pilipiliÂþ»­

You are cordially invited to theÌýGraduate Student Research SeminarÌýof theÌýDepartment of Industrial Engineering

Date: Wednesday, October 13, 2021
Time: 1:00 - 4:30 PM
Venue: Online Event

Ìý

Schedule:

1:00-1:25 PMÌýLuana Almeida
A Greedy Randomized Adaptive Search Procedure (GRASP) for the multivehicle prize collecting arc routing for connectivity problem

1:25-1:50 PMÌýPranitha Vattoni
Critical regions and locations for the road clearing and relief supplies distribution model in Vancouver Island after a Cascadia earthquake

1:50-2:15 PMÌýTanmoy Das
Modeling resource allocation of emergency response in Arctic oil spills

2:15-2:40 PMÌýLinden Smith
Identifying changes in demand for perishable products using statistical process control and machine learning forecasting

2:40-2:50 PMÌýBREAK

2:50-3:15 PMÌýWheejae Kim
Impacts on human-machine team trust, workload and accuracy by altering reliability information display of Automated Aid System Cognitive Shadow

3:15-3:40 PMÌýCeilidh Bray
A Comparative framework in healthcare decision-making of the functional characteristics and the financial costs of stockpiling two respirator types in pandemic settings

3:40-4:05 PMÌýYun Yin
Modelling and solving of the multi-calendar naval surface ship work period

4:05-4:30 PMÌýHyojae Kim
Developing a decomposition matheuristic method to solve the multi-calendar naval surface ship work period problem

Abstracts:

A Greedy Randomized Adaptive Search Procedure (GRASP) for the multivehicle prize collecting arc routing for connectivity problem
Luana Almeida, PhD Candidate

Natural disasters such as earthquakes can severely impact road networks. Depending on the disaster intensity and the size of the affected area, the network may be divided into multiple disconnected parts. In a disaster response context, decision-makers need to determine the roads that should be unblocked to facilitate relief activities such as search and rescue, evacuation, and distribution of emergency supplies. The multi-vehicle prize collecting arc routing for connectivity problem (KPC-ARCP) is a well-known problem dealing with such a scenario. A matheuristic to solve the KPC-ARCP was proposed in previous research, which tested instances with fewer than 400 vertices and 700 edges. However, it is unknown whether the matheuristic can handle larger instances. This article proposes a Greedy Randomized Adaptive Search Procedure (GRASP) metaheuristic with the hypothesis that GRASP is faster and can solve more extensive networks. Two sets of tests are performed on randomly generated instances with increasing size. The gap in the objective function values and the execution times of GRASP versus the matheuristic are compared. The results indicate that GRASP can achieve objective function values as good as the matheuristic and is significantly faster depending on the parameter settings.

Critical regions and locations for the road clearing and relief supplies distribution model in Vancouver Island after a Cascadia earthquake
Pranitha Vattoni, MASc Student

Vancouver Island lies on the Cascadia Subduction Zone, which makes the region extremely vulnerable to large-scale earthquakes and tsunamis that may follow. To improve the preparedness for such an event, the SIREN project research team has developed several models to identify: communities that may need assistance, delays in transportation operations, identification and reconstruction of damaged roads, optimizing the delivery of supplies from the mainland to the islands using ferries, barges, or helicopters, and from the ports to the communities using trucks within a specific time limit. With this work, we aim to run what-if analyses and conduct an extensive sensitivity analysis of the Road Clearing and Relief Supply Distribution model to identify critical roads, communities, and regions on the island. The inputs to the model are classified into four different types and some output parameters such as the number of communities served, the total population served, and the percentage of roads repaired were studied for changes made in the inputs. By changing the location of the road clearing teams’ depot, we have identified the ‘best’ location for the depot, as well as roads, communities, and regions on Vancouver Island that are critical to the model.

Modeling resource allocation of emergency response in Arctic oil spills
Tanmoy Das, PhD Candidate

Accidental oil spills result in significant contamination in the marine environment and postspill response recovery operations are expensive as well as time-consuming. The problem become exacerbating when the spill size is large. Hence, minimizing the consequences of oil spills is a prime concern for decision-makers. However, unified decision support tools that can estimate oil spills, rank, and allocate response resources optimally, and capture uncertainty are still underdeveloped. The overall purpose of the proposed research is to model emergency resource allocation in the form of a decision support tool (DST). This DST can be used to provide reasonable estimates of likely spill volumes quickly and response allocation and uncertainties herein, ultimately contributing to risk management. This DST will be implemented in hypothetical oil spill scenarios in Arctic Canada. This research will deliver a decision-making modeling framework with practical relevance in pollution preparedness and response risk assessment, ultimately broadening the available toolboxes. The modeling outcome includes long-term planning of resource prioritization and allocation options of an oil spill, which is helpful for the Canada Coast Guard, oil spill response organizations e.g., Eastern Canada Response Corporation.

Identifying changes in demand for perishable products using statistical process control and machine learning forecasting
Linden Smith, MASc. Student

Fast and accurate identification of changes in demand is crucial in the management of blood products. Canadian Blood Services (CBS) manages the collection and distribution of blood products in the Ottawa region of Ontario, Canada. CBS is planning a pilot project to apply pathogen reduction technology (PRT) to platelet production. The introduction of PRT is expected to change hospital demand for platelets; however, the form of this change is unknown. A lag time exists between the identification of a supply-demand imbalance and the ability to address it. The objective of this research is to determine how quickly and accurately demand changes can be detected, to help minimize lag time and thus provide better patient health outcomes. A discrete-event simulation was used to model platelet inventory and generate data representative of possible demand shift scenarios. Process control methods were used to detect and quantify shifts in demand. To improve time to detection, forecasted data points were included in the changepoint detection scheme. Several forecasting methods were tested in this role; ultimately a Generalized Additive Model, fit using splines, was selected. It was found that statistical process control methods were effective in detecting demand shifts of any form and that forecasting decreased the time to detection, while slightly raising the false alarm rate. When the magnitude of the demand shift increased, the detection rate increased and the time to detection decreased. The consequences of a hidden demand shift are substantially less for shifts of smaller magnitude, mitigating the risk due to increased detection time. These results will be useful in minimizing the patient impact of PRT. TBU.

Impacts on human-machine team trust, workload and accuracy by altering reliability information display of Automated Aid System Cognitive Shadow
Wheejae Kim, MASc. Student

Human-machine team is used in various fields that require high accuracy and high work intensity. Cognitive Shadow is an automated decision support system that is used in Simulated Combat Control System of naval air defence training program, where the operator aims to identify targets as friendly or hostile. The Cognitive Shadow learns the pattern of decision making by the individual user and alerts the user when the decision deviates from the normal pattern, with an option to display its reliability rating. Ultimately, the user makes the informed decision to accept or reject the suggestion of the Cognitive Shadow. The participant groups were divided into Control (no Cognitive Shadow), No reliability (Cognitive Shadow alerts the user, but does not display the reliability information), Offline reliability (Cognitive Shadow alerts the user, but its reliability rating is relayed verbally by the researcher), and Online reliability (Cognitive Shadow is working and displays the reliability rating every time it is triggered). Our results show a significant reduction in perceived workload in all groups as the experiment progresses. Also, there is a significant increase in accuracy in all groups as the experiment progresses. These two observations are likely due to participants becoming familiar with the research procedures. Our results showed no significant change in trust level, workload or accuracy among the groups with the different reliability display methods. Finding the reliability information display that impacts the performance and accuracy of the humanautomation team will be a point of interest for future studies.

A Comparative framework in healthcare decision-making of the functional characteristics and the financial costs of stockpiling two respirator types in pandemic settings
Ceilidh Bray, MASc. Student

SARS-CoV-2 has posed implications for personal protective equipment supply. In this research we examined if elastomeric facepiece respirators (EFRs) are efficacious substitutes for N95 respirators through comparing their functionality and cost. A review of literature shows that users favour N95 respirators for comfort but prefer EFRs for protection. EFRs are more cost effective when N95s are used as designed (single use) but mixed strategies minimize costs when N95s are reused (as practiced during shortages). SIR and SEIR compartmental models were used to compare respirator stockpiling requirements and costs depending on PPE utilization and disinfection strategies. SARS-CoV-2, the 2009 H1N1 and Spanish Flu scenarios were investigated to identify trends in preferred respirator utilization strategies. Additionally, a case study on Dartmouth General Hospital was examined. Lastly, a multi-criteria decision analysis using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) was conducted to consider the financial and functional criteria together. According to the TOPSIS results, the ideal solution is typically N95s unless cost criteria are given the greatest weight of importance. The research shows that financial and functional findings are predictive, rather than prescriptive, as results vary according to epidemiolocal characteristics of a pandemic and healthcare respiratory protection program needs. Ultimately, this research provides more sophisticated techniques for forecasting respirator stockpiling demands and offers insights for comparing functional and financial aspects from an operational research perspective.

Modelling and solving of the multi-calendar naval surface ship work period problem
Yun Yin, MASc. Student

This presentation will present the modelling and solving of the multi-calendar naval surface ship work periods problem (NSWPP). The specific network structure and constraints of NSWPP will be introduced, including the limited work period, activities of varying priority, multi-calendars activities, multi-calendar resources, precedence relationships, and timeenforced constraints. NSWPP is modeled as a resource-constrained project scheduling problem. A mixed integer linear programming model focusing on front-loading activities based on priority and duration is developed. However, RCPSP is a NP-hard problem, in which the difficulty and computational time grows with the number of activities and resources. To shorten the computational time, serial schedule generation scheme (SGS), a heuristic method, is used to narrow down the time window of activities and generate a feasible solution to warm-start the solver engine (Gurobi). The modifications made in serial SGS to adapt the network structure and constraints of NSWPP will be introduced. Results from multiple experiments using data from real large-scale refit operations will be presented.

Developing a decomposition matheuristic method to solve the multicalendar naval surface ship work period problem
Hyojae Kim, MASc. Student

In this work, a multi-calendar naval surface ship work period problem (NSWPP) is formulated as a resource-constrained project scheduling problem (RCPSP). Given that the NSWPP is NPhard Problem, it is difficult to solve within a reasonable time with computation time increasing exponentially as the number of activities increases. For solving real world problems, a decomposition matheuristic method is proposed to quickly provide feasible and near optimum solutions. The matheuristic uses a multi-step optimization procedure to minimize the weighted priority-duration of the activities. The proposed method handles multi-calendars activities, multi-calendar resources, precedence relationships, and time-enforced constraints. Results of numerical examples are discussed and performance of the model in obtaining the solution and computation time is discussed.

Contact Person:
Prof. Dr. Floris Goerlandt
email: floris.goerlandt@dal.ca