1 Inconsistency Resolution of Feature Trees within Multi-level Feature Modeling “A new approach for ensuring consistency in Feature Modeling” Erman Şen Department of Computer Engineering Middle East Technical University Ankara, Turkey erman.sen@ceng.metu.edu.tr Seyyit Alper SERT Department of Computer Engineering Middle East Technical University Ankara, Turkey alper.sert@ceng.metu.edu.tr Ali Hikmet Doğru Department of Computer Engineering Middle East Technical University Ankara, Turkey dogru@ceng.metu.edu.tr Abstract—The study is focused on a new approach for ensuring the consistency of feature trees among a multi-level feature model hierarchy. Although a single global feature tree sounds well in theory, practical cases showed that multiple feature trees, each of which standing for a specific feature of any sub-product line among a product family, worked better. Especially, dealing with domains consisting of high complexity product lines makes it inevitable to work within a multi-level framework. Meanwhile, utilizing multi-level feature modeling brings another dilemma for “possible deviations” on to the stage due to changes applied to feature models over a long period. The focus of this paper is to present a new approach for aligning feature models across these possible deviations, hence ensuring consistency among the overall modeling framework. We provide the basics of multi-level feature trees concept first. We then elaborate possible deviations on feature trees together with their conformity issue that may become cumbersome, inflexible and unmanageable over a long time. Finally, we propose a solution for dealing with the inconsistency between referring feature trees and their respective reference feature trees. Index Terms—Multi-level feature model, reference feature tree, referring feature tree, model deviations, consistency check, alignment of feature models. I. INTRODUCTION The aerospace industry is one of the manufacturing environments with high-complexity. Many independent groups and departments collaborate to accomplish their dedicated jobs. However, the integration among thousands of components of an aircraft is vital. Especially in such intricate organizational contexts, feature modeling has great potential: features may serve as a link between management, marketing and development in a single company and as a link between companies to ease communication; and they can provide a central view of variability in a wide range of development artifacts - such as requirements, design models or test cases - thus becoming the core of all variability and evolution management [4]. On the other hand, involving feature models with such a huge complexity becomes an issue in terms of flexibility and management. An effective approach is already presented to handle the management of large feature modeling framework in [3]. With this approach, the aim is to avoid a global large variability modeling by introducing multi-level feature modeling. Nevertheless, the implication of that approach also brings the concept of “deviations” on feature models as it moves on adding derivative features apart from those defined on their respective reference feature models. Hence, the feature models become very much alike from their respective reference models over a long time. Ensuring consistency among the whole modeling structure then needs the alignment of those deviations. We will start in Section I with presenting basic definitions together with some domain information – the aircraft manufacturing. Then, we will mention possible deviations in Section II. Thereafter, we will explore our contributions by defining the “alignment process” to be applied on deviated feature trees. Finally, we will mention about some related and future work. II. BACKGROUND In this section we will briefly present the basis for feature models and then introduce the “reference feature model” followed by an extension called “referring feature model”. Afterwards, Multi-level feature model will be defined.