In-situ Health Monitoring for Power Electronics Modules Prof V Pickert, Dr B Ji Newcastle University

Pictures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Motivation

Pictures

Picture Source www

Engine Management Light

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Background

Improve Reliability/ Availability/ Safety

Condition Monitoring

Diagnostic

Lifetime enhancement

Health Monitoring

Lifetime Estimator

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

Redundancy

Fail Safe/ Fault Tolerance

20.02.2014

Diagnostic

Condition Monitoring System is constantly monitored through diagnostic tools

Health Monitoring System is constantly monitored through diagnostic tools and compared with its infant state

Lifetime Estimator System is constantly monitored through diagnostic and prognostic tools

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

Part is replaced when it fails Part will be replaced after a warning has been issued Part will be replaced to the most convenient time

20.02.2014

Change in failure rate over time

Failure rate

early failure period

random failure period

wear-out failure period condition based Condition monitoring maintenance

normal operation Health Monitoring diagnostic and & prognostic Lifetime Estimator

end of life

time

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Condition monitoring

Simple condition monitor tool

Pictures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Lifetime estimator

Data Acquisition Data Manipulation State Detection Health Assessment Prognosis Assessment Advisory Generation

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Health monitoring

• Measures the status of health of a component. • Compares health relevant parameters with a baseline. • The baseline is is seen as the ideal (or healthy) parameter. • The difference between the measured value and the baseline is called degradation or ageing. • “Health risk” is identified if degradation becomes too large.

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Advantages of health monitoring

Car on crossing

• Safety is improved because a warning is flagged up prior failure

Car in garage at convenient time

• Availability is increased through maintenance scheduling

Garage reads out failure code before failure occurs Parts are changed based on their health not OEM recommendation

• Reliability is increased through collecting information on degradation

• Reduce maintenance cost

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Health monitoring techniques Model-based method

Mission Profile

Parameter Extraction

Physics of failure model

Counting Algorithm

Fatigue Accumulation

Data driven method

Pictures Static / Right-Portable power module health characterisation tester

Fusion method combination of both

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

In-situ health monitoring technique

Health monitoring in real time proposed by researchers from Newcastle University

Pictures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Power converter system failure distribution

Power device failures 31% 2. “An Industry-Based Survey of Reliability in Power Electronic Converters”, IEEE IAS, 2011

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Standard IGBT module packaging and main failures

Cover Terminal

Bond wire

Silicon chip

Isolation substrate with copper foils on both side

Silicon gel

Epoxy

Base plate (Copper)

Two major failure modes: 1) bond wire failure 2) solder fatigue

∆l = CTE ⋅ ∆T l

A power module fails due to temperature swing NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Temperature cycling induces stress symbol

type

causes

features

∆Tj

active power cycling

Power dissipation within semiconductor chips

Short cycling period, larger temperature gradient from chip to cooling plate

∆Theat sink

passive thermal cycling

Operational environment changes (e.g. ambient temperature, coolant temperature, etc.)

Long cycling period, large variation, identical temperature excursion

Pictures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

How to measure temperature?

Direct measurement

Pictures

Indirect measurement Temperature Sensitive Electric Parameters: e.g. VCE(on) and VGE(th)

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

VCE(on) as a temperature sensor C 140

G

positive temperature coefficient

120 100

E

IC [A]

SIG C 158 T120 R3 80

① negative temperature coefficient

40 20

25°C 75°C 150°C

0 0

500

1000 VCE [mV]

1500

2000

forward voltage drop

60

-2mV/K ③ ②

25°C

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

125°C X°C virtual junction temperature

20.02.2014

VCE(on) also helps to measure bondwire resistances

Vchip RKK’1 L1 RL1

Vchip RMA1

RAA’1

RKK’2

RMA2

RAA’2

RCu RL12 RCu

RCu RCC’1

RME1 RG1

REE’1 RE1

RCu RL3

RCu RL2 L2 RCC’2

RME2 RG2

REE’2 RE2

L3 VT

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

VT

20.02.2014

Bondwire degradation

Pictures

VCE(on) increase NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Solder Aging Test and Analysis

1. Accelerated aging test with air-toair thermal shock chamber

2. Evaluate solder layer conditions with SAM

Pictures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Experimental Results Sample 1

Sample 2

IR camera

0 cycle

97.37%

98.29%

98.74%

99.08%

89.32%

74.79%

86.86%

800 cycles

74.52% 1300 cycles

59.75%

73.88%

61.72%

71.6%

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

In-situ health monitoring circuit

T3

D1 n io t c e t o r p

M T2

D6

n io t c e t o r p

e iv r d e t a G

n io t c e t o r p

e iv r d e t a G

Measurement with digital isolation (optical and inductive)

e iv r d e t a G

e iv r d e t a G &

&

&

T6

D4

6

D5

D2

n io t c e t o r p

n io t c e t o r p

e iv r d e t a G

T4

T5 e iv r d e t a G

CDC

D3 n io t c e t o r p

&

&

VDC

&

T1

Selector relay network 5

100mA D

Aux.

Auxiliary switch 1

Gate drive

Controller

Relay signal

High current

SPI

Digital isolation

6

Inverter T1~T6

ISO0 PSU

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Digital isolation and protection circuit DC/DC

+VISO +VISO

REF BANK2

Voltage VISO ISO1 Reference +VISO Voltage -VISO REF ISO2 ISO1 Reference ISO2

10uF

ISO1+VISO

10uF

100nF

ISO2 A1

+VISO ISO1 ISO2 -VISO

Z D

100nF

ISO1REF

ISO2 -VISO1 ISO

R2

+VISO -VISO

REF

ISO1

VCC

VCC

DC/DC

REF VDD VIO

ISO2

SCLK REF VDD VIO A1A/D SDI A2 Converter A2 SDO SCLK A/D CSSDI A3 A4 ConverterSDO A3 A5 GND CS A6 ISO1 GND

VCC

Digital isolator IA OA Digital OB Isolators IB IC OA OC IA OD OB ID IB IC OC ID OD

R1

SCLK DO DI CS0

VEE

MCU

CS1

ISO

(a) up to 600V

BANK1

VCC

BANK2

Characteristics: based on digital isolation 16-Bit analog differential input 125k samples per channel per second bank isolated isolated from earth ground multiple gains NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

DPST (NO)

VEE

ISO

(b) 600V-2500V 20.02.2014

Generating the baselines for bond wire failures

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Operation detecting bondwire failure 300us

100us Measurement delay time

Ih Il (100mA) time

VCE(h) VCE(l) time VCE(h): Voltage drop with high current

VCE(l): Voltage drop with low current

The high current Ih is used to measure the voltage drop and the low current Il is used to measure the temperature using TSEP.

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Results IGBT with broken bond wires: Total failure with 2 broken bond wires

Pictures FWD with broken bond wires:



Voltage increases by approximately 12mV for the IGBT and 7mV for the diode with one bond wire lift while the resolution for the proposed in-situ measurement circuit is 1.2mV

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Solder fatigue results into higher chip temperatures

100%

81%

64%

49%

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Operation Thermocouple at baseplate

TSEP

Z thjr (t ) =

T j (t ) − Tr (t ) P 1 Pav = N

Itest+Isense current

N

∑V

CE ( on )

(u ) ⋅ I C (u )

u =1

1 ms Test pulse about 1 ms Duty Ratio = 94.4%

Isense (100mA) time

VCE(h) VCE(l)

time VCE(l): Voltage drop with sense current (before test pulse sequence)

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

VCE(l): Voltage drop with sense current (during test pulse sequence)

VCE(h): Voltage drop with heating current (during test pulse)

20.02.2014

Summary

• Introduction to health monitoring for power electronics • In-situ health monitoring is operating in “real-time” and can be embedded in EVs/HEVs • In-situ health monitoring can be applied to other packaging technologies • In-situ health monitoring can be applied to other power electronics devices such as capacitors for example • Received funding from KTP/TSB to increase TRL level • Received Faculty Innovator Award

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014

Thank you everyone! Prof V Pickert, Dr B Ji Newcastle University

Email: [email protected]

NEWCASTLE UNIVERSITY Power Electronics, Drives and Machines Research Group

20.02.2014