Title: Process monitoring for manufacturing attribute data using model-based approach
Authors: Angelo Santanna - Federal University of Bahia (Brazil) [presenting]
Abstract: Solving problems in industry, even inside companies known as expert in their sector, is not just a question of applying the right technique. The control chart is a traditional tool for data monitoring processes and the model-based approach has been shown to be very effective in detecting disturbances in output variables when input variables are measurable. The idea of the model-based control chart is to integrate generalized linear model and control chart tools to monitor any changes in process data. In many situations, there are variables that are nonconforming data following a Binomial distribution, and the modeling and monitoring this type data suffers serious inaccuracies in control limits specification when the rate of nonconforming is small. We propose the monitoring of nonconforming data using a Beta distribution approximation. A case study is illustrated for the proposed method to compare the results against several model-based charts and a simulation study based on Markov chain is conducted to overcome such inaccuracies and performance for process monitoring.