The role of simulation in contemporary Industrial Research – Sharing experiences

Rajendra Naik, PhDSenior Principal, GE Global Research

May 15, 2018

This is General Electric (GE)

AVIATION

TRANSPORTATION

HEALTHCARE

POWER

BAKER HUGHES –A GE COMPANY

RENEWABLES

GLOBAL RESEARCH

Intersection of Big data and physics

33

High Performance Computing

Industrial Big Data

High Fidelity Models

Industrial Big Data

Artificial Intelligence, Learning Model

Hybrid Models

Physics Based+

Data Adapted

Digital Twin – Engineering models with defined outcome

44

Operational Data

Maintenance Actions

Inspection

Sensors

Data source Asset

Domain knowledge

Performance Models

Lifing Models

Digital Twin

Control ModelLearning/AI

Business outcome

• Fuel consumption

• Power output

• System efficiency

• Remote Monitoring & Diagnostics

• Predict failure of components

• Condition based maintenance

• Life Optimizing controls

• Mission Optimization

• Asset assignment

Performance

ReliabilityIntegrated

Optimization

Digital Twin – Aviation

55

Reduce maintenance cost using turbine blade cumulative damage

models – updated per flight

Increased availability, reduce unnecessary service overhauls

Inspection time;Optimized shop time

Objective : Maximum availability of an aircraft engine by intelligent workscoping

Environmental conditions;Per-Flight data;

Prior damage;Engine Operating Mode

Digital Twin – Transportation

66

Locomotive – Trip Optimizer

Objective : Minimize fuel consumption & emissions –

generated per trip

Locomotive data;Track database;

Operating condition

Real time optimization : Optimal speed & horse power

Per asset model1 Continuously tuned – new data / insights3

Scalable – MMs assets4Business outcomes2

Adaptable – new …5

3-17% fuel savingsEnabled by system modeling, real time optimization & controls

Operator Cab

77

Novel approach to modeling, algorithms, software architecture

Objective : Determine optimal wind turbine positions to maximize Annual Energy Production (AEP) and reduce Balance of Plant costs

Constraints of geography, turbine loads, acoustic noise and mix of turbines

Heuristics + Best in class MINLP algorithms + multi threaded optimization code

Wind Farm Layout Optimization

Wind Farm Operational Optimization

88

Objective : Maximize Annual Energy production (AEP) of a wind farm

Solution : Minimize inter turbine wake losses with coordinated controls

Digital Twin of the wind farm

0.5-2% AEP improvement

Networked Controls

CommunicationsFarm-wide awarenessCo-ordinated turbine control

1% AEP improvement $ 2MM value per year for 100 MW farm

Model Based Design Cycle

99

Model of a plant

Reference : http://en.wikipedia.org/wiki/Model-based_design

Controller development

Simulations

Deployment on actual system

• Plant : Process / system to be controlled• Data driven and/or physics based

• Controller synthesis & analysis• Choose appropriate type of controller

• Offline or real time simulation• Model in the loop/ Hardware in the loop

Autocode generation

Iterate on different designs, sensors,

actuators

Model Based Design Example – IC Engine

1010

Design a controller for an IC engine to meet speed and power requirements across the operating range

Engine

Plant to be controlled

These models are taskes for GRC. These

should be models extracted from GT Power.

7

Qdot_oil

6

Qdot_water

5

Brake_Torque

4

Tleft_exh_man

3

Tright_exh_man

2

Mdot_cyl_out

1

Mdot_cyl_inRPM

Pint_man

Tint_man

Pexh_man

Vol_Ef f

Volumetric Efficiency

RPM

Pint_man

Tint_man

Pexh_man

Mf uel

Adv _Angl

Ind_Ef f

Right Bank - Indicated Efficiency

RPM

Pint_man

Tint_man

Pexh_man

Mf uel

Adv _Angl

Exh_Enrgy _Frac

Right Bank - Exhaust Energy Fraction

6

Number of active cylinders

on the right bank6

Number of active cylinders

on the left bank

RPM

Pint_man

Tint_man

Pexh_man

Mf uel

Adv _Angl

Ind_Ef f

Left Bank - Indicated Efficiency

RPM

Pint_man

Tint_man

Pexh_man

Mf uel

Adv _Angl

Exh_Enrgy _Frac

Left Bank - Exhaust Energy FractionGround6

Ground5

1/2

RPM

Pint_man

Tint_man

Pexh_man

Friction_ Torque

Friction Torque

RPM

Pint_man

Tint_man

Mf uel

Num_Activ e_Cy l_Right

Num_Activ e_Cy l_Lef t

Vol_Ef f

Mdot_cy l_out

Mdot_cy l_in

Mdot_f uel

Calculate Mass Flowrates

Mdot_cy l_out

Mdot_cy l_in

Mdot_f uel

Right_Enrgy _Frac

Tint_man

Tf uel

Num_Activ e_Cy l_Right

Num_Activ e_Cy l_Lef t

Lef t_Enrgy _Frac

Tright_exh_man

Tlef t_exh_man

Calculate Exhaust Temperatures

Right_Ind_Ef f

Mf uel

Num_Activ e_Cy l_Right

Num_Activ e_Cy l_Lef t

Lef t_Ind_Ef f

Friction_Torque

Brake_Torque

Calculate Brake Torque

7

RPM

6

Tfuel

5

Act_Inj

4

Tint_man

3

Pleft_exh_man

2

Pright_exh_man

1

Pint_man

RPM

Pint_man

Tint_man

Pright_exh_man

<RPM>

<Pint_man>

<Tint_man>

<Pright_exh_man>

<Mf uel>

<Adv _Angl>

<RPM>

<Pint_man>

<Tint_man>

<Pright_exh_man>

<Mf uel>

<Adv _Angl>

<RPM>

<Pint_man>

<Tint_man>

<Pright_exh_man>

<Mf uel>

<Adv _Angl>

<RPM>

<Pint_man>

<Tint_man>

<Pright_exh_man>

<Mf uel>

<Adv _Angl>

<Adv _Angl>

Plef t_exh_man

Pexh_man

<RPM>

<Pint_man>

<Tint_man>

<Pexh_man>

<RPM>

<Pint_man>

<Tint_man>

<Pexh_man>

<RPM>

<Pint_man>

<Tint_man>

<Mf uel>

<Mf uel>

<Num_Activ e_Cy l_Right>

<Num_Activ e_Cy l_Lef t>

<Num_Activ e_Cy l_Right>

<Num_Activ e_Cy l_Lef t>

Num_Activ e_Cy l_Lef t

Num_Activ e_Cy l_Right

Tf uel

<Tint_man>

<Tf uel>

<Num_Activ e_Cy l_Right>

<Num_Activ e_Cy l_Lef t>

<Mf uel>

Model : Physics based and data driven (Grey box

approach)Controller design

Multi-loop PID

Actual engine testing

Control Unit

Engine

Simulation studiesDesktop/ hardware

Marine Engine Turbocharger - Model Based Design

1111Model based design helped business take a decision to go for 3-Turbo configuration

Trade off matrix Three turbo Four turbo

Cost, packaging & system complexity Lower Higher

Transient response Slower but well within CTQ

Faster & well within CTQ

Three Turbo Configuration

Results from Simulink Mean Value Model

Four Turbo Configuration

Transmission and Distribution – Voltage Stability

1212

N60 Phasor Measurement Unit

XPC Real Time Target Machine

Phasor Data

RT-LABTarget PC

RT-LABCommand Station

TCP/IP

Voltage/ Current

Analog output

Voltage Stability Indices Visualization

Voltage stability monitoring and control algorithms

Control Command

Communication Latencies

“I find out what the world needs,then I proceed to invent it.”

- Thomas Alva Edison

And he didn’t have the tools that we have today……….