Partial Differential Equation

2011-07-20 9:56 pm

回答 (2)

2011-07-22 5:45 am
✔ 最佳答案
Let v(x,t)=e^t u(x,t), then u_xx= u_t + u convertes to v_xx=v_t
and v_x(0,t)=v_x(4, t)=0, v(x,0)=(8x-1)e^0=8x-1

By the separation variable method:
X" T=XT', X"/X = T'/T and X_x(0)=0, X_x(4)=0, then
X=cos(nπx/2), T= e^(-n^2 π^2 t/ 4)
v(x,t)=∑[n=0~∞] a(n) cos(nπx/2)e^(-n^2 π^2 t/4)
v(x,0)=8x-1 =∑[n=0~∞] a(n) cos(nπx/2)
Taking the fourier half cosine for 8x-1 (for x=0~2 ), then
4a(0)=2∫[0~2] (8x-1) dx , a(0)=7
2a(n)=2∫[0~2] (8x-1) cos(nπx/2) dx, a(n)= 32/(nπ)^2 *[(-1)^n -1]
so
u(x,t)={ 7-∑[n=1~∞] 64cos[(2n-1)πx/2]/[(2n-1)π]^2 } e^(-t) for x=0~2
u(x,t)={23+∑[n=1~∞] 64cos[(2n-1)πx/2]/[(2n-1)π]^2 } e^(-t) for x=2~4

Note1: the last eq.
23+∑[n=1~∞] 64cos[(2n-1)πx/2]/[(2n-1)π]^2 } e^(-t) for x=2~4
can be obtained from
7-∑[n=1~∞] 64cos[(2n-1)πx/2]/[(2n-1)π]^2 } e^(-t) for x=0~2
by symmetry(Since x=2, 8x-1=15)

Note2: Solving by the method of Laplace transform
v_xx=v_t , v(x,0)=8x-1 then V(x,s)=8/(s√s) e^(-x√s) + (8x-1)/s
thus, v(x,t)=(8x-1)- erfc[x/(2√t)]* 8/√(πt)
hence, u(x,t)=v(x,t)*e^(-t)

2011-07-21 21:59:07 補充:
thus, v(x,t)=(8x-1)- erfc[x/(2√t)]* 8/√(πt) ( "*" is convolution)
hence, u(x,t)=v(x,t)*e^(-t) ( "*" is multiplication)

2011-07-23 18:17:34 補充:
謝謝d大對於Laplace inverse of V(x,s)的答案,so,
v(x,t)=(8x-1)+16√t/√π exp(-x^2/4t)-8x erfc[x/(2√t)]

2011-07-23 18:23:04 補充:
To: d大
1. 如果題目是x=0~inf,則會考慮用Fourier transform
為何用Laplace? 因為本題是linear PDE,取Laplace可減化一個變數的導數!
缺點是其inverse 不一定容易求得!
2. 乘以e^t化簡u項,這是習慣問題,本題不化簡亦可!

2011-07-23 18:29:30 補充:
3. 本題通常採用前面的作法,故先將u項消去,求eigenfunction時較方便!

2011-07-23 18:38:59 補充:
4. 您的提問3階ode那題,不知exp(-s^2)的inverse為何?
另一題(積分收斂發散),已約略看過題目,但提問太多,可否只寫一題呢?

2011-07-24 11:49:15 補充:
L{e^(x²/4)}不存在!
2011-07-23 4:06 am
Lotus_本願山彌陀講堂,為甚麼你可以一眼看得出這條PDE能用Laplace transform去解?究竟一條PDE能用Laplace transform去解的先決條件是甚麼?只需要考慮B.C.?還是B.C.和l.C.都需要考慮?

2011-07-22 20:09:52 補充:
還有,為甚麼你要先把這條PDE轉換成Heat equation才開始用Laplace transform去解而不是直接用Laplace transform去解?直接用Laplace transform去解不可以嗎?

2011-07-23 06:09:50 補充:
Lotus_本願山彌陀講堂,其實你用Laplace transform去解的那部分有更簡化的結果,詳見http://eqworld.ipmnet.ru/en/auxiliary/inttrans/LapInv5.pdf#page=02。

所以大家有時都要相信eqworld,不要當eqworld是垃圾!

2011-07-24 10:17:22 補充:
回Lotus_本願山彌陀講堂:
根據http://en.wikipedia.org/wiki/List_of_integrals#Definite_integrals_lacking_closed-form_antiderivatives,e^(-s²)的inverse two-sided Laplace transform應為與e^(x²/4)成正比(此乃regularized value)。

2011-07-25 08:05:49 補充:
雖然我都知道L{e^(x²/4)}和B{e^(x²/4)}本身並不存在,但是如果有一條線性微分方程進行雙邊拉式轉換後得出Y = ce^(-s²),該線性微分方程確實有其中一個線性獨立解為e^(x²/4),因此可以理解成B{e^(x²/4)}的regularized value為與e^(-s²)成正比。

留意發散的積分可進行regularized calculation,求出regularized value。

2011-07-25 08:09:19 補充:
那條3階ODE問題我已續期了(http://tw.knowledge.yahoo.com/question/question?qid=1011072501482),請前往該處討論,始終打擾這條PDE問題不是太好。


收錄日期: 2021-04-22 00:35:11
原文連結 [永久失效]:
https://hk.answers.yahoo.com/question/index?qid=20110720000010KK05067

檢視 Wayback Machine 備份