PostmaShmoylova.ppt 1301KB Jun 23 2011 12:12:12 PM
Computer Algebra vs. Reality
Erik Postma and Elena Shmoylova
Maplesoft
June 25, 2009
© 2009 Maplesoft, a division of Waterloo Maple Inc.
Outline
• Introduction
• How to apply computer algebra techniques to
real world problems?
• Example
• Open discussion
© 2009 Maplesoft, a division of Waterloo Maple Inc.
2
Introduction
• Computer algebra is based on symbolic
computations
• Benefit: Result is a nice closed form solution
• Drawback: Problem itself should be nice too
© 2009 Maplesoft, a division of Waterloo Maple Inc.
3
Computer Algebra Methods
• Polynomial solvers for polynomial systems with
coefficients in a rational extension field
• Differential Groebner basis for polynomial DEs with
coefficients in a rational extension field
• Functional decomposition for multi- or univariate
polynomials over a rational extension field
• Index reduction for continuous and in some cases
piecewise-continuous models
© 2009 Maplesoft, a division of Waterloo Maple Inc.
4
Common Elements of Real-World
Problems
•
•
•
•
•
•
Floating point numbers and powers
Trigonometric and other special functions
Lookup tables
Piecewise functions
Numerical differentiators
Compiled numerical procedures (“black-box”
functions)
• Delay elements
• Random noise terms
• etc.
© 2009 Maplesoft, a division of Waterloo Maple Inc.
5
How to apply computer algebra
techniques to real-world problems?
© 2009 Maplesoft, a division of Waterloo Maple Inc.
6
Convert One Type of Difficulty into
Another
• Look-up tables into piecewise
• Almost anything into black-box function
• Approximate functions by their Taylor or Padé
series
• Smooth piecewise functions, e.g. using radial
basis functions
• Floating point numbers into rationals
© 2009 Maplesoft, a division of Waterloo Maple Inc.
7
Remove Difficulty from Model
• If a difficulty can be combined into a subsystem,
remove the subsystem from the model
– View its arguments as outputs of the model
– View its result as inputs into the model
– Use symbolic technique on the model
• Limited to techniques that can deal with arbitrary
external inputs
© 2009 Maplesoft, a division of Waterloo Maple Inc.
8
Floating Point Numbers
• Replace with rational numbers
© 2009 Maplesoft, a division of Waterloo Maple Inc.
9
Initial Conditions for Hybrid DAE
Models
• Problem:
– User does not provide all initial conditions, need
to find remaining initial conditions
• Difficulty:
– High-order DAEs have hidden constraints that may
be needed to find initial conditions
© 2009 Maplesoft, a division of Waterloo Maple Inc.
10
Simple Example
• DAEs
x1 2 x1
x1 0
x2 1
x2
( x2 1) 1 x1 0
x12 x22 4
x1 0
0 2
2
x1 ( x2 1) 1 x1 0
• ICs
x 1 0.75
x2 1.6
x 2 1
© 2009 Maplesoft, a division of Waterloo Maple Inc.
11
Identifying Mode (I)
• From constraint
1.2 x1 0
x1
0.8 x1 0
• Do not know what branch to choose
• Index reduction can be performed on both
branches
• Hidden constraint
x1 x 1 x2 x 2
0
x1 x 1 ( x2 1) x 2
x1 0
x1 0
© 2009 Maplesoft, a division of Waterloo Maple Inc.
12
Identifying Mode (II)
• Check which branch of the hidden constraint is
satisfied
x1 x 1 x2 x 2
x1 x 1 ( x2 1) x 2
•
•
x1 0
x1 0 2.5 x1 0
x1 0 0 x1 0
mode is active
x1 0.8
© 2009 Maplesoft, a division of Waterloo Maple Inc.
13
Initial Conditions for Hybrid DAEs
• To find ICs, hidden constraints are needed
• To find hidden constraints, index reduction
should be performed
• It is infeasible to perform index reduction for
all modes separately, need to know what
mode system is in
• To find mode of system, need to know the
values of all variables, i.e. ICs
© 2009 Maplesoft, a division of Waterloo Maple Inc.
14
Open Discussion:
How to apply computer algebra
techniques to real-world problems?
© 2009 Maplesoft, a division of Waterloo Maple Inc.
15
Erik Postma and Elena Shmoylova
Maplesoft
June 25, 2009
© 2009 Maplesoft, a division of Waterloo Maple Inc.
Outline
• Introduction
• How to apply computer algebra techniques to
real world problems?
• Example
• Open discussion
© 2009 Maplesoft, a division of Waterloo Maple Inc.
2
Introduction
• Computer algebra is based on symbolic
computations
• Benefit: Result is a nice closed form solution
• Drawback: Problem itself should be nice too
© 2009 Maplesoft, a division of Waterloo Maple Inc.
3
Computer Algebra Methods
• Polynomial solvers for polynomial systems with
coefficients in a rational extension field
• Differential Groebner basis for polynomial DEs with
coefficients in a rational extension field
• Functional decomposition for multi- or univariate
polynomials over a rational extension field
• Index reduction for continuous and in some cases
piecewise-continuous models
© 2009 Maplesoft, a division of Waterloo Maple Inc.
4
Common Elements of Real-World
Problems
•
•
•
•
•
•
Floating point numbers and powers
Trigonometric and other special functions
Lookup tables
Piecewise functions
Numerical differentiators
Compiled numerical procedures (“black-box”
functions)
• Delay elements
• Random noise terms
• etc.
© 2009 Maplesoft, a division of Waterloo Maple Inc.
5
How to apply computer algebra
techniques to real-world problems?
© 2009 Maplesoft, a division of Waterloo Maple Inc.
6
Convert One Type of Difficulty into
Another
• Look-up tables into piecewise
• Almost anything into black-box function
• Approximate functions by their Taylor or Padé
series
• Smooth piecewise functions, e.g. using radial
basis functions
• Floating point numbers into rationals
© 2009 Maplesoft, a division of Waterloo Maple Inc.
7
Remove Difficulty from Model
• If a difficulty can be combined into a subsystem,
remove the subsystem from the model
– View its arguments as outputs of the model
– View its result as inputs into the model
– Use symbolic technique on the model
• Limited to techniques that can deal with arbitrary
external inputs
© 2009 Maplesoft, a division of Waterloo Maple Inc.
8
Floating Point Numbers
• Replace with rational numbers
© 2009 Maplesoft, a division of Waterloo Maple Inc.
9
Initial Conditions for Hybrid DAE
Models
• Problem:
– User does not provide all initial conditions, need
to find remaining initial conditions
• Difficulty:
– High-order DAEs have hidden constraints that may
be needed to find initial conditions
© 2009 Maplesoft, a division of Waterloo Maple Inc.
10
Simple Example
• DAEs
x1 2 x1
x1 0
x2 1
x2
( x2 1) 1 x1 0
x12 x22 4
x1 0
0 2
2
x1 ( x2 1) 1 x1 0
• ICs
x 1 0.75
x2 1.6
x 2 1
© 2009 Maplesoft, a division of Waterloo Maple Inc.
11
Identifying Mode (I)
• From constraint
1.2 x1 0
x1
0.8 x1 0
• Do not know what branch to choose
• Index reduction can be performed on both
branches
• Hidden constraint
x1 x 1 x2 x 2
0
x1 x 1 ( x2 1) x 2
x1 0
x1 0
© 2009 Maplesoft, a division of Waterloo Maple Inc.
12
Identifying Mode (II)
• Check which branch of the hidden constraint is
satisfied
x1 x 1 x2 x 2
x1 x 1 ( x2 1) x 2
•
•
x1 0
x1 0 2.5 x1 0
x1 0 0 x1 0
mode is active
x1 0.8
© 2009 Maplesoft, a division of Waterloo Maple Inc.
13
Initial Conditions for Hybrid DAEs
• To find ICs, hidden constraints are needed
• To find hidden constraints, index reduction
should be performed
• It is infeasible to perform index reduction for
all modes separately, need to know what
mode system is in
• To find mode of system, need to know the
values of all variables, i.e. ICs
© 2009 Maplesoft, a division of Waterloo Maple Inc.
14
Open Discussion:
How to apply computer algebra
techniques to real-world problems?
© 2009 Maplesoft, a division of Waterloo Maple Inc.
15