Project Summary

Armada Supply Chain Solutions provides end-to-end supply chain management services for many large clients including McDonalds, Chipotle, and Chick-fil-A. They currently operate four warehouses around the US. The three older warehouses share many design similarities while the newest warehouse in Flower Mound, TX, is significantly different. Prior to the project, Armada struggled to compare their warehouses and was unclear on which design choices were best for efficiency. Armada would like to gain clarity on the impacts of each layout design choice as they are planning to build a new warehouse in the coming years.

Our team decided to utilize the DMADV approach from six sigma to structure our project. We began work with Armada to define the problem and the details of our project, including goals and scope. Armada supplied most of the data we used in our project for the measure phase, but we supplemented their info with research. We used the variety of data collected to perform analysis on several key areas of warehouse performance. These calculations covered space utilization, throughput, labor productivity, and a cost analysis. A data envelopment analysis was performed using these metrics to holistically compare the four current warehouses. Following our analysis, the focus was on delivering our client our findings and recommendations. We developed a packet containing all key findings with our explanations and the supporting calculations and figures. This is presented to the client for feedback and verification.

Analysis revealed that the distinct design choices of the Flower Mound facility, namely a flow-through warehouse with narrow-aisle racking, were less efficient than the conventional designs of the other three warehouses. A variety of metrics calculated support this finding. The shallow racking depths in Flower Mound reduced the pallet per square foot. Additionally, narrow aisles and specialized material handling equipment slowed the rates of labor processes including picking and put away. While the new warehouse in Flower Mound did have the highest number of locations available for storage, it had lower inventory volume compared to the other warehouses and lower throughput. Of the three conventional warehouses, Romeoville turned out to be the most efficient according to the data envelopment analysis.

Based on our findings, Armada should use Romeoville as their benchmark warehouse. We recommend several features for Armada to implement in their future warehouse. Utilizing a U-shaped warehouse with shipping and receiving docks on a single side of the warehouse will improve labor productivity. Matching racking depth to product velocity will allow better space utilization. Standard aisle widths of twelve feet will also allow for quicker material handling. We also found that Armada would benefit from balancing areas devoted to each temperature zone. However, additional measurements taken at Flower Mound could affect conclusions since that warehouse hasn’t been operational for much over a year. The packet completed for our client contains our findings and recommendations with all supporting calculations. This project succeeded in enhancing Armada’s understanding of their current warehouses and informing them about the creation of their future warehouse.

Project Summary

In this project, our team addresses axle damage at Alstom's West Mifflin site caused by unreliable packaging during transportation by providing three alternative packaging designs. Our team applied the DMAIC methodology to define the problem, measure the extent of damage, analyze root causes, improve packaging designs, and establish control measures. Three new alternative packaging solutions were developed, focusing on enhancing stability, reducing stress, and promoting sustainability. Alternative Three, which includes a foam-backed fall protection system, can achieve the greatest improvements: a 26% reduction in the center of mass and a 55.56% decrease in pallet stress. Financial analysis showed that although this solution slightly increases packaging costs, it is expected to significantly reduce axle damage and associated delays. Implementation plans include supplier coordination, employee training, and continuous damage monitoring to ensure long-term success. This project not only resolves the immediate packaging issue but also provides a foundation for Alstom’s future process improvements.

Project Summary

FedEx’s current Harmonized Tarif classification model relies solely on text descriptions to predict HTS codes used for international shipments, achieving only 15% accuracy for fully qualified codes. This low accuracy results in costly customs delays, product seizures, and financial penalties due to a lack of compliance. With over 5,000 highly specific HTS codes and increasingly complex global trade regulations, there is an urgent need to improve classification accuracy, especially as custoemrs often provide product images that remain unused in the current process.

To generate text descriptions from images, we focused on image captioning models within computer vision. After evaluating various models based on accuracy, speed, and flexibility, we selected the BLIP model for its superior performance. BLIP (Bootstrapping Language-Image Pretraining) is a deep learning framework capable of generating natural language from visual inputs. We fine-tuned the model using 10,000 FedEx product images and corresponding real-world descriptions. After three epochs of training, both training and validation losses decreased, confirming improved model performance and more realistic caption generation.

To test the effectiveness of the model, the team compared the performance of a control and experimental group. The control group consisted of short product captions, while the experimental group consisted of product descriptions that were generated using the model, in which only images were used as input. These product descriptions from the two groups were then input into the already existing HTS classification model, and the HTS codes that were output were recorded.  The accuracy of these HTS codes were measured, and the experimental group had a 6% higher accuracy than the control for predicting the entire HTS code correctly. The accuracy of the entire code predicted increased from 15% accuracy to 21%, which exceeds the goal of reaching 18% accuracy. 

Integrating this image classification solution with FedEx’s already existing platfroms such as FedEx Ship Manager and FDX is reccomended, enabling FedEx to capitalize on uploaded product images for more accurate HTS code generation. This implementation should be scaled globally to maximize operational benefits across FedEx’s international shipping network. Further refinement using expanded datases would significantly increase prediction accuracy beyond current levels, ultimately providing FedEx with a competitive advantage in international logistics.  

Project Summary

FedEx, one of the biggest transportation companies in the United States, operates both air and surface networks moving millions of packages across the globe daily. Those packages are supplied and demanded from different places using trailers. When a trailer loaded with goods arrives at its destination, it becomes an empty trailer. Simply redirecting those empty trailers to their origin costs a lot of money and is not very effective.

This project addresses a critical logistics optimization problem, focusing on minimizing transportation costs for a network of distribution hubs. By determining the most strategic and impactful nodes, in terms of volume processing, geographical location, and balancing trailers in FedEx’s current ground operations, this study aimed to develop the most cost-efficient trailer distribution network. So that to streamline operations, reduce expenses, and enhance logistical efficiency. The problem was divided into two phases. One is to find the shortest path for full trailers to satisfy their respective demand. Another is to balance the empty trailers across the network. Balance the empty trailers means that the same number of trailers that leaves a facility must return to the same facility. This constraint was given by FedEx

Initially, data preparation involved extensive cleaning and consolidation of trailer demand and transportation distances among multiple facilities. Additional complexities included handling directional routes, managing clusters of geographically close hubs, and ensuring all routes were comprehensively considered.

A thorough network analysis was conducted to ensure the feasibility of route optimization, including detecting and resolving cases of disconnected or one-way routes. Missing reverse routes were identified and systematically completed to maintain a balanced and fully connected transportation network.

Multiple optimization methodologies were explored, such as NetworkX and Google OR-Tools, with the aim of achieving an optimal flow allocation that minimizes total transportation cost. Ultimately, the selected approach employed a minimum cost flow optimization algorithm, which significantly improved efficiency by determining the optimal distribution of trailers across all hubs based on distance and demand.

The optimized results offer clear operational benefits, including significant cost savings, more balanced hub utilization, and improved resource allocation. This structured approach ensures operational robustness and provides a scalable framework for future logistics management and strategic decision-making.

Project Summary

In an effort to increase package visibility at their sorting locations, FedEx has recently piloted Radio Frequency Identification (RFID) technology within three facilities. However, the company is unsure about how to utilize the data received from the RFID scanners. During this semester, our team worked with FedEx to help analyze the East Cleveland facility’s RFID data and provide throughput rates of specific manual process areas along with ideas for possible uses with the technology. Throughout this process, FedEx allowed our team to use data that was from January 18 to February 14 and February 24 to March 8.

Since the team obtained the data in different segments, we first combined all of the files using Microsoft Excel and Python and created a master data set. Initial analysis was then performed, and the time spent in each process area for each package was calculated. Through data cleaning and manipulation, the team isolated each process area into its own Excel sheet, and descriptive statistics were calculated regarding the time packages spent traveling through that specific process area. Afterwards, the data for each process area was fitted to a normal distribution which allowed us to remove all values that were three standard deviations away from the mean. Descriptive statistics were then recalculated without these values. This provides FedEx with the time that the packages spent at each process area on average without the outliers.

After the analysis, the team created a training document that explains the process to FedEx. Examples of Python code and important figures are provided. This document would allow the company to replicate our work well after the completion of the project. We recommend repeating the analysis on the other two facilities with RFID scanners already installed and then repeating it for other time periods, like Peak/ Christmas season for example, where the volume is significantly different. Additionally, possible ideas for use are listed out so that FedEx can utilize the RFID data for other purposes. This includes tracking outlier packages to check for problems in other process areas, updating productivity standards, identifying faulty RFID scanners, and gauging how many process areas packages visit on average.

Project Summary

The Lunar Intelligence and Reconnaissance Initiative (LIRI), in collaboration with Lawrence Livermore National Laboratory (LLNL), aims to develop a satellite surveillance network to monitor activity on the lunar surface. As lunar exploration grows, ensuring continuous monitoring for national security, scientific research, and resource tracking is becoming increasingly important. The project is designed to track man-made objects and activities in cislunar space.

LIRI focuses on optimizing existing satellite and sensor technology to always survey a large area of the moon. Using the DMAIC methodology (Define, Measure, Analyze, Improve, Control), the team will address key challenges for selecting appropriate sensors within specific orbital configurations and various constraints. By choosing satellite placement and imaging techniques, LIRI will enhance the United States' ability to monitor and respond to lunar activities, ensuring a strategic advantage in cislunar space operations.

Keywords— Lunar surveillance, satellite imaging, cislunar space security, ground sampling distance (GSD), space reconnaissance, sensor optimization, orbital configurations, hyper spectral imaging, thermal imaging, data transmission.

Project Summary

This report shows the results of a project with Starline Holdings, LLC, a Legrand company. The warehouse is located in Canonsburg, PA, and builds 50–60 kits daily and stores over 40,000 parts. Workers face delays in getting materials to assembly lines. Problems include unclear picking steps, tight space, software issues, and frequent mistakes.

Our team used the DMAIC method. We studied the current process through observations and time studies. Most time was wasted on actions that did not add value. Delays often came from waiting on forklifts or clearing space.To solve this, our team suggested a new forklift procedure to reduce waiting times and unnecessary movements. We revised the standard work using worker feedback and updated the process map to reflect the new forklift plan. For the control step, we created a full training guide and new standard work to keep the picking process consistent across all material handlers.

Initial results showed a 15% reduction in non-value-added time. These solutions are designed to be scalable, reduce variation, and improve efficiency in warehouse operations. We recommend Starline implement the forklift procedure, revised and comprehensive standard work, and formal training system to support long-term performance improvements.

Project Summary

NuturePA is a nonprofit organization based in Pittsburgh, PA, that operates with the mission of supporting mothers and families of newly born children. Through the Nurture® program, mothers are paired with mentors who exchange text messages relating to maternal mental health, social-emotional activities, parenting knowledge, and anything the mother may be seeking assistance with. The program screens mothers for signs of Perinatal Mood and Anxiety Disorders (PMAD) by using questionnaires: the Edinburgh and GAD-7 Surveys. Currently, NurturePA uses a third-party platform, Sogolytics, to create surveys and store survey responses. NurturePA spends $1,250 annually on Sogolytics. Although Sogolytics manages the survey data, NurturePA has to manually store and analyze the survey results in a separate file, slowing the workflow and increasing the risk of errors. The solution must require an in-house survey platform that automates survey results and data analysis to reduce the time and risk of errors for NurturePA.

Microsoft products, such as Excel, Power Automate, and Power BI, were utilized to automate the survey results and provide data analysis opportunities for NuturePA. The data results are automatically added to NurturePA’s PMAD Tracker and automatically assess a mother’s depression or anxiety scenario number. This provides With Power BI, a dashboard was created with the survey results, allowing NurturePA to do further data analysis. The team recommends that NurturePA study seasonality trends throughout the year, determining if there is a portion of the year where mothers are more likely to suffer from depression or anxiety, and the trend in a mother’s score to be proactive in providing resources to her. The team also found NurturePA’s resources for mothers to be in multiple files, increasing the time to send a resource to a mother. The team categorized the varying resources into their respective categories, scenario text responses or profession-help resources, and included filtering features to allow NurturePA to select a resource based on categories such as county, resource type, and depression or anxiety scenario.

As a result, NurturePA will see a decrease in the processing time involved in surveying mothers, analyzing their scores, and responding appropriately with resources. With less time spent manually entering data, NurturePA will be able to spend more time directly engaging with families. The dashboard provides visibility into the trends of depression and anxiety seen in mothers throughout the year, and to be more proactive in providing help to mothers who are at risk of PMAD. The team recommends the  NurturePA team in the future to expand their surveys to be accessible for non-English speakers and expand resource management to include other demographics such as marital status and income level.

Project Summary

The University of Pittsburgh Medical Center’s (UPMC) Presbyterian pharmacy located in Oakland, Pittsburgh strives to ensure efficient patient medication order, verification, and delivery over all hospital units. The pharmacy, however, faces workflow disruptions and inefficiencies due to a lack of standardization, specifically in the medication delivery process carried out by delivery technicians. These technicians or pharmacy interns experience 15-45 minutes of downtime per hour per technician. This downtime and underutilization is the result of inconsistent delivery routes and a lack of predelivery organization. 

To address this, the DMAIC (Design, Measure, Analyze, Improve, Control) framework was used. Through shadowing observations and data collection via a sign-in/sign-out sheet, a multiple linear regression analysis was conducted to identify the key reasons for delivery time delays. This analysis proved that delivery time was most affected by the size of the delivery load and the use of a delivery cart. The solution and deliverables include a new Delivery Technician Handbook that lays out new, optimized delivery routes for technicians, updated technician roles, standard operating procedures, and best practices. A bin system was also redesigned, and a bin prototype was constructed, all to align with the new delivery routes and to improve predelivery organization. 

Project Summary

The purpose of this project is to present the work of the UPMC Pharmacy inventory and storage team - self-named Medicine Inventory System Supply (MISS) UPMC team - for the Industrial Engineering Senior Design class. Our team worked with UPMC Presbyterian Pharmacy personnel to identify waste and recommend improvements in their receiving process. The state of the receiving process was slow and varied due to a lack of standardization and upkeep of the work environment. Using DMAIC to guide our process, we formulated new workflows and a plan for implementation of 5S to reduce waste and create a standardized team based approach to receiving medicine. Throughout the project, the team managed a relationship with the stakeholders, pharmacy technicians who conduct the receiving operations every day. At first the stakeholders were reluctant to work with our team, but over time we built rapport which culminated in a Kaizen event between our team and the stakeholders. 

Project Summary

WE, The World is a nonprofit organization founded in 1998 by Rick Ulfick, and is now primarily operated by the Executive Director, Angel Short, and the HR Director, Jana Larsen. WE, The World’s mission is to be a platform for their partner organizations so they they may amplify charitable efforts across the globe. Leaders at WE, The World also employ the help of volunteers to acheive their goals. WE, The World currently champions eleven of their own campaigns relating to Unity, Interdependence, Environment, Economic Justice, Health, Children & Youth, Women, Human Rights, Freedom, Disarmament, and Peace. 

WE, The World currently operates on a completely virtual basis, hosting webinars and online events to further their cause. However, WE, The World faces a variety of operational challenges that pose a significant hinderance to their philanthropic goals. WE, The World lacks an impact measurement framework, without which they cannot accurately quantify the impact of their campaigns and grassroots movements. Moreover, WE, The World’s fundraising processes are beset by inefficiencies, and they lack a clear channel of communication to internal and external stakeholders, effectively disregarding the continuous upkeep of their relationships with their partner organizations. 

To assist WE, The World in advancing their mission of maximizing and accelerating global social change, our team of Industrial Engineers from the University of Pittsburgh levereaged the Theory of Change (ToC) to identify various project goals, execute a thorough analysis of WE, The World’s operations, and deliver customized solutions with projected benefits in the short, intermediate, and long term. 

Project Summary

Zoll Medical Corporation manufactures and refurbishes medical devices. One of their devices is the LifeVest, a wearable cardiac defibrillator. These devices are manufactured in Zoll’s GAMMA  Facility. Gamma also receives and inspects all material and components. Currently, Zoll’s GAMMA  Facility is facing challenges with material flow and storage. The material flow and storage is complex and not based on data. This project focuses on the production line refill process in the stockroom and the general storage capacity of the facility. To address these issues, the team used the DMAIC framework to identify areas of improvement and  provide recommendations.

This problem required data from three general categories: process, capacity, and frequency/location. Time studies and interviews were conducted to collect data on the stockroom process, both in discrete time and in process steps. Materials team documents and existing schematics were used to understand and analyze the facility’s current and expected future capacity. Stockroom transaction records were used to measure the frequency at which stock was moved in and out of the stock room.

This data was used in several analyses to identify areas of improvement. The process data was used to create a process map and value stream map of the stockroom. From this, the team identified that many of the steps were non-value added and should be targeted for improvement. The process data also allowed for a pareto analysis of the time study. This showed that extraneous, non-value added tasks, like answering non-urgent questions from employees outside of the stockroom, took up a large portion of the refill cycle time.

The capacity data enabled a comparison analysis of current state storage and proposed future storage alternatives. The facility is currently storing more pallets than it has space for. Through this is analysis, the team found that constructing a storage garage on site is the best option moving forward.

The stockroom transactions fed an FMR analysis. This analysis categorized materials by their transaction frequency from Very Frequent Moving to Slow Moving. The categorizes enable storage location recommendations that better reflect the frequency in which materials are retrieved to better utilize stockroom employees.

Based on these analyses, the team makes several recommendations. For improvement in the process, stockroom quiet hours should be established to limit interruptions and extraneous tasks during peak work hours. In addition, the process should be evaluated to remove non-essential, out-of-scope tasks. For improvement in capacity, a garage should be constructed. Once a garage is constructed and there is flexibility in storage locations, materials should be stored throughout the facility according to the categories provided by the FMR analysis. Very Frequent Moving material should be prioritized for bulk storage in or near the stockroom, while Slow Moving material can be stored further away. This will improve transportation efficiency. 

Implementing these recommendations should provide Zoll GAMMA Facility with a reduction in process time, an increase in capacity, and an increase in efficiency.