290 likes | 462 Views
Primary & Secondary Incident Management Dr. Asad Khattak Old Dominion University akhattak@odu.edu Acknowledge: H. Zhang, X. Wang, L. Zheng, K. Yang ITSVA Conference Results are preliminary. Secondary Incidents. HR ~45,000 incidents responded to yearly Secondary incidents-3% to 15%
E N D
Primary & Secondary Incident ManagementDr. Asad KhattakOld Dominion Universityakhattak@odu.eduAcknowledge:H. Zhang, X. Wang, L. Zheng, K. YangITSVA ConferenceResults are preliminary
Secondary Incidents • HR ~45,000 incidents responded to yearly • Secondary incidents-3% to 15% • 45,000*0.03 = 1350 to 6750 • Analyzed primary incident durations • Analyzed dependence of secondary incident occurrence on primary incident duration
Incident Data • Provided by Hampton Roads Smart Traffic Center (STC) • Vehicle-based record includes 43 variables • Archived as Mircrosoft Excel table • Jan. 2004 to June 2007 • Exact position of incidents not available within section
Incidents in HR • Durations • Conversion from a “vehicle file” to “incident file” • Duration of incidents is ~14 minutes, on average • Incident response times • If SSP is detection source (~90%), then zero • If non-SSP detection, then ~8 minutes
Secondary Incident Definition http://www.youtube.com/watch?v=KMwXBWML9E4 S P
Secondary Identification Secondary incident with…. • Static time/space based method • 15 min • 1 hour/1 mile, 2 hours/2 miles • Queue-based dynamic method • Secondary only occurs within incident duration and within queue • We used a version of this method • Same direction only or both
C1 C2 Section1 Section 2 Section3 Some incidents may be missed
C3 C4 C2 C1 Section 1 Section 2 Section3 Issue: 1 Pair (C1 vs. C2, C3, C4) or 2 Pairs (C1, C2) and (C3, C4)
Single-incident queuing calculation Maximum Queue length (Qmax ) caused by incident is indicated as green dash line V = Demand (vph) C = Capacity (vph) Ci = Remaining capacity due to incident (vph) T = Incident duration (hours) Total vehicle delay (TD) caused by incident is the colored triangular area between the arrival and departure curve
Demand and Remaining Capacity • AADT% distribution in urban areas is shown • Demand V= (AADT%)*AADT • Remaining Capacity derived from HCM (2000) Exhibit 22-6 (shown) according to incident severity
Identified secondary incidents Two bounds: Rubbernecking effect
Descriptive statistic for Duration (2006) Total incidents = 38614 Normal:37304 Pri:662 Sec:687
I- 64 I- 264 564 I- 664 I- 464 Individual and Secondary Incidents Frequency Distribution (2006)
Distribution Density (2006) Individual Incident Distribution Density I- 64 564 I- 264 Secondary Incident Distribution Density I- 664 I- 64 I- 464 564 I- 264 I- 664 I- 464
Relationship between variables • Type • Detection Source • Weather condition • Lanes closed? • Number of Vehicles involved • EMS response? • Right shoulder affected? • Left shoulder affected? • Ramp affected? • AADT • TOD Simultaneity Duration Sec. Incident
Regression Models • Duration Model Use linear model Duration = b0 + b1 (Detection) + b2 (weather) + b3 (Type) + b4 (Laneclose) + b5 (# of vehicles) + b6(EMS) + b7(AADT) + b8 (Rightshoulder) + b9 (leftshoulder) + b10(ramp)+b11(Resptime) + b12(Primary) + e • Secondary Incident Model Use Binary Logistic model Logit(P(SEC)) = g0 + g1 (Detection) + g2 (weather) + g3 (Type) + g4 (Laneclose) + g5(# of vehicles) + g6(EMS) + g7 (AADT) + g8 (Rightshoulder) + g9 (Leftshoulder) + g10 (ramp) + b11(Resptime) + b12(Duration) + b13(TOD)+ u
Factors associated with Duration: • Secondary incident (if occurs, then longer) • Response time (the higher, the longer) • Detection Source (CCTV, VSP, radio and phone call have longer duration compared with the SSP) • Incident Type (Accident longer duration) • Severe injuries (EMS responded implies longer) • Freeway facility damage-if lane closed, then longer • Vehicles number (more, the longer) • AADT (more, the longer)
Secondary incident occurrence model Used all normal incidents and primary incidents. Not including the secondary incidents since they are highly associated with the primary incidents. The Pseudo R2 for Log transformed model is higher
Factors associated with secondaryincident occurrence • Primary incident duration (the longer the higher) • Happened in peak hours (higher) • Freeway facility damage-lane closure • Vehicles involved (the more the higher) • Higher AADT level (the more the higher) • Weather (?)
Simultaneity • Durations of primary incidents are expected to be longer if secondary incidents occur • The secondary incidents are more likely to occur if the primary incident lasts long • Calculate the residual variable • w = SEC – P(SEC) • w is added to the original regression to test for simultaneity • The resulting equation is: • Duration = b0 + b1 (Detection) + b2 (weather) + b3 (Type) + b4 (Laneclose) + b5 (# of vehicles) + b6(EMS) + b7(AADT) + b8 (Rightshoulder) + b9 (leftshoulder) + b10(ramp)+b11(Resptime) + b12(Primary) + b13(w)
Test Simultaneity Use all “normal” incidents and “primary” incidents. Not including the secondary incidents since they are highly associated with the primary incidents. Residual is significant
Delay Distribution Delay of 163 primary incidents on I-64 WB in 2006 Nearly 80% primary incidents have no delay Total: 305,195 Vehicle-Hours Caveat: Method applies in certain conditions
Closure • Work in progress • Need to look at “rubbernecking” • Clean up the models • Use the results for better planning and operations • Future work • Prediction models • Better geo-coded data • Contact: akhattak@odu.edu • ODUTRI: http://eng.odu.edu/transportation/