Date of Award

2026-08-01

Degree Name

Master of Science

Department

Systems Engineering

Advisor(s)

Sergio Luna Fong

Abstract

Aerospace wire harness manufacturing operates under conditions that standard industrial engineering methods were not designed to address. Volumes are low, configurations change between orders, assembly is performed manually by certified labor that cannot be expanded quickly, and IPC/WHMA-A-620 Class 3 acceptance requires that every unit be verified rather than sampled. Kit assembly concentrates the resulting risk, because each item retrieved from the rack is an independent selection decision, and the attributes distinguishing a correct item from an incorrect one are frequently not visible to the operator. Errors introduced there are discovered downstream, after substantial certified labor has been invested. A manager deciding how to protect quality at this operation may add an operator, install a guided-selection control, or install automated verification, but has no analytical basis for the choice: static capacity models represent no quality consequence of staffing, and conventional discrete-event practice specifies quality as a fixed yield applied to finished output, which forecloses the evaluation of any intervention intended to change the rate at which errors occur.

This thesis develops a discrete-event simulation of a wire harness kitting operation in which material-selection errors are generated by the work performed rather than assigned to it. Each completed selection is treated as an independent error opportunity, so that kit-level release quality follows from the order mix and from the number of selections each order requires, while only the residual per-opportunity probability is supplied as an input. Four configurations - the existing manual process, pick-to-light guided selection, computer-vision-assisted verification, and a two-worker cell with sequential checking - are compared through residual escapes per completed opportunity and clean-kit release rate, measured at the boundary where the kit is released to downstream operations. Each configuration was evaluated over 300 terminating weekly replications, and the sensitivity of the comparison to its transferred parameters was examined directly.

The model reproduces a relationship that fixed-yield representations cannot: release quality degrades as the number of selections a kit carries increases, so the benefit of a control acting at the point of selection scales with order complexity. All three interventions reduce escapes relative to the existing process, raising the clean-kit release rate from 94.7 percent to 96.4 percent under guided selection and to 97.3 percent under automated verification. The ordering between those two controls, however, follows from effectiveness values transferred from unrelated studies, so the defensible result is a threshold rather than a preference: guided selection must achieve approximately 50 percent relative error reduction to match automated verification. The two-worker cell yields the largest quality improvement together with a substantial capacity increase, but its advantage over automated verification erodes as the dependence between the two checks rises and disappears when the second worker confirms rather than independently re-derives the requirement. The value of a second check is therefore governed by how the second worker's task is arranged rather than by the addition of a person. The study is comparative rather than predictive, and it concludes by specifying the facility measurements required to convert the framework into a basis for investment approval.

Language

en

Provenance

Received from ProQuest

File Size

76 p.

File Format

application/pdf

Rights Holder

Jorge Mares

Download

Share

COinS