¹úÃñ²ÊƱ

You are cordially invited to the Graduate Student Research Seminar of the Department of Industrial Engineering.

Date: Wednesday, March 26, 2025
Time: 1:00pm – 3:40pm AST
In Person:ÌýÌýÌýÌýÌýÌýÌýÌýÌýÌý Room MA310, Sexton Campus

Online:Ìý°Õ±ð²¹³¾²õ:ÌýÌýÌýÌý
(Meeting ID:Ìý226 323 499 964;ÌýPasscode:ÌýPe7Uj3ZE)

Schedule:

Abstracts:

Design of Picker-to-Parts Warehouse Fulfillment Sections Using Surrogate Machine Learning Model
Basava Sri Krishna Vamsy Lanka, MASc. Student

The design of picker-to-parts warehouse sections contains various decision parameters such as warehouse dimensions, routing policy and storage assignment policy. Assessing the holistic importance of each decision parameter cannot be easily measured or quantified due to their mutual interdependence. It is crucial to obtain this information and investigate the possible combinations of policies and warehouse specifications. To solve this problem we use a surrogate machine learning model to simulate the warehouse conditions across varying pick list size. Seasonally varying demand and pick face requirements are also considered. Dataset derived from simulation is fed to train various machine learning algorithm. The model uses Monte Carlo method and average travel distance as the output parameter to evaluate performance. The Random Forest, Decision Tree, Gradient Boost, XG Boost, LightGMB, CatBoost and a tuned Artificial Neural Network show the best performance in terms of error and fit. SHAP feature importance is calculated and analyzed. Warehouse design practitioners and fourth party logistics can easily adapt and deploy the developed warehouse simulation methodology and machine learning model to help with bid design in determining optimal warehouse parameters and policies.

A Network-based Simulation Model for Helicopter Rescue Time Estimation in the Canadian Arctic
Mahroo Mohammadi, MASc. Student

Search-and-rescue (SAR) helicopter operations in the remote Canadian Arctic encounter extreme weather conditions, limited infrastructure, and vast distances. To improve response times, this study develops a network-based simulation model aimed at supporting helicopter rescue missions in harsh environments. The model combines network analysis techniques, high-resolution meteorological data, and simulation-based evaluation to create realistic rescue scenarios and determine efficient routes from SAR bases to potential incident sites. The methodology is divided into four phases. In Phase One, weather data from Spire Global is acquired and processed from GRIB files into a structured format. Key meteorological parameters—temperature, wind speed, visibility, and precipitation—are used to classify flight conditions as favorable, unfavorable, or no-go, establishing mission feasibility under rapidly changing conditions. Phase Two generates possible routes connecting multiple SAR bases to incident locations across the Arctic using a Breadth-First Search algorithm that considers constraints such as route length, fuel capacity, and refueling requirements. In Phase Three, the weather framework is integrated into the candidate route network by dividing each route into segments and assessing meteorological conditions for every segment, capturing the impact of localized weather variability on helicopter performance. Finally, Phase Four applies discrete-event techniques to model complete SAR missions. The simulation covers helicopter dispatch, flight operations, refueling stops, and on-scene rescue procedures, yielding quantitative estimates of rescue times and revealing potential operational bottlenecks. Preliminary findings indicate incorporating local weather data with operational constraints influences route selection and mission duration, providing a tool for SAR coordinators, policymakers, and emergency planners.

A Hybrid Bi-LSTM Model for Data-Driven Maintenance Planning
Alex Noussis, MASc. Student

Modern industries depend on the reliable operation of a multitude of assets under constrained resources. The advent of Industry 4.0 has increased the use of sensors for monitoring asset/system performance, while deep learning models have allowed for accurate system health predictions, enabling more effective maintenance planning. Most existing papers on intelligent maintenance develop deep learning models solely