/[ascend]/trunk/base/generic/packages/sensitivity.h
ViewVC logotype

Diff of /trunk/base/generic/packages/sensitivity.h

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1162 by johnpye, Wed Jan 17 04:34:59 2007 UTC revision 1163 by johnpye, Wed Jan 17 06:33:30 2007 UTC
# Line 19  Line 19 
19      Sensitivity add-on for ASCEND      Sensitivity add-on for ASCEND
20      by Tom Epperly, Kirk Abbot      by Tom Epperly, Kirk Abbot
21    
22      @note This is a static package in ASCEND. It can not be built as a dynamic package      -- notes here have been moved to models/sensitivity/sensitivity.c --
     because it doesn't contain the self-registration stuff.  
   
     This file attempts to implement the extraction of dy_dx from  
     a system of equations. If one considers a black-box where x are  
     the input variables, u are input parameters, y are the output  
     variables, f(x,y,u) is the system of equations that will be solved  
     for given x and u, then one can extract the sensitivity information  
     of y wrt x.  
   
     One crude but simple way of doing this is to finite-difference the  
     given black box, i.e, perturb x, n\subx times by delta x, resolve  
     the system of equations at the new value of x, and compute  
     dy/dx = (f(x\sub1) - f(x\sub2))/(x\sub1 - x\sub2).  
     This is known to be very expensive.  
   
     The solution that will be adopted here is to formulate the Jacobian J of  
     the system, (or the system is already solved, to grab the jacobian at  
     the solution. Develop the sensitivity matrix df/dx by exact numnerical  
     differentiation; this will be a n x n\subx matrix. Compute the LU factors  
     for J, and to solve column by column to : LU dz/dx = df/dx. Here  
     z, represents all the internal variables to the system and the strictly  
     output variables y. In other words J = df/dz.  
   
     Given the solution of the above problem we then simply extract the rows  
     of dz/dx, that pertain to the y variables that we are interested in;  
     this will give us dy/dx.  
   
     @todo Do we really need 2 files called [Ss]ensitivity.[ch]?  Other one is in tcltk.  
   
     @todo Make this self-registering so that it can be compiled as a dlopenable library.  
23  */  */
24    
25  #ifndef ASC_SENSITIVITY_H  #ifndef ASC_SENSITIVITY_H
# Line 60  Line 30 
30  #include <compiler/compiler.h>  #include <compiler/compiler.h>
31  #include <general/list.h>  #include <general/list.h>
32  #include <compiler/extfunc.h>  #include <compiler/extfunc.h>
   
33  #include <solver/mtx.h>  #include <solver/mtx.h>
34  #include <solver/mtx_basic.h>  #include <solver/mtx_basic.h>
35  #include <solver/mtx_perms.h>  #include <solver/mtx_perms.h>
# Line 100  ASC_DLLSPEC int Compute_dy_dx_smart(slv_ Line 69  ASC_DLLSPEC int Compute_dy_dx_smart(slv_
69  */  */
70    
71  ASC_DLLSPEC int DoSolve(struct Instance *inst);  ASC_DLLSPEC int DoSolve(struct Instance *inst);
 ASC_DLLSPEC int DoDataAnalysis(struct var_variable **inputs,  
               struct var_variable **outputs,  
               int ninputs, int noutputs,  
               real64 **dy_dx);  
72  ASC_DLLSPEC int finite_difference(struct gl_list_t *arglist);  ASC_DLLSPEC int finite_difference(struct gl_list_t *arglist);
73  ASC_DLLSPEC struct Instance *FetchElement(struct gl_list_t *arglist,  ASC_DLLSPEC struct Instance *FetchElement(struct gl_list_t *arglist,
74                       unsigned long whichbranch,                       unsigned long whichbranch,

Legend:
Removed from v.1162  
changed lines
  Added in v.1163

john.pye@anu.edu.au
ViewVC Help
Powered by ViewVC 1.1.22