Human reliability in engineering systems

Presented by:Zaniar golabi & sina

masihabadi

Professor:Dr.jobin ghayoor

introuduction In 1958 williams recognized that human element

reliability must be included in overall system reliability. In 1960 proved that human error is the cause for %20 to

%50 of all equipment failure. In 1962,a database known as DATA STORE containing

time and human performance reliability estimates for human engineering design features was established.

In 1973,ieee transaction on reliability published a special issue on human reliability.

In 1980,a selective bibliography on human reliability was published covering the period from 1958 to 1978.

The first book on human reliability entitled human reliability with human factors appeared in 1986.

Terms and definitions Human error:this is the failure to perform a given task

that could resault in the disruption of scheduled operations or damage to property and equipment.

Human performance: this is a measure of human functions subject to specified conditions.

Continuous task: this is a task that involves some sort of tracking activity: for example, monitoring a changing situation.

Human reliability: this is the probability of accomplishing a task successfully by the human at any required stage in system operations within a stated minimum time.

Human error occurrence examples Human operator accounted for over %90 of the

documented air traffic control system errors. Over %50 of all technical medical equipment problems

are due to operator errors. Up to %90 of accidents both generally and in medical

devices are caused by human mistakes. A study of 23000 defects in the production of nuclear

components revealed that approximately %82 of the defects were due to human errors.

During the period from june 1,1973 to june 30,1975,401 human errors occurred in U.S commercial light water nuclear reactors.

Human error occurrence types:

Design errors: these types of errors are the result of inadequate design. for example the placement of controls and displays so far apart that an operator is unable to use them in an effective manner.

Operator errors :these errors are the result of operator mistakes and the conditions that lead to operator errors include lack of proper procedures ,complex tax ,poor training and operator carelessness.

Assembly errors: these errors occur during product assembly due to humans.

Inspection errors: these errors occur because of less than %100 accuracy of inspectors.

Maintenance errors: the errors occur in the field due to oversights by the maintenance personnel.

Installations errors: these errors occur due to various reasons including using the wrong installation related blueprints or instructions.

Handling errors: these errors basically occur because of inadequate storage or transportation facilities.

Causes for the occurrence of human errors: Poor motivation of involved personnel. Poor training or skill of concerned personnel. Poor equipment design. Inadequate or poorly written equipment operating and

maintenance procedures. Poor job environment: poor lighting ,high/low

temperature ,high noise level ,crowded work space, ,etc.

Inadequate work tools. Complex tasks. Poor work layout

Human error occurrence classification:

decision error: occur when the wrong decision is made after considering the situation.

Action error: are the result of no action ,incorrect action ,or the performance of correct action on the wrong object when required.

Transmission error: occur when information that must be passed or to others is not send ,sent incorrectly ,or sent to the wrong destination.

Checking error: occur when system require checks,the incorrect checks are made ,checks are omitted ,or correct checks are made on the wrong object.

Diagnostic errors :are the result of misinterpreting the actual situation when an abnormal event occurs.

Retrieval errors :occur when required information either from an individual,an individual memory,or from any other reference source is not received or the incorrect information is recieved.

Factors that influence on human performance: Reaction to stress Time at work Fatigue Group interaction and identification Social pressure Repetitive work Supervisor’s expectations Morale Social interaction Crew efficiency Idle time

Stress factors and operator stress characteristics Dissatisfied with the current job. Faced with serious financial problems. Working with individuals having unpredictable

temperaments. Low chance of promotion from current position. facing a possible work layoff. Current job below ability and experience. Conducting tasks under extremely tight time schedules. Having health problems. Excessive demands of superior. Inadequate expertise to conduct the tasks required in the

current job. Often tasks work home to meet deadlines. Having difficulties with spouse/children.

The operator stress characteristics include: short decision making time Several displays difficult to discriminate. Requirement to perform steps at high speed Poor feedback for the determination of accuracy of

actions taken. Requirement for prolonged monitoring. Very long sequence of steps required to perform a task. Requirement to make decisions on the basis of data

obtain from various different sources. Requirement to operate at high speed more than one

control simultaneously.

Human performance reliability in continuous time and mean time to human error (mtthe) measure

As humans perform time continuous tasks such as aircraft maneuvering ,scope monitoring ,and missile countdown, a general expression for human performance reliability can be developed the same way as for the development of the general classical reliability functions.

Human reliability evaluation methods

Probability tree method: this is used to perform task analysis by diagrammatically representing critical human actions and other events associated with the system.

Fault tree method: this special case has been discussed in the class.

Markov method: this is a powerful reliability engineering tool that can also be used to perform time continuous human reliability analysis.

Human reliability markov modeling Reliability analysis of a system with human

error: this mathematical model represent a system which can fail either due to hardware failure or to a human error.

Reliability analysis of a human performing a time continuous task under fluctuating environment: this mathematical method represents a human performing a time continuous task under fluctuating environment.

Human error data Human reliability predictions are only as effective as the

body of quantitative on which they are made.

The collection and maintenance of human error data are at least as important as making various types of human reliability predictions.

Means to collecting human error data Experimental studies.

Expert judgments.

Self made error reports.

Human data recorder.

Automatic data recorder.

Published literature.

Human error data Specific human error data banks and

sources

Human error data for selective tasks

Brief description of selected human eror data banks

Specific human error data banks and sources Data store Operational performance recording and evaluation data

system Nuclear plant reliability data system Aerojet general method Aviation safety reporting system Bunker ramo tables Safety related operator action program Technique for establishing personnel performance

standards

Brief description of selected human error data banks Data store: the data store was established in 1962 by

american institute for research, pittsburgh and it contains estimates for time and human performance reliability.

Aviation safety reporting system: originally this databank was developed by the national aeronautics and space administration (NASA) and contains information on civil aircraft accidents.

Operational performance recording and evaluation data system: this system was developed to collect data on operational human performance by the U.S navy electronics laboratory, san diego.

Learning from other fields: PSTN FCC-collected data on outages in the US public-switched

telephone network metric: breakdown of customer calls blocked by system

outages (excluding natural disasters). Jan-June 2001

9%

47%17%

5%

22%

Human-co.Human-ext.Hardware FailureSoftware FailureOverloadVandalism

Human error accounts for 56%56% of all blocked calls

Learning from other fields: PSTN

PSTN trends: 1992-1994 vs. 2001

Cause Trend 1992-94 2001Human error:

company 98 176

Human error: external 100 75

Hardware 49 49

Software 15 12

Overload 314 60

Vandalism 5 3

Minutes Minutes (millions of customer minutes/month)

MARKOV METHOD Assumptions: The probability of a transition occurrence from one state

to another in finite time Δt is λΔt .The parameter λ in our case is the constant human error rate.

The probability of two or more transitional occurrence in Δt is negligible.

All occurrences are independent of each other.

MARKOV METHODSystem has two states.The she state 0 represent human performing the time continuous task normally and state 1 denotes human committed error.

By using the MARKOV METHOD we can obtain the probabilites of

the system being in state 1 or 0 at time t for known human error

rate.P0(t+ Δt)=P0 (t)(1-λ Δt)

P1(t+ Δt)=P1 (t)(λ Δt)+P1(t)

MARKOV METHOD Rearranging and taking limits: dp0(t)/d(t)=- λ p0(t)

dp1(t)/d(t)= λ p0(t) At time t=0: p0(0)=1 and p1(0)=0 so: p0(t)=exp(-λ t) p1(t)=1- exp(-λ t) Thus the human reliability R(t) is given by: R(t)=p0(t)=exp(-λ t)

MARKOV METHOD By integration over the time interval [0,∞]:

MTTHE=∫R(t) dt= 1/λ

HUMAN RELIABILITY MARKOV MODELING: This mathematical model represents a system which can be

failed due to a hardware failure or a human error. Both failure rates are constant. We have the following symbols:

λh:the constant human error rate λnh:the constant nonhuman error rate Pi(t): the probability that system is in state i at time t i=0 system operating normally i=1 system failed due to nonhuman error i=2 system failed due to human error

HUMAN RELIABILITY MARKOV MODELING: By applying the Markov method: dp0(t)/d(t)+ (λh+ λnh)p0(t)=0 dp1(t)/d(t)- λnh p0(t)=0 dp2(t)/d(t)- λh p0(t)=0 Solving the above equations we have:

P0(t)=exp(-(λh+λnh)t) P1(t)= λnh/(λnh + λh)[1-exp(-(λh+ λnh)t)] P2(t)= λh/(λnh + λh)[1-exp(-(λh+ λnh)t)] The system reliability R(t) with human error is: R(t)=p0(t)=exp(-(λh+λnh)t)

HUMAN RELIABILITY MARKOV MODELING By integrating over the time interval [0,∞]:

MTTF= ∫R(t)dt=1/λh

Refrences: Dhillon,B.S,design reliability,fundamentals and

applications,CRC press,1999

J. Reason, Human Error, Cambridge, 1990

www.eliminatemistakes.com

Human error in medicine,marc green ,2003

Thanks for your attention