Vibrating Beam Inverse Problem Team K.E.Y Scott Clark ● Asya Monds ● Hanh Pham SAMSI Undergraduate Workshop 2007
Outline First Model (spring) Potential Problems How to improve Second Model (beam) Results
The first model: Spring Model We have observations : (t 1 ; y 1 ); … ; (t m ; y m ). The goal is to estimate the unknown parameters C and K.
Our cost function: Now we need to minimize the cost function. After running the script, we get: C=0.7284; K=1537.8
Checking the assumptions Homoscedasticity Assumption:
Normality assumption
Independence Assumption
Beam Model
Cost Function Minimize it. How? 7 parameters, YI, CI, ρ, etc Extremum may be dense in parameter space Find “reasonable” values Set beam to same as patch, search near given data, try to minimize a new cost function
A new cost function Needs to take into account spatial variations as well as frequency variations from the model and the data So we use a weighted least squares cost function minimized a simplex method (fminsearch). This doesn’t work. Phase change too much to overcome.
What now then? Limit the search. Fewer parameters, smaller variations. And then, it works! (kind of)
The data
The parameters found gamma air damping YI_beam ** beam -- Young's modulus CI_beam beam -- internal damping Kp Kp for beam rho_patch ** linear density of patch YI_patch ** patch -- Young's modulus CI_patch patch -- internal damping
Thank You Any Questions?