morley

Slide1: 

Array Element Localization Using Ship Noise Mike Morley Stan Dosso, Ross Chapman Pacific Rim Underwater Acoustics Conference, Vancouver, BC October 4, 2007

Array Element Localization: 

Array Element Localization AEL : Using measured acoustic travel times to estimate individual receiver positions on an array (source positions). Seabed Receivers Source c(z) xr, yr, zr xs, ys, zs

AEL inversion method: 

Iterative linearized inversion of ray-tracing eqns: Regularization: Prior position and uncertainty estimates Array and/or source track smoothness AEL inversion method Minimize: Where:

GoM experiment: 

GoM experiment VLA deployed in ~800 m of water VLA allowed to freefall to seafloor Recorded water-gun & ship-noise data along tracks Severe timing errors/clipping in water-gun data Ship noise data used instead

VLA: 

VLA 16 hydrophones, 12.5 m spacing 200 m aperture Autonomous recording and data storage 6-s record every 18 s 10 kHz sampling rate

Ship noise data processing: 

Ship noise data processing 20–600 Hz band-pass filter applied to ship noise before cross correlating Obtained relative acoustic arrival times from cross correlations of ship noise

Ship noise data processing: 

Ship noise data processing Results of cross correlating h/phs 2 & 6 along N-S track line

Data and prior uncertainties: 

Data and prior uncertainties Relative travel-time picks from cross correlation 13 rec. & 64 src. positions → 832 data 295 unknowns: (xr, yr, zr, xs, ys, zs, t0) Data uncertainty: σt = 0.2 ms Prior pos’n uncertainties: Rec: δxr = δyr = 1000 m, δzr = 100 m Src: δxs = δys = 15 m, δzs = 3 m

Results: 

Results Inversion initialized from VLA drop position Sol’n converged in 7 iterations VLA repositioned ~44 m SE of deployed position Source positions unchanged from prior estimates Plan view of src. & rec. positions

Results: 

Results Top of array deflected: ~5.5 m towards SSE agrees with avg. dir. of bottom current (ADCP) H/ph spacing: 12.6 m (inversion) 12.5 m (nominal) Water depth: 768.8 m (inversion) 769.0 m (charted) Profile view of array shape

Data residuals: 

Data residuals Fit data to χ2 = N = 832 >23 times smaller than for starting model Avg. data fit to 0.2 ms Data residuals from prior and estimated model

Model uncertainties: 

Model uncertainties Linearized uncertainty est. Absolute Monte Carlo 500 iterations of inversion Gaussian errors added to data, priors & starting model Relative Monte Carlo Trans. & rot. errors removed ¯¯ x -- y ··· z ـــ r = (x2 + y2)½

Array geometry vs. source repositioning: Synthetic study: 

Array geometry vs. source repositioning: Synthetic study Plan view of true src. & rec. positions Incl. horizontal aperture: 9 receivers 150 m separation b/t receivers 64 source positions Receiver pattern offset 70 m in X, -120 m in Y to break symmetry Gaussian rand. errors added to data and true src. & rec. positions

Array geometry vs. source repositioning: Synthetic study: 

Array geometry vs. source repositioning: Synthetic study Both src. & rec. positions improved from prior estimates

Array geometry vs. source repositioning: Synthetic study: 

Array geometry vs. source repositioning: Synthetic study Abs. errors in general agreement with linearized estimates ¯¯ Estimated uncertainty • Absolute errors Source position errors

Conclusions: 

Conclusions Relative acoustic travel times were obtained from ship noise by cross correlation of filtered time series b/t separated receiver pairs Accurate h/ph positions were obtained from inversion broadband ship noise Array with horizontal aperture allows for better source repositioning – avoid symmetry in src/rec geometry Negates need for impulsive acoustic source: Allows for use of sources of opportunity where there is worry of acoustic environmental impact where covertness is required