Event driven approximate reasoning
Krzysztof Psiuk
Silesian University of Technology,
Institute of Fundamentals of Machine Design
Goals of technical diagnostics
Increased safety
Enhanced reliability
Increased quality
Reduced costs
Assessment of technical condition
of devices and machines
2
Why to use expert systems
Object complexity
Large number of input data
Big number of investigated symptoms (output
data)
Data uncertainty and inaccuracy
Changing conditions
3
Elements of expert systems
User interface module,
Knowledge base module
Database module,
Inference module.
4
Expert systems
Static systems
they most frequently work in “off-line” mode
search for solutions in a fixed (frozen) environment
Dynamic systems
most frequently work in the “on-line” mode
search for solutions in the changing environment
realize the reasoning proces in limited time and with
limited resources
5
Diagnostic expert systems
Dynamic expert system based on
Notice boards
Intuitionistic notice boards
Notice board
Place for publishing notices,
Notices include statements
6
Statements
Statement is understood as an utterance stating
observed facts or representing a given opinion
s c, v
where:
c – statement content,
v – statement value.
7
Multilayer statement network
Graphical representation
8
Example of a multilayer statement
network structure
9
Event driven reasoning
Solution for dynamic expert systems
Starting the reasoning process by events
10
Event driven architecture
Process management,
Programming applications,
In managing and control systems.
In reasoning process.
11
Event driven systems
System based on an Event Driven Architecture
(EDA)
Event
An event is an occurrence or happening, which
originates inside or outside a system, and is
significant for, and consumed by, a system’s
component.
Event types.
Primitive events
Composite events
12
An event
e = <h, b>
where:
h – event header
b – event body
13
An event
Event Header
event name,
occurrence time,
occurrence number,
event source,
possible other elements that are included in the event
specification.
Event Body
include details of the detected state change
14
Event driven systems elements
Event producer
a subsystem that generate events,
Event transmitter
receive events communicated by the event
processing engine and react to these events,
Event processing engine
receives events, processes them and initiates actions
Event consumers
receives events from processing engine and reacts to
them
15
Modules of the System
Observer Knowledge Base
Notice board Inference module
Statement 01
Statement 02
⁞ Measurement modules
Statement k
⁞
Statement n
Analyzing modules
Notice board viewer Control modules
16
Knowledge representation
Belief networks
based on the Bayes theorem
using the Bayesian interpretation of the probability
(a measure of degree of belief)
elements as nodes
connections between nodes with directed edges
Intuitionistic statement networks
elements as statements
connections define relations between statements
17
Event producers
Control modules
Controllers placed on the object
Measurement modules
Data acquisition processors
Signal features calculators
Inference module
Determined statement values
18
Event transmitter
OPC Unified Architecture standard
an industrial communication protocol
flexible communication protocol
many industrial applications
19
Event processing engine
Notice board
Implemented as OPC UA Server
Observer
monitor statement values,
generate new statements
run inference process when necessary
20
Event processing engine
Change in statement value will be an event ei
Set of events E
ei E
21
Event processing engine
Event header
event type,
statement name,
statement value timestamp,
source of statement value,
range of statement value where the event is omitted.
Event body:
old value,
new value.
22
Event processing engine
If the difference between old value and new
value is to small (less than range of statement
value defined in event header) the event should
be omitted,
Initialize the process of evaluation of other
statements value, using some equations or other
special prepared procedures,
Initialize the inference process.
23
Event consumers
Inference module
Notice board viewer
24
Conclusions
Event driven architecture is an interesting
solution for the operation of dynamic expert
systems.
It allows better use of available resources
Enables the integration of expert systems with
data acquisition systems.
Changes in the value of conditioned data could
call the appropriate automatic reaction of expert
system
25
Further work
Improve the system for using real-time data
Improve the event processing engine
26
Acknowledgments
Described herein are selected results of study,
supported partly from the budget of Research Task
No. 4 implemented under The National Centre for
Research and Development (NCBiR) in Poland
and ENERGA S.A. strategic program of scientific
research and development entitled Advanced
technologies of generating energy
27
28
Reasoning process
Use event driven reasoning.
Start and stop the reasoning process the event
29
Event producer
Event producer: Event transmitter:
- Object controllers
- Sensors
- Calculators
Notice board:
Event processing engine: Event consumers:
30
Expert systems using event driven
architecture
Inference process could be started:
By users,
On time,
On time repetitively,
Using notice boards,
Reasoning process started by events
31
