STATEN predictions in a CDFR subassembly and the use of simulated outlet temperatures to test decision-making techniques

In order to gauge the viability of temperature noise as a blockage detection method the axially changing geometry of a CDFR subassembly outlet region is modelled using STATEN. Five linked axial stages are used, each with different velocities and assumptions of turbulence values, covering the venturi shield region and incorporating the effect of an orientation bar. A number of different intra-bundle blockage sizes, in corner and central locations, together with various cross-subassembly temperature tilts, are used to estimate the corresponding temperature profiles at the bundle exit. These are then input to STATEN to produce simulated temperature signals on the subassembly axis at its outlet. Investigations are made into the performance of an automatic decision technique - the Adaptive Learning Network (ALN) - using these simulated temperature signals. It is demonstrated that, under the assumptions made in the simulation, it is possible for one ALN, trained on data from all blockage and tilt conditions, to detect a 6% blockage with a high probability. (author)