12/2/2019

Ali M. Sahlodin
Department of Chemical Engineering
AmirKabir University of Technology
1397 S.H

 Methods of event detection and location

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 2

1
 12/2/2019

 Computational burden of Newton iterations

at each step.
 What if the current step does not contain any
event?
g(t)
Existence test g (t k ) g (t k 1 )  0
tk t* tk+1 t
Can the test fail to detect an
existing event? g(t)

Even number of events within a

time step tk tk+1 t
t1* t2*

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 3

 Case I: multiple zero- g(t)

crossings of discontinuity
function
tk tk+1t
L : g1 (t )  0

 Case II: state conditions L : ( g1 (t )  0)  ( g 2 (t )  0)

with conjunctions
g1>0 g1>0 State condition
(Multiple discontinuity functions True False False
with AND conditions) True True True
False True False
Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 4

2
 12/2/2019

Park & Barton, ACM Transactions on Modeling and

Computer Simulation 6 (2) 1996.

Steam trap valve opens only if (P  0)  ( H  H min )

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 5

 Existence test can fail

g(t)
g (tk ) g (tk 1 )  0
tk tk+1 t

 No guarantee that the Newton or bisection

method will find the first zero-crossing
 Completely incorrect state trajectories

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 6

3
 12/2/2019

Example (Park and Barton 1996)

Wrong trajectories Correct trajectories
(event-detection method (event-detection method
missing some events) detecting all events)

ACM Transactions on Modeling and Computer Simulation 6 (2) 1996.

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 7

 Rigorous root-finding using Interval Analysis

 Interval arithmetic

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 8

4
 12/2/2019

t  [tk , tk 1 ]  g p (t )  [?,?]
gU
gU

G g g
G
gL
gL
tk tk+1 tk tk+1
Non-existence test: 0  G (T )
what if 0  G (T )?
Guaranteed to have no
transition!
Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 9

Assumption: the new mode works with the same variables

 Previous mode f ( k ) ( x ( k ) , x( k ) , y ( k ) , u ( k ) , p)  0

 New mode f ( k 1) ( x ( k 1) , x( k 1) , y ( k 1) , u ( k 1) , p)  0

 How to initialize the system in the new mode?

 Transition function

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 10

5
 12/2/2019

f1 (t , x) g (t , x)  0
 How many dynamic DOF in the x = 
new mode? f 2 (t , x) g (t , x) > 0

 Natural transition: continuity of

state variables
T21  x ( k 1) (t * )  x ( k ) (t * )
Pulse
injection
 Impulsive transition: jump in of product
state variables
 Examples

T21  x( k 1) (t * )  x( k ) (t * )  x
Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 11

More complicated that ODEs!

 How many dynamic DOF in the new mode?
 Consistent initialization analysis

 What are those DOF?

Example

 Event: step change in f(t)

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 12

6
 12/2/2019

 Step in f(t)
 Keep x1 continuous  x2 must jump
 Keep x2 continuous  x1 must jump Both are wrong!
 Underlying ODE shows both x1 and x2 must jump
x1  2 f (t )  x1  2 f (t )  c1
x   f (t )  x   f (t )  c
1 2 2

 However, the following quantity remains continuous

(why?)
x1( k )  2 x2( k )  x1( k 1)  2 x2( k 1)

Copyright © Ali M. Sahlodin, Dept. of Chemical Engineering, AmirKabir Univ. of Tech. 13