Abstract

In today’s highly automated automotive paint shops, production equipment such as robots and conveyors controlled by various sensors and programmable logic controllers (PLCs) are widely used to coat vehicle bodies-in-white. A high-volume automotive paint line typically consists of miles of conveyor lines and dozens of sealing and painting robots equipped with hundreds of sensors and PLCs. It takes a body-in-white several hours to travel on conveyor tracks to complete the painting process. Throughout this process, thousands of process variables are monitored and collected from sensors and devices that control the automatic equipment. When a vehicle is painted, process variables and associated outcomes are collected in a centralized database called the Painted Surface Performance Management (PSPM) system. This database contains a complete history of each painted vehicle and its corresponding quality inspection results. In this work, machine learning (ML) models are used to select a list of the most important features that may affect the outcomes of the painted vehicles. However, this list needs to go through additional steps of feature investigations to validate true causal relationships between the selected features and the target variables. Once this step is completed, the selected features can then be used to target operational first time quality (FTQ) of products through use of such ML models to predict FTQ outcomes. An automotive paint shop data from 2020 to early 2021 was used to conduct a case study to demonstrate the process of feature investigation. Observations and recommendations from the ML model analysis are summarized for improvement.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call