AI-ESTATE Demonstration Plan

fancyfantasicAI and Robotics

Nov 7, 2013 (3 years and 7 months ago)


ESTATE Demonstration Plan

The revised AI
ESTATE standard will contain six information models to be used to
support knowledge
and historical diagnostic data exchange. The first phase of
the demonstration will focus on exchanging two of these mo

the Fault Tree Model
and the Bayesian Model.


Each diagnostic approach will be implemented as either a standalone application
or a component in a larger application, hereafter “application.” Each application
will be implemented independently, by two g
roups of one or more individuals,
with no code sharing between the groups, resulting in two implementations of
each approach and associated model.


The applications for the fault tree model
will be developed for creating and
editing diagnostic fault trees a
nd performing diagnosis with those fault trees. For
the latter, a simple user interface will be developed where by test results can be
provided to the application. No actual
hardware tests
will be


The applications for the Bayesian model
will be develo
ped for creating and
editing simple diagnostic Bayesian networks and performing diagnosis with those
Bayesian networks. The same user interface for providing test results from step 1
will be used here. In fact, as a simplifying assumption, we will use the
ordering in the
fault tree to guide the test selection process for the Bayesian


Each application will be capable of reading and writing models in XML format
according to the AI
ESTATE XML Schema. All reading will be done with a
validating par
ser using the AI
ESTATE XML Schema. Additionally, each module
will perform model validation according to the EXPRESS information model.


A set of “measures of effectiveness” will be defined for the model exchange
process. These MOEs must include success/fa
il criteria, indications of work load
(human and computational) for the exchange, and identification of any
deficiencies in the process.


The demonstration process itself will consist of the following steps (or similar)

each of the two applications


e each application to develop a fault tree and a diagnostic Bayesian
network. If possible, an existing fault tree will be acquired and entered
into each application.

If possible, an existing Bayesian network will be
acquired and entered into each applicati
on. Ideally, we would like the
Bayesian network to be of the same unit under test as the fault tree so we
can use the fault tree for test selection.


Demonstrate the ability of the application to use the fault tree
and the
Bayesian network
for diagnosis.


e each application to create AI
ESTATE conformant exchange files for
the fault tree and Bayesian network.


Pass the exchange files from the creating application to the other
application, and have the receiving application import and validate the


onstrate the ability of each application

to use the received fault tree
and diagnostic Bayesian network to perform diagnosis.


Modify each of the received diagnostic models using the editing features
of the application. Then create exchange files, pass them

back to the
original applications, and demonstrate that the original applications can
process the received, modified models.

To illustrate the process, User #1 uses the Fault Tree Application A to create and
save a fault tree as an XML file using the AI
ESTATE XML Schema. User #2 uses Fault
Tree Application B to read the model, enters test data and performs diagnosis with it.
User #2 modifies the model in Application A and saves it. User #1 uses Fault Tree
Application A to read the modified model and perf
orms diagnosis with it by entering test