Fault diagnosis in the steel coil manufacturing process

The use of advanced statistical models can help industries to design more economical and rational investment plans. Fault detection and diagnosis is an important problem in continuous hot dip galvanizing. Increasingly stringent qualityrequirements in the automotive industry also requires ongoing efforts in process control to make the process more robust. Robust methods for estimating the quality of galvanized steel coils are a tool for the comprehensive monitoring of the performance of the manufacturing process. The main objective of this paper is to use real data to identify which steel coil manufacturing process characteristics affect defective risk, and which models better fit data to evaluate the probability of failure. This study applies different statistical regression models: generalized linear models, generalized additive models and classification trees for estimating the quality of galvanized steel coils on the basis of short time histories. The data, consisting of 48 galvanized steel coils, was divided into sets. Five variables were selected for monitoring the process: steel strip velocity and four bath temperatures. The first data set was 25 conforming coils and the second data set was 23 nonconforming coils. Generalized linear models (GLM) are an extension of linear models for non-normal distributions of the response variable and nonlinear transformations. Generalised additive models (GAM) are a natural extension of GLM in the sense that they adjust nonparametric functions to study the relationship between predictive variables and the response variable. In general terms, the purpose of the analyses via classification trees is to determine a set of if-then logical (split) conditions that permit accurate prediction or classification of cases. The results show that the classification trees provide good estimates of quality coils and can be useful for quality control in the manufacturing process. We conclude that the different models used provided basically similar results, reinforcing the validity of the identification of influential characteristics.