Advanced LIGO at Stanford

At Stanford, I had the chance to work on Advanced LIGO (Laser Interferometer Gravitational-Wave Observatory), a super-ambitious project to directly detect gravitational waves. On February 11, 2016, LIGO famously announced the first ever gravity wave detection!

As a Graduate Research Assistant, I worked on several projects focusing on seismic isolation, ranging from mechanical hardware analysis and optimization to software tool development and control loop tuning.

Strut Optimization

Using FEA modal analysis, I worked on choosing optimal strut arrangements for the beam splitter structure. We chose to minimize modal mass and maximize modal frequency for the lowest two structural modes. However, we also considered other factors, such as the third structural mode, and the modal mass of higher frequency modes.

Some of the biggest challenges were accurately modeling the fixed constraints in FEA, dealing with the space constraints of existing structures, and designing for placement in different chambers.

We demonstrated several strut arrangements, all with significant improvement over the nominal design. To validate the analysis, we performed hammer tap tests on a representative structure. The experimental results agreed very well with the FEA prediction.

Check the full report for more details.

Filter Blending

To provide active seismic isolation, several sensors (inertial and displacement) are used for the position control of a two-stage active platform. Depending on the system mode of operation, these sensor signals are blended differently (in frequency space) to produce a position signal for feedback control. Switching between these sensor blends was done offline and required costly system downtime. I helped design and implement the smooth filter switching operation that switches from one sensor blend to another in real time.

To implement real-time blend-switching, we rely on a modified Simulink diagram (see image), C code to ramp between blends, and a Perl script to provide hierarchical control of the switching. I worked on the first two tasks. The final result was an intuitive GUI incorporated into the existing MEDM interface (see image), commonly used in the experimental physics community.

Check the blend switching user guide for more details.

System Identification

I worked on a system identification project for the two-stage active seismic isolation platform. Using a 4th order coupled mass-spring based ARMAX model, we were able to successfully capture the system dynamics in the desired frequency range of 100mHz to 100Hz. The model's step response was close to the experimental response.

Check the full report for more details.