24
Estimation of crew demand in S-tog Agenda ƒ The planning process ƒ Motivation for the model ƒ What is the purpose of the model ƒ Input to the model ƒ Objectives ƒ Experiments and conclusion ƒ Future work Michael Folkmann Operational Researcher Production Planning, DSB S-tog

Estimation of crew demand in S-tog Agenda ƒ The planning process ƒ Motivation for the model ƒ What is the purpose of the model ƒ Input to the model ƒ Objectives

Embed Size (px)

Citation preview

Estimation of crew demand in S-tog

Agenda

ƒ The planning processƒ Motivation for the modelƒ What is the purpose of the modelƒ Input to the modelƒ Objectivesƒ Experiments and conclusionƒ Future work

Michael FolkmannOperational Researcher Production Planning, DSB S-tog

Process of the (sequential) planning

Number of Passengers

Time Table

Draft of Product

Economystructure

Evaluation

Rolling Stock plan

Crew plan

New product and changes in the product.

Basis plan

Planning period ≈ one year

Track

Work

Track

Work Track

Work

Track

Work

Track

Work

…….. … Track

Work

Motivation for the model

Good early estimation of crew costs in reasonable time for a plan

Crew plans were time consuming

Crew targetƒ Train drivers

ƒ First generationƒ Needs Rolling Stock Planƒ Duty planning with Turni integrated ƒ Adjustment for track works

ƒ Second generationƒ Change the input

ƒ Other groups of personnelƒ No Rolling Stock Plan

What is the purpose of the model

Breifly:Cover a workload with duty template in relation to a number of constraints

DayTypesƒ Weekdaysƒ Saturdayƒ Sunday

Size of time intervals; 15 minutes

Inputsƒ Workload

ƒ Coming from Rolling Stock Plan which needs a driverƒ Duty templatesƒ Parameters for constraints

Use of the resultsƒ Not actual planƒ Crew demand

Workload

Fictitious Workload

0

10

20

30

40

50

60

03:0

0

3:45

4:30

5:15

6:00

6:45

7:30

8:15

9:00

9:45

10:3

0

11:1

5

12:0

0

12:4

5

13:3

0

14:1

5

15:0

0

15:4

5

16:3

0

17:1

5

18:0

0

18:4

5

19:3

0

20:1

5

21:0

0

21:4

5

22:3

0

23:1

5

24:0

0

24:4

5

25:3

0

26:1

5

Time

Rollin

g S

tock

s

Duty Templates

ƒWorking dayƒAmount of drivingƒCheck In/out – work – Breaks

ƒSix different template variationsƒBy hand or with model

ƒFractionel

Template Type 1

0

0,2

0,4

0,6

0,8

1

1,2

00:0

0

00:3

0

01:0

0

01:3

0

02:0

0

02:3

0

03:0

0

03:3

0

04:0

0

04:3

0

05:0

0

05:3

0

06:0

0

06:3

0

07:0

0

Time

Am

ount

Duty Templates variants

ƒ For all six templates, i.e. three sets with six templates

Template Type 2

0

0,2

0,4

0,6

0,8

1

1,2

00:0

0

00:3

0

01:0

0

01:3

0

02:0

0

02:3

0

03:0

0

03:3

0

04:0

0

04:3

0

05:0

0

05:3

0

06:0

0

06:3

0

07:0

0Time

Am

ount

Template Type 3

0

0,2

0,4

0,6

0,8

1

1,2

00:0

0

00:3

0

01:0

0

01:3

0

02:0

0

02:3

0

03:0

0

03:3

0

04:0

0

04:3

0

05:0

0

05:3

0

06:0

0

06:3

0

07:0

0

Time

Am

ount

Difference in Template Types

Template Type T1

Template Type T2

Template Type T3

Working TimeNo break Utilization % Utilization % Utilization %

Template 1 07:15 2 71% 74% 79%

Template 2 06:30 2 69% 73% 77%

Template 3 06:15 2 64% 66% 72%

Template 4 07:15 1 73% 83% 83%

Template 5 06:45 1 77% 79% 81%

Template 6 07:15 2 67% 69% 71%

Working Rules to be modelled

Each Duty - Inputƒ Length of break in dutiesƒ Duty lengthƒ Length of driving blockƒ …

All Duties - Constraintsƒ Average working time per dutyƒ Average breaks per duty

Variables

Check-in cIn

Duty Template Dt

Number of train drivers check-in in interval cIn following template Dt

x(cIn,Dt) - Integer

Average Working Time per Duty

Total amount of working time divided by the number of duties

Reformulated:

AverageWTDtcInx

DtlengthDtcInx

Dt cIn

Dt cIn

),(

))(()),((

0)))(()),(( Dt cIn

AverageWTDtlengthDtcInx

Average Breaks per Duty

Reference solution ƒ241 dutiesƒBreak Average 1.722 (=AB)

ƒTotal number of breaks in duties divided by the number of duties.

ƒReformulated constraints with variation

0))1()(()),(( Dt cIn

VariationABDtbreaksDtcInx

  One break Two breaks

  (Total 241 duties)

Variation At most At least

10.00% 109 133

5.00% 88 153

4.00% 84 157

3.00% 79 162

2.00% 75 166

1.00% 71 170

0.10% 67 174

0.01% 67 174

0.00% 67 174

Model Objective

ƒMinimise the total amount of working time in the solution (Objective1)

ƒMinimise the total number of templates used (Objective 2)

cIn Tj

cIn,Dt)x(min

Dt cIn

DtlengthDtcInx ))(()),((min

Adjustment and Experiments

Decisionƒ Size of time interval

Adjustmentsƒ Importance of the average breaksƒ Importance of the templates types

Experimentsƒ Two objectivesƒ Gap - solution quality

Runs

One daytype for one plan – one reference solution

Two fictitious plan – relativ evaluation

ƒ Running time 6:00 hoursƒ Zero gap tolerance

ƒ Cplex 8.1.0 (using Gams 2.0.23.10)ƒ Windows 2000ƒ Pentium M 1700 MHzƒ 2GB ram

Average Breaks and Templates

Three Templates Types - Objective 1

1500

1550

1600

1650

1700

0,01% 0,10% 1,00% 2,00% 3,00% 4,00% 5,00% 10,00% No limit

Average Break variation

No h

ours T1

T2

T3

Average Breaks and Templates

Three Templates Types - Objective 2

205

210

215

220

225

230

235

240

245

0,01% 0,10% 1,00% 2,00% 3,00% 4,00% 5,00% 10,00% No limit

Average Break variation

No D

uties

T1

T2

T3

Quality of solutions

Gap in solution - Objective 1

0,00%

0,50%

1,00%

1,50%

2,00%

2,50%

3,00%

3,50%

0,01% 0,10% 1,00% 2,00% 3,00% 4,00% 5,00% 10,00% No limit

Average Break variation

Gap

T1

T2

T3

Quality of solutions

Gap in solution - Objective 2

0,00%

0,50%

1,00%

1,50%

2,00%

2,50%

3,00%

3,50%

4,00%

4,50%

5,00%

0,01% 0,10% 1,00% 2,00% 3,00% 4,00% 5,00% 10,00% No limit

Average Break variation

Gap

T1

T2

T3

Gap progress

Running time for (almost) no progress in gap compared to 360 minutesƒTemplate Type T1ƒBreak Average 0.1% and 1.0%

Break Average variation

0.10%Break Average variation

1.0%

Objective 1 325 minutes 45 minutes

Objective 2 81 minutes 1.25 minutes

Results and conclusion

ƒ Decreasing in objective – as expected

ƒ Two Objectives

ƒ Average Breaks – important but not strict

ƒ Running time

Have initial parameter setting:ƒ Minimising number of duties with T1 duty templates with 15 minutes intervals and Average Breaks variation 1.0% for a short running time

Future work

More reference solution from Turni

Different templatesƒ effect of small changesƒ fewer/more templates

Develop of the template model

Other constraintsƒ difference during the dayƒ limited number of specific templates

Change the model into second generation – i.e. Public Time Table as input

Questions?