Fragmentation ammunitions, is a basic type of ammunitions which attack air
and land targets by means of fragments and explosion. Recently, the fragmentation-ammunitions
fast-design system has become a hot area and the performance analysis plays
an important role in the system (Song and Jiang, 2007;
Yuan et al., 2006).
The methodology for the performance evaluations of fragmentation ammunitions
involves coupling of the results from the estimation of internal ballistics,
the external ballistics and the terminal ballistics with empirical formulas
and models of thumb developed over the years. However, there were few existing
standards to select an appropriate performance analysis model that ensures lower
approximation errors. Designers always find it difficult and time-consuming
to get a content analysis model. It is the bottleneck in the ammunition analysis
process. Actually, the exploitation of analysis model is still based on the
engineers experience. In view of the extensive research activities in
the ammunition area, it becomes urgent to establish the performance analysis
expert system which can help the user to evaluate the ammunition easier and
faster. However, the existing analysis systems for fragmentation ammunitions
(Song and Jiang, 2007; Yuan et
al., 2006), which realized the limit of analysis models, didnt
solve the problem but only showed them without any explanations or hint. And
it made the designers more confusing and didnt help the ammunition design.
In this study, the knowledge engineering technology is applied in the performance
analysis expert system for fragmentation ammunitions. The knowledge representation
and inference mechanism fit for the performance analysis models are analyzed
as well. This system co-operates with a fast design platform and enables knowledge
ARCHITECTURE OF THE SYSTEM
The performance analysis expert system supplies calculation supporting for
ammunitions design platform. In the system, the analysis models can be chosen
automatically according to the type and structure of the designed ammunition.
The primary components are in-out module, calculation-inference module and knowledge
In-out module is a package of several interfaces. It receives the data stream
from design platforms and I/O. At the same time, it exports the result in terms
of files and curves.
|| Schematic view of the performance analysis expert system
Calculation-inference module, as the central part of the system, consists of
parameter calculation module, including the configuration parameters and performance
parameters calculation, inference mechanism module and blackboard architecture.
The knowledge module is in charge of knowledge storage and management. The schematic
view of the system is shown in Fig. 1.
IMPLEMENTATION STRATEGY OF THE SYSTEM
Acquisition of knowledge about performance analysis: The knowledge related
to the performance analysis is the basis of the performance analysis system.
Therefore, the knowledge used during the performance analysis process should
be collected, extracted and classified before the system is built. The relevant
knowledge of the system can be grouped under the following headings: parameters,
computational models, computational model rules and procedural knowledge. The
parameters contain configuration parameter, empirical parameter and main impact
The core and chief difficulties of the knowledge acquisition are the acquisition
of application rules to the computational models, including the main impact
parameters and their range. The determination of the main impact factors and
their ranges, which are the selection criterions of the analysis models, is
the result of overall consideration of the derivation of the computational models,
the design elements and the action principle of fragmentation ammunitions. After
that, the application rules of the models are confirmed by means of experiments
and simulations. For example, supposing that the target is thick target, the
main impact factors of the limit velocity for the penetration v50
should be target material, fragment material and fragment shape.
|| Common equations for the limit velocity v50
|| Application rules of equations for the limit velocity v50
The most common equations are shown in Table 1 (Huang
and Zhu, 1998; Zukas, 1990; Wu,
1997; Chen, 1993). To get the application rules
of equations, almost 200 penetration experiments with kinds of situations were
done and other experiment results were found from published reports. Those were
compared with the results calculated by the 4 equations, which were shown in
Table 1. The obtained rules are in Table 2.
The accuracy being considered as acceptable for the knowledge base was limited
Knowledge representation for performance analysis models: As knowledge
representation has strong influence on the computation and the knowledge management
efficiency of the system, it is one of the key content of the study. Hybrid
knowledge representation including production rules knowledge representation,
frame knowledge representation and object-oriented approach is employed in the
system. This kind of representation leads to the analysis models being separated
with the rules, as a result, the knowledge management becomes easier and the
computation modules are not affected.
The calculation models are represented by an object-oriented approach. And
they are described as a class named calculate object. In the class, the calculation
parameters, the class name and the empirical parameters are the data member.
Computation program, inference mechanism and process control program are defined
as the member function. The coarse definition of the class is like this:
Class calculate object:
The application rules are presented by production rules and other kinds of
knowledge are frame. They are stored in the relational database. Production
rule consists of rule name, premise and conclusion. The premise is described
as a premise chain. Each node in the chain is a necessary condition of the inference.
For example, one of the rules for the initial velocity models is described like
this: IF<detonation is point detonation> and <detonation position is
one-end-center> and <fragment pattern is integral> and <L/D>3>
and <end restrain is heavy > and <0<β<2> Then <v 0
= v 01 ()>. The E-R model (Entity-relationship model) of the database is
built around the production rules, which is shown in Fig. 2.
The knowledge base and its data sheets are constructed according to the model.
With relational database, the efficient retrieval algorithms and indexing technology
of relational database are utilized fully. Therefore, the inference efficiency
is enhanced and the knowledge management becomes convenient.
Control strategy of computation and reference flow: The flow control
of the system is a combination of process control and heuristic inference. The
computation terrace, as the primary content of the process knowledge, is summarized
and stored according to the calculation priority of parameters involved in the
performance analysis process. While starting, the system reads the computation
terrace and arranges the computation flow. In this way, the original computation
driftway will not be affected when the knowledge base is extended. This is also
a way to enhance the operative efficiency. The heuristic inference compares
the fact with the rule in base and gets the adaptive computational models.
|| E-R model of rules
|| Flowchart for inference and computation
The forward reasoning policy, which is fully grown, is naturalized. Actually,
the inference mechanism is encapsulated to the function rule in the class calculate
object. The rules and parameters in the base are read and searched with SQL
language. The collision counteraction follows the depth-first search principle.
The computation procedure displays as Fig. 3.
Data stream control: The data drove and control is executed by the blackboard
architecture. The blackboard architecture provides a framework for integrating
knowledge from several sources and serves as a global database. Knowledge and
data produced in each stage are organized separately. And the inference mechanism
consists of the agenda and the monitor of the data in the blackboard (Chau
and Albermani, 2002).
In the system, the blackboard is made up of four work areas: process zone,
inference work area, parameter list and computational models sheet. Process
zone is the storage of process index. Inference work area memorizes the intermediate
data and conclusions in the matching process with the knowledge base. Parameter
list remembers the value of the main impact factor required by premise chain,
the initial parameters related to computational models, the intermediate parameters
and the calculation results. The contents of computational models sheet are
the names of the computational models which ought to be assumed at each evaluation
procedures. The blackboard model in this system virtually integrates the computational
data and the reasoning data and makes the computation and reasoning process
EXAMPLE OF PERFORMANCE ANALYSIS EXPERT SYSTEM
To construct the structure of fragmentation ammunitions performance analysis
expert system, visual studio 2010 and exploitation language C++ are employed.
The knowledge base and the blackboard are based on MYSQL database. The performance
analysis model is memorized in DLL.
In order to demonstrate how the system works, a sample of terminal ballistics
performance analysis is laid out. The main menu screen is shown in Fig.
4a. A number of tabs represent different functions, which are grouped into
two functions: input and result display. The first tab Configuration Parameter
is the input part and others are result display.
While the system started, the information of the ammunitions is transferred
and displayed in this tab automatically. However, some parameters, such as Detonation
Points, Targets and Dynamical Status, should be set up by the engineer. After
that, the Calculate button was enabled and when it was clicked, the calculation
started. The procedures are red, the inference engine worked and the computational
models were selected automatically. The main results were send to Main Results
tab till all performance parameters were calculated, as shown in Fig.
4b. Other tabs are data sheets like initial velocity distribution etc.
||Interface of the terminal ballistics performance analysis
system, (a) Initial parameters input card, (b) Main results card, (c) Datasheet
card and (d) Curve window
the Export button is to export data files and Plot button is to plot curves.
The data sheet card and the curve window are shown in Fig. 4c-d.
Considering the operative habits of the ammunition designer, rule management
takes computational models as the object of operation in the knowledge management
module. The computational models are summarized into three groups: Internal
ballistic, external ballistic and terminal ballistic. Each class can be divided
into lots of subclasses and there are several computational models under them.
The interface is shown in Fig. 5.
The system not only selects the ana Knowledge based performance analysis expert
system for fragmentation ammunitions was developed to combine expert knowledge
with conventional performance analysis. A knowledge model was built for the
analysis models and the key technologies of the system were expounded, such
as knowledge acquisition, knowledge presentation, computation and reference
flow and data stream. The system not only select the analysis models and does
the calculation automatically according to the structure of the ammunitions;
it also stores the analysis model and settles a platform for the management
of the analysis model, which enables the users to manage the analysis model
conveniently and expand the knowledge storage while new analysis model appears
in the future. It is the basis to the development of the fast design system
of fragmentation ammunitions and of great importance to the inheritance, utilization
of the performance analysis knowledge.
This study was supported by the National Defense Basic Scientific Research