A New Application of Deep Learning

In progress!

This is an implementation of a new application for Deep Neural Networks (DNN). More Details will be added after completion.

• Period: Spring 2017 - Present
• Advised by: Mark S. Daskin

Close Project

Flexible Network Design under Correlated Uncertainties

In progress!

Traditional Supply Chain Network Design (SCND) models assume demand values at different markets and for different products are independent and the demand distributions are given. These assumptions are not accurate when correlation among the demand of different products exists. In this research, we propose a decision model for designing flexible supply chain networks (SCNs) operating under correlated uncertainties through capacity planning. We extend the current models by coordinating several SCNs of common commodities of different products. To study the correlation across uncertain parameters in the model, we employ distributionally robust optimization for identified sources of uncertainties. Initial results indicate that our model often deploys more capacity than the traditional stochastic models for a profit-maximization problem where positive correlations can lead to profit loss.

• Period: Fall 2016 - Present
• Advised by: Mark S. Daskin

Close Project

Ford-UoM Alliance project: Global Sourcing Under Uncertainty

Supported by Ford Motor Company

Managing uncertainty is one of the challenging issues in supply chains. Using only the expected value of the uncertain parameters for sourcing decisions in a deterministic model can be risky due to the uncertainties that threaten both the optimality and feasibility of the decision variables. Deviations from the expected values of key input parameters can make the deterministic formulations infeasible and/or inefficient much of the time. In this study, we address not only the need of an effective tool, but also how to implement it to incorporate potential uncertainties associated with the model parameters into sourcing decisions.

• Period: Spring 2014 - Present
• Advised by: Mark S. Daskin, Don Zhang (Ford)

Close Project

Data-driven Scenario Reduction Techniques for Network Design Problem

Supported by NSF-UW Data Science Workshop (Invited Talk)

• Period: Spring 2015 - Summer 2015
• Link: -

Close Project

An R-Package for Multi-type Functional Neuroimaging Data Analysis

Supported by NSF-UW Data Science Workshop (Invited Talk)

• Period: Summer 2016 - Fall 2016
• Installiation: (1) Load devtools in R, (2) Run install_github("nimasasa/RFCMFND").

Close Project

Retail Chain Design under Operational and Disruption Risks

Supported by a grant from the INSF

Designing robust and resilient retail networks under operational and disruption risks can create substantial competitive advantage. In this paper, a deterministic multiple set-covering model is first proposed. Then, it is extended to a possibilistic scenario-based robust model by scenario generation and disruption profiling to design a robust and resilient retail network. The developed models are validated through randomly generated examples and a real case study in retailing. Numerical results demonstrate that designing retail chains without considering operational and disruption risks is really misleading. Also, multiple covering of retail stores as the measure of redundancy increases the network’s resilience significantly.

• Period: Fall 2012 - Spring 2013
• Link: Transportation Research Part E.

Close Project

Condition-based Maintenance for Power Generation System

Supported by a grant from the INSF

Condition-based maintenance (CBM) is an increasingly applicable policy in the competitive marketplace as a means of improving equipment reliability and efficiency. Not only has maintenance a close relationship with safety but its costs also make it even more attractive issue for researchers. This study proposes a model to evaluate the effectiveness of CBM policy compared to two other maintenance policies: Corrective Maintenance (CM) and Preventive Maintenance (PM). Maintenance policies are compared through two system performance indicators: reliability and cost. To estimate the reliability and costs of the system, the proposed Markovian discrete-event simulation model is developed under each of these policies. The applicability and usefulness of the proposed Markovian simulation model is illustrated for a series-parallel power generation system. The simulated characteristics of CBM system include its prognostics efficiency to estimate remaining useful life of the equipment. Results show that with an efficient prognostics, CBM policy is an effective strategy compared to other maintenance strategies.

• Period: Spring 2011 - Fall 2012
• Link: Reliability Engineering and System Safety.

Close Project