# "PROPER ORTHOGONAL DECOMPOSITION"

 Register Blogs Members List Search Today's Posts Mark Forums Read

 December 5, 2005, 01:56 "PROPER ORTHOGONAL DECOMPOSITION" #1 dpshaka Guest   Posts: n/a DOSE ANY BODY KNOW THE DETAIL OF THE "PROPER ORTHOGONAL DECOMPOSITION", I HAVE REVIEWED A LOT OF PAPERS ,BUT I STILL CAN NOT UNDERSTAND THE DETAIL OF THE NUMERICAL process of "POD". IF SOME ONE IS FAMILIAR ABOUT "POD" , PLEASE HELP ME / REGARDS

 December 5, 2005, 21:49 Re: "PROPER ORTHOGONAL DECOMPOSITION" #2 Peter Attar Guest   Posts: n/a Essentially you collect snapshots of the solution vector at various points in time for represented "forcings". You collect these into a matrix Q and then perform a singular value decomposition on it. You then say that the POD modes are a linear combination of the snapshot vectors(snapshot method of POD) so that the modes can be represented as [R]=[Q][V] where Q is the snapshot matrix and V are the right singular vectors. For a more detailed explanation on this look for papers by Thomas,Dowell et al and Beran et al. Peter

 December 8, 2005, 14:28 Re: "PROPER ORTHOGONAL DECOMPOSITION" #3 richard Guest   Posts: n/a You may want to check the following link: http://dissertations.ub.rug.nl/FILES...azemier/c2.pdf It doesn't seem to be too badly explained by this guy. For further literature, look for the method of Sirovich, which is an approximation of the continuous (aka. classic) POD. Best regards, Richard

 December 11, 2005, 05:01 Re: "PROPER ORTHOGONAL DECOMPOSITION" #4 dpshaka Guest   Posts: n/a thank you for your feedback, but would you please give me some detail information about the numerical procedure of the POD . regards.

 December 13, 2005, 19:35 Re: "PROPER ORTHOGONAL DECOMPOSITION" #5 richard Guest   Posts: n/a Hopefully I got you right and you are looking for an implementation? Below you find a kind of step-by-step example (I am sorry for the looks. "_" and "^" are used for indecies and (f,g) is a scalar product): N vector fileds u_k=u_k(t_k) (provided by e.g. PIV measurements) at different times t_k, k = 1...N. u_k is representet by e.g. 2 matrices: u-component and v-component of velocity u_k, both of the size m x n. Snapshot method by Sirovich: 1) create a vector (one dim. array) for each vector field e.g. U=[u_11,v_11,...u_mn,v_mn], where m and n are the dimensions of the vector field. The vector U has the size of 1 x 2nm. 2) Calculate the 2-point cross-correlation matrix Q: Q_ij = 1/N (U_i,U_j). 3) Solve the eigen value problem "Q q_i = lambda Q" to find the eigen vectors q_i. There are N eigen vectors q_i (i = 1...N) with the size of 1 x 2mn. 4) Calculate the orthonormal base, phi_i, (aka. eigen modes) of the POD: phi^i = SUM[n = 1...N](q_n^i U_n). 5) Reconstitute a vector field in the new orthonormal base: U_new = SUM[i = 1...N]((U_old,phi^i)phi^i). I hope this will help you. Regards

 December 13, 2005, 22:31 Re: "PROPER ORTHOGONAL DECOMPOSITION" #6 dpshaka Guest   Posts: n/a Dear richard: Thank you very much for your feedback.It gives me so much help. I have seached a lot of paper, but the method that they used is different from what you have said. they all solve a eigenvalue problem in a integral equation form. so would you mind give me your email address, i still have some problem about this POD technique. ztdep@163.com best wishes your DP PRC XJTU

 December 17, 2005, 02:38 Re: "PROPER ORTHOGONAL DECOMPOSITION" #7 ztdep Guest   Posts: n/a Dear richard: i have send a email to you ,have you received it ? regards? yours

 Thread Tools Display Modes Linear Mode

 Posting Rules You may not post new threads You may not post replies You may not post attachments You may not edit your posts BB code is On Smilies are On [IMG] code is On HTML code is OffTrackbacks are On Pingbacks are On Refbacks are On Forum Rules

 Similar Threads Thread Thread Starter Forum Replies Last Post Blob Main CFD Forum 0 June 5, 2009 09:32 mike Main CFD Forum 8 June 10, 2006 03:39 Mehmet SEN Main CFD Forum 7 May 19, 2006 12:38 R.D.Prabhu Main CFD Forum 0 March 15, 2000 18:31 Rudolf Pfeiffer Main CFD Forum 6 December 10, 1998 11:20

All times are GMT -4. The time now is 08:02.