Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Mechanical and Civil Engineering

First Advisor

Chiradeep Sen

Second Advisor

Hamidreza Najafi

Third Advisor

Beshoy Morkos

Fourth Advisor

Siddhartha Bhattacharyya


Function modeling plays an essential role in academy design studies, yet a lack of acceptance in industry. A possible reason is that the designer must have well understand of the controlled vocabularies and grammars to utilize this method. The research presented in this dissertation is to fill this gap so that designers can use function-based design method without relevant knowledge. In graph-based function models, the function verbs and flow nouns are usually chosen from predefined vocabularies. The vocabulary class definitions, combined with function modeling grammars defined at various levels of formalism, enable function-based reasoning. However, the text written in plain English for the names of the functions and flows is presently not exploited for formal reasoning. This dissertation presents a formalism (representation and reasoning) to support semantic and physics-based reasoning on the information hidden in the plain-English flow terms, esp. for automatically decomposing black-box function models and to generate multiple design alternatives. First, semantic reasoning infers the changes of flow types, flow attributes, and the direction of those changes between the input and output flows attached to the black-box. Then, a representation of qualitative physics is used to determine the material and energy exchanges between the flows and the function features needed to achieve them. Finally, the topological layer provides reasonings to infer multiple options of composing those function features into topologies and to thus generate multiple alternative decompositions of the functional black-box. The data representation formalizes flow phases, flow attributes, qualitative value scales for the attributes, and qualitative physics laws. A three-layer algorithm manipulates this data for reasoning. The dissertation shows four validation case studies to demonstrate the workings of this formalism.