Deterioration diagnosis method and deterioration diagnostic device

Abstract

To provide a deterioration diagnosis method and a deterioration diagnostic device capable of appropriately detecting deterioration of a power module.SOLUTION: The deterioration diagnosis method for diagnosing deterioration of a power module including a semiconductor element and a copper pattern bonded to the semiconductor element using a sintered material, includes: a first step of obtaining an X-ray CT image of the power module actually used for a prescribed period with an X-ray CT apparatus; and a second step of diagnosing deterioration of the power module actually used on the basis of the image and an X-ray CT image obtained by imaging wavy deformation in the copper pattern generated when a power module having the same specification as the power module is subjected to repeated temperature changes.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to a degradation diagnosis method and a degradation diagnosis device.

Background Art

[0002] Patent Document 1 describes the following power module. That is, the power module described in Patent Document 1 includes a first solder portion that joins a first semiconductor chip and a wiring portion, and a second solder portion that is more likely to crack with respect to the thermal history than the first solder portion and joins a second semiconductor chip and the wiring portion. According to this power module, for example, when a crack occurs in the second solder portion, it can be presumed that the possibility of a crack occurring in the first solder portion has increased, and it is possible to detect a sign of a crack occurring in the first solder portion.

[0003] In addition, Non-Patent Document 1 discloses the following matters related to the present disclosure. That is, as a result of evaluating a sintered joint using silver nanoparticles in a power module by a thermal shock cycle test, it is described that due to the large deformation of the copper electrode joined to the sintered joint, microvoids in the joint expand, and in some parts with large deformation, the microvoids may connect to form a crack penetrating between the semiconductor chip and the substrate.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Non-Patent Documents

[0005]

Non-Patent Document 1

SUMMARY OF THE INVENTION

PROBLEMS TO BE SOLVED BY THE INVENTION

[0006] In the power module described in Patent Document 1, for example, when the usage environment such as the peripheral temperature of the second solder part is more severe than that of the first solder part, even if a crack occurs in the second solder part, it is possible that the possibility of a crack occurring in the first solder part does not necessarily increase, and there is a problem that a sign of a crack occurring in the first solder part may not be appropriately detected.

[0007] The present disclosure has been made to solve the above problems, and an object thereof is to provide a degradation diagnosis method and a degradation diagnosis device capable of appropriately detecting degradation of a power module.

MEANS FOR SOLVING THE PROBLEMS

[0008] In order to solve the above problems, a degradation diagnosis method according to the present disclosure is a method for diagnosing degradation of a power module including a semiconductor element and a copper pattern joined to the semiconductor element with a sintering material, and includes a first step of acquiring an X-ray CT image of the power module actually used for a predetermined period using an X-ray CT device, and a second step of diagnosing degradation of the actually used power module based on the image and an X-ray CT image obtained by photographing a wavy deformation generated in the copper pattern when a repeated temperature change is applied to a power module having the same specifications as the power module.

[0009] The degradation diagnosis device according to the present disclosure is a device for diagnosing the degradation of a power module including a semiconductor element and a copper pattern joined to the semiconductor element with a sintering material, and includes an acquisition unit that acquires an X-ray CT image of the power module actually used for a predetermined period, taken using an X-ray CT device, and a diagnosis unit that diagnoses the degradation of the actually used power module based on the image and an X-ray CT image that captures the wavy deformation that occurred in the copper pattern when a repeated temperature change was applied to a power module having the same specifications as the power module.

Effect of the Invention

[0010] According to the degradation diagnosis method and the degradation diagnosis device of the present disclosure, the degradation of the power module can be appropriately detected.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Embodiments for Carrying Out the Invention

[0012] Hereinafter, a deterioration diagnosis method and a deterioration diagnosis apparatus according to embodiments of the present disclosure will be described with reference to FIGS. 1 to 14. FIG. 1 is a block diagram showing a configuration example of a deterioration diagnosis apparatus according to an embodiment of the present disclosure. FIG. 2 is a flowchart showing an operation example of the deterioration diagnosis apparatus according to an embodiment of the present disclosure. FIG. 3 is a side view schematically showing a schematic cross-sectional configuration of a power module according to an embodiment of the present disclosure. FIG. 4 is a schematic diagram for explaining a power cycle test according to an embodiment of the present disclosure. FIG. 5 is a schematic diagram for explaining a trained machine learning model according to an embodiment of the present disclosure. FIG. 6 is a perspective view showing a configuration example of a semiconductor element or the like according to an embodiment of the present disclosure. FIGS. 7 to 12 are diagrams showing examples of X-ray CT images according to embodiments of the present disclosure. FIG. 13 is a schematic diagram for explaining another trained machine learning model according to an embodiment of the present disclosure. FIG. 14 is a schematic block diagram showing the configuration of a computer according to an embodiment of the present disclosure. In addition, the same or corresponding configurations in each figure are denoted by the same reference numerals, and the description thereof will be appropriately omitted.

[0013] (Configuration of Deterioration Diagnosis Apparatus) The deterioration diagnosis device 1 according to the embodiment of the present disclosure shown in FIG. 1 can be configured using one or more computers such as a server or a personal computer. Further, part or all of the one or more computers constituting the deterioration diagnosis device 1 may be virtually configured on a network. The deterioration diagnosis device 1 is configured as a functional configuration including a combination of hardware such as a computer constituting the deterioration diagnosis device 1 and peripheral devices of the computer, and software such as a program executed by the computer, and includes an acquisition unit 11, a diagnosis unit 12, a machine learning unit 13, and a storage unit 14. The storage unit 14 stores a learned machine learning model 15. Note that the machine learning unit 13 and the learned machine learning model 15 may be integrally configured, for example.

[0014] The deterioration diagnosis device 1 shown in FIG. 1 is a device for diagnosing the deterioration of a power module (also referred to as a power semiconductor module or the like) 3 actually used in the field. The deterioration diagnosis device 1 diagnoses the deterioration of the power module 3 based on an X-ray CT image 4 of the power module 3 actually used in the field, which is non-destructively taken by an X-ray CT (Computed Tomography) device 2. When taking the X-ray CT image 4, the power module 3 is temporarily removed from equipment such as a moving body or a device such as a vehicle on which the power module 3 is mounted, for example, during a regular inspection, and then mounted on the X-ray CT device 2. Further, when there is no problem with the power module 3 for which the X-ray CT image 4 has been taken, it is returned to the original equipment or the like.

[0015] Figure 3 schematically shows a configuration example of the power module 3 targeted for deterioration diagnosis in this embodiment. The power module 3 shown in Figure 3 includes one or more semiconductor elements 31 and one or more copper patterns 33 each joined to the respective semiconductor element 31 with a sintered material 32. The semiconductor element 31 is a wide bandgap (WBG) power semiconductor element such as SiC (silicon carbide) or GaN (gallium nitride), for example. Each semiconductor element 31 constitutes, for example, a transistor such as a MOSFET (metal oxide semiconductor field effect transistor), IGBT (insulated gate bipolar transistor), or bipolar transistor, or a diode. The sintered material 32 is a joining member and is, for example, a nanoparticle sintered material such as Ag (silver) or Cu (copper). The copper pattern 33 is, for example, a copper electrode formed on the surface of the insulating substrate 35. The copper pattern 34 and the copper pattern 36 are also, for example, copper electrodes formed on the surface of the insulating substrate 35. The heat dissipation member 37 is inserted between the copper pattern 36 and the base plate 38 and is a member that ensures heat dissipation from the copper pattern 36 to the base plate 38. A terminal 41 is joined to the copper pattern 33. The copper pattern 34 is connected to the upper surface electrode of the semiconductor element 31 with a wire 40 and is joined to a terminal 42. A case 39 is fitted to the base plate 38. The case 39 is filled with a sealing material 43. Note that the power module 3 can include configurations not shown in the figure, such as a control circuit, a protection circuit, and a temperature sensor.

[0016] The acquisition unit 11 shown in Figure 1 acquires, for example, an X-ray CT image 4 of the power module 3 actually used in the field for a predetermined period, which was taken using the X-ray CT apparatus 2, by using a communication line, a detachable recording medium, or the like. There is no limitation on the predetermined period, and it can be set appropriately according to the usage environment of the power module 3, for example.

[0017] The diagnostic unit 12 diagnoses the degradation of the power module 3 (hereinafter also referred to as the power module 3) to be diagnosed that has actually been used in the field for a predetermined period, based on the X-ray CT image 4 of the power module 3 and the X-ray CT image of the wavy deformation generated in the copper pattern 33 when repeated temperature changes were applied to the power module 3 having the same specifications as the power module 3. The wavy deformation generated in the copper pattern 33 is the deformation generated in the copper pattern 33 due to thermal stress caused by repeated temperature changes, and includes, for example, wavy deformation, uneven deformation, expansion of voids (holes), crack progression, and the like. Further, the repeated temperature changes applied to the power module 3 can be applied, for example, by a power cycle test or by a temperature cycle test. In the present embodiment, as an example, repeated temperature changes are applied to the power module 3 by a power cycle test. The power cycle test is a test in which a constant current is repeatedly applied and interrupted to the power module 3.

[0018] FIG. 4 shows an example of the time change of the junction temperature (bonding temperature) of the semiconductor element 31 and the time change of the energizing current flowing through the semiconductor element 31 in the power cycle test. The time Tcycle is the time of one cycle, a constant (substantially constant) current Ion is applied for a time Ton within the time Tcycle, and the energizing current is interrupted during the time Toff.

[0019] In the present embodiment, the diagnostic unit 12 uses the learned machine learning model 15 to compare the X-ray CT image 4 of the power module 3 with the X-ray CT image (referred to as the X-ray CT image with undulation) of the power module 3 in a state where degradation has occurred by a power cycle test, and calculates the degree of degradation based on the degree of coincidence between the X-ray CT image 4 of the power module 3 and the X-ray CT image with undulation. Note that in the present embodiment, the degree of degradation means the presence or absence of degradation or the degree of degradation.

[0020] As shown in FIG. 5, the learned machine learning model 15 is a model machine-learned by the machine learning unit 13. Taking the X-ray CT image 4 of the power module 3 as the first image, and taking the X-ray CT images of the power module 3 with the same specification given repeated temperature changes as the second images, a plurality of second images respectively labeled with information indicating the presence or absence or the degree of the wavy deformation generated in the copper pattern 33 are used as the teacher data 5 for machine learning. The first image is input, and a value indicating the presence or absence or the degree of deterioration of the power module 3 (a value representing the degree of deterioration) is output. The value representing the degree of deterioration can be, for example, a value representing the probability (degree of coincidence) of being determined to be an image with deterioration (coinciding with the image with deterioration) when the second image is labeled with either the presence or absence of deterioration. Alternatively, when the second image is labeled with a value representing the degree of the wavy deformation (for example, an index, etc.), the degree of deterioration can be, for example, a value representing the degree of deformation.

[0021] Note that the learned machine learning model 15 is, for example, a model composed of neural networks. The weight coefficients between the neurons of each layer of the neural network are optimized by machine learning so that a solution (degree of deterioration) obtained for a large number of input data (for example, each data representing each pixel value of an image) is output. The learned machine learning model 15 is composed of, for example, a combination of a program that performs operations from input to output and the weight coefficients (parameters) used in the operations.

[0022] In the example shown in FIG. 5, the training data 5 is an X-ray CT image in which undulating deformation is confirmed after performing a power cycle test for a predetermined cycle, and is an X-ray CT image 51 with undulation that is labeled with a value indicating the presence or degree of undulation, and an X-ray CT image in which undulating deformation is not confirmed, and is an X-ray CT image 52 without undulation that is labeled as having no undulation (the degree of undulation is zero). Note that the X-ray CT image 51 with undulation and the X-ray CT image 52 without undulation include images with different numbers of cycles of the power cycle test. Also, the labeling of each image can be performed, for example, according to the judgment and instruction of an operator, or automatically or semi-automatically using existing image processing techniques.

[0023] FIGS. 7 to 12 show examples of X-ray CT images taken after performing a power cycle test for a predetermined cycle. Note that the correspondence of directions in FIGS. 6 and 7 to 12 is indicated by the arrows XYZ. FIG. 7 shows an X-ray CT image 4a of the region A1 (the surface region of the copper pattern 33) shown in FIG. 6. In the X-ray CT image 4a, no undulating deformation is observed in the copper pattern 33. FIG. 8 shows an X-ray CT image 4b of the region A1 shown in FIG. 6. In the X-ray CT image 4b, an undulating deformation RA11 is observed in the copper pattern 33. FIG. 9 shows an X-ray CT image 4c of the region A1 shown in FIG. 6. In the X-ray CT image 4c, an undulating deformation RA12 is observed in the copper pattern 33.

[0024] FIG. 10 shows an X-ray CT image 4d of the region A2 shown in FIG. 6. In the X-ray CT image 4d, no undulating deformation is observed in the copper pattern 33. FIG. 11 shows an X-ray CT image 4e of the region A2 shown in FIG. 6. In the X-ray CT image 4e, an undulating deformation RA21 is observed in the copper pattern 33. FIG. 12 shows an X-ray CT image 4f of the region A3 shown in FIG. 6. In the X-ray CT image 4f, an undulating deformation RA31 is observed in the copper pattern 33.

[0025] The diagnosis unit 12 inputs the X-ray CT image 4 of the power module 3 into the learned machine learning model 15, compares the value representing the degree of degradation output by the learned machine learning model 15 with a predetermined threshold value, and determines that the power module 3 is degraded when the value representing the degree of degradation exceeds the threshold value.

[0026] (Operation example of the degradation diagnosis device) FIG. 2 shows an operation example of the degradation diagnosis device 1 shown in FIG. 1. It is assumed that the learned machine learning model 15 machine-learned by the machine learning unit 13 as described with reference to FIG. 5 is already stored in the degradation diagnosis device 1.

[0027] In the operation example shown in FIG. 2, first, for example, according to the operation of an operator, an X-ray CT image 4 of the power module 3 actually used for a predetermined period is taken using the X-ray CT device 2, and the taken X-ray CT image 4 is acquired by the acquisition unit 11 (step S1).

[0028] Next, based on the acquired X-ray CT image 4 and an X-ray CT image (wavy X-ray CT image 51) of the wavy deformation generated in the copper pattern 33 when a power cycle test is repeatedly performed on the power module 3 of the same specification to give a repeated temperature change, the diagnosis unit 12 diagnoses the degradation of the actually used power module 3 (step S2).

[0029] According to the above operation, the degradation diagnosis device 1 of the present embodiment can detect degradation based on the initial stage of degradation of the sintered material 32 or the wavy deformation generated in the copper pattern 33 that occurs as a sign of degradation of the sintered material 32. Therefore, the degradation of the power module 3 can be appropriately detected.

[0030] (Another configuration example of the learned machine learning model) Regarding the present disclosure, when a power cycle test was conducted, it was confirmed that when the energization current was constant, the temperature of the semiconductor element 31 increased according to the number of cycles. As an example of the temperature increase mode, when the same current was applied, the maximum temperature of the semiconductor element 31 was 225°C immediately after the start of the test, 250°C at 1000 cycles, and 270°C at 2000 cycles. From this phenomenon, it is inferred that the progress of the deterioration of the power module 3 with the increase in the number of cycles affects the temperature increase of the semiconductor element 31. Therefore, regarding the machine learning model for calculating the degree of deterioration of the power module 3, by using the temperature of the semiconductor element 31 in addition to the X-ray CT image as an explanatory variable of the machine learning model, it is expected to more appropriately calculate the degree of deterioration, which is the target variable.

[0031] FIG. 13 shows a model (as the learned machine learning model 15a) in which the temperature of the semiconductor element is added to the input of the learned machine learning model 15 described above. The learned machine learning model 15a is a model machine-learned by the machine learning unit 13. When the X-ray CT image 4 of the power module 3 is used as the first image and the X-ray CT image of the same-specification power module 3 given repeated temperature changes is used as the second image, information indicating the presence or absence or degree of the wavy deformation generated in the copper pattern 33 and information related to the temperature of the semiconductor element 31 obtained during the temperature change are respectively labeled for a plurality of second images (image 51a or image 52a) as teacher data 5a and machine-learned. The first image and information related to the temperature of the semiconductor element obtained during actual use are input, and a value indicating the presence or absence or degree of deterioration of the actually used power module is output.

[0032] Note that the information related to the temperature of the semiconductor element obtained during actual use can be, for example, a value representing the maximum temperature (or average temperature) of, for example, the junction temperature of the semiconductor element 31 calculated from the temperature values obtained during a predetermined period immediately before diagnosis using a temperature sensor in the power module 3 or a temperature sensor in a unit such as an inverter including the power module 3.

[0033] According to the learned machine learning model 15a, in addition to the X-ray CT image 4, it is possible to appropriately detect the degradation of the power module 3 in consideration of information related to the temperature of the semiconductor element 31.

[0034] (Function and effect) According to the degradation diagnosis method and the degradation diagnosis device of the present embodiment, an X-ray CT image 4 of the power module 3 actually used for a predetermined period is acquired using the X-ray CT device 2, and based on the image and an X-ray CT image 4 that captures the wavy deformation that occurred in the copper pattern 33 when repeated temperature changes were applied to the power module 3 having the same specifications as the power module 3, the degradation of the actually used power module 3 is diagnosed, so that the degradation of the power module 3 can be appropriately detected.

[0035] (Other embodiments) As described above, the embodiments of the present disclosure have been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and design changes and the like within the scope not departing from the gist of the present disclosure are also included. For example, the calculation of the degree of degradation (the calculation of the degree of coincidence with an image including degradation) may be configured to calculate the degree of coincidence with an X-ray CT image with undulations using an image processing technique such as pattern matching instead of or in combination with a configuration using a learned machine learning model.

[0036] 〈Computer configuration〉 FIG. 14 shows a schematic block diagram showing the configuration of a computer according to at least one embodiment. The computer 90 includes a processor 91, a main memory 92, a storage 93, and an interface 94. The above deterioration diagnosis device 1 is implemented in a computer 90. And the operations of the above-described respective processing units are stored in a storage 93 in the form of a program. The processor 91 reads out the program from the storage 93, expands it in the main memory 92, and executes the above processing according to the program. Further, the processor 91 secures a storage area corresponding to each of the above-described storage units in the main memory 92 according to the program.

[0037] The program may be for realizing a part of the functions to be exhibited by the computer 90. For example, the program may exhibit functions by combination with other programs already stored in the storage or combination with other programs implemented in other devices. In other embodiments, the computer may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to or instead of the above configuration. Examples of the PLD include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), FPGA (Field Programmable Gate Array), and the like. In this case, part or all of the functions realized by the processor may be realized by the integrated circuit.

[0038] Examples of the storage 93 include HDD (Hard Disk Drive), SSD (Solid State Drive), magnetic disk, magneto-optical disk, CD-ROM (Compact Disc Read Only Memory), DVD-ROM (Digital Versatile Disc Read Only Memory), semiconductor memory, and the like. The storage 93 may be an internal medium directly connected to the bus of the computer 90, or may be an external medium connected to the computer 90 via the interface 94 or a communication line. Further, when this program is distributed to the computer 90 via a communication line, the computer 90 that has received the distribution may expand the program in the main memory 92 and execute the above processing. In at least one embodiment, the storage 93 is a non-transitory tangible storage medium.

[0039] <Appendix> The degradation diagnosis method and the degradation diagnosis apparatus 1 described in the embodiments of the present disclosure are understood as follows, for example.

[0040] (1) The degradation diagnosis method according to the first aspect is a method for diagnosing the degradation of the power module 3 including the semiconductor element 31 and the copper pattern 33 joined to the semiconductor element 31 with the sintered material 32, including a first step (S1) of acquiring an X-ray CT image 4 of the power module 3 actually used for a predetermined period using the X-ray CT apparatus 2, and a second step (S2) of diagnosing the degradation of the actually used power module 3 based on the image and an X-ray CT image 4 obtained by photographing a wavy deformation generated in the copper pattern 33 when a repeated temperature change is applied to the power module 3 having the same specifications as the power module 3. According to this aspect and the following aspects, since the degradation can be detected based on the wavy deformation generated in the copper pattern 33 at the initial stage of the degradation of the sintered material 32 or as a sign of the degradation of the sintered material 32, the degradation of the power module 3 can be appropriately detected.

[0041] (2) The degradation diagnosis method according to the second aspect is the degradation diagnosis method of (1), wherein the X-ray CT image of the power module actually used during the predetermined period is used as the first image, and the X-ray CT image of the power module of the same specification repeatedly subjected to the temperature change is used as the second image. In the second step (S2), a plurality of the second images in which information indicating the presence or absence or the degree of the wavy deformation generated in the copper pattern 33 is labeled respectively are used as the training data 5 for machine learning. Using the learned machine learning model 15 that inputs the first image and outputs a value indicating the presence or absence or the degree of degradation of the power module, the degradation of the actually used power module is diagnosed. According to this aspect, by appropriately generating the learned machine learning model 15, the degradation of the power module 3 can be appropriately detected.

[0042] (3) The degradation diagnosis method according to the third aspect is the degradation diagnosis method of (2), wherein the temperature change is given by repeatedly energizing and interrupting a constant current to the power module 3 of the same specification. The machine learning model (learned machine learning model 15a) is machine-learned using a plurality of the second images in which information indicating the presence or absence or the degree of the wavy deformation generated in the copper pattern 33 and information related to the temperature of the semiconductor element 31 obtained during the temperature change are labeled respectively as the training data 5a. The first image and the information related to the temperature of the semiconductor element 31 acquired during actual use are input, and a value indicating the presence or absence or the degree of degradation of the actually used power module 3 is output. According to this aspect, in addition to the X-ray CT image 4, considering the information related to the temperature of the semiconductor element 31, the degradation of the power module 3 can be appropriately detected.

Explanation of Reference Numerals

[0043] 1…Degradation diagnosis device 11…Acquisition unit 12…Diagnosis unit 13…Machine learning unit 14…Storage unit 15, 15a…Learned machine learning model 3…Power module 31…Semiconductor element 32…Sintered material 33, 34, 36…Copper pattern 35…Insulating substrate 37…Heat dissipation member 38…Base plate 4, 4a~4f…X-ray CT images 5, 5a…Teacher data

Claims

1. A method for diagnosing degradation of a power module including a semiconductor element and a copper pattern joined to the semiconductor element with a sintering material, comprising: a first step of acquiring an X-ray CT image of the power module actually used for a predetermined period using an X-ray CT apparatus; a second step of diagnosing degradation of the actually used power module based on the image and an X-ray CT image obtained by photographing a wavy deformation generated in the copper pattern when a repeated temperature change is applied to a power module having the same specifications as the power module; The degradation diagnosis method including the above.

2. When the X-ray CT image of the power module actually used for the predetermined period is defined as a first image, and the X-ray CT image of the power module having the same specifications to which the temperature change is repeatedly applied is defined as a second image, in the second step, a plurality of the second images each labeled with information indicating the presence or absence or degree of the wavy deformation generated in the copper pattern are used as teacher data for machine learning, the first image is input, and a learned machine learning model that outputs a value indicating the presence or absence or degree of degradation of the power module is used to diagnose the degradation of the actually used power module. The degradation diagnosis method according to claim 1.

3. The temperature change is given by repeatedly energizing and interrupting a constant current to the power module having the same specifications, and the machine learning model is machine-learned using, as teacher data, a plurality of the second images each labeled with information indicating the presence or absence or degree of the wavy deformation generated in the copper pattern and information related to the temperature of the semiconductor element obtained during the temperature change, inputs the first image and information related to the temperature of the semiconductor element acquired during actual use, and outputs a value indicating the presence or absence or degree of degradation of the actually used power module. The deterioration diagnosis method according to claim 2.

4. An apparatus for diagnosing deterioration of a power module including a semiconductor element and a copper pattern joined to the semiconductor element with a sintering material, an acquisition unit configured to acquire an X-ray CT image of the power module actually used for a predetermined period, which is taken using an X-ray CT apparatus; and a diagnosis unit configured to diagnose deterioration of the actually used power module based on the image and an X-ray CT image of a wavy deformation generated in the copper pattern when a repeated temperature change is applied to a power module having the same specifications as the power module. A deterioration diagnosis apparatus comprising the above.

Images