Industrial Projects

2016-2017 Projects

Data Analysis Case Study: Nanocomp Technologies

Sponsor: Nanocomp Technologies, Inc. (Merrimack, NH)
Advisor: Randy Paffenroth
Student: Jonathan Orr (MS-Industrial Mathematics)

In this project we demonstrate several techniques for achieving a deeper understanding of the various steps within the production process of Nanocomp's Miralon yarns. The primary goals of the project were to determine the between lot and the within lot variations, and to develop con dence intervals that enable Nanocomp to set accurate tolerances for the MiralonTM yarns provided to their customers.

Process Mining the Credit Suisse Advisory Process

Sponsor: Credit Suisse (Zurich, Switzerland)
Advisors: Randy Paffenroth and Stephan Sturm
Students: Shannon Feeley, Ian Jacoway, and Gina Rios

Our Major Qualifying Project is to show the business benefits of process mining, and our sponsor is Credit Suisse. Credit Suisse uses many different processes for its employees to onboard, advise, and manage its clients. Each activity that an employee completes in the process is logged in the form of an event. An event consists of a timestamp, a case ID, and an activity. A collection of events that are part of the same process is called an event log. Event logs can be analyzed, or "mined," to discover a process model. The derivation of a process model is called "process mining." Through the use of process mining, a company can compare the process model to the intended business model to identify deviations. We consult with the Credit Suisse team in charge of the advisory process to compare the process model derived from the advisory process data to their intended business model of the advisory process. The advisory process is Credit Suisse’s five step process to assist their clients in creating investment portfolios. We are able to identify deviations from the intended model and suggest methods of reducing operational costs by eliminating redundant tasks and points of congestion to increase productivity.

Improving Student Placement in IQP Centers via Preference Matching

Sponsor: IGSD (WPI)
Advisor: Andrew C. Trapp (MG)
Student: Camila Siqueira Dias (IE), Lin Jiang (ID), and Elizabeth K. Karpinski (IE)

The goal of this MQP is to improve Worcester Polytechnic Institute’s (WPI) Interdisciplinary and Global Studies Division (IGSD) off-campus IQP placement process. The IQP is a cornerstone of the WPI plan, however, placements have become increasingly complex and demanding as the number of applicants grow. Our team proposes an Excel-based decision support tool to improve the matching of students and IQP project center directors based on expressed preferences, thereby assisting IGSD by recommending student-IQP site placements. The decision-support tool uses VBA to build and solve an optimization model. In addition to reducing the time and effort required to place students, we believe that the decision-support tool will increase overall participant satisfaction with placements.

Modeling-Backed Microwave Imaging in Closed Systems: Reconstruction of a Spherical Inhomegeneity

Sponsor: MACOM
Advisor: Vadim Yakovlev
Student: Taylor York

Despite a wide practical use of sophisticated imaging devices based on ultrasound, X-Ray, and MRI, the need of safe, accurate and cost-efficient imaging technologies is still acute. This project addresses the problem of development of reliable and robust computational techniques required for processing of data in microwave imaging performed inside closed metal cavities. We outline a computational procedure for the use in microwave imaging of a spherical inhomogeneity in a dielectric sample. This procedure utilizes an artificial neural network capable of full reconstructing geometrical and material parameters of the inclusion. The network uses data from the FDTD model of a multiport microwave system. A series of computational experiments is reported for the standard four-port waveguide element (known as a Magic Tees Junction) containing a rectangular Teflon sample with a hidden dielectric or metal inclusion. The error in reconstruction of four geometrical parameters of a dielectric sphere is shown to be on the level of 1.7-3.3%, whereas the error in determination of complex permittivity of the inclusion is about 9.8%. The project makes a solid theoretical background for the upcoming experimental program exploring the resources of multiport closed systems for practical microwave imaging applications.

Maintained by webmaster@wpi.edu
Last modified: Jun 20, 2010, 09:03 EDT