Calculation times for complex FTAs in Robin
Building the complete set of Fault Tree Analyses needed to substantiate the safety of an integrated system is a long process. FTs keep growing in size and complexity, until you can’t hardly manage to see the trees through the forest unless supported by a solid platform. To better manage complex systems FTA sets, Robin offers a collection of handy features from the usability point of view: full screen option, transfer gates, treeview tab, primary events data list, etc.
With regards to calculation, Robin FTA implements two paths: exact method or MCS. The algorithm for the exact method computation has been optimized to achieve even better results than the approximate MCS for medium to low complexity FTs, see Direct Evaluation solution competing with Minimal Cut Sets methods. For MCS, Robin FTA offers the possibility to choose the Cut Set cutoff which is set by default to 5. For most FTAs, this will work nicely, providing a reasonable computation time. How about complex FTAs?
Introducing Artificial Intelligence in Robin
Robin team is developing a smart approach to predict the time that a specific fault tree lasts to be computed. As it could be expected, there are several factors that affect a computation, such as:
The type of computation (exact method vs Minimal Cut Sets)The number of unique primary eventsThe number of repeated eventsThe cut set cutoff orderThe intrinsic performance of the server(s)
Since the release of Robin version 1.3, we are monitoring and collecting these and other computation parameters. Hence, we are building a common database that store the execution time of each fault tree calculation accompanied by the aforementioned parameters. This will allow, from Robin version 1.4 onwards, that any customer could look up to the historical data of execution times (anonymous data coming from all customers) besides knowing its own calculation durations. The provided charts will look similar to the following figure:
The above graph includes the scatter execution time points of each type of computation and smooth 3-D surface splines, which are surfaces that smoothly suit the provided scatter data based on a Radial Basis function. Besides providing these statistic graphs, this spline function will be used, since version 1.4, to give the user an idea of how much time a specific calculation would last depending on the aforementioned parameters. This will support the user on deciding specific input calculation parameters, such as the Cut Set cutoff order.
The path to a smarter RAMS companion
Although the main objective of this new feature is to provide the user with more information before computing a specific fault tree, this is also a big step towards implementing Artificial Intelligence in Robin for the first time. In the near future, when more data is available, other AI models will be tested (e.g. Random Forest) to perform a deeper benchmark of the best model for our purposes.
We look forward to providing you this smarter Robin to speed up your path towards complex systems Fault Tree Analysis sets.