Towards Automatically Generating Constraint-based Product Configuration Knowledge

Product configuration is considered as one of the most successful applications of knowledgebased approaches in the past decade. Knowledge-based configurations can be classified into three different approaches, namely rule-based, model-based, and case-based approaches. Past research has mainly focused on the development of reasoning techniques for mapping requirements to configurations. Despite the success of certain conventional approaches, the acquisition of configuration knowledge is usually done manually. This work explores fundamental issues in product configuration system, and proposes a novel approach based on data mining techniques to automatically discover configuration knowledge in constraint-based configurations. Given a set of product data comprising product requirements specification and configuration information, we adopted an association rule mining algorithm to discover useful patterns between requirement specification and product components, as well as the correlation among product components. These patterns are translated into constraints for configuration reasoning. The significance and effectiveness of the proposed approach is demonstrated by its incorporation in our configuration system prototype. Our configuration system takes XML-based requirement specification as input and produces product configuration as output consisting of a list of selected components and the structure and topology of the product. Three modules are developed, namely product data modelling, configuration knowledge generation, and product configuration generation module. The product data modelling module generates generic product family models where product data may be instantiated. The proposed approach is implemented in the configuration knowledge generation module. The configuration generation module realizes a resolution of constraint satisfaction problem (CSP) to generate the output configuration. A case study is conducted based on product data from a local water heater manufacturer. The results are discussed in detail in the article.