Damage assessment system, damage assessment device, damage assessment method, and computer program

Abstract

To provide a damage determination system, a damage determination device, a damage determination method, and a computer program that can detect the deterioration in a bonding part inside a semiconductor device with high accuracy.SOLUTION: A damage determination system includes one or more semiconductor devices, a plurality of sensors, and a damage determination unit. The one or more semiconductor devices include one or more bonding parts that bond a heat dissipation substrate to another constituent member. The sensors detect elastic waves generated in the one or more bonding parts of the one or more semiconductor devices. The damage determination unit determines the damage of the one or more bonding parts on the basis of the elastic wave that satisfies the condition based on temperature among a plurality of elastic waves detected by the respective sensors.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] An embodiment of the present invention relates to a damage determination system, a damage determination device, a damage determination method, and a computer program. [Background technology]

[0002] In industrial systems and infrastructure, technologies are known for monitoring the condition of structures in order to take measures before reliability and safety problems occur. For example, one known technology for monitoring the condition of structures is the acoustic emission (AE) method, which uses highly sensitive sensors to detect elastic waves generated by the initiation and propagation of fatigue cracks or material deformation.

[0003] Semiconductor devices such as power modules repeatedly undergo power cycles due to power on / off operations and thermal cycles due to changes in environmental temperature. Therefore, the joints under semiconductor chips and the joints under mounting substrates such as DBC (Direct Bonding Copper) boards in semiconductor devices are susceptible to thermal fatigue due to differences in the linear expansion coefficients between the components. Due to the effects of such thermal fatigue, semiconductor devices are prone to failures such as fractures in joints and disconnections in wiring. Conventionally, semiconductor device failure signs have been detected by detecting voltage fluctuations and thermal resistance fluctuations. However, conventional methods have sometimes been unable to accurately detect deterioration of joints within semiconductor devices. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-144131 [Patent Document 2] Patent No. 6742448 [Non-patent literature]

[0005] [Non-Patent Document 1] Chanyang Choe, Chuantong Chen, Shijo Nagao, Katsuaki Suganuma, “Real-Time Acoustic Emission Monitoring of Wear-Out Failure in SiC Power Electronic Devices During Power Cycling Tests”, IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 36, NO. 4, (2021). [Non-Patent Document 2] H. C. Huang, T. Y. Hung, S. Y. Lin, K. H. Liao, C. C. Wang, K. N. Chiang, “Reliability Assessment of the Temperature Profiles Effect on the Power Module”, 37th National Conference on Theoretical and Applied Mechanics (37th NCTAM 2013) & The 1st International Conference on Mechanics (1st ICM), Procedia Engineering 79 (2014), p.333-338. [Non-Patent Document 3] Akira Kano, Tomoko Monda, Tomoyuki Suzuki, Hideaki Uehara, Tomoya Fumikura, Kenji Hirohata, “PROGNOSTIC HEALTH MONITORING METHOD FOR THERMAL FATIGUE FAILURE OF POWER MODULES BASED ON FINITE ELEMENT METHOD-BASED LAGRANGIAN NEURAL NETWORKS”, Proceedings of the ASME 2021 International Mechanical Engineering Congress and Exposition, IMECE2021-70783, (2021) Summary of the Invention [Problem to be solved by the invention]

[0006] The problem to be solved by the present invention is to provide a damage determination system, a damage determination device, a damage determination method, and a computer program that can detect deterioration of a joint inside a semiconductor device with high accuracy. [Means for solving the problem]

[0007] A damage determination system according to an embodiment includes one or more semiconductor devices, a plurality of sensors, and a damage determination unit. The one or more semiconductor devices have one or more bonding portions that bond a heat dissipation substrate to another component. The plurality of sensors detect elastic waves generated at the one or more bonding portions of the one or more semiconductor devices. The damage determination unit determines, from among the plurality of elastic waves detected by each of the plurality of sensors, The condition is that the retention time, which indicates the time during which a state in which the temperature change is small, is equal to or greater than the first threshold value and equal to or less than the second threshold value, and the temperature is equal to or greater than the reference temperature. Temperature-based No. 1 conditions or a second condition based on temperature, which requires that the temperature change is small and that the temperature is equal to or higher than the reference temperature. Damage to the one or more joints is determined based on the elastic waves that satisfy the above condition. [Brief explanation of the drawings]

[0008] [Figure 1]1 is a schematic cross-sectional view of a semiconductor device according to an embodiment. [Figure 2] FIG. 1 is a diagram showing an example of the configuration of a system used in a temperature cycle test. [Figure 3] FIG. 10 is a graph showing the relationship between temperature during a thermal cycle, the number of detected elastic waves, and the peak amplitude obtained by a temperature cycle test. [Figure 4] FIG. 1 is a diagram showing the configuration of a damage determination system according to a first embodiment. [Figure 5] FIG. 2 is a diagram showing an example of the arrangement of a plurality of sensors in the first embodiment. [Figure 6] FIG. 2 is a schematic block diagram illustrating the functions of a signal processing unit according to the first embodiment. [Figure 7] FIG. 2 is a schematic block diagram illustrating the function of an AFE according to the first embodiment. [Figure 8] FIG. 2 is a schematic block diagram illustrating the functions of a control unit according to the first embodiment. [Figure 9] 4A and 4B are diagrams showing an example in which damage occurs in a joint under a heat dissipation substrate in the first embodiment, and an image of the location of an elastic wave source. [Figure 10] FIG. 3 is a sequence diagram showing the flow of damage assessment processing performed by the damage assessment system according to the first embodiment. [Figure 11] FIG. 3 is a sequence diagram showing the flow of damage assessment processing performed by the damage assessment system according to the first embodiment. [Figure 12] FIG. 10 is a diagram showing the configuration of a damage determination system according to a second embodiment. [Figure 13] FIG. 10 is a sequence diagram showing the flow of damage assessment processing performed by the damage assessment system according to the second embodiment. [Figure 14] FIG. 10 is a sequence diagram showing the flow of damage assessment processing performed by the damage assessment system according to the second embodiment. [Figure 15] FIG. 10 is a diagram showing the configuration of a damage determination system according to a third embodiment. [Figure 16] FIG. 11 is a sequence diagram showing the flow of damage assessment processing performed by a damage assessment system according to a third embodiment. [Figure 17]FIG. 11 is a sequence diagram showing the flow of damage assessment processing performed by a damage assessment system according to a third embodiment. [Figure 18] FIG. 10 is a diagram showing the configuration of a damage determination system according to a fourth embodiment. [Figure 19] 10A and 10B are diagrams for explaining a correction method performed by a correction unit in the fourth embodiment. [Figure 20] FIG. 10 is a diagram showing a configuration of an inverter device according to a fifth embodiment. [Figure 21] FIG. 10 is a diagram showing the configuration of a damage determination system according to a fifth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, a damage determination system, a damage determination device, a damage determination method, and a computer program according to embodiments will be described with reference to the drawings.

[0010] (overview) First, an overview of a damage assessment system according to an embodiment will be described. The damage assessment system according to an embodiment applies an acoustic emission (AE) method to detect damage to joints within a semiconductor device, such as a power module. That is, the damage assessment system according to an embodiment detects damage to joints within a semiconductor device nondestructively. FIG. 1 shows a schematic cross-sectional view of a semiconductor device 1 according to an embodiment. As shown in FIG. 1, the semiconductor device 1 is mounted on a heat sink 2. The heat sink 2 is a component that absorbs heat generated in the semiconductor device 1 and dissipates (radiates) the absorbed heat into the air, thereby cooling the semiconductor device 1. A lubricant G, such as grease, is filled between the semiconductor device 1 and the heat sink 2. This prevents air or other gaps from forming between the semiconductor device 1 and the heat sink 2, allowing heat generated by the semiconductor device 1 to be smoothly transferred to the heat sink 2, thereby cooling the semiconductor device 1.

[0011] The semiconductor device 1 includes one or more power semiconductors 3, a heat dissipation substrate 4, and a Cu base plate 5. The heat dissipation substrate 4 is bonded onto the Cu base plate 5 via a bonding portion 8, and the power semiconductor 3 is bonded onto the heat dissipation substrate 4 via a bonding portion 9. In this manner, the bonding portions 8 and 9 bond the heat dissipation substrate 4 to other components (for example, the power semiconductor 3 or the Cu base plate 5). The bonding portions 8 and 9 may be made of adhesive, solder, or the like. The semiconductor device 1 in the embodiment is not limited to the structure shown in FIG. 1 as long as it is a device configured with a layered structure via one or more bonding portions. The semiconductor device 1 does not necessarily have to include the Cu base plate 5.

[0012] The power semiconductor 3 is a semiconductor package such as an SOP (Small Outline Package). The power semiconductor 3 may be any semiconductor package used for any purpose as long as it is a semiconductor package. The heat dissipation substrate 4 is a substrate used to dissipate heat generated by the power semiconductor 3. The heat dissipation substrate 4 is, for example, a DBC substrate or a DBA (Direct Bonding Aluminum) substrate. In the following explanation, a case where the heat dissipation substrate 4 is a DBC substrate will be explained as an example. The heat dissipation substrate 4 is composed of Cu 6-1, 6-2 and a ceramic insulating substrate 7.

[0013] Cu 6-1 and 6-2 are copper and are bonded to each surface of the ceramic insulating substrate 7. For example, as shown in FIG. 1 , Cu 6-1 is bonded to the first surface (upper portion) of the ceramic insulating substrate 7, and Cu 6-2 is bonded to the second surface (lower portion) of the ceramic insulating substrate 7. The ceramic insulating substrate 7 is an insulating circuit board made of ceramics such as aluminum nitride ceramics or silicon nitride ceramics. The Cu base plate 5 is a heat sink for dissipating heat generated by the power semiconductor 3 to the heat sink 2. The Cu base plate 5 is fixed to the heat sink 2 with bolts B or the like. This allows the semiconductor device 1 to be fixed to the heat sink 2. Instead of the Cu base plate 5, another heat sink may be used for the semiconductor device 1.

[0014] The joints 8 and 9 shown in FIG. 1 are susceptible to thermal fatigue due to differences in the linear expansion coefficients of the components caused by temperature loads (power cycles) associated with device operation and environmental temperature changes. It is known that long cycles of power on / off operation cause heat generated by the elements to propagate throughout the semiconductor device 1, primarily damaging the joints 8 below the heat dissipation substrate 4. The inventors conducted a temperature cycle test to measure elastic waves during the thermal cycle. The temperature cycle test simulates an environment in which the temperature repeatedly changes due to the external environment or self-heating of the electronic components, and applies thermal stress caused by temperature changes to verify resistance. By heating at ambient temperature, the entire semiconductor device 1 is heated, making it possible to conduct tests under conditions similar to those of a long-period power cycle.

[0015] The configuration of the elastic wave measurement system used in the temperature cycle test is shown in Figure 2. Figure 2 is a diagram showing an example of the configuration of the system used in the temperature cycle test. A semiconductor device 1, four sensors 10, and a temperature sensor 20 were arranged on an aluminum plate AL. Specifically, four sensors 10 (sensors CH1 to CH4) were arranged on the aluminum plate AL so as to surround the semiconductor device 1, and a temperature sensor 20 was arranged near the semiconductor device 1. The sensors 10 detect elastic waves generated from the semiconductor device 1. Each sensor 10 was connected to a signal processing unit 40 by a cable via an amplifier 30, and the temperature sensor 20 was connected directly to the signal processing unit 40 by a cable. The signal processing unit 40 measured elastic waves and the measurement results were displayed on a damage determination device 50.

[0016] Figure 3 shows the relationship between temperature, the number of detected elastic waves, and peak amplitude during a thermal cycle obtained through a temperature cycle test. In Figure 3, the horizontal axis represents time, the left vertical axis represents temperature, and the right vertical axis represents peak amplitude. Each point P1 in Figure 3 represents a detected elastic wave, and each line L1 represents the change in temperature measured by the temperature sensor 20. Note that the period from rise to fall on one line L1 represents the temperature history of one thermal cycle. As shown in Figure 3, elastic waves were detected during all thermal cycles. This is thought to be due to secondary AE generated by material deformation, friction, or wear, rather than primary AE generated when cracks occur. Here, the temperature history of one cycle is divided into six stages, from stage A to stage F, and the trends in the number of detected elastic waves and peak amplitude for each stage are explained.

[0017] Process A Temperature rise process 1: Temperature rises rapidly (large temperature change) Process B Temperature increase process 2: Temperature increases slowly Process C Temperature holding time (high temperature side): Small temperature change Process D: Temperature drop process 1: Temperature drops rapidly (large temperature change) Process E: Temperature drop process 2: The temperature drops slowly Process F: Temperature holding time (low temperature side): Small temperature change

[0018] Figure 3 shows that when the temperature changes rapidly (when the temperature change is large), as in processes A and D, the peak amplitude is large and many elastic waves are detected. In processes B and E, elastic waves are detected, but the peak amplitude is small. Furthermore, it can be seen that fewer elastic waves are detected in process E, which is at a low temperature. In processes C and F, elastic waves are detected, but the peak amplitude is small. Furthermore, it can be seen that fewer elastic waves are detected in process F, which is at a low temperature. Elastic waves are generated not only when cracks occur and propagate, but also due to elastic or inelastic deformation. Since elastic waves with large amplitudes are detected when the temperature change is large in processes A and D, it is thought that in processes A and D, elastic waves generated by elastic deformation of the heat dissipation substrate 4 and Cu base plate 5, and elastic and inelastic deformation of the joints 8 and 9 due to mismatches in the linear expansion coefficients of each component are mainly detected.

[0019] On the other hand, elastic waves are also generated in processes C and F, even though the impact of mismatch in the linear expansion coefficients of the components is smaller than in processes A and D. Here, a detailed explanation will be given using Non-Patent Document 2. On page 337 of Non-Patent Document 2, a diagram showing the cumulative creep strain of Sn-Ag solder versus time obtained by FEM analysis is shown (for example, Fig. 7). In Fig. 7 of Non-Patent Document 2, dwell at 125 corresponds to process C, and dwell at -40 corresponds to process F.

[0020] In Figure 7 of Non-Patent Document 2, slight inelastic strain (creep strain) occurs at dwell times of 125 and -40. This is due to the creep characteristics of the joining material (e.g., Sn-Ag-Cu solder), and the higher the temperature, the more pronounced the inelastic deformation due to creep. Thus, the elastic waves detected during processes C and F, where the temperature change is small, are thought to correspond to the amount of inelastic deformation due to creep of the joining material. This suggests that by measuring elastic waves only during the high-temperature holding time when inelastic deformation due to creep becomes significant, it is possible to eliminate noise elastic waves generated from components other than joints 8 and 9 and concentrate on detecting elastic waves generated from joints 8 and 9.

[0021] Therefore, in the damage determination system of the embodiment, by starting measurement of elastic waves during the period of process C shown in Fig. 3, it becomes possible to detect with high accuracy the deterioration of the bonding portions 8 and 9 inside the semiconductor device 1. The specific configuration will be described below.

[0022] (First embodiment) 4 is a diagram showing the configuration of a damage determination system 100 in the first embodiment. The damage determination system 100 is used to determine damage that has occurred at the bonding portions 8 and 9 inside the semiconductor device 1. Note that the semiconductor device 1 used in the damage determination system 100 may be a device other than a power module.

[0023] The damage assessment system 100 includes a semiconductor device 1, a heat sink 2, a plurality of sensors 10-1 to 10-n (n is an integer equal to or greater than 2), a temperature sensor 20, an operation control device 35, a signal processing unit 40, and a damage assessment device 50. The plurality of sensors 10-1 to 10-n and the signal processing unit 40 are connected by wire. The signal processing unit 40 and the damage assessment device 50 are connected by wire or wirelessly. The temperature sensor 20 and the operation control device 35 are connected by wire. The operation control device 35 and the signal processing unit 40 are connected by wire. In the following description, when there is no need to distinguish between the sensors 10-1 to 10-n, they will be referred to as sensors 10.

[0024] Sensor 10 has a piezoelectric element and detects elastic waves generated from semiconductor device 1. More specifically, sensor 10 detects elastic waves generated due to damage to bonding portions 8 and 9 inside semiconductor device 1. Bonds 8 and 9 gradually peel off due to repeated thermal cycles. Sensor 10 detects elastic waves generated when damage such as interfacial peeling occurs. Sensor 10 may be installed so as to surround bonding portion 8 or bonding portion 9 inside semiconductor device 1, and may be installed, for example, on heat sink 2, heat dissipation substrate 4, or Cu base plate 5.

[0025] 5 is a diagram showing an example of the arrangement of a plurality of sensors 10 in the first embodiment. When a plurality of sensors 10 are arranged to surround the semiconductor device 1 as shown in FIG. 5, if an elastic wave is generated due to damage to the bonding portion 8, the generated elastic wave reaches each sensor 10 via the Cu base plate 5 and the heat sink 2. If an elastic wave is generated due to damage to the bonding portion 9, the generated elastic wave reaches each sensor 10 via the heat dissipation substrate 4, the bonding portion 8, the Cu base plate 5, and the heat sink 2.

[0026] Returning to FIG. 4, the explanation will continue. Sensor 10 converts the detected elastic waves into an electrical signal. Sensor 10 outputs the electrical signal to signal processing unit 40. Sensor 10 uses a piezoelectric element having sensitivity in the range of 10 kHz to 1 MHz, for example. Sensor 10 comes in various types, such as a resonance type that has a resonance peak within a frequency range and a wideband type that suppresses resonance, but any type of sensor 10 is acceptable. Sensor 10 can detect elastic waves using a voltage output type, a resistance change type, or a capacitance type, but any detection method is acceptable. Note that an acceleration sensor may be used instead of sensor 10. In this case, the acceleration sensor detects elastic waves generated from bonding portions 8 and 9 inside semiconductor device 1. The acceleration sensor then converts the detected elastic waves into a voltage signal by performing the same processing as sensor 10.

[0027] The temperature sensor 20 measures the temperature in the environment in which the semiconductor device 1 is installed. The temperature sensor 20 outputs temperature information, which is the measurement result, to the operation control device 35. The temperature sensor 20 may be newly installed on the heat dissipation substrate 4 or the Cu base plate 5, or a thermistor built into the heat dissipation substrate 4 or the Cu base plate 5 may be used.

[0028] The operation control device 35 controls the operation of the signal processing unit 40 based on the temperature information output from the temperature sensor 20. The operation control device 35 determines whether or not activation conditions for activating the signal processing unit 40 are satisfied based on the temperature indicated by the temperature information, and activates the signal processing unit 40 if the activation conditions are satisfied. For example, if the activation conditions are satisfied, the operation control device 35 outputs an activation signal to the signal processing unit 40 to activate the signal processing unit 40. Activating the signal processing unit 40 means causing the signal processing unit 40 to perform signal processing. In other words, the signal processing unit 40 does not perform signal processing on elastic waves until the activation signal is acquired from the operation control device 35.

[0029] The activation condition is that the retention time t is equal to or greater than the first threshold value Th1 and equal to or less than the second threshold value Th2 (Th1≦t≦Th2), and the temperature T is equal to or greater than the reference temperature T1 (T≧T1). Here, the retention time represents the time during which the temperature change ΔT remains less than the third threshold value Th3. In other words, the retention time t can be considered as the time during which a state with little temperature change is maintained. When the retention time t is equal to or greater than the first threshold value Th1 and equal to or less than the second threshold value Th2, the state corresponds to process C or F shown in FIG. 3. Note that the period during which this activation condition is satisfied corresponds to process C shown in FIG. 3. The activation condition is one aspect of a temperature-based condition.

[0030] As an example, consider a case where the temperature is 60°C at time t1, 61°C at time t2, 64°C at time t3, 70°C at time t4, and 64°C at time t5, and the third threshold Th3 is 5°C. In this case, the temperature changes ΔT are 1°C (temperature change from time t1 to time t2), 3°C (temperature change from time t2 to time t3), 6°C (temperature change from time t3 to time t4), and -6°C (temperature change from time t4 to time t5), respectively. The time during which the temperature change ΔT remains at 5°C is from time t1 to time t3. In this case, the retention time t is 3 seconds, from time t1 to time t3.

[0031] The first threshold Th1 is the lower limit of the elapsed time in the activation conditions, for example, 5 seconds. The first threshold Th2 is the upper limit of the elapsed time in the activation conditions, for example, 3600 seconds. The third threshold Th3 is a value indicating that the temperature change ΔT is small, for example, 5°C. The reference temperature T1 is a temperature that is considered to be high, corresponding to process C shown in FIG. 3, for example, 70°C. Note that the values ​​of the first threshold, second threshold, third threshold, and reference temperature T1 shown above are merely examples, and the values ​​may be changed depending on the usage situation.

[0032] Referring to the above example, the operation control device 35 determines that the activation condition is satisfied when the temperature measured by the temperature sensor 20 is 70°C or higher and the temperature is maintained at 70°C or higher for a period of 5 to 3600 seconds. The operation control device 35 activates the signal processing unit 40 when the maintenance time reaches a first threshold value (e.g., 5 seconds), and stops the signal processing unit 40 when the maintenance time reaches a second threshold value (e.g., 3600 seconds) or when the temperature falls below the reference temperature T1.

[0033] The signal processing unit 40 transitions to an operating state based on an operating signal output from the operation control device 35. The signal processing unit 40 is in an inactive state until transitioning to an operating state. The inactive state is a state in which the signal processing unit 40 is inactive. The inactive state is a state in which power consumption is reduced compared to the operating state. The inactive state may be, for example, a state in which the signal processing unit is activated but does not perform signal processing, a sleep state, or a stopped state in which the power is turned off.

[0034] When the signal processing unit 40 transitions to an operating state, it performs signal processing on the electrical signals of each elastic wave output from each sensor 10. The signal processing performed by the signal processing unit 40 includes, for example, noise removal and parameter extraction. The signal processing unit 40 generates transmission data including the processed digital signal and outputs the generated transmission data to the damage determination device 50. The signal processing unit 40 is configured using an analog circuit or a digital circuit. The digital circuit is realized, for example, by an FPGA (Field Programmable Gate Array) or a microcomputer. The digital circuit may also be realized by a dedicated LSI (Large-Scale Integration). The signal processing unit 40 may also be equipped with a non-volatile memory such as a flash memory or a removable memory.

[0035] The damage determination device 50 identifies the location where damage has occurred in the semiconductor device 1 and estimates the area of ​​the damage, based on the transmission data obtained from the signal processing unit 40.

[0036] 6 is a schematic block diagram showing the functions of the signal processing unit 40 in the first embodiment. The signal processing unit 40 includes a plurality of AFEs (Analog Front Ends) 41, a control unit 42, a communication unit 43, an operation unit 44, and a power supply unit 45.

[0037] The AFE 41 performs filtering and analog-to-digital conversion on the electrical signal output from the sensor 10. The AFE 41 outputs the signal after filtering and analog-to-digital conversion to the control unit .

[0038] The control unit 42 controls the entire signal processing unit 40. The control unit 42 is configured using a processor such as a CPU (Central Processing Unit) and a memory. For example, the control unit 42 generates transmission data based on the signal output from the AFE 41 after filtering and analog-to-digital conversion.

[0039] The communication unit 43 transmits the transmission data generated by the control unit 42 to the damage determination device 50.

[0040] When the operation unit 44 receives an operation signal from the operation control device 35, it puts the control unit 42 and the power supply unit 45 into an operation state. For example, when the control unit 42 is in a sleep state, the operation unit 44 puts the control unit 42 into an operation state so that signal processing is possible in the control unit 42. For example, when the control unit 42 is in a stopped state, the operation unit 44 puts the control unit 42 into an operation state by causing the power supply unit 45 to supply power to the control unit 42. Note that, similarly, when the AFE 41 or the communication unit 43 is not powered on, the operation unit 44 instructs the power supply unit 45 to supply power to the unpowered functional units, causing the power to be supplied.

[0041] The power supply unit 45 supplies power to each functional unit in accordance with instructions from the operation unit 44. Specifically, the power supply unit 45 supplies power to the AFE 41, the control unit 42, the communication unit 43, and the operation unit 44. The power supply unit 45 is a unit that receives power supplied from an external power source, a primary battery, a secondary battery, a solar cell, an energy harvester, etc., and supplies power from there to each functional unit in accordance with instructions from the operation unit 44.

[0042] 7 is a schematic block diagram showing the functions of the AFE 41 in the first embodiment. The AFE 41 is composed of a receiving unit 411, a first filter 412, an analog-to-digital conversion unit 413, and a second filter 414. The receiving unit 411 receives an electrical signal transmitted from the sensor 10. The receiving unit 411 outputs the received electrical signal to the first filter 412. It is assumed that the electrical signal is provided with time information detected by the sensor 10.

[0043] The first filter 412 removes noise from the electrical signal received by the receiving unit 411. For example, the first filter 412 removes, as noise, frequency bands other than a specific frequency band from the electrical signal. The first filter 412 is, for example, a band-pass filter. The first filter 412 outputs the analog signal after noise removal (hereinafter referred to as the "noise-removed analog signal") to the analog-to-digital conversion unit 413.

[0044] The analog-to-digital converter 413 quantizes the noise-removed analog signal output from the first filter 412 to convert the analog signal into a digital signal. The analog-to-digital converter 413 outputs the digital signal to the second filter 414.

[0045] The second filter 414 removes noise from the digital signal output from the analog-to-digital conversion unit 413. The second filter 414 is a filter for removing noise. The second filter 414 outputs the digital signal after the noise removal (hereinafter referred to as the "noise-removed digital signal") to the control unit 42. In the following description, the processing performed in the AFE 41 will be referred to as preprocessing.

[0046] 8 is a schematic block diagram showing the functions of the control unit 42 in the first embodiment. The control unit 42 executes a program to function as an event signal generation unit 421, a feature extraction unit 422, and a transmission data generation unit 423. The programs for realizing the event signal generation unit 421, the feature extraction unit 422, and the transmission data generation unit 423 may be installed in the signal processing unit 40 at the time of shipment, or may be installed separately.

[0047] The event signal generation unit 421 receives the noise-removed digital signal output from the second filter 414. The event signal generation unit 421 generates a gate signal indicating whether the waveform of the received noise-removed digital signal is sustained. The event signal generation unit 421 is realized by, for example, an envelope detector and a comparator. The envelope detector detects the envelope of the noise-removed digital signal. The envelope is extracted, for example, by squaring the noise-removed digital signal and performing predetermined processing (for example, processing using a low-pass filter or a Hilbert transform) on the squared output value. The comparator determines whether the envelope of the noise-removed digital signal is equal to or greater than a predetermined threshold.

[0048] When the envelope of the noise-removed digital signal becomes equal to or greater than a predetermined threshold, the event signal generation unit 421 outputs a first gate signal indicating that the waveform of the noise-removed digital signal is continuing to the feature extraction unit 422. The output of the first gate signal indicates that an event has occurred. On the other hand, when the envelope of the noise-removed digital signal becomes less than the predetermined threshold, the event signal generation unit 421 outputs a second gate signal indicating that the waveform of the noise-removed digital signal is not continuing to the feature extraction unit 422. The output of the second gate signal indicates that the event has ended. ChangeFinder, AIC (Akaike's Information Criterion), or the like may be used to detect the occurrence of an event, i.e., to determine whether the envelope has become equal to or greater than a predetermined threshold.

[0049] The feature extraction unit 422 receives the gate signal output from the event signal generation unit 421 and the noise-removed digital signal output from the second filter 414. Based on the received gate signal and noise-removed digital signal, the feature extraction unit 422 extracts a feature from the noise-removed digital signal when the signal waveform is continuous. The feature is, for example, a parameter such as the amplitude [mV] of the waveform of the noise-removed digital signal, the rise time [usec] of the gate signal, the duration [usec] of the gate signal, the number of zero-cross counts [times] of the noise-removed digital signal, the energy [arb.] of the waveform of the noise-removed digital signal, and the frequency [Hz] of the noise-removed digital signal.

[0050] The amplitude of the noise-reduced digital signal is, for example, the maximum amplitude value in the noise-reduced digital signal. The rise time of the gating signal is, for example, the time it takes for the gating signal to rise from zero and exceed a predetermined value. The duration of the gating signal is, for example, the time from when the gating signal starts to rise until the amplitude becomes smaller than a predetermined value. The zero-cross count of the noise-reduced digital signal is, for example, the number of times the noise-reduced digital signal crosses a reference line passing through zero. The energy of the waveform of the noise-reduced digital signal is, for example, the value obtained by integrating the square of the amplitude at each time point over time. Note that the definition of energy is not limited to the above example, and may be approximated using, for example, the envelope of the waveform.

[0051] The feature extraction unit 422 outputs the extracted feature to the transmission data generation unit 423. When outputting the feature to the transmission data generation unit 423, the feature extraction unit 422 associates a sensor ID with the feature. The sensor ID represents identification information for identifying the sensor 10 installed in the semiconductor device 1. The feature extraction unit 422 outputs the feature associated with the sensor ID to the transmission data generation unit 423.

[0052] The transmission data generation unit 423 receives as input the feature amounts associated with the sensor IDs output from the feature amount extraction unit 422. The transmission data generation unit 423 generates transmission data including the feature amounts associated with the input sensor IDs. The transmission data generation unit 423 outputs the generated transmission data to the communication unit 43.

[0053] 4, the damage determination device 50 includes a communication unit 51, a control unit 52, a storage unit 53, and a display unit 54. The communication unit 51 receives transmission data output from the signal processing unit 40.

[0054] The control unit 52 controls the entire damage determination device 50. The control unit 52 is configured using a processor such as a CPU and a memory. The control unit 52 executes a program to function as an acquisition unit 521, a position determination unit 522, a damage determination unit 523, and a notification unit 524. The programs for realizing the acquisition unit 521, the position determination unit 522, the damage determination unit 523, and the notification unit 524 may be installed in the damage determination device 50 at the time of shipment, or may be installed separately.

[0055] The acquisition unit 521 acquires the transmission data output from the signal processing unit 40. For example, when the signal processing unit 40 and the damage assessment device 50 communicate wirelessly, the acquisition unit 521 functions as a communication interface and acquires the transmission data by performing wireless communication with the signal processing unit 40. For example, when the signal processing unit 40 and the damage assessment device 50 communicate wired, the acquisition unit 521 functions as a communication interface and acquires the transmission data by performing wired communication with the signal processing unit 40.

[0056] The positioning unit 522 locates the position of the source of the elastic wave (hereinafter referred to as the "elastic wave source") based on the sensor position information and the sensor ID and time information included in the transmission data. Transmission data related to elastic waves acquired during the period when the activation conditions are satisfied as described above is input to the damage determination device 50. The elastic waves acquired during the period when the activation conditions are satisfied are basically elastic waves generated due to damage in the joints 8 and 9. Therefore, the positioning unit 522 locates the position of the elastic wave source in the joints 8 and 9 based on the elastic waves acquired during the period when the activation conditions are satisfied. The sensor position information includes information related to the installation position of the sensor 10 associated with the sensor ID. The sensor position information includes information related to the installation position of the sensor 10, such as latitude and longitude, or horizontal and vertical distances from a reference position of the semiconductor device 1 or heat sink 2. The positioning unit 522 stores the sensor position information in advance. The sensor position information may be stored in the position locating unit 522 at any timing before the position locating unit 522 locates the position of the elastic wave source.

[0057] The damage determination unit 523 determines damage to the joints 8 and 9 based on the elastic waves acquired during the period when the activation condition is satisfied. Specifically, the damage determination unit 523 estimates the sound area, the damaged area, and the percentage of the damaged area (hereinafter referred to as the "damaged area ratio") of the joints 8 and 9 based on the results of position location performed by the position location unit 522 based on the elastic waves acquired during the period when the activation condition is satisfied. The sound area is the area of ​​the joints 8 and 9 where no peeling has occurred. For example, in the initial state (no damage), the sound area is the area of ​​the entire joint 8 or the entire joint 9. The damaged area is the area of ​​the joints 8 and 9 where peeling has occurred. The damaged area ratio is calculated by dividing the sound area by the sound area in the initial state. This makes it possible to determine the percentage of peeling that has occurred since the initial state.

[0058] Then, the damage determination unit 523 determines whether the estimated damaged area ratio has increased by a predetermined value (e.g., 20%) or more from the sound area in the initial state. Information on the sound area in the initial state may be stored in advance. If the estimated damaged area ratio has increased by a predetermined value (e.g., 20%) or more from the sound area in the initial state, the damage determination unit 523 determines that damage has occurred in the joints 8 and 9. If the estimated damaged area ratio has not increased by a predetermined value (e.g., 20%) or more from the sound area in the initial state, the damage determination unit 523 determines that no damage has occurred in the joints 8 and 9. If the damage determination unit 523 determines that damage has occurred, it causes the notification unit 524 to issue an alert.

[0059] The notification unit 524 issues an alert in response to control from the damage determination unit 523. That is, the notification unit 524 issues an alert when the damaged area ratio estimated by the damage determination unit 523 increases by a predetermined value (for example, 20%) or more from the healthy area in the initial state. The notification unit 524 may issue an alert in the damage determination device 50, or may issue an alert to another information processing device that is set in advance. The alert may include information indicating that damage to the semiconductor device 1 has increased.

[0060] The storage unit 53 stores the transmission data acquired by the acquisition unit 521. The storage unit 53 may also store information on the activation conditions, the installation position of the sensor 10, and the healthy area in the initial state. Each storage unit 53 is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device.

[0061] The display unit 54 is an image display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The display unit 54 displays the position information of the elastic wave source and the determination result under the control of the control unit 52. The display unit 54 may be an interface for connecting an image display device to the damage determination device 50. In this case, the display unit 54 generates a video signal for displaying the determination result and outputs the video signal to the image display device connected to the display unit 54.

[0062] FIG. 9 is a diagram showing an example of damage occurring in the joint 8 below the heat dissipation substrate 4 in the first embodiment, and an image of the location of the acoustic wave source. FIG. 9 shows a state in which the heat dissipation substrate 4 is soldered to the Cu base plate 5. In FIG. 9, region R1 indicates the region where the heat dissipation substrate 4 and the Cu base plate 5 are joined (region where no delamination has occurred), and region R2 indicates the region where the heat dissipation substrate 4 and the Cu base plate 5 are not joined (region where delamination has occurred). Region R2 in FIG. 9 is a region where fatigue cracks have progressed in the joint 8. In FIG. 9, the number of thermal cycles increases from the top to the bottom. For example, the top of FIG. 9 shows the initial state (no damage), the middle of FIG. 9 shows a state after more thermal cycles than the top of FIG. 9, and the bottom of FIG. 9 shows a state after more thermal cycles than the middle of FIG. 9.

[0063] Fatigue cracks progress as thermal cycles are repeated. As the fatigue crack progresses, the proportion of region R2 increases, as shown in Figure 9. Since inelastic strain due to creep in the joint 8 is concentrated at the crack tip, the crack tip is considered to become the elastic wave source SR. As the damage progresses, the position of the elastic wave source SR also moves. By tracking changes over time, it is possible to visualize the expansion of the deteriorated region, making it possible to estimate the location and area of ​​damage during use of the semiconductor device 1. Region R2 is the damaged area. For example, the damage determination unit 523 estimates the damaged area ratio by dividing the area of ​​region R1 by the area of ​​region R1 shown in the upper part of Figure 9 (the healthy area in the initial state). The damage determination unit 523 determines whether or not damage exists based on the estimated damage area ratio.

[0064] 10 and 11 are sequence diagrams showing the flow of damage determination processing performed by the damage determination system 100 in the first embodiment. It is assumed that the signal processing unit 40 is in a resting state when the processing in FIGS. 10 and 11 starts. The operation control device 35 acquires temperature information measured by the temperature sensor 20 (step S101). The operation control device 35 continues the process of step S101 for a predetermined period. As a result, the operation control device 35 acquires temperature information for the predetermined period. The operation control device 35 determines whether or not the activation condition is satisfied based on the acquired temperature information for the predetermined period (step S102). If the activation condition is not satisfied (step S102-NO), the operation control device 35 repeatedly executes the process of step S101.

[0065] If the activation condition is satisfied (step S102-YES), the operation control device 35 generates an operation signal. The operation control device 35 outputs the generated operation signal to the signal processing unit 40 (step S103). Upon receiving the operation signal from the operation control device 35, the operation unit 44 of the signal processing unit 40 instructs the power supply unit 45 to supply power, thereby putting each functional unit into an operating state (step S104). This allows the signal processing unit 40 to obtain the electrical signal output from the sensor 10.

[0066] The AFE 41 performs preprocessing on the electrical signal transmitted from the sensor 10 (step S105). Specifically, the AFE 41 performs filtering and analog-to-digital conversion on the electrical signal. The AFE 41 outputs the digital signal to the control unit 42. The control unit 42 receives the digital signal output from the AFE 41 as input and extracts features from the digital signal (step S106).

[0067] The feature extraction unit 422 associates the extracted feature with a sensor ID and outputs the resulting feature to the transmission data generation unit 423. The transmission data generation unit 423 generates transmission data including the feature output from the feature extraction unit 422 (step S107). The transmission data generation unit 423 outputs the generated transmission data to the communication unit 43. The communication unit 43 transmits the transmission data output from the transmission data generation unit 423 to the damage determination device 50 (step S108).

[0068] The communication unit 51 receives transmission data transmitted from the signal processing unit 40. The acquisition unit 521 stores the acquired transmission data in the storage unit 53 (step S109). The positioning unit 522 locates the elastic wave source using the multiple transmission data stored in the storage unit 53 (step S110). Specifically, the positioning unit 522 first extracts transmission data for one event from the multiple transmission data stored in the storage unit 53. An event refers to an elastic wave generating event that occurred in a steel structure. In this embodiment, the elastic wave generating event is damage to the joints 8 and 9 inside the semiconductor device 1. When one event occurs, multiple sensors 10 detect elastic waves at approximately the same time. In other words, the storage unit 53 stores multiple transmission data related to elastic waves detected at approximately the same time. Therefore, the positioning unit 522 sets a predetermined time window and extracts all transmission data whose arrival times fall within the time window range as transmission data for one event.

[0069] The time window range Tw may be determined using the elastic wave propagation velocity v in the steel structure of interest and the maximum sensor spacing dmax so that it is in the range of Tw≧dmax / v. In order to avoid erroneous detection, it is desirable to set Tw to as small a value as possible, so in practice Tw=dmax / v can be used. The elastic wave propagation velocity v may be determined in advance.

[0070] Next, the position locating unit 522 locates the position of the elastic wave source based on the sensor ID and time information included in the transmission data for the extracted one event and pre-stored sensor position information. For example, the position locating unit 522 calculates the difference in arrival time of the elastic wave at each of the multiple sensors 10. Next, the position locating unit 522 locates the position of the elastic wave source using the sensor position information and information on the difference in arrival time. The position locating unit 522 locates the positions of multiple elastic wave sources by repeating this process multiple times.

[0071] The position determination unit 522 outputs information on the location result to the damage determination unit 523. The damage determination unit 523 determines damage based on the location result output from the position determination unit 522 (step S111). Specifically, first, the damage determination unit 523 estimates a damage area based on the position of the elastic wave source indicated as the location result. Next, the damage determination unit 523 estimates a damage area ratio based on the damage area. If the damage determination unit 523 determines that the estimated damage area ratio has increased by a predetermined value (e.g., 20%) or more from the initial state, it determines that damage has occurred. If the damage determination unit 523 does not determine that the estimated damage area ratio has increased by a predetermined value (e.g., 20%) or more from the initial state, it determines that damage has not occurred.

[0072] The damage determination unit 523 determines whether or not the alert notification condition is satisfied (step S112). Specifically, the damage determination unit 523 determines that the alert notification condition is satisfied when it determines that damage has occurred, and determines that the alert notification condition is not satisfied when it determines that no damage has occurred. When it determines that the alert notification condition is satisfied (step S112-YES), the damage determination unit 523 causes the notification unit 524 to notify an alert. The notification unit 524 notifies the alert in accordance with control from the damage determination unit 523 (step S113).

[0073] After the process of step S113, or if it is determined that the alert notification conditions are not satisfied (step S112-NO), the damage determination device 50 executes the processes of steps S109 to S113 until the signal processing unit 40 is stopped. If the activation conditions are no longer satisfied after being satisfied, the operation control device 35 generates a stop signal to stop the signal processing unit 40. The operation control device 35 transmits the generated stop signal to the signal processing unit 40. The signal processing unit 40 transitions to a sleep state based on the stop signal transmitted from the operation control device 35.

[0074] The damage assessment system 100 configured as described above can detect deterioration of the bonded portions 8 and 9 inside the semiconductor device 1 with high accuracy. Specifically, the damage assessment system 100 assesses damage to the bonded portions 8 and 9 based on elastic waves acquired during a period when an activation condition is satisfied, among the multiple elastic waves detected by each of the multiple sensors 10. In this manner, the damage assessment system 100 activates the signal processing unit 40 and starts measuring elastic waves when a condition indicating a high-temperature holding time (when temperature change is small) is satisfied. This allows elastic waves generated from other than the bonded portions 8 and 9 due to temperature change to be removed, and elastic waves generated from the bonded portions 8 and 9 to be efficiently detected. The elastic waves detected in this manner are elastic waves generated from the bonded portions 8 and 9. Therefore, the damage assessment system 100 can accurately detect deterioration of the bonded portions 8 and 9 inside the semiconductor device 1 by assessing damage based on the detected elastic waves.

[0075] (Second embodiment) In the second embodiment, a configuration will be described in which temperature information is estimated based on the drive current of a semiconductor device without using a temperature sensor. 12 is a diagram showing the configuration of a damage determination system 100a in the second embodiment. The damage determination system 100a is used to determine damage that has occurred in the bonding portions 8 and 9 inside the semiconductor device 1.

[0076] The damage determination system 100a includes a semiconductor device 1, a heat sink 2, a plurality of sensors 10-1 to 10-n, an operation control device 35a, a signal processing unit 40, a damage determination device 50, and a storage device 60. The plurality of sensors 10-1 to 10-n and the signal processing unit 40 are connected by wire. The signal processing unit 40 and the damage determination device 50 are connected by wire or wirelessly. The operation control device 35a and the storage device 60 are connected by wire. The operation control device 35a and the signal processing unit 40 are connected by wire.

[0077] Damage assessment system 100a differs in configuration from damage assessment system 100 in that it does not include temperature sensor 20, that it includes operation control device 35a instead of operation control device 35, and that it newly includes memory device 60. The other configurations of damage assessment system 100a are the same as those of damage assessment system 100. Therefore, an explanation of damage assessment system 100a as a whole will be omitted, and only the differences from damage assessment system 100 will be explained.

[0078] A conversion equation for converting current into temperature is stored in the storage device 60. Because there is a correlation between temperature and current, the relationship between the temperature and current of the semiconductor device 1 is investigated in advance through experiments or unsteady heat transfer analysis using the finite element method. A conversion equation for converting current into temperature is determined based on the previously obtained relationship between the temperature and current of the semiconductor device 1, and this conversion equation is stored in the storage device 60. The storage device 60 is configured using a storage device such as a magnetic storage device or a semiconductor storage device. The storage device 60 may be a server on the cloud, or may be provided inside the operation control device 35a.

[0079] The operation control device 35a controls the operation of the signal processing unit 40 based on the drive current of the semiconductor device 1 and a conversion formula stored in the storage device 60. Specifically, the operation control device 35a refers to the conversion formula stored in the storage device 60 and converts the drive current of the semiconductor device 1 into a temperature value. The operation control device 35a determines whether a startup condition for starting the signal processing unit 40 is met based on the converted temperature value, and starts the signal processing unit 40 if the startup condition is met. The value of the drive current of the semiconductor device 1 may be obtained from the current supplied to the semiconductor device 1 or may be obtained by notification from outside. The startup condition is the same as in the first embodiment.

[0080] Figures 13 and 14 are sequence diagrams showing the flow of damage determination processing performed by the damage determination system 100a in the second embodiment. At the start of the processing in Figures 13 and 14, it is assumed that the signal processing unit 40 is in a sleep state. In Figures 13 and 14, the same processes as in Figures 10 and 11 are assigned the same reference numerals as in Figures 10 and 11, and their explanations will be omitted.

[0081] The operation control device 35a acquires the value of the drive current of the semiconductor device 1 (step S201). The operation control device 35a refers to a conversion formula stored in the storage device 60 and converts the acquired drive current value into a temperature value (step S202). The operation control device 35a continues the processes of steps S101 and S102 for a predetermined period. As a result, the operation control device 35a acquires temperature information for the predetermined period. The operation control device 35a determines whether the start-up condition is satisfied based on the acquired temperature information for the predetermined period (step S203). If the start-up condition is not satisfied (step S203-NO), the operation control device 35a repeatedly executes the processes of steps S201 and S202. If the start-up condition is satisfied (step S203-YES), the processes from step S103 onwards are executed.

[0082] According to the damage determination system 100a configured as described above, even without installing the temperature sensor 20, the signal processing unit 40 is activated to start measuring elastic waves when a condition indicating that the high-temperature holding time (when temperature change is small) is met, as in the first embodiment. This makes it possible to remove elastic waves generated from places other than the bonded portions 8 and 9 due to temperature change and efficiently detect elastic waves generated from the bonded portions 8 and 9. The elastic waves detected in this manner are elastic waves generated from the bonded portions 8 and 9. Therefore, the damage determination system 100a can accurately detect deterioration of the bonded portions 8 and 9 inside the semiconductor device 1 by determining damage based on the detected elastic waves.

[0083] (Third embodiment) In the third embodiment, a configuration in which the signal processing unit is constantly activated will be described. 15 is a diagram showing the configuration of a damage determination system 100b according to the third embodiment. The damage determination system 100b is used to determine damage that has occurred in the bonding portions 8 and 9 inside the semiconductor device 1.

[0084] The damage determination system 100b includes a semiconductor device 1, a heat sink 2, a plurality of sensors 10-1 to 10-n, a temperature sensor 20, a signal processing unit 40b, a damage determination device 50, and a filter 65. The plurality of sensors 10-1 to 10-n and the signal processing unit 40b are connected by wires. The signal processing unit 40b and the filter 65 are connected by wires. The temperature sensor 20 and the filter 65 are connected by wires. The damage determination device 50 and the filter 65 are connected by wires.

[0085] Damage assessment system 100b differs in configuration from damage assessment system 100 in that it does not include operation control device 35, that it includes signal processing unit 40b instead of signal processing unit 40, and that it newly includes filter 65. The other configurations of damage assessment system 100b are the same as those of damage assessment system 100. Therefore, an explanation of damage assessment system 100b as a whole will be omitted, and only the differences from damage assessment system 100 will be explained.

[0086] The temperature sensor 20 measures the temperature in the environment in which the semiconductor device 1 is installed. The temperature sensor 20 outputs the measurement result, ie, temperature information, to the filter 65.

[0087] Unlike the signal processing unit 40, the signal processing unit 40b is in a constantly operating state. The signal processing unit 40b performs signal processing on the electrical signal of the elastic wave output from the sensor 10. The signal processing unit 40b generates transmission data including the processed digital signal, and outputs the generated transmission data to the damage determination device 50 via the filter 65.

[0088] The filter 65 performs filtering based on the temperature information output from the temperature sensor 20. Specifically, the filter 65 determines whether the filter conditions for outputting the transmission data output from the signal processing unit 40b to the damage determination device 50 are satisfied based on the temperature indicated by the temperature information, and outputs the transmission data acquired during the period when the filter conditions are satisfied to the damage determination device 50. That is, the filter 65 does not output the transmission data output from the signal processing unit 40b to the damage determination device 50 during the period when the filter conditions are not satisfied. For example, the filter 65 discards the transmission data output from the signal processing unit 40b during the period when the filter conditions are not satisfied.

[0089] The filter conditions are that the value of the temperature change ΔT is less than the third threshold Th3 (ΔT < Th3) and the temperature satisfies the condition of being equal to or higher than the reference temperature T1. Thus, the filter conditions are a state where the temperature change is small and the temperature satisfies the condition of being equal to or higher than the reference temperature. Note that the values of the third threshold and the reference temperature T1 shown above are examples, and the values may be changed according to the usage situation. The filter conditions are an aspect of the conditions based on temperature. The period during which these filter conditions are satisfied corresponds to the state of process C shown in FIG. 3.

[0090] FIGS. 16 and 17 are sequence diagrams showing the flow of the damage determination process performed by the damage determination system 100b in the third embodiment. In FIGS. 16 and 17, the same processes as those in FIGS. 10 and 11 are denoted by the same reference numerals as in FIGS. 10 and 11, and the description thereof is omitted.

[0091] The AFE 41 of the signal processing unit 40b performs preprocessing on the electrical signal transmitted from the sensor 10 (step S301). Specifically, the AFE 41 performs filter processing and analog-to-digital conversion processing on the electrical signal. The AFE 41 outputs a digital signal to the control unit 42. The control unit 42 extracts a feature amount from the digital signal using the digital signal output from the AFE 41 as an input (step S302).

[0092] The feature extraction unit 422 associates the extracted feature with a sensor ID and outputs the resulting feature to the transmission data generation unit 423. The transmission data generation unit 423 generates transmission data including the feature output from the feature extraction unit 422 (step S303). The transmission data generation unit 423 outputs the generated transmission data to the communication unit 43. The communication unit 43 transmits the transmission data output from the transmission data generation unit 423 to the damage determination device 50 (step S304). The transmission data transmitted from the signal processing unit 40b to the damage determination device 50 is input to the filter 65.

[0093] The filter 65 acquires temperature information measured by the temperature sensor 20 (step S305). The filter 65 continues the process of step S305 for a predetermined period of time. As a result, the filter 65 acquires temperature information for the predetermined period of time. For convenience of explanation, the process of step S305 is described after steps S301 to S304, but the process of step S305 is performed while steps S301 to S304 are being executed. The filter 65 determines whether the filter condition is satisfied based on the acquired temperature information for the predetermined period of time (step S306). If the filter condition is not satisfied (step S306-NO), the filter 65 discards the transmission data acquired during the period when the filter condition was not satisfied (step S307).

[0094] On the other hand, if the filter condition is satisfied (step S306-YES), the filter 65 outputs the transmission data acquired during the period in which the filter condition is satisfied to the damage assessment device 50 (step S308). As a result, the damage assessment device 50 acquires only the transmission data acquired during the period in which the filter condition is satisfied. The period in which the filter condition is satisfied corresponds to the period of process C shown in FIG. 3. Thereafter, the damage assessment device 50 executes the processes from step S109 onwards.

[0095] The damage determination system 100b configured as described above requires more power consumption than the first and second embodiments due to continuous measurement, but it is possible to detect elastic waves that were overlooked in the first and second embodiments. This makes it possible to detect deterioration of the bonding portions 8 and 9 inside the semiconductor device 1 with higher accuracy than the first and second embodiments.

[0096] (Modification of the third embodiment) The damage determination system 100b may be configured to acquire temperature information as in the second embodiment. When configured in this manner, the damage determination system 100b includes a storage device 60 instead of the temperature sensor 20. The filter 65 acquires the value of the drive current of the semiconductor device 1 using a method similar to that of the operation control device 35a. The filter 65 refers to a conversion formula stored in the storage device 60 and converts the acquired drive current of the semiconductor device 1 into a temperature value. The filter 65 determines whether the filter condition is satisfied based on the converted temperature value. The subsequent processing is the same as that shown in the third embodiment.

[0097] (Fourth embodiment) In the fourth embodiment, a configuration will be described in which a fatigue life distribution prediction method is combined with the damage detection method of any one of the first to third embodiments to improve the accuracy of life prediction.

[0098] First, an overview of the fourth embodiment will be described. A conventional method has been proposed in which an inelastic strain range is estimated using a surrogate model based on a constantly measured temperature history, and a fatigue life distribution is predicted using a cycle counting method and a cumulative damage law (see, for example, Non-Patent Document 3). In this way, the fatigue life distribution of the joints 8 and 9 can be estimated using the conventional method. The vertical axis of the fatigue life distribution represents the probability density, and the horizontal axis represents the cumulative damage value. A cumulative damage value exceeding 1 indicates damage. In the fourth embodiment, the conventional method is used to calculate the fatigue life distribution and the average cumulative damage value in advance for both the joint 9 below the power semiconductor 3 and the joint 8 below the heat dissipation substrate 4. This information is stored in a damage determination device in advance. When damage is determined to have occurred using the damage detection method of any of the first to third embodiments, the average cumulative damage value of the joint 9 below the power semiconductor 3 and the average cumulative damage value of the joint 8 below the heat dissipation substrate 4, both of which are stored in the damage determination device, are corrected. A specific configuration will be described below.

[0099] Fig. 18 is a diagram showing the configuration of a damage determination system 100c according to the fourth embodiment. The damage determination system 100c is used to determine damage that has occurred at bonding portions 8 and 9 inside a semiconductor device 1. The damage determination system 100c includes a semiconductor device 1, a heat sink 2, a plurality of sensors 10-1 to 10-n, a temperature sensor 20, an operation control device 35, a signal processing unit 40, and a damage determination device 50c. The plurality of sensors 10-1 to 10-n and the signal processing unit 40 are connected by wire. The signal processing unit 40 and the damage determination device 50c are connected by wire or wirelessly. The temperature sensor 20 and the operation control device 35 are connected by wire. The operation control device 35 and the signal processing unit 40 are connected by wire.

[0100] Damage assessment system 100c differs in configuration from damage assessment system 100 in that damage assessment system 100c is provided instead of damage assessment device 50. Other configurations of damage assessment system 100c are the same as those of damage assessment system 100. Therefore, an explanation of the damage assessment system 100c as a whole will be omitted, and only the differences from damage assessment system 100 will be explained. Note that damage assessment system 100c shown in FIG. 18 assesses damage to joints 8 and 9 using the damage detection method shown in the first embodiment.

[0101] The damage determination device 50c includes a communication unit 51, a control unit 52c, a storage unit 53c, and a display unit 54. The configurations of the communication unit 51 and the display unit 54 perform the same processing as the functional units included in the damage determination device 50, and therefore, a description thereof will be omitted.

[0102] The memory unit 53c stores the transmission data acquired by the acquisition unit 521, the fatigue life distribution and average cumulative damage value of the joints 9 below the power semiconductor 3 calculated by the conventional method, and the fatigue life distribution and average cumulative damage value of the joints 8 below the heat dissipation substrate 4. The memory unit 53c may also store information on the activation conditions, the installation position of the sensor 10, and information on the healthy area in the initial state. The memory unit 53c is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device.

[0103] The control unit 52c controls the entire damage determination device 50c. The control unit 52c is configured using a processor such as a CPU and a memory. The control unit 52c executes a program to function as an acquisition unit 521, a position determination unit 522, a damage determination unit 523, a notification unit 524, and a correction unit 525. The programs for realizing the acquisition unit 521, the position determination unit 522, the damage determination unit 523, the notification unit 524, and the correction unit 525 may be installed in the damage determination device 50c at the time of shipment, or may be installed separately.

[0104] When the damage determination unit 523 determines that damage has occurred, the correction unit 525 corrects the average cumulative damage value of the joints 9 under the power semiconductor 3 and the average cumulative damage value of the joints 8 under the heat dissipation substrate 4 stored in the memory unit 53c.

[0105] Here, a specific method of correction performed by the correction unit 525 will be described with reference to Fig. 19. Fig. 19 is a diagram for explaining the correction method performed by the correction unit 525 in the fourth embodiment. In Fig. 19, the joints under the power semiconductor 3 (before correction) and the joints under the heat dissipation substrate 4 (before correction) represent the average cumulative damage values ​​stored in the memory unit 53c. Damage occurs when the cumulative damage value D reaches 1, and damage to the joints 8 under the heat dissipation substrate 4 is detected (it is determined that damage has occurred) when the average cumulative damage value is 0.86.

[0106] In this case, the correction unit 525 determines that the cumulative damage value corresponding to ΔD=1−0.86 was added for some unexpected reason, and divides 0.53, which is the average cumulative damage value of the joints 9 below the remaining undamaged power semiconductors 3, by 0.86, which is the average cumulative damage value of the joints 8 below the heat dissipation substrate 4. The correction unit 525 then divides 0.53 by 0.86, to obtain a new average cumulative damage value of the joints 9 below the power semiconductors 3. In response to detecting damage to the joints 8 below the heat dissipation substrate 4 (determining that damage has occurred), the correction unit 525 sets the new average cumulative damage value of the joints 8 below the heat dissipation substrate 4 to 1. In this way, the correction unit 525 corrects the average cumulative damage value of the joints 9 below the power semiconductors 3 and the average cumulative damage value of the joints 8 below the heat dissipation substrate 4. It should be noted that after the correction, the cumulative damage value of the joints 8 under the heat dissipation substrate 4 exceeds 1, but this damage is the crack initiation life and does not mean the product life accompanied by crack propagation. Correction unit 525 also corrects the fatigue life distribution of the joints 9 under the power semiconductors 3 and the fatigue life distribution of the joints 8 under the heat dissipation substrate 4 based on the corrected average cumulative damage value of the joints 9 under the power semiconductors 3 and the average cumulative damage value of the joints 8 under the heat dissipation substrate 4.

[0107] According to the damage assessment system 100c configured as described above, the average cumulative damage value of the joints 8 and 9 is corrected at the timing of issuing an alert notification. This allows the fatigue life distribution of the joints 9 below the power semiconductor 3 and the fatigue life distribution of the joints 8 below the heat dissipation substrate 4 to be corrected to more accurate results. In this way, it becomes possible to improve the accuracy of life prediction.

[0108] (Modification of the fourth embodiment) In the above example, the damage assessment system 100c is configured to assess damage to the joints 8 and 9 using the damage detection method described in the first embodiment. The damage assessment system 100c may be configured to assess damage to the joints 8 and 9 using the damage detection method described in the second embodiment or the damage detection method described in the third embodiment.

[0109] (Fifth embodiment) In the fifth embodiment, a configuration will be described in which a fatigue life distribution prediction method is combined with the damage detection method of any one of the first to third embodiments to improve the accuracy of life prediction.

[0110] First, an overview of the fifth embodiment will be described. In the fifth embodiment, the accuracy of life prediction is improved for an inverter device 80 incorporating multiple semiconductor devices 1 as shown in FIG. 20. As shown in FIG. 20, multiple sensors 10 are arranged in each semiconductor device 1 included in the inverter device 80. Note that the device targeted in the fifth embodiment is not limited to the inverter device 80, but may be any other device incorporating multiple semiconductor devices 1. In the fifth embodiment, the accuracy of life prediction is improved by using a conventional fatigue life distribution as a prior probability distribution and calculating a posterior probability distribution by Bayesian estimation. A specific configuration will be described below.

[0111] FIG. 21 is a diagram showing the configuration of a damage determination system 100d according to the fifth embodiment. The damage determination system 100d is used to determine damage that has occurred at bonding portions 8 and 9 inside a semiconductor device 1. The damage determination system 100d includes a semiconductor device 1, a heat sink 2, a plurality of sensors 10-1 to 10-n, a temperature sensor 20, an operation control device 35, a signal processing unit 40, and a damage determination device 50d. The plurality of sensors 10-1 to 10-n and the signal processing unit 40 are connected by wire. The signal processing unit 40 and the damage determination device 50d are connected by wire or wirelessly. The temperature sensor 20 and the operation control device 35 are connected by wire. The operation control device 35 and the signal processing unit 40 are connected by wire.

[0112] Damage assessment system 100d differs in configuration from damage assessment system 100 in that damage assessment system 100d includes damage assessment device 50d instead of damage assessment device 50. Other configurations of damage assessment system 100d are the same as those of damage assessment system 100. Therefore, an explanation of the entire damage assessment system 100d will be omitted, and only the differences from damage assessment system 100 will be explained. Note that damage assessment system 100d shown in FIG. 21 assesses damage to joints 8 and 9 using the damage detection method shown in the first embodiment.

[0113] The damage determination device 50d includes a communication unit 51, a control unit 52d, a storage unit 53d, and a display unit 54. The configurations of the communication unit 51 and the display unit 54 perform the same processing as the functional units included in the damage determination device 50, and therefore, a description thereof will be omitted.

[0114] The memory unit 53d stores the transmission data acquired by the acquisition unit 521, and the fatigue life distribution of the bonding portions 9 below the power semiconductors 3 of the plurality of semiconductor devices 1 and the fatigue life distribution of the bonding portions 8 below the heat dissipation substrates 4 of the plurality of semiconductor devices 1 calculated by a conventional method. The memory unit 53d may also store information on the activation conditions, the installation position of the sensor 10, and information on the sound area in the initial state. The memory unit 53c is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device.

[0115] The control unit 52d controls the entire damage determination device 50d. The control unit 52d is configured using a processor such as a CPU and a memory. The control unit 52d executes a program to function as an acquisition unit 521, a position determination unit 522, a damage determination unit 523, a notification unit 524, and a calculation unit 526. The programs for realizing the acquisition unit 521, the position determination unit 522, the damage determination unit 523, the notification unit 524, and the calculation unit 526 may be installed in the damage determination device 50d at the time of shipment, or may be installed separately.

[0116] When the damage determination unit 523 determines that damage has occurred, the calculation unit 526 calculates a posterior probability distribution by Bayesian estimation using the fatigue life distribution of the joints 9 under the power semiconductors 3 of the multiple semiconductor devices 1 stored in the memory unit 53c and the fatigue life distribution of the joints 8 under the heat dissipation substrates 4 of the multiple semiconductor devices 1 as pre-confirmed distributions.

[0117] Here, a specific description will be given of the processing performed by the calculation unit 526. Consider a case where damage occurs in k (k is an integer of 1 or more) semiconductor devices 1 out of m (m is an integer of 2 or more) semiconductor devices 1 operated in the inverter device 80. If this is event A, the probability that event A occurs in a semiconductor device 1 with reliability R is expressed by equation (1).

[0118]

number

[0119] In this case, reliability R represents the probability that a device or its components will perform the required function under given conditions for a specified period of time, and is expressed by equation (2).

[0120]

number

[0121] In equation (2), P frepresents the failure probability. The failure probability is expressed as the sum of the probability densities where the cumulative damage value exceeds 1. The posterior probability distribution f(R|A) is expressed by equation (3).

[0122]

number

[0123] The calculation unit 526 calculates the posterior probability distribution f(R|A) based on the above formula (3). This calculated posterior probability distribution f(R|A) serves as a highly accurate life prediction result.

[0124] According to the damage assessment system 100d configured as above, it is possible to use Bayesian estimation to make estimations with higher accuracy than conventional fatigue life predictions.

[0125] (Modification of the fifth embodiment) In the above example, the damage assessment system 100d is configured to assess damage to the joints 8 and 9 using the damage detection method described in the first embodiment. The damage assessment system 100d may be configured to assess damage to the joints 8 and 9 using the damage detection method described in the second embodiment or the damage detection method described in the third embodiment.

[0126] According to at least one of the embodiments described above, by having one or more semiconductor devices 1 each having one or more bonding portions 8, 9 that bond the heat dissipation substrate 4 to other components, a plurality of sensors 10 that detect elastic waves generated at one or more bonding portions 8, 9 of the one or more semiconductor devices 1, and a damage determination unit 523 that determines damage to one or more bonding portions 8, 9 based on elastic waves that satisfy temperature-based conditions among the multiple elastic waves detected by each of the multiple sensors 10, it is possible to detect deterioration of bonding portions within the semiconductor device with high accuracy.

[0127] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention described in the claims and their equivalents. [Explanation of symbols]

[0128] 1... semiconductor device, 2... heat sink, 3... power semiconductor, 4... heat dissipation substrate, 5... Cu base plate, 6-1, 6-2... Cu, 7... ceramic insulating substrate, 10, 10-1 to 10-n... sensor, 20... temperature sensor, 35, 35a... operation control device, 40, 40b... signal processing unit, 41... AFE, 42... control unit, 43... communication unit, 44... operation unit, 45... power supply unit, 50, 50c, 50d... damage determination device, 51... communication unit, 52, 52c, 52d... control unit, 53, 53c... Memory unit, 54...display unit, 60...storage device, 65...filter, 80...inverter device, 100, 100a, 100b, 100c, 100d...damage determination system, 411...receiving unit, 412...first filter, 413...analog-digital conversion unit, 414...second filter, 421...event signal generation unit, 422...feature extraction unit, 423...transmission data generation unit, 521...acquisition unit, 522...positioning unit, 523...damage determination unit, 524...notification unit, 525...correction unit, 526...calculation unit

Claims

1. one or more semiconductor devices having one or more bonding portions for bonding the heat dissipation substrate to other components; a plurality of sensors for detecting elastic waves generated at the one or more bonding portions of the one or more semiconductor devices; a damage determination unit that determines damage to the one or more joints based on elastic waves detected by the plurality of sensors, the elastic waves satisfying either a first temperature-based condition that requires a retention time, which indicates a time during which a state in which temperature change is small, is equal to or greater than a first threshold value and equal to or less than a second threshold value, and the temperature is equal to or greater than a reference temperature, or a second temperature-based condition that requires a state in which temperature change is small and the temperature is equal to or greater than a reference temperature; A damage determination system comprising:

2. a position locating unit that locates a position of a source of an elastic wave in the one or more joints based on an elastic wave that satisfies either the first condition based on temperature or the second condition based on temperature; The damage determination system according to claim 1, wherein the damage determination unit estimates the proportion of the damaged area in the one or more joints based on the location results from the position determination unit, and determines that damage has occurred if the estimated proportion of the damaged area is equal to or greater than a threshold value.

3. A signal processing unit that performs signal processing on the multiple elastic waves detected by each of the multiple sensors; an operation control device that controls the operation of the signal processing unit based on the temperature; Furthermore, The damage determination system according to claim 1 or 2, wherein the operation control device activates the signal processing unit when the retention time becomes equal to or greater than the first threshold value when a first condition based on the temperature is satisfied.

4. The damage determination system according to claim 3 , wherein the operation control device stops the signal processing unit when a first condition based on the temperature is no longer satisfied.

5. a temperature sensor for measuring a temperature around the one or more semiconductor devices; The damage determination system according to claim 3 , wherein the operation control device determines whether the first temperature-based condition is satisfied based on temperature information measured by the temperature sensor.

6. a storage device that stores a conversion formula for converting a current into a temperature; The damage determination system of claim 3, wherein the operation control device acquires a current value supplied to the one or more semiconductor devices, converts the acquired current value into a temperature using the conversion formula stored in the memory device, and determines whether the first condition based on the temperature is satisfied based on the converted temperature information.

7. A signal processing unit that performs signal processing on the multiple elastic waves detected by each of the multiple sensors; a filter unit that, when the second temperature-based condition is satisfied, extracts elastic waves that have been signal-processed by the signal processing unit while the second temperature-based condition is satisfied, and discards elastic waves that have been signal-processed by the signal processing unit while the second temperature-based condition is not satisfied; The damage determination system according to claim 1 or 2, further comprising:

8. a temperature sensor for measuring a temperature around the one or more semiconductor devices; The damage determination system according to claim 7 , wherein the filter unit determines whether the second temperature-based condition is satisfied based on temperature information measured by the temperature sensor.

9. a storage device that stores a conversion formula for converting a current into a temperature; 8. The damage determination system according to claim 7, wherein the filter unit acquires a current value supplied to the one or more semiconductor devices, converts the acquired current value into a temperature using the conversion formula stored in the storage device, and determines whether the second condition based on the temperature is satisfied based on the converted temperature information.

10. a memory unit in which an average cumulative damage value for each joint is stored; a correction unit that corrects the average cumulative damage value for each joint stored in the storage unit when the damage determination unit determines that damage has occurred; The damage determination system according to claim 1 or 2, further comprising:

11. the one or more semiconductor devices are a plurality of semiconductor devices, a conversion device incorporating the plurality of semiconductor devices; a storage unit that stores a prior probability distribution of fatigue life of each of the bonding portions of the plurality of semiconductor devices; a calculation unit that calculates a prior probability distribution of fatigue life of each of the bonding portions of the plurality of semiconductor devices when the damage determination unit determines that damage has occurred; The damage determination system according to claim 1 or 2, further comprising:

12. a damage determination unit that determines damage to one or more junctions based on elastic waves detected by a plurality of sensors that detect elastic waves generated at one or more junctions of one or more semiconductor devices having a heat dissipation substrate and another component, and that satisfy either a first temperature-based condition that a retention time representing a time during which a state in which there is little temperature change is maintained, which is equal to or greater than a first threshold value and is equal to or less than a second threshold value, and the temperature is equal to or greater than a reference temperature, or a second temperature-based condition that a state in which there is little temperature change and the temperature is equal to or greater than a reference temperature; A damage determination device comprising:

13. detecting an elastic wave generated at one or more bonding portions of one or more semiconductor devices having one or more bonding portions that bond a heat dissipation substrate to another component; A damage determination method for determining damage to one or more joints based on elastic waves that satisfy either a first temperature-based condition, which requires that the retention time, which represents the time during which a state of little temperature change is maintained, among multiple detected elastic waves, is a time between a first threshold value and a second threshold value and the temperature is above a reference temperature, or a second temperature-based condition, which requires that the temperature is in a state of little temperature change and the temperature is above a reference temperature.

14. A computer program for causing a computer to execute a determination step of determining damage to one or more junctions based on elastic waves detected by each of a plurality of sensors that detect elastic waves generated at one or more junctions of one or more semiconductor devices having one or more junctions that join a heat dissipation substrate to other components, where the retention time, which represents the time during which a state of little temperature change is maintained, is a time between a first threshold and a second threshold and is equal to or greater than a reference temperature, or a second temperature-based condition, which is a state of little temperature change and is equal to or greater than a reference temperature.

Images