Message ID: 15926
Entry time: Tue Mar 16 19:13:09 2021
In reply to: 15902
Reply to this: 15930

Author:

Paco, Anchal

Type:

Update

Category:

SUS

Subject:

First success in Input Matric Diagonalization

After jumping through few hoops, we have one successful result in diagonalizing the input matrix for MC1, MC2 and MC3.

Code:

Attachment 2 has the code file contained. For now, we can only guarantee it to work on Donatella in conda base environment. Our code is present in scripts/SUS/InMatCalc

We mostly follow the steps mentioned in 4886 and the matlab codes in scripts/SUS/peakFit.

Data is first multiplied with currently used inpur matrix to get time series data in DOF (POS, PIT, YAW, SIDE) basis.

Then, the peak frequencies of each resonance are identified.

For getting these results, we did not attempt to fit the peaks with lorentzians and took the maxima point of the PSD to get the peak positions. This is only possible if the current input matrix is good enough. We have to adjust some parameters so that our fitting code works always.

TF estimate between the sensor data w.r.t UL sensor is taken and the values around the peak frequencies of oscillations are averaged to get the sensing matrix.

This matrix is normalized along DOF axis (columns in our case) and then inverted.

After inversion, another normaliation is done along DOF axis (now rows).

Finally we plot the comparison of ASD in DOF basis when using current input matrix and when using our calculated inpur matrix (diagonalizing).

You can notice in Attachment 1 that after the diagonalization, each DOF shows resonance at only one and its own resonance frequency while earlier there was some mixing shown.

Absolute value of the calculated DOF might have changed and we need to calibrate them or apply appropriate gain factors in the DOF basis filter chains.

Next steps:

We'll complete our scripts and make them more general to be used for any optic.

We'll combine all of them into one single script which can be called by medm.

In parallel, we'll start onwards from step 2 in 15881.

Anything else that folks can suggest on our first result. Did we actually do it or are we fooling ourselves?