DE102021211948A1 - Computer-implemented method and system for determining a printed circuit board layout and/or information for placing groups of components on a printed circuit board - Google Patents

Abstract

The invention relates to a computer-implemented method and system for determining a layout (10) of a printed circuit (1) and/or information for placing component groups (CG1-CGn) on a printed circuit (1). The invention also relates to a computer-implemented method for providing a first machine learning classification algorithm (A1) for determining a degree of similarity (M) of first component and/or layout data (D3) to second component group and/or layout data (D4 ), a computer program and a computer-readable data carrier.

Description

The invention relates to a computer-implemented method for determining a layout of a printed circuit board and/or information for placing groups of components on a printed circuit board.

The invention further relates to a computer-implemented method for providing a first machine learning classification algorithm for determining a degree of similarity of first component and/or layout data to second component group and/or layout data stored in a database.

The invention further relates to a system for determining a layout of a printed circuit and/or information for placing components on a printed circuit, a computer program and a computer-readable data carrier.

US2013091481A discloses a computer-implemented method for reusing layout patterns or layout constraints by identifying sub-circuits with identical or similar schematic structure based on a topology comparison strategy.

The selected sub-circuit is transformed into a topology that represents the relative positions between the instances of the selected sub-circuit. Based on the topology, one or more sub-circuits with identical or similar topology are detected and identified in a predefined area of a circuit diagram. Accordingly, the layout or layout constraint of the selected sub-circuit is copied and linked to each of the identified sub-circuits. In addition, the identified sub-circuits are listed on a user interface with notations after their identification.

Further revealed US2019042684A1 a method for schematic controlled analog circuit layout automation. The method consists of a user providing input to the schematic of an analog circuit represented in a circuit layout tool to group components into component groups.

In response to the grouping of components, the circuit layout tool can automatically create connections between components in each group in accordance with the circuit diagram. Based on user input, the groups can be moved to desired locations within a physical layout map. After that, the circuit layout tool can automatically create connections between each group according to the circuit diagram. A physical layout plan can then be provided in response to the full generation of connections between the groups.

A user-friendly and efficient identification of layouts and/or groups of components similar to those used in the past for determining a new layout of a printed circuit and automatic placement of created components on the printed circuit is only possible to a limited extent by the publications mentioned above.

It is therefore an object of the invention to provide an improved method and system for determining a layout of a printed circuit board and/or information for placing groups of components on a printed circuit board.

The object is achieved with a computer-implemented method for determining a layout of a printed circuit board and/or information for placing component groups on a printed circuit board with the features of patent claim 1.

In addition, the object is achieved with a computer-implemented method for providing a first machine learning classification algorithm for determining a degree of similarity of first component and/or layout data to second component group and/or layout data stored in a database with the features of claim 12.

Furthermore, the object is achieved with a system for determining a layout of a printed circuit and/or information for placing components on a printed circuit with the features of claim 13.

Furthermore, the object is achieved with a computer program having the features of patent claim 14 and a computer-readable data carrier having the features of patent claim 15.

Disclosure of Invention

The present invention provides a computer-implemented method for determining a printed circuit board layout and/or information for placing groups of components on a printed circuit board.

The method includes receiving data of a circuit diagram of the printed circuit or identification data of a circuit diagram of the printed circuit stored in a first database and retrieving the layout of the printed circuit associated with the identification data in the first database.

The method further comprises extracting first component group and/or layout data from the circuit diagram data of the printed circuit board layout.

The method further includes determining a degree of similarity of the extracted first component group and/or layout data to second component group and/or layout data stored in the first database and/or in a second database using a first machine learning classification algorithm.

The method also includes outputting a data set comprising a predetermined number of second component groups and/or layouts which have a predetermined degree of similarity.

The present invention further creates a computer-implemented method for providing a first machine learning classification algorithm for determining a degree of similarity of first component and/or layout data to second component group and/or layout data stored in a database.

The method includes receiving first training data of extracted component and/or layout data of a printed circuit and receiving second training data of a similarity measure of the extracted component and/or layout data to second component group and/or layout data of the printed circuit stored in a database.

In addition, the method includes training the first machine learning classification algorithm using an optimization algorithm, which calculates an extreme value of a loss function.

The present invention also creates a system for determining a layout of a printed circuit and/or information for placing components on a printed circuit, having a computing device which is set up to carry out the method according to the invention.

The present invention also creates a computer program with program code to carry out the method according to the invention when the computer program is run on a computer and a computer-readable data carrier with program code of a computer program to carry out the method according to the invention when the computer program is run on a computer.

An idea of the present invention is to provide a method for representing a PCB layout using a similarity algorithm to calculate the similarity between a new proposed layout and previous layouts.

A user interface is also provided through which a user can upload the new PCB layout and receive recommendations from the system for similar layouts that satisfy constraints such as component area, cooling area, etc.

Identifying similar layouts is accomplished by providing a methodology for extracting features from PCB layouts. The extracted features are provided as input for the similarity search algorithm.

The output of the similarity search algorithm is displayed to a user who can specify a layout as relevant or not relevant. The algorithm is further optimized with user input and learns from user feedback.

Advantageous embodiments and developments result from the dependent claims and from the description with reference to the figures.

A PCB layout is defined as an ASCII file consisting of information about design, placement, connectors and the like. The ASCII file also contains component-specific information. The most relevant of these are the package name, the placement information for a component, and the mesh information that defines the connections between the components.

A component is one of the basic units of a circuit. In a PCB layout, a component is identified with a package name and reference designation. The package name represents the properties of a component. Therefore, components within a circuit can have the same package name if they share the same properties.

To achieve the overall functionality of an electronic device, the circuit consists of multiple modules or groups of components that are interconnected based on domain constraints.

Component groups are identified by a main component, which is typically an integrated circuit. This main component, connected to other components (referred to as peripheral components) with some naming convention, forms a component group.

The component placement information consists of x, y coordinates, angle and placement side. Because a circuit board has an A side and a B side, a component can be placed on either side.

According to a preferred development, it is provided that component data, component pin data, network data, layout data and/or layout metadata are extracted from the data, in particular an ASCII file, of the circuit diagram of the printed circuit. Thus, component groups and/or a layout can be extracted from the ASCII file in an advantageous manner.

According to a further preferred development, it is provided that regular or irregular shapes of an area of the printed circuit covered by the component group or the layout are converted into a polygon represented by a sequence of points in a 2D coordinate system. In this way, an exact area calculation of the area of the printed circuit covered by the component groups can be made possible in an advantageous manner.

According to a further preferred development, it is provided that the 2D coordinate system comprises a large number of square fields, in particular with a side length of between 0.001 and 0.1 mm, with the area of the printed circuit covered by the component group or the layout being determined for estimating which of the square fields are arranged inside the polygon. The small dimensioning of the fields thus allows a fine resolution in the estimation of the area of the printed circuit covered by the component group or the layout.

According to a further preferred development, an outer boundary is created for each component group and/or each layout, with a ray-casting algorithm being used for each square field to determine whether a center point of the field is located within the polygon. This determination enables one to determine whether the square field is located substantially inside or outside the polygon.

According to a further preferred development, it is provided that the degree of similarity of the extracted first component group and/or layout data to the second component group and/or layout data stored in the second database is determined by calculating a distance measure, in particular a weighted Euclidean distance. Thus, the similarity of the extracted first component group and/or layout data to the second component group and/or layout data stored in the second database can be calculated exactly using geometric aspects.

According to a further preferred development, it is provided that the data set output by the first machine learning classification algorithm is received by a second machine learning classification algorithm, which outputs a relevance score for each of the predetermined number of component groups and/or layouts, on the basis of which a weighting of the first classification algorithm is adapted to machine learning. Outputting the relevance score thus enables component groups and/or layouts to be listed according to their relevance.

According to a further preferred development, it is provided that a user selects third component group data from the extracted first component group data, with fourth component group data being determined for determining the degree of similarity of the extracted first component group data to second component group and/or layout data stored in the second database main component is identical to a main component of the third component group data selected by the user.

By restricting the determination of the degree of similarity to component groups whose main component is identical to the main component of the third component group data selected by the user, a preselection can be made in an advantageous manner, which enables a more exact similarity determination of the subsequent comparison method.

According to a further preferred development, it is provided that the second components stored in the second database assigning a numeric value to each of the group data and the user selected third component group data, converting the respective component group data to a vector representation using the numeric value, and normalizing a respective component group vector to a value between zero and one.

The conversion of the respective component group data into a vector representation enables easier processing and comparability of the component group data for determining the similarity between different data sets.

According to a further preferred development, it is provided that a similarity between a vector of all component group data and a vector of the third component group data selected by the user is calculated, with a similarity score between the third component group data selected by the user and the second component group data stored in the database , in particular using a cosine similarity based method. Thus, a similarity score can be efficiently calculated between the vector of all component group data and the vector of the third component group data selected by the user.

According to a further preferred development, it is provided that the user selects fifth component group data from the component group data provided with the similarity score, with an assignment being carried out between the fifth component group data and the third component group data selected by the user by identifying components with the same package designation .

The package designation can thus be used in an advantageous manner to enable an improved assignment of the fifth component group data selected by the user to the third component group data selected by the user.

The configurations and developments described can be combined with one another as desired.

Further possible configurations, developments and implementations of the invention also include combinations of features of the invention described above or below with regard to the exemplary embodiments that are not explicitly mentioned.

In an advantageous development, it can be provided that the printed circuit is produced according to this layout or according to this information for placing groups of components. That is, a manufacturing machine may be provided which is operated to manufacture the printed circuit board according to this layout or according to this information for placing groups of components

character list

The accompanying drawings are provided to provide a further understanding of embodiments of the invention. They illustrate embodiments and, together with the description, serve to explain principles and concepts of the invention.

Other embodiments and many of the foregoing advantages will become apparent by reference to the drawings. The illustrated elements of the drawings are not necessarily shown to scale with respect to one another.

• 1-4 ein Ablaufdiagramm eines computerimplementierten Verfahrens zum Bestimmen eines Layouts einer gedruckten Schaltung gemäß einer bevorzugten Ausführungsform der Erfindung;
• 5 ein Ablaufdiagramm eines computerimplementierten Verfahrens zum Bereitstellen eines ersten Klassifikationsalgorithmus maschinellen Lernens zum Bestimmen eines Ähnlichkeitsmaßes von ersten Komponenten- und/oder Layoutdaten zu in einer Datenbank gespeicherten, zweiten Komponentengruppen- und/oder Layoutdaten;
• 6-8 ein Ablaufdiagramm eines computerimplementierten Verfahrens zum Bestimmen von Informationen zum Platzieren von Komponentengruppen auf einer gedruckten Schaltung gemäß der bevorzugten Ausführungsform der Erfindung;
• 9 eine schematische Darstellung eines Systems zum Bestimmen eines Layouts einer gedruckten Schaltung und/oder von Informationen zum Platzieren von Komponenten auf einer gedruckten Schaltung.

Show it:

• 1-4 a flowchart of a computer-implemented method for determining a layout of a printed circuit board according to a preferred embodiment of the invention;
• 5 a flowchart of a computer-implemented method for providing a first machine learning classification algorithm for determining a degree of similarity of first component and/or layout data to second component group and/or layout data stored in a database;
• 6-8 Figure 12 is a flowchart of a computer-implemented method for determining information for placing groups of components on a printed circuit board in accordance with the preferred embodiment of the invention;
• 9 FIG. 12 is a schematic representation of a system for determining a printed circuit board layout and/or information for placing components on a printed circuit board.

In the figures of the drawings, the same reference symbols denote the same or functionally the same Items, parts or components unless otherwise noted.

1 FIG. 12 shows a flow chart of a computer-implemented method for determining a layout of a printed circuit board according to a preferred embodiment of the invention.

A draft layout and component placement information is reusable as many of the sub-modules are shared across projects. A PCB layout design is a complex structure with various levels of detail included in the design file.

Therefore, a method is provided to browse previous layouts and recommend relevant layouts that a layout artist can use to solve a new product requirement.

A layouter or user U uploads a new ASCII file from which features f1 to fn or data D1 of a circuit diagram of the printed circuit board 1 are extracted and compared with the historical layouts to find the 10 most important/similar projects. Alternatively, the number of most important/similar projects can be less or more than 10, for example.

A parser extracts the relevant information or data listed below from the ASCII file.

The component information is extracted, such as housing information, component placement, geometric coordinates of the component and/or a height of the component, and component pin information is extracted, such as a number of pins for a component and/or a number of actual connected pins.

In addition, there is an extraction of the net information, i.e. the connection information between the components on the circuit board, such as the pin to which the net is connected, the type of net, voltage and current flowing through the net, an extraction of the information at the layout level like number of conductive layers, cooling layers with exact coordinates on the board as well as an extraction of meta-information like conductor thickness and/or CDCI values.

These results are then reused by the layout artist to take the existing design and make changes to it, rather than creating it from scratch.

Alternatively, a layout developer can enter identification data D2 stored in a first database DB1 or a historical project ID of a circuit diagram of the printed circuit 1, on the basis of which the layout 10 of the printed circuit 1 assigned to the identification data D2 is retrieved in the first database DB1.

The weight W of the features f1 to fn is initially determined by domain experts. The pipeline also enables a user feedback mechanism where these weights are updated over time based on user behavior.

A similarity measure M of the extracted first component group and/or layout data D3 to the second component group and/or layout data D4 stored in the first database DB1 and/or in a second database DB2 is also determined S4 using a first classification algorithm A1 machine learning .

In addition, a data set D5 is output S5 comprising a predetermined number of second component groups and/or layouts 10, which have a predetermined degree of similarity M.

2 Figure 12 shows a flow chart of a partial aspect of the computer-implemented method for determining the layout of the printed circuit board in accordance with the preferred embodiment of the invention.

Regular or irregular shapes of an area of the printed circuit board 1 covered by the component group 12 or the layout 10 are converted into a polygon PG represented by a sequence of points Pxy in a 2D coordinate system.

The 2D coordinate system comprises a large number of square fields 14, in particular with a side length of between 0.001 and 0.1 mm, preferably 0.01 mm determines which of the square fields 14 are located within the polygon.

An outer boundary is also established for each component group 12 and/or each layout 10, and for each square field 14 it is determined by a ray-casting algorithm whether a center point of the field 14 is located within the polygon.

To achieve the final goal of identifying the trapped region, we propose the following algorithm.

The square fields 14 are represented by 4 corner points and a center. For each square field 14 we use the ray casting algorithm to find out whether the center of the field 14 is inside the geometry or not. With the Ray Casting algorithm, an infinite line is drawn from the selected point to an infinite space. If the number of intersections with the polygon is odd, the point is inside the polygon. If the number of intersections with the polygon is an even number, the point is outside the polygon.

The area enclosed by a layer could be determined by a union of all geometries within the layer. This information could be used to calculate the area covered by a layer, component, cooling area, etc. and/or the number of overlaps between two layers.

3 Figure 12 shows a flow chart of a partial aspect of the computer-implemented method for determining the layout of the printed circuit board in accordance with the preferred embodiment of the invention.

A degree of similarity M of the extracted first component group and/or layout data D3 to the second component group and/or layout data D4 stored in the first database DB1 and/or in a second database DB2 is determined using a first machine learning classification algorithm A1.

The similarity measure M used here is a distance measure. The further the distance D, the lower the similarity. In view of the weight W of the features f1-fn, a weighted Euclidean distance is used as a measure.

In the formula, f = layout features and w _i = weight of feature "i".

Alternatively, for example, another similarity calculation algorithm could be used instead of the Euclidean distance-based method, which takes into account weights specified by the domain expert.

4 Figure 12 shows a flow chart of a partial aspect of the computer-implemented method for determining the layout of the printed circuit board in accordance with the preferred embodiment of the invention.

The feature weights W are provided by the domain experts based on the interpretation of the feature importance. These weights W are used to recommend similar layouts 10 to a user.

If the user considers the given recommendation relevant, he can mark it as relevant Y. He can also mark the recommendation as irrelevant N if it does not satisfy his need for information.

This user feedback of marking recommended layouts as relevant/irrelevant is then used to optimize the model weights W.

The data set collected with the user feedback is used to train a classifier, in particular a second machine learning classification algorithm A2, which can recognize two layouts whether they are relevant as a binary classification problem or not.

The training can be carried out with various learning algorithms and can be completed with a gradient boosted tree algorithm, for example. The updated feedback is continuously fed into the algorithm.

5 shows a flow chart of a computer-implemented method for providing a first machine learning classification algorithm for determining a degree of similarity of first component and/or layout data to second component group and/or layout data stored in a database.

The method includes receiving S1' first training data TD1 extracted component and/or layout data of a printed circuit board and receiving S2' second training data TD2 of a degree of similarity M of the extracted component and/or layout data to second component groups and stored in a database DB2 /or layout data D4 of the printed circuit 1.

The method also includes training S3′ of the first machine learning classification algorithm A1 using an optimization algorithm A3, which calculates an extreme value of a loss function.

6 12 shows a flow chart of a computer-implemented method for determining information for placing groups of components on a printed circuit board in accordance with the preferred embodiment of the invention.

User U uploads a new layout file. When a new layout, ie data of a circuit diagram of the printed circuit board 1, is uploaded, it is parsed by a parser P and the component groups CG1-CGn are extracted together with the main component and stored in a third database DB3.

The extracted component groups CG1-CGn are then displayed to the user U and he selects the component groups CG1-CGn for which similar component groups CG1-CGn are to be identified.

A component group CG1-CGn consists of a main component and a set of peripheral components. In order to identify the similar component group CG1-CGn, all previous component groups CG1-CGn whose main component is identical to the main component CG1-CGn entered by the user are therefore first filtered. In this way, all earlier component groups CG1-CGn filtered in this way become candidates for the similarity search.

For example, if the user has selected main component CG1 with main component package or package IC-500, the database is first searched for component groups CG1-CGn whose main component package is IC-500. All such component groups CG1-CGn in the database become candidates for the similarity search.

The user U selects third component group data D6 from the extracted first component group data D3. To determine the degree of similarity M of the extracted first component group data D3 to the second component group and/or layout data D4 stored in the second database DB2, fourth component group data D7 is determined, the main component of which is identical to a main component of the third component group data D6 selected by the user U.

A numerical value V is assigned to the second component group data D4 stored in the second database DB2 and to the third component group data D6 selected by the user U. The respective component group data D4, D6 are converted into a vector representation using the numerical value V. A respective component group vector is then normalized to a value between zero and one.

Further, a similarity between a vector of each component group data and a vector of the third component group data D6 selected by the user U is calculated.

A similarity score SC between the third component group data D6 selected by the user U and the second component group data D4 stored in the database is calculated, in particular using a cosine similarity-based method 18 . The candidate component groups CG1-CGn are then listed according to the cosine similarity score.

The calculated ranking of the candidate component groups CG1-CGn is displayed to the user. This then selects a source component group SCG from which the placement information is exported to a target component group.

7 Figure 1 shows a flow chart of a partial aspect of the computer-implemented method for determining information for placing component groups CG1-CGn on the printed circuit board according to the preferred embodiment of the invention.

After the most relevant component group CG1-CGn has been identified, the next step is to identify the mapping between components in the source and target component groups SCG, TCG.

The source component group SCG is the one that the user has selected as the most relevant component group. The target component group TCG is the input component group selected by the user from his uploaded file.

Not all components of the target component group TCG may exist in the source component group SCG. Therefore, this type of mapping is required.

The assignment or mapping is achieved by identifying the components with the same package name in both the source and target component groups TCG.

That is, for each peripheral component in the target component group TCG, package names are compared with the package name of the peripheral component in the source component group SCG.

In scenarios where more than one component with the same package name exists, the schematic's network topology is created based on the most-common-neighbor strategy.

8th FIG. 12 shows a flow chart of another partial aspect of the computer-implemented method for determining information for placing groups of components on the printed circuit board in accordance with the preferred embodiment of the invention.

The export and import of component placement information or data is in 8th shown.

In the illustration above left, the relative position of the peripheral component in relation to the main component is calculated in the source component group SCG.

The illustration at the top right shows a peripheral component in the target component group TCG being moved to the origin.

In the illustration below left, the peripheral component is moved from the origin to the location to the main component.

The illustration below right shows using the calculated relative position to calculate the position of a peripheral component with respect to the main component.

• x_m, y_m: Position der Hauptkomponente;
• x_p, y_p: Position der peripheren Komponente; und
• x_p_ref, y_p_ref: Relative Position der peripheren Komponente zur Hauptkomponente in der Quellkomponentengruppe.

In the 8th Abbreviations used denote the following:

• x_m, y_m: position of the main component;
• x_p, y_p: position of the peripheral component; and
• x_p_ref, y_p_ref: Relative position of the peripheral component to the main component in the source component group.

9 shows a schematic representation of a system 2 for determining a layout 10 of a printed circuit board 1 and/or information for placing components on a printed circuit board.

The system 2 has a computing device 16 which is set up to execute the method according to the invention for determining a layout 10 of a printed circuit board 1 and/or information for placing component groups CG1-CGn on a printed circuit board 1.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• US2013091481A [0004]
• US2019042684A1 [0006]

Claims (15)

Computer-implemented method for determining a layout (10) of a printed circuit board (1) and/or information for placing component groups (CG1-CGn) on a printed circuit board (1), comprising the steps: Receiving (S1) data (D1) of a circuit diagram of the printed circuit (1) or identification data (D2) stored in a first database (DB1) of a circuit diagram of the printed circuit (1) and calling up (S2) the identification data (D2) associated layouts (10) of the printed circuit (1) in the first database (DB1); extraction (S3) of first component group and/or layout data (D3) from the data (D1) of the circuit diagram of the layout (10) of the printed circuit board (1); Determining (S4) a degree of similarity (M) of the extracted first component group and/or layout data (D3) to the second component group and/or layout data (D4) stored in the first database (DB1) and/or in a second database (DB2). ) using a first classification algorithm (A1) machine learning; and Outputting (S5) a data set (D5) comprising a predetermined number of second component groups and/or layouts (10) which have a predetermined degree of similarity (M). Computer-implemented method claim 1 , Component data, component pin data, network data, layout data and/or layout metadata being extracted from the data (D1), in particular an ASCII file, of the circuit diagram of the printed circuit (1). Computer-implemented method claim 1 or 2 , converting regular or irregular shapes of an area of the printed circuit board (1) covered by the component group (CG1-CGn) or the layout (10) into a polygon (PG) represented by a sequence of points (Pxy) in a 2D coordinate system will. Computer-implemented method claim 3 , wherein the 2D coordinate system comprises a plurality of square fields (14), in particular with a side length of between 0.001 and 0.1 mm, wherein for estimating the area covered by the component group (CG1-CGn) or the layout (10). of the printed circuit (1) it is determined which of the square fields (14) are arranged within the polygon. Computer-implemented method claim 4 , wherein an outer boundary is created for each component group (CG1-CGn) and/or each layout (10), wherein for each square field (14) it is determined by a ray-casting algorithm whether a center point of the field (14) located inside the polygon. Computer-implemented method according to one of the preceding claims, wherein the degree of similarity (M) of the extracted first component group and / or layout data (D3) to the second database (DB2) stored, second component group and / or layout data (D4) by calculating a Distance measure, in particular a weighted Euclidean distance is determined. Computer-implemented method according to one of the preceding claims, wherein the data set (D5) output by the first classification algorithm (A1) of machine learning is received by a second classification algorithm (A2) of machine learning, which for each of the predetermined number of component groups (CG1-CGn) and / or layouts (10) outputs a relevance score (RS), on the basis of which a weighting (W) of the first classification algorithm (A1) is adapted to machine learning. Computer-implemented method according to one of the preceding claims, wherein a user (U) selects third component group data (D6) from the extracted first component group data (D3), wherein for determining the degree of similarity (M) of the extracted first component group data (D3) to in the second database (DB2) stored second component group and/or layout data (D4) fourth component group data (D7) whose main component is identical to a main component of the third component group data (D6) selected by the user (U). Computer-implemented method according to one of the preceding claims, wherein the second component group data (D4) stored in the second database (DB2) and the third component group data (D6) selected by the user (U) are each assigned a numerical value (V), the respective component group data (D4, D6) is converted into a vector representation using the numeric value (V) and wherein a respective component group vector is normalized to a value between zero and one. Computer-implemented method claim 8 or 9 , where a similarity between a vector of all component group data and a vector of the third component group data (D6) selected by the user (U) is calculated, with a similarity score (SC) between the third component group data (D6) selected by the user (U) and the second component group data (D4 ), in particular using a cosine similarity based method (18). Computer-implemented method claim 10 , The user (U) selects fifth component group data (D8) from the component group data provided with the similarity score (SC), wherein between the fifth component group data (D8) and the third component group data (D6) selected by the user (U) an assignment is performed by identifying components with the same package designation. Computer-implemented method for providing a first classification algorithm (A1) machine learning for determining a degree of similarity (M) of first component and / or layout data (D3) stored in a database, second component group and / or layout data (D4), with the steps : receiving (S1') first training data (TD1) of extracted component and/or layout data of a printed circuit board (1); Receiving (S2') second training data (TD2) of a degree of similarity (M) of the extracted component and/or layout data to second component group and/or layout data (D4) of the printed circuit (1) stored in a database (DB2); and Training (S3') the first classification algorithm machine learning (A1) by an optimization algorithm (A3), which calculates an extreme value of a loss function. System (2) for determining a layout (10) of a printed circuit (1) and/or information for placing components on a printed circuit (1), comprising a computing device (16) which is set up to carry out the method according to one the Claims 1 until 11 to execute. Computer program with program code to implement the method according to one of Claims 1 until 11 to be performed when the computer program is run on a computer. Computer-readable data carrier with program code of a computer program for the method according to one of Claims 1 until 11 to be performed when the computer program is run on a computer.

Images