Author |
Message |
anddychen (anddy)
Member Username: anddy
Post Number: 4 Registered: 10-2009
| Posted on Saturday, October 10, 2009 - 10:46 pm: | |
What is the meaning of Running Error"Timestep has fallen below 1e-009 of starting value!You may have a temporal discontinuity in parameters or you may simply need a smaller BY clause in the TIME section" Re-edited script, Thank you for your help.
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1299 Registered: 06-2003
| Posted on Sunday, October 11, 2009 - 03:27 pm: | |
FlexPDE dynamically controls the time step size to find a range in which the solution progresses smoothly in time. Some errors in the description of a problem can cause the timestep to be cut endlessly, because there is no smooth solution. FlexPDE has a test for this occurrence, and stops the problem when the step size has been cut to 1e-9 times the initial timestep. You still have no boundary conditions specified in your script. The I equation is therefore underspecified, as only deriviative BCs are applied by default. This equation is first-order in z, so you need a value condition on the input Z surface to make the solution unique.
|
Aditya Putranto (adityaptr)
New member Username: adityaptr
Post Number: 1 Registered: 12-2009
| Posted on Sunday, December 06, 2009 - 11:39 pm: | |
Hi Robert, I try to use flex PDE for parameter estimation to obtain parameter (D) from unsteady mass and heat balance. For this purpose, I need to fit the model (from flex PDE) and experimental data. However, the experimental data is dependent on time (unsteady). I have followed examples (heaterssi and heaterti) by imposing afitting equation. In my case I wrote: dt(xc)=setpoint-xc where: setpoint=(0.0095*(t/3600)^4 - 0.1743*(t/3600)^3 + 1.2215*(t/3600)^2 - 4.4566*(t/3600) + 9.0378) xc=integral(u)/integral(1) However, it was said : time has fallen below 1e-9. What's wrong with this? Could you help me with this? Many thanks
|
Marek Nelson (mgnelson)
Moderator Username: mgnelson
Post Number: 164 Registered: 07-2007
| Posted on Monday, December 07, 2009 - 12:01 am: | |
This means you have a discontinuity in your initial conditions or your initial timestep is too large. I can't comment more without seeing your script. If you post your script (.pde file), I can take a look and maybe help further.
|
Aditya Putranto (adityaptr)
New member Username: adityaptr
Post Number: 2 Registered: 12-2009
| Posted on Monday, December 07, 2009 - 12:08 am: | |
Robert, I attach the script below. Could you please help me what's wrong with the script? Thanks
|
Marek Nelson (mgnelson)
Moderator Username: mgnelson
Post Number: 165 Registered: 07-2007
| Posted on Monday, December 07, 2009 - 02:08 pm: | |
You have many nonlinear terms in the system and they all feedback onto one another, so it is difficult to say what is the cause of the problem. I suggest starting with a simplified system with constant coefficients and add the nonlinear terms one at a time. Plot all the variables and nonlinear terms so that you can see how they are behaving. Think about what they mean to the equations and how they will effect the values of the variables. I have attached a simplified model (no nonlinear terms in the equations - and the global D removed.) It runs and plots some of your terms.
|
Aditya Putranto (adityaptr)
Junior Member Username: adityaptr
Post Number: 3 Registered: 12-2009
| Posted on Tuesday, December 08, 2009 - 01:21 am: | |
Hi Marek Nelson, In this case, I need to estimate parameter D so that global variables cannot be eliminated. I tell you the problem I want to solve: I have data of average moisture content vs time at different time, I want to obtain D from this experimental data. For this purpose, I need to generate equations (mass and heat balance) and impose a objective function. In this case, the objective function is written: dt(xc)=setpoint-xc (as example hearterti). I used set point as function of time which I generated from data of moisture content versus time (experimental). However, it cannot work. I attach the simplified script for estimation of D. What's wrong with this? Do you have any ideas to determine D from data of moisture content vs time (several points)?What equation should I use for objective function? Thanks for your help Aditya
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1309 Registered: 06-2003
| Posted on Wednesday, December 09, 2009 - 02:53 pm: | |
You have several problems with this script. 1. Your equation for U has no sources, so at large time U will simply equilibrate to the boundary value. D controls how rapidly this equilibration occurs, but it cannot cause U to be larger or smaller than the boundary value (except insofar as U may start out higher or lower than the BC and transition through your desired mean on its way to equilibrium). 2. Xc is defined simply as the average value of U. You cannot apply a time derivative to this quantity, as it is already defined by the U equation. And your equation is not, in fact, the same as that in our Heaterti sample. In that example we control the power input, which does have the capability of raising and lowering U. In our sample, the time derivative is applied to power, not to the average of U. 3. D controls the rate of equilibration of U to the boundary value, but you want to track a moving setpoint. If the setpoint moves faster than the equilibration speed, you will have a pursuit problem with poor definition. 4. If you want to solve for a value of D that is constant over time, this formulation will not succeed, because D will be time dependent.
|