This thesis is concerned with the design of multivariate deviations-from-nominal (DNOM) control charts for short production runs or job-shop processes in which many small batches of differing part types are run and the number of quality characteristics varies with the batch.
Two multivariate DNOM charts for the process mean using individual observation and sample mean are proposed for the cases where the process parameters are known and unknown. For the parameter unknown case, the updating formulas to estimate variance-covariance matrix are also proposed. These charts are all plotted in a standardized normal scale, and therefore the processes producing different part types can be monitored with only one control chart. Both multivariate DNOM charts are compared with corresponding multiple univariate charts in terms of average run length performance. Under various correlation structures, the proposed charts are found to be more effective than multiple univariate charts.
