Systems and methods for predicting a quality of a printed circuit board assembly

1 . A computer-implemented method of predicting a quality of a printed circuit board (PCB) assembly, the computer-implemented method comprising:

obtaining production data relating to production of the PCB assembly, the production data including solder paste information;

mapping the production data, including the solder paste information, onto a latent vector of a latent space of a trained adaptive algorithm, wherein the trained adaptive algorithm is trained on real X-ray images of PCB assemblies, serves for generating X-ray images of the PCB assemblies, or a combination thereof; and

determining a subspace of the latent space related to the latent vector, the subspace indicating a quality of the PCB assembly;

generating, by the trained adaptive algorithm, based on the latent vector, an X-ray image of the PCB assembly, such that a quality of the PCB assembly is determined; or

a combination thereof,

wherein the solder paste information includes a formation, a size, and a volume of one or more solder depots applied to the PCB assembly,

wherein the PCB assembly comprises a plurality of regions of interest, each region of interest of the plurality of regions of interest comprising one or more soldering points, the respective region of interest corresponding to an area of the PCB assembly, an area of an image of the PCB assembly, or the area of the PCB assembly and the area of the image of the PCB assembly, the production data relating to the respective region of interest,

wherein the computer-implemented method further comprises selecting, for each region of interest of the plurality of regions of interest, a corresponding trained adaptive algorithm for determining the quality of the PCB assembly, and

wherein the plurality of regions of interest differ regarding a number of the one or more soldering points, positions of the one or more soldering points, sizes of the one or more soldering points, or any combination thereof.

2 . The computer-implemented method of claim 1 , wherein the mapping comprises mapping based on a trained regression algorithm.

3 . The computer-implemented method of claim 1 , wherein the subspace indicates a quality of a region of interest of the PCB assembly.

4 . The method of claim 1 , further comprising outputting, by the trained adaptive algorithm, the generated X-ray image, such that a quality of the PCB assembly is determined.

5 . The method of claim 1 , further comprising determining a quality indicator for the PCB assembly based on the latent vector, the subspace of the trained adaptive algorithm, the generated X-ray image, or any combination thereof.

6 . The method of claim 1 , wherein the production data comprises:

component information;

information of one or more properties of the PCB;

a residual oxygen level, a temperature profile of a reflow oven, a degree of aging, or any combination thereof of the PCB, components mounted on the PCB, or a combination thereof; or

any combination thereof.

7 . The method of claim 6 , wherein:

the solder paste information includes formation, size, volume, or any combination thereof of one or more solder depots applied to the PCB;

the component information includes co-planarity information of pins of a component;

the component information includes a material of a substrate of the PCB, a solder resist application process type, a production site, or any combination thereof; or

any combination thereof.

8 . The method of claim 1 , further comprising:

inspecting the generated X-ray image of the PCB assembly based on a first computer vision algorithm using predetermined static criteria;

inspecting the generated X-ray image of the PCB assembly based on a second computer algorithm using a trained machine learning model; or

a combination thereof.

9 . An apparatus comprising:

a processor; and

a memory,

wherein the processor is configured to:

obtain production data relating to production of a PCB assembly, the production data including solder paste information;

map the production data, including the solder paste information, onto a latent vector of a latent space of a trained adaptive algorithm, wherein the trained adaptive algorithm is trained on real X-ray images of PCB assemblies, serves for generating X-ray images of the PCB assemblies, or a combination thereof, and

determine a subspace of the latent space related to the latent vector, the subspace indicating a quality of the PCB assembly;

generate, by the trained adaptive algorithm, based on the latent vector, an X-ray image of the PCB assembly, such that a quality of the PCB assembly is determined; or

a combination thereof,

wherein the solder paste information includes a formation, a size, and a volume of one or more solder depots applied to the PCB assembly,

wherein the PCB assembly comprises a plurality of regions of interest, each region of interest of the plurality of regions of interest comprising one or more soldering points, the respective region of interest corresponding to an area of the PCB assembly, an area of an image of the PCB assembly, or the area of the PCB assembly and the area of the image of the PCB assembly, the production data relating to the respective region of interest, and

wherein the processor is further configured to select, for each region of interest of the plurality of regions of interest, a corresponding trained adaptive algorithm for determining the quality of the PCB assembly, and

wherein the plurality of regions of interest differ regarding a number of the one or more soldering points, positions of the one or more soldering points, sizes of the one or more soldering points, or any combination thereof.

10 . The apparatus of claim 9 , wherein the apparatus is an inspection station.

11 . A computer-implemented method of obtaining an adaptive algorithm for predicting a quality of one or more printed circuit board (PCB) assemblies, the computer-implemented method comprising:

obtaining real X-ray images of one or more PCB assemblies;

training an autoencoder capable of reconstructing the real X-ray images input;

obtaining the adaptive algorithm for predicting the quality of the one or more PCB assemblies, the obtaining of the adaptive algorithm comprising identifying a decoder part of the autoencoder,

wherein the decoder part serves as a latent space interpreter,

wherein the adaptive algorithm is configured to receive a latent space vector as input and reconstruct or generate an X-ray image based on the received latent space vector as output, the latent space vector being based on production data,

wherein the production data includes solder paste information, the solder paste information including a formation, a size, and a volume of one or more solder depots applied to the one or more PCB assemblies,

wherein a PCB assembly of the one or more PCB assemblies comprises a plurality of regions of interest, each region of interest of the plurality of regions of interest comprising one or more soldering points, the respective region of interest corresponding to an area of the PCB assembly, an area of an image of the PCB assembly, or the area of the PCB assembly and the area of the image of the PCB assembly, the production data relating to the respective region of interest,

wherein the computer-implemented method further comprises selecting, for each region of interest of the plurality of regions of interest, a corresponding trained adaptive algorithm for determining the quality of the PCB assembly, and

wherein the plurality of regions of interest differ regarding a number of the one or more soldering points, positions of the one or more soldering points, sizes of the one or more soldering points, or any combination thereof.

12 . The computer-implemented method of claim 11 , wherein obtaining the real X-ray images of one or more PCB assemblies comprises obtaining the real X-ray images of one or more PCB assemblies from an X-ray inspection system.

13 . The computer-implemented method of claim 11 , wherein identifying the decoder part of the autoencoder comprises removing an encoder part of the autoencoder.

14 . The method of claim 6 , wherein the production data comprises the residual oxygen level, the temperature profile of the reflow oven, and the degree of aging.

15 . The computer-implemented method of claim 1 , wherein a first of the regions of interest and a second of the regions of interest differ regarding function of components attached to the PCB assembly within the first region of interest and the second region of interest, respectively.