TL;DR
This paper introduces KADBASE, a novel interface designed to seamlessly connect disparate expert systems (ESs) and databases within complex engineering environments. By providing a flexible, knowledge-based intermediary, KADBASE enables these systems to communicate and share information, overcoming the challenges of heterogeneous data models and distributed computing.
What Problem Does This Paper Address?
Modern Computer-Aided Engineering (CAE) environments are increasingly integrating diverse software components, including traditional algorithmic programs and newer knowledge-based expert systems (ESs). While a central database management system (DBMS) often serves as the backbone for algorithmic tools, ESs introduce new challenges. ESs often operate with limited data and lack sophisticated data management capabilities. Furthermore, integrating these ESs with existing, potentially heterogeneous databases (using different data models, implementations, and hardware) is a significant hurdle. Simply forcing all data into a single, unified database is often impractical or undesirable. This paper addresses the need for a flexible interface that can manage communication and data exchange between multiple ESs and multiple, diverse databases in a distributed engineering computing environment.
Methods
KADBASE employs a knowledge-based interface architecture that acts as a central hub. It leverages a frame data model to represent various types of knowledge, including schema descriptions and mapping procedures. The core of KADBASE's approach involves:
- Decoupled Components: Each ES and DBMS is linked only to the KADBASE interface, not directly to each other. This modularity allows for easy addition or removal of components.
- Global Schema: A unified global schema, based on the frame data model, integrates the data spaces of individual components.
- Syntactic and Semantic Mapping: Requests from ESs are translated through a two-step process:
- Syntactic Mapping: Translates requests between the component's native data manipulation language (e.g., QUEL, SQL) and KADBASE's internal representation. This is handled by specialized processors.
- Semantic Mapping: Translates the meaning of data structures and operations between different data models. This is a crucial step for handling heterogeneity and is performed by an application-independent knowledge module.
- Layered Schema Description: KADBASE utilizes three levels of schema description:
- Local Schema (LS): Describes the component's native data structure.
- Local Frame-Based Schema (LFBS): Represents the LS in the frame data model, potentially grouping data from multiple LS structures.
- Global Schema (GS): The unified view of all integrated data.
- Mapping Knowledge: Specific knowledge sources (LFBM, LIM, GDSM, GIM) are used to define how to translate between these schema levels.
Datasets
Not discussed in the paper. The paper focuses on the architecture and methodology for integrating existing databases and knowledge-based systems, rather than introducing new datasets for training or evaluation.
Key Insights and Ideas
- Decoupling is Key to Flexibility: By making each ES and DBMS communicate only with the interface, KADBASE achieves remarkable flexibility. Adding a new ES or database requires only defining its interface to KADBASE, not reconfiguring existing connections.
- Knowledge-Based Mapping is Essential for Heterogeneity: The paper highlights that simply translating syntax isn't enough. Semantic mapping, using knowledge about data structures and their relationships, is critical for bridging the gaps between diverse data models and representations in engineering.
- Frame Data Model as a Unifying Representation: The use of a frame data model provides a consistent way to represent diverse engineering data and knowledge, facilitating the integration process.
- Separation of Concerns for Mapping: The clear division of schema description (LS, LFBS, GS) and mapping knowledge (LFBM, LIM, etc.) makes the system more understandable and maintainable.
Key Results
The paper describes KADBASE as a prototype that has been tested with two knowledge-based structural engineering applications:
- SPEX: A structural component design system that used KADBASE to access a database of standard structural steel members.
- HICOST: A knowledge-based cost estimator for building designs that leveraged KADBASE to access multiple databases (building design, project management, unit costs).
The implementation was done using Franz Lisp on a distributed computing environment (DEC VAX 11/750 and MicroVAXs) and utilized the INGRES DBMS. The success of these demonstrations indicates that KADBASE can effectively facilitate the integration of ESs with existing databases in practical engineering scenarios.
Limitations and Future Work
- Manual Schema Integration: The current prototype requires a global database administrator to manually define the global schema and relate it to local views. The paper acknowledges that an intelligent schema integrator was not attempted.
- Prototype Status: KADBASE is presented as a prototype, suggesting that further development and refinement would be necessary for production-level deployment.
- Future Work: The authors imply that the development of an intelligent schema integrator would be a valuable next step. Additionally, generalizing KADBASE to serve all components of an engineering application (both algorithmic and non-algorithmic) is a stated goal, paving the way for large-scale, diverse engineering environments.
Summary updated