Live Sit Liv e Site Demon e Demonstr traD aDon R on Resu - PowerPoint PPT Presentation

Live Sit Liv e Site Demon e Demonstr traD aDon R on Resu esults lts Tobyhanna To a Army y Depot t (TOAR) ) Formerly Used De ormerly Used Defen ense Sit se Site (FUDS), e (FUDS), Pennsylvania % Pe % Military MuniMons Support Services (M2S2) Webinar April 21, 2016 MaOhew Barner, P.G. David Wright 1

In InvesD esDgaD aDon on p purp rpose ose To demonstrate performance of advanced EMI sensor in detecMon of subsurface metal and classificaMon of derived targets as TOI under challenging field condiMons. • Dynamic Survey – performed to idenMfy locaMons of subsurface metal and derive targets for follow-up cued interrogaMon • Cued InterrogaMon – performed to gather data at target locaMons, which will be used to derive extrinsic and intrinsic properMes of metal objects 2

Gener Gen eral s site l e loc ocaDon on % % I-380 MRS R04A West Operational Grids ESTCP Demonstration Area State Game Lands Road Prospect Street/PA 423 3

Demon Demonstr traD aDon ar on area ea % % • ~11 acres • 100’ x 100’ Grids • Suspected impact area west of Grids 78/49 and 78/48 HISTORY: • Originally established as Camp Sumerall in 1909 • Machine gun & arMllery training in 1913 • Field arMllery pracMce from 1913 to 1949 • Live cannon fire 1919 to 1932 • Explosives storage depot ~10 months in 1919 4

Rec ecover ered ed* mu * muniD iDon ons-r s-rela elated it ed items a ems at t MR MRS 04A S 04A W Wes est t • 75mm shrapnel projecMle • 75mm HE projecMle • 155mm shrapnel projecMle • 155mm HE projecMle • PD fuze Note: 37mm projecMle use reported onsite but at different part of TOAR FUDS *Previously recovered by USACE contractor performing site invesMgaMons 5

Sit Site e condiDo ndiDons ns % % • Densely wooded • Variable underbrush thickness • Impact craters • Large boulders • Variable relief • HunMng area • Remote locaMon (Access by UTV) • Poison plants, Mcks, bees 6

Technology Te 7

Approa Ap oach ch • Conduct G-858 transect survey to select ~2 acres of grids • Prepare the site • Conduct dynamic TEMTADS survey (0.5 m lane spacing) • Process data / select target locaMons • Flag target locaMons / conduct cued interrogaMon • Analyze data /derive prioriMzed dig list • Intrusively invesMgate all target locaMons 8

In Instru trumen ment V t Verific erificaD aDon Strip on Strip 9

In Instru trumen ment V t Verific erificaD aDon Strip on Strip % % T-005 (blank) T-004 T-003 T-002 T-001 Line 2 End 10

Dynamic TEMT Dyn amic TEMTADS 2x2 su ADS 2x2 survey y • 8 days to collect 4 grids (~1 acre total) • 3 days for gap-fill (aOempt to get 100% coverage in one grid) 11

Dyn Dynamic su amic survey r y resu esults – w lts – wes est grid t grids s 12

Dynamic su Dyn amic survey r y resu esults – eas lts – east grid t grids (in s (iniD iDal c al cover erag age) e) 13

Dyn Dynamic su amic survey r y resu esults – eas lts – east grid t grids (ad s (addiD iDon onal al co cove verage) ) Grid with addi,onal coverage 14

Dynamic su Dyn amic survey r y resu esults – eas lts – east grid t grids (t s (tar arget t selecDons) 15

Cu Cued ed TEM TEMTADS ADS 2x2 s 2x2 survey y % % • 264 targets in east grid pair • 165 targets in west grid pair • Targets flagged by surveyor • 7 days data collecMon 16

Fu FuncD cDon t on tes ests ts 17

Dynamic IV Dyn amic IVS su S survey p y perf erforman ormance ce • Horizontal targets are more difficult to posiMon using amplitude response peak detecMon. 18

Cu Cued IV ed IVS su S survey p y perf erforman ormance ce 19

Dynamic IV Dyn amic IVS su S survey p y perf erforman ormance ce % % • Targets selected in dynamic survey using amplitude response peak detecMon have errors associated with coverage gaps • Cued target posiMons are derived from dipole fit analyses MQO MQO 40 cm 25 cm 20

Perf erforman ormance ob ce objecD jecDves an es and r d resu esults – d lts – dyn ynamic su amic survey % % Performance Minimum Acceptable Metric Data Required Result Objective Criteria Initial dynamic survey Accuracy of derived Derived target positions Derived positions within ± 25 cm of Fail (one outlier data positioning target positions from initial measurements the ground truth from horizontal at the instrument targets – at 31 cm) verification strip (IVS) Ongoing dynamic Precision of derived Derived target positions Derived positions within ±25 cm of Fail (two outliers survey data target positions from daily measurements at the average positions during from horizontal positioning the IVS ongoing daily measurements targets – all within 30 cm) Along line Point to point sample Mapped survey data 98% ≤ 25 cm; no gaps >40 cm Fail, (Pass for 100% measurement spacing distance unless obstruction or hazard is coverage area) present Dynamic survey Effective footprint Mapped survey data 100% at ≤ 75 cm cross-track Fail, (Pass for 100% spatial coverage coverage measurement spacing with intended coverage area) spacing of 50 cm Detection of TOI Percent of seed items Seed item locations 100% of seeded items within a 40 Fail (all detected, detected cm halo of ground truth but coverage gaps Geo-referenced anomaly resulted in 2 list distance failures) 21

Perf erforman ormance ob ce objecD jecDves an es and r d resu esults – cu lts – cued su ed survey % % and c and classific lassificaDo aDon n % % Performance Minimum Acceptable Metric Data Required Result Objective Criteria Initial cued survey data Accuracy of dipole-fit Target fit positions from IVS item fit locations within ±25 cm Pass positioning derived target positions initial measurements at the of ground truth locations IVS Ongoing cued survey Precision of dipole-fit Target fit positions from IVS item fit locations within ±20 cm Pass data positioning derived target positions daily measurements at the of average fit locations during IVS ongoing daily measurements. Initial cued sensor Accuracy of dipole fit Dipole-fit derived Library Match metric ≥ 0.9 to initial Pass polarizability accuracy derived intrinsic target polarizabilities from initial polarizabilities for each set of features measurements at the IVS inverted polarizabilities Ongoing cued sensor Precision of dipole fit Dip-ole-fit derived Match metric ≥ 0.95 to initial Pass polarizability precision derived intrinsic target polarizabilities from daily polarizabilities at the IVS for each features measurements at the IVS set of inverted polarizabilities from daily measurements Cued interrogation Instrument position Cued data 100% of anomalies where the center Pass anomaly coverage of the array is positioned within 30 cm of anomaly location Correct classification of Number of TOI correctly Ranked anomaly lists 100% of all seeded targets Pass TOI identified Scoring reports from 100% of all TOI categorized as ESTCP Program Office “digs” or “Can’t Analyze” Model results support Number of anomalies Modeling fit coherence ≥ 90% of targets have fit coherence > Pass classification decision classified as “Can’t results 0.80 Analyze” 22

Pr Prelimin eliminar ary R y ROC cu OC curve e 23

Lessons learned Lesso ns learned • Increased prism height was factor in lower posiMoning precision (more pronounced in detecMon survey) • Use of RTS presents challenges with line-of-sight in wooded condiMons: – Refresh rate of RTS a?er re-establishing lock with prism – Total loss of prism by RTS / need for it to search for prism • Weight of sensor in tandem mode / personnel faMgue much greater than with system on its wheels • DetecMon survey presented greater technical challenges, but classificaMon sMll successful 24

This invesMgaMon was completed as ESTCP Project MR-201314 and under contract W912HQ-13-C-0039 Special thanks to: 25