PowerPoint Presentation · •Condition Analysis Tachometer •Local Data Concentrator -...

Post on 27-Mar-2020

7 views 0 download

Transcript of PowerPoint Presentation · •Condition Analysis Tachometer •Local Data Concentrator -...

EMPOWERINGMAINTENANCE ENGINEERS & BUSINESSES WITH ACTIONABLE DATA.

HPUs Winch HPU

E-motor/shaft

Gearboxes

What can go wrong ?

Source: O&M cost-based FMECA results for average scenario

D. Cevasco et al 2018 J. Phys.: conf. Ser 1102

Why should you care ?

Note: offshore turbines, gearless systems

Why not all turbines are monitored ?

1st barrier:

Cost of installations: hardware, wiriring

Retrofitting

Sensor types «current, vibration, oil, speed»

2nd barrier:

Vibration sensors: 100 ksps

Acoustic sensors: 2 000 ksps

Human machinery diagnostics repeatability ?

Failure severity quantifiability ?

Human diagnostics digitalisability ?

3rd barrier:

Complex installation Data bottleneck, speed and cost Human data interpretation

4th barrier:

Data ownership

Complex contract

Manufacturer solution vs warranty

Why not all turbines are monitored ?

1st barrier:

Cost of installations: hardware, wiriring

Retrofitting

Sensor types «current, vibration, oil, speed»

2nd barrier:

Vibration sensors: 100 ksps

Acoustic sensors: 2 000 ksps

Human machinery diagnostics repeatability ?

Failure severity quantifiability ?

Human diagnostics digitalisability ?

3rd barrier:

Complex installation Data bottleneck, speed and cost Human data interpretation

4th barrier:

Data ownership

Complex contract

Manufacturer solution vs warranty

Main Shaft

MainBearing

3-stage Gearbox

Generator

•17 Bearings•9 Gears•8 Shafts

Low Speed Shaft

Int. Speed Shaft

High Speed Shaft

typical drivetrain layoutComplex installation:

• Condition Analysis Sensors

• Condition Analysis Tachometer

• Local Data Concentrator - RS-485/Ethernet Bridge

• Cloud Server - Host Database/User Display

System installation: NRG, TPhD

Why not all turbines are monitored ?

1st barrier:

Cost of installations: hardware, wiriring

Retrofitting

Sensor types «current, vibration, oil, speed»

2nd barrier:

Vibration sensors: 100 ksps

Acoustic sensors: 2 000 ksps Human machinery diagnostics repeatability ?

Failure severity quantifiability ?

Human diagnostics digitalisability ?

3rd barrier:

Complex installation Data bottleneck, speed and cost Human data interpretation

4th barrier:

Data ownership

Complex contract

Manufacturer solution vs warranty

AnalogDigital

Analog:

Full-scale range: ±500 gRange: 0 Hz – 50 KHzUltralow noise density: 125 μg/√HzOperation: −40°C to +125°CComplete electromechanical self-test

Digital:

ADC: 24 bits Sampling rate: 400 to 100 000 Hz Faraday cageIP 67RAM: flash 512 MB, SDRAM: 32 MBMicrocontroller: 100 MHzCommunication: 2 MBPSBused architecture

State of the art sensingDiagnostics within the sensor package

100 000 sps

Processing sampling speed of 100 ksps on the fly, output 5 kbytes.

Data bottleneck

Why not all turbines are monitored ?

1st barrier:

Cost of installations: hardware, wiriring

Retrofitting

Sensor types «current, vibration, oil, speed»

2nd barrier:

Vibration sensors: 100 ksps

Acoustic sensors: 2 000 ksps Human machinery diagnostics repeatability ?

Failure severity quantifiability ?

Human diagnostics digitalisability ?

3rd barrier:

Complex installation Data bottleneck, speed and cost Human data interpretation

4th barrier:

Data ownership

Complex contract

Manufacturer solution vs warranty

EDGY SENSORSAnalog to digital & onboard analytics

performed directly on the sensor.

Condition Indicators

CIs

Component level

Raw

dat

a

Shaft analysisBearing analysisGear analysisSpeed analysisMotor current analysisPump analysisMotor vibration analysis

Har

dw

are

Soft

war

e

Physics based models

CLOUDData warehousingAlarm thresholding

Prognostics

Dig

ital

(Kb

its)

Concatenate CI Intomachine Health Indicators (HIs)

(multi feature data fusion)CI -> HI

Alarms thresholding

Prognostics

Statistics (machine learning)

Prognostics(state space model)

Data centric models

Automated monitoring algorithms: edge & cloud

Turbine 1 Turbine 2 Turbine 3

► First level: Fleet overview with traffic light

display

► On click, component level showing

quantified normalised component Health

Index (HI)

► On click, mechanical diagnostics

24/7 Normalised, Quantified, Digitalised machinery heath status

Turbine 4 Turbine 5 Turbine 6

Turbine 7 Turbine 8 Turbine 9

Faults automatically detected

► Remaining useful life (RUL) prognostics models target warnings at -250 hours lifetime and have confidence intervals associated with them.

Automated prognostics:

Add on algorithms: Speed analysis – why should I care ?

▪ Wind speed changes with height

▪ As the Blade Reaches Top of it’s Arc, Delivers More Power

▪ RPM Speeds Up

▪ On This Machine

▪ 0.01 to 0.05% Change in RPM

Barthelmie et al, Riso National Labs

• Each Blade is Sensitive to Changes In Wind Shear

Speed analysis: food for thoughts

➢ Could this be used to evaluate blade

efficiency?

Detect icing

Differences in icing will effect blade lift

Blade pitch error

Why not all turbines are monitored ?

1st barrier:

Cost of instalations: harware, wirering

Retrofittting

Sensor types «current, vibration, oil, pressure»

2nd barrier:

Vibration sensors: 100 kHz

Acoustic sensors: 2 000 KHz

Human machinery diagnostics repetability ?

Failure severity quantifiability ?

Human diagnostics digitalibility ?

3rd barrier:

Complex instalation Data bottleneck, speed and cost Human data interpretation

4th barrier:

Data ownership

Complex contract

Manufacturer solution vs warantee

Since then

• Generator monitoring, algorithms developed

Electrical monitoring

• Digital twin feedStructural

monitoring

• Algorithms developed Oil

monitoring