labor production rates and the sensitivity of i d e n s e d influencing factors. ...
both risk analysis and sensitivity analysis of construction labor productivity. The.
University of Alberta
Productivity Studies Using Advanced ANN Models BY
Ming Lu
@
A thesis submitted to the Faculty of Graduate Studies and Research in p h a l fdultillment
of the requirements for the degree of Doctor of Philosophy
in Construction Engineering and Management
The Department of Civil and Environmental Engineering University of Alberta Edmonton, Alberta, Canada Spring 2001
National Library
Bibliathèque nationale
du Canada
Acquisitions and Bibliographie Services
Acquisitions et services bibliographiques
395 Wellington Street Ottawa ON K1A ON4 Canada
395. rue Wellington Ottawa O N K IA ON4
Canada Your file votre rëférence
Our fi& Notre rdfdrence
The author has granted a nonexclusive licence allowing the National Library of Canada to reproduce, ioan, distribute or sell copies of this thesis in rnicroform, paper or electronic formats.
L'auteur a accordé une licence non exclusive permettant a la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/fih, de reproduction sur papier ou sur format électronique.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fkom it may be printed or otherwise reproduced without the author's permission.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Estirnating labor productivity is one of the most difficult aspects of prepming an estimate, or a control budget based on the estimate for labor-intensive activities in construction. The primary objective of research is deveioping artificial neural neavork or
ANN based C$i.Kna~g tools
CO
offer estimators valuable information about labor
productiviv in bidding nem jobs.
In conjunction with a major Canadian industrial contractor, the thesis research presents case studies on the theoretical basis and practical considerations for m e a s k g and analyzing labor productivity in industrial construction. Two important activities of process piping mere investigated: pipe installation in the field and spool fabrication in the fabrication shop. Emerging cornputer modeling techniques such as data warehouses and
ANN were researched fiom an academic perspective and irnplemented in industry to meet the challenges in productivity studies. The thesis research has addressed: (1) how to quannfv labor productivity in indusmal construction Erom a contractor's point of view; (2) how to measure actual labor productivity in industrial construction based upon onsite control practices; and (3) how to utilize ANN to analyze the vmiability of actual labor production rates and the sensitivity of i d e n s e d influencing factors.
Using actual data, the proposed ANN models were proven to be effective in both risk analysis and sensitivity analysis of construction labor productivity. The developed data warehouses and - W - b a s e d decision-support tools have been
lmplemented or are in the process ofimplementation at the involved Company. The h a 1 results of the research not only assist estîmaton to improve the accuracy of e s t i m a ~ g labor production rates for studied activities in biddlig new jobs, but also offer the management a precise and integrated view of corporate productiviq information
spanning across many business divisions. The esperience and lessons levned Gom the successful, productive and rnumally beneficial collaboration benveen academh and industry in the diesis research wX potentialiy benefit other universiq-indusq joint research projects in the hture.
This thesis is organized in a paper format, consishg of five main chapters and five appendices. Every chapter is an independent paper and c m be read separately. However, î11 the chapters are logically coherent and pertinen~to the theme of thesis. Each ;ippendix is a user manual for one computer program that was developed in house
in the thesis research. Chapter 1 o v e ~ e w the s mhole thesis by introducing background information, problem statements, research objectives, methodologies used, and contributions achieved. Chapter 2 discusses a case study of industrial construction Iabor productivïty, which depicts the settings of the research. Chapter 3 presents a probabilistic neural nenvork classification model along with its application in estimating the production rates of field pipe installation. Chapter 4 presents a sensitivity analysis method of back propagation neural networks along with its application in estimating the production rates of shop spool fabrication. Chapter 5 surnmarizes what has been done thus far and recomrnends what to do in the future research. Appendk A is for the PINN trainer program based on the mode1 described in Chapter 3. Appendis B is for the FabMaster program, which is the data marehouse for the fabrication facilities. Appendis C is for Fab-OLAP, which is an on-line analytical processing program in cornpanion wïth FabMaster. Appendïx D is for the PipinghIaster program, whïch is the data warehouse for the field construction systems.
Appendks E is for the SensitiveNN
program based on the model as described in Chapter 3.
First and siacerely, 1 would like to thank my universitg adssor, Dr. S. bI. AbouRizk, mithout whose visions, guidance and encouragement this academic achievement would not have become a redty.
Especially, 1 would like to ha&
my industry advisor, b k U. H. Hermann of
PCL, whose professionalïsm and enthusiasm have set the pace and the standard for the mhole work. 1 am exaemely gratefd to PCL Industriai Constructors, Inc. for s p o n s o ~ g the research hnancially and allowing me to use its actual data for developing problems and validaMg solutions diroughout the thesis.
Findy, 1 wodd like to dia& my wife, Duojia, for her understanding, love and assistance in making this thesis fiom thoughts to finish. 1 dedicate this work to her and o u arriving W u a n Hum.
Table of Contents
.................................................................................................................
CHAPTER 1 INTRODUCTION 1 BACKGROUNDS .......................................................................................................................................... 1 Industrial Construction ....................................................................................................................... 1 Prodrictivity Strtdies .............................................................................................................................. -3 Prodrictiviry Modeis ............................................................................................................................. 3 An@cial Neural Nehvorks .................................................................................................................... 5 .............................................................................................................................. PROBLEM STATEMENTS 8 Procirictivity Stridies .............................................................................................................................. S A NN Models ........................................................................................................................................ 13 RESEARCH OBJECI-IVES ............................................................................................................................ 15 Prodrictivity Strtdies ............................................................................................................................ 16 Probnbilistic Neural Network Modeling ............................................................................................. 16 16 Sensitivity Analysis of Neural Nehvorks.............................................................................................. M ~ O D O L O G ..................................................................................................................................... ES 17 Reviewing Literatrtre to Recognize Issues ........................................................................................... 17 Identihing Factors froni Brairlstorrriing by Donrairz Erperts .............................................................. 17 Using Data Warehorlse tu Qrrantitative Data ..................................................................................... 18 Qriestionnaire Srirvey .......................................................................................................................... 19 21 Cornputer Progrnmniing ..................................................................................................................... AcADEMICCONTR[B UTIONS ..................................................................................................................... 21 22 INDUSTRIAL CONTRBUTIONS................................................................................................................... CONCLUS~ONS .......................................................................................................................................... 23 ............................................................................................................................................... 23 REFRENCE~ CHAPTER 2 A CASE STUDY OF INDUSTRIAL CONSTRUCTION LABOR PRODUCTIVITY INTRODUCTION ......................................................................................................................................... .................................................................................................................... INDUSTRIAL CONSTRUCTION FIELD PIPEINSTALLATION ........................................................................................................................ Prodrictivity Qiianrificcntion................................................................................................................. Producrivity Measrrrerrtenr .................................................................................................................. Input Factors ....................................................................................................................................... Probabilistic neriral nenvork trrodeling ............................................................................................... ....................................................................................................................... SHOPSPOOL FABRICATION Prodrictivity Quantification ................................................................................................................. Prodrtctivity Measurernent .................................................................................................................. lnprit Factors ....................................................................................................................................... Sensitivity Analysis of Ittflrienci~zgFactors ......................................................................................... CONCLUSIONS .......................................................................................................................................... REFRENCES ...............................................................................................................................................
.....................................................................................................................................
28 28 30 32 33 34 36 40 42
43 46 47 51 57
58
CHAPTER 3 ESTIMATING LABOR PRODUCTIVITY USING PROBABILITY INFERENCE NEURAL NETWORK 6 0 ....................................................................................................................................... INTRODUC~ION 6 0 Probleni Dorrlairz ................................................................................................................................. 61 Review of NN Applications .................................................................................................................. 64 P R O B A B INFERENCE ~ ~ NEURALL'WWORK ( P I W ) MODEL ..................................................................... 64 Introduction of the PINN Mode1 ......................................................................................................... 64 Overview of the PlNN Topology and Process ........................ . . . ...................................................... 67 Data Pre-Processing ........................................................................................................................... 70 Orrtpilt Zone Setzip............................................................................................................................... 74 Processing Elements ( PE) crt Kotionen Layer ................................................................................... 74 NN Learning Process .......................................................................................................................... 75
...........................................................................................................................
List of Tables Table 2-1: Sample of pipe installation unit labor rates ............................................................. 33
...................................................... Table 2-2: Input factors to pipe installation productioi~
39
Table 2-3: Sample of degree-of-difficulty factors for converting welds into units ............... 4 4 Table 2-4: Explanatory factors ro spool fabrication productivity ........................................... 50 Table 3-1: Input Factors and Data Type of PINN Mode1...................................................... 72 Table 3-2: Inpur Data Sample of PINN Mode1.........................................................................
73
Table 3-3: Scaled Input Vector and Initial Weight Vectors ...................................................
78
Table 3-4: Updating Kreight Vectors in Firs t Leaming Stage ..................................................
79
Table 3-5: Updathg Weight Vectors in Second Leaning Stage............................................... 81 Table 3-6: Trained PINN Ready to R e c d for A Given Input Vecmr ............................... ....85 Table 3-7: Recall Calculations at Bayesian Layer ....................................................................... 86 Table 4-1 : Data Set for Testing B P N N and Regression Analysis ......................................... 117
aN1
Table 4-2: Pan5al Derivative (Slope) (-)
NP
at Four Input Points ...................................... 119
aN. ........................................
Table 4-3: Statisucs of Partial Derivative (Slope) Values: (-)
~ S P
119
Table 4-4: Input Factors of Spool Fabrication Labor Productivity .................................. 128 Table 5-1: PINN vs . BP NN ......................................................................................................
144
Table B-1: Size Range Codes ...................................................................................................... 159 Table B-2: Material Type Codes.................................................................................................
160
Table B-3: Item Codes for Spool Level Data Compilation ...................................................
160
Table B-4: Sarnple cf Fabhlaster Outputs ................................................................................
170
Table D-1 : Sample of Quantity Calculation Summary Table in PipinghIaster ................... 190
List of Figures Figure 1-1: Sample Q u e s h o ~ a i r efor Findïng Facts about Spool Fabrication .................... 20 Figure 2-1: Output interface of PINN recall program .............................................................
41
Figure 2-2: Sensitivity Analysis of Spool Fabrication BPNN Mode1.................................... 51 Figwe 2-3: Tes ting Sensitivity of BPNN to Matenal Type ...................................................... 57 Figure 3-1: Topology of PINN Mode1........................................................................................
66
Figure 3-3: Operations at Bayesian Layer in R e c d ................................................................... 80 Figure 3-3: Cornparison of PINN and Back Propagation NN .......................................... 88 Figure 3-4: PINN Output for the Base Case Scenario ............................................................. S9 Figure 3- 5: PINN Output for Scenario 1..............................................................................
91
Figure 3-6: P W Output for Scenario 3 ................................................................................... 92 Figure 4-1: Stnicme of Back-Propagation NN Mode1.......................................................... 103 Figure 4-2: Illustration for Node and Laper Representations ................................................ 108 Figure 4-3: Distributions for Input Sensitivity.....................................................................
120
Figure 4-4: Sensitivity Analysis of Spool Fabrication BPNN Mode1.................................... 123 Figure 4-5: T e s ~ Sensitivïty g of BPNN to Material Type .................................................... 132 Figure A-1 : Select an identifier key of one previous mal ........................ ......
.................. 147
Figure A-2: User selects data table.............................................................................................
149
Figue A-3: Flag status of records ..............................................................................................
150
Figure A-4: Setup structure and leaming parameters for PINN .......................................... 152 Figure A-5: SpecZy training iteraüons and train-test PINN .................................................. 153 Figure A-6: Check training results .............................................................................................
151
. . Figure A-7: Detected noise m training data..............................................................................
155
Figure A-8: Global report for a train-test t r i a l ........................................................................
156
Figure A-9: PINN Trainer on-line help ....................................................................................
157
Figure B-1: Program Flow Chart of FabMaster ...................................................................... 165 Figure B-2: Main User Interface of FabMaster........................................................................
128
Figure C-1: Select one ratio.........................................................................................................
181
Figure C-2: T d on "number of pipe pieces per foot" ..........................................................
182
Figure C-3: View details of data ........ .........................................................................................
183
s Raw Data Tables for A Project ..................................................... 185 Figure D-1: S t r u c ~ e of Figure D-2: Main user interface of FabMaster ........................................................................
186
Figure D-3: Handikg: S-Reference Information Inte@o/ Check ....................................... 188 Figure D-4: Welding: X-Re ference Information In t e g i t y Check ....................................... 189 Figure D-5: Productivity Analvsis Page for Pipe HandlLig .................................................. 191 Figure D-6: Sample of Pipe Handling Ques tiomaire ............................................................. 192 Figure D-7: Program Flow Chart of Pipinghfas ter ................................................................. 193
Figure E-1: Splash Screen of SensitiveNN program ...............................................................
195
Figure E-2: Program Switchboard.............................................................................................
197
Figure E-3: Open FFBPNN-mdb First ..................................................................................... 197 Figure E-4: Select data source table ........................................................................................... 198 Figure E-5: Examine details of data and edit record status ................................................... 199 Figure E-6: Program main interface of SensitiveNN............................................................. 2 0 1
Figure E-7: Check leaming results when NN training temiinates ........................................ 202 Figure E-8: Check s Input Sensitivitp for each input-output pair ........................................ 204
Chapter 1: Introduction
Industrial Construction Barrie et al (1992) described industrial construction as:
"Indusmal construction covers a wide range of construction projects that are essential to our ualities and basic industries, such as petroleum r e h e n e s and petrochemical plants, synthetic fuel plants, fossil fuel and nuclear power plants, off sbore
oil/gas
production
facilities, cryogenic plants
etc.
Industrid
construction generalIy Çeatures large amounts of hghly cornples process piping, mechanical, electrical, and instrumentation mork; bo th design and consuuction require the highest level of engineering expertise Gom multiple disciplines."
In particular, the installation of process piping systems in indusuial construction is selected for productivïty studies because it accounts for the bulk of direct labor hours of an ind~soialcontractor. Process piping is used to transport fltids between storage tanks and processing units. Instdation of piping systems generally consists of nvo processes: (1) spool fabrication in a commercial pipe shop; (2) pipe installation in the field. Although the trvo processes are inseparable and can be integrated to optimize the econornics of a particular situation, they are treated independent of each otl~erin the
thesis because of the cuxent estimaMg and control practices of the involved company. The productivity studies described in the thesis are conducted to support the management's decision-rnaliing in the context of the company's m e n t management systems, as opposrd to radically changing these systems.
Productivity Studies In a construction task that is performed by hand labor, productivity is commonly espressed as the labor production rate (man-hours per installed unit), which measures a key dimension of performance and is a critical factor to estimating, scheduling and control of the project (hlfeld, 1988). Little information could be found in literature on the theoretical basis and practical considerations for measuring and analyzing labor
productivity of indusmial construcuon. In conjunction wïth a major indusnial concractor (refened to as "the company" hereafter), productivity smdies were conducted for tsvo Lnportmt activities in industeid construction: pipe installation in the field and spool fabrication in the fabrication shop.
In general, productivity studies encompass
three tasks:
(1) developing special
, to measure methods and techniques to quanti@ labor productivity for e s ü m a ~ g and actual labor productivïty for on-site control, (2) identifjmg input factors that cause the vmiability in productiviq, and (3) analyzing the relationships benveen input factors and productivitg to enhance the accuracy of productivity estimating or Mprove the on-site performance dkectly. The focus of investigation is the average labor production rates (man-houn per unit) of these activities 3t the end of a project, rather than the d d y hbor production rates, because the primary objective of reseaech is developing ANN-based
estimating tools to offer estimators valuabie information about labor productivity in bidding new jobs, rather than assessing and improving the crem performance in the field.
Productivity Models Several established models for studying productivity c m be found in the literature, including work study techniques, expectancy model, action-response model, regression model, expert systems, and arnhcial neural networks (NN).
Work-study techniques were adopted in a nurnber of productivity models, in which only a fenr factors related to work method were included (Thomas and D d y , 1983). Such work-study models cannot be used to model esternai and management factors. Thomas et al. (1990) and Thomas et ai. (1991) discussed additional drawbacks of work-study techniques for construction productivity modeling.
The Espectancy model and action-response rnodel are tsvo alternative techniques proposed to exphin variations in construcuon productivity. In the elrpectancy model, the effort that an individual is &g
to evert accounts for the ciifferences in job
performance or productivity (Maloney and McFillen 1985). The action-response model graphically depicts the interaction of a number of factors that lead to the loss of productiviq- (Halligan et. al. 1994). Both models contribute to understanding the variations in productivity; however, neither can be used to quanufy the influences of
multiple factors on construction productivity (Sommez and RoMngs, 1998). Sanders and Thomas (1993) developed an additive linear regression model to study the effect of sis project-related variables on masonry productivity based on data
obtaioed fiom 11 projects. Eight binary variables mere used in the model to represeat the variations in productivity due to temperature and hurnidity. The effect of crew size was also taken into account in the model. The results of this regression model suggested higher productiviq rates for crew with femer members. Thomas and Sakarcan (1994) conànued the reseaxch of Sandea and Thomas (1993) by developing the additive Linear regression model for the purpose of forecasckg labor productivity. They only included job condition variables chat describe the work content and the physical components of the work. The focus of both studies was to determine the coefficient of condition variables, or the effect of a present condition on the activity productiviry rate based on the results of historical study; such coefficients were derived independendy of other inputs vithout accounting for combined effects. In addition, the determined coefficients are constants based upon the average values of historical data, and do not reflect the real situations in wiiich the values of such coefficients may vary with the specific job conditions.
Esqxrt systems is another technique applied co model labor productivity in
trvo
studies found in literature. Hendnckson e t al. (1987) developed an hvo-stage espert system named ''MASON" to estimate acuvity durations for m a s o q construction. First, the maximum espected productivity was estirnated. Nest, this rate was adjusted for various characteristics of job or site. The masimum productivity estimates and the followhg adjustments mere based o n the knowledge obtained Erom interviews with a professional mason and a supportkg laborer. Christian and Hachy (1995) developed an esTert system to estimate the production rates for concrete pouring. The expert system relied on the knowledge extracted Erom experts and data coLlected £rom seven
construction sites. The user simply queried the expert system for an estimate through a question-and-ansmer routine. In both e-upert systems, productiviq mas estimated through previously dehned decision d e s obtained from domain experts. Because the nature of forrnulating rules is subjective, the resultant rules may be inconsistent. h o t h e r disadvantage of analyzing productivirp based on expert systems is expert systems do not perforrn Functional input-output mapping, i.e. quantitative evaluation of the impact of job condiuons on productivity.
In the follow-.ï.ng subsection background information about r \ N N models d be introduced and the technique of modeling productirity using IWN d be discussed.
Artificial Neural Networks htïficial Neural Networks (ANN) research involves multiple disciplines including biology, xtificial inteiligence, cornputer science, and mathematics and evolves
with the developments in each related discipline. Kohonen (1995) dehned MJN as
"
massively pardel htercomected network of simple (usudy adaptive) elements and their hierarchical organizations, intended to interact with the objects of the real world in the same way as the biological nervous systems do."
Sirnply put, an A N N mode1 is an
analytical mode1 that sirnulates the cognitive learning process of the h u m a . brain, and is automaticdy constructed feom leaming esamples or data by
ttid
and error Gthout
heuiristic design or other human intervention.
ANN deds effkctively mith ill-structured problems, in which the algorithms required to solve them CaMOt be given in a precise and explicit fashion, or the data for a
partïcular problem are either not complete or cannot be s p e d e d preasely (Widman et. al., 1989). ANN has been found to be capable of perfomilig pardel computations on
different tasks, such as pattern recognition, 1inea.r optimization, speech recognition, and predicuon (Mukhejee and Deshpande 1995). In short, the s p e d leaming algonthms of
ANN are capable of performing hg.
dimensional, non-lineu input-output mapping and
extracMg hidden patterns and predictive information from observing the leuning esamples.
In recent years, ANN has been rescarched and applied as a convenient dedsionsupport tool in a rarieq of application areas in civil engineering, including modulv l (Flood and consmiction decision making &Iurt;iza and Fisher, 1993), s t n i c ~ a analysis Katirn, 1994), e s t i m a ~ gconstruction productivity (Portas and AbouRizk, 1997), mode choice analysis of beight mansport market (Sayed and Razavi, 1999), construction m&p
estünating (LIet al 1999), measurkg organizauonal effecùveness (Sinha and
hfcI
Total Position \Veld / Design Weld Vol.
O
AS (Moy)
Total Roll Weld / Design \Veld DiaIn
1
AS (Moy)
Total Roll Weld / Design Weld Equiv.DiaIn
1
-AS (rvloy)
Total Roll Weld / Design Weld Vol.
1
AS (Moy) AS (Noy)
T o d SV7 1'01.
AS ( M ~ Y ) T o d Mulu-Staaon Roll Weld / Design \Veld DiaIn AS (Moy) Total bluia-Station Roll 'LVeld / Design WeId Equiv-DiaIn
S.34724GE-03 8.347246E-03 2.78841 GE-O3
0.4257095 0.4257095
AS W o y )
Total Uula-Station Roll Weld / Design Weld Vol.
AS (May) -4s (Moy)
Total
Single-Station Roll Weld / Design Weld DiaIa
0.5743905
r o ta1 Single-Station Roll Weld / Design Weld Equïv.DiaIn
0.5742905
AS (Ailey)
Total
Single-Stauon RolI \Veld / Design \Veld TroL
0.567347
AS (Moy)
rotal
- ï'ig Process Weld / Design WeId Vol.
0.432653
0.2067211
Material AS (Moy) AS (Noy)
Ratio Description Size Totai Mig Process \Veld / Design Weld Vol.
Total F=\V
Process Weld / Design Wdd Vol-
AS W o y ) AS (Moy)
Total Stick Process !Veld / Design Wdd Vol.
AS (Mo).)
Total Rotoweld Process Weld / Design Weld 1701-
AS (Mo).)
Total Repait: Rate (No. of R / R+A)
AS (Moy)
Total No. of Cut Sheet Revision / No. oESpool
AS (rvloy)
Total RT rate /Spool
AS (Moy)
Total MT rate /Spool
AS (Aioy)
Total PT rate /Spool
AS (Moy)
Total Ph11 rate /Spool
AS (Noy)
Total PWHT rate /Spool
Lis (Moy)
Total FT rate /Spool
AS (-Uoy)
Total B H N rate /Spool
Lis (hlloy) Lis (Moy)
Total 'T rate /Spool
AS (Uoy)
Total Non-LVelded Spool/Welded Spool (Weiglit)
AS (rvloy)
Total Non-LVelded Spool/LVelded SpooI (Units)
AS (Moy)
6-14" No. of Pipe Pieces / Footage
AS (Noy)
6-14" No. of Pipe Pieces / DialnFt
AS (May)
6-14" 30.of Pipe Pieces / Ton
AS (Moy)
6-14" No. of Pipe Pieces / Unit
AS (May) ,AS (Moy)
(5-14" No. of Pipe Pieces Over 3 Ft / Footzge
AS (Moy)
6-14" No. of Pipe Pieces Over 3 Ft / Ton
AS (Moy)
6-14"
Total Subhrc Process \Veld / Design Weld VOL
Total U T rate /Spool
6- 14" 'JO.of Pipe Pieces Over 3 ft / DïaInFt
AS W ~ Y )6-14" AS (AUoy) 6-14" AS (Moy)
6- 14"
AS (ALioy)
6-14"
AS (LUloy) 6- 14" AS (Alloy)
6- 14"
AS (Alioy)
6-14"
AS (Uoy)
6-14"
AS (May) AS (Alloy)
6-14" 6-14"
AS P ~ Y )6-14" AS (Aiioy) 6-14" AS (Alloy)
6-14"
So. of Pipe Pieces over 3 ft / Unit
/ Footage 'To- of Fhnges / DiaInFr 30. of Fhnges / Ton No. of Fhnges / Unit Mo. of In-Line Fitlingj / Footage No. of In-Line Fittings / DialnFt No. of In-line Fittings / Ton qo. of In-Line Piltings / Unit go. of Non-In-Lïne Fitaogs / Footage qo. of Non-la-Line Fittligs / DiaInFt go. of Non-In-Line Fittings / Ton go. of Non-In-Line Filtings / Unit \JO. of Valves / Footage Mo. of Fianges
Ratio
Materid
AS (Moy)
Size Ratio Description 6-14" No. of Valves / DialnFt No. of Valves / Ton
AS (Moy)
6-14"
AS (Alloy)
6- 14" No-of Trdves / Unit
AS (Ailey)
6-24"
No. of Supports / Footage
AS (r\Uoy)
6-14"
No. oESupports / DiaInFt
AS (AlIoy)
6-14"
No. of Supports / Ton
AS (Moy)
6-14"
No. of Supports / Unit
AS (,-Vloy)
6-14"
No. of Mise. / Foomge
AS (~Uoy) 6-14" No. of hlisc. / D i a I f i t
/ Ton AS (,-Vloy) 6 1 4 " No-of hiisc. / Unit AS (Alloy) 6-14" Pipe Weight / Spool Weight 6-14" Non-Pipe \Veight / Spool LVeight AS (Moy) AS (AUoy)
6- 14" No. of bLisc.
AS (-Uoy)
6-14"
No. of Connections (Design \Velds) / Foocage
AS (Moy)
6-14"
No. ofcomections (Design \Velds) / DiaInFt
-4s (Alloy)
6-14"
No. of C o ~ e c t i o n s(Design \Xelds) / Ton
AS (-Uoy)
6- 14" No. of Comrctions (Design Welds)
AS (Moy)
6-14"
No. of hldti-Station Roll \Velds / Foomge
AS (Auoy)
6-24"
No- of bfulu-Scation Roll \Velds / DiaInFt
AS GWoy)
6- 13" No. of M d t i - k a o n RoU \Velds
AS (Moy)
6-14"
A S (Moy)
6-13"
AS (AUoy)
6-14"
AS (Alloy)
6-14"
AS (,illoy)
6-14"
AS (Moy)
6-14"
AS (Moy)
6-14"
/ Ton No. of hfulu-Station Roll \Velds / Unit No. of Repaired \Velds / Footage No. of RepaLed Welds / DiaInFt No. of Repaired \Velds / Ton No. of RepaL \Velds / Unit B\V DiaIn / Design \Veld DiaIn B\V Equiv-DiaIn / Design \Veld Equiv.Dia1n
AS W o y ) AS (Alloy)
6-14"
B\V Vol. / Design \Veld Trol.
6-14''
S\V DiaIn / Design \Veld DiaIn
Lis (Auoy)
6-14"
SW Equiv.Didn / Design Weld Equiv-Ddn
AS (Lüioy)
6-14"
S\V Vol. / Design \Veld Vol.
AS W O Y ) AS (,.Vloy)
6-13"
OL DiaIn / Design \Veld DiaIn
6-13"
OL Equiv.DiaIn / Design Weld Equiv.DiaIn
AS P O Y ) 6-14" 6-14'' AS (Moy)
/ Unit
OL Vol. / Design \Veld Vol. Pressure A t t a c h e n t / Design Weld Dialn
G- 14"
Pressure Attachment / Design Weld Equiv.DiaEn
6-14"
Pressure Attachrnent / Design \Veld VOL
AS (May)
6- 14"
Non Pressure Attachmenr / Design 'WeId DiaIn
AS (May)
G- 14"
Non Pressure Attachment / Design WeId Equiv.Dialn
AS W o y ) AS (Moy)
Ratio
Material AS (LLUoy)
Size Ratio Description 6-14" Non Pressure Attachent / Design \Veld Vol.
Ratio 3-170402E-03
-4s (Moy)
6-14"
AS (+Uoy)
6- 14"
AS (Uloy)
6-14''
AS (Moy)
6- 14"
AS (-Uoy)
6-13"
-4s (Moy)
6- 14"
AS (,-Uioy)
6- 14"
AS (Uoÿ)
6-14"
AS (Alloy)
6-14''
'4s (*Uoy)
6-14"
AS (Moy)
6-14"
/ Design \Veld Diain Position LVeld / Design Weld Equiv-DiaIn Position \Veld / Design \Veld Vol. Roll \Veld / Design \Veld D d n RoU \Veld / Design LVeld EqukDiaIn Roll Weld / Design \Veld VOL blulti-Station Roll \Veld / Design \Veld DiaIn hlulti-Staaon Roll \Veld / Design Weld Equiv-DiaIn hlulti-Station Roll Weld / Design \Veld Vol. Single-Station Roll \Veld / Design \Veld DiaIn Single-Station Roll \Veld / Design Weld Equiv-Dialn Single-Station RoU \Veld / Design \Veld Vol. rig Process Weld / Design \Veld Vol. 4% Process Weld / Design Weld Vol. FCALV Process \Veld / Design Weld VOL Stick Process \Veld / Design \Veld Vol. SubArc Process \Veld / Design Weld Vol. Xotoweld Process Weld / Design \Veld Vol. XepaL Rate (No. of R / Rt-\) go. of Cut Sheet Revision / No. of Spool XT rate /Spool
AS (Noy)
6-14"
b1T rate /Spool
100
AS (Alloy)
6-14"
?Trate /Spool
O
AS (Alloy)
6-14"
'MI rate /Spool
AS (-Uoy)
6-14"
?\VHT rate /Spool
1
AS (Noy)
6-14"
T'rate /Seo01
O
,AS (Mo).)
6-14"
3HhT rate /Spool
100
'4s (-Uoy)
6-14"
4T rate /Spool
1O0
AS (-Uoy)
6-14"
JT rate /Spool
100
AS (Moy)
6-14"
\ion-Welded Spool/CVelded Spool (YVeight)
O
AS (floy)
G-14"
\ion-ivelded Spool/Welded Spool (Uni=)
O
AS (U~OF) 6-14" 6-14" AS (Moy)
AS (Moy)
6- 14"
AS (MoY) AS (Uloy)
6-14"
AS (Moy)
6- 13"
A4S (,.Ulo:.)
6-14"
AS (Moy)
6-14"
AS (AUoy-)
6-14"
-4s (-Uoy)
6-14"
6-14"
Position !Veld
O O O
1 1 1 0.1257095 0.4257095 0.432653 0.5742905 0.5732905 0.567347 0.106721 1 O O
0.893279 O
O 2.933985E-02 O 100
1CO
FabMaster processed project data individually and warehoused the item-coded project information in an easy-to-access format. Fab-OLAP provides the functionaliq of vieming and analyzing productiviq-related information across
dl projects that have
been processed bp FabMaster. Fab-OLAP is an O n - h e Andytical Processing System custom-developed for the Fabrication FaUlities of PCL Indusmal Constructors, Tnc. The system feanires dynamic query, graphic presentation, and the functionality of statistical anaiysis o n 105 ratios of labor productivity/spool configuraton compleSq/qualiq control. It is an advanced decision-support tool for management to grasp the trend in the historical project data and identifg exceptional problems in the work at hand.
Figure C-1: Select one ratio
Step 1. Load the program and select one ratio User selects one ratio from the "Select Ratio" dropdomn Est, which includes all the 105 ratios computed in FabMaster, as shown in Figure C-1.
Step 2. Apply Filters on Material Type and Size Range Fab-OLAP uses the standard codes of the Company for the material types and size ranges. Fab-OWP helps user e-xplore data in decision-oriented mays and allows user to view data and get at them fomi different perspectives dong the dimension of 181
Figure C-2: Trial on ccnumberof pipe pieces per foot"
material and size. The histogram d o n g &th
~ t a t i ~ t analysis id results for the selected
ratio is presented o n screen and updated automatically. Figure C-2 shows the nial based on "carbon steel 6-14 inch spool, nurnber of pipe pieces per foot of pipe".
Step 3. Drill into details of data Fab-OUP
d o m s user to drill d o m to details of data by clicking the T i e w
Data" button. Figure C-3 shows the data behind the selected ratio.
17ûû250/ CS (Carbon]
-
1700204j CS (Carbon] 17002551CS [Carbon) 17002651CS (Carbon) 17002341CS (Carbon) 17004781CS (Carbon]
i
0.13346798!
i No- af Pipe Pieces / Faotage !No. of Pipe Pieces / Footage
1
O.OU4265; 4.8767 17E-02;
1
: 3.660536~-02; '
No. of Pipe Reces / Foatage
0.0488468 3 6.061 019E-M! O. 06563953
16-14"
No. of Pipe Pieces / Foatage
'i No. of Pipe Pieces / Footage
6-14"
j No. of Pipe Pieces / Footage
17004621CS [Carbon]
6-14"
:No. of Pipe Pieces / Footage
1700466j CS [Carbon]
6-14"
i No. of Pipe Pieces / Footage
16.762063E-021 7.231822E-Mf
No. of Pipe Pieces / Fooiage
/ 7.610802E-021
1700242! CS [Carbon) 1700206jCS [Carbon] 1700481j CS [Carbon) '
1 6-14" 1 6-14" 1 6-14"
j No- af Pipe Pieces / Footage I+ No. of Pipe Pieces / Footage
/ 6-14"
I
1700205; CS (Carbon] 1700474[Cç [Carbon]
'
6-14" 6-14"
i
/ 6-14" 1 6-14"
i 6-14"
l
i
b
1
p
6.587098~-021 1
No. of Pipe Pieces / Footage
7 ï29314~-021
No. af Pipe Pieces / Footage
7.750466E.O2]
1
1700211 CS [Caban)
j 6-14"
No. af Pipe Pieces / Fmtage
.8.051168E-021
1700464[ CS [Cabon)
1 6-14"
'No. of Pipe Pieces / Footage
1 6-14"
' No.
of Pipe Pieces / Footage
8 496705E-Mi 8 673514E-02:
16-14''
:No. of Pipe Pieces / Footage
1 8.859606E-021
1700491!Cç (Carbon) 1
1700213;CS (Caban]
i
Figure C-3: View details of data
Step 4. Prht out the trial and statistical analysis resdts User clicks the 'Trint out" button to psint a hard copy of the current trial and stausticd analysis results including the histograrn for record.
&PENDIX
D: USER'SMANUAL FOR PIPINGMASTER
PipingMaster is a historical project data =-arehoushg system customized for the field construction systems of PCL Industrial Constructors, Inc. It is an automated data processing tool to estract ram data from Labor Cost Control System, Estimating System,
and Quality Contcol Spstem, and convert raw data into useful pruductiviq information based on embedded expert d e s . Pipe handling and welding are processed by Pipinghlaster independendy, but in simila fashions including the user interfaces and prograrn logic. Thus, Pipe handling is selected to illustrate the program flow in the following steps.
Step 1. Impoa raw data in standard format User irnports three ram data tables for each project into the database manudy to
dow
for
PipingMaster
RDJroject#Hand
table
to
the quantity of
piping
for pipe hanrlling, RD-Project#Detl
components, RD-Project#Weld
Figure D-1.
calculate
work, for
namely,
pipe
work
for pipe melding. The table structures are shown in
RD-Pro j#Hand
RD-Pro j#Detl
RD-Pro j#Weld
Project #
Project #
Project #
Nominal Size
Detl Type
Nominal Size
Schedule
Nominal Size
Schedule
Ciassifïcation
Classifïcation
Joint Type
hhterial Type
Mzterial Type
Classification
Length (fi)
Quantity
Material Type # \cVelds
Es tUnitPvLH
Es tUnitlhIH
Figure D-1: Structures of Raw Data Tables for A Projcct
T h e detailed quantity take-off (in footage) for pipe handling of one project is available in the project estimate only. L'sually information is known and complete o n the size, the thickness, the material tgpe, and the location classification of each individual pipe
section. The detailed quantity take-off in number of welds for pipe melding of one project is availaole either in the project estimates or in the field quality control system. In most cases the pipe ske, pipe thickness, pipe material type, location classification and meld joint type are knomn for each individual weld.
Installation of other piping moxk components (or piping details) includes pipe supports, bolt-ups, valces, screw joints, and misceUaneous items like flanges, specialties, 185
elbows, cuts and bevels. The number and type ofwork components and estimated unit manhours for one project
me
available in the project estimates. However, information on the
size, material type, location dassihcation may not be found in the estimate. Therefore, we need to check the ram data integrïty of the piping work components prior to processing.
Step 2 Raw Data Integrity Check The raw data integrity check is controlled by the entered project setting regarding the raw data integiq and methods of actual mm-hour cost coding as shown in Figure D-2. User
Matuid Type -
Sue R a g e [16'1 OIoaae Ho If Co& To Total Lwd
y=
----
ri NO
Figure D-2: Main user interface of FabMaster
enters the project number to be processed and answer a number of Yes/No questions about
the project Nesq user clicks "Check Raw Data Integritg" button to s t m the program. User
d l be prompted to correct any problems due to failure to pass the checks. The PipingMaster is capable of identïfyïng missing data or incorrect data in the raw data tables. For esample, if actual labor hours in the labor cost system were tracked to the level of various classifications of location, then a null in the "Classihcation" field of the raw data tables d be detected as invalid data and must be corrected for W e r processing. Three valid types ofmeld joinh i.e. BW putt Weld), SW (Socket Weld), OL (Olet Weld) and five valid types of piping work components are allowed lo the RD-Project#Ded table, 1.e. bolt-up, valve, screw joint, support, and misc.
Step 3 Check cross-reference Data for Quantity Calculation User hrst chooses one of four options and then click the "Check Cross Reference Integriy" button to perform the check for the selected option. Four options should be checked through one by one. User d be prompted to correct raw data or update cross reference tables in case Pipinghlaster finds a problem. The "Action" button d only be activated when aLi the ram data checks and cross reference checks are passed.
In PipingMaster, a number of cross-reference tables are involved to c a l d a t e the quantitg in various units of measurement, i.e. five units of measurement for pipe handling: DiametePLength (Tnch*Feet), Equivalent Diameter*Length (Inch*Feet), Length (Fee t), Weight (p.lI.T), and Base Manhours (MH); five units of measurement for pipe welding: Diameter (Inch), Equivalent Diameter (Inch), Volume (Cubic Inch), Volume/Thickness (Square Inch), and Base bfanhours ($El).Cross-reference integrity check is performed to ascertain that each record in the raw data table can h d the needed information in the conesponding cross-reference tables so as to calculate accurate quantities. The forrnulae 187
used are commonly found in an industriai maoual o r piping handbook-The relationships between raw data tables and cross-reference tabIes are shomn in Figure D-3 and Figure D-4.
NominalSize Schedule RD-Proj#Hand Project # Nominal Size Schedule Classiacation M a t e d Type Length
OuterDiameter (inch)
Figure D-3: HandLuig: X-Refercncc Information Integrity Check
1-
Schedule Thicknes (inch)
Schedule
Project # Nominal Size Schedule Joint Type Classification Ma terid Type # Welds EstUnitkfH
JT 1s OL tblOletDim Nominal Oudet Dimension B
Figure D-4: Welding: X-Reference Information Integrity Check
Step 4 Generate Aggregate Cost Codes and Calculate Quantities User hits the "Action" button to generate aggregate cost codes to the level of project nurnber, classi£ication of location, material type, size range, activity, and unit of measure. The total quantities and quantities breakdown for size ranges, d o n g with the generated cost codes 1
.be
appended a summary table called "tblQuanutyLMaster". Table 1 shows sample
records in the summary table for one relatively s m d job.
Table D-1: Sample of Quantity Calculation Surnmary Table in PipingMaster Project# Material Class CostCode CS
410
Description Welding Total Volume/Thickness
CS
410
Handling Total Feet
CS
460
Handiing Total Feet
CS
460
Welding Tota1 Volurne/Thickness
AS
460
Handling Total Feet
AS
460
Welding Total Volume/Thickness
SS
460
Handling Total Feet
SS
460
LVeIdrng T otri Volurne/Thickness
TOT
31O
Hand Tot Mati Tot Size Ft
TOT
460
Hand Tot M a t l Tot Sizc Ft
Step 5 Enter ActuaI Hours and Compute Actual Degree-of-difficulty Factors
Folloming generating the cost codes and calculating the quantities, PipingbIaster
Show Gnnpïkd Records aid A
Proje&
d lnActual Mhs h m LCS Report
1 Materiaflypa 1 Classification 1 CostCode 1 QtyTotal 1 BaseMH 1 Adual M h n 1
Statris
[
Shan U35 Act MHr la Girrcnt Roiect Show Campikd HanMuPipiicrt: 1 Proje- lateria4 Classl CostCode 1 QtyBetow2 ( Qty2TolS 1 QtyAbovelS 1 QtyTotal 1 ES!MUI~M u l t l 15L30486 CS 410 302151-02 11 5623 O 5634 100 388
Statur No
Figure D-5: Productivity Analysis Page for Pipe Handiing
sliifis focus to productiviry analysis page as shown in Figure D-5. User reads acrual manhours and enters into the "actuai manhours" column for corresponding records. Nesr, user hits the "Analyze" button to let PipingMaster figure out the actud labor hours for pipe handling based on the project setting about actual labor cost t r a c h g practice and the ernbedded espert d e s for handling different scenarios. Evennidy, the acmal degree-ofdifficulty factors are computed for each record and listed against the factors estimators have used for comparison. After comparison, user decides on which records are valid for NN to use by switching the status of one record from N o to Yes.
Step 6 Make Questionnaires for Valid Records
T L INDUSTRIAL CONTRACTORS INC. Pipe Handling Report Prtpaied By:
1.Handling Job Group by
I # i t d i Sotaert
Prqect #
1500484_007
Report Date:
15/7/59
CosiCode:
m-
Figure D-6: Sample of Pipe Handling Questionnaire
PipingMaster m&es questiomaires for those valid records as conhmied by user.
Figure D-6 shows a sampIe questionnaire. Followïng the above six steps, Pipïngbkister processes one project and convert ram data into accurate cost-coded productiviy information for f5uther productivity analysis. Figure D-7 shows the flom chart of the whole program.
1
Check Raw Data Intecgrity N
1s Data Complete and
1
[
Correct Missîng/
1
Incorrect Data
A
Correct? 1s Piping Work Component
N
included in the "Pipe Install~'Cost Code? Y k/
v
1s ~ o c a t i &Classification, Matenal Type, and Size Ranges Known For Piping Work Component?
Y I
Generate Indices for Piping \Vork Components
N
\1/ Compile Estimated Manhours for Pipe Work Component from the Raw Data
Prorate Estimated Manhours for Pipe Work Component Based On Pipe
Extract Estimated Manhours fcom "Pipe Instd" Cost-Coded MA-hours & Genqrate Indices