Using the Weibull Distribution: Reliability, Modeling, and Inference

Understand and utilize the latest developments in Weibullinferential methods

While the Weibull distribution is widely used in science andengineering, most engineers do not have the necessary statisticaltraining to implement the methodology effectively. Using theWeibull Distribution: Reliability, Modeling, and Inferencefills a gap in the current literature on the topic, introducing aself–contained presentation of the probabilistic basis for themethodology while providing powerful techniques for extractinginformation from data.

The author explains the use of the Weibull distribution and itsstatistical and probabilistic basis, providing a wealth of materialthat is not available in the current literature. The book begins byoutlining the fundamental probability and statistical concepts thatserve as a foundation for subsequent topics of coverage,including:

Optimum burn–in, age and block replacement, warranties andrenewal theory Exact inference in Weibull regression Goodness of fit testing and distinguishing the Weibull from thelognormal Inference for the Three Parameter Weibull

Throughout the book, a wealth of real–world examples showcasesthe discussed topics and each chapter concludes with a set ofexercises, allowing readers to test their understanding of thepresented material. In addition, a related website features theauthor′s own software for implementing the discussed analyses alongwith a set of modules written in Mathcad®, and additionalgraphical interface software for performing simulations.

With its numerous hands–on examples, exercises, and softwareapplications, Using the Weibull Distribution is an excellentbook for courses on quality control and reliability engineering atthe upper–undergraduate and graduate levels. The book also servesas a valuable reference for engineers, scientists, and businessanalysts who gather and interpret data that follows the Weibulldistribution.

Preface xiii

1. Probability 1

1.1 Sample Spaces and Events, 2

1.2 Mutually Exclusive Events, 2

1.3 Venn Diagrams, 3

1.4 Unions of Events and Joint Probability, 4

1.5 Conditional Probability, 6

1.6 Independence, 8

1.7 Partitions and the Law of Total Probability, 9

1.8 Reliability, 12

1.9 Series Systems, 12

1.10 Parallel Systems, 13

1.11 Complex Systems, 15

1.12 Crosslinked Systems, 16

1.13 Reliability Importance, 19

References, 20

Exercises, 21

2. Discrete and Continuous Random Variables 23

2.1 Probability Distributions, 24

2.2 Functions of a Random Variable, 26

2.3 Jointly Distributed Discrete Random Variables, 28

2.4 Conditional Expectation, 32

2.5 The Binomial Distribution, 34

2.5.1 Confidence Limits for the Binomial Proportion p, 38

2.6 The Poisson Distribution, 39

2.7 The Geometric Distribution, 41

2.8 Continuous Random Variables, 42

2.8.1 The Hazard Function, 49

2.9 Jointly Distributed Continuous Random Variables, 51

2.10 Simulating Samples from Continuous Distributions, 52

2.11 The Normal Distribution, 54

2.12 Distribution of the Sample Mean, 60

2.12.1 P[X < Y] for Normal Variables, 65

2.13 The Lognormal Distribution, 66

2.14 Simple Linear Regression, 67

References, 69

Exercises, 69

3. Properties of the Weibull Distribution 73

3.1 The Weibull Cumulative Distribution Function (CDF),Percentiles, Moments, and Hazard Function, 73

3.1.1 Hazard Function, 75

3.1.2 The Mode, 77

3.1.3 Quantiles, 77

3.1.4 Moments, 78

3.2 The Minima of Weibull Samples, 82

3.3 Transformations, 83

3.3.1 The Power Transformation, 83

3.3.2 The Logarithmic Transformation, 84

3.4 The Conditional Weibull Distribution, 86

3.5 Quantiles for Order Statistics of a Weibull Sample, 89

3.5.1 The Weakest Link Phenomenon, 92

3.6 Simulating Weibull Samples, 92

References, 94

Exercises, 95

4. Weibull Probability Models 97

4.1 System Reliability, 97

4.1.1 Series Systems, 97

4.1.2 Parallel Systems, 99

4.1.3 Standby Parallel, 102

4.2 Weibull Mixtures, 103

4.3 P(Y < X), 105

4.4 Radial Error, 108

4.5 Pro Rata Warranty, 110

4.6 Optimum Age Replacement, 112

4.6.1 Age Replacement, 115

4.6.2 MTTF for a Maintained System, 117

4.7 Renewal Theory, 119

4.7.1 Block Replacement, 121

4.7.2 Free Replacement Warranty, 122

4.7.3 A Renewing Free Replacement Warranty, 122

4.8 Optimum Bidding, 123

4.9 Optimum Burn–In, 124

4.10 Spare Parts Provisioning, 126

References, 127

Exercises, 128

5. Estimation in Single Samples 130

5.1 Point and Interval Estimation, 130

5.2 Censoring, 130

5.3 Estimation Methods, 132

5.3.1 Menon s Method, 132

5.3.2 An Order Statistic Estimate of x0.10, 134

5.4 Graphical Estimation of Weibull Parameters, 136

5.4.1 Complete Samples, 136

5.4.2 Graphical Estimation in Censored Samples, 140

5.5 Maximum Likelihood Estimation, 145

5.5.1 The Exponential Distribution, 147

5.5.2 Confidence Intervals for the ExponentialDistribution Type II Censoring, 147

5.5.3 Estimation for the Exponential Distribution IntervalCensoring, 150

5.5.4 Estimation for the Exponential Distribution Type ICensoring, 151

5.5.5 Estimation for the Exponential Distribution The ZeroFailures Case, 153

5.6 ML Estimation for the Weibull Distribution, 154

5.6.1 Shape Parameter Known, 154

5.6.2 Confidence Interval for the Weibull ScaleParameter Shape Parameter Known, Type II Censoring, 155

5.6.3 ML Estimation for the Weibull Distribution ShapeParameter Unknown, 157

5.6.4 Confidence Intervals for Weibull Parameters Completeand Type II Censored Samples, 162

5.6.5 Interval Censoring with the Weibull, 167

5.6.6 Confidence Limits for Weibull Parameters Type ICensoring, 167

References, 177

Exercises, 179

6. Sample Size Selection, Hypothesis Testing, and Goodness ofFit 180

6.1 Precision Measure for Maximum Likelihood (ML) Estimates,180

6.2 Interval Estimates from Menon s Method of Estimation,182

6.3 Hypothesis Testing Single Samples, 184

6.4 Operating Characteristic (OC) Curves for One–Sided Tests ofthe Weibull Shape Parameter, 188

6.5 OC Curves for One–Sided Tests on a Weibull Percentile,191

6.6 Goodness of Fit, 195

6.6.1 Completely Specified Distribution, 195

6.6.2 Distribution Parameters Not Specified, 198

6.6.3 Censored Samples, 201

6.6.4 The Program ADStat, 201

6.7 Lognormal versus Weibull, 204

References, 210

Exercises, 212

7. The Program Pivotal.exe 213

7.1 Relationship among Quantiles, 216

7.2 Series Systems, 217

7.3 Confidence Limits on Reliability, 218

7.4 Using Pivotal.exe for OC Curve Calculations, 221

7.5 Prediction Intervals, 224

7.6 Sudden Death Tests, 226

7.7 Design of Optimal Sudden Death Tests, 230

References, 233

Exercises, 234

8. Inference from Multiple Samples 235

8.1 Multiple Weibull Samples, 235

8.2 Testing the Homogeneity of Shape Parameters, 236

8.3 Estimating the Common Shape Parameter, 238

8.3.1 Interval Estimation of the Common Shape Parameter, 239

8.4 Interval Estimation of a Percentile, 244

8.5 Testing Whether the Scale Parameters Are Equal, 249

8.5.1 The SPR Test, 250

8.5.2 Likelihood Ratio Test, 252

8.6 Multiple Comparison Tests for Differences in ScaleParameters, 257

8.7 An Alternative Multiple Comparison Test for Percentiles,259

8.8 The Program Multi–Weibull.exe, 261

8.9 Inference on P (Y < X), 266

8.9.1 ML Estimation, 267

8.9.2 Normal Approximation, 269

8.9.3 An Exact Simulation Solution, 271

8.9.4 Confi dence Intervals, 273

References, 274

Exercises, 274

9. Weibull Regression 276

9.1 The Power Law Model, 276

9.2 ML Estimation, 278

9.3 Example, 279

9.4 Pivotal Functions, 280

9.5 Confidence Intervals, 281

9.6 Testing the Power Law Model, 281

9.7 Monte Carlo Results, 282

9.8 Example Concluded, 285

9.9 Approximating u∗ at Other Stress Levels, 287

9.10 Precision, 289

9.11 Stress Levels in Different Proportions Than Tabulated,289

9.12 Discussion, 291

9.13 The Disk Operating System (DOS) Program REGEST, 291

References, 296

Exercises, 296

10. The Three–Parameter Weibull Distribution 298

10.1 The Model, 298

10.2 Estimation and Inference for the Weibull LocationParameter, 300

10.3 Testing the Two– versus Three–Parameter WeibullDistribution, 301

10.4 Power of the Test, 302

10.5 Interval Estimation, 302

10.6 Input and Output Screens of LOCEST.exe, 307

10.7 The Program LocationPivotal.exe, 309

10.8 Simulated Example, 311

References, 311

Exercises, 312

11 Factorial Experiments with Weibull Response 313

11.1 Introduction, 313

11.2 The Multiplicative Model, 314

11.3 Data, 317

11.4 Estimation, 317

11.5 Test for the Appropriate Model, 319

11.6 Monte Carlo Results, 320

11.7 The DOS Program TWOWAY, 320

11.8 Illustration of the Influence of Factor Effects on theShape Parameter Estimates, 320

11.9 Numerical Examples, 327

References, 331

Exercises, 332

Index 333

JOHN I. McCOOL, PhD, is Professor of Systems Engineeringat Penn State Great Valley School of Graduate Professional Studies.A Fellow of the American Society for Quality, Dr. McCool previouslyserved as principal engineering scientist at SKF Industries Inc.,where he conducted corporate as well as federally sponsoredresearch projects with the Wright–Patterson Air Force Base, theOffice of Naval Research, the Naval Air Propulsion Center, theDepartment of Energy, and the Air Force Office of ScientificResearch.