Date of Award
Doctor of Philosophy in Industrial and Systems Engineering
Mechanical, Industrial and Systems Engineering
Today's consumers' demands are continuously increasing, sometimes surpassing commercial and operational advancements. Both individual and business customers expect to get goods faster, more flexibly, and at low or no delivery cost. These high expectations along with more customized manufacturing are burdening the supply chain industry. Many report analyses underscore the vulnerability of current supply chain systems, particularly in their agility and reaction times, which was so apparent during the COVID-19 pandemic, and the direct succession of the Russo-Ukrainian conflict. These analyses emphasize the necessity of automation with swift and accurate planning. Judging from the aforementioned facts, this dissertation follows two directions to improving the supply chain industry, namely, the logistic planning and the manufacturing process. Obtaining high-speed planning algorithms is necessary for nowadays' logistic systems to meet the increasing consumer demands, and quickly respond to unpredicted circumstances, which directly improves the trade flow and, in turn, reflects on the economy. Whereas achieving technological migration inside manufacturing facilities from a functional operation to a comprehensively linked data and network model, will help optimize the manufacturing process in terms of speed, quality, consumption of raw material, and energy perspectives.
With regard to improving logistic planning methodologies, this study proposes a novel approach to advance a class of path planning algorithms called the vehicle routing problem (VRP). The proposed approach utilizes the unique surging computing hardware technology and introduces a genetic algorithm (GA)-based approach running entirely on graphics processing units (GPUs). These processing units are known for their capabilities of performing massive arithmetic and logic computations in parallel at a high execution speed. The study introduces two algorithms, one that runs on a single GPU platform for small and medium size VRP problems. The other algorithm runs on multi-GPU high-performing computers (HPCs) that proved to process large-size problems up to 20,000 customer nodes. The algorithms show improvements in the execution speed compared with their conventional CPU counterparts at a factor reaching 1,700.
The second direction of this study integrates smart sensors, actuators, and controllers into advanced communication protocols where a huge amount of data is exchanged between the devices. The study adopts cutting-edge technologies such as artificial intelligence (AI) and automation in the industrial internet of things (IIoT) framework and introduces a lab-scale smart manufacturing system. This IIoT system involves different manufacturing stations that mimic a real-world manufacturing facility. Despite the merits obtained by IIoT like controllability and automation, the broadly distributed and entangled nature of the IIoT networks makes them vulnerable to serious security issues. Therefore, communications between units are secured by hashing and digital certification techniques. Moreover, a comprehensive systematic literature review on cybersecurity approaches in IIoT systems is developed and almost 500 research works in the field have been reviewed over the last decade. Based on this review, an intelligent anomaly detection approach for a typical IoT system is discussed where five machine learning (ML) algorithms are applied to a benchmark IoT dataset to analyze the behavior of different processes and classify it into normal behavior, and seven different attack types.
Abdelatti, Marwan Fouad, "Empowering Industrial IoT Systems with GPU-Powered Optimization Algorithms and Cybersecurity Tools" (2022). Open Access Dissertations. Paper 1478.