Chapter
5
Conclusions and future works
In this work several model was constructed enable to simulate the different dynamic behaviour of a substructure between a simulated ideal state and a simulated test state, when it is attached to its assembly and when it is attached to a shaker to test it, respectively. In the first part an harmonic analysis was carried out on simple models, with 1DOF per node, using the software Matlab. The substructure (box), was attached to the assembly (engine), at one point on the real configuration and it was attached to the shaker at one point on the test configuration. The mounting point response, where the box is connected to the engine, used as input for the shaker, can be measured either with or without the box mounted on it. Using the mounting point response without the box mounted on the engine, the resulting behaviour for the substructure was different between test and real configurations. In some frequency ranges the displacement of the box was higher on the test configuration than on the real configuration leading to over testing, while in other frequency ranges under testing occurred. Over and under testing happened mainly close to the neo natural frequencies from the connected engine and box system. The amount of over or under testing depended on the mounting point of the engine and on the type of bracket used. The largest differences were found for the box fixed on a flexible part of the engine by a bracket without an anti-vibration system. However using the mounting point response from the engine with the box mounted on it there was no difference between the two configurations but most of the time this vibration environment is not available for intellectual properties. Afterwards the box was attached at more that one point to the engine to better simulate the ideal state. The over or under testing originated one more from the different input spectrum for the shaker, which was caused by the presence of the box on the mounting points. Another issue was raised
for these configurations, in fact, even if the box was attached at more than one point to the engine and to the shaker, the shaker can only reproduced one displacement for all mounting points with the same amplitude. The displacement control method, whose the aim is to find an input for the shaker that enables the target point to have the same response on the test configuration as on the real configuration, was implemented for the box attached at four point to the engine. Obviously for this kind of test there was no over or under testing for the point of interest, target point, but different behaviour occurred for the other point of the substructure.
Afterwards a part of aircraft engine was consider as assembly, Combustion Camber Outer Casing (CCOC) and its accessories as substructure, (probe and electronic control unit). As first step a modal analyses on the CCOC for different boundary conditions were done using the software Ansys. By these analyses, and following the suggestion given from Rolls Royce PLC, the suitable constraints enable to simulate the real behaviour of the CCOC into the complete engine, are simple supports. This means the casing is essentially simply supported on one flange plane, with the other flange being translationally restrained except for the axial direction which is unrestrained. All rotational DOFs in the two flange planes remain unrestrained. Next step was an harmonic analysis carried out for the accessories attached using different ways to the casing, for the simulated ideal state, and to the shaker, for the simulated test state. There was not an mitigation on the under and over testing using different kinds of mounts but there was a reduction on the vibration environment transmitted to the accessories. For instance, using simple bracket instead of rigid mount or even better adding anti-vibrator mounts between the brackets and the box. The issue is that in both cases the max amplitude is close to the natural frequencies regarding the modes shape of the mounts, whether of the brackets or of the anti-vibrator mounts. This means high stress for the mounts which became critical substructure for the assembly. Finally the harmonic analysis on the two configurations showed and quantified the over or under testing. This originated one more from the different vibration environment which the accessory is subjected to when it is on the casing or when it is on the shaker. To sum up: the shaker is a very stiff body with its own dynamic, while the assembly has a totally different dynamic behaviour which influences the behaviour of the mounted substructure; the shaker can apply a same displacement for all mounting points whereas the assembly applies different displacement at each point; the assembly applies to the substructure six displacements for each mounting point: three rotations and three translation the shaker can only apply one of these in general a translational displacement.
In a further work an Real-time dynamic substructuring should be applied to the studied models to reproduce more accurately the vibration environment for the substructure on the test configuration. The Real-time dynamic substructuring, [7], [8], [9], is a hybrid numerical-experimental testing technique that combines a critical element, tested experimental, with a numerical model of the remainder of the system being considered. The combination of the two parts of the system during a real-time test is intended to mimic the behaviour of the complete or emulated system. To carry out a dynamic substructuring test the substructure is identified and fixed into an experimental test rig. The interface interaction between the substructure and the numerical model is typically provided by electric or hydraulic actuators, which apply displacements on the substructure. The actuators act as a transfer system and are designed to follow the appropriate output displacements calculated by the numerical model. To complete the coupling between the substructure and the numerical model the forces imposed by the substructure on the transfer system at the interface are measured and included within the numerical model. The whole testing process must take place in real-time to simulate the dynamic behaviours of the emulated system. Typically, for mechanical systems several delays occur between the two parts of the model, such as measurement, signal processing and computation delays, which are large enough to have a significant influence on the overall dynamics of the substructured system. For all these reasons many research are developed in this field, [10], [11], to improve the Real-time dynamic substructuring and to allow the study on a complex model, particularly useful in Earthquake Engineering too.
