1© 2015 The MathWorks, Inc.

Managing Performance and

Safety in Multi-Domain

Complex Systems

Juan Sagarduy

2

Managing Performance & Safety in Multi-Domain Systems

Performance

size-weight

autonomy

Safety

emergency response

fault management

Electric Vehicle

Active Safety

3

Managing Performance & Safety in Multi-Domain Systems

1. System design

(integration, optimization)

4

1. System design

(integration, optimization)

2. Development of

control algorithms

Managing Performance & Safety in Multi-Domain Systems

5

1. System design

(integration, optimization)

2. Development of

control algorithms

3. Hardware-based

physical emulation

(real-time testing)

Managing Performance & Safety in Multi-Domain Systems

6

1. System Design

Explore – Integrate – Optimize

7

Active Safety – System Design

Step 1. Model configuration (no supervisory logic)

to understand physical behaviour and add controllability

* models available upon request

8

Active Safety – System Design

• Refine requirements

• Set-up test scenarios

• Report & align with others

9

Active Safety >> System Design -> Battery Cooling

Battery

Temperature

(K)

Dissipated

Heat

(kW)

10

Active Safety >> System Design -> Vehicle Dynamics

Tire model

assumption

(model fidelity)

Tractive

torque

11

2. Algorithm Development

Regulate – Tune – Supervise

12

Active Safety – Algorithm Development

Structure and threshold values

are critical to supervisory logic design.

Physical state of all components is decisive.

13

Structure and threshold values are critical to supervisory logic design.

Physical state of all components is decisive.

Active Safety – Algorithm Development > Supervisory Logic

14

Are physical boundaries trespassed?

Is safety guaranteed?

Active Safety – Algorithm Development -> braking

What is the trade-off

between hydraulic

and regenerative braking?

Hydraulic brake

capacity

Regenerative

motor

currents

15

Active Safety - Prognosis > fault/degradation signature(s)

16

Active Safety - Prognosis > fault/degradation signature(s)

Normal Eccentricity

Phase current Phase current

17

How is the signature quantity

i.e. current affected

by deviations/uncertainties

In physical/control properties?MonteCarlo simulations

(Sensitivity Analysis with

Parallel Computing)

agressive PI gains

enhance

the fault signature

Motor flux constant

& inductance highly

influence current

Active Safety - Prognosis > Sensitivity & Calibration

18

3. Testing

Verify your solution in real-time

19

Active Safety - Real-time Testing

Prepare model for real-time testinga. golden reference >> controller (fixed-step) + Plant (variable-step)

b. controller + Plant (fixed-step solver)

c. optimal solution >> robustness vs. speed

d. test on hardware (Speedgoat)

20

Conclusions

21

Physical model

& algorithmsPerformance Safety

Configurability

Insight - Flexibility

22

MATLAB/Simulink offers a unique environment

for data analytics & embedded development

23

Partnership with MathWorks

reduces risk and accelerates the adoption process