cern, be-abp-cc3 jürgen pfingstner verification of the design of the beam-based controller jürgen...

CERN, BE-ABP-CC3

Jürgen Pfingstner Verification of the Design of the Beam-based Controller

Verification of the Design of the Beam-based Controller

Jürgen Pfingstner

2. June 2009

CERN, BE-ABP-CC3


Content

1. Analysis of the contoller with standard control engineering techniques

2. Uncertanty studies of the response matix and the according control performance

CERN, BE-ABP-CC3


Analysis of the contoller with standard control engineering techniques

• Daniel developed a controller, with common sense and feeling for the system

• I tried to verify this intuitive design with, more abstract and standardized methods:

• Standard nomenclature

• z – transformation

• Time-discrete transfer functions

• Pole-zero plots

CERN, BE-ABP-CC3


The model of the accelerator

1.) Perfect aligned beam line

BPMiQPi

Laser-straight beam

2.) One misaligned QP

yixi

Betatron- oscillation

(given by the beta function of the

lattice)

yjxj(=0)

a.) 2 times xi -> 2 times amplitude -> 2 times yj b.) xi and xj are independent

Linear system without ‘memory’

2

1

2221

1211

2

1

x

x

rr

rr

y

y

Rxy

y … vector of BPM readingsx … vector of the QP displacementsR … response matrix

CERN, BE-ABP-CC3


Mathematical model of the controlled system

(R*)-1+ + z-1 R +

z-1

+

vi

ni

yiri xixi+1

uiui+1 G(z)

C(z)

-

ui, ui+1 … controller state variablesxi, xi+1 … plant state variables

(QP position)

C(z) … ControllerG(z) … Plant

ri … set value (0) … BPM measurementsyi … real beam position vi … ground motionni … BPM noise

iy

iy

CERN, BE-ABP-CC3


System elements(SISO analogon!)

• I controller

z

RzG )(

1)()( 1*

z

zRzC

Be aware about the mathematical not correct writing of the TF (matrix instead of scalar)

• simple

• allpass

• non minimum phase

CERN, BE-ABP-CC3


Stability and Performance

• Stability • necessary attenuation at high frequencies

• all poles at zero

• Performance of the interesting transfer functions

•

•

•

))(1(

11

)(1

)(

)(

)(:)(

1*

RRzz

zR

zL

zG

zv

zyzV

))(1(

1

)(1

1

)(

)(:)(

1*

RRz

z

zLzn

zyzN

))(1(

1)(

)(1

)(

)(

)(:)(

1*1*

RRzRR

zL

zL

zr

zyzR

1

1)()()(:)( 1*

zRRzGzCzL

z

1

CERN, BE-ABP-CC3


Important transfer functions

)(zV)()( zNzR

(ground motion behavior)(set point following and

measurement noise)

CERN, BE-ABP-CC3


Conclusions• Controller is:

• very stable and robust (all poles at zero)

• integrating behavior (errors will die out)

• good general performance

• simple (in most cases a good sign for robustness)

• measurement noise has a strong influence on the output

Further work• H∞ optimal control design

CERN, BE-ABP-CC3


Uncertanty studies of the response matix and the according control performance

Controller is robust, but is it robust enough?

yes no

• Plan A:• Use methods from robust control to adjust the controller to the properties of the uncertainties (e.g. pole shift)

• Plan B:• Use adaptive control techniques to estimate R first and than control accordingly

done

• Answer to that question by simulations in PLACET

CERN, BE-ABP-CC3


Tests in PLACET

• Script in PLACET where the following disturbances can be switched on and off:

• Initial energy Einit

• Energy spread ΔE• QP gradient jitter and systematic errors• Acceleration gradient and phase jitter• BPM noise and failures • Corrector errors• Ground motion

• Additional PLACET function PhaseAdvance • 2 Test series (Robustness according to machine drift):

• Robustness regarding to machine imperfection with perfect controller model

• Robustness regarding to controller model imperfections with perfect machine

• Analysis of the controller performance and the resulting R

CERN, BE-ABP-CC3


Test procedure

1.) Misalign the QP at the begin of the simulation to create an emittance growth at the end of the CLIC main linac

2.) Observe the feedback action in respect to the resulting emittance over time

3.) Change certain accelerator parameter and repeat 1 and 2.

=> The test focus on the stability and convergence speed more than on the steady-state emittance growth and growth rate. (see metric)

CERN, BE-ABP-CC3


Performance metric

0

max

st

t

dtt

m4.0

0 0max

0)(:

)( initialEm

]20[ mspulsestime ][GeVEinitial

Normalized emittance ][)( radnmt

Allowed range for iswhere

initialE05.0)( max mEm initial

maxm

… Emittance produced by the DR and RTML … Emittance at the end of the main linacTheoretical minimum for

0)(t

01.002.05.0min m

CERN, BE-ABP-CC3


ResultsQuantity Acceptable values Nominal values

Einit 8.6 – 9.4 GeV 9.0 GeV ( ± 1 %)

σE 0.0 – 9.0 % 2.0 % (± <1%)

QP gradient jitter 0.0 - 0.25 % < 0.1%

QP gradient error (syst.) 0.0 – 0.25 %

Acceleration gradient variation

0.0 – 0.3 ‰ < 0.3 ‰

Corrector jitter 0.0 – 8.0 nm (std) < 10 nm (std)

BPM noise 0.0 – 500 nm (std) < 50 nm (std)

BPM error distributed *2 110 of 2010 (5%)

BPM block errors *2 6 of 2010

Ground motion 0.68 x 10-3nm/min ( 5d)*1 … Values are according to the multi-pulse projected emittance*2 … Failed BPMs that deliver zeros are worse then the one giving random values.

CERN, BE-ABP-CC3


Conclusions

• Controller is very robust in respect to stability

• It is sufficient in respect to disturbances

CERN, BE-ABP-CC3


Thank you for your attention!

CERN, BE-ABP-CC3


z – Transformation • Method to solve recursive equations• Equivalent to the Laplace – transformation for

time-discrete, linear systems • Allows frequency domain analysis

i

i

zigzG

0

][)(dTiez

z – transfer function

G(z)

Discrete Bode plot][ig

Discrete

System

z

1

CERN, BE-ABP-CC3


Measures for the performance

• Goal: Find properties of Rdist that correspond with the controller performance

i jijnorm

i jij

norm

nomdist

r

m

m

RRM

,

normabs

Rdist … disturbed matrixRnom … nominal matrixabsnorm … absolut matrix norm

normev

)(max1

)(

)( 1*

ii

norm evev

LeigEV

RREL

evnorm … eigenvalue normL … matrix that determines the poles of the control loop

CERN, BE-ABP-CC3


Information from absnorm and evnorm

evnormabsnorm

Controller works well for:• absnorm = 0.0 – 0.4

Controller works well for:• evnorm = 0.5 – 1.0

cern, be-abp-cc3 jürgen pfingstner verification of the design of the beam-based controller jürgen...

Documents