Intro To Reliability Fundamentals IEEE Rev2

Introduction to Reliability Fundamentals
Donald G. Dunn – Principal Consultant
© 2019 D2 Training
Presenter
• Donald Dunn
• W.S. Nelson
• Senior Consultant
• Long Time IEEE Houston Section Officer
• Numerous other leadership positions in IEEEE including sections, regions,
institute, societies and standards
• Numerous leadership positions in ISA including institute and standards
• Leadership positions or participation in API, IEC, NFPA and PIP
© 2019 D2 Training
Reliability Fundamentals Overview
▫ Why Reliability ▫ Inspection Interval Data
▫ The Problem and Course Data
▫ Reliability = Safety ▫ WEIBAYES
▫ The Process from 50K ▫ Crow – AMSAA Reliability Growth
▫ Weibull History Model
▫ Reliability Definitions ▫ Data Set Issues
▫ Weibull Analysis ▫ Example Analysis
▫ Dirty Data ▫ SAP/Meridium Methods
▫ Data Inconsistencies How to Rearrange a Historical Multiyear
Work Order (SAP/Meridium)
© 2019 D2 Training
Why Should I be Concerned with Reliability?
4 © 2019 D2 Training
Filxborough - 1974
© 2019 D2 Training
Filxborough - 1974
© 2019 D2 Training
Bhopal - 1984
© 2019 D2 Training
Piper Alpha - 1988
© 2019 D2 Training
Henderson, Nevada – 1988
© 2019 D2 Training
Milford Haven, UK – 1994
© 2019 D2 Training
11
Milford Haven, UK – 1994
© 2019 D2 Training
Toulouse, France - 2001
© 2019 D2 Training
13
Toulouse, France - 2001
© 2019 D2 Training
Texas City - 2005
© 2019 D2 Training
Texas City - 2005
© 2019 D2 Training
Texas City - 2005
© 2019 D2 Training
Puerto Rico - 2009
© 2019 D2 Training
Puerto Rico - 2009
© 2019 D2 Training
Puerto Rico - 2009
© 2019 D2 Training
Torrance, CA - 2015
© 2019 D2 Training
Crosby, TX - 2017
© 2019 D2 Training
Superior, WI - 2018
© 2019 D2 Training
The Problem
The “Swiss Cheese” Theory of Risk
© 2019 D2 Training
Accident Trajectory
Major Loss
© 2019 D2 Training
The Problem
• Management measured by KPI’s
▫ One numerical indicators budgets
▫ Maintenance, engineering and reliability programs
• Invoke budget cuts by arbitrary percentages
▫ Without due consideration to the effects
▫ Without reflecting on impact of cost cutting
• Significant Compensation Component
▫ Bonus & Stock Options
• Growing Experience/Knowledge GAP
• Management never directly request cuts on safety!
• Unfortunate few grasps reliability tied safety
© 2019 D2 Training
The Problem
• Few accidents occur when things run smoothly and the
equipment has a high level of reliability.
• Accidents occur in reactive maintenance realm
▫ Where assets fail unexpectedly
▫ There is a rush to get the process running
• Manufacturing plants are asset-intensive
▫ Reliability increases so do safety.
▫ When assets become more reliable
▫ End-users service them less
▫ Workforce is exposed to hazards less
© 2019 D2 Training
Reliability = Safety
• Three maintenance environments
• Reactive Maintenance - all behavior is reactive

▫ Reliability is nonexistent
▫ No planned maintenance or failure mode activities
▫ All activities are reactive or breakdown maintenance
▫ Safety is instinctive
© 2019 D2 Training
• Planned Maintenance - behavior is guided through the planning and

scheduling process
▫ Safety procedures are embedded in maintenance plan and thus enforced
▫ Reliability is scheduled, scheduled maintenance activities can support
reliability initiatives.
▫ Focusing maintenance activities on failure modes
© 2019 D2 Training
• Pro-active Maintenance - behaviors are a self-disciplined
▫ Focus on continuous reliability improvement
▫ Embedded safety all aspects maintenance/production
▫ Maintenance focus on how they can be improved
▫ Strong focus on condition monitoring
Only intervening when failure mode trends up
▫ Reliability & Safety are embedded & pro-active
Second nature where responsibility is everyone’s role
Viewed value add to the business
© 2019 D2 Training
• Studies on Maintenance-related major accidents
▫ 30-40 percent determined maintenance was a factor
Includes preparation and performance of maintenance
Contains incidents caused by lack of proper maintenance
▫ 76% occurred during the maintenance activity
▫ 24% occurred during site preparation or transition to or from production
activities
▫ Maintenance activities expose workers to higher risks
Studies by Okoh & Haugen, “Maintenance-related major accidents: Classification of causes and case study”
© 2019 D2 Training
Three Legs to the Stool
Safety
Reliability Operating Excellence

(OE)
© 2019 D2 Training
The Process from 50K
Reliability Management Program
Excellence in Asset Management Drives Competitive Edge
Faster
Financial Market
Advantage Advantage
Be
pe
tte
ea
Ch
r
Business
Value
Technology & Community
Employment Advantage
Advantage
r
Cl
e
rt
an e
a
Sm
r e
Safer
© 2019 D2 Training
Overall Equipment Effectiveness (OEE)
Scheduled work
Weekends/Holidays
• Uptime Planned Breaks/Lunches
A = Total Available Time

• Throughput Losses Meetings
Housekeeping
• First-pass Yield
Key Manufacturing Losses

Equipment breakdown
Lack of materials
Operational Absenteeism
B = Scheduled Time
Losses Set-up time
Changeover time
Product lost that could

Speed have been sold at the
Losses
correct run rate.
C = Uptime
Quality Product that did not
Losses meet first pass quality
criteria.
D = Good
Good product made on
Product
Good
Product time and right the first
Manufactured time.
© 2019 D2 Training
Maintenance Inefficiencies:
North American Average Wrench Time = 25%
© 2019 D2 Training
World Class -- Key Performance Indicators
Manufacturing: Maintenance:
OEE/Asset Utilization 85/95% Mtce Costs, % ARV* <1-3%
On-time Deliveries 99% Unplanned Downtime Loss <1-2%
Customer Complaints 0.01% Planned Maintenance >90%
Process Quality - Cpk >2 Reactive Maintenance <10%
OSHA Injuries per 200k hrs $ARV per Mtce Tech >$6-8M
Recordables <0.5 % Maintenance Rework <1%
Lost Time <0.05
Stores:
Human Resources: Parts Stock out Rate <1%
Overtime <5% Parts Inventory Turns >2
Skills Training ($/yr) $2-3K Value as a % of ARV <0.5%
(hrs/yr) >40 *ARV= Plant Asset Replacement Value
How does your plant compare to these?

© 2019 D2 Training
Metrics – Leading & Lagging
Behavioral Metrics
Results Metrics
(Recommended Target)
• % Labor-Hours PM (15%) • Asset Health % (+90)
• PM Coverage % (+95%) • Overall Equipment Effectiveness (OEE)
• PM Route Compliance (+95%) • Maintenance Availability
• PM Route Adherence (+90%) • Wrench Time 45-55%
• Mean Time To Repair (MTTR)
• % Labor-Hours PdM (15%)
• MTBF
• Mean Time to Implement (<45 days)
• Reliability %
• PdM Coverage % (Top 2nd Quartile)
• Maintenance Costs/Unit Shipped
• PdM Route Compliance (+95%)
• Labor Utilization (+95%)
• Ready Backlog (>2 Crew Weeks)
• Total Backlog (<6 Crew Weeks)
• Schedule % (100%)
• Schedule Compliance % (+95%)
• Proactive Work % (40%)
• Planned Work % (+90%)
• Overdue Work Order % (<10%)
• % Rework (<5%)
© 2019 D2 Training
Criticality
• Purpose
▫ Develop relative ranking of equipment for use during:
Equipment Maintenance Plan (EMP) creation
Input for corrective work planning and execution
Prioritization of assets to review/create bill of materials (BOMs)
Input for critical spare identification yielding data for stock versus non-stock
parts and service level agreements (SLAs)
© 2019 D2 Training
Criticality
• Ranking items to determine which items get preferential treatment with
respect to resource allocation
• Items are generally assets and the resources are labor, materials and schedule
priority
• Criticality analysis be performed at:
▫ System
▫ Asset
▫ Failure Mode
• Level of effort increase, with the system level requiring least amount of time.
© 2019 D2 Training
Criticality Analysis – Biggest Opportunity
• Opportunities to reduce criticality • When to periodically review?
▫ Install redundant asset ▫ New Equipment
▫ Complete BoMs ▫ About every 3 years
▫ Adjust part plan or establish SLAs ▫ Change in Operating Context
Change
▫ Change in failure probability
Successful RCA and RCM projects
© 2019 D2 Training
Defining Asset Reliability
Asset Reliability – the probability that an
asset will perform without failure within a
given period of time, under known conditions.
Asset Performance – the state of asset

operation that adds value to the organization.
Asset Accountability – the responsibility of

an organization for asset performance.
Source: ISO PC 251 – ISO 55000 Asset Management

© 2019 D2 Training
What Causes the Defects?
23% • Operating above/below design parameters
• Improper operating procedures
Production • Operator errors
22% • Undefined/unclear specifications

• Improper design
Engineering • Undocumented changes
17% • Poorly designed maintenance strategies

• Improper maintenance procedures
Maintenance • Craft errors
12% • Providing out-of-specification materials

• Improper storage and handling procedures
Procurement • Material degradation in Stores
15% as a result of Sales/Customer expectations exceeding Production capabilities.

11% as a result of current Management philosophies (i.e. postponing capital replacement).
Source: A.T. Kearney study of U.S. Manufacturing. Originally published by Newsweek in 1999.
Corroborated by studies performed by Keith Mobley and the Plant Performance Group.
© 2019 D2 Training
Failure Patterns and PM Tasks
Equipment Maintenance Strategy:

• 89% of failures are random in nature
• 11% are a function of time
Source: John Moubray, Nowlan & Heap
© 2019 D2 Training
What is a Failure Mode? Why Important?
• As defined by Stan Nowlan and Howard Heap, a Failure Mode is the specific
manner of failure; the circumstances or sequence of events which leads to
functional failure.
• Understanding failure modes is the only way to design a complete maintenance
strategy that will ensure the inherent designed reliability of your equipment.
• A failure modes based strategy is focused on dealing with causes instead of
effects!
© 2019 D2 Training
Failure Modes Analysis
• Understanding the failure modes present in each piece of equipment, and how
to identify those failure modes, should dictate your Design, PM and PdM
activities
• If a task is being performed that:
▫ Doesn’t have the ability to identify a failure mode
▫ Isn’t a regulatory or statutory requirement
▫ Doesn’t prevent a failure mode
• Then…we are simply wasting our valuable craft time!
© 2019 D2 Training
Failure Mode - Definition
• Part
▫ Bearing
▫ Breaker
▫ Shaft
• Defect
▫ Fatigued
• Reason
▫ Misalignment
© 2019 D2 Training
Defect
• Not just the Mode
▫ Worn
▫ Not descriptive enough
• Has to include Failure Mechanism

▫ When the mechanism is descriptive
▫ Knowing the mechanism might change the countermeasure (task)
▫ Example: Adhesive Wear versus Abrasive Wear
© 2019 D2 Training
Reason
• Describes the action or condition that was physically responsible for the
problem
▫ Example: Improper Installation
• NOT another failure

▫ Cannot reference another failure mode as the reason why this failure mode is
present
▫ These are called “Secondary Effects”
▫ Example: Bearing – Abrasive Wear – Seal Failed
© 2019 D2 Training
Reliability Improvement Philosophy
I P Low
Failure
Initiated
Ultrasonic Energy
Detected
Vibration Analysis
Fault Detection
P-F Curve
Oil Analysis
Detected
Hi
Audible Noise
gh
Equipment Condition
Hot to Touch
Mechanically
Loose
Ancillary
Critical
Damage
Catastrophic
PRECISION PREDICTIVE PREVENTIVE RUN TO FAILURE Failure
Time F
© 2019 D2 Training
Reliability Waste Cause and Effect
© 2019 D2 Training
The “Reliability” Journey
© 2019 D2 Training
Reliability Systems Model
5 – OPERATIONAL
EXCELLENCE
4 – ENGINEERED
RELIABILITY
3 – ORGANIZATIONAL
ENGAGEMENT
2 – PROACTIVE
ENABLERS
1 – GOVERNANCE
© 2019 D2 Training
Weibull History
Weibull History
• WaloddiWeibull - 1887 - 1979

▫ Professor at Royal Institute of Technology - England
▫ Weibull distribution first introduced by W. Weibull in 1937
When he was studying the issue of structural strength and life data analysis
▫ Hallmark paper in 1951 described a probability distribution function that we now
call Weibull distribution
▫ Much of Weibull’s work involved life prediction of ball bearings
• In probability theory and statistics, Weibull distribution is one of most important
continuous probability distributions
• Data could select the distribution and fit the parameters
• Weibull analysis is leading method for fitting life data
© 2019 D2 Training
Weibull History
• Dr. Bob Abernathy

▫ Co-Patent holder of Pratt Whitney’s SR71 “Blackbird” Jet Engine
▫ Author of the ASAF Weibull Analysis Handbook - 1981
▫ Author of the New Weibull Handbook – 1993, 1996, 1998, 2000
▫ Main proponent of Weibull Analysis in areas outside of aviation / aerospace industries
▫ Author of “The New Weibull Handbook” (used to develop this course)
© 2019 D2 Training
Reliability Definitions
(1)
• Availability: A measure of the degree to which an item is in an operable and committable
state at the start of the mission when the mission is called at an unknown and random time.
© 2019 D2 Training
(2)
• Achieved Availability (AA): includes the measure of preventative and corrective maintenance. It
is defined as:
© 2019 D2 Training
(3)
• Inherent Availability (AI): Availability in the ideal state. It is defined as:
• Operational Availability (AO): Operational availability includes the time for corrective action
and response to system faults. It is defined as:
© 2019 D2 Training
(4)
• Average Effectiveness Level: A figure of merit for quantifying system effectiveness. It is defined
as:
• Benign failure: A failure which does not affect performance of the system and goes unnoticed
as a problem until the system receives an in-depth inspection.
• Burn-in: An accelerated test under higher than normal stress to screen-out infantile mortality
problems in a short time interval and thus stabilize the failure pattern under normal stress
conditions.
• Capability: A quantification of an item’s ability to achieve mission objectives considering the
conditions during mission.
• Censored data: Data which has been “cur-out” of the life analysis for a variety of reasons. See
suspended data for more details.
© 2019 D2 Training
(5)
• Cost effectiveness: Spending small amounts of money to prevent expenditure of large amounts
with due consideration to the time value of money.
▫ Small amounts are usually spent on improvements in the design, production, or maintenance of a system to
avoid larger expenditures resulting from short life, high maintenance costs, human injuries, replacement costs,
etc.
• Dependability: A quantification of the degree to which an item is operable and capable of
performing its required function at any random time during a specified mission profile based
on its availability at the start of the mission. It is defined as:
• Derating: Upgrading component reliability by applying stresses below rated values to gain an
advantage in life expectancy for improving reliability of a component or system.
© 2019 D2 Training
(6)
• Duane Plot: A learning curve type of graph showing the rate at which the cumulative MTTF or
the cumulative failure rate can be expected to change over time when plotted on log-log paper.
• Durability: A measurement of useful life which equals the probability that an item will survive
its projected service life, overhaul point, or rebuild point without a durability failure.
• Effectiveness: The potential or actual probability of a system to perform a mission for a given
level of performance under specified operating conditions. It is defined as:
© 2019 D2 Training
(7)
• Exponential failure law: The probability of survival of equipment showing only a chance
failure rate such that the die-off is the same always for any size of surviving unfailed population.
)
It is defined by the reliability function: !(#) = & '(*+ = & '(-))
Where: R = exponential reliability
e = 2.718281828… or Naperian logarithm
t = Time of usage by which reliability is measure in cycles, time, etc.
. = Mean life which is measured in the same units as time, t
/ = Failure rate in some specified time interval = 1/.
• Failure: An event which renders equipment as non-useful for the intended or specified
purpose during a designated time interval.
▫ The failure can be sudden, partial, one-shot, intermittent, gradual, complete, or catastrophic. The degree of
failure can be degradation or gradual, sudden or one-shot, from weakness, from imperfections, from misuse
and so forth.
• Failure mechanism: A variety of physical processes which results in failure from chemical,
electrical, thermal, or other mechanisms. © 2019 D Training 2
(8)
• Failure mode: The result of a mechanism caused by a type of imperfection which contributes
to failure.
▫ In general, the three major failure modes are infantile mortality, chance failures, and old age wear-out. In
specific, the failure mode may be due to open circuit, short circuit, fatigue, wear, and so forth.
• Failure mode and effect analysis (FMEA): The study of potential failures that might occur in any
part of a system to determine the probable effect of each on all other parts of the system and
on probable operational success.
• Failure mode and effect and criticality analysis (FMECA): A widely used and effective design
reliability analysis considers from the bottom up each potential mode of failure that might
occur of every component of a system and determines the probable effects on system
operation of each failure mode in turn on probable operational success and the results of which
are ranked in order of seriousness.
▫ FMECA can be performed from different viewpoints such as safety, mission success, availability, repair cost,
failure mode, reliability reputation , and so forth.
© 2019 D2 Training
(9)
• Fault tree analysis (FTA): Top down study of sequence of events for failure of a system using
diagrammatic method of symbols and considering multiple failures as a matter of course.
• Failure rate: While operating under stated conditions it is the total number of failures of an
item within a population divided by the total number of life units expended by that population
during a particular measurement interval.
• Hazard: A set of existing conditions with the potential for initiating an accident sequence in the
operation of a product or system.
▫ Hazards may be classified as catastrophic, critical, controlled/marginal, or negligible.
• Hazard function: It is a measure of proneness to failure and a descriptor of the force of
mortality.
▫ The hazard rate or hazard function is the conditional probability of failure in the interval x to (x + dx) given
that no failures occurred prior to time interval X. Hazard functions are a general approach to derivation of
time-to-failure distributions.
© 2019 D2 Training
(10)
• Hazard rate: The number of units of failures per unit of time.
▫ It is the failure rate of the survivors to time t in the very next instant following t. Hazard rate failure rate, and
instantaneous failure rate are equivalent terms. Hazard rate, h(t), is defined as follows:
© 2019 D2 Training
(11)
• Infant mortality: Failures/deaths at an early age. Infant morality is often due to inadequate
burn-in or stress screening; production problems, mis-assembly, quality control difficulties;
overhaul problems; solid state electronic failures; inadequate test specifications, components,
engineering; inadequate materials; improper handling or packaging; or subsystem interactions.
▫ Overhauling components surviving from a population with infant mortality problems is not appropriate
because very old parts are better than new (unless new parts infant mortality has been fixed).
• Life units: A measure of the duration of use applicable to the item including events such as
operating hours, cycles, distance, rounds fired, attempts to operate, starts/stops,
takeoffs/landings, shelf/storage time.
▫ Time at high temperature, time at high stress, and other counting criteria used for measuring life.
© 2019 D2 Training
(12)
• Log-normal distribution: A mathematical probability distribution with long tails to the right
for modeling certain types of life data such as metal fatigue and electrical insulation life, and
for describing distributions of repair times of equipment. It is related to the normal
distribution when logs are taken of the data.
▫ The descriptors in the equations often cause much confusion and should not be confused with the normal distribution. For
example, for the log-normal case, m is the mean of the log of life, not of life and it is called the log mean. The parameter s
is the standard deviation of the log of life, not life and it is called the log standard deviation. Note also the logarithms can be
base 10 for log or the natural base e for ln. The log (or ln) mean m determines the median point and thus the spread of the
distribution. The log (or ln) standard deviation s determines the shape of distribution. Both the log mean and the log
standard deviation are pure dimensionless numbers and do not have the same units as the life data.
▫ The log-normal model is usually too optimistic when projecting results to small percentiles as compared to a projection
based on a Weibull distribution: log normal and Weibull are competing failure distributions. The precise model that leads to
a log-normal distribution is called a multiplicative (or proportional) growth model. The model undergoes a random
increase of degradation that is proportional to its present state and thus the multiplicative effect of all the random effects
growth and thus builds up to failure.
© 2019 D2 Training
(13)
• Maintainability: The measure of the ability of an item to be retained in or restored to specified
condition when maintenance is performed by personnel having specified skill levels, using
prescribed procedures and resources, at each prescribed level of maintenance and repair.
• Maintenance action: An element of a maintenance event. One or more tasks (i.e., determine
the approximate location of a fault and the extent necessary to effect repair, servicing and
inspection) necessary to retain an item or restore it to a specified condition.
• Maintenance, corrective/repair: All actions performed as a result of failure including restoring
an item to a specified condition.
▫ Corrective/repair maintenance can include any or all of the following steps: determining the approximate
location of a fault and the extent necessary to effect repair, disassembly, removing items to be replaced and
installing replacement items, reassembly, alignment, and checkout.
© 2019 D2 Training
(14)
• Maintenance event: One or more maintenance actions required to effect corrective and
preventative maintenance due to any type of failure or malfunction, false alarm, or scheduled
maintenance plan.
• Maintenance, preventive: All actions performed in an attempt to retain an item in specified
condition by providing systemic inspection, detection, and prevention of incipient failures.
• Maintenance, scheduled: Preventive maintenance performed at prescribed points in the item’s
life.
• Mean time between failures (MTBF for repairable equipment): A basic measure of
reliability for repairable items during which the parts perform within their specified limits
under stated conditions. Mean time between failures is often used as a synonym for MTTF. It is
defined as:
© 2019 D2 Training
(15)
• Mean time between maintenance (MTBM): A measure of system effectiveness and reliability
taking into account maintenance policy. It is defined as:
• Mean time between maintenance actions (MTBMA): A measure of system reliability related to
demand for maintenance manpower. It is defined as:
• Mean time to failure (MTTF for non-repairable equipment): A basic measure of

reliability for non-repairable items during which the parts perform within their specified limits
under stated conditions. Mean time between failures is often used as a synonym for MTBF. It is
defined as:
© 2019 D2 Training
(16)
• Mean time to repair (MTTR): A measure of system maintainability equal to the average
system repair time, and this value is the reciprocal of repair rate in the exponential case. It is
defined as:
• Mission: An event which is often considered a single cycle, depending on how it is specified,
and which is the ultimate output of a system for which the design objective is intended to
accomplish.
• Mission profile: A time-phased description from start to finish for all usage and operation
cycles required by the system to perform throughout the life cycle for which reliability is
specified.
▫ It includes: a criteria to judge success or failure, modes of the task or missions, operation requirements, task
length, system environment for helping access operational reliability. Tasks for the mission profile must
describe in detail how the system complexity is described such as: multifunctional, single function,
continuous, cyclic, or a one-time function.
© 2019 D2 Training
(17)
• Probability: The likelihood of an occurrence on a scale from 0 (zero change for an occurrence)
to 1 (100% certainty for an occurrence) attached to a random event based on a particular
mode for which the event can occur.
• Probability paper: Special graph paper with divisions on one scale apportioned to a specific
statistical distribution so that life units from the distribution will plot as a straight line.
• Random failure: A failure whose cause and/or mechanism make the exact time of failure
unpredictable but the failure may be anticipated in a probabilistic or statistical sense.
▫ Random failures are independent of time and this failures show lack of memory. Causes of random failures are:
insufficient/inadequate design margins, misapplication/overstress, part used in wrong environment,
predictable failure levels from a statistical chance occurrence basis, non-pattern failures, maintenance/human
errors, and failures due to nature/foreign objects. Surviving parts from a population suffering from random
failures cannot be corrected by overhauls since an old part is as good as a new part (unless the new one has
been fixed from the assignable cause for random failures).
© 2019 D2 Training
(18)
• Redundancy: The existence of more than one way for performing a given function whereby a
provision is made for more than one element to share a load in order to improve life
performance.
• Redundancy, active: Redundancy where all means for performing a given function are
operating simultaneously.
• Redundancy, standby: Redundancy where alternative means for performing a given function
are inoperative until needed.
• Reliability:
▫ As a general sense, reliability is the ability of an item to perform a required function under stated
conditions for a stated period of time.
▫ As a characteristic, reliability denotes the probability of success or the success ratio.
▫ As a measure of quality, reliability exists by design as an objective or a requirement of a product
from its inception to the end of its working life.
© 2019 D2 Training
(19)
• Reliability (continued):
▫ As a probabilistic statement: reliability is concerned with the probability of future events based on
past observations.
▫ As a concept, reliability is a special development of engineering industries for the collective name of
measures of quality that reflect the effect of time in storage or time in use of a product as distinct
from measures that show the state of the product at time of delivery. Some industries are concerned
about the time aspect of product quality and the reliability concept is expressed in other terms.
▫ As a basic concept, durable and high probability of failure-free performance under stated conditions
including all item life units, not just mission time and all failure with the item, not just mission-
critical failures at the time level of assembly.
• Reliability block diagram: A method of using reliability analysis based on functional black box
diagrams to understand the reliability relationship of components in a system by which the
black box are connected in a manner not unsimilar to analyzing electrical circuits.
© 2019 D2 Training
(20)
• Reliability engineering: The application of appropriate engineering disciplines, techniques,
skills, and improvements to achieve the required reliability, maintainability, serviceability
exchangeability, and availability of both products and processes at a cost that is affordable for
the business.
• Reliability growth: Improvement in reliability caused by successful learning experiences
during correction of faults or deficiencies in the design, manufacture, sales, and service of a
product or component.
• Repair: see definition of maintenance, corrective.
• Series system: A system or subsystem with components arranged whereby the failure of one
item results in the failure of the system.
• Series parallel system: A system with active redundancy which uses redundant elements in
series along with units in series for enhanced reliability.
© 2019 D2 Training
(21)
• Standby system: A system with components arranged in parallel and usually connected with a
switch to detect failure in the main component so that the standby unit can be quickly
switched into service to function as active redundancy for improving reliability.
▫ The standby element does not undergo degradation or consume power as if it were operating in “spinning
reserve”.
• Suspension: A test or operational unit that has not failed at the time of the life analysis. Testing
on the un-failed unit may been stopped for practical or economical reasons.
▫ Or the unit may be suspended from analysis because the failure mode is not the same as the units under
analysis. Suspended data is also known as censored data.
• Suspension, left: Test data which is censored before the first failure time.
• Suspension, right: Test data which is censored after the last failure time.
• Test, accelerated: A test which increases a stress parameter on the component above the
normal operating conditions such as temperature, pressure, voltage, chemicals, etc. so the
deterioration in life can be observed in a short time. © 2019 D Training 2
(22)
• Test, substantiation: A test demonstrating that a redesigned part or system has eliminated or
significantly improved a known failure mode.
• Test, reliability: A test demonstrating that a reliability requirement has been met.
• Unreliability: The probability that a device will fail to perform a required or intended
function under stated conditions for a specified period of time. It is the complement of
reliability. It is defined as:
• Wear-out failure: A malfunction resulting from equipment deterioration because of applied

stresses whereby the failure rate increases with age which results in decreasing reliability
during the period of deterioration.
▫ Causes of wear-out are: Time, aging, incipient stresses, limited-life components, wear, scratching, inadequate/improper preventive
maintenance, misalignment/interferences, low cycle fatigue, bearing fatigue failures, corrosion, erosion, mechanical fatigue, and material
property deterioration with time/stress/cycles. Overhauls and inspections may be cost effective for parts which are near the end of their
useful life, and replacements may be required where overhaul is not appropriate. © 2019 D2 Training
(22)
• Wear-out failure: A malfunction resulting from equipment deterioration because of applied
stresses whereby the failure rate increases with age which results in decreasing reliability
during the period of deterioration.
▫ Causes of wear-out are: Time, aging, incipient stresses, limited-life components, wear, scratching, inadequate/improper
preventive maintenance, misalignment/interferences, low cycle fatigue, bearing fatigue failures, corrosion, erosion,
mechanical fatigue, and material property deterioration with time/stress/cycles. Overhauls and inspections may be cost
effective for parts which are near the end of their useful life, and replacements may be required where overhaul is not
appropriate.
• Weibull distribution: A mathematical distribution valuable in reliability application
because of its versatility to model different failure modes.
▫ The family of curves associated with the distribution are determined by selecting various values for the three parameters:
h is the scale parameter known as characteristic life, b is the scale parameter or slope of the data when plotted on 1:1
probability paper, and t0 is the location parameter which shifts the starting point (either in the + or – direction) for the
distribution. In the Weibull failure model many small defect sites compete with each other to be the one that causes the
earliest time of failure, thus its designation as failure of the weakest link.
© 2019 D2 Training
What is Reliability?
Basics…
• Reliability ≠ Availability
• Lusser’s Equation: *
"($ ) = ' ()+, = ' ((-*) = ' (.

t = Mission time
/ = MTBF
0 = Failure rate
N = # failures during mission
• One year = (365 days) * ( 24 hrs/day) = 8760 hours = t

• Availability measures proportion of time system is alive and well
• Reliability measures probability for failure free operation
© 2019 D2 Training
Real world
• Plant with assorted equipment
▫ Base Case (Reactive)
Reactive Maintenance: Breakdown list, Overtime list
▫ Optimize Manpower Use (Planned)
Planned Maintenance: Schedules, Job Plans, Work Orders
▫ Implement Reliability Improvements (Proactive)
Pro-active Maintenance: Pareto Cost List, Failure Data & Analysis, MTBF Improvements,
Root Cause Analysis, Life Cycle Analysis, Cost of Unreliability Analysis
© 2019 D2 Training
Real world
• Uptime = 6730 hrs • Uptime = 7760 hrs • Uptime = 8560 hrs
• Downtime = 2030 hrs • Downtime = 1000 hrs • Downtime = 200 hrs
• Availability = 76.83% • Availability = 88.58% • Availability = 97.72%
• Failures = 12 • Failures = 10 • Failures = 2
• Reliability = 0.00061% • Reliability = 0.0045% • Reliability = 13.53%
• Hours per failure ~ 169 • Hours per failure - 100 • Hours per failure ~ 100
© 2019 D2 Training
(1)
• Availability: A measure of the degree to which an item is in an operable and committable
state at the start of the mission when the mission is called at an unknown and random time.
© 2019 D2 Training
Availability Problem
Availability Calculation Example 1
Component MTBF (hours) MTTR (hours) Availability in %

Power Supply 100,000 8 0.9999200 99.99200
Fan 100,000 8
System Board 300,000 8
memory 1,000,000 8
CPU 500,000 8
Network Interface Controller 250,000 8
© 2019 D2 Training
Availability Probelm
© 2019 D2 Training
Availability Calculation Example 1
Component MTBF (hours) MTTR (hours) Availability in %

Power Supply 100,000 8 0.9999200 99.99200
Fan 100,000 8 0.9999200 99.99200
System Board 300,000 8 0.9999733 99.99733
memory 1,000,000 8 0.9999920 99.99920
CPU 500,000 8 0.9999840 99.99840
Network Interface Controller 250,000 8 0.9999680 99.99680
© 2019 D2 Training
Availability Caculation Example 2
Plant Availability Uptime (hours) Downtime (hours) Availability in %

Plant A 6,815 1,945 0.7779680 77.79680
Plant B 7,612 1,148
Plant C 8,471 289
Plant D 8,550 210
Plant E 8,606 154
Plant F 8,715 45
© 2019 D2 Training
Availability Caculation Example 2
Plant Availability Uptime (hours) Downtime (hours) Availability in %

Plant A 6,815 1,945 0.7779680 77.79680
Plant B 7,612 1,148 0.8689498 86.89498
Plant C 8,471 289 0.9670091 96.70091
Plant D 8,550 210 0.9760274 97.60274
Plant E 8,606 154 0.9824201 98.24201
Plant F 8,715 45 0.9948630 99.48630
© 2019 D2 Training
Reliability is focused on:
• Avoiding events called failures
• Calculated based on lack of failures
• Involves uncertainty since time of future failures are unknown
• Reliability concepts apply to both products sold or products used
• Measure of time oriented performance avoiding failures
• When failures occur and modes of failure are very important
• Failures in industry always have cost impacts, thus cost reductions are motivators for
making improvements
• Probabilities are involved in failures, you never know precisely when a failure will occur
© 2019 D2 Training
What is failure?
• All reliability problems have failures
• Failure: Event which renders equipment non-useful for intended or specified purpose during
designated time interval. Includes:
▫ Stoppage due to malfunction
▫ Unexpected occurrence that interrupts routine operation of a system
▫ Cessation of component function
▫ Cessation of meeting predetermined quality, quantity, or cost expectations
• Failures have classes, degrees, and/or grades. Examples:
▫ Critical – causing downtime 8 hours or greater
▫ Major – causing downtime between 4 to 8 hours
▫ Incidental – causing downtime between ½ to 4 hours
▫ Nuisance – causing downtime less than 30 minutes
• Failure for one case may be ignored in another situation…use good judgement when failures
• When recording failures be consistent…correct data is essential
© 2019 D2 Training
Failure Example:
• How many failures can you afford?
! (#) = & '(

N = # failures during mission
© 2019 D2 Training
95
Series Networks
• Series components – all must function.
• R = (RA ) (RB ) (RC) (multiply R’s)
- ( l A + l B +...+ l n ) t
• R= e (add l’s)
A B C
© 2019 D2 Training
96
Parallel Networks
• Parallel components – all must fail for system to fail.
A
• R = RA + RB – (RARB)
C
• R = 1 – (1 – RA) (1 – RB) (1 – RC)…

▫ (n components)
© 2019 D2 Training
97
Reliability and Redundancy
© 2019 D2 Training
98
Series and Parallel Networks
• Figure 12.10- Reduce parallel

blocks to equivalent series
element.
© 2019 D2 Training
The Probabilistic/Deterministic View
• Study of probabilities of failure over time intervals
▫ Chances for survival or complement chances for failure ! (#) = 1 − ((#)
▫ Statistical models frequently used:
Continuous distributions – normal, log-normal, exponential, and Weibull
Discrete distributions – Poisson
• Failures and reliability focus on three methods:
▫ Accommodate failures – probabilistic viewpoint
Measure rates of failure on broad statistical basis
Plan for redundancies to increase reliability for sustaining operations while accommodating failures
▫ Measure and predict failures – deterministic viewpoint
Find and solve the root causes of individual failures with general view of preventing failures to increase reliability
▫ Detecting and reducing failures to increase reliability
Expected end result of both viewpoints when used with predictive methods for finding potential problems
© 2019 D2 Training
Availability Reliability
Lost Time Lost Time Lost Time Failures Per 10 Failures Per 100
Availability (hours) (minutes) (seconds) Reliability Failures Per Year Years Years
60.00% 3504 10.00% 2.3
65.00% 3066 20.00% 1.61
70.00% 2628 30.00% 1.2
75.00% 2190 40.00% 0.92
85.00% 1314 50.00% 0.69
90.00% 876 60.00% 0.51
95.00% 438 70.00% 0.36
96.00% 350.4 80.00% 0.22 2.23
97.00% 262.8 90.00% 0.11 1.05
98.00% 175.2 95.00% 0.05 0.51
99.00% 87.6 99.00% 0.01 0.1 1.01
99.50% 43.8 99.50% 0.005 0.05 0.50
99.90% 8.76 525.6 99.90% 0.001 0.01 0.10
99.99% 0.876 52.6 3153.6 99.99% 0.0001 0.001 0.01
99.999% 0.0876 5.3 315.36 99.999% 0.00001 0.0001 0.001
99.9999% 0.00876 0.5 31.536 99.9999% 0.000001 0.00001 0.0001
99.99999% 0.000876 0.1 3.1536 99.99999% 0.0000001 0.000001 0.00001
Time Lost Number of Failures
© 2019 D2 Training
What causes the Defects?
• Production 23% • Sales/Customer 15%
▫ Operating above or below design parameters ▫ Expectations exceed production capabilities
▫ Improper operating procedures • Procurement 12%
▫ Operator error
▫ Providing out of specification materials
• Engineering 22% ▫ Improper storage and handling procedures
▫ Undefined or unclear specifications ▫ Material degradation in stores
▫ Improper design
• Management Philosophies 11%
▫ Undocumented changes
▫ Postponing capital replacement
• Maintenance 17% ▫ Maintenance/reliability budget cuts
▫ Poorly designed maintenance strategies
▫ Improper maintenance procedures
▫ Craft errors
Source: A.T. Kearney study of U.S. Manufacturing published by Newsweek

in 1999
© 2019 D2 Training
Reliability/People/Costs
• People who make products and use products affect reliability by introducing large sources of
uncertainty and cause failures
• People influence the environment where products must operate and they introduce variability
which cannot be easily quantified
• People who have the role of reliability, maintenance, and process engineers influence reliability
and failures by their actions
• People strongly influence reliability costs (+ and -) by means not anticipated by designer
• Reliability must be view like Safety

• Safety - All Incidents Are Preventable
• Relaiblity – All Failures are Preventable
© 2019 D2 Training
Reliability/People/Costs
• Equipment Maintenance Strategy:
▫ 89% of failures are random in nature
▫ 11% are a function of time
Source: John Moubray, Nowlan & Heap
© 2019 D2 Training
Reliability System Model
• Governance OE
• Proactive Asset Engagement
Reliaiblity
• Organizational Engagement
Engineering
• Reliability Engineering
• Operational Excellence
Organizational
Engagement
Proactive Asset
Engagement
Governance
© 2019 D2 Training
What is Reliability? OE
Reliaiblity
Engineering
Reliability System Model Organizational
Engagement
• Governance: Proactive Asset
Engagement
▫ Leadership, organizational structure, roles & responsibilities, alignment
Governance
• Proactive Asset Engagement
▫ Asset catalog, work identification/initiation, planning & scheduling, maintenance plans, work execution & review, spare parts,
data management systems, shutdown/turnaround/outage, asset criticality, failure mode analysis, condition monitoring,
preventative maintenance, and KPI’s
• Organizational Engagement
▫ Operator care, craft skills, training, data mining, failure modes and effects analysis (FMEA)
• Reliability Engineering
▫ Reliability, availability, maintainability (RAM) modeling, lifecycle cost analysis (LCC), reliability centered maintenance (RCM),
reliability centered design
• Operational Excellence
▫ Standardization, overall equipment effectiveness (OEE), continuous improvement, asset management, audit
© 2019 D2 Training
Reliability & Total Productive Maintenance (TPM) Programs
• How to reduce unreliability by uncovering hidden defects:
▫ Eliminate accelerating deterioration by cleaning, lubricating, & tightening
Problems are found early for low cost corrections
▫ Eliminate accelerated deterioration by using equipment as designed
Do not abuse, operate above design criteria
▫ Restore equipment to its optimal condition
Perform maintenance repairs to reverse deterioration
▫ Restore processes to their optimum conditions
Renew process and process equipment to restore deterioration
▫ Extend equipment lifetime by correcting design deficiencies
Improve equipment for a longer life
▫ Eliminate unexpected failures by improving operations/maintenance skillset
Train to improve skills and reduce errors
© 2019 D2 Training
Reliability & Total Productive Maintenance (TPM) Programs
• How to reduce unreliability by uncovering hidden defects (continued):
▫ Reduce dispersion in failure intervals so PM replacements are effective
▫ Prevent accelerated deterioration so as to extend life and reduce variation
▫ Lengthing equipment life by avoiding operating and maintenance errors
▫ Periodic servicing, diagnostics, and inspection to restore deterioration
▫ Predict equipment life from its conditions and replace before failure
© 2019 D2 Training
How to Quantify Reliability
• Establish reliability requirements and objectives
▫ Define the overall reliability required for process or product
• Reliability budgeting or allocation
▫ Allocate the overall reliability objective to assets
• Reliability prediction
▫ Estimate asset reliability using data and experience
• Reliability analysis
▫ Identify defective components or systems for implementing improvements
© 2019 D2 Training
How to Quantify Reliability
• Establish reliability requirements and objectives
▫ Define the overall reliability required for process or product
• Reliability budgeting or allocation
▫ Allocate the overall reliability objective to assets
• Reliability prediction
▫ Estimate asset reliability using data and experience
• Reliability analysis
▫ Identify defective components or systems for implementing improvements
© 2019 D2 Training
Reliability Statistics

Weibull Analysis - What is it good for?
• Life prediction with very few failures
• Life prediction with very small sample sizes
• Prediction of when is the first failure expected - Weibayes
• Determination of: Has a problem been fixed? - improving / deteriorating failure
rate - Crow - AMSAA Reliability Grow Modeling
© 2019 D2 Training
Normal Distribution
• Events where the Probability NORMAL DISTRIBUTION
of occurrence follows a “Bell
Curve” 1
PROBABILITY DENSITY
• Failures are not age dependent 0.8
FUNCTION F(x)
▫ Probability of occurrence does 0.6
not change with time
0.4
• Not typically used as left tail
0.2
extends to negative infinity
▫ Could cause incorrect modeling 0
of negative times to failure 0 2 4 6 8 10
MTBF
© 2019 D2 Training
Reliability Statistics NORMAL DISTRIBUTION
Normal Distribution 1
PROBABILITY DENSITY
• Mode, median & mean 0.8
FUNCTION F(x)
• If, mean of data > 0 & variation relatively 0.6
low, useful for modeling certain types of life 0.4
0.2
data
0
0 2 4 6 8 10
MTBF
& +& (
!(#) = * ((#)
√()
Where: f(x) = probability density function

x = event
) = 3.14159
e = natural log (2.718)
© 2019 D2 Training
Weibull Analysis
Log-Normal Distribution
• Events where the Probability of
occurrence follows a
“Exponential Curve”
• Failures are dependent with
time (age)
▫ Probability of occurrence increases
/ decreases with time
© 2019 D2 Training
Weibull Analysis
Log-Normal Distribution
• Will predict lower average failure rates at earlier times
than Weibull distribution
▫ If extrapolate beyond the range of sample data
• Failures caused by chemical reactions or corrosion
usually model with log-normal distribution
./ #))
& ,(-
)')
1)
!(#) = +
#'√)*
Where: f(x) = probability density function

x = event, x >= 1
* = 3.14159
Sigma (') = shape parameter, sigma > 0
© 2019 D2 Training
Weibull Distribution
• Events where the Probability of
occurrence follows a “Bathtub
Curve”
• Failures are age dependant
▫ Probability of occurrence changes
(increases or decreases) with time (age)
• Describes well the failure rate of
real world components
Infant Random Early Old Age
© 2019 D2 Training
Time to Failure (Weibull Distribution)
Weibull Distribution 0.4
0.35
0.3
Probability
0.25
0.2
0.15
0.1
0.05
0
0 5 10 15 20 25 30 35
MTBF
,
!(#) = & − ()(#/+)
Where: f(t) = cumulative fraction failing

t = failure time
eta (+) = characteristic life
beta (,) = slope (shape parameter)
© 2019 D2 Training
Time to Failure (Weibull Distribution)
0.4
0.35
0.3
Probability
0.25
0.2
0.15
0.1
0.05
0
0 5 10 15 20 25 30 35
Infant Random Early Old Age
MTBF
© 2019 D2 Training
Weibull Paper
• Weibull Paper is a ln X vs. lnlnY scale
• Weibull probability function plotted on Weibull Paper produces a straight line
with:
▫ Beta as the slope (shape parameter)
▫ Eta as the value of Y = 63.2%
• Even Engineers can draw a straight line
© 2019 D2 Training
© 2019 D2 Training
© 2019 D2 Training
Failure Modes
• Beta < 1 - infant mortality

• Beta = 1 - random
• 1 < Beta < 4 - early wear out
• Beta > 4 - old age wear out
© 2019 D2 Training
Infant Mortality
• Beta < 1
▫ Inadequate burn-in or stress screening
▫ Production problems
Assembly Error
▫ Overhaul (maintenance) problems
▫ Solid state electronic failures
▫ Overhaul not appropriate - old parts are better than new parts
© 2019 D2 Training
Random Failures
• Beta = 1
▫ Random - failures independent of time
▫ Identical to Exponential Distribution - of those that survive, a
constant % fail in the next unit time.
Maintenance errors, human error
Failures due to nature, foreign object damage, lightening strikes
Mixed failure modes (3 or more)
▫ Overhauls not appropriate - repair what breaks
© 2019 D2 Training
Early Wear Out
• 1.0 < Beta < 4.0
▫ Low Cycle Fatigue
▫ Most Bearing Failures
▫ Corrosion, erosion
▫ Overhauls or parts replacement at low B lives may be cost effective - this restores the
machine to start over again
© 2019 D2 Training
Old Age Wear Out
• Beta > 4.0
▫ Steep Beta mean things are wearing out rapidly
Bad news if trying to run longer
Good news if well past overhaul / replace interval
▫ Stress Corrosion
▫ Material properties (high cycle fatigue)
▫ Brittle materials like ceramics
▫ Some forms of erosion
▫ Inspection and overhaul with parts replacement may be cost effective
© 2019 D2 Training
Weibull Analysis - Example 1
Baseline Rivet Data
Serial Number Failure Time (Minutes) Remarks
1 90 Failure
2 96 Failure
3 100 Flare loosened
4 30 Failure
5 49 Failure
6 45 Flare loosened
7 10 Lug failed
8 82 Failure
© 2019 D2 Training
• Rank Order the data by failure / suspension time
• Determine the Median Rank using (Weibull.com)
• Plot the Data on Weibull Paper
• Draw the Weibull Line
• Determine Eta and Beta
© 2019 D2 Training
Ranked Rivet Data

Order Number (i) Failure Time - t Median
(Minutes) (X) Rank % (Y)
1 30 12.94
2 49 31.38
3 82 50
4 90 68.62
5 96 87.06
Median Rank From Weibull.com, Sample size 5
© 2019 D2 Training
© 2019 D2 Training
Suspended Data
• Non-failed units or units that fail due to a different failure mode are
“censored” or “suspended” units
• Non-failed units, while not having a failure time associated with them have
at least made it to some life without failure of the evaluated failure mode
• Such suspended data must be accounted for in the analysis
• To include suspended data - the rank must be adjusted to account for the
data
Adjusted Rank = (Inverse Rank) x (Previous Adjusted Rank) + (N + 1)

(Inverse Rank) + 1
© 2019 D2 Training
Adjusted Rank Rivet Data

Rank Order Failure Time - t Inverse Adjusted Rank (i) Median Rank
Number (Minutes) (X) Rank
1 10S 8 Suspended
2 30F 7 [(7x0) + (8+1)] / (7 + 1) = 1.125 9.80%
3 45S 6 Suspended
4 5
5 4
6 3
7 2
8 1
Median Rank from Weibull.com, Sample size of 8 using Adjusted Rank (i) and interpolating the table
© 2019 D2 Training

Rank Order Failure Time - t Inverse Adjusted Rank (i) Median Rank
Number (Minutes) (X) Rank
1 10S 8 Suspended
2 30F 7 [(7x0) + (8+1)] / (7 + 1) = 1.125 9.80%
3 45S 6 Suspended
4 49F 5 [(5x1.125) + (8+1)] / (5 + 1) = 2.438 25.50%
5 82F 4 [(4x2.438) + (8+1)] / (4 + 1) = 3.750 41.10%
6 90F 3 [(3x3.750) + (8+1)] / (3 + 1) = 5.063 56.70%
7 96F 2 [(2x5.063) + (8+1)] / (2 + 1) = 6.375 72.30%
8 100S 1 Suspended
Median Rank from Weibull.com, Sample size of 8 using Adjusted Rank (i) and interpolating the table
© 2019 D2 Training
© 2019 D2 Training
Weibull Plot Information
• Eta - Characteristic Life
• Beta - Slope (Shape Factor)
• B1, B10 Lives - % of population that has failed - 1%, 10%, etc.
• R **2 - Goodness of fit, line-to-data 1.00 is perfect fit
© 2019 D2 Training
Benard’s Approximation
• To prevent interpolation of Appendix I Benard determined an
approximation for the Median Rank
▫ Benard’s Median Rank = (i - 0.3) / (N + 0.4)
▫ Where: i = Adjusted Rank, N = Sum of failures and suspensions
▫ The Median Rank is converted to %’s for plotting by multiplying by 100
▫ Accurate to 1% for N = 5 and 0.1% for N = 50
© 2019 D2 Training
Rank Order Failure Time - t Inverse Adjusted Rank (i) Median Rank Median Rank
Number (Minutes) (X) Rank Interpolated Table Benard's Approx.
1 10S 8 Suspended
2 30F 7 1.125 9.77% (1.125 - 0.3) X 100 / (8 + 0.4) = 9.82%
3 45S 6 Suspended
4 49F 5 2.438 25.34%
5 82F 4 3.75 41.03%
6 90F 3 5.063 56.73%
7 96F 2 6.375 72.42%
8 100S 1 Suspended
Median Rank from Interpolated Appendix I compared to Benard's Approximation
© 2019 D2 Training
Rank Order Failure Time - t Inverse Adjusted Rank (i) Median Rank Median Rank
Number (Minutes) (X) Rank Interpolated Table Benard's Approx.
1 10S 8 Suspended
2 30F 7 1.125 9.77% (1.125 - 0.3) X 100 / (8 + 0.4) = 9.82%
3 45S 6 Suspended
4 49F 5 2.438 25.34% (2.438 - 0.3) X 100 / (8 + 0.4) = 25.45%
5 82F 4 3.75 41.03% (3.75 - 0.3) X 100 / (8 + 0.4) = 41.07%
6 90F 3 5.063 56.73% (5.063 - 0.3) X 100 / (8 + 0.4) = 56.70%
7 96F 2 6.375 72.42% (6.375 - 0.3) X 100 / (8 + 0.4) = 72.32%
8 100S 1 Suspended
Median Rank from Interpolated Appendix I compared to Benard's Approximation
© 2019 D2 Training
Suspended Data
• The Characteristic Life (Eta) increases with suspensions
• Beta remains fairly unchanged
• To ignore suspensions is to ignore survival data - results in an overly pessimistic
answer
© 2019 D2 Training
Weibull Data Regression Methods
(Drawing the Line)
• Median Rank Regression (RR)
▫ Time-to-failure as dependant variable (X on Y)
▫ Median Ranks as the dependant variable (Y on X)
• Maximum Likelihood Estimation (MLE)
• Grouped or interval data
▫ Coarse Data
▫ Inspection Data
• Interval Analysis Cover Later
▫ Destructive Testing
▫ NDE Data
© 2019 D2 Training
Median Rank Regression
• Uses least squares regression curve fitting method to minimize the sum of
squares of the dependant variable
• For Weibulls, the X variable time-to-failure has the greatest scatter and error than
the Y variable median ranks
• Best method for Weibull plots is to regress X on Y
© 2019 D2 Training
If Scatter is in the Y direction, Regress Y on X
But if the Scatter is in the X direction, Regress X on Y
Scatter in Y Scatter in X
25 25
20 20
15 15
Y
Y
10 10
5 5
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
X X
Weibull scatter is in the X direction,

Regress X on Y
© 2019 D2 Training
Regression Comparison
Comparison of X on Y versus Y on X Regression - Most Accurate in Bold Face
Based on 1000 Data Sets - Weibull True Values Eta=1000, Beta=3.0 B1=216.0
Sample X on Y Y on X X on Y Y on X X on Y Y on X
Size Eta Eta Beta Beta B1 B1
N=4 994 1009 3.02 2.75 213 190
N=10 996 1014 2.95 2.79 210 193
N=30 999 1011 2.98 2.86 212 204
N=100 1000 1006 2.98 2.92 212 208
N=500 1000 1002 2.99 2.97 214 213
N=1000 1000 1001 2.99 2.98 215 213
© 2019 D2 Training
Maximum Likelihood Estimation
• MLE - Uses statistics to calculate Beta and Eta that maximizes the “likelihood” of
obtaining the observed data
• MLE is a function of the data
• Requires complex computer driven calculations
• Provides better estimation on large data sets
© 2019 D2 Training
Regression vs. MLE Comparison
Comparison of X on Y Regression with Maximum Likelihood Estimate
Most Accurate is Boldfaced
Based on 1000 Data Sets - Weibull True Values Eta=1000, Beta=3.0 B1=216.0
Sample X on Y MLE X on Y MLE X on Y MLE

Size Eta Eta Beta Beta B1 B1
N=4 994 981 3.02 4.00 213 312
N=10 996 990 2.95 3.32 210 245
N=30 999 998 2.98 3.11 212 227
N=100 1000 999 2.98 3.03 212 219
N=500 1000 999 2.99 3.00 214 216
N=1000 1000 999 2.99 3.00 215 215
© 2019 D2 Training
Weibull Data Methods
• Use Rank Regression - X on Y for small to medium data sets

• Use MLE for:
▫ large samples (>500 points)
▫ real random errors are suspected except for small data sets (<30 points)
• Use Rank Regression -Y on X only with special inspection method (Probit data)
© 2019 D2 Training
Parts Failure / Run Time Data
Serial # Time On Part (hours) Remarks
831 9.00 Fail
832 6.00 Fail
833 14.60 Susp
834 1.10 Fail
835 20.00 Fail
836 7.00 Susp
837 65.00 Fail
838 8.00 Susp
© 2019 D2 Training
• Rank Order the data by failure / suspension time
• Calculate the Adjusted Rank (i)
▫ [( IR X Prev Adj Rank) + (N+1)] / (IR + 1)
• Calculate the Median Rank using Benard’s Approximation
▫ [( i - 0.3) / (N + 0.4)] X 100
• Plot the Data on Weibull Paper
© 2019 D2 Training
• What is Eta?
• What is Beta?
• What is failure mode?
• What is B1? What is B10?
• What % of the parts will fail by 10 hours?
• What is the reliability at 50 hours?
© 2019 D2 Training

Serial Rank Order Failure Time - t Fail / Susp Inverse Adjusted Rank (i) Median Rank (Benard's Approx.)
Number Number (Hours) (X) Rank [(IR X Prev Adj Rank) + (N+1)] / (IR + 1) [( i - 0.3) / (N + 0.4)] X 100
834 1 1.1 F 8 [(8x0) + (8+1)] / (8 + 1) = 1.000 [(1.0 - 0.3) / (8 + 0.4)] x 100 = 8.33%
2 7
3 6
4 5
5 4
6 3
7 2
8 1
© 2019 D2 Training

Serial Rank Order Failure Time - t Fail / Susp Inverse Adjusted Rank (i) Median Rank (Benard's Approx.)
Number Number (Hours) (X) Rank [(IR X Prev Adj Rank) + (N+1)] / (IR + 1) [( i - 0.3) / (N + 0.4)] X 100
834 1 1.1 F 8 [(8x0) + (8+1)] / (8 + 1) = 1.000 [(1.0 - 0.3) / (8 + 0.4)] x 100 = 8.33%
832 2 6 F 7 [(7x1) + (8+1)] / (7 + 1) = 2.000 [(2.0 - 0.3) / (8 + 0.4)] x 100 = 20.24%
836 3 7 S 6 Suspended
838 4 8 S 5 Suspended
831 5 9 F 4 [(4x2) + (8+1)] / (4 + 1) = 3.400 [(3.4 - 0.3) / (8 + 0.4)] x 100 = 36.90%
833 6 14.6 S 3 Suspended
835 7 20 F 2 [(2x3.4) + (8+1)] / (2 + 1) = 5.267 [(5.267 - 0.3) / (8 + 0.4)] x 100 = 59.13%
837 8 65 F 1 [(1x5.267) + (8+1)] / (1 + 1) = 7.133 [(7.133 - 0.3) / (8 + 0.4)] x 100 = 81.34%
© 2019 D2 Training
© 2019 D2 Training
• What is Eta? - 29 hours
• What is Beta? - 0.8
• What is failure mode? - Infant (<1)
• What is B1? - 0.08 hours
• What is B10? - 1.75 hours
• What % of the parts will fail by 10 hours? - 35%
• What is the reliability at 50 hours? - 20%
© 2019 D2 Training
Failure Modes Cover
Other Modes
• The first (shortest life) failure mode “governs” the Weibull
▫ Weibull line covers all other (later) Weibulls
• All units fail due to the shortest failure mode and never get a chance to
encounter the second failure mode
• There is always additional failure modes to the right of a given Weibull
▫ There is always some new failure mode
© 2019 D2 Training
Failure Modes Cover
© 2019 D2 Training
Dirty Data
Weibull Analysis meets

the Real World
© 2019 D2 Training
Dirty Data Problems
• Small Sample - few failures

• Poor Fit to data - Goodness-Of-Fit
• Suspensions - lack data of failures
• Curved Weibulls - three parameter Weibull
• Curve Weibulls - Log Normal Distribution
© 2019 D2 Training
Dirty Data Problems
• Data Inconsistencies
▫ Low Time Failures
▫ Close Serial Numbers
▫ Mixed Failure Modes
▫ Suspension times unknown (unknown life of those left in the fleet)
▫ Inspection or Interval Data
▫ Steep Slope (Beta > 10)
© 2019 D2 Training
90% Confidence Bounds
© 2019 D2 Training
© 2019 D2 Training
Sample Size
• The larger the sample size:

▫ Failures
▫ Suspensions
• Improves reliability (reduces probability bounds)
© 2019 D2 Training
Goodness -Of-Fit
• Correlation Coefficient (r)
▫ Measures the strength of a linear relationship to two variables
Closer to 1.00 the better the fit
• Critical Correlation Coefficient (CCC)
▫ Monte Carlo simulation of an ideal Weibull
▫ Uses a 90% bound for goodness of fit
▫ If “r” is > CCC - Good Fit
• Critical Coefficient of Determination
▫ The square of “r” (r2)
▫ Closer to 1.0 the better the fit
▫ Good quick measure of Goodness-Of-Fit
© 2019 D2 Training
Goodness -Of-Fit
• Tables of CCC vs Number of Failures

▫ Used to determine if calculated line fits the data based on the number of failures with
an ideal distribution (Weibull or Log Normal)
• WinsmithWeibull automatically evaluates “r” against the CCC
© 2019 D2 Training
Suspensions
• Suspensions occur two ways
▫ Unit has not failed yet
▫ Unit failed due to some other failure mode
• Suspensions increase the Characteristic Life (Eta) but
have little effect on Slope (Beta)
• Suspensions past the last failure have a slight effect in
reducing the Slope (Beta)
© 2019 D2 Training
Data Inconsistencies
Common Data Problems
© 2019 D2 Training
Low Time Failures
• Failures that occur very early (short life) often after substantial number are
operating (late suspensions) well past current failures
• Suspect Batch Problem
▫ Change in manufacturing
▫ Change (error) in overhaul or maintenance
© 2019 D2 Training
Low-Time Failures
© 2019 D2 Training
Close Serial Numbers
• Anytime failures occur in a group of serial numbers - batch problems should be

suspected
• Failures can occur anytime in fleet life
▫ Mid production change
▫ Change in suppliers
• Implications are the same as Low-Time Failures - look for cause of batch
grouping
© 2019 D2 Training
Mixed Failure Modes
• Not sorting data by failure mode results in a Beta = ~1.0 with Eta
= ~MTTF which is the Exponential Distribution
• Sorting data by the part that failed is NOT sufficient (i.e. bearing
failures)
▫ Must sort by failure mode (i.e. lack of lubrication, overload,
brinelling, false brinelling, contamination, etc.)
▫ All failures NOT of the mode being evaluated must be suspended
© 2019 D2 Training
Mixed Failure Modes
• Getting data in the failure mode specific format is the biggest challenge
in getting good Weibull results
▫ Takes discipline to determine true root cause failure mode for each failure and
to record with failure code
• Getting the suspension lives is the second biggest challenge (more to
come on this issue)
• Two failure modes in the data is a common problem - Called Bi-
Weibull
© 2019 D2 Training
Bi-Weibull
• A Weibull Plot with a “dogleg bend” in the data (data is telling a story)
• Engineering basis for two different failure modes must be determined
• Segregation of the data sets into two Weibulls is justified if two different
failure modes can be determined
• Steep slope followed by a shallow slope indicates a Batch Problem
(perpetual survivors)
© 2019 D2 Training
Bi-Weibull
© 2019 D2 Training
Bi-Weibull
© 2019 D2 Training
Suspension Times Unknown
• Have time of failures

• Don’t know the run time on the units that have NOT failed (don’t know
suspensions)
• May not even know the number of units that require suspension (unknown fleet
size)
© 2019 D2 Training
Dauser Shift
• Suspension Times Unknown
• Do know some information
▫ Time on each failure
▫ Average time to failure
▫ From the initial Weibull - know Beta
• Assumes future failures of un-failed units will on average have the same
MTTF
• Last Resort
▫ No mathematical support for this approach
▫ Even a gross estimation of suspension times (histogram) is better than Dauser Shift
▫ Some data is better than no data
© 2019 D2 Training
Dauser Shift
© 2019 D2 Training
Dauser Shift
MTTF
© 2019 D2 Training
Inspection Interval Data
and Course Data
When Benign Failures Occur
© 2019 D2 Training
• Failure detected and noted on inspection
▫ Called Interval data
▫ Failure could have occurred anytime after the last inspection - time of exact
failure is unknown but bracketed between the inspections
▫ Unknown failure time results in less certainty in the Y (% failures) than the
inspection time - should regress Y on X
• Probit Data Options
▫ Option 1 - Use only the last failure from the inspection to determine the
Weibull
▫ Option 2 - Regress Y on X
© 2019 D2 Training
© 2019 D2 Training
Interval Data - Inspection Option
© 2019 D2 Training
Interval Data
• Heat Exchanger Tube Failures (cracking)
▫ Fixed number of tubes inspected - 4000
▫ Cumulative failure is the total # of tubes plugged after each inspection (#plugs this inspection +
previous # of plugs)
▫ Inspection interval every 18 months
• Use “Probit” data (regress Y on X)
▫ Median Rank not used to plot data - use Cumulative % Failure for plotting position
▫ Data has three components
Inspection Time
Constant number inspected
Cumulative number failed
© 2019 D2 Training
Interval Data - Inspection Option
© 2019 D2 Training
Course Data
• When data not recorded in the correct time frame
▫ Recorded annually instead of monthly
▫ Quarterly instead of weekly, etc.
• Same problems as Inspection Data
• Data will have vertical data sets
• Handle the same way as Inspection Data
© 2019 D2 Training
Steep Beta (Slope)
• Hide data problems
▫ Failures are compressed by the scale
▫ Hides data issues such as
Curved data
Outlier data points
• Indicative of possible data selection problems (liars figure and
figures lie)
• If real - failures are very rapid (sudden death)
© 2019 D2 Training
Steep Beta
© 2019 D2 Training
Interpreting the Plot
• Look at each figure

▫ Is it a good Weibull?
▫ Is the data trying to tell us something?
▫ What steps (if any) should be taken to improve the analysis?
▫ Anything to be concluded?
© 2019 D2 Training
WEIBAYES
WHEN WEIBULLS ARE UNCERTAIN

OR IMPOSSIBLE
© 2019 D2 Training
Weibayes
N
• Eta = [ Σ tiBeta / r ] 1/ Beta

i=1
• ti = time / cycle on unit
• r = number of failed units
• Eta = MLE of Characteristic Life
• N = total number of failures + suspensions
© 2019 D2 Training
Weibayes
• Weibull Analysis with an assumed Beta
▫ From historical or test data
▫ Assumes first (next) failure is imminent
• With defined failure modes - significantly improved accuracy over small
data set Weibulls
• Useful under two conditions
▫ No failures
▫ With failures (typically very few data sets)
© 2019 D2 Training
Weibayes with Failure
© 2019 D2 Training
Weibayes Without Failure
© 2019 D2 Training
Weibayes
• Past experience good source of Betas

• Weibull data sources
▫ Paul Barringer - http://www.barringer1.com/
▫ Dr. Bob Abernethy / Mr. Wes Fulton - http://www.weibullnews.com/
▫ Reliability Engineering - http://www.Weibull.com
© 2019 D2 Training
Crow - AMSAA
Reliability Growth Model
When Weibull and Weibayes Won’t Work
© 2019 D2 Training
• Used when:
▫ Have multiple failure modes
▫ Don’t know the size of the fleet
▫ Constant change (improving / solving)
• Provides graphic of status - better / worse than past
• Provides method to predict next failure
• Manager friendly!!!
© 2019 D2 Training
• Plots Cumulative Failures vs. Cumulative Time

• Log - log curve fit (line)
• Parameters (Beta, Lambda) determined from fit line
• Developed by Larry Crow at the Army Material Systems Analysis Activity
(AMSAA)
© 2019 D2 Training
• Rho (t) = Lambda x Beta x t(Beta -1)

• Log of cumulative failures N(t) versus log cumulative time (t) is linear if
model applies
• N(t) = Lambda x tBeta
• ln N(t) = ln (Lambda) + ln (t) which is a straight line
© 2019 D2 Training
• Cumulative failure rate C(t) = N(t) / t

• C(t) = Lambda x t(Beta - 1)
• Rho (t) = Lambda x Beta x t(Beta -1)

• MTTF = 1 / Rho (t)
• Beta < 1.0 indicates that failure rate is decreasing (things getting better)
• Beta = 1.0 indicates that failure rate is the same (status quo - nothing changing)
• Beta > 1.0 indicates failure rate is increasing (getting worse - NOT good)
© 2019 D2 Training
CPO Safety Growth Model
© 2019 D2 Training
DATA SET ISSUES
When Mixed Failure Modes Occur
© 2019 D2 Training
Data Set Issues
• When evaluating Weibulls that clearly have mixed failure modes
▫ Have Engineering justification to segregate the data
▫ How do we know that the data sets are really different sets?
• Don’t want to be fooling ourselves into seeing differences that
are not statistically significant
© 2019 D2 Training
Single Data Set
© 2019 D2 Training
Data Set Segregated
© 2019 D2 Training
Segregated Data Set
© 2019 D2 Training
Segregated Data Set
• Notice that the 90% confidence bounds overlap

▫ Data sets are NOT significantly different
▫ No statistical justification to segregate the data
▫ May not have two different failure modes
© 2019 D2 Training
Meridium Problems
• Data is limited to since SAP came on line - no previous MWO data
• Data is very dirty, has all work orders
▫ Mixed failure modes
▫ Some data is flat wrong
• Difficult to cull data from a Query
• Analysis tools set is limited in ability to handle dirty data / inspection
data issues
© 2019 D2 Training
Meridium Solutions
• Use Meridium to Query SAP and generate a data set

• Export data set with descriptions (short text, etc.) to Excel
• Clean data set of extraneous errors
• Use Excel to calculated TOF by subtracting dates - generate a complete
data set
• Paste data into WinSmith Weibull
© 2019 D2 Training
Example Analysis
Carburization Failures of Olefin Cracking

Furnace Tubes
© 2019 D2 Training
Cracking Furnace Design
© 2019 D2 Training
TUBE CONDITIONS
• Operating = 6 - 20 psig @ 2,000 oF

• Decoke = 6 - 30 psig @ 2,050 oF
• Alloys used (% Cr / % Ni)
- 25/20 (HK 40)
- 25/35 (HP45 and HP-MOD)
- 35/45
© 2019 D2 Training
Original Operation
• Heavy Feedstock
• Substantial Steam-to-Feed Ratio
• Low to Moderate Feed Rates (50-75%)
• Moderate Severity (outlet temperature)
© 2019 D2 Training
Revised Operation
• Lighter Feedstock
• Higher Feed Rates (125%)
• Lower Steam-to Feed Ratio
• Slight Increase in Severity
© 2019 D2 Training
Weibull Analysis
• Evaluate 6” Outlet Tubes
> Highest Temperature
> Data on originals and replacements
> Catastrophic failure could result in
loss of furnace
© 2019 D2 Training
FURNACE OUTLET TUBE FAILURE
99
95 5
W/rr/insp1
90 13
% 80 7
60
70 74
7
C 50 8
10
O 40 15
30 12
I 103
20 43
L 10
S 10
18
F 5
A 4
I 2
L 1
E 3
D .5
Eta Beta r^2 n/s
.2
64.04698 2.584 0.970 210/62
.1
1 10 100 1000
COIL LIFE (Months)

© 2019 D2 Training
Initial Weibull Analysis
Eta Beta r^2

64 months 2.584 0.970
• Good fit with inspection data
• Weibull Distribution
• Visual Evaluation - Slightly Curved Data
© 2019 D2 Training
99
95 W/rr/insp1/c=fmD90 5
90 6
% 80
70 13 13
60 12 7
C 50 3
40 10 73
O REPLACEMENT TUBES 42
30 10 7
I
20 2
L 18
S 10 10
ORIGINAL TUBES
5 4
F
A 2
I 2
3
L 1
E
D .5
Eta Beta r^2 n/s
79.84531 6.662 0.959 96/36
.2
45.69361 2.771 0.959 114/26
.1
1 10 100 1000
COIL LIFE (MONTHS)

© 2019 D2 Training
Second Weibull Analysis
Set - Eta (months) Beta Wear Out
Original 79.8 6.662 Old Age
Repl. 45.7 2.771 Early
© 2019 D2 Training
90% Likelihood Ratio Contours
9
B 6
e
t y 5
a Original Tubes
4
Replacement Tubes
3
1997
2 0721
1 REK
30 40 50 60 70 80 90 100
Eta (x)
© 2019 D2 Training
99 5
95 W/rr/insp1/c=lrD90
90 13
% 80
70
60 7
C 50 WALL TEMP. CHANGE 3
40 7
O 7
30
I
20 2
L 10 ORIGINAL TUBES
S 10
F 5
A 2
I 2
L 1
E
D .5
Eta Beta r^2 n/s
79.84531 6.662 0.959 96/36
.2
45.7 6.662
.1
10 100 1000
COIL LIFE (MONTHS)
© 2019 D2 Training
99
95 W/rr/insp1/c=fmD90 5
90 6
% 80
70 13 13
60 12 7
C 50 3
40 10 73
O REPLACEMENT TUBES 42
30 10 7
I
20 2
L 18
S 10 10
ORIGINAL TUBES
5 4
F
A 2
I 2
3
L 1
E
D .5
Eta Beta r^2 n/s
79.84531 6.662 0.959 96/36
.2
45.69361 2.771 0.959 114/26
.1
1 10 100 1000
COIL LIFE (MONTHS)

© 2019 D2 Training
Root Cause
• Carburization Rate is limited by SiO2 scale barrier in 25/20 and 25/35 alloys
• Reduction of steam-to-feed ratio destabilizes the SiO2 layer
• Explains changes in Beta
© 2019 D2 Training
OPTIMAL REPLACEMENT INTERVAL
U 6" FURNACE OUTLET TUBE
N .6
I
T
.5
C
O
S .4
T
.3
P
E
R
.2
M
O .1
N
T
H 0
0 20 40 60 80 100 120 140
TUBE LIFE (MONTH)

© 2019 D2 Training
Maintenance Philosophy
• Optimum replacement verifies low penalty for extending tube

replacements
• Use Characteristic Life to predict tube replacement.
• Limit repairs to two attempts before requiring replacement.
© 2019 D2 Training
Other Weibulls
Item Eta Beta

4”- Ethane 67.9 3.139
4”- Oil 118.1 2.901
3”- Ethane 231 1.254
3”- Oil 234 6.815
© 2019 D2 Training
Other Weibulls
Item Eta Beta

3” to 4”- Ethane 117 2.754
3” to 4”- Oil 185 6.815
4” to 6”- Ethane 46* 1.157*
* Log-Normal distribution (Mu, SigF)
© 2019 D2 Training
Weibayes
• Weibayes is a Weibull without (or few) failures

• Must have a known failure mode - good estimate of Beta
• Use where few or no failures exist
• Uses the ages of non-failed components to estimate the Weibayes
© 2019 D2 Training
6” Ethane Furnace Tube
© 2019 D2 Training
6” Oil Furnace Tube
© 2019 D2 Training
Conclusions
Weibull Analysis Provides

• Clear method to evaluate tube life
• Quantifies maintenance vs replacement decisions
• Insight into changes of failure mechanisms
• Excellent means to include the $ side of the solution.
© 2019 D2 Training
References Utilized to Develop these Materials
• New Weibull Handbook
▫ Available at: http://www.barringer1.com/
• PlayTIME With SuperSMITH - Source of most example problems
© 2019 D2 Training

Intro To Reliability Fundamentals IEEE Rev2

Uploaded by

Copyright:

Available Formats

Intro To Reliability Fundamentals IEEE Rev2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Intro To Reliability Fundamentals IEEE Rev2

Uploaded by

Copyright:

Available Formats

Introduction to Reliability Fundamentals

Donald G. Dunn – Principal Consultant

Milford Haven, UK – 1994

Toulouse, France - 2001

• Reactive Maintenance - all behavior is reactive

• Planned Maintenance - behavior is guided through the planning and

Reliability Operating Excellence

A = Total Available Time

Key Manufacturing Losses

Product lost that could

How does your plant compare to these?

Asset Performance – the state of asset

Asset Accountability – the responsibility of

Source: ISO PC 251 – ISO 55000 Asset Management

22% • Undefined/unclear specifications

17% • Poorly designed maintenance strategies

12% • Providing out-of-specification materials

15% as a result of Sales/Customer expectations exceeding Production capabilities.

Equipment Maintenance Strategy:

• Has to include Failure Mechanism

• NOT another failure

• WaloddiWeibull - 1887 - 1979

• Dr. Bob Abernathy

• Mean time to failure (MTTF for non-repairable equipment): A basic measure of

• Wear-out failure: A malfunction resulting from equipment deterioration because of applied

Where: R = exponential reliability

• One year = (365 days) * ( 24 hrs/day) = 8760 hours = t

Availability Calculation Example 1

Component MTBF (hours) MTTR (hours) Availability in %

Availability Calculation Example 1

Component MTBF (hours) MTTR (hours) Availability in %

Availability Caculation Example 2

Plant Availability Uptime (hours) Downtime (hours) Availability in %

Availability Caculation Example 2

Plant Availability Uptime (hours) Downtime (hours) Availability in %

! (#) = & '(

Where: R = exponential reliability

• R = 1 – (1 – RA) (1 – RB) (1 – RC)…

• Figure 12.10- Reduce parallel

Time Lost Number of Failures

Source: A.T. Kearney study of U.S. Manufacturing published by Newsweek

• Reliability must be view like Safety

Source: John Moubray, Nowlan & Heap

110 © 2019 D2 Training

low, useful for modeling certain types of life 0.4

Where: f(x) = probability density function

Where: f(x) = probability density function

Infant Random Early Old Age

Where: f(t) = cumulative fraction failing

• Beta < 1 - infant mortality

Ranked Rivet Data

Adjusted Rank = (Inverse Rank) x (Previous Adjusted Rank) + (N + 1)

Adjusted Rank Rivet Data

Adjusted Rank Rivet Data

Weibull scatter is in the X direction,

Sample X on Y MLE X on Y MLE X on Y MLE

• Use Rank Regression - X on Y for small to medium data sets

Parts Failure / Run Time Data

Parts Failure / Run Time Data