IGBT fault diagnosis method based on stray inductance and collector-emitter voltage

By constructing an IGBT fault diagnosis method based on stray inductance and collector-emitter voltage, and utilizing a support vector machine model, the accuracy and cost issues of IGBT bonding wire fault diagnosis in existing technologies are solved, achieving efficient and low-cost fault detection.

CN118444123BActive Publication Date: 2024-11-15UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410906672.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-15
Estimated Expiration
2044-07-08

AI Technical Summary

Technical Problem

Existing IGBT bonding wire fault diagnosis technologies cannot provide detailed connection status information, and the detection methods are costly or have strict environmental requirements, making it difficult to accurately detect minor faults.

Method used

By calculating the stray inductance and emitter voltage of the IGBT module under different bond wire breakage conditions, a support vector machine classification model is constructed. Feature data is obtained using double pulse testing, and the emitter voltage and current signals are trained to achieve fault diagnosis.

Benefits of technology

It enables accurate diagnosis of IGBT bond wire breakage. The detection method is simple, low-cost, highly adaptable, and has small error, making it suitable for fault diagnosis of modules with similar structures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118444123B_ABST
    Figure CN118444123B_ABST
Patent Text Reader

Abstract

The present invention discloses an IGBT fault diagnosis method based on stray inductance and collector-emitter voltage, which belongs to the technical field of IGBT bonding lead fault diagnosis, specifically: using double pulse test to obtain the stray inductance of the IGBT module to be tested under different IGBT bonding lead fracture situations; repeating the test, taking the stray inductance and corresponding collector-emitter voltage signal obtained in each test as a group of sample points, adding the IGBT bonding lead fracture situation label, and constructing a training set; constructing a classification model and training; obtaining the stray inductance and corresponding collector-emitter voltage signal of the IGBT module to be fault diagnosed, and using the trained classification model to output the diagnosis result of the IGBT bonding lead fracture situation. The present invention is more intuitive in data acquisition, has higher data measurement accuracy, can directly detect the connection status information of the bonding lead, and diagnose and obtain accurate IGBT bonding lead fracture situations, and has the advantages of simple detection method, low cost, and low requirements on the detection environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of IGBT bonding wire fault diagnosis, and in particular relates to an IGBT fault diagnosis method based on stray inductance and collector-emitter voltage. Background Art

[0002] In power electronic converters, power semiconductor devices are one of the key components, and insulated gate bipolar transistors (IGBTs) have quickly become the dominant power semiconductor devices due to their large capacity, fast switching speed, easy drive, low on-state voltage drop, and high input impedance. However, during long-term operation, IGBT modules will generate dynamic thermal stress at the connection between the bonding wire and the chip due to temperature cycling, leading to thermal mechanical fatigue failure. Therefore, bond wire breakage is one of the most prominent failure types of IGBT modules.

[0003] At present, there are three main detection methods for IGBT bond wire fracture: resistance detection, capacitance detection, and power detection. The resistance detection method is simple and low-cost, and can quickly determine whether the bond wire is open or short-circuited, but it cannot provide detailed information about the connection status of the bond wire, and has limited detection capabilities for minor faults or changes in electrical performance. The capacitance detection method can detect the insulation between the bond wires and evaluate possible short circuits or other capacitance anomalies, but it also cannot provide detailed information about the connection status of the bond wires, and has insufficient detection capabilities for small changes in electrical performance faults. The power detection method can detect some minor faults or anomalies by monitoring the power consumption of the IGBT device, such as the increase in resistance caused by loose connection of the bond wire, but it requires special test equipment support, is costly, and has strict requirements on the test environment. Therefore, it is necessary to propose a more accurate and efficient IGBT bond wire fracture fault diagnosis technology. Summary of the invention

[0004] In view of the problems of existing IGBT bonding wire fault diagnosis technology, the present invention provides an IGBT fault diagnosis method based on stray inductance and collector-emitter voltage, which improves the reliability of IGBT modules and the stability of power electronic converters through accurate and efficient fault diagnosis.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage includes the following steps:

[0007] S1. Calculate the stray inductance L of the IGBT module under test under different IGBT bonding wire breakage conditions, specifically:

[0008] The IGBT bonding wire breakage scenario includes a complete IGBT scenario and a scenario where different numbers of IGBT bonding wires are cut off;

[0009] Performing double pulse tests on the IGBT modules to be tested under different IGBT bonding wire breakage conditions, testing the collector-emitter voltage signal Vce of the IGBT module to be tested, and the current Ic flowing through the IGBT module to be tested, and then calculating the stray inductance L of the IGBT module to be tested;

[0010] S2. Repeat the double pulse test under different IGBT bonding wire breakage situations, use the stray inductance L and the corresponding collector-emitter voltage signal Vce obtained from each double pulse test as a group of sample points, and add labels corresponding to the IGBT bonding wire breakage situation to the sample points; then, multiple groups of sample points under all IGBT bonding wire breakage situations are jointly constructed to obtain a training set;

[0011] S3, build a classification model;

[0012] S4. Based on the training set, train the classification model to obtain a trained classification model;

[0013] S5. Perform a double pulse test on the IGBT module to be diagnosed for fault, obtain the stray inductance L and the corresponding collector-emitter voltage signal Vce of the IGBT module to be diagnosed for fault, input them into the trained classification model, and output the diagnosis result of the IGBT bonding wire breakage of the IGBT module to be diagnosed for fault.

[0014] Furthermore, during the double pulse test, when the pulse signal applied to the IGBT module to be tested is turned off, the collector-emitter voltage signal Vce forms a voltage spike, at which time the current Ic gradually decreases, and then the stray inductance L of the IGBT module to be tested is calculated based on the collector-emitter voltage signal Vce and the current Ic.

[0015] Furthermore, during the double pulse test, a voltage spike is formed when the pulse signal applied to the IGBT module to be tested is turned off for the second time.

[0016] Furthermore, the calculation formula of the stray inductance L is:

[0017] .

[0018] Furthermore, a classification model is constructed based on support vector machine (SVM).

[0019] Furthermore, the classification model uses a decision function mapped by a Gaussian kernel function f ( x ), the formula is:

[0020] ;

[0021] In the formula,sign (·) represents a symbolic function; N Represents the total number of sample points input into the classification model; x is any sample point; x i Indicates i sample points; i Indicates i The Lagrange multiplier of the sample points; y i Indicates i The label of the IGBT bonding wire fracture situation corresponding to each sample point; the superscript " T ” indicates transposition; K (·) represents Gaussian kernel function; b represents the coefficient matrix.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1. The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage proposed in the present invention makes full use of the electrical characteristics of the IGBT, directly obtains the collector-emitter voltage signal Vce and current Ic under different IGBT bonding wire fracture conditions through double pulse testing, and then constructs a training set to train a classification model for bonding wire fault prediction; compared with other IGBT fault detection methods, the training set constructed by the present invention is more intuitive in data acquisition, and the measurement accuracy of the training set data is higher. It can directly detect the connection status information of the bonding wire, diagnose the accurate IGBT bonding wire fracture condition, and the prediction error is smaller;

[0024] 2. The present invention has the advantages of simple detection method, low cost and low requirements on the detection environment;

[0025] 3. Preferably, the present invention uses a support vector machine to analyze and learn the training set, the required training set sample size is small, the model converges quickly, the operation is simple, and the IGBT bonding wire breakage of the IGBT module to be fault diagnosed can be accurately and efficiently predicted. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a schematic diagram of an IGBT module to be tested used in Embodiment 1 of the present invention;

[0027] Figure 2 The double pulse test circuit diagram used in Example 1 of the present invention;

[0028] Figure 3 is the double pulse signal V obtained by the double pulse test in Example 1 of the present invention ge , Collector-Emitter Voltage V ce and current I c Schematic diagram of

[0029] Figure 4 In the case of a complete IGBT in Embodiment 1 of the present invention t Double pulse signal V near time 3 ge , Collector-Emitter Voltage V ce and current I c The change curve of

[0030] Figure 5 In the case of cutting off an IGBT bonding wire in Embodiment 1 of the present invention t Variation curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3;

[0031] Figure 6 In the case of cutting off two IGBT bonding wires in Embodiment 1 of the present invention t Variation curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3;

[0032] Figure 7 In the case of cutting off three IGBT bonding wires in Embodiment 1 of the present invention t Variation curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3;

[0033] Figure 8 is a line graph showing the variation of the stray inductance L with the number of broken IGBT bonding wires in Example 1 of the present invention;

[0034] Fig. 9 Schematic diagram of collector-emitter voltage signal Vce under different numbers of broken IGBT bonding wires in Example 1 of the present invention;

[0035] Fig.10 This is a diagram showing the diagnostic accuracy results of the classification model after training in Example 1 of the present invention. DETAILED DESCRIPTION

[0036] The following is a further complete description of the technical solutions in the specific implementation of the present invention in conjunction with the accompanying drawings and embodiments. The described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Example 1

[0038] The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage proposed in this embodiment includes the following steps:

[0039] S1. Calculate the stray inductance L of the IGBT module under test under different IGBT bonding wire breakage conditions, specifically:

[0040] The schematic diagram of the IGBT module to be tested used in this embodiment is as follows Figure 1 As shown, there are four IGBT bonding wires, and the IGBT bonding wire breakage situations include four situations, namely, a complete IGBT situation, a situation where one IGBT bonding wire is cut off, a situation where two IGBT bonding wires are cut off, and a situation where three IGBT bonding wires are cut off;

[0041] Double pulse tests are performed on the IGBT modules to be tested under different IGBT bonding wire breakage conditions. The double pulse test circuit used in this embodiment is as follows: Figure 2 As shown, the IGBT module is characterized by a transistor and a diode connected in parallel. An IGBT module that is kept turned off is used as a freewheeling diode, and another IGBT module is used as the IGBT module to be tested. The freewheeling diode, the IGBT module to be tested and a voltage source are connected in series to form a loop; an inductor L1 is connected in series at both ends of the freewheeling diode; a double pulse signal Vge is applied to the gate of the IGBT module to be tested, a high-voltage isolation probe is used to obtain the collector-emitter voltage signal Vce at both ends of the IGBT module to be tested, and a Rogowski coil current probe is used to obtain the current Ic near the IGBT module to be tested in the loop;

[0042] In the double pulse test process, the double pulse signal Vge is applied to the gate of the IGBT module to be tested in sequence, and the collector-emitter voltage signal Vce and current Ic of the IGBT module to be tested are tested, and the following are obtained: Figure 3 The signal schematic diagram shown, specifically:

[0043] exist t At time 0, the first pulse drive signal is applied to the IGBT module to be tested, Vge is at a high level, the transistor of the IGBT module to be tested is saturated and turned on, Vce is 0, and the inductor current increases linearly with the time of applying the first pulse drive signal. At this time, the detected Ic also increases linearly;

[0044] exist t At moment 1, Ic increases to the maximum, and then the pulse signal applied to the IGBT module to be tested is turned off, Vge is at a low level, Vce at both ends of the IGBT module to be tested is the voltage source voltage, the inductor current is continued by the freewheeling diode and decays slowly, and at this time Ic also gradually decays to 0;

[0045] exist tAt time 2, the second pulse drive signal is applied to the IGBT module to be tested, Vge is high, the transistor of the IGBT module to be tested is saturated and turned on again, Vce is 0, the freewheeling diode enters reverse recovery, and the generated reverse recovery current and the inductor current flow into the transistor of the IGBT module to be tested. At this time, the detected Ic includes the reverse recovery current and the inductor current, and then a current spike appears first, and then increases linearly;

[0046] exist t At time 3, Ic increases to the maximum, and then the pulse signal applied to the IGBT module to be tested is turned off again, and Vge is at a low level. Since Ic is large at this time, due to the existence of stray inductance L, a voltage spike appears on Vce, and Ic gradually decays to 0;

[0047] Furthermore, for t 3. The voltage spike appears at Vce at time 3. Using Vce and Ic at this time, calculate the stray inductance L of the IGBT module to be tested:

[0048] ;

[0049] S2. Repeat the double pulse test under different IGBT bonding wire breakage conditions. In this embodiment, the double pulse test is performed 100 times for each IGBT bonding wire breakage condition with a repetition interval of 5 to 10 minutes.

[0050] Figure 4 For complete IGBT t The changing curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3, Figure 5 To cut an IGBT bonding wire t The changing curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3, Figure 6 To cut two IGBT bonding wires t The changing curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3, Figure 7 To cut three IGBT bonding wires t Variation curves of the double pulse signal Vge, collector-emitter voltage signal Vce and current Ic near time 3;

[0051] It can be seen that under different IGBT bonding wire fracture conditions, t There will be a Vce voltage spike near time 3, and Ic will gradually decay to 0;

[0052] Calculate the stray inductance L of the IGBT module under different IGBT bond wire breakage conditions and plot the following: Figure 8The stray inductance L shown in the figure changes with the number of broken IGBT bonding wires. Fig. 9 The schematic diagram of the collector-emitter voltage signal Vce under different numbers of broken IGBT bonding wires is shown (where the complete IGBT is represented as "unbroken"), and it can be seen that the stray inductance L and the collector-emitter voltage signal Vce under different IGBT bonding wire broken situations have significant characteristics, which can be used to diagnose the IGBT bonding wire broken situation;

[0053] Therefore, the stray inductance L and the collector-emitter voltage signal Vce are used as features to diagnose and predict the IGBT bond wire fracture situation. Specifically, the stray inductance L and the corresponding collector-emitter voltage signal Vce obtained from each double pulse test are used as a group of sample points. Some sample points under different IGBT bond wire fracture situations are taken to jointly construct a test set containing 48 groups of sample points. For the remaining groups of sample points, the labels corresponding to the IGBT bond wire fracture situation are added to each sample point, and a training set is jointly constructed, with a total of 352 groups of sample points.

[0054] S3. Based on support vector machine, build a classification model and use Gaussian kernel function mapping decision function f ( x ), the formula is:

[0055] ;

[0056] In the formula, sign (·) represents a symbolic function; N Represents the total number of sample points input into the classification model; x is any sample point; x i Indicates i sample points; i Indicates i The Lagrange multiplier of the sample points; y i Indicates i The label of the IGBT bonding wire fracture situation corresponding to each sample point; the superscript " T ” indicates transposition; K (·) represents Gaussian kernel function; b represents the coefficient matrix;

[0057] S4, training the classification model based on the training set, optimizing the classification model parameters through cross-validation, and obtaining the trained classification model;

[0058] S5. Based on the test set, the diagnostic accuracy of the trained classification model is evaluated and verified. Specifically:

[0059] The test set is input into the trained classification model, and the diagnosis results of the IGBT bonding wire fracture corresponding to each group of sample points in the test set are output. The results are compared with the labels of the known real IGBT bonding wire fracture conditions, as shown in the following figure. Fig.10 As shown in the diagnostic accuracy result diagram, it can be seen that the IGBT fault diagnosis method based on stray inductance and collector-emitter voltage proposed in this embodiment has a high accuracy rate of 93.75%, and can be used to diagnose the IGBT bonding wire fracture of the IGBT module to be diagnosed for the fault having the same structure as the IGBT module to be tested;

[0060] S6. Perform a double pulse test on the IGBT module to be diagnosed for fault, obtain the stray inductance L and the corresponding collector-emitter voltage signal Vce of the IGBT module to be diagnosed for fault, input them into the trained classification model, and output the diagnosis result of the IGBT bonding wire breakage of the IGBT module to be diagnosed for fault.

[0061] It should be noted that these are only specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto, and any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. An IGBT fault diagnosis method based on stray inductance and collector-emitter voltage, characterized in that: The following steps are involved: S1. Calculate the stray inductance L of the IGBT module under test under different IGBT bonding wire breakage conditions, specifically: The IGBT bonding wire breakage scenario includes a complete IGBT scenario and a scenario where different numbers of IGBT bonding wires are cut off; Double pulse tests are performed on the IGBT modules to be tested under different IGBT bonding wire breakage conditions. When the pulse signal applied to the IGBT module to be tested is turned off for the second time, the collector-emitter voltage signal Vce of the IGBT module to be tested forms a voltage spike. The stray inductance L of the IGBT module to be tested is calculated through the test collector-emitter voltage signal Vce and the current Ic flowing through the IGBT module to be tested at this time. S2. Repeat the double pulse test under different IGBT bonding wire breakage situations, use the stray inductance L and the corresponding collector-emitter voltage signal Vce obtained from each double pulse test as a group of sample points, and add labels corresponding to the IGBT bonding wire breakage situation to the sample points; then, multiple groups of sample points under all IGBT bonding wire breakage situations are jointly constructed to obtain a training set; S3, build a classification model; S4. Based on the training set, train the classification model to obtain a trained classification model; S5. Perform a double pulse test on an IGBT module to be diagnosed that has the same structure as the IGBT module to be tested, obtain the stray inductance L and the corresponding collector-emitter voltage signal Vce of the IGBT module to be diagnosed, input them into the trained classification model, and output the diagnosis result of the IGBT bonding wire breakage of the IGBT module to be diagnosed.

2. The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage according to claim 1, characterized in that: The calculation formula of the stray inductance L is:

3. The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage according to claim 1, characterized in that: Based on support vector machine, a classification model is constructed.

4. The IGBT fault diagnosis method based on stray inductance and collector-emitter voltage according to claim 3 is characterized in that: The classification model uses the decision function f(x) mapped by the Gaussian kernel function, and the formula is: In the formula, sign(·) represents the sign function; N represents the total number of sample points of the input classification model; x is any sample point; x i represents the i-th sample point; λ i represents the Lagrange multiplier of the i-th sample point; y i Indicates the label of the IGBT bonding wire fracture corresponding to the i-th sample point; the superscript "T" indicates transposition; K(·) indicates the Gaussian kernel function; and b indicates the coefficient matrix.