Class AdamsBashforthIntegrator
- java.lang.Object
-
- org.apache.commons.math4.legacy.ode.AbstractIntegrator
-
- org.apache.commons.math4.legacy.ode.nonstiff.AdaptiveStepsizeIntegrator
-
- org.apache.commons.math4.legacy.ode.MultistepIntegrator
-
- org.apache.commons.math4.legacy.ode.nonstiff.AdamsIntegrator
-
- org.apache.commons.math4.legacy.ode.nonstiff.AdamsBashforthIntegrator
-
- All Implemented Interfaces:
FirstOrderIntegrator
,ODEIntegrator
public class AdamsBashforthIntegrator extends AdamsIntegrator
This class implements explicit Adams-Bashforth integrators for Ordinary Differential Equations.Adams-Bashforth methods (in fact due to Adams alone) are explicit multistep ODE solvers. This implementation is a variation of the classical one: it uses adaptive stepsize to implement error control, whereas classical implementations are fixed step size. The value of state vector at step n+1 is a simple combination of the value at step n and of the derivatives at steps n, n-1, n-2 ... Depending on the number k of previous steps one wants to use for computing the next value, different formulas are available:
- k = 1: yn+1 = yn + h y'n
- k = 2: yn+1 = yn + h (3y'n-y'n-1)/2
- k = 3: yn+1 = yn + h (23y'n-16y'n-1+5y'n-2)/12
- k = 4: yn+1 = yn + h (55y'n-59y'n-1+37y'n-2-9y'n-3)/24
- ...
A k-steps Adams-Bashforth method is of order k.
Implementation details
We define scaled derivatives si(n) at step n as:
s1(n) = h y'n for first derivative s2(n) = h2/2 y''n for second derivative s3(n) = h3/6 y'''n for third derivative ... sk(n) = hk/k! y(k)n for kth derivative
The definitions above use the classical representation with several previous first derivatives. Lets define
(we omit the k index in the notation for clarity). With these definitions, Adams-Bashforth methods can be written:qn = [ s1(n-1) s1(n-2) ... s1(n-(k-1)) ]T
- k = 1: yn+1 = yn + s1(n)
- k = 2: yn+1 = yn + 3/2 s1(n) + [ -1/2 ] qn
- k = 3: yn+1 = yn + 23/12 s1(n) + [ -16/12 5/12 ] qn
- k = 4: yn+1 = yn + 55/24 s1(n) + [ -59/24 37/24 -9/24 ] qn
- ...
Instead of using the classical representation with first derivatives only (yn, s1(n) and qn), our implementation uses the Nordsieck vector with higher degrees scaled derivatives all taken at the same step (yn, s1(n) and rn) where rn is defined as:
(here again we omit the k index in the notation for clarity)rn = [ s2(n), s3(n) ... sk(n) ]T
Taylor series formulas show that for any index offset i, s1(n-i) can be computed from s1(n), s2(n) ... sk(n), the formula being exact for degree k polynomials.
The previous formula can be used with several values for i to compute the transform between classical representation and Nordsieck vector. The transform between rn and qn resulting from the Taylor series formulas above is:s1(n-i) = s1(n) + ∑j>0 (j+1) (-i)j sj+1(n)
where u is the [ 1 1 ... 1 ]T vector and P is the (k-1)×(k-1) matrix built with the (j+1) (-i)j terms with i being the row number starting from 1 and j being the column number starting from 1:qn = s1(n) u + P rn
[ -2 3 -4 5 ... ] [ -4 12 -32 80 ... ] P = [ -6 27 -108 405 ... ] [ -8 48 -256 1280 ... ] [ ... ]
Using the Nordsieck vector has several advantages:
- it greatly simplifies step interpolation as the interpolator mainly applies Taylor series formulas,
- it simplifies step changes that occur when discrete events that truncate the step are triggered,
- it allows to extend the methods in order to support adaptive stepsize.
The Nordsieck vector at step n+1 is computed from the Nordsieck vector at step n as follows:
- yn+1 = yn + s1(n) + uT rn
- s1(n+1) = h f(tn+1, yn+1)
- rn+1 = (s1(n) - s1(n+1)) P-1 u + P-1 A P rn
[ 0 0 ... 0 0 | 0 ] [ ---------------+---] [ 1 0 ... 0 0 | 0 ] A = [ 0 1 ... 0 0 | 0 ] [ ... | 0 ] [ 0 0 ... 1 0 | 0 ] [ 0 0 ... 0 1 | 0 ]
The P-1u vector and the P-1 A P matrix do not depend on the state, they only depend on k and therefore are precomputed once for all.
- Since:
- 2.0
-
-
Nested Class Summary
-
Nested classes/interfaces inherited from class org.apache.commons.math4.legacy.ode.MultistepIntegrator
MultistepIntegrator.NordsieckTransformer
-
-
Field Summary
-
Fields inherited from class org.apache.commons.math4.legacy.ode.MultistepIntegrator
nordsieck, scaled
-
Fields inherited from class org.apache.commons.math4.legacy.ode.nonstiff.AdaptiveStepsizeIntegrator
mainSetDimension, scalAbsoluteTolerance, scalRelativeTolerance, vecAbsoluteTolerance, vecRelativeTolerance
-
Fields inherited from class org.apache.commons.math4.legacy.ode.AbstractIntegrator
isLastStep, resetOccurred, stepHandlers, stepSize, stepStart
-
-
Constructor Summary
Constructors Constructor Description AdamsBashforthIntegrator(int nSteps, double minStep, double maxStep, double[] vecAbsoluteTolerance, double[] vecRelativeTolerance)
Build an Adams-Bashforth integrator with the given order and step control parameters.AdamsBashforthIntegrator(int nSteps, double minStep, double maxStep, double scalAbsoluteTolerance, double scalRelativeTolerance)
Build an Adams-Bashforth integrator with the given order and step control parameters.
-
Method Summary
All Methods Instance Methods Concrete Methods Modifier and Type Method Description void
integrate(ExpandableStatefulODE equations, double t)
Integrate a set of differential equations up to the given time.-
Methods inherited from class org.apache.commons.math4.legacy.ode.nonstiff.AdamsIntegrator
initializeHighOrderDerivatives, updateHighOrderDerivativesPhase1, updateHighOrderDerivativesPhase2
-
Methods inherited from class org.apache.commons.math4.legacy.ode.MultistepIntegrator
computeStepGrowShrinkFactor, getMaxGrowth, getMinReduction, getNSteps, getSafety, getStarterIntegrator, setMaxGrowth, setMinReduction, setSafety, setStarterIntegrator, start
-
Methods inherited from class org.apache.commons.math4.legacy.ode.nonstiff.AdaptiveStepsizeIntegrator
filterStep, getCurrentStepStart, getMaxStep, getMinStep, initializeStep, resetInternalState, sanityChecks, setInitialStepSize, setStepSizeControl, setStepSizeControl
-
Methods inherited from class org.apache.commons.math4.legacy.ode.AbstractIntegrator
acceptStep, addEventHandler, addEventHandler, addStepHandler, clearEventHandlers, clearStepHandlers, computeDerivatives, getCounter, getCurrentSignedStepsize, getEvaluations, getEventHandlers, getExpandable, getMaxEvaluations, getName, getStepHandlers, initIntegration, integrate, setEquations, setMaxEvaluations, setStateInitialized
-
-
-
-
Constructor Detail
-
AdamsBashforthIntegrator
public AdamsBashforthIntegrator(int nSteps, double minStep, double maxStep, double scalAbsoluteTolerance, double scalRelativeTolerance) throws NumberIsTooSmallException
Build an Adams-Bashforth integrator with the given order and step control parameters.- Parameters:
nSteps
- number of steps of the method excluding the one being computedminStep
- minimal step (sign is irrelevant, regardless of integration direction, forward or backward), the last step can be smaller than thismaxStep
- maximal step (sign is irrelevant, regardless of integration direction, forward or backward), the last step can be smaller than thisscalAbsoluteTolerance
- allowed absolute errorscalRelativeTolerance
- allowed relative error- Throws:
NumberIsTooSmallException
- if order is 1 or less
-
AdamsBashforthIntegrator
public AdamsBashforthIntegrator(int nSteps, double minStep, double maxStep, double[] vecAbsoluteTolerance, double[] vecRelativeTolerance) throws IllegalArgumentException
Build an Adams-Bashforth integrator with the given order and step control parameters.- Parameters:
nSteps
- number of steps of the method excluding the one being computedminStep
- minimal step (sign is irrelevant, regardless of integration direction, forward or backward), the last step can be smaller than thismaxStep
- maximal step (sign is irrelevant, regardless of integration direction, forward or backward), the last step can be smaller than thisvecAbsoluteTolerance
- allowed absolute errorvecRelativeTolerance
- allowed relative error- Throws:
IllegalArgumentException
- if order is 1 or less
-
-
Method Detail
-
integrate
public void integrate(ExpandableStatefulODE equations, double t) throws NumberIsTooSmallException, DimensionMismatchException, MaxCountExceededException, NoBracketingException
Integrate a set of differential equations up to the given time.This method solves an Initial Value Problem (IVP).
The set of differential equations is composed of a main set, which can be extended by some sets of secondary equations. The set of equations must be already set up with initial time and partial states. At integration completion, the final time and partial states will be available in the same object.
Since this method stores some internal state variables made available in its public interface during integration (
AbstractIntegrator.getCurrentSignedStepsize()
), it is not thread-safe.- Specified by:
integrate
in classAdamsIntegrator
- Parameters:
equations
- complete set of differential equations to integratet
- target time for the integration (can be set to a value smaller thant0
for backward integration)- Throws:
NumberIsTooSmallException
- if integration step is too smallDimensionMismatchException
- if the dimension of the complete state does not match the complete equations sets dimensionMaxCountExceededException
- if the number of functions evaluations is exceededNoBracketingException
- if the location of an event cannot be bracketed
-
-