라플라스 변환: 두 판 사이의 차이

잔글편집 요약 없음
잔글편집 요약 없음
8번째 줄: 8번째 줄:
를 '''양방향 라플라스 변환(Bilateral laplace transform)'''이라 한다. 이 문서에서는 단방향 라플라스 변환에 대해서만 다룬다.
를 '''양방향 라플라스 변환(Bilateral laplace transform)'''이라 한다. 이 문서에서는 단방향 라플라스 변환에 대해서만 다룬다.


== 라플라스 변환의 존재성 ==
== 존재성 ==
라플라스 변환이 언제나 존재하는 것은 아니다. 예를 들어 <math>f(t)=e^{t^2}</math>라면
라플라스 변환이 언제나 존재하는 것은 아니다. 예를 들어 <math>f(t)=e^{t^2}</math>라면
: <math>F(s)=\int_0^{\infty}e^{t^2-st}dt=\infty</math>
: <math>F(s)=\int_0^{\infty}e^{t^2-st}dt=\infty</math>
15번째 줄: 15번째 줄:
를 만족하면 ''f''는 지수적 차수(exponential order) α를 가진다고 한다. 만약 ''f''가 <math>[0,\infty)</math>에서 조각적 연속이고 지수적 차수 α를 가진다면, ''f''의 라플라스 변환은 <math>\operatorname{Re}s > \alpha</math>에서 존재하고 [[절대수렴]]한다.
를 만족하면 ''f''는 지수적 차수(exponential order) α를 가진다고 한다. 만약 ''f''가 <math>[0,\infty)</math>에서 조각적 연속이고 지수적 차수 α를 가진다면, ''f''의 라플라스 변환은 <math>\operatorname{Re}s > \alpha</math>에서 존재하고 [[절대수렴]]한다.


== 라플라스 변환의 유일성 ==
== 유일성 ==
어떤 라플라스 변환에 대응되는 함수는 일반적으로 유일하지 않다. 예를 들어, 함수 ''f'', ''g''를 다음과 같이 정의하자.
어떤 라플라스 변환에 대응되는 함수는 일반적으로 유일하지 않다. 예를 들어, 함수 ''f'', ''g''를 다음과 같이 정의하자.
: <math>f(t)=e^{-t}</math>
: <math>f(t)=e^{-t}</math>
138번째 줄: 138번째 줄:
{{ㅊ|이걸 어떻게 다 외워? 살려줘~~}}
{{ㅊ|이걸 어떻게 다 외워? 살려줘~~}}


== 라플라스 변환의 활용 ==
== 활용 ==
=== 미분방정식의 풀이 ===
=== 미분방정식의 풀이 ===
라플라스는 미분방정식을 푸는 데 유용하다. 미분방정식
라플라스는 미분방정식을 푸는 데 유용하다. 미분방정식

2015년 5월 12일 (화) 15:58 판

틀:학술 관련 정보

정의

[math]\displaystyle{ 0\le t\lt \infty }[/math]에서 정의된 함수 f에 대해,

[math]\displaystyle{ F(s)=\mathcal{L}(f(t))=\int_0^{\infty}f(t)e^{-st}dt\quad(s\in \mathbb{C}) }[/math]

f(t)라플라스 변환(Laplace transform), 또는 단방향 라플라스 변환(Unilateral laplace transform)이라 한다. 그러면 "양방향"도 있을 것 같은데, 실제로 쓰고 있다.
함수가 [math]\displaystyle{ -\infty\lt t\lt \infty }[/math]에서 정의되어 있을 때,

[math]\displaystyle{ \mathcal{B}(f(t))=\int_{-\infty}^{\infty}e^{-st}f(t)dt }[/math]

양방향 라플라스 변환(Bilateral laplace transform)이라 한다. 이 문서에서는 단방향 라플라스 변환에 대해서만 다룬다.

존재성

라플라스 변환이 언제나 존재하는 것은 아니다. 예를 들어 [math]\displaystyle{ f(t)=e^{t^2} }[/math]라면

[math]\displaystyle{ F(s)=\int_0^{\infty}e^{t^2-st}dt=\infty }[/math]

이므로 f의 라플라스 변환은 존재하지 않는다. 함수 f에 대해 양수 M, t0와 실수 α가 존재하여 임의의 tt0에 대해

[math]\displaystyle{ |f(t)|\le M e^{\alpha t} }[/math]

를 만족하면 f는 지수적 차수(exponential order) α를 가진다고 한다. 만약 f[math]\displaystyle{ [0,\infty) }[/math]에서 조각적 연속이고 지수적 차수 α를 가진다면, f의 라플라스 변환은 [math]\displaystyle{ \operatorname{Re}s \gt \alpha }[/math]에서 존재하고 절대수렴한다.

유일성

어떤 라플라스 변환에 대응되는 함수는 일반적으로 유일하지 않다. 예를 들어, 함수 f, g를 다음과 같이 정의하자.

[math]\displaystyle{ f(t)=e^{-t} }[/math]
[math]\displaystyle{ g(t)=\begin{cases} e^{-t} & (0\le x \lt 1, 1\lt x\lt \infty)\\ 1 & (x= 1) \end{cases} }[/math]

그러면 [math]\displaystyle{ F(s)=G(s)=\dfrac{1}{s+1} }[/math]이다. 따라서 [math]\displaystyle{ H(s)=\dfrac{1}{s+1} }[/math]인 함수 h는 유일하지 않다. 이 결과는 라플라스 변환을 기껏 정의해놓고서 못 쓰는 건 아닌가 하는 불안을 불러일으킨다. 다행히 역라플라스 변환으로 (조각적) 연속함수를 찾는다면 그 연속함수는 유일하기 대문에 걱정하지 않아도 된다. 다시 말하면, 정의역[math]\displaystyle{ [0,\infty) }[/math]연속함수 f,g에 대해 [math]\displaystyle{ F(s)=G(s) }[/math]이면 f=g이다. 더 나아가, 정의역이 [math]\displaystyle{ [0,\infty) }[/math]인 조각적 연속함수 f,g에 대해 [math]\displaystyle{ F(s)=G(s) }[/math]이면 불연속인 점을 제외했을 때 f=g이다.

역라플라스 변환

[math]\displaystyle{ F(s)=\mathcal{L}(f(t)) }[/math]이면 f(t)F(s)역라플라스 변환(Inverse laplace transformation)이라고 한다. 만약 f[math]\displaystyle{ [0,\infty) }[/math]에서 연속이고 t<0일 때 f(t)=0이며, 지수적 차수 α를 가지고 f' [math]\displaystyle{ [0,\infty) }[/math]에서 조각적 연속이면, 역라플라스 변환은 다음 공식으로 나타낼 수 있다.

[math]\displaystyle{ f(t)=\lim_{y\to\infty}\frac{1}{2\pi i}\int_{x-iy}^{x+iy}e^{ts}F(s)ds }[/math] (푸리에-멜린 반전공식)

그런데 이 공식을 잘 쓰려면 유수 정리 같은 복소함수론의 정리를 잘 이용해야 하므로 복소함수론을 배우지 않았으면 이 공식 대신 라플라스 변환표에서 변환에 맞는 함수를 찾아서 쓴다. 적분을 직접 하고 싶다? 직접 해보자. 어떤 연속함수 f의 라플라스 변환이 다음과 같이 주어졌다고 하자.

[math]\displaystyle{ F(s)=\frac{2as}{(s^2+a^2)^2} }[/math]
Contourintla.png

그러면 [math]\displaystyle{ F(s) }[/math][math]\displaystyle{ s=\pm ai }[/math]에서 이차극점을 가진다. 이때 그림과 같이 복소평면 위에 점 (x,0)을 지나고 허수축과 평행인 직선 ED와, 원점을 중심으로 하고 경로가 DGFKE인 부분원 [math]\displaystyle{ C_R }[/math]을 생각하자. 이때 [math]\displaystyle{ R^2=x^2+y^2 }[/math]이다. [math]\displaystyle{ R\gt a }[/math]이 되도록 충분한 R을 설정할 수 있고, 이때 극점 [math]\displaystyle{ \pm ai }[/math]는 닫힌 경로 EDGFKE의 내부에 속한다. 이때 이 경로를 [math]\displaystyle{ \Gamma }[/math]라고 하자. 그러면

[math]\displaystyle{ \int_{\Gamma}e^{ts}F(s)ds=\int_{ED}e^{ts}F(s)ds+\int_{C_R}e^{ts}F(s)ds }[/math]

이다. 이때

[math]\displaystyle{ \lim_{R\to\infty}\int_{C_R}e^{ts}F(s)ds=0 }[/math]

이고[1] 유수 정리에 의해

[math]\displaystyle{ \int_{\Gamma}e^{ts}F(s)ds=2\pi i(\operatorname{Res}(ai)+\operatorname{Res}(-ai)) }[/math]

이다. 계산을 열심히 하면

[math]\displaystyle{ \operatorname{Res}(ai)=-\frac{1}{2}ite^{iat},\quad \operatorname{Res}(-ai)=\frac{1}{2}ite^{-iat} }[/math]

이므로

[math]\displaystyle{ \begin{align}\int_{\Gamma}e^{ts}F(s)ds&=2\pi i\left(-\frac{1}{2}ite^{iat}+\frac{1}{2}ite^{-iat} \right)\\ &=2\pi i t\left(\frac{e^{iat}-e^{-iat}}{2i}\right)\\ &=2\pi i t\sin at \end{align} }[/math]

이다. 이 식을 푸리에-멜린 반전공식에 대입하면

[math]\displaystyle{ f(t)=t\sin at }[/math]

를 얻는다. 그런데 이거 라플라스 변환표에 기본적으로 나와있는 것이므로, 꼭 필요한 때가 아니라면 이 문서처럼 쓸데없이 길게 적분하지 말고 변환표 외우자.

성질

  • 선형성
[math]\displaystyle{ \mathcal{L}(c_1 f_1(t)+c_2f_2(t))=c_1\mathcal{L}(f_1(t))+c_2\mathcal{L}(f_2(t)) }[/math]
  • (고계)도함수
[math]\displaystyle{ \mathcal{L}(f'(t))=s\mathcal{L}(f(t))-f(0) }[/math]
[math]\displaystyle{ \mathcal{L}(f''(t))=s^2\mathcal{L}(f(t))-sf(0)-f'(0) }[/math]
[math]\displaystyle{ \mathcal{L}(f^{(n)}(t))=s^n\mathcal{L}(f(t))-s^{n-1}f(0)-\cdots-f^{(n-1)}(0) }[/math]
  • tn과 함수의 곱
[math]\displaystyle{ \mathcal{L}(t^n f(t))=(-1)^n F^{(n)}(s) }[/math]
  • 적분으로 정의된 함수
[math]\displaystyle{ \mathcal{L}\left(\int_0^t f(u)du\right)=\frac{F(s)}{s} }[/math]
  • 주기함수
[math]\displaystyle{ f(t+T)=f(t) }[/math]일 때,
[math]\displaystyle{ \mathcal{L}(f(t))=\dfrac{\int_0^T e^{-st}f(t)dt}{1-e^{sT}} }[/math]
  • 합성곱
[math]\displaystyle{ (f * g)(t)=\int_0^{\tau} f(\tau)g(t - \tau)d\tau }[/math]일 때,
[math]\displaystyle{ \mathcal{L}((f * g)(t))=F(s)G(s) }[/math]

라플라스 변환표

공업수학에서 필수적으로 이 관련 표가 나온다. 적어도 이 표는 죄다 외워주는게 시험보기 편하다.

이걸 어떻게 다 외워? 살려줘~~

[math]\displaystyle{ f(t) }[/math] [math]\displaystyle{ \mathcal{L}(f) }[/math] 수렴범위
1 [math]\displaystyle{ \frac{1}{s} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ t^p }[/math] (단, p>-1) [math]\displaystyle{ \frac{\Gamma(p+1)}{s^{p+1}} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ \sin(at) }[/math] [math]\displaystyle{ \frac{a}{s^2+a^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ \cos(at) }[/math] [math]\displaystyle{ \frac{s}{s^2+a^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ t\sin(at) }[/math] [math]\displaystyle{ \frac{2as}{(s^2+a^2)^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ t\cos(at) }[/math] [math]\displaystyle{ \frac{s^2-a^2}{(s^2+a^2)^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt 0 }[/math]
[math]\displaystyle{ \sinh(at) }[/math] [math]\displaystyle{ \frac{a}{s^2-a^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt |a| }[/math]
[math]\displaystyle{ \cosh(at) }[/math] [math]\displaystyle{ \frac{s}{s^2-a^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt |a| }[/math]
[math]\displaystyle{ e^{at} }[/math] [math]\displaystyle{ \frac{1}{s-a} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt a }[/math]
[math]\displaystyle{ e^{at}\sin(bt) }[/math] [math]\displaystyle{ \frac{b}{(s-a)^2+b^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt a }[/math]
[math]\displaystyle{ e^{at}\cos(bt) }[/math] [math]\displaystyle{ \frac{s-a}{(s-a)^2+b^2} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt a }[/math]
[math]\displaystyle{ t^n e^{at} }[/math] (단, n은 자연수) [math]\displaystyle{ \frac{n!}{(s-a)^{n+1}} }[/math] [math]\displaystyle{ \operatorname{Re} s\gt a }[/math]
[math]\displaystyle{ H_c(t) }[/math] (단, c>0) [math]\displaystyle{ \frac{e^{-cs}}{s} }[/math]
[math]\displaystyle{ \delta_c(t) }[/math] (단, c>0) [math]\displaystyle{ e^{-cs} }[/math]
참고사항.
  1. [math]\displaystyle{ \Gamma(t) }[/math]감마함수.
  2. [math]\displaystyle{ H_c(t) }[/math]헤비사이드 함수.
  3. [math]\displaystyle{ \delta_c(t) }[/math]디랙 델타 함수.

이걸 어떻게 다 외워? 살려줘~~

활용

미분방정식의 풀이

라플라스는 미분방정식을 푸는 데 유용하다. 미분방정식

[math]\displaystyle{ \frac{d^2x}{dt^2}+2\beta \frac{dx}{dt}+w_0^2 x = 0 }[/math][2]

이 주어졌다고 하자. 이때 양변의 라플라스 변환은

[math]\displaystyle{ (s^2F(s)-sf(0)-f'(0))+2\beta(sF(s)-f(0))+w_0^2F(s)=0 }[/math]

이고, 식을 F(s)에 대해 나타내면

[math]\displaystyle{ F(s)=\dfrac{(s+2\beta)f(0)+f'(0)}{s^2+2\beta s +w_0^2}=\frac{(s+\beta)f(0)}{s^2+2\beta s +w_0^2}+\frac{\beta f(0)+f'(0)}{s^2+2\beta s +w_0^2} }[/math]

이다. [math]\displaystyle{ F_1(s),F_2(s) }[/math]를 다음과 같이 정의하자.

[math]\displaystyle{ F_1(s)=\frac{(s+\beta)f(0)}{s^2+2\beta s +w_0^2} }[/math]
[math]\displaystyle{ F_2(s)=\frac{\beta f(0)+f'(0)}{s^2+2\beta s +w_0^2} }[/math]

[math]\displaystyle{ w_0^2\gt \beta^2 }[/math]라고 가정하자. 그러면

[math]\displaystyle{ F_1(s)=f(0)\frac{s+\beta}{(s+\beta)^2+(\sqrt{w_0^2-\beta^2})^2}=f(0)\mathcal{L}(e^{-\beta t}\cos(\sqrt{w_0^2-\beta^2}t)) }[/math]
[math]\displaystyle{ F_2(s)=\frac{\beta f(0)+f'(0)}{\sqrt{w_0^2-\beta^2}}\frac{\sqrt{w_0^2-\beta^2}}{(s+\beta)^2+(\sqrt{w_0^2-\beta^2})^2}=\frac{\beta f(0)+f'(0)}{\sqrt{w_0^2-\beta^2}}\mathcal{L}(e^{-\beta t}\sin(\sqrt{w_0^2-\beta^2}t)) }[/math]

따라서

[math]\displaystyle{ x(t)=f(0)e^{-\beta t}\cos(\sqrt{w_0^2-\beta^2}t)+\frac{\beta f(0)+f'(0)}{\sqrt{w_0^2-\beta^2}}\cdot e^{-\beta t}\sin(\sqrt{w_0^2-\beta^2}t) }[/math]

를 얻는다.어때요, 정말 쉽죠?

이상적분의 계산

라플라스 변환을 이용하면 부정적분초등함수로 나타나지 않는 함수의 이상적분을 계산할 수 있다. 이상적분

[math]\displaystyle{ \int_0^{\infty}\frac{e^{-x}-e^{-2x}}{x}dx }[/math]

를 계산해보자. [math]\displaystyle{ f(x)=\dfrac{e^{-x}-e^{-2x}}{x} }[/math]라 하면 [math]\displaystyle{ xf(x)=e^{-x}-e^{-2x} }[/math]이다. f의 라플라스 변환을 F(s)라 하면

[math]\displaystyle{ - \frac{dF(s)}{ds}=\frac{1}{s+1}-\frac{1}{s+2} }[/math]

이다. 정리하면

[math]\displaystyle{ F(s)=\ln(s+2)-\ln(s+1) }[/math]

을 얻는다. 따라서

[math]\displaystyle{ \ln(s+2)-\ln(s+1)=F(s)=\int_0^{\infty}e^{-sx}\frac{e^{-x}-e^{-2x}}{x}dx }[/math]

이며, [math]\displaystyle{ s=0 }[/math]을 대입하면

[math]\displaystyle{ \int_0^{\infty}\frac{e^{-x}-e^{-2x}}{x}dx=\ln 2 }[/math]

를 얻는다.

점화식의 일반항 계산

라플라스 변환을 이용해 점화식으로 주어진 수열의 일반항을 구할 수 있다.

[math]\displaystyle{ \mathcal{L}(a^{\lfloor t \rfloor})=\frac{1-e^{-s}}{s(1-ae^{-s})} }[/math]

[3]을 이용하여 점화식

[math]\displaystyle{ a_{n+2}-a_{n+1}-a_n=0,\;a_0=0,\;a_1=1 }[/math]

으로 주어진 피보나치 수열의 일반항을 구해보자. 함수 f

[math]\displaystyle{ f(t)=a_n\quad(n\le t\lt n+1) }[/math]

로 정의하자. 그러면 주어진 점화식은

[math]\displaystyle{ f(t+2)-f(t+1)-f(t)=0 }[/math]

으로 나타낼 수 있다. 이때,

[math]\displaystyle{ \begin{align} \mathcal{L}(f(t+2))&=\int_0^{\infty}e^{-st}f(t+2)dt\\ &=\int_2^{\infty}e^{-s(u-2)}f(u)du\\ &=\int_0^{\infty}e^{-s(u-2)}f(u)du-\int_0^2e^{-s(u-2)}f(u)du\\ &=e^{2s}\mathcal{L}(f(t))-e^{2s}\left(\frac{e^{-s}-e^{-2s}}{s}\right)\\ &=e^{2s}\mathcal{L}(f(t))-\frac{e^s-1}{s} \end{align} }[/math]

이고

[math]\displaystyle{ \mathcal{L}(f(t+1))=e^s \mathcal{L}(f(t)) }[/math]

이다. 그러므로

[math]\displaystyle{ \begin{align} \mathcal{L}(f(t))&=\frac{e^s-1}{s(e^{2s}-e^s-1)}\\ &=\frac{e^s-1}{\sqrt{5}s}\left(\frac{1}{e^s-\frac{1+\sqrt{5}}{2}}-\frac{1}{e^s-\frac{1-\sqrt{5}}{2}}\right)\\ &=\frac{1}{\sqrt{5}}\left(\frac{1-e^{-s}}{s(1-\frac{1+\sqrt{5}}{2}e^{-s})}-\frac{1-e^{-s}}{s(1-\frac{1-\sqrt{5}}{2}e^{-s})}\right)\\ &=\frac{1}{\sqrt{5}}\left(\mathcal{L}\left(\left(\frac{1+\sqrt{5}}{2}\right)^{\lfloor t \rfloor}\right)-\mathcal{L}\left(\left(\frac{1-\sqrt{5}}{2}\right)^{\lfloor t \rfloor}\right)\right) \end{align} }[/math]

이므로 [math]\displaystyle{ \bigcup_{m\in\mathbb{N}}(m,m+1) }[/math]에서[4]

[math]\displaystyle{ f(t)=\frac{1}{\sqrt{5}}\left(\left(\frac{1+\sqrt{5}}{2}\right)^{\lfloor t \rfloor}-\left(\frac{1-\sqrt{5}}{2}\right)^{\lfloor t \rfloor}\right) }[/math]

이고, [math]\displaystyle{ t=n+\frac{1}{2} }[/math]을 대입하면

[math]\displaystyle{ a_n=\frac{1}{\sqrt{5}}\left(\left(\frac{1+\sqrt{5}}{2}\right)^n-\left(\frac{1-\sqrt{5}}{2}\right)^n\right) }[/math]

을 얻는다.나는 괄호가 싫어요

같이 보기

각주

  1. Joel L. Schiff (1999). The Laplace Transform: Theory and Applications. Springer. pp.154-155. ISBN 0387986987
  2. 조화 단진자의 감쇠진동을 나타나는 운동방정식이다. Stephen T. Thornton · Jerry B. Marion (2011). 강석태 옮김. 『일반역학』(제5판). Cengage Learning. p.118. ISBN 9788962183009
  3. Joel L. Schiff (1999). The Laplace Transform: Theory and Applications. Springer. pp.109-110. ISBN 0387986987
  4. [math]\displaystyle{ 0\in \mathbb{N} }[/math]으로 간주한다.