Data analytics in semiconductor industry

Batch-wise manufacturing processes are found in many industries like chemical, pharmaceutical, bio-technical and semi-conductor. Typical examples include PVC polymerization, fermentations, beer brewing, and wafer etching. Batch process usually has a finite duration, from initialization to completion, and the trajectories of batch process variables describe dynamic time dependency. Batch processes give rise to data tables that are different from the two-way data structures. Measured data from batches are handled in three-way matrices (Figure 1). There might be more than one block of batch process data. Initial conditions data are given by one data table (often called the Z-matrix), information of relevance here would be characteristics of raw material or environment. In the second data matrix, batch evolution data are gathered. This matrix is often called the X-matrix. Naturally, the last block of data, often called the Y-matrix, is composed of results and product quality data for each batch.