Date of Award

12-2018

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical and Civil Engineering

First Advisor

Beshoy Morkos

Second Advisor

Chiradeep Sen

Third Advisor

Ju Zhang

Fourth Advisor

Aldo Fabregas

Abstract

Requirements play critical role in the design process as objective statements of stakeholders’ expectations. Design process is iterative, and requirements are also constantly changed and updated to reflect stakeholders’ expectations, design changes, regulations, resource limitations, etc. Managing requirement change is one of the most important requirement management tasks. Since requirements are driving factors in product development from initial concept development stage to final production, mismanaged requirement changes can adversely affect project health leading to monetary and time losses. The ability to assess a requirement change and predict its propagation early in the design process will enable engineers to make informed decisions regarding change implementation. To mitigate issues arising from requirement change propagation, prior research performed by Morkos culminated in the Automated Requirement Change Propagation Prediction (ARCPP) tool. This tool employs a retrospective change propagation method in which engineering changes (ECs) are mapped to respective requirement changes. The syntactic natural language data, part of speech (POS) elements, extracted as requirement relators from requirement statements are utilized to create requirement relationships to form a requirement network from which change propagation due to an initial change is predicted. The tool was later refined to identify that noun POS relators representing physical domain are more useful in predicting change propagation and to recommend the number of noun POS relators to extract from each requirement statement. Based on this premise that change propagation can be predicted in the requirement domain, this research aims to assess requirement change propagation based on requirement change volatility. Requirement change volatility is determined using four volatility classes a requirement may belong to: (1) multiplier, (2) absorber, (3) transmitter, and (4) robust and their respective metric values, multiplicity, absorbance, transmittance, and robustness indicating the level of belongingness to the volatility classes. The volatility class metric values of each requirement are determined from requirement relationships of effective requirement networks produced by the refined ARCPP tool and requirement change data of industrial case studies. Complex network metrics of each requirement are calculated. Computational methods, specifically, regression analysis (RA), artificial neural networks (ANNs), and multilabel learning (MLL) methods are performed to determine if complex network metrics can be employed to determine requirement change volatility class metrics. The results and conclusions are presented along with the recommendations for the future direction of this research.

Comments

Copyright held by author

Share

COinS