[OpendTect_Users] Well logs and neural networks.

Thu Jul 28 14:30:44 CEST 2011

Hello to all,

Thanks for the replies and the suggestions. I'll try to find an auxiliary software to perform the log-to-log NN as I'll be waiting for this new funcionallity in the next development release of OpendTect. Nevertheless, dear Huck, is there a plan to develop the direct log-to-log NN in OpendTect?

Best regards,

Alex.

Prof. Alex Francisco Antunes 
Geologia Estrutural e Interpretação Sísmica 
Laboratório de Geologia e Geofísica de Petróleo 
Depto. Geologia - Prog. Pós-Grad. Geodinâmica e Geofísica / UFRN

---------- Original Message -----------
From: Arnaud Huck <arnaud.huck at dgbes.com> 
To: Alex Antunes <alex at geologia.ufrn.br> 
Cc: "users at opendtect.org" <users at opendtect.org> 
Sent: Thu, 28 Jul 2011 09:25:48 +0200 
Subject: Re: [OpendTect_Users] Well logs and neural networks.

> Dear OpendTect users,
> 
> The "Create" button of the well manager allows to create anycomposite log based on other logs and/or constant values, for freesince this module in part of OpendTect. One may also use logicalstatements since the syntax is fully similar to the mathematicsattribute, like "litho == 1 ? equation1 : litho == 2 ? equation2 :undef". This formula suggests you have a lithology log, and willapply one equation for the lithology 1 and another for the lithology2. It will return undef in other parts of the well (not covered bythe lithology log or for other lithologies).
> 
> New: In the current development release a new function has beenadded in the right-click menu of the wells. This module creates 3Dvolumes in a user-defined radius along a deviated well track, andoutputs a chosen log as amplitudes in this 3D volume. It is part ofOpendTect so free. The volumes from multiple wells can be merged asusual in the Seismic manager using the option "Merge cube parts intoone cube".
> This extra step creates the missing first part for performinglog-log neural network predictions using the OpendTect neuralnetwork plugin. All well logs used for the training (wells with alllogs) and the application (well with missing target) should beconverted to one merged 3D volume per log type in OpendTect, priorto being used as input in the neural network OpendTect plugin. Thequantity to predict remains a log on disk (does not need to be a 3Dvolume). As it is already the case you can apply the neural networkdirectly to a new composite log, using the "Create attribute log"option in the right-click menu of the well, for multiple wells inone go.
> This development version is expected to be released in the incomingweek.
> 
> Best regards,
> Arnaud Huck.
>-- Senior 
Geoscientist
-- dGB Earth 
Sciences
-- Nijverheidstraat 11-2, 7511 JM Enschede, The 
Netherlands
-- mailto: arnaud.huck at dgbes.com, http://www.dgbes.com
-- Tel: +31 53 4315155 , Fax: +31 53 
4315104

> On 07/27/2011 05:33 PM, Christopher P. Ross wrote:Hi Alex,
> 
> As Andres has said, the way to do this is with multiple wells. I am not sure if OpendTect will do this, but I do this all thetime with other software when I am missing key logs or logintervals and I have sufficient wells with sufficient logsacquired.
> 
> Chris
> 
> On 7/27/2011 9:42 AM, Gordon, Andres Cesar wrote:Hi Alex, I did in the past and work very fine for me but 
you need several wells in the area and several 
logs.
You need to have a neural network software and you need to train the network 
using several logs in wells that you have sonic (for example, SP, GR, DENSITY, 
NPHI, PEF, SONIC, RESISTIVITIES) and all the existing logs that are present in 
all the wells. After training the network, you can apply the found relationship 
to the unknown well. If you have many wells to train and several logs in those 
wells this methodology works 
fine
Regards
Andres

-----Original 
Message-----
From: users-bounces at opendtect.org [mailto:users-bounces at opendtect.org] 
On Behalf Of Alex 
Antunes
Sent: quarta-feira, 27 de julho de 2011 
11:14
To: users at opendtect.org
Subject: [OpendTect_Users] Well logs and neural 
networks.

Hello 
everybody,

I'm needing to compute a sonic log in a well that doesn't have it, but I don't 
want to use the Gardner's 
approximation.

I think that this is possible to perform by using the neural networks in 
OpendTect, using a sonic log plus a set of other logs from a reference well, but 
I didn't find any tip on it in the OD 
documentation.

Does anyone know how to do 
it?

Regards,

Alex 
Antunes.

Prof. Alex Francisco 
Antunes
Geologia Estrutural e Interpretação Sísmica 
Laboratório de Geologia e Geofísica de Petróleo Depto. 
Geologia - Prog. Pós-Grad. Geodinâmica e Geofísica / 
UFRN

_______________________________________________
Users mailing 
list
Users at opendtect.org
http://lists.opendtect.org/mailman/listinfo/users
******************************************************************
THIS E-MAIL AND ANY MATERIALS TRANSMITTED WITH IT MAY CONTAIN CONFIDENTIAL OR 
PROPRIETARY MATERIAL FOR THE SOLE USE OF THE INTENDED RECIPIENT. ANY REVIEW, 
USE, DISTRIBUTION OR DISCLOSURE BY OTHERS IS STRICTLY PROHIBITED. IF YOU ARE NOT 
THE INTENDED RECIPIENT, OR AUTHORIZED TO RECEIVE THE INFORMATION FROM THE 
RECIPIENT, PLEASE NOTIFY THE SENDER BY REPLY E-MAIL AND DELETE ALL COPIES OF 
THIS 
MESSAGE.
******************************************************************

_______________________________________________
Users mailing 
list
Users at opendtect.org
http://lists.opendtect.org/mailman/listinfo/users

> 
> -- 
>

CrossQI Sig

> 
> Christopher P. Ross
> 830.896.7656
> "Survivors seem to function best, whenperil is at hand." (J. Buffett,2004) 
_______________________________________________
Users mailing 
list
Users at opendtect.org
http://lists.opendtect.org/mailman/listinfo/users

------- End of Original Message -------

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opendtect.org/pipermail/users/attachments/20110728/ce42edee/attachment.html>