detection of precursors to aviation safety incidents due to human factors i. melnyk, p. yadav, m....

21
Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk , P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department of Computer Science & Engineering University of Minnesota, Minneapolis ICDM Workshop on Domain Driven Data Mining December 7, 2013

Upload: nigel-bell

Post on 06-Jan-2018

223 views

Category:

Documents


6 download

DESCRIPTION

3 Fatal accidents and onboard fatalities [Boeing Report ’12] Introduction

TRANSCRIPT

Page 1: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

Detection of Precursors to Aviation Safety Incidentsdue to Human Factors

I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee

Department of Computer Science & EngineeringUniversity of Minnesota, Minneapolis

ICDM Workshop on Domain Driven Data MiningDecember 7, 2013

Page 2: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

2

• Introduction

• Related Work

• Proposed Approach• HMM vs HSMM

• Experimental Results

• Conclusion

Overview

Page 3: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

3

• Fatal accidents and onboard fatalities 2003-2012 [Boeing Report ’12]

Introduction

Page 4: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

4

• Estimated two-fold increase in air traffic by 2025 [Sheridan ’06]

• Congestion in air and airports• Load on pilots and traffic controllers• Greater chance to make error

• Our objective• Detect precursors to aviation safety incidents due to human factors

– Analysis and modeling of pilot actions– Data generation and evaluation

Introduction

Page 5: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

5

• Hidden Markov Models [Srivastava ’05]• Observations modeled using N-dim binary vector (switches in cockpit)• Cluster data to get smaller class of observations; build HMM over reduced data

• Clustering approach [Budalakoti et al. ’09]• Cluster pilot action sequences using k-medoids based on nLCS• Rank order sequences to identify anomalous sequences

• One-class SVM [Das et al. ’10]• Detects anomalies in both continuous and discrete sequences• Employed Multiple Kernel learning: LCS for discrete, SAX for continuous

• Dynamic Bayesian Networks [Saada et al. ’12]

• Hidden nodes – pilot actions; Observable nodes – aircraft sensors• Detects pilot errors in the past given current instrument data

Related Work: Aviation Safety

Page 6: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

6

• Given database of normal pilot action sequences

• Actions come from finite alphabet:• Examples: “raise landing gear”, “lower flaps”, “decrease throttle”, etc.• Construct model of a normal sequence from data

• Assign anomaly score to a test sequence

• Entire sequence is anomalous (offline anomaly detection)• Specific action is anomalous (online anomaly detection)• Examples: “unusual order of actions”, “forgotten action”, etc.

Problem Formulation

Page 7: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

7

• Flight phases

• Example: landing phase pilot actions

Analysis of Pilot Actions

Action ID

Time

descent touch down braking on runway

stage duration

time btw actionsNull action

Page 8: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

8

• Flight phases

• Example: landing phase pilot actions• Simplification: ignore time duration between actions

Analysis of Pilot Actions

Action ID

Time

descent touch down braking on runway

stage duration

Page 9: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

9

• Hidden states• Stages of aircraft operation• Examples: initial descent, touch down, braking, etc.

• Observations• Pilot actions• Example: initial descent – reduce throttle, lower flaps, lower landing gear, etc.

• Model parameters• Prob. distributions: Transition , observation , prior

• Drawback• Geometric state-duration distribution – encourages fast state switching• Inability to model arbitrary state durations

Hidden Markov Model (HMM)

Page 10: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

10

• Additional hidden variable• State duration• Forces hidden state to last time steps

• Model parameters• Probability distributions:

• Duration

• Transition

• Observation , initial distributions ,

Hidden Semi-Markov Model (HSMM)

Page 11: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

11

• Estimate probability distributions (conditional probability tables)• Duration• Transition• Observation• Initial distributions

• Use database of normal pilot action sequences

• Select parameters which maximize likelihood of data

• Non-convex problem without closed-form solution• Use Expectation Maximization (EM) [Dempster et al. ’77]

• Similar to Baum-Welch algorithm for HMM [Baum et al. ’70]

HSMM: Model Parameters Estimation

Page 12: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

12

• Detect if a test sequence is anomalous

• Entire sequence is anomalous (offline anomaly detection)• Normalized joint log likelihood

• Specific action is anomalous (online anomaly detection) • Conditional probability

• Computational complexity • Computation uses Junction Tree algorithm for inference

• Cost:• For comparison, complexity for HMM:

Anomaly Detection Methodology

- sequence length- number of hidden states- maximum state duration

Page 13: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

13

• Compared HMM and HSMM to detect duration anomalies• Data:

Results: Synthetic Data

Normal Anomalous

Training 200 0Testing 25 25

Page 14: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

14

Flight Simulator

• FlightGear flight simulator

• Simulator setup• Landing flight phase• Cessna 172 Skyhawk landing at Half Moon Bay, CA airport• Aircraft controlled using keyboard

• Keystrokes interpreted as pilot actions• 12 commands to control aircraft

Page 15: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

15

Pilot Actions for Landing

Page 16: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

16

Flight Simulator Data

• Generated Data

• Types of anomalies• 1 – Throttle kept constant, flaps are not lowered; rest is normal• 2 – No initial throttle increase; rest is normal• 3 – Flaps are not lowered; rest is normal• 4 – At the end of flight brakes are not applied; rest is normal• 5 – Pilot overshoots runway and lands behind it; rest is normal

Normal Anomalous

Type 1 2 3 4 5

# of sequences

110 10 10 10 10 10

Page 17: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

17

• AUC based on 11-fold cross-validation

• 110 normal sequences split into 11 parts• 10 parts used for training• 1 part + 50 anomalous sequences used for testing• Initialization was fixed across runs• Performance metric: area under ROC curve (AUC)

AUC Results: HSMM vs HMM

Page 18: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

18

• Dependency of AUC on initialization

• Selected 10 random model initializations• Split of dataset was fixed across runs

• Training: 100 normal sequences• Testing: 10 normal + 50 anomalous sequences

• Performance metric: area under ROC curve (AUC)

Effect of Initialization: HSMM vs HMM

Page 19: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

19

• Detected anomalies in streaming data1: No initial throttle increase2: Incorrect usage of rudder3: Mistakenly used elevator control after touch down

Offline vs Online Anomaly Detection

Anomaly scoring: Anomaly scoring:

Page 20: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

20

• Baseline methods [Chandola et al. ’08]

• EFSA: Extended Finite State Automata• t-STIDE: Threshold-based sequence time delay embedding• WIND: Window-based anomaly detection

• Setup• Training: 100 normal sequences; Testing: 60 sequences• Sliding window length (history length) was varied from 1 to 20• Selected model with best AUC on training set; Evaluated on test set

HSMM vs Other Methods

Type 1 Type 2 Type 3 Type 4 Type 5

HSMM 1.00 0.97 0.77 0.88 1.00HMM 0.87 0.60 0.71 0.95 0.99WIND 0.9 0.60 0.70 0.87 1.00EFSA 0.85 0.67 0.67 0.91 0.98

t-STIDE 0.84 0.67 0.68 0.92 1.00

Page 21: Detection of Precursors to Aviation Safety Incidents due to Human Factors I. Melnyk, P. Yadav, M. Steinbach, J. Srivastava, V. Kumar and A. Banerjee Department

21

• Proposed framework to model discrete pilot actions• HMM

• Hidden states – stages of aircraft operation• Observations – pilot actions• Drawback – inability to model arbitrary state durations

• HSMM• Introduces additional hidden variable to model state durations

• Evaluated model performance• Synthetic data• Flight simulator data• Compared HSMM to HMM and other anomaly detection algorithms

Summary

Thank you!