Presented by

Brian St Rock, PjL

UTRC PI

Veronica Adetola, Larry Zeidner

ARPA-E Co-design

Workshop

July 26, 2019

System Co-design through Co-Simulation

Defining what’s next:

UTRC is UTC’s

innovationengine

Berkeley, CA

East Hartford, CT

ShanghaiChina

CorkIreland

Rome & TrentoItaly

Definenew

frontiers

Co-developnew

technologies

Solvetough

problems

Serveas hub for

technical

interchange

Leverageglobal network

of innovation

MonetizeUTC

intellectual

property

Aglobal

presence

Globalcollaborations

Concurrent System Design – Motivating Example

High performance, next generation systems require increased integration and adaptiveness

This page contains no technical data subject to the EAR or the ITAR.

3

Lesson Learned: Cannot assess value of new architectures without early consideration of performance, robustness, & controls

Fundamental controllability issuesLate learning & engineering changesHardware over-specification

More-Electric Aircraft

Outcome:

Control Interdependency Matrix (1 subsystem shown)

Industry-leading fuel burn achieved through highly integrated and adaptive power and thermal management

Conventional More-Electric

Conventional Design

Metric Conventional More-Electric

Control Loops ~5 ~40

Lines of Code 200k 600k

System and Supervisory Control Co-design

Hybrid refrigeration for Truck/Trailer • Design output: Battery size, voltage, control set-points, parameters and limits

• Multi-objective optimization1. Cost (Battery +Power electronics)2. Battery capacity loss

• System-level constraints1. Maintain cargo temperature: 2. One hour autonomous battery operation3. At least 15 years of battery life

• Variabilities – cooling load, usage• Model uncertainties

12.5 13 13.5 14 14.54

6

8

10

12

14

16

18

20

PE

+ b

atte

ry c

ost in

$K

% Battery capacity loss @ 15 years

Optimization Progress and Pareto Front

Benefits: Fuel savingsReduced emission & noiseReduced food waste

Simulation-based design - Bayesian optimization based on Gaussian process

UTC proprietary. No technical data subject to the EAR or the ITAR.4

Co-design: Sensors, Actuators and Control

Synergetic system design for maximalachievable performance

UTC proprietary. No technical data subject to the EAR or the ITAR.5

DISCOVER: Computational Intelligence for System Design

Automated exploration and evaluation of system (plant & control) architectures

UTC proprietary. No technical data subject to the EAR or the ITAR.7

Problem Def Exploration Machine Learning

Optimization

Inputs

Video

Improvements: 5X Performance, 2X Volume, >15X Cycle Time

• Requirements• System-level metrics• Technology options• Flow types • Mission segments• Composition rules

DISCOVER: Computational Intelligence for System Design

Model-Based Exploration of Safety (FHA), Protective Controls, and Reconfigurability

UTC proprietary. No technical data subject to the EAR or the ITAR.7

Flight Phase 1

OM 2

OM5

OM4

OM8

OM

10

OM7

OM6

OM 1

OM 3

OM9

“Dark

Aircraft”

(Start & End)All OMs must be

reachable from and reachable to OM1

SPT- & Timing-Feasible

Transition Path (TP)

Transition

Modes

UnreachableOperating Mode

Timing-Infeasible

Transition Paths(take too long)

0,1,7

0,3,5

0,2,3

0,3,5

0,2,7

0,3,7

0,1,5

0,1,5

Supported

Failure Modes

Off

Flight Phase 2 Flight Phase 3

OM

11

0,2,5

Component-Failure

Transition Paths

Flight-Phase

ChangeTransition

Paths

Compliance by Construction Generates only safety-compliant system architectures

Source PrioritiesFinds all Operating Modes and the Transition Paths between them

Architecture = Union(OMs)

Builds Operating Mode Graph, spanning all of the mission’s flight phases.

Reconfigurability: Operating Modes

Discrete Supervisory Control Exploration Dynamic Controls Assessment

Off-DesignModels need to capable of sizing, off-design performance, and dynamics

DynamicsAuto generation of piecewise-defined linear dynamic models

Linear Model & Dynamic RGA