Abstract:
|
The magnetic diagnostics subsystem of the LISA Technology Package (LTP)
on board the LISA PathFinder (LPF) spacecraft includes a set of four tri-axial
fluxgate magnetometers, intended to measure with high precision the magnetic
field at their respective positions. However, their readouts do not provide a
direct measurement of the magnetic field at the positions of the test masses, and
hence an interpolation method must be designed and implemented to obtain the
values of the magnetic field at these positions. However, such an interpolation
process faces serious difficulties. Indeed, the size of the interpolation region is
excessive for a linear interpolation to be reliable while, on the other hand, the
number of magnetometer channels do not provide sufficient data to go beyond
the linear approximation. We describe an alternative method to address this
issue, by means of neural network algorithms. The key point in this approach is
the ability of neural networks to learn from suitable training data representing
the behaviour of the magnetic field. Despite the relatively large distance
between the test masses and the magnetometers, and the insufficient number
of data channels, we find that our artificial neural network algorithm is able
to reduce the estimation errors of the field and gradient down to levels below
10%, a quite satisfactory result. Learning efficiency can be best improved by
making use of data obtained in on-ground measurements prior to mission launch
in all relevant satellite locations and in real operation conditions. Reliable
information on that appears to be essential for a meaningful assessment of
magnetic noise in the LTP. |