Data stream processing in factory automation

Abstract

Data stream processing is a valuable technique to solve demanding problems that also occur in factory automation, such as continuous data processing with high throughput and real-time output, and distributed data acquisition and processing. However, the intricacies of data stream processing techniques make its application difficult in real-life scenarios. One particularly challenging situation arises when changing conditions necessitate a modification in processing logic of system operators. This is especially difficult in the presence of streaming data and transient internal states of the system. Since downtimes are expensive, an efficient solution has to be provided for updating the processing logic. In this paper, strategies for on-the-fly adaptation of data stream queries are presented and experimentally evaluated with examples from the domain of condition-based maintenance. Techniques for state preservation allow for a fast transition to new processing logic. The results show that our strategies are well suited for demanding applications in factory environments.