- #1
Ulver48
- 12
- 2
Hello guys,
I try to use the Newton - Chord technique in order to solve a nonlinear system and find it's equilibrium points.This method requires the inverse of the Jacobian Matrix of the nonlinear system. After the linearization around the given starting point x0, I create a linear approximation of the initial problem and then I compute the Jacobian Matrix. Let's call this matrix J. The problem is that this matrix is badly scaled. The system of ODEs consists of 5 equations, 3 of which describe the rate of change for some probabilities P1,P2,P3 (so the range of their values are somewhere between 0 and 1) and two equations that describe the rate of change for an Electric Field ( the real and the imaginary part, who are normalized in such a way that they take values of the order of 1e17).
I use Matlab which tells me that the resulting Jacobian is very badly scaled ( for instance rcond = 1e-18). I tried replacing the Electric Field by a scaled version F, which is equal to the electric field divided by the value 1e18, but it didn't work. So I wanted to ask for your advice. What would someone do in a situation like this ? If you want further information then I will give you more details about my particular set of equations and the values of the resulting Jacobian. I wanted to avoid a really long post and I know that most of you are busy with your own work. Time is a precious thing.I hope for some general guidelines.that could help me deal with my problem. Thank you all for your time.
I try to use the Newton - Chord technique in order to solve a nonlinear system and find it's equilibrium points.This method requires the inverse of the Jacobian Matrix of the nonlinear system. After the linearization around the given starting point x0, I create a linear approximation of the initial problem and then I compute the Jacobian Matrix. Let's call this matrix J. The problem is that this matrix is badly scaled. The system of ODEs consists of 5 equations, 3 of which describe the rate of change for some probabilities P1,P2,P3 (so the range of their values are somewhere between 0 and 1) and two equations that describe the rate of change for an Electric Field ( the real and the imaginary part, who are normalized in such a way that they take values of the order of 1e17).
I use Matlab which tells me that the resulting Jacobian is very badly scaled ( for instance rcond = 1e-18). I tried replacing the Electric Field by a scaled version F, which is equal to the electric field divided by the value 1e18, but it didn't work. So I wanted to ask for your advice. What would someone do in a situation like this ? If you want further information then I will give you more details about my particular set of equations and the values of the resulting Jacobian. I wanted to avoid a really long post and I know that most of you are busy with your own work. Time is a precious thing.I hope for some general guidelines.that could help me deal with my problem. Thank you all for your time.