EP4732186A1 - Image based hot spot detection based on void distance or aspect ratio of a defective area - Google Patents

Abstract

A Method for determining a locally increased a thermal resistance (R1, R2) of a defective area (A1, A2) of a connection layer (L) is described. The defective area comprises defective spots (S1, S2). In a first alternative, a thermal resistance (R1, R2) assigned to a defective area (A1, A2) is dependent on the distance (d1, d2) between the spots (S1, S2) of the defective area, wherein the thermal resistance assigned to a defective area (A1, A2) increases with decreasing distance (d1, d2) between the spots (S1, S2) of the defective area (A1, A2). In a second alternative, a thermal resistance (R1, R2) assigned to a defective area (A1, A2) is dependent on the aspect ratio of the defective area (A1, A2), wherein the thermal resistance (R1, R2) assigned to a defective area (A1, A2), increases with decreasing aspect ratio of width vs. height (w1, h1), in recumbent orientation (RO). The method can be carried out by an artificial neural network.

Description

Description

Image based hot spot detection based on void distance or aspect ratio of a defective area

Power electronics applications require to conduct high currents generate a substantial amount of heat. Thus, reliable connections for transferring heat and/or high currents are vital to power electronic modules. Usually, power electronic elements have a flat surface attached to a heat sink (directly or via a substrate). A connection layer (eg. sintered layer or solder layer) connects this flat surface to a heat sink (directly or via a substrate). Even small failures within the connection between the heat source I power electronic element to a heat sink or conducting substrate lead to a locally increased thermal resistance. The locally increased thermal resistance between the heat source and the heat sink leads to a hot spot within the power electronic element. As a result, the power electronic element is locally degraded during operation.

It is known to inspect images of the connection layer in order to discard products with large defects while product with acceptable defects are not discarded.

Since high scrap rates significantly reduce the economical success of a product, it is an object of the invention to distinguish critical defective areas in the connection layer from acceptable defective areas in a more reliable way.

This object is solved by the subject-matter of claim 1 . Further features, embodiments and advantages are given with the dependent claims, the description and the figures.

It has been discovered that a more precise prediction of the seventy of hot spots can be achieved by considering also the areal density of defective spots or shape of the defective area formed by the defective spots. It has been discovered that defective spots, which are close together, lead to a locally increased thermal resistance higher than the thermal resistance of defective spots, which are further apart from each other. In other words, the hot spot occurring at defective spots, which are close together, leads to a higher failing risk than the hot spot at defective spots, which are located further apart from each other. This finding relates to the same defective area size, ie. to the same total size of the defective spots (same size of the defective area consisting of the defective spots).

Further, it has been discovered that a defective area (consisting of defective spots), which form a line or another shape with high aspect ratio, leads to a locally increased thermal resistance lower than the thermal resistance of a defective area, which has a more concentrated shape like a circle or a square or another shape with a lower aspect ratio. In other words, the hot spot occurring at a defective area having a concentrated shape (circle, square, regular polygon), leads to a higher failing risk than a defective area with a more elongated shape. This finding relates to the same defective area size.

Therefore, it is proposed to determine the thermal resistance of a defective area taking into account the distance between the defective spots of a connection layer, to take into account the shape (in particular the aspect ratio) of the defective area of a connection layer, or both. In addition, the size of the defective area and/or the closeness to a boundary of the connection layer can be taken into account. The locally increased thermal resistance of all defective areas or of the defective areas over a given size of the connection layer can be determined.

In one embodiment, the determined thermal resistances are used for detecting connection layers to be discarded. According to the determined thermal resistances, the connection layer (or the product containing the connection layer) can be labelled or marked as defective, or not. In case one or more of the determined locally increased thermal resistances are higher than a threshold, the connection layer (or the product containing the connection layer) is labelled or marked as defective.

Other embodiments are focused on the determined locally increased thermal resistance or resistances. The determined locally increased thermal resistance or resistances can be provided as a technical information about locally increased thermal resistances in the sense of a measurement, prediction or analysis of the individual thermal characteristics of the connection layer. The method described herein determines increased thermal resistance or resistances and, therefore, provides thermal characteristics of the connection layer, comparable to a high power test of the contact layer, in which at least one local hot spot is measured by a thermal imaging method, providing the information about at least one defective area or defective spot. Consequently, the method described herein determines at least one locally increased a thermal resistance of a contact layer, wherein information characterizing the thermal resistance can be output, memorized or forwarded. The locally increased thermal resistance can be given as an absolute thermal resistance value (eg. normalized by the area size) or can be given by a signal or value characterizing the locally increased a thermal resistance. For example, the locally increased thermal resistance can be given as an indication, whether the locally increased thermal resistance is above a given threshold, or not. Further, the locally increased thermal resistance can be given as an indication, whether the locally increased thermal resistance leads to a high risk hot spot, or not. In particular, the locally increased thermal resistance can be given as an indication, whether there is a high failure probability due to the locally increased thermal resistance, or not. The locally increased thermal resistance can be given as an information characterizing the locally increased thermal resistance itself (eg. as an absolute value), can be given as a relative value, eg. in form of a value characterizing the increase resulting from the defective area or spots, can be given as a failure probability (interval), or can be given as a binary value characterizing, whether thermal specifications are met, or not.

The method described herein is used for determining a locally increased a thermal resistance of a defective area of a connection layer. The connection layer can be an attachment layer, by which a power electronics element is attached to a heat sink. The power electronics element can be attached directly to a heat sink or can be attached indirectly, eg. via at least one substrate, a joining element, a thermal interface layer or others. The power electronics element can be a bare die or can be a housed element, in particular a power semiconductor, eg. a power transistor, a power thyristor, a power diode or others. The connection layer can be the layer between power electronics element and the heat sink. The connection layer can be a least partly solidified layer, eg. a solidified soldering or brazing layer or a sintered layer. The connection layer can be the layer between power electronics element and the heat sink. The connection layer can be a least partly solidified layer, eg. a solidified soldering or brazing layer or a sintered layer. The connection layer can be an unmounted layer for attachment of an electronic element, in particular in form of a green body, eg. a green body of soldering paste or sintering paste (paste comprising unmolten soldering particles, unfired sintering particles). The connection layer can be in form of a dried green body of sintering paste. The connection layer is preferably dedicated to transport heat (from one main surface to the other) and can be an interface between a heat source and a heat sink and can be denoted as thermal transport layer. Further, the connection layer can be a solder preform. In particular, the connection layer can be a soldered solder preform, for example a soldered preform (eg. a preform layer solidified after a soldering step) attaching a power electronics element to a heat sink. Further, the connection layer can be a (yet) unsoldered solder preform, which is intended to be soldered for attaching a power electronics element to a heat sink.

At least one defective area is given within the connection layer. The defective area comprises defective spots. Particularly, the defective area consists of defective spots. Defective spots are spots, in which the connection layer does not have sufficient thickness or are void spots (lacking of connection layer material). Defective spots are spots, in which the connection layer material provides an increased thermal resistance (in comparison to non-defective spots, ie. spots with sufficient thickness or with sufficient connection layer material). This relates to a connection layer by which an element is mounted on a heat sink as well as to a connection layer in form of a green body, which transforms (by soldering or sintering or solidifying) into a soldered or sintered connection layer, the solidified (soldered or sintered) connection layer having defective spots where the green body had defective spots before transformation.

A thermal resistance is assigned to a defective area having defective spots. The thermal resistance is dependent on the distance between the spots. The thermal resistance assigned to a defective area increases with decreasing distance between the spots of the defective area. In other words: A first thermal resistance assigned to a defective area having defective spots located in a first distance to each other, the first thermal resistance being greater than a second thermal resistance, which is assigned to a defective area having defective spots located in a second distance to each other, which is longer than the first distance. This also relates to (nonoverlapping) groups of defective spots, groups of spots being defined as spots within the same area section of the connection layer. A function can be given, which assigns distinct distances between defective spots to distinct thermal resistances or distinct values or signals characterizing distinct thermal resistances. The function can be partly or completely continuous or can be provided by a discontinuous function. The function is preferably decreasing: the higher the distance among the defective spots, the lower the thermal resistance.

Alternatively or in combination, the shape of defective areas can be taken into account: distinct thermal resistances are assigned to defective areas with distinct shapes. A thermal resistance assigned to a defective area can be dependent on the aspect ratio of the defective area. The thermal resistance assigned to a defective area increases with decreasing aspect ratio of width vs. height. This relates to a recumbent orientation of the defective area. In other words: A first thermal resistance is assigned to a defective area having a first aspect ratio, wherein the first thermal resistance is greater than a second thermal resistance, which is assigned to a defective area of a second aspect ratio being greater than the first aspect ratio. A function can be given, which assigns distinct aspect ratios to distinct thermal resistances or distinct values or signals characterizing distinct thermal resistances. The function can be partly or completely continuous or can be provided by a discontinuous function. The function is preferably increasing: the higher the aspect ratio, the lower the thermal resistance.

The aspect ratio relates to the defective area in recumbent orientation, ie. an orientation, in which the width is longer than the height. Preferably, this relates to an orientation within a system, in which the width is horizontal and the height (same or smaller than the width) is vertical. This particularly relates to rectangular defective areas or defective areas, which can be approximated by a rectangular shape. In particular, this relates to a shape, which is a incircle or an excircle of the defective area. The defective area can comprise one or more non-defect spots. In case that non-defect spots are included or surrounded by the defective area (defective spots), the shape relates to the circumference of the defective area. The aspect ratio can be given as an (approximated) value characterized by the ratio of circumference of the defective area and square root of the area size (surface size) of the defective area. The aspect ratio can be given for the defective area itself or for an approximated shape (regular polygon, eg. rectangle). The aspect ratio can be given for an area approximated to the defective area, eg. for an (rectangular, hexagonal, ... ) area inscribing or circumscribing the defect area. The aspect ratio can be given as a measure characterizing, how concentrated (low aspect ratio) or how elongated/stretched (high aspect ratio) the shape of the defective area is. The defective area can geometrically enclose one or more non-defect spots; preferably, the non-defective spots cover an area size of the defective area not more than 20%, 10%, 5% or 2% of the overall area size of the defective area. A step of outputting the determined thermal resistance or its increase can be provided. A step of inputting the (surface view or image of) the connection layer can be provided, wherein the thermal resistance is determined for this input.

Further embodiments provide that the thermal resistance is a thermal resistance per area (area size) covered by the defective spots. Alternatively, the defective area is normalized by the size of the respective defective area. The defective area to which the thermal resistance is assigned to (depending on the distance or depending on the aspect ratio) can relate to the same area size or can be normalized to the same area size. In this way, the shape of the defective area and/or the distance between defective spots has a significant impact on the thermal resistance even if the sizes of defective spots or defective areas vary significantly.

The thermal resistance can be given as the aggregated thermal resistance of the defective spots within the defective area. Further, the thermal resistance can be approximated as the maximum thermal resistance of the defective spots within the defective area. The thermal resistance can be given by information characterizing the locally increased thermal resistance itself (eg. as an absolute value), can be given as a relative value, eg. in form of a value characterizing the increase resulting from the defective area or spots, can be given as a failure probability (interval), or can be given as a binary value characterizing, whether thermal specifications are met, or not. In particular, it can be given as a defectiveness grade, in which distinct grades (excellent, good, satisfactory, sufficient, insufficient) are given to distinct levels of increase of thermal resistance, to distinct levels of thermal resistance, to distinct levels of defectiveness or to distinct levels of hot spot risk.

In order to locate the defective spots and/or the defective area, an image of the connection layer can be provided. The aspect ratio and/or the distance between defective spots can be determined by using the image, in particular by identifying the defective spots/area and by measuring the distance and/or the aspect ratio of the defective spots/area given in the image. An imaging method can be used to providing the image of the connection layer. The imaging method can be an x-ray imaging method, can be an optical imaging method, can be an ultrasonic imaging method, or similar. The image provided by the imaging method is preferably a top view image (showing the connection layer in top view).

The locally increased thermal resistance is preferably compared to a given threshold. The connection layer is preferably marked as defective in case the resistance is greater than the threshold and is marked as approved (non-defective) otherwise. Further, a quality measure (in particular a value characterizing the quality) can be assigned to the connection layer reflecting the amount of locally increased thermal resistance(s) within the connection layer, based on the defective spots/area. A quality measure (in particular a value characterizing the quality) can be assigned to the connection layer reflecting the probability of failure (eg. for a certain number of thermocycles or for a given live span of the connection layer), based on the defective spots/area. Preferably, an object or product (in particular a power electronics module) having a connection layer, which is marked as defective or to which a quality measure below a given quality standard is assigned, is discarded.

The method (ie. the determination of the locally increased thermal resistance) can be carried out (at least partly) by a neural network, by a statistical evaluation scheme (in particular by a regression model) or statistical algorithm or by an evaluation algorithm. The network, scheme or algorithm preferably has a classifier, which carries out the step of assigning the thermal resistance or an equivalent value to the defective area. By having this classifier, the thermal resistance assigned to a defective area increases with decreasing distance and/or increases with decreasing aspect ratio. The step of assigning is preferably carried out by the network, scheme or algorithm. In particular, the classifier distinguishes between distinct distances between the spots and/or distinguishes between distinct aspect ratios of the defective area. Instead of a classifier, a relationship (in particular of a regression model) can be used, the relationship representing the dependency between the thermal resistance and the distance and/or between the thermal resistance and the aspect ratio as given herein. The step of evaluating the thermal resistance, eg. by comparison to a given threshold (for marking the connection layer as defective or for discarding the connection layer), can be carried out within the neural network, the statistical evaluation scheme (in particular a regression model) or statistical algorithm or the evaluation algorithm, or can be carried out in an additional step. Further, step of determining the locally increased thermal resistance is preferably implemented by the neural network, the statistical evaluation scheme (in particular a regression model) or statistical algorithm or the evaluation algorithm.

In case that method (in particular the determination) is carried out by a neural network or in case that the step of assigning is carried out by the neural network (generally: a system capable of machine learning), the neural network is a neural network trained on thermal simulations of defective areas having distinct distances between the defective spots (of the defective area). Alternatively or in combination therewith, the neural network is a neural network is trained on thermal simulations of defective areas having distinct aspect ratios. In particular, the neural network predicts the thermal resistance of the defective area. Further, the neural network can be trained on thermal measurements or captured thermal images. The measurements I captured thermal images are provided by a thermal imaging step of the connection layer, the thermal measurement providing a measured image of the defective areas or spots (a thermal image of the connection layer during or after actively operating heat sources like power semiconductors on the connection layer or a substrate thermally connected to the connection layer). Further, an artificial neural network is described. The artificial neural network is adapted to carry out the method described herein or is at least adapted to carry out at least the step of assigning the thermal resistance. The neural network has or is based on a classifier, in particular a classifier as given above. The classifier distinguishes between distinct distances between the spots and/or distinguishes between distinct aspect ratios of the defective area. The artificial neural network is preferably trained using the distance and/or the aspect ratio as classifier. The artificial neural network uses the distance and/or the aspect ratio as feature vector. At least one class of the neural network is the distance and/or the aspect ratio (upon training.

The artificial neural network can be an artificial neural network trained by the distance and/or the aspect ratio as classifier, feature vector and/or class. The artificial neural network can be an artificial neural network trained with training data representing defective areas/spots and pertaining thermal simulations or thermal measurements of the (locally increased) thermal resistance of defective areas/spots. The training data is propagated within the artificial neural network and/or used for optimizing the neural network); distance and aspect ratio are thereby taken into account (as classifier, feature vector and/or class). Preferably, the training data represents distinct distances between the defective spots and/or represents distinct aspect ratios, together with the pertaining (and locally increased) thermal resistance.

The training data can represent the thermal resistance for the contact layer or at least for the defective spots/areas. The training data can represent the areal thermal resistance distribution and/or the resulting areal temperature distribution in operation of the contact layer and give R(x,y) or T(x,y), R being a thermal resistance and T being a temperature of distinct point having the coordinates (x,y) on the contact layer. The coordinates (x,y) are preferably based on a two-dimensional (cartesian) coordinate system. Upon operation, the local temperature reflects the local thermal resistance: the local thermal resistance increases and decreases with the local thermal resistance. The thermal resistance and/or the temperature can be given in absolute values or can be given as relative values. Relative values can characterize the (locally increased) resistance or the (locally increased) temperature (both relating to defective areas/spots) in relation to average resistance or average temperature, respectively, or, as an alternative, resistance or temperature at defective areas/spots in relation to resistance or temperature at non-defective areas. Upon carrying out the method, thermal simulations of the connection layer or thermal images of the connection layer (in operation) are used as input for the neural network (or for a statistical evaluation scheme or an evaluation algorithm).

The neural network (the statistical evaluation scheme or the evaluation algorithm provides the (locally increased) thermal resistance assigned to a defective area/spot as a result on this input. In general, the method of determining provides the thermal resistance as output, in particular in form of information characterizing the locally increased thermal resistance itself (e.g. as an absolute value), in form of a relative value, e.g. in form of a value characterizing the increase resulting from the defective area or spots, in form of a failure probability (interval), in form of a binary value characterizing, whether thermal specifications are met, or not. The output can be given as an image of the determined thermal resistance, the image two- dimensionally showing the connection layer and the respective areal thermal resistance distribution (in a third dimension - color, height, ... ), or can be given as (local) max. value(s) of form of scalar(s) or a binary value (defect connection layer / non-defect connection layer).

A software or program product can be provided, which includes the neural network, and in particular software interface(s) adapted for inputting an image (of the connection layer) and/or for outputting the thermal resistance or a signal characterizing the layer as defect/non-defect or characterizing a decision to discard/not discard a product linked with the image. The software is adapted to carry out the method described herein, in particular when running on a processor. A data carrier can be provided on which the software is memorized. The data carrier is preferably adapted to be accessed to for running the software in a processor. The software assigns (a) the thermal resistance or (b) a signal characterizing the layer as defect/non-defect or (c) a signal characterizing a decision to discard/not discard a product linked to the image of the connection layer and therefore changes or creates a property of the connection layer. In general, the software carrying out the method assigns a thermal property to the connection layer (ie. the connection layer under test) or provides an output doing so.

Figures 1 and 2 show exemplarily connection layers with defective spots of distinct distance and defective areas with distinct aspect ratios for exemplifying aspects of the method described herein.

Fig. 1 shows a first and a second connection layer L1 , L2 in form of rectangles. The first connection layer L1 (left) comprises two defective spots S1 , S2. This is also the case for the second connection layer L2 (right). The two defective spots S1 , S2 of the first connection layer L1 are spaced apart by a distance d1 . The two defective spots S1 , S2 of the second connection layer L2 are spaced apart by a distance d2. The distance d2 is larger than distance d1 . In other words, the defective spots S1 , S2 of connection layer L1 are closer to each other than the defective spots S1 , S2 of connection layer L2. Thus, the defective spots S1 , S2 of connection layer L2 are further apart from each other than the defective spots S1 , S2 of connection layer L1 . In the example of Fig. 1 , the spots S1 and S2 of connection layer L2 are similar to the spots S1 , S2 of connection layer L2, apart from the mutual distance between each other. In particular, the (area) sizes of the spots S1 , S2 within Fig. 1 are the same. Further, their shape is the same. In layer L1 , the defective spots S1 , S2 constitute the defective area A1. In layer L2, the defective spots S1 , S2 constitute the defective area A2.

In Fig. 1 , the first and the second connection layer L1 , L2 lie in a plane, wherein two- dimensional coordinates x, y describe the location of points/spots/areas of the respective layer L1 , L2. In Fig. 1 , the thermal resistance R1 , R2 is depicted in dependence of the location x, wherein the diagram showing the thermal resistance R1 , R2 depending on x relates to the thermal resistance along line C1 , C2. Line C1 is a cut through the defective spots S1 , S2 of connection layer L1 along the x-axis. For connection layer L1 , diagram D1 shows the thermal resistance R1 along the x- axis. For connection layer L2, diagram D2 shows the thermal resistance R2 along the x-axis. Line C2 is a cut through the defective spots S1 , S2 of connection layer L2 along the x-axis. It can be seen that maxima R1 M, R2M of the thermal resistance R1 , R2 (within the diagrams D1 , D2) occur at the location of the defective spots S1 , S2. It is shown that the maximum thermal resistance R1 M for the spots S1 , S2 of layer L1 (distance: d1 ) is greater than the maximum thermal resistance R2M for the spots S1 , S2 (distance: d2; d2 > d1 ) of layer L2. Thus, the local increase of the thermal resistance R1 M is greater for defective spots S1 , S2, which are close together (distance: d1 ), than the local increase of the thermal resistance R2M for defective spots S1 , S2, which are further apart from each other (distance d2 > d1 ).

In other words, the thermal resistance R1 M, R2M (in particular the locally increased thermal resistance or the increase itself) of defective spots S1 , S2 depends on the distance d1 , d2 between the defective spots S1 , S2. The thermal resistance R1 , R2 (in particular, its maximum R1 M, R2M) assigned to defective spots S1 , S2 increases with decreasing distance (d1 < d2) between the spots S1 , S2.

Fig. 2 shows a first and a second connection layer L1 , L2 in form of rectangles. The first connection layer L1 (left) comprises a (first) defective area A1 . This is also the case for the second connection layer L2 (right) having a (second) defective area A2. The defective area A1 of the first connection layer L1 has a width w1 and a height hi , resulting in an aspect ratio of w1 / hi . The defective area A2 of the first connection layer L2 has a width w2 and a height h2, resulting in an aspect ratio of w2 I h2. Since the defective area A1 is in the shape approximated to an elongated rectangle, its aspect ratio (width vs. height in recumbent orientation RO) is larger than the aspect ratio of defective area A2 being in the shape approximated to a square. The aspect ratio of defective area A1 (of layer L2) is greater than the aspect ratio of defective area A2 (of layer L2). In the example of Fig. 2, the areas A1 , A2 are similar, apart from their aspect ratio (shape). In particular, the area sizes of the areas A1 and A2 are the same. Herein, the aspect ratio relates to a recumbent orientation RO in which the direction of the width is along the longest side and in which the direction of the height is perpendicular thereto. It can be seen from defective area AT, that in case of a shape, which is not orientated along a horizontal axis or y-axis, the recumbent (lying) orientation RO is along the direction of extension of the longest dimension of the area AT, which is the width wT. The height hT of area A1 ’ is perpendicular thereto. In this way, the aspect ratio relates to the maximum of all aspect ratios for all orientations (rotations) of the respective defective area.

In Fig. 2, the first and the second connection layer L1 , L2 lie in a plane, wherein two- dimensional coordinates x, y describe the location of areas of the respective layer L1 , L2. In Fig. 2, the thermal resistance R1 , R2 is depicted in dependence of the location x, wherein the diagram showing the thermal resistance R1 , R2 depending on x relates to the thermal resistance along line C1 , C2. Line C1 is a cut through the defective area A1 of connection layer L1 along the x-axis. Line C2 is a cut through the defective area A2 of connection layer L2 along the x-axis.

For connection layer L1 , diagram D1 of Fig. 2 shows the thermal resistance R1 along the x-axis. For connection layer L2, diagram D2 of Fig. 2 shows the thermal resistance R2 along the x-axis. It can be seen in figure 2 that maxima R1 M, R2M of the thermal resistance R1 , R2 are distinct and depend on the aspect ratio or shape of the defective area A1 , A2. It is shown that the maximum thermal resistance R1M for the defective area A1 of layer L1 (high aspect ratio of approx. 6) is less than the maximum thermal resistance R2M for the defective area A2 of layer L2 (low aspect ratio of approx. 1 ). Thus, the local increase of the thermal resistance R1 M is lower for defective area A1 having an elongated shape (high aspect ratio) of w1 / hi , than the local increase of the thermal resistance R2M for defective area A2 having a concentrated, square shape (low ratio) of w2 I h2.

In other words, the thermal resistance R1M, R2M (in particular the locally increased thermal resistance or the increase itself) of defective areas A1 , A2 depends on the aspect ratio of the respective defective areas. The thermal resistance R1 , R2 (in particular, its maximum R1 M, R2M) assigned to the defective areas A1 , A2 increases with decreasing aspect ratio (w2/h2 < w1/h1 ) of the defective areas A1 , A2.

In Fig. 2, the axes showing the thermal resistance R1 , R2 have the same scale. This also applies to Fig. 1 .

Claims

Claims

1 . Method for determining a locally increased a thermal resistance (R1 , R2) of a defective area (A1 , A2) of a connection layer (L), wherein the defective area comprises defective spots (S1 , S2), wherein a) a thermal resistance (R1 , R2) assigned to a defective area (A1 , A2) is dependent on the distance (d1 , d2) between the spots (S1 , S2) of the defective area, wherein the thermal resistance assigned to a defective area (A1 , A2) increases with decreasing distance (d1 , d2) between the spots (S1 , S2) of the defective area (A1 , A2). and/or wherein b) a thermal resistance (R1 , R2) assigned to a defective area (A1 , A2) is dependent on the aspect ratio of the defective area (A1 , A2), wherein the thermal resistance (R1 , R2) assigned to a defective area (A1 , A2), increases with decreasing aspect ratio of width vs. height (w1 , hi ), in recumbent orientation (RO).

2. Method according to claim 1 , wherein the thermal resistance (R1 , R2) is a thermal resistance per area covered by the defective spots (S1 , S2) or wherein the defective area (A1 , A2) is normalized by the size of the respective defective area (A1 , A2).

3. Method according to claim 1 or 2, wherein the defective area (A1 , A2) to which the thermal resistance is assigned to depending on the distance or depending on the aspect ratio relates to the same area size or is normalized to the same area size.

4. Method according to claim 1 , 2 or 3, wherein the thermal resistance is given as the aggregated resistance of the defective spots within the defective area (A1 , A2), or wherein the thermal resistance is approximated as the maximum thermal resistance of the defective spots (S1 , S2) within the defective area (A1 , A2).

5. Method according to one of the preceding claims, further comprising: providing an image of the connection layer (L) by using an imaging method.

6. Method according to claim 5, wherein the imaging method is an x-ray imaging method, is an optical imaging method or is an ultrasonic imaging method.

7. Method according to one of the preceding claims, wherein the locally increased thermal resistance (R1 , R2) is compared to a given threshold, and wherein the connection layer (L) is marked as defective, in case the resistance (R1 , R2) is greater than the threshold and is marked as approved otherwise.

8. Method according to claim 7, wherein an object having a connection layer (L) marked as defective is discarded.

9. Method according to one of the preceding claims, wherein the connection layer is an unmounted layer for attachment of an electronic element and is in form of a green body or in form of an at least partly solidified layer.

10. Method according to one of the preceding claims, wherein the method is carried out by a neural network, a statistical evaluation scheme or an evaluation algorithm having a classifier, which carries out the step of assigning the thermal resistance (R1 , R2) or an equivalent value to the defective area (A1 , A2).

11 . Method according to claim 10, wherein the classifier distinguishes between distinct distances (d1 , d2) between the spots (S1 , S2) and/or between distinct aspect ratios of the defective area (A1 , A2).

12. Method according to claim 10 or 11 , wherein the method is carried out by a neural network, and wherein the neural network is a neural network trained on thermal simulations or captured thermal images of defective areas (A1 , A2) having distinct distances (d1 , d2) between the spots (S1 , S2) of the defective area (A1 , A2) and/or trained on thermal simulations or captured thermal images of defective areas (A1 , A2) having distinct aspect ratios, and wherein the step of assigning is carried by the neural network, which predicts the thermal resistance (R1 , R2) of the defective area (A1 , A2).

13. Artificial neural network adapted to carry out the method of one of claims 1 - 12, the neural network having a classifier distinguishing between distinct distances (d1 , d2) between the spots (S1 , S2) and/or between distinct aspect ratios of the defective area (A1 , A2).