Fall 2022 Projects
Insightin Health is a healthcare consultant company that makes data-driven recommendations to solve healthcare challenges. These recommendations are typically given to healthcare payers to deliver member value. By surveying Medicare patients from December 14, 2020, through August 17, 2022, Insightin Health identified that 6.6% of respondents indicated it was difficult to get any care, tests, or treatment that they needed, and the factors that affect patients’ access to care prior to this project were unknown. The goals of this project were to identify what combinations of factors are most influential in a patient having access to care, develop a machine learning application using these factors, and develop a list of recommendations to improve patients’ access to the treatment, testing, and care they need. To identify the most impactful factors in patients’ access to care, the student team tested several combinations of parameters in a feature selection greedy algorithm. After doing so, the team found that using forward feature selection and random forest classifier as the estimator resulted in the best measures of accuracy, precision, and recall. Twenty-four features selected through this method were included in a final supervised machine learning model that reached 93.9% accuracy, 90.0% precision, and 98.6% recall. After identifying the impactful features, the team created a document listing data-driven and research-based recommendations for improving patients’ access to care. The list includes the features found to be significantly impactful to a patient having trouble receiving the care they need, what can be done to reduce the impact of these metrics, and any applicable resources used to develop the solutions. The document of recommendations and machine learning model have been shared with Insightin Health to assist their team in addressing the factors that impact Medicare patients’ access to care.
Paragon Foods is a rapidly growing company that prepares fresh produce for restaurants, universities, and hospitals. At Paragon Foods in the JustCut operations, the Main Line for production produces made-to-order precut and prepackaged vegetables. The Main Line can produce a wide variety of final products; however, the current process is chaotic due to line flexibility, product variability and lack of staffing. The main goals of this project were to establish baseline data, create a simulation of the process in SIMIO, and combine these two resources in order to analyze the process and optimize worker allocation on the Main Line. Controlled time studies were conducted to obtain the appropriate data needed to construct an accurate simulation. In order to manage product variability, group technology was utilized to develop three major part families that account for eighty-five percent of the total production volume. Simulations were built to model each of the part families for them to be analyzed. Using the simulation, a sensitivity analysis was performed to see how the models would react to variable changes. Looking at the results of the experiments, it was determined that certain variables had much larger impacts on the system when they were adjusted in the simulation. Therefore, it was recommended that these variables be increased or decreased appropriately. It was also recommended that scheduling is done so that vegetables that are in the same part family go through the system in sequence with one another.
JADCO Manufacturing, a steel fabrication company located in Harmony, PA, currently has issues reporting defects throughout their entire production process. An older company that has experienced recent significant growth, JADCO has outgrown their legacy defect reporting system and asked our team to help digitize the process. Our team worked closely with JADCO’s senior management and quality assurance representatives to not only achieve this goal, but also identify root causes for recurring issues. Our approach consisted of designing multiple digital form design options, performing a cost analysis on jobs that had a defect, creating a tableau dashboard to visualize and monitor the defects, and generating recommendations based on our root cause analysis.
Based on our data analysis and root cause analysis, we recommend that JADCO focuses their effort on reducing defects related to drawing/blueprint errors and communication failures, because these issues contribute disproportionately to the costs incurred by defects. They can reduce these defects by implementing our digital defect reporting system and Communicating Changes SOP, which details who needs to be notified when a specific issue occurs. If JADCO implements our recommendations, they can expect to see a reduction in costs due to defects, better visibility of defects in real-time, and improved employee satisfaction.
JADCO Manufacturing focuses on manufacturing quality impact and abrasion resistant steel products. In recent JADCO history, a lack of comprehensive safety documents and procedures has led to ambiguity among workers in terms of job performance and safety measures. The goals of the project include creating a dynamic dashboard in excel for safety and processes as well as the redesign of a material handling equipment rack. These will first be used in the welding department at JADCO. Within the created dashboard, fields such as which work area within the welding department, what the work task is, the severity and probability scores, and what PPE is required can all be changed depending on the job. Once hazard considerations and severity and probability scores are selected, a hazard level will show with a connected color. In order to eliminate safety hazards from the process, the redesigned rack no longer requires operators to perform lifts that present safety risks. As well, the dashboard has a field to select what type of material handling equipment is needed and the weight of the part to tell the operator what spot on the rack the necessary piece of equipment is located. This as well eliminates safety risks like choosing the wrong clamp out of the process by designing it into the dashboard. In the short term, the dashboard will be used in the welding department but in the long term, can be used in the entire JADCO plant. Another long term recommendation is to change from paper copies to tablets so the operators can fill out the dashboard themselves rather than the current way of having the safety manager do all forms. In conclusion, the designed dashboard increases worker understanding of safety in normal everyday tasks and associated risks with welding, grinding, etc.; redesigning the rack allowed for the connection between the dashboard and the rack to reduce misuse of equipment; and the dashboard proves the need for software such as this in manufacturing applications.
Exacerbated by pandemic constraints, MSA faces challenges ensuring on-time and consistent delivery from their suppliers. While MSA currently has robust services for monitoring and mitigating the risk of their larger suppliers, they do not have methods for monitoring and reporting risk for their smaller suppliers. The Pitt Senior Design Team was tasked with creating a dashboard to dynamically monitor MSA’s smaller suppliers. MSA desired a dashboard that would incorporate their current system, Dunn & Bradstreet (DnB), as well as various other risk scores. The goal was to build an intuitive and accessible dashboard that can be frequently updated to reflect a supplier’s current status.
To effectively achieve this dashboard, the Pitt Senior Design Team needed to first establish what types of risk were most relevant to be highlighted by the dashboard. Through a series of stakeholder interviews and surveys, five critical risks were identified. Those five risks were product importance, number of parts supplied, total spend, geopolitical, and weather. Supplier specific data such as product importance, number of parts supplied, and total spend were all provided to the Pitt team by MSA directly. The other metrics, which are region specific, were found using trusted online databases.
The final deliverable presented to the MSA is a web-based Dashboard created using Tableau. The Dashboard displays normalized and weighted risk scores for each supplier that MSA is currently using. The Dashboard has two main areas for further analysis. The first is a list of each supplier and their overall risk score with the ability to find the location of each supplier on the map as well as display a graphic highlighting how they performed in each risk category. The second is a map highlighting each supplier location with a weather risk panel, allowing the team to determine which suppliers are most at risk for specific weather disasters.
Through the use of this tool, we believe the MSA will be able to significantly streamline their supplier tracking operations. The Dashboard effectively fills the visibility gaps between large and small suppliers, creating a one-stop-shop for the MSA team to perform supplier-level risk mitigation.
IDL Worldwide is a leading global brand experience agency that defines, creates, produces and transforms brand experiences. IDL works as a job shop that can design, print, cut, and finish graphics that allow companies to advertise their products as needed. IDL has recognized that its demand is expected to grow in the next few years and believes that they will not be able to hire the labor necessary to meet these rising demands. In response, they are considering two different opportunities to investigate automation options to increase their fulfillment capacity without the need to hire new labor. The Indigo Digital Offset printer has been identified as one possibility to replace, as the current printer is underutilized due to its limited ability to only print on a small selection of materials. The second area of opportunity is a replacement or addition to the digital plotter tables for product finishing and cutting, as the current plotter tables cut at a much slower rate than manual cutting methods. In order to determine if IDL should invest in new technologies for these two areas, our team first had to gather information on the current usage of the Indigo Printer and Plotter using datasets provided by IDL. For the indigo, we considered available technologies that could increase the number of orders that could be printed from a digital offset printer method, and for the plotter we searched for faster plotter tables that could automatically load, process, and unload sheets. After finding some options and comparing them to the current means of production, it was concluded that the Canon imagePRESS V1370 would be the most fiscally viable option to replace the indigo, saving 57 labor hours per week and decreasing IDL’s need to outsource materials, and the Elitron TAV-R was identified as the best addition to the plotter capacity by adding 130 automated hours to the finishing department. With a 7-year payback period and 10% MARR, the Canon impagepress1370 net present value comes to $452,175 and the Elitron TAV-R comes to $149,348. Combined, these two recommendations come to a total present value of $533,341.
LMI, or Logistics Management Institute, is a large government consulting agency headquartered in Tysons, Virgina. They are dedicated to powering a high functioning government with their complex digital and analytical solutions, logistics, and management consultancy services. One of the key business units of the organization is the Enterprise Technology Service Management team. They are responsible for the procurement and distribution of laptops and other technology throughout the organization. An issue they have recently identified as a target is the lack of laptop supply and inability to keep up with the frequency of new hires, upgrades, and other needs for laptop fulfillment within the company. Based on the information provided there were three main factors contributing to this; variable and low visibility of lead times from suppliers, interdepartmental communication, and externally, the inconsistency of the global supply chain. We collected and cleaned hiring, termination, invoice, and lead time datasets from the client and client’s supplier which allowed us to conduct a statistical analysis on the flow of laptops throughout the internal supply chain of the LMI. A linear and centered-moving-average based regression was run on each month for hiring and terminations datasets. This determined that there were monthly effects in January and September for hiring, and April for terminations. From these findings, we applied the regression equations to a model that optimized ordering time, quantity, and inventory levels. The model also accounted for other variables presented by user input which included lead time, date, inventory, hiring/acquisition buffer, supplier distribution, and a safety stock buffer. It then took those inputs and returned two order schedules based on long lead times exceeding 4 weeks and normalized lead times for the proceeding three months. This model was developed as an Excel workbook using macros and VBA (Visual Basic) code. Its main purpose is to increase visibility on laptop demand and create an optimized ordering schedule for LMI to use as a guide for their laptop procurement method and was implemented to do so. In conclusion, we were able to save LMI 25% on ordering costs, deliver a functional excel tool with elements of flexibility and simplicity, and decrease laptop supply stockouts. We recommended that the client maintain a schedule of evaluating ordering on the first of the month and keeping 2-3 weeks of inventory on hand to account for any extreme unexpected cases.
Stoelzle Glass is having issues with their material delivery process. From the inaccurate quantity of movements per shift to the communication and documentation process of material moves, it affects the picking and delivery times of pallets. This causes confusion and wasted time spent working on non-value added activities. For our approach, we decided to focus on three simple aspects that would have the biggest impact. To be able to implement easy solutions that would be flexible to change and fully supported by sound analysis. In regards to the main findings, we found that there is a lack of visibility when it comes to the flow of material. This refers to the delayed key in of material moves into Oracle that creates uncertainty in how much and when material is needed. This results in the disorganization of inventory and excessive amount of packaging materials on the production floor. Consequently, this eventually leads to inefficiencies in transportation and inventory management, negatively affecting overall material flow in the facility. For our recommendations, we believe in implementing a new inventory warehouse, a Kanban system, and 5S methodology to solve these issues.
This semester we worked closely with the data and quality team at Kennametal. Kennametal is a manufacturing company that creates carbide machining tools. They have recently implemented "smart factory" grinding machines that collect an abundance of data that can be overwhelming and impossible to draw conclusions from. A system is needed to allow Kennametal to simplify and explore their historical data, draw conclusions regarding capability, and understand patterns that can be used to improve future performance. To address this problem, we created a suite of tools that gives the user a complete picture of their process.
Functions of System:
- Easy preprocessing functions to correct the input data from machines immediately
- Current state analysis displays
- Comparative visuals to help identify processes of interest
- Ability to zoom in to the machine, part id, order, or piece level
- Analysis at the part-by-part level to give insight into sources of variability
This tool will be used by Kennametal to better understand their grinding processes and act as a basis for future root cause analysis projects. Using our system, we found one major source of variability and were able to explain the mature of the source. In grinding operations, tool wear occurs rapidly and must be compensated for. Currently, the grinding machines use two compensation cycles that act independently of each other but do not account for how the process changes when their resets occur simultaneously. We found that if possibility of simultaneous resets is accounted for, Kennametal can increase capability of their highest volume processes by up to 96%. In the few cases we explored, we found opportunities for major improvement in Kennametal’s manufacturing process using this system. We recommend Kennametal applies this system to their entire manufacturing process to find and explain other sources of variation in their processes.
This semester our group worked with Penn United Technologies Inc., an employee owned high precision metal manufacturing company. We worked specifically with their carbide manufacturing division in the areas where they package and reclaim their graded material. Our problem statement is as follows: To redesign and refine material handling processes in the powder department to improve safety and efficiency. Currently in there areas, employees are at risk for ergonomic related hazards due to lifting overexposure and unsafe lifting loads. In addition to reclaim and packaging, storage reorganization was also necessary to allow for more working floorspace.
We were able to break up the three problem areas, storage, reclaim and packaging into three separate recommendations. Within storage we recommend implementing a gravity flow rack and a workstation crane to decrease the storage footprint from approximately 1,500 sq ft to 225 sq ft and provide aided lifting in the storage area where there previously wasn’t any. For reclaim, we recommend moving all reclaim processes into the same area to concentrate all equipment together and provide space for the new packaging solution. We also recommend the use of automated guided vehicles to automate the process of reclaim pickup which reduces employee overexposure to heavy lifting. Lastly, we redesigned the packaging process so that material is loaded into buckets using a conveyor system which removes the necessity for any heavy lifting. Before implementation, RULA, REBA, and NIOSH scores for the lifting done in the area indicated that employees were at a very high risk of erogonomic hazard. Should Penn United move forward with our solution, little to no lifting would be required within this area.
The ability to regulate the mechanical stiffness in a large range could be crucial for soft robots to conform, grasp, and move while interacting with the environment. Stimuli-responsive materials can be used to achieve shape morphing. Liquid crystal elastomers (LCE) are heat-responsive materials with regular deformation properties when raised or reduced temperature. Polyether ether ketone (PEEK) is a semicrystalline thermoplastic with excellent mechanical and chemical resistance properties that are retained to high temperatures.
In this project, our team designs, tests, adjusts, and optimizes the “LCE cylinder with PEEK disks tethered to a string” structure. Our team successfully demonstrated that when heat is applied to the structure, it would shrink in the axial direction, and the disks inside would jam together to create a rigid cylinder. When heat is removed and cooled, the structure returns to its original versatile state. Our desired goal was to satisfy at least a 10% stiffness difference between the soft and rigid states through jamming phenomena.
Our team was able to design a series of 12 experiments. We mainly explored two independent factors, the number of disks and the level of excessive compression, and found the relationship between these factors and stiffness difference. Through our experiments, we are confident that when weight is applied to our structure, at least 10% stiffness change requirement is met, with stiffness differences varying from 37.93% to 185.85%. The structure with 6 PEEK disks in the LCE with a 10% excessive level of compression design has the best overall performance. More replications can be conducted to increase the stability and accuracy of the results in the future.
Our project provides a heat-responsive system that changes stiffness on command repeatedly. This opens the door for a variety of soft robotics and applications in different industries.
PNC recently acquired approximately 700 new ATMs from BBVA, and would like to ensure that the BBVA ATMs are cost effective and operate similarly to the PNC ATMs. In order to successfully integrate the new BBVA ATMs into PNC’s current ATM network, PNC has assigned the team the task of comparing both ATMs in terms of maximum holding amount, service frequency, market, demand distributions, and residual percentage distributions, as well as performing a cost analysis between BBVA’s ATMs and PNC’s ATMs.
To begin the comparisons, the team clustered the BBVA and PNC data together according to the K-means cluster method in R. From the cluster analysis, it was determined that 83% of all ATMs fell into three significant clusters: 1, 2 and 6. Through the cost analysis, the team was able to conclude that the BBVA ATMs have less total cost than the PNC ATMs, which is attributed to lower demand. Due to this, BBVA has a more aggressive approach to cash handling and relies more on Emergency Cash Orders compared to PNC. In turn, PNC’s conservative management strategy allows them to maintain less Out-of-Cash instances than BBVA but with higher total costs.
Using this knowledge and additional insight, the team suggests that PNC define a standardized emergency cash reorder point of $10,000.00 for cluster 1 ATMs. As for cluster 2, the team advises PNC to continue to manage the ATMs using the current management strategy and protocol and extend this management style to the recently acquired BBVA ATMs with similar performance metrics. Additionally, the team suggests increasing BBVA’s vendor delivery frequency to improve overall ATM performance. Lastly, for cluster 6, the team suggests PNC use a standard order of the maximum holding amount for each ATM while maintaining the same vendor delivery schedule.