This talk was delivered to a panel from the LBL ER-LTT Office on
25 Aug 1995.  At that time, all ER-LTT projects were being "re-ranked"
for the purposes of project RIF due to an anticipated Congressional
recission of FY95 funds.


Advanced Flux Visualization and Virtual Reality for Reservoir Engineering

1. Project Background

In response to the Call for Participation for the Advanced Computational
Technology Initiative (ACTI) put forth by President Clinton in mid-1994,
this project's participants responded with a proposal titled "Advanced
Flux Visualization and Virtual Reality for Reservoir Engineering."  The
philosophy behind the ACTI program is to apply capabilities within the
National Laboraties to problems pertaining to domestic energy production.
In that spirit, the objective of this project is to identify and 
commercialize techniques which facilitate visualization and subsequent
interpretation of "flux" data.  Flux data is a vector field which corresponds
to the movement of fluids, such as oil and water, and gas within a
petroleum reservoir.

Attached is the four-page project proposal which, in a competitive
ranking by a panel of industry experts, was ranked 9th of 122 proposals
from the nine National Laboratories, and was the highest-ranked
proposal from LBNL.  Based upon this favorable ranking, a two-year
funding amount was awarded by DOE.  The plan calls for $205K in
FY95 and $60K in FY96.  Actual funds received in FY95 were $100K.

The technical team working on this project consists of a computer
scientist from LBNL, two reservoir engineers from BP Exploration,
and a software engineer from Western Atlas Software.  

LBNL (Berkeley, CA) brings to the project extensive experience in applying 
techniques to a variety of scientific disciplines, as well as 
demonstrated successes in applying Virtual Reality technology to 
scientific visualization and scientific computing.  Part of our
ongoing mission is to develop and apply advanced visualization 
techniques and user interfaces (VR) to scientific visualization
and computing.  The results of our efforts have been published in
numerous papers as well as demonstrated widely at conferences and
at in-house events. 

The reservoir engineers from BP Exploration (Houston, TX) bring years of 
in interpreting flux (and other) data.  The job of reservoir engineers
is to review various production scenarios, production histories, results
from simulations (the primary source of flux data) as well as geological
and geophysical data and to make recommendations about production
strategies.  Reservoir simulators have proven to be highly valuable to 
this group in asking "what if" types of questions.  Such simulators 
produced vast amounts of data which must be quickly and efficiently
interpreted.  The most effective way of performing this task is through
the use of computer visualization of results, whereby abstract data is
rendered into images which may be rapidly interpreted.  To date,
effective software tools for visualization of flux data do not exist,
thus the motivation for this project.

Western Atlas Softare (Houston, TX) is a software developer providing
software tools for the oil industry.  In particular, this company produces
a reservoir simulator, called "VIP", which is widely used within the
oil industry, as well as a visualization product, called "3DVIEW", which
is used to visualize the results from the simulation.  3DVIEW, prior to
this project, had no capabilities for the display of flux data.  Western's
motivation for participating in this project is to have the capability
to perform effective flux visualization in its product family.

2. Scope of Project

Given that the objective of the project is to make available effective
techniques for flux visualization in the form of a commercial product,
the technical team working on this project operates in a collaborative
way to achieve the objective.  The reservoir engineers who must use
these tools on a daily basis provide much of the design goals for the
tools.  The visualization expertise of both LBNL and Western balances
these design goals against current and anticipated workstation
capabilities.  Prototypes of techniques are rapidly implemented in
the Lab environment, and subsequently subjectively evaluated by the
team for efficacy.  Agreed-upon techniques are then implmented and
integrated into the commercial product by the software company.

The focus of the first year of the project is on flux visualization
techniques, and in deploying them in a commercially available and
supported software product.  The focus of the second year is to dually refine
the user interface to these techniques and to improve upon the
techniques, based upon user feedback from the initially-deployed set
of tools.

In the first year, the visualization techniques are rapidly prototyped
in a visualization environment in the Lab environment.  This environment
permits rapid development of visualization tools, at the expense of
optimal performance.  The visualization modules will be released to
the public at the conclusion of the first year of the project.  The
algorithms of the tools themselves do not constitute patentable 
material, as the notion of iconic representations of data has been
in use for over 30 years.  Techniques for vector field visualization are
drawn primarily from the aerospace industry.  These have been widely 
published over the course of the last decade.  We expect that the
prototype tools developed as part of this project will be useful in
visualizing results of fluid-flow studies currently underway at LBNL and 
at other DOE laboratories.

Based upon the evaluation of the prototype visualization techniques,
the software company will implement them in their own environment then
commercialize the tools through their own product line.  Unlike the 
prototype tools, in which the primary concern is rapid deployment for
the purposes of evaluation, a primary concern of the software company
will be optimizing the algorithm for their particular environment.  The
completion of this phase of the study is timed to coincide with
this company's annual product release cycle.

A paper describing the results of this study will be published in an
appropriate technical journal.  Of interest will be the set of techniques
which we considered, the process used in evaluating them, the acceptance or 
rejecting of each of the tools and the rationale for doing so.

The plans for the second year include two broad tasks.  The first is
to continue the study of applicable visualization techniques for flux
visualization.  We expect useful feedback from the user community that
will guide enhancements of existing techniques as well as the introduction
of new techniques.  Second, we will focus on the user interface to
the visualization process.  Virtual Reality input devices, those that
generate three or six dimensional information directly, will be coupled
with the visualization tools with the intent of increasing the efficacy
of the software.  The results of this effort will be included in the
next annual release of the commercial software.

3. Project Technical Performance/Progress

The first task was to build a prototype which implemented a technique for
flux visualization whereby at each grid point, an oriented cone was placed.
The direction of the cone corresponds to the directional component of the
flux vector, and the size of the cone is proportional to the flux 
magnitude.  This technique results in images which are incomprehensible
due to the sheer volume of geometry present in the image.  This type of
image led us to consider techniques for data reduction.

Data reduction is being approached in two different ways.  One way is to
somehow combine the three (or more) component phases into a few number
of phases.  The industrial partners are working indenpendantly to derive
two different representations: one is to combine all liquid phases into a
single flux value, and the other method is to combine all liquid and
gaseous phases into a single flux value.  Another way to reduce the
amount of data is to spatially "average" flux vectors.  We built a prototype
which will average together, at each (i,j) grid location, flux vectors
from like "simulation layers."  This technique was deemed successful and
is being implemented in the commercial product.

In another direct representation method, it was suggested to "paint"
a two-dimensional icon onto cuberillized grid blocks of saturation, for
example.  The two dimensional icon represents the flux magnitude and
direction in two dimensions.  We are still exploring the applicability
of this technique.

Other methods of visualization involve an abstract representation of data,
whereby a picture is generated which represents something about the data,
rather than directly representing the data itself.  One commonly used
technique in the aerospace industry is the use of "streamlines."  A
streamline is a curve which is tangential to the vector field along its
length.  We built a prototype streamlines module which takes into account
generalizations drawn from typical reservoir grid topology.

Building on the use of streamlines, we prototyped several different 
visualization techniques which use the streamlines computation as a base.
We explored the use of "backwards" streamlines which allow the engineer
to see where flow came "from" rather than where it goes "to."  We explored
the use of depth-colored streamlines, whereby each streamline was color
coded according to its initial depth in the reservoir.  We also explored
the use of "time markers" along the streamline itself.  This technique is
analagous to "dropping a bread crumb" at a user-specified time interval,
and is useful for quickly identifying areas of stagnation and acceleration
in a vector field.  Each of these techniques is slated for inclusion in
the commercial product.

Particular to the streamlines computation is the notion of a "seed point."
From the seed point, the streamlines computation proceeds through space.
The subject of much of the previous work in combining VR and scientific
visualization has been allowing the user to arbitrarily position a seed
point in 3-space using a VR input device.  We created a "virtual well"
tool which permits the user to position a cylinder of seed points at
an arbitrary position within the reservoir.  The initial interface
was keyboard-based; the user selects an (i,j) location in plan view.
A subsequent version permits the user to attach the VR input device
to the virtual well, freely moving the well throughout the reservoir
model.  The virtual well will be included in this year's commercial
product, while the use of the VR input device for positioning will
be released next year.

We have used the World Wide Web as a collaborative tool for sharing
results of the prototypes, as well as email and telephone as a primary
means of communication.  When a new prototype technique is developed
and tested, the results are placed on a Web Server here at LBNL and
subsequently viewed by the partners in Houston and Anchorage.

There has been overwhelmingly positive support from industry for the amount 
and significance of progress made on this project.  They are very pleased
with the results and are excited about the prospects for future

4. Project Administrative Performance/Progress

Each of BP Exploration and Western required LBNL to sign non-disclosure
agreements prior to contributing data and source code, respectively.

There has been a great deal of effort put toward obtaining an "agreement"
between LBNL and the industry partners, however.  We decided that a
CRADA was inappropriate for this project, for a number of reasons.  These
include the fact that no new intellectual property is being generated and
that there is no money coming into the Lab from industry.  Next, we looked
at a so-called "Data Exchange Agreement."  This type of agreement is
essentially a non-disclosure agreement, and seemed more appropriate for
this project than a CRADA.  However, there is a technicality which prohibits
its use - a DEA cannot be used if commercialization is involved, which
there is in this case.  At present, the Technology Transfer Department
at LBNL is drafting a "Memorandum of Understanding" which will outline
the objectives of the project, major milestones and deliverables, the
expectations of each of the participants as well as reference the two
already-signed non-disclosure agreements.

5. Expected Benefits of Project

The tools which we have developed directly benefit a wide variety of
research programs at LBNL.  The Earth Sciences Division, for example, makes
use of many of our software tools in its day-to-day visualization activities.

A suite of tools for deploying VR interfaces to scientific visualization
from the desktop has been developed and disseminated (as experimental
software) to the international user community.  An award for this work,
citing a high level of innovativeness, was received by LBNL in April 1995.

This project brings to the laboratory 1/2 FTE and equipment dollars in
difficult financial times.  This year's ACTI funding has been crucial to
the continued operation of our group.

Using the new tools for flux visualization, working reservoir engineers
are able to devise better strategies for domestic energy production.  This
benefit is one reason that this project was favorable reviewed by the 
industrial review panel last fall.  These tools translate directly into
increased domestic oil production and reduced dependance on foriegn 
energy sources.

This project involves a methodical study of new techniques for flux
visualization by combining the expertise of a diverse but complementary
group of technical disciplines.

The tools are being commercialized by a US company.

Due to the significant amount of technical progress made on this project,
many in industry have the favorable impression that the National Laboratories
are assets to the country.

-------------------- Outline of live demonstration

Direct Flux Vis techniques

Oriented Cones
	(make an image with zillions of cones)

Markers on Boxes
	barbs, triangles, scaling, etc.

	data/depth averaging

	basic streamlines - seed points from slicing
	virtual well streamlines 
		- dummy device
		- VR input device
	variations on streamlines:
		- radial bore of virtual well
		- color coded by depth
		- time markers
		- backwards streamlines

Demonstrate what "product" will look like:
	cuberillized pressure/sat (from recurrent map files?)
	along with vector vis.  add isosurfaces for constant flux