Abstract
Batch synchronization has been widely misunderstood as being only needed when variable trajectories have uneven
length. Batch data are actually considered not synchronized when the key process events do not occur at the
same point of process evolution, irrespective of whether the batch duration is the same for all batches or not.
Additionally, a single synchronization procedure is usually applied to all batches without taking into account the
nature of asynchronism of each batch, and the presence of abnormalities. This strategy may distort the original
trajectories and decrease the signal-to-noise ratio, affecting the subsequent multivariate analyses. The approach
proposed in this paper, named multisynchro, overcomes these pitfalls in scenarios of multiple asynchronisms.
The different types of asynchronisms are effectively detected by using the warping information derived from
synchronization. Each set of batch trajectories is synchronized by appropriate synchronization procedures, which
are automatically selected based on the nature of asynchronisms present in data. The novel approach also includes
a procedure that performs abnormality detection and batch synchronization in an iterative manner. Data from
realistic simulations of a fermentation process of the Saccharomyces cerevisiae cultivation are used to illustrate
the performance of the proposed approach in a context of multiple asynchronisms.