Prevention of sheetbreaks using multivariate statistical analysis in an expert system environment

Sheetbreaks in papermaking processes are diagnosed with the help of Principal Component Analysis (PCA) and Projection to Latent Structures (PLS) techniques. These methods are capable of retaining and compressing huge amounts of data into low dimensional spaces. The data are incorporated into expert systems to monitor the operational variables and thereby provide advice on remedial measures. A study of the process data from two mills in North America revealed that the three variables contributing most to breaks were lumpbreaker loadings, vacuum system under the wire and dryer steam pressure.