Transcript Research Implications for Construction Management and Education

Knowledge Management
And Its Research Implications In Construction
Management and Education
Presented by Ken-Yu Lin, PhD
Assistant Professor
Dept. of Construction Management
College of Built Environments
1
About the Speaker
• 1993-1997
• 1997-1999
• 2000-2005
• 2005-2006
• 2006-2007
• 2008 to date
B.S., CE, NTU
M.S., CE, NTU
Ph.D., CEE, U of I
Fulbright Exchanged Scholar
Consultant, Ming-Jian Company
Post-doc, CE, NTU
Assistant Professor, CM, UW
2
University of Washington
3
University of Washington
• Fall 2008 Enrollment – 41,405 students
University
Total No. of
Student
Total No. of
Undergraduate
Students
No. of Graduate
Students
NCKU (2007)
21,184
10,451
10,733
UW (2008)
41,405
29,304
12,101
• Ranked Top 16 of World Universities by
Shanghai Jiao Tong University (2008)
4
Dept. of Construction Management
• College of Built Environments
– Architecture
– Construction Management
• 10 full-time faculty members
• 13 lecturers
• 30-ish industry advisory council members
– Landscape Architecture
– Urban Design and Planning
5
Dept. of Construction Management
• Pacific Northwest Center for Construction Research
and Education – Virtual Construction Lab
6
Dept. of Construction Management
• Pacific Northwest Center for Construction Research
and Education – Methods and Materials Lab
7
Dept. of Construction Management
• Pacific Northwest Center for Construction Research
and Education - Education Lab & Collaboration Suites
8
CM Graduate Programs
• M.S. Program in Construction Management
• Distance Learning M.S. Program in
Construction Engineering
• Ph.D. Program in the Built Environments
Last revised 12.21.2008
9
CM Graduate Programs
• M.S. in CM
– Core courses (9 credits)
– CM focus areas (27 or 33 credits)
•
•
•
•
•
Integrated Project Delivery Systems
Sustainable Built Environment
Infrastructure Development
International Construction
Virtual Design and Construction
– Thesis or research paper (9 or 3 credits)
Last revised 12.21.2008
10
CM Graduate Programs
• Ph.D. Program in the Built Environments
– Core Courses (21 credits)
• History, Theory, and Ethics
• Colloquium-Practicum
• Research Methods and Design
– Areas of Study (30 credits)
• Sustainable Systems and Prototypes
• Computational Design and Research
– Examination/Dissertation (30 credits)
11
What is Knowledge?
• Data, Information, Knowledge
– “Данные”
– “82%”
• Valuable information for the human mind (Davenport,
1997)
• Information only becomes knowledge when it is put
into a logical and understandable context which can
be verified and recalled from human experiences
(Gunnlaugsdottir, 2003)
Last revised 07.15.2009
12
What is Knowledge Management?
• Knowledge Classification
(Nonaka and Takeuchi, 1995)
– Explicit
• Knowledge that can be articulated in formal languages
– Tacit
• Subtle level of understanding often rooted in practice,
expressed through skillful execution, and transmitted
by apprenticeship
13
Data Mining
Knowledge-assisted
Information Retrieval
Ontology
Knowledge Sharing
and Reuse
14
Data Mining for Construction
15
Data Mining for Construction
• Fort Wayne, IN - Flood Control Project*
– Four phases
– Significant project delay (54%) for the
6”-42” drainage pipeline installation
– Main sub-activities
•
•
•
•
Excavating the ground
Installing pipelines
Erosion protection
Backfilling compacted material
* Source - Soibelman, L, and Kim, H., “Knowledge Discovery for Project Delay Analysis”
Bauingenieur, Springer VDI Verlag, February 2005.
16
Data Mining for Construction
Understand
and Define
the Problem
Collect
Data
Explore
Data
Enhance
Data
Original Data
Pre-Processing Data
Post-Processing Data
Pre-computing the Data
Clean
Data
Select
Data
Attribute
Mine
the
Data
Data Base Systems
AI Tools
Data
Analysis
Evaluate
the
Result
Assess
Model
Visualization Tools
Decision Support
Expert Knowledge
Figure 1. Data mining framework
17
Data Mining for Construction
Inaccurate site survey
appeared to be more
influential
Weather was not the
main cause
Figure 2. Analysis for the Data Mining Example Project
18
Data Mining for Construction
• Predicting Activity Duration
– Adjusting the duration for 320-unit drainage pipe
installation with 10 available workers
• Empirical Data from RS Means
– Industry productivity average = 10 units/worker/day
– 3.2 days
• Neural Network prediction
– 4.96 - 6.86 days
• Monte Carlo Simulation
– 3.51 – 6.85 days
19
Data Mining
Knowledge-assisted
Information Retrieval
Ontology
Knowledge Sharing
and Reuse
20
Knowledge-assisted Information Retrieval
Legend
UF: Use-for Term
BT: Broader Term
RT: Related Term
Skylight
BT
Natural
light
RT
UF
Daylighting
panels
Translucent roof
panels
UF
Translucent
roof assemblies
RT
Fiberglass
UF
Translucent
roof systems
Figure 3. An example knowledge representation
21
Knowledge-assisted Information Retrieval
Grow
Query expansion Query expansion
by lexical
by semantic
components
components
Trim
Pooling
Figure 4. Illustration of the crawling and ranking strategy
22
Knowledge-assisted Information Retrieval
• P-Norm (Salton et. al., 1983)
Daylighting
(0,1)
Goal (1,1)
Daylighting
(0,1)
d1
d3
(0,0)
(1,1)
d1
d2
Panels (1,0)
d2
Null (0,0)
Panels (1,0)
Figure 5. Illustration of the ranking model
23
Knowledge-assisted Information Retrieval
• Performance : # of Relevant Documents Found
35
30
25
20
This Research
15
Google
10
5
0
Data set Data set Data set Data set Data set
1
2
3
4
5
Figure 6. Performance evaluation in terms of the # of relevant documents found
24
Knowledge-assisted Information Retrieval
• Performance: # of Manufacturers Found
40
35
30
25
This Research
20
Google
15
SWEETS
10
5
0
Data set Data set Data set Data set Data set
1
2
3
4
5
Figure 7. Performance evaluation in terms of the # of manufacturers found
25
Knowledge-assisted Information Retrieval
Last revisedAcquisition
07.15.2009
Figure 8. Information
Took Kit
26
Data Mining
Knowledge-assisted
Information Retrieval
Ontology
Knowledge Sharing
and Reuse
27
What is Ontology?
• According to Gruber 1993
– A formal, explicit specification of
conceptualization
– An agreement to use a vocabulary in a way that is
consistent with respect to the theory specified by
an ontology
28
Identify Purpose
and Scope
Determine the domain
and scope
Reuse existing ontologies
Ontology Capture
Identify Concepts
Produce Definition
of concepts
Identify Entities
and Relations
Enumerate important terms
Define the classes
and the class hierarchy
Define the property of classes
Define the facets of the slots
Ontology Coding
(Method by Uschold and Gruninger, 1996)
Create instances
(Method by Noy and McGuinness, 2001)
Figure 9. Ontology Development Procedure Exemplars
29
Objective 1. Domain definition
Task 1.1 Determining green purchasing processes
Step 1.1 Mapping purchasing procedure
using the process mapping method
Task 1.2 Identifying needed information
Step 1.2 Designing and implementing
Task 1.3 Specifying how project stakeholders
questionnaire survey
gather and utilize the information
Objective 2. Green purchasing semantic model development
Step 2.1 Defining model scope
Step 2.2 Exploiting existing ontologies
Task 2.1 Reusing existing ontologies
Step 2.3 Deriving concepts/terminologies
Task 2.2 Building green purchasing taxonomies
Step 2.4 Structuring concept hierarchies
Task 2.3 Specifying concept restrictions and
enumerations
Step 2.5 Assigning concept restrictions
and enumerations
Task 2.4 Applying taxonomy concepts to model
the green purchasing information needs
Step 2.6 Semantic modeling
Step 2.7 Model coding and visualization
Objective 3. Model evaluation and improvement
Task 3.1 Completeness evaluation
Step 3.1 Scope and scenario verification
Task 3.2 Consistency evaluation
Step 3.2 Logic checking via reasoner
Task 3.3 Conciseness evaluation
Step 3.3 Interviews and surveys with AEC
practitioners
Task 3.4 Iterative model revision
Step 3.4 Feedback and model revision
Figure 10. Green Purchasing Ontology Development Procedure
30
Figure 11. Developing Domain Ontology Using Domain Handbooks
(*Source: H.T. Lin, S.H. Hsieh, K.W. Chou and K.Y. Lin, “A Statistical Method for Constructing
Engineering Domain Ontology through Extraction of Knowledge from Domain Handbooks”, Advanced
Engineering Informatics, under review. )
31
Data Mining
Knowledge-assisted
Information Retrieval
Ontology
Knowledge Sharing
and Reuse
32
KM in CM Education
• Drowning in Data
– How Could Construction Students Find Useful Online
Resources for Learning and Researching Needs?
• Knowledge Is Not Power, Sharing Is
– How Can One Student Share the Captured Construction
Online Resources with Other Students Who Have Similar
Learning or Researching Interests?
33
KM in CM Education
A
Resource Y
C
Concept Z
Resource X
B
Figure 15. Learning through socialization
34
KM in CM Education
Domain
Knowledge
Setting the
Contexts
Online
Resources
Searching
Resource
Repository
Socializing
Knowledge
Coding
Figure 16. Sustainable e-learning model
35
KM in CM Education
• A Pilot Study
Basic Electrical
Concepts: Volts,
Amps, Ohms
Wiring
Components: Hot,
Neutral, Ground
Current Kills,
Not the Voltage
Common Hazards:
Shocks, Burns,
Explosions, Fire
Breaker for
Equipment
Protection
Safety
Prevention
Reverse Polarity
Using GFCI Outlets
for Personnel
Protection
Grounding
Tool Maintenance
and Inspection
Figure 17. e-Learning exemplar on construction S&H
36
KM in CM Education
• Feasibility Study
– Electrical safety in
construction
– Students selected 10
concepts to plot mind
maps and then
furnished each
concept with 3 online
resources
Figure 18. Data Processing Interface
37
KM in CM Education
• Preliminary Data Analysis
– Responses from 28 Students Were Analyzed
– 717 Online Resources Were Identified
– 363 Online Resources Were Reused
– Most Cited
• OSAH, Wikipedia, YouTube, WA L&I, and CDC/NIOSH
38
KM in CM Education
Criteria
Effective (before adjustment)
Cited by more than one student
(before adjustment)
Number of Links
717
363 (= 50% of 717)
Effective (after adjustment)
Cited by more than one student
(after adjustment)
Unique
599
154 (= 25% of 599)
528
39
Thank You
40