the emergence of data-driven technologies for … · the emergence of data-driven technologies for...

Post on 23-Apr-2018

222 Views

Category:

Documents

4 Downloads

Preview:

Click to see full reader

TRANSCRIPT

The Emergence of Data-Driven Technologies for

Operational Oceanographic ApplicationsSimon Foster / Rizwan Sheikh

www.fugro.com2

Metocean Measurement

Metocean Consultancy

Environmental Services

Weather Forecasting

Integrated Offshore Monitoring Systems

Structural Monitoring

Metocean Services and Systems

www.fugro.com3

Operational Oceanography

Operational

Oceano-

graphy

Weather

Forecasting

Real-time

Monitoring

Operational

Statistics

www.fugro.com4

Data ScienceData Science is a multidisciplinary field combining

mathematics, statistics, computer science and

programming. In practice, Data Scientists manage,

process and interpret often large/complex datasets

in search of meaning.

AnalyticsAnalytics is the “science of analysis” where in the

context of Data Science “data products” are

developed to process data and return

algorithmically generated outputs, commands or

actions.

Machine LearningMachine Learning refers to an array of techniques

to algorithmically make a prediction/estimation

(Supervised Learning) and/or algorithmically

decipher patterns in data (Unsupervised Learning)

Data Driven Technologies

Image courtesy of HHI

www.fugro.com5

Data-driven Operational Oceanography

PAST PRESENT FUTURE

Hindcast

MeasurementsReal-time Systems Forecasts

Example:

Smart Mooring Integrity Monitoring

Operational Criteria

- Operability

- Weather windows

- OLS

Monitoring

- Metocean

- Helideck

- Structural Integrity

Weather

- Tow route

- Installation

- Site specific

www.fugro.com6

Example 1: Smart Mooring Integrity Monitoring

Motivation:

Numerous mooring line failures worldwide, many initially undetected.

Existing hardwired monitoring systems are expensive and often unreliable.

Opportunity:

Extract that hidden value from existing datasets and systems.

Reduce cost and increase reliability of mooring integrity monitoring.

Solution:

Use Machine Learning algorithms applied to real-time measurements from sensors on the

facility topside: Metocean, position, response, draft & structural.

Algorithms learn the normal intact behaviour of the system in place and are capable of

detecting anomalies synonymous with line failure.

Applications:

Potentially any permanently moored vessel. System anomaly detection.

www.fugro.com7

Asset Integrity: Mooring Monitoring

Wind Speed (knots)

18Buoy Roll (o)

1.3

Current Speed (knots)

0.4Current Direction (oT)

182Wind Direction (oT)

035Buoy Pitch (o)

1.5Buoy Tilt (o)

7.5Buoy Tilt Direction (oT)

198Hawser Tension (tonnes)

91.5Buoy Heading (oT)

047Buoy Offset (m) Buoy Offset Direction (oT)

202Mooring 1 Mooring 2 Mooring 3

Mooring 4Mooring 5 Mooring 6

16.6

www.fugro.com8

Data-driven Operational Oceanography

PAST PRESENT FUTURE

Hindcast

MeasurementsReal-time Systems Forecasts

Example:

Optimized Weather Forecasting

Operational Criteria

- Operability

- Weather windows

- OLS

Monitoring

- Metocean

- Helideck

- Structural Integrity

Weather

- Tow route

- Installation

- Site specific

www.fugro.com9

Weather Forecasting: Current Practice

X

t

Present

FuturePast

Present Present

Real-time

Obs

Forecast

New

Forecast

2. Forecast bulletin issued based on output from numerical models and forecaster

1. Location with real-time observations of Metocean parameter, X (e.g. wave, wind)

3. New observations acquired – user qualitatively evaluates forecast performance

4. Prior forecast superseded by updated forecast bulletin

5. More observations acquired with forecasts updated on 6/12/24hr intervals …

www.fugro.com10

Weather Forecasting: Optimized Forecasts

X

t

Present

FuturePast

Present Present

Real-time

Obs

Forecast

New

Forecast

Optimized

Forecast

New

Optimized

Forecast

2. Forecast bulletin produced based on output from numerical models (and forecaster)

1. Location with real-time observations of Metocean parameter, X (e.g. wave, wind)

3. Forecast optimized using recent observations and trained ML algorithm

4. More observations acquired

5. New forecast produced by models (and forecaster)

6. New optimized forecast produced and issued…Process continues…

www.fugro.com11

Data integration is an enabler for data-driven

Operational Oceanographic technologies.

Data-driven Operational Oceanography

PAST PRESENT FUTURE

Hindcast

MeasurementsReal-time Systems Forecasts

Operational Criteria

- Operability

- Weather windows

- OLS

Monitoring

- Metocean

- Helideck

- Structural Integrity

Weather

- Tow route

- Installation

- Site specific

www.fugro.com12

Closing Thoughts

• The renewed focus on technological innovation.

• Domain knowledge (Science & Engineering) together with

Data Science (Analytics) is vitally important in developing

data-driven Operational Oceanographic technologies.

• Industry forums, like the FOO and others, are among the

best places for minds to meld and spark ideas.

Thank You

s.foster@fugro.com

top related