Review of ODE

Signals and systems——ODE

• ODE
  • Time-Domain
    • Classical Method
    • State Variable Method
    • Numerical Method
      • Euler Method
        • RK-4 Method
  • S-domain
    • LT
    • FT

Time Domain

Classical Method

   The essence of solving linear system is to solve linear constant coefficient system differential equation. This method is called classical method.
   For a general form of constant coefficient differential equations:
$$C_0\frac{d^n}{d{t}^n}r(t)+C_1\frac{d^{n-1}}{d{t}^{n-1}}r(t)+\cdots +C_{n-1}\frac{d}{dt}r(t)+C_{n}r(t)=E_0\frac{d^m}{d{t}^m}e(t)+E_1\frac{d^{m-1}}{d{t}^{m-1}}e(t)+\cdots +E_{m-1}\frac{d}{dt}e(t)+E_{m}e(t)$$
   The solution is of the form
$$r(t)=r_h(t)+r_p(t)$$
where $r_h(t)$ is the homogeneous solution, $r_p(t)$ is the particular solution.
For the homogeneous equation
$$C_0\frac{d^n}{d{t}^n}r(t)+C_1\frac{d^{n-1}}{d{t}^{n-1}}r(t)+\cdots +C_{n-1}\frac{d}{dt}r(t)+C_{n}r(t)=0$$
We can solve it by analysing the characteristic equation first so that we derive the characteristic roots ${\lambda }_1,{\lambda }_2...{\lambda }_{n-1},{\lambda }_n$ from it.
If these roots are not equal to each other,
$$r_h(t)=\sum _{i=1}^n{A}_i{e}^{{\lambda }_{i}t}$$
If there are some multiple roots (Supposing ${\lambda }_1$ is a $k$ multiple number root )
$$r_h(t)=\left(\sum _{j=1}^k{B}_j{t}^{k-j}\right)e^{{\lambda }_{1}t}+\sum _{i=k+1}^n{A}_i{e}^{{\lambda }_{i}t}$$
The particular solution $r_p(t)$ depends on the specific form of the activation function

$e(t)$	$r_p(t)$
$E$	$B$
$t^p$	$B_1{t}^p+B_2{t}^{p-1}+...+B_{p}t+B_{1+p}$
$e^{at}$	$B{e}^{at}$
$cos(\omega t)$ $or$ $sin(\omega t)$	$B_{1}cos(\omega t)+B_{2}sin(\omega t)$
$t^p{e}^{at}cos(\omega t)$ $or$ $t^p{e}^{at}sin(\omega t)$	$(B_1{t}^p+B_2{t}^{p-1}+...+B_{p}t+B_{1+p})e^{at}cos(\omega t)+(D_1{t}^p+D_2{t}^{p-1}+...+D_{p}t+D_{1+p})e^{at}sin(\omega t)$
e.g.$$\frac{d^{2}r(t)}{d{t}^2}+2\frac{dr(t)}{dt}+3r(t)=\frac{de(t)}{dt}+e(t),e(t)=t^2$$
Solve the characteristic equation
$${\lambda }^2+3\lambda +2=0$$
so we can obtain
$${\lambda }_1=-1,{\lambda }_2=-2$$
Thus,
$$r_h(t)=C_1{e}^{-t}+C_2{e}^{-2t}$$
as $e(t)=t^2$ ,we suppose $r_p(t)=B_1{t}^2+B_{2}t+B_3$
Using method of undetermined coefficients,
$$r_p(t)=t^2-2t+2$$

State Variable Method

$$\left[\begin{array}{l}{\dot{x}}_1\\ {\dot{x}}_2\\ ⋮\\ {\dot{x}}_n\end{array}\right]=\left[\begin{array}{cccc}{a}_{11}& a_{12}& \dots & a_{1n}\\ a_{21}& a_{22}& \dots & a_{2n}\\ ⋮& ⋮& ⋮& ⋮\\ a_{n1}& a_{n2}& \dots & a_{nn}\end{array}\right]\left[\begin{array}{l}{x}_1\\ x_2\\ ⋮\\ x_n\end{array}\right]+\left[\begin{array}{cccc}{b}_{11}& b_{12}& \dots & b_{1m}\\ b_{21}& b_{22}& \dots & b_{2m}\\ ⋮& ⋮& ⋮& ⋮\\ b_{n1}& b_{n2}& \dots & b_{nm}\end{array}\right]\left[\begin{array}{l}{f}_1\\ f_2\\ ⋮\\ f_m\end{array}\right]$$
For example, you could define the $x$ variable the current
$$\left[\begin{array}{l}{y}_1\\ y_2\\ ⋮\\ y_n\end{array}\right]=\left[\begin{array}{cccc}{c}_{11}& c_{12}& \dots & c_{1n}\\ c_{21}& c_{22}& \dots & c_{2n}\\ ⋮& ⋮& ⋮& ⋮\\ c_{n1}& c_{n2}& \dots & c_{nn}\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\\ ⋮\\ x_n\end{array}\right]+\left[\begin{array}{cccc}{d}_{11}& d_{12}& \dots & d_{1m}\\ d_{21}& d_{22}& \dots & d_{2m}\\ ⋮& ⋮& ⋮& ⋮\\ d_{n1}& d_{n2}& \dots & d_{nm}\end{array}\right]\left[\begin{array}{l}{f}_1\\ f_2\\ ⋮\\ f_m\end{array}\right]$$
By this you can solve the whole current equation.

Numerical Method

Each differential equation corresponds to a curve.
The numerical solution of differential equations is essentially to calculate the shape of a curve. This particular curve will have a particular starting point and a particular differential equation. Now suppose we know a starting point, and we can use the equation to figure out the slope of the tangent line of the curve at that point, and then we can follow the tangent line with a small enough displacement to get to the next point, and then use that point as the starting point of the second calculation until we have drawn the whole curve. So, the difference between different numerical methods is how do you get from one point to the next

Euler Method

Take the following differential equation
$$y^{\prime }(t)=f(t,y(t))y\left(t_0\right)=y_0$$
at $(t_0,y(t_0))$, the equation of the tangent line passing this point is
$$y-y\left(t_0\right)=y^{\prime }\left(t_0\right)\left(t-t_0\right)=f\left(t_0,y\left(t_0\right)\right)\left(t-t_0\right)$$
At the next moment $t_1$, points $(t_1,y(t_1))$ on the curve can be replaced by points on the tangent line
$$y\left(t_1\right)=y\left(t_0\right)+f\left(t_0,y\left(t_0\right)\right)\left(t_1-t_0\right)$$
Iteration form
$$y_{n+1}=y_n+hf\left(t_n,y_n\right)$$

RK-4 Method

Iteration form
$$\begin{array}{l}{y}_{n+1}=y_n+\frac{h}{6}\left(k_1+2k_2+2k_3+k_4\right)\\ k_1=f\left(t_n,y_n\right)\\ k_2=f\left(t_n+\frac{h}{2},y_n+\frac{h}{2}{k}_1\right)\\ k_3=f\left(t_n+\frac{h}{2},y_n+\frac{h}{2}{k}_2\right)\\ k_4=f\left(t_n+h,y_n+h{k}_3\right)\end{array}$$
code implementation

def function(x, y):
    f = y**2 - x / y
    return f

def rk4(x0, y0, t, N):
    n = 1
    while (n != N):
        x1 = x0 + t
        k1 = function(x0, y0)
        k2 = function(x0 + t / 2, y0 + t * k1 / 2)
        k3 = function(x0 + t / 2, y0 + t * k2 / 2)
        k4 = function(x1, y0 + t * k3)
        y1 = y0 + t * (k1 + 2 * k2 + 2 * k3 + k4) / 6
        print('%f, %f' %(x1, y1))
        n = n + 1
        x0 = x1
        y0 = y1
        
rk4(1, 2, 0.01, 6)

#lisp-racket 实现龙格库塔算法计算阻尼单摆
(define rk4-pendulum%
  (class pendulum%
    (inherit-field G/L DT Q FD OmegaD)
    (super-new)
    (define/private (domega/dt t omega theta)
      (- 0 (* (sin theta) G/L) (* Q omega) (* -1 FD (sin (* t OmegaD)))))
    (define/private (domega/dt-with-theta theta)
      (lambda (t omega) (domega/dt t omega theta)))
    (define/public (create-pos theta omega t)
      (let ([new-omega (rk4 omega t DT (domega/dt-with-theta theta))])
        (make-pos
         (+ theta (* DT new-omega))
         new-omega
         (+ t DT))))))

Frequency domain

The analysis in the Frequency domain is a little bit more crude than the Time domain analysis ,we just need to take the Fourier transform or the Laplace transform of both sides of this equation, and we get the solution in the Frequency domain , and then we get the solution in the Time domain by using the inverse transformation.

LT

Definition
$$L[f(t)]={\int }_0^{\infty }f(t)e^{-st}dt=F(s)$$
Inverse Laplace transform
$$L^{-1}[f(t)]=\frac{1}{2\pi j}{\int }_{\sigma -j\infty }^{\sigma +j\omega }F(s)e^{st}dt=f(t)$$
The basic law
1.linear law
$$\begin{array}{l}L\left[a{f}_1(t)+b{f}_2(t)\right]\\ =aL\left[f_1(t)\right]+bL\left[f_2(t)\right]\\ =a{F}_1(s)+b{F}_2(s)\end{array}$$
2.differentiation law
$$\begin{array}{l}L\left[\frac{df(t)}{dt}\right]=sF(s)-f(0)\\ \\ L\left[\frac{d^{2}f(t)}{d{t}^2}\right]=s^{2}F(s)-sf(0)-f^{\prime }(0)\\ \\ ...\\ \\ L\left[\frac{d^{n}f(t)}{d{t}^n}\right]=s^{n}F(s)-s^{n-1}f(0)-s^{n-2}{f}^{\prime }(0)-\cdots \cdots f^{(n-1)}(0)\end{array}$$
3.Integral law
$$\begin{array}{c}L\left[{\int }_0^{t}f(t)dt\right]=\frac{1}{s}F(s)+\frac{1}{s}{f}^{(-1)}(0)\\ \\ L\left[{\iiint }_{s}f(t)(dt{)}^2\right]=\frac{1}{s^2}F(s)+\frac{1}{s^2}{f}^{(-1)}(0)+\frac{1}{s}{f}^{(-2)}(0)\\ \\ ...\\ \\ L[\underset{n}{\underbrace{\left[\int \dots \int f(t)(dt{)}^n\right]}}=\frac{1}{s^n}F(s)+\frac{1}{s^n}{f}^{(-1)}(0)+\cdots +\frac{1}{s}{f}^{(-n)}(0)\end{array}$$

4.lag theorem
$$L[f(t-T)1(t-T)]=e^{-Ts}F(s)$$
5.Attenuation theorem
$$L\left[f(t)e^{-at}\right]=F(s+a)$$
6.Boundary theorem
$$\underset{t\to \infty }{\lim }f(t)=\underset{s\to 0}{\lim }sF(s)$$
and
$$\underset{t\to 0}{\lim }f(t)=\underset{s\to \infty }{\lim }sF(s)$$

$F(s)-Table$

$f(t)$	$F(s)$
$\delta (t)$	$1$
$1(t)$	$\frac{1}{s}$
${\delta }_r(t)={\sum }_{n=0}^{\infty }\delta (t-nT)$	$\frac{1}{1-e^{-Ts}}$
$t$	$\frac{1}{s^2}$
$\frac{t^2}{2}$	$\frac{1}{s^3}$
$\frac{t^n}{n!}$	$\frac{1}{s^{n+1}}$
$e^{-at}$	$\frac{1}{s+a}$
$a^{\frac{t}{T}}$	$\frac{1}{s-(1/T)\ln a}$
$t{e}^{-at}$	$\frac{1}{(s+a{)}^2}$
$1-e^{-at}$	$\frac{a}{s(s+a)}$
$e^{-at}-e^{-bt}$	$\frac{b-a}{(s+a)(s+b)}$
$\sin (\omega t)$	$\frac{\omega }{s^2+{\omega }^2}$
$\cos (\omega t)$	$\frac{s}{s^2+{\omega }^2}$
$e^{-at}\sin (\omega t)$	$\frac{\omega }{(s+a{)}^2+{\omega }^2}$
$e^{-at}\cos (\omega t)$	$\frac{s+a}{(s+a{)}^2+{\omega }^2}$

$F(s)⟶f(t)$
$$F(s)=\frac{B(s)}{A(s)}=\frac{b_m{s}^m+b_{m-1}{s}^{m-1}+\cdots +b_{1}s+b_0}{a_n{s}^n+a_{n-1}{s}^{n-1}+\cdots +a_{1}s+a_0}(n>m)$$
$\text{ (1) }A(s)=0$ no multiple root
$$F(s)=\frac{c_1}{s-s_1}+\frac{c_2}{s-s_2}+\cdots +\frac{c_i}{s-s_i}+\cdots +\frac{c_n}{s-s_n}=\sum _{i=1}^n\frac{c_i}{s-s_i}$$
$$\begin{array}{l}{L}^{-1}[F(s)]=f(t)=L^{-1}\left[{\sum }_{i=1}^n\frac{C_i}{s-s_i}\right]={\sum }_{i=1}^n{C}_i{e}^{s_{i}t}\\ \\ C_i={\lim }_{s\to s_i}(s-s_i)\cdot F(s)\end{array}$$
$\text{ (2) }A(s)=0$ with multiple roots
$$F(s)=\frac{B(s)}{{\left(s-s_1\right)}^r\left(s-s_{r+1}\right)\cdots \left(s-s_n\right)}$$

$$\begin{array}{c}{c}_r={\lim }_{s\to s_1}{\left(s-s_1\right)}^{r}F(s)\\ \\ c_{r-1}={\lim }_{s\to s_1}\frac{d}{ds}{\left(s-s_1\right)}^{r}F(s)\\ \\ ......\\ \\ c_{r-j}=\frac{1}{j!}{\lim }_{s\to s_1}\frac{d^{(j)}}{ds}{\left(s-s_1\right)}^{r}F(s)\\ \\ c_1=\frac{1}{(r-1)!}{\lim }_{s\to s_1}\frac{d^{(r-1)}}{d{s}^{(r-1)}}{\left(s-s_1\right)}^{r}F(s)\end{array}$$

FT

Not commonly used