Back to search

Georgia Institute of Technology

Cognitive Intelligent Decision-Making for Digital Transformation of Reconfigurable Production Systems

Abstract

dc:description.abstract

Cognitive intelligent decision-making for the digital transformation of Reconfigurable Production Systems (RPS) lies at the core of this dissertation. Modern manufacturers must deliver high-variety products under tight lead times while operating hybrid shop floors that mix legacy lines with limited sensing and highly instrumented lines. Classical planning and control architectures, which assume homogeneous resources and static routings, are no longer sufficient to coordinate these heterogeneous, partially digitalized assets and the humans who operate them. This dissertation investigates how digital transformation, production reconfiguration, and decision-making can be integrated into a coherent framework, and develops four main pillars: a graph-based representation for managing product and process variety, a online data-driven method for reconfigurable manufacturing loads balancing and adaptive task allocation, a sociotechnical game-theoretic model for supervision–compliance dynamics on the shop floor, and a blockchain-enabled Information-as-a-Service (IaaS) architecture for trusted information sharing in crowdsourced production environments. These elements are instantiated and evaluated through case studies on an automotive connector family and tank-trailer manufacturing. The dissertation first formalizes a conceptual model of RPS that spans physical production assets and their reconfigurable structure, information flows arising from sensors, digital twins, and legacy systems, and decision responsibilities distributed across strategic, tactical, and operational roles. From a lifecycle perspective, the work distinguishes three stages: transforming customer orders into manufacturable product and process definitions; planning and allocating manufacturing loads across heterogeneous lines; and executing tasks in a human–cyber–physical environment where operator behavior and supervision policies affect outcomes. This analysis highlights fundamental issues in variety management, information services, load balancing, and sociotechnical dynamics. Within this conceptual foundation, the dissertation develops mathematical, computational, and systems models that support: Within this foundation, the dissertation develops models that jointly support: (1) a configuration function that unifies product variety, routing logic, and execution constraints in a single graph-based representation; (2) a planning function that leverages predictive models to perform reconfigurable manufacturing loads balancing and task allocation; (3) a behavioral modeling function that captures supervision–compliance synergy between engineers and operators through evolutionary game-theoretic constructs; and (4) an information service function that employs distributed ledger technology to provide secure, auditable data services for crowdsourced and reconfigurable production. Together these functions define a cognitive decision-making architecture that links configuration knowledge, operational planning, human behavior, and information infrastructure. First, to manage product and process variety, a graph-based Generic Bill of Materials and Processes (GBOMP) formalism is proposed for production configuration management. The formalism generalizes generic BOM and routing ideas into a typed, attributed graph in which generic components, configured product instances, operations, subassemblies, and constraint nodes are connected by incidence, precedence, and variety relations. Variety axes are modeled as attributes with explicit domains and propagation rules, while production schemata define how configuration decisions activate nodes and edges and maintain consistency. On this basis, a configuration algorithm resolves a configured product input into a consistent set of components and operations. The graph-based model is further linked with a knowledge graph constructed from robot programs and accessed via graph-augmented large language models, enabling generative queries and explanations for engineers working with high-variety connector families. Second, to address planning and control in RPS, a method for reconfigurable manufacturing loads balancing and predictive task allocation is developed. The shop floor is modeled as a set of production lines with differing digitalization levels: conventional lines are captured by max-plus models identified from event data, while advanced lines rely on discrete-event simulation and digital twins to provide finish-time predictions. At the upper, leader level, a decision model allocates manufacturing loads and selects line assignments to maximize revenue and capacity utilization; at the lower, follower level, line-specific scheduling and job release plans are optimized to reduce tardiness and operational disruption. This hierarchical problem is formulated as a bilevel optimization model and addressed with a nested metaheuristic that couples genetic algorithms with model predictive control. A connector assembly case demonstrates that the proposed approach exploits differences in line capability and digitalization to improve throughput and due-date performance. Third, to capture human-centric execution dynamics, the dissertation develops a sociotechnical modeling framework and a fuzzy evolutionary game model for supervision–compliance synergy on the shop floor. Interactions between supervisors and operators are viewed as a cooperation–competition relationship in which both sides adapt strategies in response to incentives, monitoring, workload, and perceived fairness. A two-player evolutionary game is formulated at the machine level with payoff structures encoding productivity, quality, and supervision costs, and is then extended to multiple machines to capture population-level dynamics. Stability analysis reveals how different incentive and monitoring regimes lead to compliant, opportunistic, or unstable behavioral patterns. To incorporate uncertainty, payoffs are represented using triangular fuzzy numbers and a fuzzy Moran process simulation is developed to study stochastic evolution of strategy adoption, showing how adjusted supervision and incentive parameters can promote resilient, mutually beneficial patterns. Fourth, to provide a trusted information backbone for reconfigurable and crowdsourced production, the dissertation proposes a blockchain-enabled Information-as-a-Service architecture tailored to manufacturing. A layered design integrates a blockchain network for tamper-resistant transaction logging, smart contracts for enforcing task and data-sharing agreements, and a distributed file system for decentralized storage of manufacturing data such as digital twin records and quality documentation. A prototype instantiated in a tank-trailer manufacturing case demonstrates how trusted, auditable information services can support traceability, transparency, and automated enforcement of agreements, complementing the configuration, planning, and behavioral models and enabling secure, cross-organizational integration of reconfigurable production resources.

Degree

thesis:*
Name thesis:degree_name
Mechanical Engineering, PhD
Department thesis:degree_department
Mechanical Engineering
Grantor
Georgia Institute of Technology
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Song, Mulang
Advisor dc:contributor.advisor
  • Jiao, Jianxin (Roger)
Committee members dc:contributor.committeemember
  • Liang, Steven
  • Ding, Yu
  • Zhang, Fan
  • Choi, Seung-Kyum

Rights

Language dc:language.iso
English

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/81765

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Song, Mulang. Cognitive Intelligent Decision-Making for Digital Transformation of Reconfigurable Production Systems. Georgia Institute of Technology, 2026. https://hdl.handle.net/1853/81765