Based on the proximity matrix and a preset threshold, it is determined whether there is a connection relationship between electronic components. The feature vector of electronic components (nodes) is represented by geometric feature vectors. If there is a connection relationship, there is an edge between the nodes, thus constructing a graph structure, which reflects the geometric features and position information of electronic components.

[0107] Step 3: Transformer pre-training:

[0108] Each electronic component is used as the starting token. Based on the existence of connections (determined by the adjacency matrix) and in reverse order of distance from the center point, other electronic components are used as tokens to construct the Transformer's input sequence. This input sequence consists of multiple subgraphs, each representing a sequence of feature vectors from the target component entity / electronic component entity model (the first token) and neighboring component entity models / (surrounding component entity models) (other tokens). This input sequence is then concatenated with positional information to form the Transformer's input representation. This representation considers both the structural and topological features of each electronic component entity model and its spatial distribution on the PCB board, helping the Transformer model better understand the relationships between different electronic components during processing. Masking prediction is employed during Transformer model training. Masking prediction technology uses random masked parts of the input representation. The Transformer model needs to predict the masked data (masked tokens) based on the unmasked parts (unmasked tokens). In addition, the Transformer can also predict the adjacency matrix of electronic components, thus enabling the Transformer model to learn more representative and general feature representations.

[0109] Step 4: Fine-tuning Transformer classification training:

[0110] The Transformer model is fine-tuned for classification training to achieve the recognition of classification labels for electronic components. The first token of the input sequence of the Transformer represents an electronic component. Based on the classification label of each electronic component, an MLP network is added to the output of the Transformer. The MLP network can perform classification fine-tuning training on the first token in each subgraph data in the input sequence to obtain the final recognition result (classification label) of the electronic component.

[0111] In summary, the embodiments of the present invention, based on graph neural networks and Transformer self-attention mechanism, capture the geometric features and spatial distribution information (positional arrangement information) of each component entity, solving the difficulty of identifying electronic component assemblies. Furthermore, the physical property parameters required for simulation can be assigned based on the classification labels of the identified electronic components, thereby efficiently and accurately realizing electronic heat dissipation simulation of printed circuit boards.

[0112] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0113] Reference Figure 2 This is a structural block diagram of an electronic component classification and prediction device provided in an embodiment of the present invention, such as... Figure 2 As shown, the device 20 may specifically include the following modules:

[0114] The 3D circuit board model acquisition module 201 is used to acquire a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein, the simplified 3D circuit board model includes physical models of electronic components.

[0115] The classification label acquisition module 202 is used to input the simplified 3D circuit board model into the trained deep learning model to obtain the classification label of the electronic component entity model in the 3D circuit board model output by the deep learning model; the deep learning model extracts geometric features and arrangement position information from the boundary representation data features of the electronic component entity model in the simplified 3D circuit board model, so as to predict the classification label of the electronic component entity model in the simplified 3D circuit board model based on the geometric features and the arrangement position information; the deep learning model is a deep learning model combined with a graph neural network.

[0116] In one embodiment of the present invention, the classification label acquisition module 202 is used for:

[0117] The encoder is used to extract the geometric features of the electronic component entity model in the simplified 3D circuit board model; wherein, the geometric features include geometric information and topological structure;

[0118] The simplified 3D circuit board model and the geometric features are input into the neural network model to obtain the classification labels of the electronic component entity models in the simplified 3D circuit board model output by the neural network model.

[0119] In one embodiment of the present invention, the apparatus further includes: a model training module, used for:

[0120] Obtain a sample electronic component entity model; wherein, the sample electronic component entity model has corresponding actual geometric features, the actual geometric features including actual geometric information and actual topological structure;

[0121] The encoder to be trained is self-supervised using the physical model of the sample electronic components.

[0122] When the encoder reaches the preset training conditions after self-supervised training, the encoder that has completed training is obtained.

[0123] In one embodiment of the present invention, the apparatus further includes: a model training module, used for:

[0124] Obtain a sample 3D circuit board model of a sample printed circuit board; wherein, the sample 3D circuit board model includes sample electronic component entity models, and the sample electronic component entity models have corresponding actual classification labels;

[0125] Calculate the center-to-center distance between the center points of the sample electronic component solid model;

[0126] Calculate the affinity matrix between the sample electronic component entity models based on the distance between the center points;

[0127] The connection relationships between the sample electronic components are determined based on the affinity matrix between the sample electronic component entity models and a preset threshold.

[0128] The physical model of the sample electronic components is converted into a coding unit;

[0129] For each coding unit of the sample printed circuit board, in reverse order of the distance between the center points of the coding units, the geometric feature vectors corresponding to the coding units and other coding units that have a connection relationship are used to construct sub-graph data, and all the sub-graph data of the sample printed circuit board are combined into an input sequence;

[0130] The input sequence is used to pre-train the neural network model to be trained.

[0131] In one embodiment of the present invention, the model training module is configured to:

[0132] When the affinity matrix between the sample electronic component entity models is greater than the preset threshold, it is determined that there is a connection relationship between the sample electronic component entity models.

[0133] In one embodiment of the present invention, the model training module is configured to:

[0134] The encoding units in the input sequence are randomly masked to obtain the masked input sequence;

[0135] The masked input sequence is input into the neural network model to be trained, so that the neural network model can predict the predicted classification label of the masked coding unit based on the unmasked coding unit in the input sequence.

[0136] Calculate the loss value based on the predicted classification label and the actual classification label corresponding to the encoded unit after masking;

[0137] When the loss value reaches the preset convergence condition, the pre-trained neural network model is obtained.

[0138] In one embodiment of the present invention, the neural network model includes a multilayer perceptron, which is used to classify the neural network model based on the feature vector corresponding to the first encoding unit in the input sequence output by the neural network model, and to fine-tune the pre-trained neural network model by outputting the predicted classification label corresponding to the first encoding unit, so as to obtain a trained deep learning model.

[0139] In this embodiment of the invention, a simplified three-dimensional circuit board model of the printed circuit board to be simulated for electronic heat dissipation is obtained. The simplified three-dimensional circuit board model may include physical models of electronic components. Furthermore, the simplified three-dimensional circuit board model has fewer geometric features than the unsimplified three-dimensional circuit board model. The simplified three-dimensional circuit board model is input into a trained deep learning model combined with a graph neural network. The deep learning model can output the classification labels corresponding to the physical models of electronic components in the simplified three-dimensional circuit board model. The deep learning model can extract geometric features and arrangement position information from the boundary representation data features of the physical models of electronic components in the simplified three-dimensional circuit board model. Thus, the deep learning model can predict the classification labels of the physical models of electronic components in the simplified three-dimensional circuit board model based on the geometric features and arrangement position information. In the embodiment of this invention, when performing electronic heat dissipation simulation on a printed circuit board, the deep learning model can accurately predict the geometric features and arrangement information of the electronic component entity model in the simplified three-dimensional circuit board model. Then, based on the geometric features and arrangement information, the classification labels of the electronic components in the printed circuit board are determined, which improves the identification accuracy of the electronic components. Furthermore, based on the identified classification labels, the physical property parameters required for the simulation of the electronic components can be accurately determined, thereby ensuring the accuracy and reliability of the electronic heat dissipation simulation.

[0140] The above-described apparatus embodiments are basically similar to the method embodiments, so they are described in a relatively simple manner. For relevant details, please refer to the description of the method embodiments.

[0141] This invention also provides an electronic device, such as... Figure 3 As shown, it includes a processor 301, a communication interface 302, a memory 303, and a communication bus 304, wherein the processor 301, the communication interface 302, and the memory 303 communicate with each other through the communication bus 304.

[0142] Memory 303 is used to store computer programs;

[0143] When the processor 301 executes the program stored in the memory 303, it implements the electronic component classification and prediction method described in any of the above embodiments, specifically including:

[0144] Obtain a simplified three-dimensional circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein, the simplified three-dimensional circuit board model includes solid models of electronic components;

[0145] The simplified 3D circuit board model is input into the trained deep learning model to obtain the classification label of the electronic component entity model in the 3D circuit board model output by the deep learning model; the deep learning model extracts geometric features and layout position information from the boundary representation data features of the electronic component entity model in the simplified 3D circuit board model, so as to predict the classification label of the electronic component entity model in the simplified 3D circuit board model based on the geometric features and the layout position information.

[0146] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0147] The communication interface is used for communication between the aforementioned terminal and other devices.

[0148] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0149] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0150] In another embodiment of the present invention, a computer-readable storage medium is also provided, which stores instructions that, when executed on a computer, cause the computer to perform any of the electronic component classification and prediction methods described in the above embodiments.

[0151] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the electronic component classification and prediction methods described in the above embodiments.

[0152] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state disk (SSD)).

[0153] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0154] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0155] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. An electronic component classification prediction method characterized by comprising: The method includes: Obtain a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein the simplified 3D circuit board model includes solid models of electronic components, and the simplified 3D circuit board model has fewer geometric features than the unsimplified 3D circuit board model. The simplified 3D circuit board model is input into a trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model. The deep learning model extracts geometric features and arrangement position information from the B-Rrep boundary representation data features of the electronic component entity models in the simplified 3D circuit board model, and predicts the classification labels of the electronic component entity models in the simplified 3D circuit board model based on the geometric features and arrangement position information. The deep learning model is a deep learning model that combines a Transformer neural network model with a graph neural network model. The deep learning model includes an encoder and a neural network model. Before inputting the simplified 3D circuit board model into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, the method further includes: Obtain a sample 3D circuit board model of a sample printed circuit board; wherein, the sample 3D circuit board model includes sample electronic component entity models, and the sample electronic component entity models have corresponding actual classification labels; Calculate the center-to-center distance between the center points of the sample electronic component solid model; Calculate the affinity matrix between the sample electronic component entity models based on the distance between the center points; Determining the connection relationship between the sample electronic components based on the affinity matrix between the sample electronic component entity models and a preset threshold includes: determining the existence of a connection relationship between the sample electronic component entity models when the affinity matrix between the sample electronic component entity models is greater than the preset threshold; the preset threshold is adjusted according to different simulation task requirements; The physical model of the sample electronic components is converted into a coding unit; For each coding unit of the sample printed circuit board, in reverse order of the distance between the center points of the coding units, the geometric feature vectors corresponding to the coding units and other coding units that have a connection relationship are used to construct sub-graph data, and all the sub-graph data of the sample printed circuit board are combined into an input sequence; Pre-training the neural network model to be trained using the input sequence includes: The encoding units in the input sequence are randomly masked to obtain the masked input sequence; The masked input sequence is input into the neural network model to be trained, so that the neural network model can predict the predicted classification label of the masked coding unit based on the unmasked coding unit in the input sequence. Calculate the loss value based on the predicted classification label and the actual classification label corresponding to the encoded unit after masking; When the loss value reaches the preset convergence condition, the pre-trained neural network model is obtained.

2. The method of claim 1, wherein, The simplified 3D circuit board model is input into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, including: The encoder is used to extract the geometric features of the electronic component entity model in the simplified 3D circuit board model; wherein, the geometric features include geometric information and topological structure; The simplified 3D circuit board model and the geometric features are input into the neural network model to obtain the classification labels of the electronic component entity models in the simplified 3D circuit board model output by the neural network model.

3. The method of claim 2, wherein, Before inputting the simplified 3D circuit board model into the trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model, the method further includes: Obtain a sample electronic component entity model; wherein, the sample electronic component entity model has corresponding actual geometric features, the actual geometric features including actual geometric information and actual topological structure; The encoder to be trained is self-supervised using the physical model of the sample electronic components. When the encoder reaches the preset training conditions after self-supervised training, the encoder that has completed training is obtained.

4. The method of claim 1, wherein, The neural network model includes a multilayer perceptron, which is used to classify the model based on the feature vector corresponding to the first encoding unit in the input sequence output by the neural network model, and to fine-tune the pre-trained neural network model by outputting the predicted classification label corresponding to the first encoding unit, so as to obtain a trained deep learning model.

5. An electronic component classification prediction device characterized by comprising: The device includes: A 3D circuit board model acquisition module is used to acquire a simplified 3D circuit board model of the printed circuit board to be simulated for electronic heat dissipation; wherein, the simplified 3D circuit board model includes solid models of electronic components; the simplified 3D circuit board model has fewer geometric features than the unsimplified 3D circuit board model. The classification label acquisition module is used to input the simplified 3D circuit board model into a trained deep learning model to obtain the classification labels of the electronic component entity models in the 3D circuit board model output by the deep learning model. The deep learning model extracts geometric features and arrangement position information from the B-Rrep boundary representation data features of the electronic component entity models in the simplified 3D circuit board model, so as to predict the classification labels of the electronic component entity models in the simplified 3D circuit board model based on the geometric features and the arrangement position information. The deep learning model is a deep learning model that combines a Transformer neural network model with a graph neural network model. The deep learning model includes an encoder and a neural network model. The device further includes: a model training module, used for: Obtain a sample 3D circuit board model of a sample printed circuit board; wherein, the sample 3D circuit board model includes sample electronic component entity models, and the sample electronic component entity models have corresponding actual classification labels; Calculate the center-to-center distance between the center points of the sample electronic component solid model; Calculate the affinity matrix between the sample electronic component entity models based on the distance between the center points; Determining the connection relationship between the sample electronic components based on the affinity matrix between the sample electronic component entity models and a preset threshold includes: determining the existence of a connection relationship between the sample electronic component entity models when the affinity matrix between the sample electronic component entity models is greater than the preset threshold; the preset threshold is adjusted according to different simulation task requirements; The physical model of the sample electronic components is converted into a coding unit; For each coding unit of the sample printed circuit board, in reverse order of the distance between the center points of the coding units, the geometric feature vectors corresponding to the coding units and other coding units that have a connection relationship are used to construct sub-graph data, and all the sub-graph data of the sample printed circuit board are combined into an input sequence; Pre-training the neural network model to be trained using the input sequence includes: The encoding units in the input sequence are randomly masked to obtain the masked input sequence; The masked input sequence is input into the neural network model to be trained, so that the neural network model can predict the predicted classification label of the masked coding unit based on the unmasked coding unit in the input sequence. Calculate the loss value based on the predicted classification label and the actual classification label corresponding to the encoded unit after masking; When the loss value reaches the preset convergence condition, the pre-trained neural network model is obtained.

6. An electronic device, comprising: It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method of any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that The program, when executed by the processor, implements the method of any one of claims 1-4.

Images