Engineered Software

Preventive Maintenance


Home

Optimization Inspection Schedules

Predictive Maintenance Schedules

Engineered Software Home Page

A basic question for a production or maintenance manager is how often to perform preventive maintenance, or how often to replace components. The time to fail for an item or system is variable and can be represented by some probability density function f(t). Referring to the figure below, the cost of failures per unit time decreases as preventive maintenance is done more often, but the cost of preventive maintenance per unit time increases. There exists a point where the total cost of failures and preventive maintenance per unit time is at a minimum; the optimum schedule for preventive maintenance.

The assumptions for this routine are:

  1. Preventive maintenance is performed on an item at time = T where T is the optimum schedule between preventive maintenance.

  2. Each time maintenance is performed it is returned to its initial state, that is, the item is "as good as new."

  3. The time to fail follows a Weibull distribution.

The optimum time between maintenance actions is found by minimizing the total cost per unit time.

where Cp is the cost of preventive maintenance,
Cf is the cost of a failure, and
T is the time between preventive maintenance actions.

Minimizing the equation above is tedious, and numerical routines are usually required. Dodson (1994), developed a tabular solution for this problem given the following assumptions.

  1. The time to fail follows a Weibull distribution.
  2. Preventive maintenance is performed on an item at time T at a cost of Cp.
  3. If the item fails before time = T, a failure cost of Cf is incurred.
  4. Each time preventive maintenance in performed, the item is returned to its initial state; that is, the item is "as good as new."

The optimum time between preventive maintenance actions is

T = mq + d

where m is a function of the ratio of the failure cost to the preventive maintenance cost and the value of the shape parameter, and is given in the table below.

q is the scale parameter of the Weibull distribution, and

d is the location parameter of the Weibull distribution.

ß

Cf/Cp

1.5

2.0

2.5

3.0

4.0

5.0

7.0

10.0

2.0

2.229

1.091

0.883

0.810

0.766

0.761

0.775

0.803

2.2

1.830

0.981

0.816

0.760

0.731

0.733

0.755

0.788

2.4

1.579

0.899

0.764

0.720

0.702

0.711

0.738

0.777

2.6

1.401

0.834

0.722

0.688

0.679

0.692

0.725

0.766

2.8

1.265

0.782

0.687

0.660

0.659

0.675

0.713

0.758

3.0

1.158

0.738

0.657

0.637

0.642

0.661

0.702

0.749

3.3

1.033

0.684

0.620

0.607

0.619

0.642

0.687

0.739

3.6

0.937

0.641

0.589

0.582

0.600

0.627

0.676

0.730

4.0

0.839

0.594

0.555

0.554

0.579

0.609

0.662

0.719

4.5

0.746

0.547

0.521

0.526

0.557

0.591

0.648

0.708

5

0.676

0.511

0.493

0.503

0.538

0.575

0.635

0.699

6

0.574

0.455

0.450

0.466

0.509

0.550

0.615

0.683

7

0.503

0.414

0.418

0.438

0.486

0.530

0.600

0.671

8

0.451

0.382

0.392

0.416

0.468

0.514

0.587

0.661

9

0.411

0.358

0.372

0.398

0.452

0.500

0.575

0.652

10

0.378

0.337

0.355

0.382

0.439

0.488

0.566

0.645

12

0.329

0.304

0.327

0.357

0.417

0.469

0.550

0.632

14

0.293

0.279

0.306

0.338

0.400

0.454

0.537

0.621

16

0.266

0.260

0.288

0.323

0.386

0.441

0.526

0.613

18

0.244

0.244

0.274

0.309

0.374

0.430

0.517

0.605

20

0.226

0.230

0.263

0.298

0.364

0.421

0.508

0.598

25

0.193

0.205

0.239

0.275

0.343

0.402

0.492

0.584

30

0.170

0.186

0.222

0.258

0.328

0.387

0.478

0.573

35

0.152

0.172

0.207

0.245

0.315

0.374

0.468

0.564

40

0.139

0.160

0.197

0.234

0.304

0.364

0.459

0.557

45

0.128

0.151

0.187

0.225

0.295

0.356

0.451

0.550

50

0.119

0.143

0.179

0.217

0.288

0.348

0.444

0.544

60

0.105

0.130

0.167

0.204

0.274

0.335

0.432

0.534

70

0.095

0.120

0.157

0.193

0.264

0.325

0.422

0.526

80

0.087

0.112

0.148

0.185

0.255

0.316

0.415

0.518

90

0.080

0.106

0.141

0.177

0.248

0.309

0.407

0.513

100

0.074

0.101

0.135

0.172

0.241

0.303

0.402

0.507

150

0.057

0.082

0.115

0.150

0.217

0.278

0.379

0.487

200

0.047

0.071

0.103

0.136

0.203

0.263

0.363

0.472

300

0.035

0.058

0.087

0.119

0.182

0.243

0.343

0.454

500

0.025

0.045

0.071

0.100

0.161

0.219

0.319

0.431

1000

0.016

0.032

0.054

0.079

0.135

0.190

0.288

0.403

Example
The cost of failure for an item is $1000.  The cost of preventive maintenance for this item is $25. The following Weibull distribution parameters were determined from time to fail data: b =2.5, q = 181 days, d = 0. How often should preventive maintenance be done?

Solution
The ratio of failure cost to PM cost is

Cf/Cp = 1000/25 = 40

Entering the table above with this ratio and a shape parameter of 2.5, gives 0.197 for the value of m.

A PM should be done every

T = (0.197)(181) + 0 = 35.7 days.