A1088
Title: Implementing arrays for fault detection in R software
Authors: Ulrike Groemping - Berliner Hochschule fuer Technik (Germany) [presenting]
Abstract: Covering arrays (CAs) have been designed for so-called system behaviour testing. They ensure that the test runs cover each level combination of any set of up to $t$ relevant factors at least once. Signing off a system based on a fault-free pass through such a CA provides protection against faults that arise from the combination of up to $t$ factors. In a large system, if something fails in such a CA, it may still be challenging to figure out the root cause of the failure. This gives rise to locating arrays (LAs) that are CAs extended to make sure that there are no perfect ties between different level combinations of $t$ factors. As per existing awareness, there are no R packages for CAs or LAs. Requests for suitable orthogonal arrays for system behaviour testing have been made, while CAs or LAs would often require substantially fewer runs for that purpose. The plan is to expand R's capabilities to include CAs and LAs, including design creation and possibly also subsequent analysis, as well as sequential steps like expanding a CA with the goal of resolving ambiguities. Preliminary research into the topic is presented with this implementation goal in mind. Experts in the field of CAs and LAs will not learn anything new but are invited to contribute their advice to the project of creating an R package as useful as possible for providing practitioners with design - and possibly analysis - features around CAs and LAs.
