A Structure Based Configuration Tool: Drive Solution Designer – DSD

AAAI 2002

TL;DR

This paper introduces the Drive Solution Designer (DSD), a novel configuration tool for complex industrial drive systems. DSD leverages a structure-based configuration approach, moving beyond traditional constraint-based methods to efficiently generate consistent and optimized solutions, significantly reducing development time and errors for sales engineers.

What Problem Does This Paper Address?

The industrial sales landscape for complex products like drive systems is increasingly challenging. Sales engineers face a deluge of product evolution, intricate customer-specific solutions, and stringent optimality criteria (price, delivery time). This complexity stems from the vast number of components, their numerous parameters, and the intricate, often non-obvious, physical and technical correlations between them. Traditional configuration methods, like simple constraint satisfaction, struggle with this scale, leading to lengthy offer preparation times, a high incidence of inconsistent or faulty solutions, and difficulty in incorporating new product innovations. The core problem is enabling sales engineers to efficiently and accurately configure highly customized, complex technical solutions on-site with customers.

Methods

The DSD employs a structure-based configuration approach, a departure from traditional strict constraint satisfaction. This method is built upon three key knowledge representations:

  • Ontology: A domain-specific ontology models products, their taxonomic relationships (ISA-specialization), and partonomic relationships (has-parts), along with their attributes. This is represented using a proprietary language, EngConKR, which handles complex data types like discontinuous intervals.
  • Constraints: Dependencies between objects, relations, and attributes are captured using a meta-constraint modeling language. This includes functional/predicate constraints, extensional constraints (tuples of values), and custom Java-Constraints for integrating complex calculations.
  • Strategies: Declarative control knowledge defines the configuration process. This includes the order of configuration steps (decomposition, specialization, parameterization), calculation methods, and priority rules, guiding the user through the process.

The system uses a domain-independent configuration engine called EngCon. EngCon performs a depth-first search, aiming to produce a single, consistent solution rather than exploring the entire solution space. It integrates with external constraint solvers and uses variant tables for managing component options.

Datasets

Not discussed in the paper. The paper focuses on the development and application of the DSD tool for the Lenze AG product catalog, implying the use of their internal product data and specifications as the knowledge base.

Key Insights and Ideas

  • Structure-Based Configuration is Key for Complexity: The paper argues that for highly complex, interconnected domains like drive systems, a structure-based approach is superior to pure constraint satisfaction. This method's depth-first search and guided process are better suited to managing the vast solution space and intricate dependencies.
  • Domain-Independent Engine is Crucial for Reusability: The development of EngCon as a domain-independent engine is a significant insight, allowing for its application to various technical and even non-technical domains by simply swapping out the knowledge base and GUI.
  • Declarative Control for Expert Guidance: Modeling the configuration process using declarative strategies allows for capturing and transferring expert knowledge about the order and methods of parameterization, which is critical for generating high-quality solutions and assisting less experienced engineers.
  • Balancing Automation and User Control: DSD aims to assist, not replace, the sales engineer. It proposes solutions and guides the process but leaves key decisions to the user, ensuring that the final configuration aligns with individual customer needs.
  • The Importance of a Maintainable Knowledge Base: Given the rapid evolution of technical domains, the development of a Web-based tool (K-Build) for easily building and maintaining the knowledge base is essential for the long-term success and relevance of the configuration system.

Key Results

  • Drastic Time Reduction: The time required for sales engineers to prepare an offer has been reduced from an average of 4 hours (ranging from 3 to 30 hours) to an average of 30 minutes. In simple cases, offers can be generated in as little as 15 minutes.
  • Near-Zero Error Rate: The introduction of DSD has reduced incorrect or inconsistent solutions to nearly 0 percent, significantly improving the quality and reliability of offers.
  • Improved Quality and Adoption of New Products: The system facilitates the use of standardized and newly developed components, leading to more cost-effective and higher-performing solutions.
  • Significant Cost Reduction: By promoting the use of standard components, DSD helps reduce production and maintenance costs, providing a competitive advantage.
  • High Adoption Rate: Approximately 150 sales engineers at Lenze AG are actively using the DSD for their daily tasks.

Limitations and Future Work

  • Conflict Explanation: While DSD ensures consistency, providing detailed explanations for conflicts (both taxonomic and constraint-based) to users with appropriate background knowledge remains an open challenge. The authors are investigating plan recognition methods to aid in explanation generation.
  • User Modeling: The paper suggests that automatically learning user models would be beneficial, even though user-specific defaults can be defined.
  • Reconfiguration (CBR): While partial solutions can be reused for variants, the ability to reconfigure complete solutions in the sense of Case-Based Reasoning (CBR) is not yet supported.
  • Constraint Solver Privacy: The specific constraint solver used is subject to privacy, implying potential limitations or dependencies.
  • Future Applications: The paper highlights ongoing development of DSD applications in other technical domains (industrial robots, pumps, sensors) and service domains (financial planning, insurance), indicating the broad applicability of the EngCon engine.

Summary updated

BibTeX
@inproceedings{2002structure,
  title = {{A Structure Based Configuration Tool: Drive Solution Designer – DSD}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2002},
}
Figure