Author ORCID Identifier

https://orcid.org/0000-0002-2519-7837

Date Available

11-24-2020

Year of Publication

2021

Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation

College

Engineering

Department/School/Program

Electrical and Computer Engineering

First Advisor

Dr. YuMing Zhang

Abstract

To meet the increasing requirements for production on individualization, efficiency and quality, traditional manufacturing processes are evolving to smart manufacturing with the support from the information technology advancements including cyber-physical systems (CPS), Internet of Things (IoT), big industrial data, and artificial intelligence (AI). The pre-requirement for integrating with these advanced information technologies is to digitalize manufacturing processes such that they can be analyzed, controlled, and interacted with other digitalized components. Digital twin is developed as a general framework to do that by building the digital replicas for the physical entities. This work takes welding manufacturing as the case study to accelerate its transition to intelligent welding by building its digital twin and contributes to digital twin in the following two aspects (1) increasing the information analysis and reasoning ability by integrating deep learning; (2) enhancing the human user operative ability to physical welding manufacturing via digital twins by integrating human-robot interaction (HRI).

Firstly, a digital twin of pulsed gas tungsten arc welding (GTAW-P) is developed by integrating deep learning to offer the strong feature extraction and analysis ability. In such a system, the direct information including weld pool images, arc images, welding current and arc voltage is collected by cameras and arc sensors. The undirect information determining the welding quality, i.e., weld joint top-side bead width (TSBW) and back-side bead width (BSBW), is computed by a traditional image processing method and a deep convolutional neural network (CNN) respectively. Based on that, the weld joint geometrical size is controlled to meet the quality requirement in various welding conditions. In the meantime, this developed digital twin is visualized to offer a graphical user interface (GUI) to human users for their effective and intuitive perception to physical welding processes.

Secondly, in order to enhance the human operative ability to the physical welding processes via digital twins, HRI is integrated taking virtual reality (VR) as the interface which could transmit the information bidirectionally i.e., transmitting the human commends to welding robots and visualizing the digital twin to human users. Six welders, skilled and unskilled, tested this system by completing the same welding job but demonstrate different patterns and resulted welding qualities. To differentiate their skill levels (skilled or unskilled) from their demonstrated operations, a data-driven approach, FFT-PCA-SVM as a combination of fast Fourier transform (FFT), principal component analysis (PCA), and support vector machine (SVM) is developed and demonstrates the 94.44% classification accuracy. The robots can also work as an assistant to help the human welders to complete the welding tasks by recognizing and executing the intended welding operations. This is done by a developed human intention recognition algorithm based on hidden Markov model (HMM) and the welding experiments show that developed robot-assisted welding can help to improve welding quality. To further take the advantages of the robots i.e., movement accuracy and stability, the role of the robot upgrades to be a collaborator from an assistant to complete a subtask independently i.e., torch weaving and automatic seam tracking in weaving GTAW. The other subtask i.e., welding torch moving along the weld seam is completed by the human users who can adjust the travel speed to control the heat input and ensure the good welding quality. By doing that, the advantages of humans (intelligence) and robots (accuracy and stability) are combined together under this human-robot collaboration framework.

The developed digital twin for welding manufacturing helps to promote the next-generation intelligent welding and can be applied in other similar manufacturing processes easily after small modifications including painting, spraying and additive manufacturing.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2021.033

Share

COinS