Statistical Process Control (SPC) methods like control charts and capability indices have been an integral part of process monitoring and optimization in mass production for many years. Samples are taken from the process to assess its stability and its performance based on the comparison between process variation and specification limits. In small batch and job production, SPC methods are seldom used because they require large batch sizes to provide statistically significant results.

In this thesis, it is shown that the clustering of similar features from different products can be used to create virtual mass production processes that satisfy the batch size requirements.

To ensure that meaningful results can be extracted from the SPC analysis of virtual mass production processes, features are defined as similar if their statistical process behavior is homogeneous. A procedure is presented to identify factors such as material, tool, etc. that exert a significant influence on the statistical behavior. Based on this, clusters of similar features are determined.

The main part of the research is concerned with the development of an algorithm that analyzes factors and creates statistically homogeneous clusters automatically, based on historic process data. The core of the algorithm uses statistical hypothesis tests for homogeneity between levels of a factor with regard to mean and standard deviation. Based on the results, cluster analysis techniques are applied to identify the clusters of similar features.

The algorithm is validated by using simulated process data as well as through the application in three different industrial examples of small batch and job production. The results show that virtual mass production processes can be established in a majority of cases while the error rate is on par with comparable statistical tests.