condition monitoring in power electronics€¦ · underpinning research what is condition...
TRANSCRIPT
Underpinning Research
Condition Monitoring in Power Electronics
Dr Li RanProfessor of Power Electronics - Systems
University of Warwick
26th November 2015
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
What is condition monitoring?
● Detect impeding fault at an early stage.● Detection is obtained without affecting
normal operation.● The objective is to prevent sudden and
catastrophic failures & cascaded damages.● Appropriate actions are suggested based
on prognosis.
● Condition monitoring is usually appliedin systems that are safety critical orexpensive to repair after faults.
● It would be ideal to use signals that arealready available for control or protectionto constrain the cost.
Underpinning Research
What is condition monitoring?
Condition monitoring vs Protection
● Condition monitoring is to prevent something really bad from happening.- It can be passive or active;- It is usually difficult of detect the weak early signatures;- It can be slow in response taking into consideration lifetime interest.
● Protection is waiting for something really bad to happen.- It is generally passive;- It usually needs to be very fast and automatic.
Underpinning Research
Incentives of condition monitoring
Safety critical
O&M cost accountfor >30% of energyProduction cost
Loss of productionis very expensive
Underpinning Research
Tasks in Condition Monitoring
● It is necessary to understand the monitored plant to decide what signals to transduce and whatsignatures to capture.
● Signal processing depends on knowing the operating condition of the monitored plant; issues ofvariable operating condition are of concern in many systems such as wind turbines and Evs.
● The change and trending of the monitored signature should be combined with prognosis to give judgement of severity of impeding fault.
● Condition based maintenance scheduling is the traditional operational strategy, but direct stressmanagement through power electronic control is becoming attractive.
(Tavner, Ran, Penman, Sedding, Condition Monitoring of Rotating Electrical Machines, IET Press, 2008)
Underpinning Research
Arrangement of Data Acquisition
● Raw signals are conditioned and processed locally, and it is usually the high level informationabout the condition monitored that is transmitted over longer distance to central points.
● Data acquisition channels are multiplexed and computational power can be shared withother functionalities in the system.
(Tavner, Ran, Penman, Sedding, Condition Monitoring of Rotating Electrical Machines, IET Press, 2008)
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
Instrumentation
What will be measured for power electronics condition monitoring?
● I am not sure… But perhaps: current, voltage, temperature, flow rating (water/oil cooling), strain gauges.● Performance requirement: appropriate bandwidth, range, precision, accuracy, stability over time, and cost.● Some measurement points can be difficult to access, e.g. internal temperature, device switching current
and voltages. Condition monitoring based on external (w.r.t. component package) temperatures,converter level voltage and current, and/or gate driver signals may be more favourable.
● Measurement is always subject to environmental impacts.
Reference junction compensation for a thermocouple
Underpinning Research
Distributed Temperature Sensing
The Raman-scattered component in the reflected lightis sensitive to the thermally influenced molecularvibration. According to the time when the signatures inthe back scattered light are received, the temperaturealong the optic fibre can be detected.
Underpinning Research
Current Sensing
● Bandwidth: DC to over 100 MHz.● Current range: 1500 A (but expensive).● The same principle is used to measure voltage.
● Bandwidth: AC up to 10’s of MHz.● Not affected by external magnetic effects.● No saturation.
Underpinning Research
Signal Processing
● At the foundation of signal processing is Fourier series.
(Lucas Barbosa, Wikipedia)
Underpinning Research
Sampling Requirement
● ‘Seeing is believing’ Is it?
● Shannon’s sampling theorem: In order for a band-limited signal with bandwidth fB to be
reconstructed fully, it must be sampled at a rate fs≥2fB. A signal sampled at 2fB is said to be
Nyquist sampled, and 2fB is called the Nyquist frequency. No information is lost if a signal
is sampled at the Nyquist frequency, and no additional information is gained by sampling faster than this rate.
● Misunderstanding of the sampling theorem has often caused aliasing.
Underpinning Research
Prevention of Aliasing
● Sampling theorem is sometimes misunderstood as to sampling at twice thefrequency of the interested component.
● Actually the signal to be sampled must be band limited to half of the sampling frequency.
Underpinning Research
Illustration of Picket Fence Leakage
● This figure shows a frequency resolution of 0.1 Hz, corresponding to a sampling length of 10 seconds.● But the frequency components of interest may be at 59.83 Hz, 59.85 Hz and 59.87 Hz, in addition to
59.80 Hz and 59.90 Hz. Components not at Fourier output frequencies need to be estimated.● This can happen in converter-machine systems with complex control modulation processes.
Underpinning Research
Variable Operating Condition
● Wavelet transform may be used to analyse non-stationary time-domain signals.
dt
a
bttf
abaW )()(
1),( *
where a is the scaling factor and b the time-shift.
Mother wavelet could be chosen as
0a
)]2sin()2[cos()( 2/2
tjtet t
● As b changes, the analysis is focused on different parts of the time-domain signal.● As a changes, the mother wavelet is squeezed or stretched. Signatures of different
frequencies in the time-domain signal can be picked up.
Underpinning Research
Example of Wavelet Analysis
rotor resistance: 0.685 Ohm
rotor resistance: 0.35 Ohm
Underpinning Research
Example of Wavelet Analysis
signature contained in starting current
signature extraction from
starting current
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
Dominant Failure Mechanisms
solder
1. Bond wire lift off
2. Solder cracking
• both mechanisms are due to temperature excursion,mismatch of CTEs and
elevated temperature towards the end of lifeRoot cause:
(source: Semikron)
Underpinning Research
Stresses during Temperature Cycling
The copper substrate material properties were taken to be the following: E = 110 GPa; α = 17 x10-6 K-1. For the Si die, E is about 162 GPa and α is 4.1 x10-6 K-1, while the E and α values for the SiC die are 501 GPa and 4.3 x10-6 K-1.
E: Young‘s modulusα: CTE
ΔTj=50 K
Underpinning Research
Monitoring Capacitor Ageing
● A difficulty is to place the current sensor ina sandwiched capacitor arrangement.
● The differential mode EMI (conducted) wouldgo through the reservoir capacitor, and beaffected by ESR.
● Loss of capacitance (farad) may also be detected in slow switching converters.
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
(TSEPs)
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
TSEPs – On-state Characteristics
● Infineon 1700 V, 1000 A power module.● The static V-I characteristics depend on temperature.● Solder fatigue would increase junction temperature, and higher on-state loss.● Bond wire lift-off would further increase on-state resistance.
Underpinning Research
TSEPs – Switching Edges
Positive current
A
Vg+
Vg-
Vdc
+
-
P
N
t
td td
ton
toff
Vg+
t
t
t
Vg-
VAN*
VAN Vdc - Vsat
-Vd + - -
T1
T2
D2
D1
i
● Voltage discrepancy:
500 V x 50 ns x10 kHz = 0.25 V
Underpinning Research
TSEPs – Gate Drives
● Source of figures: Niu and Lorenz● Turn-on (left): The Miller gate plateau (amplitude and duration) is Tj dependent.● Turn-off (right): The time delay between the two pulses of veE is Tj dependent.
Underpinning Research
TSEPs – Gate Drives
● Source: Rodriguez, Claudio, Theilliol and Velan, 2007.● Miller gate plateau rises and shortens with Tj.● Also sensitive to electrical operating point.
● Luo, Chen, Sun, Li and He, 2014.● The delay increases with Tj.● Also sensitive to electrical
operating point.
Underpinning Research
TSEPs for SiC Devices
TSEP Advantages Disadvantages
Body diode forward
voltage (with low
current)
Good temperature
sensitivity
Easy to measure
Less susceptible to noise
Good linearity with
temperature
Doubtful application on line
On-state voltage
drop
It can be done on-line
No modification to the
module
Not linear in SiC
Threshold voltage Can be done on-line
Module modification required
Not very temperature
sensitive in SiC
Susceptible to noise
Miller capacitance
discharge time
Can be done on-line
No module modification is
required
High precision circuitry
required (ns)
Susceptible to noise
Small temperature sensitivity
Turn-off and Turn-on
times
Can be done on-line
No module modification
is required
Low temperature sensitivity
Very susceptible to noise
Complex behaviour in SiC
Gate current
Can be done on-line
No module modification
is required
Temperature sensitivity?
• Different TSEPs have been investigated for SiC power devices
• Temperature dependencies of different parameters have been investigated
• Based on initial measurements, condition monitoring strategy will be pursued
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
(TSEPs)
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
On-line Measurement of Vce
● Source: Aalborg University, Denmark.● D1 and D2 must have large voltage blocking capability.● D1 and D2 must have very close characteristics.● Results are dependent on load current.
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 1
Inside Power Module
● External (e.g. case, ambient) temperatures are to be measured.● For water cooling, flow rate and water temperatures are to be measured.● The electrical operating point is also to be monitored.
Underpinning Research
• Aging was emulated by increasing junction-case thermal resistance by 20% of the original total.
• Junction temperature is increased by about 10 K.
• Case-above-ambient temperature is increased by 1.59 K, which is detectable.
Simulation using validated Semikron model:
Ta (◦C) Tc (◦C) Tj_d (◦C) Tj_tr (◦C) P_d (W) P_tr (W) P_tot (W)
Before aging 43.3 98.1 101.7 101.9 2.04 3.57 33.66
After aging 43.3 99.69 111.5 111.7 2.12 3.65 34.63
steady-state simulation results
Algorithms Using Converter Level
Measurements, Example 1
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 1
• Inverter is subject to a step load demand.
• Case-above-ambient temperature shows clear distinction before and after aging.
• Thermal transient due to heat sink is slow and long, making the conception application questionable under dynamic loading.
Step-input test:
experimen
tsimulation
experimen
tsimulation
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 1
●Testing adapted algorithm under variable operating condition.
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 1
Issues examined:
Demonstration by deployment into a
commercial converter.
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 2
Positive current
A
Vg+
Vg-
Vdc
+
-
P
N
t
td td
ton
toff
Vg+
t
t
t
Vg-
VAN*
VAN Vdc - Vsat
-Vd + - -
T1
T2
D2
D1
i
7th harmonic
fundamental harmonic
5th harmonic
i
ΔV
3rd harmonic
t
t
● The non-ideal characteristics are sensitive to Tj.● The delay of the turn-off edge causes discrepancy in inverter output voltage.● The discrepancy contains harmonics which can be detected. ● This is like uncompensated on-state voltage in inverter output.
Underpinning Research
Algorithms Using Converter Level
Measurements, Example 2
● Increase the sensitivity of signature detection in closed-loop control.● This is highly dependent on the application.
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
(TSEPs)
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
Monitoring Capacitors
● Source: E Wolfgang.● Capacitors are another major contributor to unreliability.● Ageing is usually accompanied by loss of capacitance and increase of resisatnce.● Measuring capacitor current in a sandwiched +Vdc/-Vdc busbar arrangement
can be difficult.
Underpinning Research
Monitoring Capacitors
● Source: Ran, Gokani, Clare, Bradley, Christopoulos, 1998.● It may be possible to measure the HF, differential mode
EMI (conducted) across the dc busbars, which depends onthe ESR of capacitor.
● The result will be highly sensitive to output cabling, loadterminal condition, and EMI filter if present, which needsto be maintained constant.
Underpinning Research
Monitoring Capacitors
● In converters or converter modules operating at lowswitching frequency, it may be possible to detect thechange of capacitor voltage ripple, after the capacitancehas lost by a few %.
(5 ms/div)
Underpinning Research
Contents
1. Introduction to condition monitoring
2. Instrumentation and signal processing
3. Failure mechanisms of power modules
4. Temperature sensitive electrical parameters
(TSEPs)
5. Condition monitoring algorithms
6. Monitoring capacitors
7. Condition based maintenance
8. Summary
Underpinning Research
Condition Based Maintenance
● Source: Held, Nicoletti, Scacco, Poech, 1997● Schedule maintenance for: completing a critical mission, capturing the next period of high production etc.● A central issue of maintenance scheduling is lifetime modelling and stressor estimation.● Accelerated lifetime testing is conducted by device manufacturers; effect of stresses during
normal operation is hard to estimate (Huang, Mawby).
Underpinning Research
Condition Based Maintenance
● Source: Lai, Chen, Ran, Alatise, Xu, Mawby, 2015.● Low ΔTj during operation would normally not causing ageing to an
unaged power module.● However, the initially same low ΔTj will cause further ageing on an aged
module.
Underpinning Research
Summary
● Power module failures could be a result of ageing (wear out) processes. Therefore the concept ofcondition monitoring can be beneficially applied to power module.
● Provided that the power modules are used in converters according to their specifications, the ageing processes are more likely to happen to the packaging structure: e.g. solder delaminationand bond wire lift-off.
● Power module ageing would cause the module to operate at higher internal temperatures. Thereforecondition monitoring very much relies on temperature sensitive electrical parameters (TSEPs) ofthe device.
● Condition monitoring could be directly detecting the TSEPs such as Vce_on, or it could detect theconsequence of the changed electrical parameters under certain operating condition, such aspower losses and external temperatures.
● Signatures in gate drive circuitry are currently of high interest in developing condition monitoringsolutions; techniques based on converter level measurements also sound attractive.
● Condition monitoring power electronics is a rapidly developing area. Therefore I have includedsome general information about instrumentation and signal processing. This may be relevant infuture development. Sincere apologies if this is not,
Thank you and have some good time at Warwick!