A.V. Petkevich, N.A. Yarmosh AN APPROACH TO RECOGNITION AND REPRESENTATION OF FORM FEATURES IN MECHANICAL DESIGNS

A.V. Petkevich, N.A. Yarmosh AN APPROACH TO RECOGNITION AND REPRESENTATION OF FORM FEATURES IN MECHANICAL DESIGNS

Automatic interpretation of design data stored in CAD models is one of the key elements for the representation of product designs in terms of features - high-level, conceptually significant for specific applications entities. Features provide a natural means to associate domain knowledge with representations of objects and ensure linking and integration of all the activities and applications in a product life cycle. In the field of mechanical engineering a solution of this problem requires an analysis of the geometric data and topological relationships of the object specifying the most objective and stable attributes for the description and classification of parts.

Unlike existing approaches to representation of the geometric information based on manual coding, the offered method of form features identification is founded on feature recognition from a boundary representation (B-rep) of a solid model. Because of a pure B-rep solid model represents geometric entities and topological relationships at a low level only, we propose to transform it into new representation using a feature recognition approach.

Such transformation requires realization of some feature recognition scheme. The defining this scheme is performed according to the system approach and principles of decomposition and synthesis. That is at the feature level the part is viewed as a complex system consisting of the set of volumetric components. Hierarchical relations between components are defined by syntax whereas the meaning of these components for a domain expert or application corresponds their semantics.

We use a graph-based approach to feature recognition. The B-rep of a part is translated into the labeled graph that is defined as a tuple

G = (F_N , L_N , FL, LL, F_T , F_R , L_T , LL_T , , , ),

where

• 
  F_N , L_N are the sets of nodes of type face and of type loop respectively;
  
• 
  FL is the set of edges of G wich connect a node in F_N to a node in L_N ;
  
• 
  LL is the set of edges of G wich connect pairs of nodes in L_N;
  
• 
  F_T, F_R and L_T are the sets of labels of nodes in F_N and L_N respectively;
  
• 
  is the sets of labels of edges in LL and : LL is a labeling for edges in LL;
  
• 
  : L_N  and : F_N  F_T  F_R are labeling for nodes in L_N and F_N respectively;
  

In order to accomplish the decomposition of the object we distinguish three kinds of the syntactical relations between volumetric components - local modification, global modification and connection. These relations are intrinsic for both prismatic and rotational parts and they can be detected by traversal of G. Feature instances and relations between them are represented in the object-oriented form. In addition to classes that are abstractions of features, we use classes representing explicitly dependencies between features with associated semantics and constraints.

Such representation enables natural decomposition and hierarchical structuring of object form knowledge and it may be mapped into conceptual structure of an object-oriented database. Feature extraction is a first step in building a comprehensive layered engineering knowledge base which incorporates various kind of information besides geometry within a single framework. To achieve tight integration between design and other downstream activities it is necessary to take into consideration the compatibility of the form feature information model with standards for the representation and exchange of product model data.

P.I. Sosnin, E.P. Sosnina MONITORING OF PRACTICAL REASONING IN DECISION-MAKING

This paper considers the logical activity and logic outgoing from the understanding of reasonings as a control facilities of any work (activity). The approach focuses on the human-computer activity and is aimed at the development of computer toolkit supporting the human reasonings in decision-making processes. Our interpretation specifies that the reasonings should carry question-answer (dialogue) character.

It was developed and researched the computer presentation of the controlling function of reasoning (monitoring of reasoning) used in real-time decision-making and in cases of infinities. The bases of the research are:

• 
  the idea of specialized question-answer protocolling of the current activity of the decision-maker;
  
• 
  the search and the development of effective methods and means for the processing, systematization and analysis of structure and content of these question-answer protocols (QA-protocols) currently registered and used for monitoring.
  

Our researches led us to the presentation of QA-protocols as event net graphs. We consider that each node of the graph codes a question or answer as the event (in a database structure) of a current reasoning. The graph is a dynamic formation developing during the decision-making.

Our group has developed and researched the additional useful and important transformations of the graph structure. The basic graph structure added by the formal grammar of names, by the applied logics (causal, temporal, interrogative) and by its PETRI-net and PERT-net versions carries out the function of mathematical model of QA-process in the decision-making.

Theoretical and Practical rezults

QA-protocoling and its control including into the decision-making process were investigated and the theory Th of QA-process was built.

The theory Th consists of two independent subtheories Th1 and Th2. Th1 is the theory of detection, identification and coding of question and Th2 is the theory of QA-control of the decision-making processes.

Object of researches in the theory Th1 is the question Q as constructive formation generated during the comparison of the current situation with the reference experience of the last decisions. The theory Th2 we present through a program product providing controlling functions of dialogue in the decision-making processes.

The description of the modeling complex was used as a specification for the development of the Monitoring Processor of dialogue structures WIQA (Working In Questions and Answers). WIQA is used for operative construction and development of the current model of the decision-making process with the recording of the current condition of a task and for operative control. An environment of its realization is Windows NT, C++. The development of the system is completed and applicable.

Methods and means offered, investigated and realized by the authors open some new posibilities for the control of the decision-making processes such as: the effective monitoring of the process; analysis of an opportunity of the parallel coordination of the subworks (with the purpose of distribution of work between the executives in the typical design decisions); demonstration (on suitable speed) of the development of events on the certain interval of the decision-making; demonstration of the current condition of decision-making at the certain moment of time; training the typical decisions (the samples of different decision-makings) and development of skills in the realization of these typical decisions; personification of the events for the subsequent definition of authorship and contribution of the members of collective to an acceptance of the decision.