I am trying to reduce the number of parameters of the ODE system: \begin{align} \label{eq:Iprime} I'(t)&= \sigma B-\mu I\\ B'_1(t)&=r_1 B_1\left(1-\frac{B_1 }{K_1}\right)-d_1IB_1-m (B_1-B_2) \\ \label{eq:B2prime} B'_2(t)&=r_2 B_2\left(1-\frac{B_2 }{K_2}\right)-d_2IB_2-m (B_2-B_1) \end{align} (where $B=B_1+B_2$) by applying nondimensionalization. In all the examples I have seen this reduced the number of parameters. However if I apply the technique to my problem the number of parameters remain the same! Why is this the case?

Work thus far

We assume the following scaling: $$I=\widetilde{I}\widehat{I} \quad B_1=\widetilde{B}_1\widehat{B}_1 \quad B_2=\widetilde{B}_2\widehat{B}_2 \quad t=\tilde{t} \hat{t} $$

where $\widetilde{I}, \widetilde{B}_1, \widetilde{B}_2, \tilde{t}$ are constants (dimension-carrying), to be chosen, and $\widehat{I}, \widehat{B}_1, \widehat{B}_2, \hat{t}$ are the dimensionless variables. Subbing these into our ODEs leads to \begin{align} \frac{d(\widetilde{I}\widehat{I})}{d(\tilde{t} \hat{t} )}&=\sigma ( \widetilde{B}_1\widehat{B}_1 + \widetilde{B}_2\widehat{B}_2)-\mu \widetilde{I}\widehat{I}\\ \frac{d \widehat{I}}{d \hat{t}}&=\frac{\tilde{t}\sigma}{\widetilde{I}} ( \widetilde{B}_1\widehat{B}_1 + \widetilde{B}_2\widehat{B}_2)-\frac{\mu \tilde{t}}{\widetilde{I}} \widehat{I} \\ \frac{d \widehat{I}}{d \hat{t}}&=\left[\frac{\tilde{t}\sigma}{\widetilde{I}} \widetilde{B}_1\right] \widehat{B}_1 + \left[ \frac{\tilde{t}\sigma}{\widetilde{I}} \widetilde{B}_2 \right] \widehat{B}_2-\left[\frac{\mu \tilde{t}}{\widetilde{I}}\right] \widehat{I} \\ \end{align}

\begin{align} \frac{d(\widetilde{B}_1\widehat{B}_1)}{d(\tilde{t} \hat{t} )}&=r_1 \widetilde{B}_1\widehat{B}_1\left(1-\frac{\widetilde{B}_1\widehat{B}_1 }{K_1}\right)-d_1\widetilde{I}\widehat{I}\widetilde{B}_1\widehat{B}_1-m (\widetilde{B}_1\widehat{B}_1-\widetilde{B}_2\widehat{B}_2) \\ \frac{d \widehat{B}_1}{d \hat{t}}&=r_1 \tilde{t}\widehat{B}_1\left(1-\frac{\widetilde{B}_1 }{K_1}\widehat{B}_1\right)-d_1\tilde{t }\widetilde{I} \widehat{I}\widehat{B}_1-m\frac{\tilde{t}}{\widetilde{B}_1} (\widetilde{B}_1\widehat{B}_1-\widetilde{B}_2\widehat{B}_2) \\ \frac{d \widehat{B}_1}{d \hat{t}}&=\left[ r_1 \tilde{t}\right]\widehat{B}_1\left(1-\left[ \frac{\widetilde{B}_1 }{K_1}\right]\widehat{B}_1\right)-\left[ d_1\tilde{t } \widetilde{I}\right]\widehat{I}\widehat{B}_1-\left[m \tilde{t} \right] \widehat{B}_1+\left[m \frac{\tilde{t} \widetilde{B}_2}{\widetilde{B}_1} \right]\widehat{B}_2 \end{align} (the procedure for $B_2$ is the same)

The square brackets are to indicate the number of independent choices for the scales. In all of the examples I have seen so far the number of independent choices for the scales are the same as the number of the orginal parameters for the system. At maximum I can only choose four of these choices to be $1$. Meaning that I will have $13-4=9$ parameters remaing. The same as what I started with. What am I doing wrong/why can't I reduce the number of parameters in my system? If you need more clarification on this problem please ask.


$\def\d{\mathrm{d}}$Take$$ \widetilde{t} = \frac{1}{m},\ \widetilde{I} = μ \widetilde{t} = \frac{μ}{m},\ \widetilde{B}_1 = \widetilde{B}_2 = \frac{\widetilde{I}}{σ \widetilde{t}} = \frac{μ}{σ}, $$ then the system is reduced to\begin{align*} \frac{\d \widehat{I}}{\d \widehat{t}} &= \widehat{B}_1 + \widehat{B}_2 - \widehat{I},\\ \frac{\d \widehat{B}_1}{\d \widehat{t}} &= \frac{r_1}{m} \widehat{B}_1 \left( 1 - \frac{μ}{σK_1} \widehat{B}_1 \right) - \frac{μd_1}{m^2} \widehat{I}\widehat{B}_1 - (\widehat{B}_1 - \widehat{B}_2),\\ \frac{\d \widehat{B}_2}{\d \widehat{t}} &= \frac{r_2}{m} \widehat{B}_1 \left( 1 - \frac{μ}{σK_2} \widehat{B}_1 \right) - \frac{μd_2}{m^2} \widehat{I}\widehat{B}_1 - (\widehat{B}_2 - \widehat{B}_1), \end{align*} which has six parameters.

  • $\begingroup$ Can you explain the process you used to get to this point? I want to better understand the process of how you got to where you did. $\endgroup$ – AzJ Mar 15 '18 at 23:52
  • 1
    $\begingroup$ @AzJ First, $\widetilde{B}_2$ should be taken the same as $\widetilde{B}_1$ because they're quantities of the same dimensions. Next, since $r_1$ may not be equal to $r_2$ (the same situation for $K_1, K_2$ and $d_1, d_2$), I tried to normalize the rest parameters. $\endgroup$ – Saad Mar 15 '18 at 23:57
  • $\begingroup$ @AzJ If you find my answer useful, please accept it :) $\endgroup$ – Saad Mar 16 '18 at 0:16
  • $\begingroup$ Okay one more question before I award the bounty is there any greater reason (some kind of structure in ODEs) that this any other ODEs like it can not be reduce to systems of n-m parameters, where n is the number of original parameters, and m is the number of state variables? Or is there just an unusual case that has no pattern? $\endgroup$ – AzJ Mar 16 '18 at 0:21
  • $\begingroup$ @AzJ Actually I don't know much insight on nondimensionalization, but in my opinion, systems like yours can be reduced because variables only appear in polynomial fashion, i.e. $\widehat{B}_1$ and $\widehat{B}_1^2$. If it has terms like $\widehat{B}$ and $\sin\widehat{B}_1$ simultaneously, nondimensionalization doesn't seem to work. $\endgroup$ – Saad Mar 16 '18 at 0:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.