Helsinki University of Technology Systems Analysis Laboratory INFORMS International, Hong Kong 2006 Supporting Road Asset Supporting Road Asset Management with Portfolio Management with Portfolio Decision Analysis Decision Analysis Pekka Mild, Juuso Liesiö and Ahti Salo Systems Analysis Laboratory Helsinki University of Technology P.O. Box 1100, 02015 TKK, Finland http://www.sal.tkk.fi [email protected]
18
Embed
Supporting Road Asset Management with Portfolio Decision Analysis
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Helsinki University of Technology Systems Analysis Laboratory
Helsinki University of Technology Systems Analysis Laboratory
6INFORMS International, Hong Kong 2006
RPM-FrameworkRPM-Framework
Based on additive weighting model– Relatively simple and widely used
Incomplete information on weights and scores– Information set for feasible values of these parameters
Non-dominated portfolios and project-specific Core Indexes– Robust decision recommendations w.r.t. incomplete information
Framework paper– Liesiö, J., Mild, P., Salo, A., (2006). Preference Programming for Robust
Portfolio Modeling and Project Selection, EJOR (forthcoming, available on-line)
Helsinki University of Technology Systems Analysis Laboratory
7INFORMS International, Hong Kong 2006
Non-dominated Portfolios and Core IndexNon-dominated Portfolios and Core Index
Non-dominated portfolios (NDP)– No other feasible portfolio gives higher
overall value with all feasible weights and
scores
Project’s Core Index (CI) conveys
the share of non-dominated
portfolios which contain the project – Core proj. are included in all NDP (CI=1)
– Exterior proj. not included in any (CI=0)
– Borderline proj. included in some (0<CI<1)
6
3
10
21
7
4
5
8
9
AB
C
A
Ove
rall
val
ue a
t ex
trem
e po
int 1
B
C
E
D
B
Ove
rall
val
ue a
tex
trem
e po
int 2
C
A
D
E
D
Ove
rall
val
ue a
tex
trem
e po
int 3
A
E
C
B
*
**
*
*
*
Helsinki University of Technology Systems Analysis Laboratory
8INFORMS International, Hong Kong 2006
Core Index AnalysisCore Index Analysis
Core Index is a relative measure of how well the project fits
into the portfolio; it accounts for:– Incomplete information on weights and scores
– Project’s cost and competing proposals
– Budget and other feasibility
constraints
Focus on classification– Correlations in project’s
scores lead to larger sets
of core and exterior projects
– Trade-offs not the primary focus
Hundreds of
projects
Multiple criteria
Portfolio-level
constraints
Borderline proj
→ focus
Exterior proj
→ reject
Core projects
→ accept
Computenon-dominated portfolios
Incomplete information
Helsinki University of Technology Systems Analysis Laboratory
9INFORMS International, Hong Kong 2006
Bridge Repair Program DataBridge Repair Program Data
Total of 313 project prosals– Different bridges calling for repair
Full bridge stock from one road district– Equivalent analysis for another district, too, with analogous findings
Bridges in need of repair and rehabilitation – Backlog of bridges at end of their life-cycle
– Pre-emptive maintence not considered here
Repair program for next three years– “Priority bundle”, not scheluded year by year
Helsinki University of Technology Systems Analysis Laboratory
10INFORMS International, Hong Kong 2006
Evaluation Criteria and CostsEvaluation Criteria and Costs
Six criteria indicating urgency for repair– Sum of Damages (“SumDam”)
– Repair Index (“RepInd”)
– Functional Deficiencies (“FunDef”)
– Average Daily Traffic (“ADTraf”)
– Road Salt usage (“RSalt”)
– Outward Appearance (“OutwApp”)
Value functions for measurement data (SumDam & RepInd), scoring tables based on recorded characteristics for the others– Scoring on a scale of 1-5, higher score indicating higher priority
Costs based on standardized condition-class and size
Helsinki University of Technology Systems Analysis Laboratory
11INFORMS International, Hong Kong 2006
Incomplete Weight InformationIncomplete Weight Information
Interval statements and/or (incomplete) rank-orderings– Define a feasible weight set
Rank-ordering:
{SumDam,RepInd}
≥ {FunDef, ADTraf}
≥ {RSalt,OutwApp}
Sum of weights = 1
0.35
0.070.03
0.120.18
0.25
0.00
0.20
0.40
0.60
0.80
1.00
SumDam RepInd FunDef ADTraf RSalt OutwApp
Wei
gh
t
Feasible weight set A feasible weight vector
Helsinki University of Technology Systems Analysis Laboratory
12INFORMS International, Hong Kong 2006
Program ConstraintsProgram Constraints
Budget of 9,000,000€– Based on current annual funding allocated to bridge repairs
Program can contain maximum of 90 bridges– Proxy for limited availability of equipment and personnel etc.
Program must repair the total sum of damages
by minimum of 15,000 units– Merit pay threshold
Helsinki University of Technology Systems Analysis Laboratory
13INFORMS International, Hong Kong 2006
Results (1/2)Results (1/2)
Found more than 10,000 non-dominated programs– Approximative algorithm based on random sampling of weighted
min-max norm and integer linear programming
– Programs’ total costs between 8.75 – 9.00 M€
– All programs contained 90 bridges → binding constraint
Proposals listed by their Core Index value– 39 core, 112 borderline and 162 exterior projects
– Tentative but not binding priority list
– Found practical and transparent by the programming managers
– Scores, costs and other characteristics displayed
Helsinki University of Technology Systems Analysis Laboratory
14INFORMS International, Hong Kong 2006
Results (2/2)Results (2/2)
Consistent with the
earlier system– Condition factors are
still main determinants
of priority
– Use of road salt and the
traffic volume affect
explicitly → appreciated
Adds to current tools– Framework and results
deemed acceptable
BRIDEGES' SCORES
Bridge number and name Core Index DamSum RepInd FunDef ADTraf Rsalt OutwApp Cost