Network Considerations for a Cloud Manufacturing Marketplace
Open Access
- Author:
- Peck, William
- Graduate Program:
- Industrial Engineering (PHD)
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- September 28, 2023
- Committee Members:
- Steven Landry, Program Head/Chair
Soundar Kumara, Co-Chair & Dissertation Advisor
Saurabh Basu, Major Field Member
Simon Miller, Outside Field Member
Russell Barton, Outside Unit Member
Daniel Finke, Co-Chair & Dissertation Advisor - Keywords:
- Industry 4.0
Cloud Manufacturing
Computer Vision
KAZE Descriptors
Discrete-Event Simulation
Tukey's Honestly Significant Difference - Abstract:
- Industry 4.0 describes advances in information technology that aid in developing modern, flexible, automatic, and innovative manufacturing systems and is an opportunity to change from production-oriented to service-oriented architectures. Cloud manufacturing helps realize this service-oriented goal by creating virtual platforms shared across the internet that allows for consumers to interact with suppliers. Resource providers (suppliers) interact with the platform by virtually encapsulating their machining capabilities and capacities and are matched with service requestors (consumers) whose product contains a number of functional requirements manifested through manufacturing features and other part specifications. The realization of a cloud manufacturing platform involves the networking of many different systems to create a seamless end-to-end architecture that allows service requestors to achieve service-oriented production. However, these networking linkages imply challenges as to the automatic recognition of manufacturing features, the identification of feasible vendors, and the simulation of production capacity. Recent advances in artificial intelligence and data analytics create new opportunities to fulfill consumer requirements in an automatic way across the cloud. The objective of this dissertation is to develop algorithms that automatically recognize manufacturing features, match part characteristics to machining capabilities, and simulate part fabrication to achieve production estimates. Therefore, the cloud manufacturing process inputs a 3D CAD model submitted by the service requestor and outputs cost and cycle-time for every feasible vendor. This research enables and assists in 1) the automatic extraction and detection of different manufacturing features created through subtractive processes, 2) the identification of feasible vendors based on granular machining capabilities that exist on the shop floor, and 3) robust estimation of cycle-time and cost for all feasible vendors. The accomplishments are summarized as follows: • Automatic recognition of manufacturing features using keypoint detection: A computer vision framework using KAZE descriptors able to extract and detect variety of manufacturing features located on a variety of parts that is able to overcomes challenges identified in literature. • Identification of feasible vendors: A matching algorithm that uses the number and type of manufacturing features and part specifications to identify feasible vendors able to fulfill all requirements of the service provider where a granular approach is used to represent machine capabilities present on the shop-floor. • Robust estimates of cost and cycle-time: A discrete-event simulation is automatically created using machine capabilities where the production capacity of feasible vendors is simulated to achieve cost and cycle-time estimates made robust through the inclusion of input uncertainty considerations.