Development of a Post-Consequence Model (PCoM) for Aircraft Accident Severity Assessment

Transcription

1 Thesis Defense Presentation Rutgers University Development of a Post-Consequence Model (PCoM) for Aircraft Accident Severity Assessment by Denise Marie Andres Thesis Advisor: Dr. James Luxhøj

2 Introduction Research Problem Research Objective Literature Survey Outline Research Methodology Modeling Results Contributions and Areas of Future Research

3 Introduction Aviation Since 1992, more than 60 airlines have begun service Over the last decade the number of passengers flying over the United States increased over half a billion Technology will present opportunities to transform and update the aviation industry New technologies offer tremendous opportunities for improved safety, security, and efficiency

4 Other commercial aviation Airlines General aviation Bicycles Introduction Aviation Safety Flying is safest mode of transportation United States National Safety Council states that it is 22 times safer flying in a commercial jet than traveling by car Rail transportation Recreational boating and other marine Highway Number Of Fatalities 41,800 United States Transportation Fatalities-2000 (Source: NTSB, 2000) Commercial aviation accident rate dropped dramatically, but has remained low and flat Anticipated increases in air traffic is a concern

5 Introduction Aviation Safety and Security Program Safety Research Prevent inadvertent or hazardous events by developing prevention and intervention technologies and strategies to mitigate contributory factors to aviation accidents Strategic Plan Performance Metrics Security Research Improve technologies and strategies that increase the aviation system s resistance against terrorists attacks Reduction or aircraft fatal accident rate by 80% in ten years and 90% in 25 years TARGETED GOAL: Develop and demonstrate products that contribute to a 50% reduction in the fatal accident rate

6 Introduction Safety Product Suite Vehicle Safety Technologies Weather Safety Technologies System Safety Technologies Aircraft Self- Environmental System Vulnerability Protection & Hazards Awareness Discovery & Preservation & Mitigation Management Synthetic Vision Systems (SVS) Single Aircraft Accident Prevention (SAAP) Accident Mitigation (AM) Weather Accident Prevention (WxAP) Aircraft Icing (AI) Aircraft System Monitoring & Modeling (ASMM) System Wide Accident Prevention (SWAP)

7 Introduction Aviation System Risk Model Probabilistic Decision Support System Prototype developed by Luxhøj, et al. Combination of analytic generalization from case studies and modeling sessions with experts Utilizes Human Factors Analysis and Classification System (HFACS) to determine accident precursors Evaluates the impact of the AvSP products on aviation system s component of likelihood 20 models have been developed to address precrash technologies

8 Introduction Assessment of AvSP Products Post- Accident Technologies Pre-Crash Technologies

9 Introduction Accident Mitigation Technology Suite AM Suite Project Goal Improve crash survivability and fire hazard mitigation in fatal and serious accidents by eliminating or reducing accident consequences Aircraft crashworthiness Occupant protection Fire safety Evacuation

10 Introduction Accident Mitigation Technology Suite AM-1: Next-Generation Crash Analysis Codes AM-4: Next-Generation Crashworthiness Design Guidelines Off-board Technologies AM-5: Fuel Tank Fire Prevention and Fire Suppression System Technologies AM-2: Energy Absorbing Seats, Restraints and Structures On-board Technologies AM-6: Cargo Hold Detection and Detection Design Guidelines AM-7: Elevated Flash Point Fuel Technologies

11 Research Problem European Transport Safety Council (1996) estimates 40% of people who should survive, die in survivable accidents 55% resulting from impact trauma 45% resulting from fire-related fatalities AM 2, 5, 6, and 7 designed to reduce the severity level of accidents by eliminating or drastically reducing the probability or outcome severity of the consequence A systematic analytical method for severity assessment impact of Accident Mitigation products on human survivability is required

12 Research Problem The Post-Consequence Model (PCoM) is an EXTENSION of ASRM PCoM is decoupled from ASRM ASRM PCoM Human Loss Event Types PCoM Human Loss Events Classifier Probability of Event Occurrence 35% 30% 25% 20% 15% 10% 5% 0% Fatality (Smoke and Serious Injury Fatality (Initiation of Fire) (Smoke and Fire) Evacuation) Serious Injury (Initiation of Evacuation) Fatality (Trauma) Serious Injury (Trauma) Without AM Technology Implementation

13 Research Objective Evaluation of Accident Mitigation Technologies: Research Project 1: Development of a Post-Consequence Analytical Framework to Assess the Impact of Accident Mitigation Products 2, 5, 6 and 7 Research Project 2: Development of a Post-Consequence Taxonomy for Runway Incursion and Engine Failure Accident Categories Research Project 3: Application of the Analytical Framework to Runway Incursion and Engine Failure Accident Categories Research Project 4: Development of a Prototype Probabilistic Decision Support Tool to visually display mitigating effects of AM products

14 Literature Survey Summary of Hazard Analysis Techniques Heinrich s Theory of Accidents Social Environment Undesirable Traits Unsafe Acts Accident Injury Preliminary Risk Analysis HAZARDS Accidents Portrayed as A Sequence of Events Focus on the events involved rather than the environmental conditions Popular in the aviation community Top-down qualitative approach Postulate unplanned undesired events Investigates event sequence that transforms to potential hazard

15 Literature Survey Summary of Risk Analysis Techniques Analysis Tool Advantages Limitations Fault Tree Analysis (FTA) A P A = AND B C P B =0.01 OR P C =0.43 D E o o Great for working overall probability of undesired event. Probabilities easily determined. o o Does not account for dependencies between events. Linear causality emphasized. P D =0.11 P E =0.32 Fire Starts Event Tree Analysis (ETA) P=0.7 Fatalities 1 Yes P=0.4 No 2 Yes Damage P=0.3 P=0.2 No Yes Fire Controlled 3 P=0.6 No Fire Contained P=0.8 4 (Adapted from Institution of Electrical Engineers, 2004) o o Multiple resulting events analyzed. Probabilities easily calculated. o o o Does not account for dependencies between events. Confined to binary logic modeling. Difficult to incorporate nonlinear relationships. Bayesian Belief Networks (BBN) X 1 X 2 X 3 X 4 X 5 X 6 o o o o o Readily handles situations where data is limited or inaccessible. Models causal relationships. Robust probabilistic inference. Strengths of relationships between elements readily represented. Allows for combination of subjectivity and objectivity. o o Combinatorial explosion if not modeled carefully. All assessments must be well documented to avoid becoming non-traceable.

16 Methodology Runway Incursion and Engine Failure Accident Categories Prior research concentrated on Runway Incursion accident category Runway Incursion and Engine Failure accidents have observable post-consequence events

17 Methodology Development of Taxonomy HAZARD ANALYSIS Examine Accident Reports and Briefs Identify Relevant Post-Consequence Events

18 Methodology Development of Taxonomy HAZARD ANALYSIS Observe Post-Consequence Sequence of Events for each accident Create Database for RI Accidents

19 Loss Initiating Events Methodology Development of Taxonomy SEQUENCE OF POST-CONSEQUENCE EVENTS Hazardous Conditions Human Loss

20 Methodology Development of Post-Consequence Model X 2 X 4 X 5 X 1 X 3 Bayesian Belief Networks X 6 Causal Networks Set of directed links and connecting nodes Links signify direct causal dependencies Nodes signify a single event/variable Parent and child nodes Directed Acyclic Graph (DAG) Graphically represent the relationships among variables of interest

21 Methodology Development of Post-Consequence Model Modeling Techniques Undirected Relations: introduction of a binary variable to establish relationship among two or more variables A B C Divorcing A 1 A 2 A 1 A 2 A 3 A 4 D B A 3 A 4 C B Bayesian Updating a priori probabilities- based on frequency data a posteriori probabilities- probabilities updated with expert judgment and/or new information

22 Loss Initiating Events Modeling Results Development of ACA Post-Consequence Model ACA PCoM without Technology Mapping Hazardous Conditions Human Loss

23 Detect Patterns Modeling Results Application of Modeling Techniques Introduce Node

24 Modeling Results Revised ACA PCoM Pattern Remains the Same Introduction of New Node

25 Node Modeling Results Revised ACA PCoM CPT Size before Modifications CPT Size after Modifications Reduction Abort Takeoff 2 Node eliminated 2 Runway Excursions Collision with Object Collision with Aircraft Ground Impact Initiation of Evacuation Forces Inducing Trauma Smoke and Fire Loss of Fuel Integrity Serious Injury from Evacuation Fatality from Evacuation Serious Injury from Smoke and Fire Fatality from Smoke and Fire Trauma-related Serious Injury Trauma-related Fatality Total Elicitations

26 Modeling Results Development of EF Post-Consequence Model EF PCoM without Technology Mapping Total Elicitations = 162

27 Post-Accident Event AM-2 AM-5 AM-6 AM-7 Initiation of Evacuation X X Modeling Results Mapping of AM Technologies ACA PCoM with Technology Mapping Forces Inducing Trauma X Smoke and Fire X X Loss of Fuel Integrity X EF PCoM with Technology Mapping Post-accident Event AM-2 AM-5 AM-7 Breach of Fuel Tank Integrity X X Fuselage Integrity/Rapid Decompression X Force Impact Trauma X

28 Results Quantifying PCoM: a priori probabilities PROCESS: Converting events of accident reports to frequency information Analyze Database of 96 Airport-Centric Accident (ACA) Reports 28 Engine Failure Accidents Evaluate Events Utilizing Developed Taxonomy Collision with Aircraft Collision with Object Runway Excursions Ground Impact Loss of Fuel Integrity Trauma Induced Forces Smoke and Fire etc. Complete Spreadsheet Determine Conditional Probabilities

29 Results Quantifying PCoM: a priori probabilities 1 = YES (DID OCCUR) RI Cases 0 = NO (DID NOT OCCUR) Event Date Aircraft RI Collision with Object Collision with Aircraft Runway Excursions Ground Impact Loss of Fuel Integrity Trauma Induced Forces Nov-96 Aircraft 1: Beech 1900C Aircraft 2: Beech A May-96 SA Dec-95 Cessna Jul-95 EMB-120 Brasilia Mar-95 Raytheon Mar-95 Raytheon Jan-95 Cessna Nov-93 SA May-93 CE

30 Results Quantifying PCoM: a priori probabilities Loss Initiating Events treated as Bernoulli Variables Conditional Probability Table for Collision with Aircraft

31 Results Quantifying PCoM: a priori probabilities Trauma-related Serious Injury CPT Forces Inducing Trauma CPT Collision Yes No Forces Inducing Trauma Runway Excursions Occur Did not occur Occur Did not occur Occur 8/15= /71=0.55 0/2=0 5/8=0.63 Did not occur 7/15= /71=0.45 2/2=1 3/8=0.38

32 Results Probabilistic Decision Support Prototype The Post-Consequence Model PCoM Accident category Loss Initiating Event Hazardous Condition Human Loss Engine Failure Collision with Aircraft Ground Impact Loss of Fuel Integrity Fire and Smoke Trauma Induced Forces Serious Injury: Fire and Smoke Fatality: Fire and Smoke Serious Injury: Trauma Fatality: Trauma Principal Investigator: Dr. James Luxhoj, Rutgers University Software Developer: Denise Marie Andres, Rutgers University ENTER

33 Preliminary Results Probabilistic Decision Support Prototype

34 ACA PCoM Results and Conclusions Propagation of PCoM: a posteriori probabilities Post-accident Event Forces Inducing Trau m a Initiation Evacu ati on Smoke and Fire Serious Injury Evacu ati on Fatality Evacuation Serious Injury (Traumarelated) Serious Injury (Smoke and Fire) Fatality (Smoke and Fire ) Occurrence Probability: Baseline 33% 54% 35% 30% 5% 10% 7% Occurrence Probability: AM 2, 5, 6, 7 29% 47% 22% 27% 4% 8% 4% 18% 11% Percent Relative Change 12% 13% 37% 10% 20% 20% 43% 39% Human Loss Events 30% 25% Probability of Event Occurrence 20% 15% 10% 5% 0% Serious Fatality Injury (Evacuation) (Evacuation) Serious Injury (Traumarelated) Fatality (Traumarelated) Serious Injury (Smoke and Fire) Without AM Technology Implementation With AM 2, 5, 6, and 7 Implementation Fatality (Smoke and Fire)

35 EF PCoM Results and Conclusions Propagation of PCoM: a posteriori probabilities Force Fuselage Integrity/ Breach of Serious Serious Serious Injury Fatality Smoke Impact Rapid Fuel Tank Emergency Injury Fatality Injury Fatality Fire and Fire and Post-accident Event and fire Trauma Decompression Integrity Evacuation Evacuation Evacuation Trauma Trauma Smoke Smoke Occurrence Probability: Baseline 49% 45% 46% 42% 56% 39% 6% 19% 39% 9% 15% Occurrence Probability: AM 2, 5, 7 39% 33% 36% 24% 50% 35% 5% 14% 28% 7% 11% Percent Relative Change 20% 27% 22% 43% 11% 10% 17% 26% 28% 22% 27% Human Loss Events 40% 35% 30% Probability of Event Occurrence 25% 20% 15% 10% 5% 0% Serious Injury (Evacuation) Fatality (Evacuation) Serious Injury (Trauma) Fatality (Trauma) Serious Injury (Fire and Smoke) Fatality (Fire and Smoke) Without AM Technology Implementation With AM 2, 5, 7 Implementation

36 ACA and EF PCoM Results and Conclusions Propagation of PCoM: a posteriori probabilities Post-accident Event ACA PCoM Forces Inducing Trau m a Serious Injury (Traumarelated) Fatality (Traumarelated) Occurrence Probability: Baseline 33% 10% 6% Occurrence Probability: AM 2 29% 8% 5% Percent Relative Change 12% 20% 17% Force Impact Trauma Breach of Fuel Tank Integrity EF PCoM Emergency Evacuation Serious Injury Evacuation Serious Injury Trauma Fatality Trauma Post-accident Event Occurrence Probability: Baseline 45% 42% 56% 39% 19% 39% Occurrence Probability: AM 2 33% 24% 50% 35% 14% 28% Percent Relative Change 27% 43% 11% 10% 26% 28% Human Loss Events Human Loss Events Probability of Event Occurrence 30% 25% 20% 15% 10% 5% 0% Serious Fatality Injury (Evacuation) (Evacuation) Serious Injury (Traumarelated) Fatality (Traumarelated) Serious Injury (Smoke and Fire) Without AM Technology Implementation With AM 2 Implementation Fatality (Smoke and Fire) 40% 35% 30% Probability of Event 25% Occurrence 20% 15% 10% 5% 0% Serious Fatality Injury (Evacuation) (Evacuation) Serious Injury (Trauma) Fatality (Trauma) Without AM Technology Implementation With AM 2 Implementation Serious Injury (Fire and Smoke) Fatality (Fire and Smoke)

37 ACA and EF PCoM Results and Conclusions Propagation of PCoM: a posteriori probabilities ACA PCoM Post-accident Event Serious Injury Evacuation Fatality Evacuation Occurrence Probability: Baseline 30% 5% Occurrence Probability: AM 5 27% 4% Percent Relative Change 10% 20% Post-accident Event Occurrence Probability: Baseline Occurrence Probability: AM 5 EF PCoM Fuselage Integrity/ Rapid Decompression Fatality Evacuation 46% 6% 36% 5% Percent Relative Change 22% 17% Human Loss Events Human Loss Events Probability of Event Occurrence 30% 25% 20% 15% 10% 5% 0% Serious Fatality Injury (Evacuation) (Evacuation) Serious Injury (Traumarelated) Fatality (Traumarelated) Serious Injury (Smoke and Fire) Without AM Technology Implementation With AM 5 Implementation Fatality (Smoke and Fire) Probability of Event Occurrence 40% 35% 30% 25% 20% 15% 10% 5% 0% Serious Injury (Evacuation) Fatality (Evacuation) Serious Injury (Trauma) Fatality (Trauma) Without AM Technology Implementation With AM 5 Implementation Serious Injury (Fire and Smoke) Fatality (Fire and Smoke)

38 Contributions 1. The PCoM research methodology provides a novel and structured method for modeling post-accident events. the methodology will continue to provide a systematic analytical risk assessment approach as new technologies emerge. 2. Emerging from the methodology is the development of a postaccident taxonomy for Airport-Centric and Engine Failure accident Categories 3. Addresses NASA s and the National goal of reducing aviation fatalities and injuries by focusing on the events that lead to human loss. Applying this knowledge to the development and design of technologies will enhance the quality of the products.

39 Areas of Future Research Area of Research 1: Ranking and prioritizing of the technologies

40 Areas of Future Research Area of Research 2: Integration of ASRM and PCoM PDS Tools PCoM ASRM Risk Assessment Area of Research 3: Distinguishing Parts 121 and 135 Accident Categories Aircraft Accidents Part 135 Part 121