Yulin’s research introduction

1

A Human Factors-Based Safety Management Model for Aviation Maintenance Safety

Yu-Lin Hsiao, PhD

Department of Industrial and Systems Engineering

2

How do we manage safety?

• The way we currently do– Reactive way: Accident & Incident Investigation

– Proactive way

• Daily- or Periodic-Based: Audit, LOSA, FOQA

• Behavior-Oriented: ASAP, LOSA

• Event- or Consequence-Based: MEDA, Risk Matrix

• Each system has its own advantages, but also probably goes its own and unique way

3

How do we manage safety?• If there is a integrated method to manage safety from

a macro-system viewpoint?

– According to the purpose and philosophy of the ICAO Safety Management System (SMS) as well

• Can we integrate all these safety programs or methods by using the same language?

– Based on the same common concepts

– Quantitative and Data-driven Method

4

How do we manage safety?• Furthermore, can we use these safety data to

evaluate the safety status and manage the risk?

– To assist upper-level management and decision-making

– For long-term and continued safety management

– But not just focus on single event analysis or case by case

5

Answer: Human Factors (HF)

• It is the major cause of flight accidents– Implicated in most accidents and incidents

• Most safety programs are related to human factors in some way

• Connected with current risk and safety management concepts– To eliminate or mitigate specific human error

6

Steps to establish the HF model

• 1. Data Transformation– From Qualitative Documents

• audit, investigation, or voluntary reports

– To Quantitative Data

• Human Factors Rates

• Use the Human Factors Analysis and Classification System – Maintenance Audit (HFACS-MA)

– Set up an internal review board

• To analyze data sourced from various systems

• Extract consensus results to calculate the quantitative rates

7

UnsafeSupervision

SupervisionDysfunction

SupervisionDisobedience

PlanningOrganizing

ControllingCorrecting

Routine ExceptionalLeading

Coordinating

OrganizationalInfluences

OrganizationalFunctionality

OrganizationalSafety Climate

OperationsProcedure

ExecutionSafetyCulture

ResourceManagement

SafetyPolicies

SafetyOversight

8

UnsafeActs

Errors Disobedience

Skill-basedErrors

DecisionErrors

Routine Exceptional

PreconditionsFor

Unsafe Acts

Conditions ofOperators

Conditions ofTask / Environment

AdverseStates

TeamworkTask

DemandsHardware &

SoftwareLimitations

PhysicalEnvironment

9

Human Factors Rates• Each error type was accumulated for monthly period

• Multiply by the weights of severity degree– Weights were developed by aviation authority– Based on Analytic Hierarchy Process (AHP) method

• Human Factors Rate =

W : The highest weight of severity degree (W = 11, the designated weight of Finding ) wi : The weight of the severity degree, i={I, R, C, F} n : The sum of the human failures with all severity degree per month ni : The sum of the human failures with specific severity degree, i={I, R, C, F}

nW

nwi

ii

10

Future Incident Rate

Incident Rate =

Incident : The number of incidents per month

Departure: The number of flight departure per month

000,1Departure

Incident

Airline DepartureTimes

Accident Incident AccidentRate

Incident Rate

A 81,448 1 73 0.012 0.90

B 134,814 2 192 0.015 1.42

11

Steps to establish the HF model

• 2. Develop the mathematical model– Use Neural Network method

– Verify the prediction performance of the model

• 3. Start using the model to evaluate safety status– Output: Future Incident Rate

– Different company might have different prediction performance or time range (month or quarter)

12

Steps to establish the HF model

• 4. Detect the uprising trend of future incident rate – Self- or Expert-decided warning threshold

• 5. Find out the root causes and the original data sources– Based on real and reliable data collected by different

safety systems or programs

– To support the corresponding safety management activities

13

• Current Achievement: – Succeed in the prediction of future incident rate using

human factors analysis

– Based on real safety audit data from aviation authority (data-driven & practical)

– Prove the causality of human error and safety

– General Accuracy

• Correlation Coefficient: 0.6 ~ 0.75

• R2: 0.35 ~ 0.58

HF-Based Safety Management Model

14

Prediction Result

0

2

4

6

8

10

Months

Inc

ide

nt

Ra

tes

Actual Simulate

Airline A

Airline B

60

0

1

2

3

4

5

Month

Inc

ide

nt

Ra

tes

Actual Simulate

15

• Advantages– Can integrate various data sources from different safety

programs using the same standard

• Human Factors concept

• As an integrated part of SMS

– Can become a safety management tool to detect risk associated with uprising incident rate from a systematic perspective

• Find out the root causes related to incident rate

• Conduct corresponding management activities to control the risk

HF-Based Safety Management Model

16

HF-Based Safety Management Model

• Improvement to safety management system– Help manager’s decision making regarding the safety

management priority

– Integrate and utilize various safety data to improve safety management in a quantitative way

– Focus on both active and latent human factors such as safety climate which could affect the safety performance

16