Statistical Methods Can Confirm Industry-sponsored University Design Project Results

Abstract

An industry-sponsored project was recently developed to automatically inspect soup mix packages. The industry sponsor had determined that its highest customer complaint was the absence of a flavor packet within the soup mix package. It partnered with Indiana UniversityPurdue University Indianapolis (IUPUI) to develop an automatic system to detect the missing flavor packet and remove it from the production line before the package was bulk-packed for shipment. The system was designed, built and installed by a team of Electrical Engineering Technology (EET) and Mechanical Engineering Technology (MET) students. A four-hour production test confirmed that the percentage of soup mix bags without flavor packets detected by the machine was nearly the same as the total percentage of bags without flavor packets returned by customers the previous year. But how reliable was the system over a longer period? This paper describes a semester-long IUPUI project to determine how well the inspection system performed on its production line for a ten-month period. An honors-student project was devised to use multiple statistical methods to determine whether the automatic inspection system actually improved the overall quality of the soup mix shipments; leading to reduced customer complaints. Customer complaint data for four-million units were analyzed to determine whether a significant difference of complaints existed between the production line with the inspection system and the one without. These data were analyzed using a Two Proportion Hypothesis Test to determine if there is a difference, and a Confidence Interval to estimate the size of difference. The student concluded with 95% confidence that customer complaints were significantly lower on the production line with the inspection system.