Scholarly article on topic 'Framework for prioritizing infrastructure user expectations using Quality Function Deployment (QFD)'

Framework for prioritizing infrastructure user expectations using Quality Function Deployment (QFD) Academic research paper on "Civil engineering"

CC BY-NC-ND
0
0
Share paper
Keywords
{"Quality Function Deployment (QFD)" / Markov / Hidden / "Risk management" / Sustainability}

Abstract of research paper on Civil engineering, author of scientific article — Aman A. Bolar, Solomon Tesfamariam, Rehan Sadiq

Abstract Customer involvement in infrastructure maintenance activities is a complex process due to various decision-making parameters surrounding maintenance. Compared to manufacturing and other disciplines where QFD is widely used, expectations of the infrastructure user as a customer are truly dynamic given the changing economic conditions, technologies, environmental regulations, etc. While such dynamic or changing customer expectations can be addressed by repeated surveys and constant communication, having indicators to predict customer response would be a valuable tool and aid the QFD decision-making process. In this study, a framework that utilizes hidden Markov model (HMM) is proposed for evaluating customer expectation by using probabilities of focus areas that are of interest to the infrastructure user as hidden parameters. The focus areas are based on sustainability parameters and include economic, social, technological, maintenance efficiency, safety and environmental conditions. Probabilities that represent the probability of transition from current state (of the focus area) to next possible state were generated based on expert opinion of the authors. Using the 2005 customer survey by California Transportation, a case study is presented in order to demonstrate the application which concludes that the proposed methodology can be successfully implemented for infrastructure maintenance.

Academic research paper on topic "Framework for prioritizing infrastructure user expectations using Quality Function Deployment (QFD)"

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

International Journal of Sustainable Built Environment (2017) xxx, xxx-xxx

hosted by

Gulf Organisation for Research and Development International Journal of Sustainable Built Environment

ELSEVIER

ScienceDirect www.sciencedirect.com

Original Article/Research

Framework for prioritizing infrastructure user expectations using Quality Function Deployment (QFD)

Aman A. Bolar *, Solomon Tesfamariam, Rehan Sadiq

Okanagan School of Engineering, The University of British Columbia, 3333 University Way, Kelowna, BC V1V1V7, Canada

Received 7 August 2016; accepted 8 February 2017

Abstract

Customer involvement in infrastructure maintenance activities is a complex process due to various decision-making parameters surrounding maintenance. Compared to manufacturing and other disciplines where QFD is widely used, expectations of the infrastructure user as a customer are truly dynamic given the changing economic conditions, technologies, environmental regulations, etc. While such dynamic or changing customer expectations can be addressed by repeated surveys and constant communication, having indicators to predict customer response would be a valuable tool and aid the QFD decision-making process. In this study, a framework that utilizes hidden Markov model (HMM) is proposed for evaluating customer expectation by using probabilities of focus areas that are of interest to the infrastructure user as hidden parameters. The focus areas are based on sustainability parameters and include economic, social, technological, maintenance efficiency, safety and environmental conditions. Probabilities that represent the probability of transition from current state (of the focus area) to next possible state were generated based on expert opinion of the authors. Using the 2005 customer survey by California Transportation, a case study is presented in order to demonstrate the application which concludes that the proposed methodology can be successfully implemented for infrastructure maintenance.

© 2017 The Gulf Organisation for Research and Development. Production and hosting by Elsevier B.V. All rights reserved.

Keywords: Quality Function Deployment (QFD); Markov; Hidden; Risk management; Sustainability

1. Introduction

1.1. Infrastructure report cards

Infrastructure condition anywhere in the world is representative of the prosperity, management capability and financial health. The maintenance of infrastructure is

* Corresponding author. E-mail addresses: aman.bolar@alumni.ubc.ca (A.A. Bolar), solomon. tesfamariam@ubc.ca (S. Tesfamariam), rehan.sadiq@ubc.ca (R. Sadiq). Peer review under responsibility of The Gulf Organisation for Research and Development.

gaining attention given that existing infrastructure is subject to increasing population growth, low economic growth and effects of climate change (Miller, 2013). A study done by the McKinsey Global Institute (MGI, 2013) has estimated that about $57 Trillion is required by the year 2030 just to keep up with gross domestic product (GDP) growth which is a standard measure of economy. In North America, Infrastructure report cards are currently utilized to provide a grade, analogous to schooling system, wherein the evaluated condition and performance of infrastructure is reported.

http://dx.doi.org/10.1016/j.ijsbe.2017.02.002

2212-6090/© 2017 The Gulf Organisation for Research and Development. Production and hosting by Elsevier B.V. All rights reserved.

2 A.A. Bolar et al. /International Journal of Sustainable Built Environment xxx (2017) xxx-xxx

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

Nomenclature

Xt parameter that is a stochastic process dependent O set representing observed parameters

on time t P Si Sj probability of going from state st to Sj

t time p 1 1 01 S probability that the observed parameter is oj gi-

P probability ven that the current state is st

j condition state C ^re credibility factor (confidence of expert estimat-

m number of condition states ing probabilities)

P 0) j probability of being in condition state j at time t F initial probability matrix

n number of time steps V observed probability matrix

S set representing condition states

For the United States, the American Society of Civil Engineers (ASCE) introduced the infrastructure report card in the year 1988. The report card contains an evaluation using five ratings "A" through "D" and "F" (where A = Exceptional; F = Failing), for each infrastructure category in the country. The 1988 report had defined eight infrastructure categories and since then five report cards have been published, the most recent in the year 2013 (ASCE, 2013). Moreover, in the year 2011, ' ' Failure to Act" documents were published by the ASCE with the objective of reporting the economic implications of current US investment trends in key infrastructure sectors. The 2013 report card has sixteen infrastructure categories and the overall rating for America's infrastructure is ' 'D+" (Poor, but leading to Mediocre). The cost to improve infrastructure is estimated to be $3.6 Trillion in a 7-year period which translates into a budget of about $514 billion per year. Compared to the $3561 billion budget, a $158 billion annual deficit exists in the US infrastructure budget. Given this situation, cost-allocation for maintenance can be a challenging task.

In Canada, the first report card was published in the year 2012 for municipal infrastructure that includes water/wastewater systems and municipal roads excluding bridges and culverts, etc. (Felio, 2012; www.canadainfras-tructure.ca). The rating systems used were: very good, good, fair, poor, very poor. The condition of municipal roads was rated as fair (meaning requires attention), while water/ wastewater, and storm water received good and very good ratings, respectively. The replacement cost of the deficient infrastructure alone is $171 Billion. Prior to 2012, reports published by the Federation of Canadian Municipalities (FECM) provided information about the condition of infrastructure. A report published in 2007, titled Danger Ahead: The coming collapse of Canadian Infrastructure" stated that deferred investments in maintenance has caused the deficiency to reach $123 Billion compared to about $12 Billion in the year 1985 (Mirza, 2007). Also recently, a model framework for Canada's Core Public Infrastructure's (CPI) performance measures was published by

1 http://www.cbo.gov/publication/25116.

Félio and Lounis (2009), which included roads, transit, bridges, and water and wastewater infrastructure. Customer Requirements in Decision-Making Process.

Among infrastructure categories, bridges are perhaps one of the most complicated assets, given their inspection and maintenance requirements and therefore bridges have always been a starting point for the development of infrastructure management systems (Adey et al., 2010). Therefore, for explaining the importance of customer requirements in the decision-making process, bridges have been quoted in this section as an example. However, the same philosophy would apply for the maintenance of other infrastructure types.

In the case of bridges, Bridge Maintenance Systems (BMS) aim prioritization and decision making for funding allocations among an inventory of bridges (also known as network level analysis). Input from network level analysis facilitates examining bridge repair strategies at the component level (also known as project level analysis). Similar to other infrastructure categories, bridge maintenance involves collecting a large body of data from various sources and requires expert judgement in making a maintenance decision (Bolar et al., 2012).

However, limited literature is published, wherein input from the infrastructure-user and stakeholders is incorporated in the maintenance decision-making process. With infrastructure condition evaluation results, a decision-maker has to choose mostly between ' do-nothing',' rehabilitate' or replace' options. With consumers (infrastructure users) becoming more involved in sustainability issues, including them in the decision-making process would prove valuable to all the stakeholders involved (Bolar, 2014). Studies involving public participation in sustainability are well documented (e.g., Kasemir et al., 2003). For example, a decision to either replace or rehabilitate a bridge can affect individuals and departments in various ways. In this context, a bridge user is defined as anyone that is affected by modifications to the bridge. A business located close to such a bridge might be affected by traffic volumes and so would be fire fighting departments, etc. The input of all these can affect decision making and including them in the process would be relevant. Also, most governments

A.A. Bolar et al. / International Journal of Sustainable Built Environment xxx (2017) xxx-xxx 3

123 world-wide are proceeding towards Public-Private Partner-

124 ships (PPPs) in infrastructure maintenance in order to har-

125 ness the expertise of private organizations who in most

126 cases assume financial and technical risk. However, ulti-

127 mately, the consumer (tax payer or the infrastructure user)

128 is the means of financing and is the end-user (Miller, 2013).

129 Therefore, including consumer requirements in any deci-

130 sion related to infrastructure is not only relevant, but

131 imperative. This process of involving customer require-

132 ments in the decision-making process can be facilitated

133 by adopting a widely used method in manufacturing, called

134 Quality Function Deployment (QFD).

135 1.2. Hidden Markov model in civil engineering problems

136 In general, the applications of hidden Markov model

137 (HMM) can be widely found in electronics and telecommu-

138 nication industry for image sequencing, recognition tech-

139 niques, etc. (Rabiner, 1989). However, not many

140 applications are seen within the civil-structural engineering

141 domain. In the structural health monitoring discipline, an

142 application of HMM for damage accumulation and prop-

143 agation was proposed by Rammohan and Taha (2005),

144 whereas a model for pavement deterioration was proposed

145 by Lethanh and Adey (2013). In the transportation disci-

146 pline, a recognition method for lane change intention was

147 proposed by Xu et al. (2011), and a HMM for vehicle

148 recognition by image processing was used by Shen and

149 Bai (2006). In hydrological research, HMM was applied

150 by Mallya et al. (2013) in their study focusing on drought

151 characteristics. A few applications are also seen in mechan-

152 ical maintenance systems. Condition-based maintenance

153 using HMM has been studied by Bunks et al. (2000) and

154 Yu (2012).

155 1.3. Motivation

156 A primary input to the QFD process is customer

157 requirements that are normally based on surveys and ques-

158 tionnaire. However, these customer requirements are

159 dynamic in the sense that depending on time and space

160 the expectation could vary. For example, in a good eco-

161 nomic condition if a customer is questioned about improv-

162 ing aesthetics on a bridge, the answer may be positive, but

163 if the economic condition is bad a natural answer expected

164 would be to "hold on" to that activity and perhaps priori-

165 tize among safety and other hazardous issues. Customer

166 requirements can therefore change over time and keeping

167 up with customer requirements can pose challenges. In

168 order to keep up with changing requirements frequent

169 and repeated customer surveys can be conducted, but they

170 involve cost, time and effort in collecting the information

171 and generating results. A hidden Markov model (HMM)

172 can aid in the determination of future probabilities using

173 currently available probabilities as "known parameters"

174 and by modelling related time-dependent entities as hidden

175 parameters. In the case of infrastructure, given the

responses of infrastructure users as "known parameters" 176

and related focus areas of interest to the customer as 177

time-dependent parameters, future response of the cus- 178

tomer can be predicted. 179

1.4. Objective 180

The objective of this study is to present a methodology 181

wherein dynamic customer (infrastructure-user) require- 182

ments can be predicted using hidden Markov parameters 183

for use in decision-making within QFD. A study done in 184

Bolar et al. (2014) considered five focus areas as relevant 185

to infrastructure user requirements. These focus areas are 186

issues related to economic, social, safety driven, technolog- 187

ical factors, maintenance efficiency, and environmental 188

issues. In this paper, the same focus areas that were based 189

on sustainability and engineering parameters are consid- 190

ered as hidden parameters in a HMM. These hidden 191

parameters can be used to identify future customer require- 192

ments and update the house of quality (HOQ). The pro- 193

posed framework is highlighted in Fig. 1 and explained in 194

the following sections. An application has been presented 195

in the current study for demonstrating practical use of 196

the methodology. 197

2. Decision-making using Quality Function Deployment 198

(QFD) 199

2.1. Customer requirements in decision-making process 200

t In order to define requirements for Asset Management, 201

the required level of performance of the asset has to be 202

established. In the year 1965, the concept of level of service 203

(LOS) for highways was introduced. The LOS is defined as 204

"a qualitative measure describing operational conditions 205

within a traffic stream, generally in terms of such service 206

measures as speed and travel time, freedom to manoeuvre, 207

traffic interruptions, comfort and convenience" (Bhuyan 208

and Nayak, 2013). While the definitions of LOS would 209

apply to studies involving transportation, the current study 210

is focused on infrastructure maintenance activities. The 211

Federation of Canadian Municipalities (FECM) and 212

National Research Council Canada (NRCC) (FCM and 213

NRC, 2003) have published a National Guide to Sustain- 214

able Municipal Infrastructure wherein best practices for 215

maintenance levels of service are outlined in detail. In the 216

guide, level of service is defined "as a composite indicator 217

that reflects the social and economic goals of the commu- 218

nity and may include any of the following parameters: 219

safety, customer satisfaction, quality, quantity, capacity, 220

reliability, responsiveness, environmental acceptability, 221

cost, and availability." Customer satisfaction/perception 222

is considered to be a key activity for achieving level of ser- 223

vice. Similarly, infrastructure asset management plan in 224

Southern Australia (LGASA, 2006) that is based on the 225

international infrastructure management manual (IIMM) 226

also identify "customer research and expectations" as an 227

Fig. 1. Hidden Markov analysis for infrastructure customer requirements in QFD.

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

A.A. Bolar et al. linternational Journal of Sustainable Built Environment xxx (2017) xxx-xxx

important component of levels of service. Most infrastructure agencies define criteria for levels of service in their maintenance quality assurance (MQA) process. For example, the Washington State Department of Transportation (WSDOT, 2012) has levels of service rated A through F, where A represents a very high service level and F representing a very low service level. LOS A indicates that all systems are operational and users experience no delays whereas F indicates significant delays occurring on a regular basis.

In business terminology, the words customer" and ' ' consumer" are often used interchangeably. Customer2 is normally the purchaser of the product and consumer3 is the end user of the product. For infrastructure, citizens are the end users, but these citizens can be regarded as customers by the way of tax payments. Therefore, in order to avoid ambiguities in definition, only the terms customer" or infrastructure user" have been used in this paper both representing individual(s) making use of the infrastructure.

Infrastructure maintenance involves collecting a large body of data from various sources and requires expert judgement in making a maintenance decision. However, limited literature is published, wherein input from the infrastructure-user and stakeholders is incorporated in the maintenance decision-making process. With infrastructure condition evaluation results, a decision-maker has to choose mostly between do-nothing', rehabilitate' or replace' options. With customers (infrastructure users) becoming more involved in sustainability issues, including them in the decision-making process would prove valuable to all the stakeholders involved. Studies involving public participation in sustainability are well documented (e.g.,

2 http://www.businessdictionary.com/definition/customer.html.

3 http://www.businessdictionary.com/definition/consumer.html.

Kasemir et al., 2003). For example, a decision to either replace or rehabilitate a bridge can affect individuals and departments in various ways. In this context, a bridge user is defined as anyone that is affected by modifications to the bridge. A business located close to such a bridge might be affected by traffic volumes and so would be fire agencies, etc. The input of all these can affect decision making and including them in the process would be relevant. Ultimately, the customer (tax payer or the infrastructure user) is the means of financing and is the end-user (Miller, 2013). Therefore, including customer requirements in any decision related to infrastructure is not only relevant, but imperative.

This process of involving customer requirements in the decision-making process can be facilitated by adopting a widely used method in manufacturing, called Quality Function Deployment (QFD).

2.2. QFD background

Quality Function Deployment (QFD) is a tool that offers many facets with an end goal of ensuring that customer requirements are satisfied. Therefore, the phrase voice of customer" is normally associated with the term QFD. In the process of satisfying customer requirements, QFD can also be depicted as a tool providing factual engineering specifications wherein customer requirements are documented and means of achieving the customer requirements is addressed.

QFD came into existence in the late 1960s in the Shipyards at Kobe, Japan during construction of super tanker cargo ships. The Japanese government at the request of Mitsubishi contacted universities to come up with a logistic where each step of the construction process was linked to a specific customer requirement. This led to the development

A.A. Bolar et al. /International Journal of Sustainable Built Environment xxx (2017) xxx-xxx 5

293 of what is today known as Quality Function Deployment

294 (QFD). Very soon, QFD made way into the automotive

295 industry and has since then been adopted by a wide range

296 of disciplines such as aerospace, defence, education, lifecy-

297 cle analysis, logistics, software, process engineering,

298 telecommunications, health care, etc. (Bolar et al., 2014).

299 In fact, the extent of areas where QFD has been researched

300 has become so exhaustive that Carnevalli and Miguel

301 (2008) investigated the research done in QFD as a research

302 topic itself. Comprehensive literature reviews have been

303 published by Sharma et al. (2008) and Chan and Wu

304 (2002).

305 In engineering, QFD has been popular especially in the

306 aerospace and automotive industry, but has slowly gained

307 attention among other disciplines. The use of QFD for

308 Civil Engineering Problems has not been exhaustive

309 although specific disciplines such as construction manage-

310 ment have incorporated QFD as a tool for mainly prioritiz-

311 ing and identifying customer requirements. Topics where

312 QFD has been considered, are related to pavement con-

313 tractor performance management (Yasamis-Speroni et al.

314 2012), soil tillage (Milan et al., 2003), construction manage-

315 ment (Ahmed et al. 2003; Lee and Arditi, 2006), architec-

316 tural design (Kamara et al. 1999; Van Luu et al., 2009),

317 environmental requirements (Utne, 2009; Zhang et al.,

318 1999), bridge conceptual design (Malekly et al., 2010), oil

319 and gas (Yang et al., 2011) and drinking water quality

320 management (Francisque et al., 2011). While most of these

321 applications deal with new greenfield construction (what

322 would be termed as design or product development by

323 mechanical and other engineering disciplines), not many

324 applications can be found in maintenance (brownfield)

325 management especially related to civil infrastructure prob-

326 lems. Bolar (2014) demonstrated an application of QFD to

327 civil infrastructure maintenance decision-making. Sarja

328 (2004) has also reported the use of QFD within LIFECON,

329 which is a European Life-Cycle Management System

330 (LMS). Junhai et al. (2007) in China also reported Bridge

331 lifecycle design based on QFD.

332 2.3. The QFD process

333 A flow chart depicting the various steps involved in

334 QFD is provided in Fig. 2. The actual QFD process is

335 shown in a typical HOQ in Fig. 3, and various terms and

336 their role in QFD is described below:

337 • House of Quality (HOQ): The HOQ (Fig. 3) is a term

338 associated with QFD which is a matrix that documents

339 and establishes all the processes in implementing QFD

340 as shown in Fig. 2. The various terms are defined in rela-

341 tion to infrastructure maintenance as follows:

342 • WHATs: The primary input in the HOQ is a prioritized

343 list of basic customer demands (requirements and needs)

344 that are usually expressed in vague and imprecise terms

(e.g., riding comfort on a bridge, environmentally 345

friendly materials, etc.). Each demand is documented 346

as a WHAT and prioritized as represented by WHAT- 347

1, ..., WHAT-n in Fig. 2. 348

• HOWs: Once the WHATs are generated, the means of 349 achieving these WHATs is identified and termed as 350 HOWs. Therefore, HOWs are the design (or technical 351 or product) characteristics that serve to meet the 352 WHATs (e.g., maintenance-aimed at no potholes on a 353 bridge, green products or construction for customer 354 requiring environmental compliance, etc.). For each 355 WHAT, a corresponding HOW is identified as repre- 356 sented by HOW-1,..., HOW-n in Fig. 2. 357

• Relationship matrix: Indicates how product characteris- 358 tics or decisions affect the satisfaction of each customer 359 need. It consists of relationships existing between each 360 WHAT and each HOW attributes (i.e., WHAT vs. 361 HOW as shown in Fig. 2). 362

• Absolute weights and ranking of HOWs: contains results 363 of the prioritization of product characteristics to satisfy 364 customer requirements. It represents the impact of each 365 HOW attribute on the WHATs and is the final step 366 before ranking of the weights for decision-making as 367 shown in Fig. 2. 368

• Correlation matrix: is the roof of the HOQ and repre- 369 sents the interdependencies among HOWs as shown in 370 Fig. 2. It can play an important role in deciding on 371 the number of HOWs that directly affect the cost, prior- 372 itizing WHATs and HOWs. 373

3. Dealing with dynamic customer requirements in QFD 375

Companies dealing with product development have 376

adopted QFD in order to identify and address customer 377

requirements. These customer needs can be termed 378

dynamic as their needs can vary over time. Chong and 379

Chen (2010) provided a review of the current state of 380

research related to dynamic customer requirements. In 381

conclusion of the review, there was a recommendation 382

to develop solutions that could automatically handle 383

dynamic customer requirements. Similar literature review 384

was done by Sepideh and Aaghaie (2011), but focusing 385

on the use of HMM alone. According to their findings, 386

in customer relationship management (CRM), about 387

27% of published articles adopted HMM. A novel 388

method integrating QFD with HMM was published by 389

Shieh and Wu (2009) by adopting HMM to update 390

QFD technical measures. The current study is based on 391

the approach used by Shieh and Wu (2009), but has been 392

extended for application to engineering, especially civil 393

engineering problems where including customer require- 394

ments is still at infancy and the customer expectations 395

have a wide variety of parameters especially with more 396

consciousness in sustainability issues. 397

6 A.A. Bolar et al. /International Journal of Sustainable Built Environment xxx (2017) xxx-xxx

Fig. 2. QFD process flowchart.

398 3.1. States and transition in Markov chain

Epj =1 8

The Markov process can exhibit a finite number of states depending on the intrinsic nature of the problem. For example, if damage on a structural beam is to be defined in terms of probabilistic states, we can have cracked, corroded, spalled, etc., condition states for the beam. We can call these condition states as state 1, 2, 3 ... m recorded at a given time t. Consider State 1 as "cracked" state and State 2 as "uncracked" state. Furthermore, let j be the condition state at time t for parameter xt and i be the condition state at t _ 1 represented by parameter xt_1. In other words, at a given time t, j or i can have one state among 1, 2, 3 ... m. The probability of xt being in condition state j given that the condition state was i in xt—1 is termed as transition probability and in matrix form is also called as the transition probability matrix (TPM).

p [x, = j X-1 = i]=n

417 Since each i should transit into one of the states repre-

418 sented by j, for any given i, the sum of all transition prob-

419 abilities should be equal to 1.

In general, for n time steps,

: p(0) p" ij

P(0) is also called as the initial probability matrix and what follows from the above is that given the TPM and initial probabilities, probabilities at any future time step n can be determined by the product of initial probability matrix and the TPM raised to the power n.

3.2. Implementing hidden Markov model for QFD

Hidden Markov model consists of a stochastic process that is not observable (hidden), but can be observed through an associated stochastic process and thereby generate the sequence of observed symbols (Rabiner and Juang 1986). Similar implementation of HMM to QFD customer requirements is also provided in Shieh and Wu (2009). The following section provides relevant theory in implementing a HMM methodology.

IJSBE 152 30 March 2017 ARTICLE IN PRESS No. of Pages 14

A.A. Bolar et al. /International Journal of Sustainable Built Environment xxx (2017) xxx-xxx 7

Consider that condition states are represented by a set S = {s1, s2, s3 ... sn} and the observed parameters are represented by set O = {o1, o2, o3......on}. Let {X} be a Markov

chain with probability values in state S and J states observed. Let Pj0) be the absolute probability that sj is in time t0. Therefore

p [x, = jjx,-1 = i]

The transition probabilities represented by matrix F, is the probability of going from state si to sj is:

Pstls1 Ps1|s2 Ps1 |s3 PS1 |sm

PS2h PS2|S2 PS2 |S3 PS2 |sm

f = PS3h PS3|S2 PS3 |S3 PS3 |sm (5)

- PSm 1st PSm |S2 Psm|s3 Psm |sm -

and the associated observed probabilities i.e. probability 455 that the observed parameter is oj given that the current 456

state is si.

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

A.A. Bolar et al. /International Journal of Sustainable Built Environment xxx (2017) xxx-xxx

2-step transformation, [P(2)] = XP|1)TAiJ = x(Xpi0)TAlJ)TAiJ

= E WE TAjW = 53 P(0)TAf

p01 |S1 P01|S2 P011 S3 p011 Sm

P 02 |s1 P 02 | S2 P 021 S3 P 021 Sm

V Cre P03 |S1 p031 S2 P03 | S3 p03 | Sm (6)

-P01|S1 P0l | S2 P0l | S3 P0m Sm

where Cre is a credibility factor applied that represents the confidence of the expert estimating the probabilities. Also,

= 1 and CrePo,

For brevity, represent transition matrix with initial probability as F = {f(0)} with state sj and observed probability V = {Pj-0)} with observed symbols oj associated with state sj. From the definition of conditional probabilities,

(n) f(n)

fmn)] = [f 1n-1) f2n-1)

f (n-1) IF

for each n e N

[pf p® ... p|i)] = {[f ® f?

for each i s N U{0}} Considering (5) and (6)

[p1n) p2n) ... P((n)]={[f 1n) f(n)

f((i)]V

(8) (9)

2 ••• f(n)]V

for each i s NU{0}}

= [f 1n-1) f2n-1) Similar to (7), for nth-step,

(n) (n) (n)

fmn-1)]Fv

[ p1n) p2n) ... p((n)] = [f 10) f20) ... fa(0)]FnV (11)

(0) (0)

m ], matrices F & V can be docu-

mented from surveys. If V is square and invertible,

[ f 10) f!

p(0) p(0) p1 p2

Eqs. (6) and (8)). Therefore

[ p1n) p(n)

p(0) ]V-1 (using

[p10) p20) ... p(0)]V-1FnV

where V_1FnV is the transformation matrix of observed symbols. Representing this transformation matrix by TA (n), after 1-step transformation, TA(1) = TA. Eq. (12) becomes:

[p((1)] = [p10)TA„ + p20)TA2i + p20)TA2i +

= Y^p(0) TAij for 1 < j < l

^PP|0)(V-1F2V)

where TA^ = ¡TAij)TAij is the 2-step transition-observation transformation matrix. For n-steps,

(^t^^ta^ = (v-1fnv)

4. Application OF QFD-HMM to infrastructure maintenance

4.1. Customer response analysis: case study

In order to demonstrate the evaluation of predicting customer response based on the above observed parameters, data were adopted from the California Department of Transportation's (2005) maintenance customer survey available at: http://dot.ca.gov/hq/maint/externaLsurvey/ 2005_survey/index. htm.

The data available were in the form of pie charts for each question, wherein the questions were regarding maintenance management by the department and responses in linguistic expression. Therefore, data from the pie charts that were in percentages were mapped into the three categories adopted for the current study, namely High, Medium and Low. Mapping of the data is presented in Table 1. The QFD customer requirements (CR) were classified on three scales as follows - High (H) with a weight 5, Medium (M) with a Weight 3 and Low (L) with a weight 1. The scales are also shown graphically in Table 2.

An Illustrative Example: The solution for HMM for estimating future customer requirements was performed using Microsoft-Excel® Visual Basic Application (VBA) programming. However, for demonstration purposes, a solved example is provided for one focus area and 2-steps of future customer requirements. Using the VBA program, any number of future states can be referred to using a drop down box.

4.2. Observed probabilities for focus areas

Customer requirements in maintenance management are truly dynamic in nature. The reason being that customer requirements certainly vary with time. For example, if customer input is sought for a design and build project, the

A.A. Bolar et al. / International Journal of Sustainable Built Environment xxx (2017) xxx-xxx 9

Table 1

Mapping of CALTRANS linguistic data to numerical rating system.

CALTRANS question3 Answer Choices1 Response Percentage1 (%) Rating Initial Probability for HMM (%)

1 How long do you feel it should take to repair/replace safety barriers, guard Same day 19 High 82

rails, and median barriers? Up to two days Two days to a week 33 30

Two weeks 2 Medium 4

Two to four 2

Depends on other 7 Low 9

priorities

No Opinion 2

2 How long should it take Caltrans to repair signs (excluding Stop and Yield Same day 25 High 86

signs) after they have been damaged? Up to two days Two days to a week 32 28

Two weeks 7 Medium 9

Two to four 2

Depends on other 5 Low 6

priorities

No Opinion 1

3 How long should it take Caltrans to remove graffiti from areas other than Same day 8 High 53

signs? Up to two days Two days to a week 16 29

Two weeks 15 Medium 23

Two to four 8

Depends on other 21 Low 24

priorities

No Opinion 3

4 How would you grade the job Caltrans is doing repairing potholes on the Excellent 11 High 46

highway you drive most often? Good 35

Fair 32 Medium 32

Poor 21 Low 22

No Opinion 1

5 What is the biggest concern you have regarding California's bridges and Structural Safety 40 High 40

overpasses? Traffic Congestion 33 Medium 33

Poor riding 13 Low 27

pavement

Other 14

a http://dot.ca.gov/hq/maint/external_survey/2005_survey/SurveyResults/StatewideResults.pdf.

541 customer requirements would be valid up to the timeframe

542 during which the project is designed, executed and put into

543 service. Once in service, the customer could exhibit varying

544 levels of expectations and addressing that would be chal-

545 lenging. Furthermore, customer surveys which would form

546 the basis of knowing what is required may have to be

547 repeated when conditions surrounding the intended use

548 of the project change. By using a HMM, some of these

549 challenges can be 4addressed by adopting the customer

550 requirements as hidden parameters and be predicted using

551 focus areas defined in the study as observed parameters.

552 From maintenance perspective of a civil engineering struc-

553 ture, five key focus areas were proposed as those that

554 would cause greatest impact to customer concerns (Bolar

555 et al., 2014). In the current study, these focus areas are

556 adopted as indicators or observed parameters for predict-

ing customer response. The observed parameters also 557

require a score or value in order to be mathematically used 558

in prediction. Therefore, the observed parameters are 559

assigned a rating value and are explained below: 560

Economic factors: The customer response to a survey 561

conducted during good economic condition may be differ- 562

ent compared to situations when the economy is weaker. 563

Customer requirements can therefore vary depending on 564

economic conditions and can therefore act as an indicator 565

to predict customer response. In the current study, the eco- 566

nomic condition is rated 0 to 1.0 (poor to good). 567

Social factors: From the point of view of civil infrastruc- 568

ture, changes to social conditions can affect customer per- 569

ception. For example, influx of population may be good 570

to the social condition of a city, but may add more traffic 571

causing congestion which could have a negative response 572

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

A.A. Bolar et al. linternational Journal of Sustainable Built Environment xxx (2017) xxx-xxx

Table 2 Customer requirement states.

STRONG WEAK

Value 5 3 1

Relationship HIGH MEDIUM LOW

in a customer survey. Another example could be that opening a new residential sub-division in the city may result in reduced traffic condition. So rather than waiting for the survey to be conducted and for the customer to respond, a change in social condition can aid in predicting customer response. In the current study, social conditions causing poor-to-good situation on a structure is rated on a scale 0-1.

Safety driven factors: While one could argue that there is no compromise on safety so there should not be a better or weaker safety condition. However, adopting for example, a new technology can result in a better management of safety related issues. Improvement on safety would definitely be seen by customers as a positive step. Safety driven conditions are rated on a scale 0-1 for poor-to-good safety condition.

Technical Factors: An agency might be able to improve on technical engineering conditions by adopting new technologies, designs, materials and technical expertise. Such expertise could be in the form of personnel, software, etc. Therefore, technical conditions can be seen to improve or worsen with time depending on situations. For example, loss of personnel in an agency can result in reduced technical expertise. In the current study, technical conditions ranging from poor-to-good situation are rated on a scale 0-1.

Maintenance efficiency: Similar to technical factors, adopting new technologies, management systems, materials, etc can result in improved maintenance. In the current

study, maintenance efficiency is rated 0-1 ranging from worse to improved condition.

Environmental factors: With customers becoming more conscious over issues involving sustainability, environmental factors surrounding a maintenance decision can be of great concern to the customer. Improved environmental conditions can hold satisfactory value to the customer, whereas degrading environmental conditions may be unsatisfactory. Therefore, predicted environmental factors can correlate positively with customer responses in a survey. In the current study, predicted environmental factors are rated 0-1 ranging from worse to improved condition.

The rating system explained for each of the above observed parameters requires expert input in order to be assigned. Few parameters, economic condition for example, may already have predictions by various experts and even on a daily basis. Social factors could be extrapolated from government agencies dealing with such issues. Safety Driven, Technical Factors, Maintenance Efficiency and Environmental Factors could be rated by the management authority of the agency seeking customer input. Since there can be variation in the rating depending on the confidence of the expert assigning the rating value, a credibility factor is applied to the ratings obtained. For the current study, the rating is obtained based on expert judgement of the authors. The ratings, credibility factors and reasoning have been provided in Table 3.

Having established the QFD customer requirements scale, transition probabilities, and emission probabilities in Table 3, all these parameters can be input into equations 1-12 for estimating future probabilities of customer requirements. The CRs can then be used in QFD for prioritization of HOWs using QFD relationship and correlation matrices. The QFD relationship matrix and correlation

Table 3

Expert (authors) opinion on transition/emission probabilities.

Credibility

Transition Probabilities Emission Probabilities

Summary

Factors Weak Average Strong Weak Average Strong

Economic 0.8 Weak 0.50 0.40 0.10 0.40 0.40 0.02 Transitions probabilities are assigned for each of the

Average 0.30 0.60 0.10 0.70 0.30 0.00 categories transiting from Weak-to- Strong/Average/

Strong 0.30 0.60 0.10 0.30 0.60 0.10 Weak, Average-to- Strong/Average/Weak, Weak-to-

Social 0.7 Weak 0.60 0.40 0.00 0.70 0.30 0.00 Strong/Average/Weak.

Average 0.30 0.60 0.10 0.20 0.60 0.20 For each of the Categories, credibility factors have been

Strong 0.00 0.20 0.80 0.00 0.40 0.60 assigned based on the confidence with which the expert

Safety 0.65 Weak 0.60 0.40 0.00 0.70 0.30 0.00 could have assigned the probabilities. For example, the

Average 0.30 0.60 0.10 0.20 0.60 0.20 economic expert may have a 80% confidence in judging

Strong 0.00 0.20 0.80 0.00 0.40 0.60 the transitions, whereas an expert on social issues many

Technical 0.75 Weak 0.60 0.40 0.00 0.70 0.30 0.00 have only 70% confidence in the social conditions

Average 0.30 0.60 0.10 0.20 0.60 0.20 predicted.

Strong 0.00 0.20 0.80 0.00 0.40 0.60

Maintenance 0.9 Weak 0.60 0.40 0.00 0.70 0.30 0.00

Efficiency Average 0.30 0.60 0.10 0.20 0.60 0.20

Strong 0.00 0.20 0.80 0.00 0.40 0.60

Environmental 0.85 Weak 0.60 0.40 0.00 0.70 0.30 0.00

Average 0.30 0.60 0.10 0.20 0.60 0.20

Strong 0.00 0.20 0.80 0.00 0.40 0.60

A.A. Bolar et al. / International Journal of Sustainable Built Environment xxx (2017) xxx-xxx

RE1AT1VE IMPORTANCE 7.6 8.2 8.2 7.6 10.0 14.8 14.8 14.2 14.2 13.7 9.7 13.6 15.4 13.3 11.2 13.6 13.3 13.0 12.8

CALTRANS QUESTION No. (TABLE 1) ECONOMIC SOCIAL SAFETY ! X MAINTENANCE EFFICIENCY ENVIRONMENTAL COMBINED Increase Taxes Borrowing Toll Customer Satisfaction Aesthetics Heritage Value Traffic Signs t Safety during Maintenance Pedestrian Safety Technical Investigation Vibration/Movement Seismic Couter-measures Improved Access Uneven Surface No Potholes Green Products Green Design Green Construction

1 4.18 2.54 3.21 3.21 3.21 3.21 3.26 9 9 9 9 7 7 0 5 0 1 7 7 7 7 7 5 7 7 7

2 4.26 4.26 4.26 4.26 4.26 4.26 4.26 7 7 7 9 7 9 5 7 0 5 9 7 7 5 5 5 5 5 5

3 3.80 3.80 3.80 3.80 3.80 3.80 3.80 7 7 1 9 1 7 0 1 0 1 1 1 1 0 0 0 0 0 0

4 3.40 3.40 3.47 3.40 3.40 3.52 3.43 7 7 7 9 1 1 0 7 0 7 7 7 7 7 7 7 7 7 7

5 3.02 3.02 3.02 5.08 3.33 3.02 3.42 7 7 7 9 9 3 0 3 0 3 3 3 3 5 5 5 5 5 5

133.7 133.7 110.9 163.5 90.63 101.4 21.3 84.21 0 62.64 99.25 90.73 90.73 85.25 85.25 78.72 85.25 85.25 85.25

22 21 10 17 15 15

Economic Social Safety Technical Maintenance Efficiency Environmental

Fig. 4. Customer requirement (relative importance) obtained using Hidden Markov analysis.

matrix in Fig. 4 were obtained from expert judgment of the authors using a 5 tier scale between 1 and 9, where 1 represents weak relationship and 9 represents strong relationship. For example, most of the questions in Table 3 are related to customer satisfaction and therefore a rating '9' has been assigned to the HOW ''customer satisfaction." Question No. 3 in Table 3 is related to removing graffiti from areas and the use of "Toll" for removing graffiti would be a very unlikely situation and therefore a relationship rating of "one" has been assigned in this case. Using economic condition probabilities from Table 3.

Customer Requirement States, s

0 0 10

v -1f1

Transition Probability Matrix, FtJ

Emission Probability Matrix, v

0.5 0.4 0.1

0.3 0.6 0.1

0.3 0.6 0.1

0.6 0.7 0.3

-0.3 1.2 1.7 -0.8 -6.3 7.2

(Note that n = 1 for step 1, therefore Pn

0.69 0.30 0.01 V-lFV = \ 0.49 0.5 0.01 [ 1.29 -0.3 0.01

f 1jV -1fn v ={ 0.67 0.32 0.01 }

SF yV -1fn v = 4.35

Therefore, the weight of customer requirement in the future after 1-step is evaluated as 4.35. For step two, the transition probability matrix has to be raised to the power 2, therefore, Pn = P2

0.24 0.64 -0.96

0.66 0.26 1.86

0.1 0.1 0.1

662 663

668 669

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

A.A. Bolar et al. linternational Journal of Sustainable Built Environment xxx (2017) xxx-xxx

Table 4

Summary of customer requirements from 2-step hidden Markov analysis.

CALTRANS question No. (Ref Table 1) Economic Social Safety Technical Maintenance Efficiency Environmental COMBINED

1 4.18 2.54 3.21 3.21 3.21 3.21 3.26

2 4.26 4.26 4.26 4.26 4.26 4.26 4.26

3 3.80 3.80 3.80 3.80 3.80 3.80 3.80

4 3.40 3.40 3.47 3.40 3.40 3.52 3.43

5 3.02 3.02 3.02 5.08 3.33 3.02 3.42

681 682

0.64 0.35 0.01

V-1F2 = V-1FFV = { 0.60 0.39 0.01 v 0.76 0.23 0.01

F1jV-1FnV = { 0.63 0.35 0.01 }

SF 1jV-1 FnV = 4.18

Therefore, after two time steps, the customer requirement in future is evaluated to be 4.18.

5. Discussion

Predicting the customer's opinion from a cognitive point of view may be a complex process and outside the scope of this research. However, starting with available customer responses as input, the focus areas of attention for the customer (economic, social, safety driven, technical, maintenance efficiency and environmental issues) were adopted for estimating future customer requirements. In the above worked example, existing customer requirements for CAL-TRANS question 1 (time taken to repair/replace safety barriers, guard rails, and median barriers) were systematically used in a 1-step HMM followed by two steps, in order to demonstrate how future customer requirements at two time steps can be evaluated. The customer requirements from 2-step hidden Markov analysis are provided in Table 4. Using a simple VBA macro, the same calculation methodology has been extended to any number of time steps possible. Furthermore, the remaining focus areas can also be included and future customer requirements can be determined. These future requirements can be used, for example, in prioritization, for identifying at what future state, which QFD HOW (action for satisfying customer requirement) would be of importance. Using the prioritization scheme in Bolar (2014), the future customer requirements are input to QFD House of Quality for updating prioritizations evaluated. Fig. 4 shows the HOQ with CRs evaluated using the HMM.

6. Conclusions

A framework that utilizes hidden Markov model has been employed for dealing with dynamic customer requirements in an application involving infrastructure maintenance. Customer involvement can be minimal in civil engineering applications leading up to design stages. However, in engineering maintenance, customer requirements

can be considered truly dynamic - as time proceeds, main- 724

tenance needs of a structure change and so does the cus- 725

tomer requirements. Capturing customer requirements 726

constantly may not be possible as conducting customer sur- 727

veys, collecting information, synthesis, and generating out- 728

comes involve time, human efforts and cost. Using a hidden 729

Markov model, expert opinion can be sought from individ- 730

uals, government agencies or departments with regards to 731

focus areas for the customer. Using those expert probabil- 732

ities for the focus area as hidden Markov parameters, esti- 733

mation of future customer requirements have been 734

circumvented thereby eliminating the need for repeated 735

customer surveys. 736

7. Scope for further research 737

While a solved example and case study have been pre- 738

sented in this study for demonstrating the application, var- 739

ious improvements are recommended for investigation. To 740

start with, additional focus areas that could be of interest 741

to the customer using infrastructure can be studied from 742

a social sciences perspective. Relevant research on infras- 743

tructure users by disciplines such as psychology, for exam- 744

ple, could provide valuable input on infrastructure user 745

attitudes. Such research may aid in conducting probabilis- 746

tic studies of infrastructure user attitude for each of the 747

focus areas thereby leading to better prioritization and 748

decision-making. Next, use of the proposed model at a rel- 749

evant maintenance agency could uncover further areas of 750

improvement and further validate the methodology. For 751

example, successive surveys could be performed at a few 752

time intervals and the results could aid in calibrating the 753

probabilities proposed by experts in each of the focus 754

areas. 755

A HMM is based on the assumption that steady state 756

probability distribution is reached thereby forming the 757

basis for using initial probabilities in evaluating future 758

states. Further investigation into the application of 759

non-stationary hidden Markov models is therefore rec- 760

ommended for improvement. Boudaren et al. (2012) 761

has presented an application in digital signal processing 762

for fusion of multistationary signals in nonstationary 763

Markovian context. Cassidy et al. (2002) has imple- 764

mented a variational Bayesian algorithm for biomedical 765

signal analysis. Similar research can be carried over for 766

infrastructure applications involving hidden Markov 767

modelling. 768

IJSBE 152 ARTICLE IN PRESS No. of Pages 14

30 March 2017

A.A. Bolar et al. l International Journal of Sustainable Built Environment xxx (2017) xxx-xxx 13

Finally, the estimated probabilities are assumed to be free of uncertainties, and therefore the model may be improved to handle uncertainties. Given the advances in uncertainty modelling, due consideration should be given to methodologies involving Bayesian modelling (or evidence theory) coupled with hidden Markov models. Methods such as Sugeno fuzzy measures and Choquet integrals have been used especially in the field of Bioinformatics. Furthermore, there are mathematically complexities if there are dependencies in the data being considered. Robust techniques for dealing with such issues using Evidential Reasoning (ER) were presented by Hughes (2009). More exact formulations using choquet integrals have been developed by Soubaras (2009). All these studies can be explored for application in infrastructure maintenance. In addition to mathematical challenges, including these advanced modelling techniques would require vast amounts of computer programming and simulation that could prove valuable in the long run.

8. Uncited references

Cariaga et al. (2007), Gargione (1999), Kobayashi et al. (2012), Lair et al. (2004), Lami and Vitti (2011), Lethanh et al. (2015), Radharamanan et al. (2008), ReVelle et al. (1998), Soderqvist and Vesikari (2003), and Rivenbark and Ballard (2011).

References

Adey, B.T., Klatter L., Kong J, S., 2010. The IABMAS Bridge management committee overview of existing bridge management systems. http://128.180.11.237/IABMAS/bodies/IABMAS-BMC-BMS-Report-20100806%5B1%5D.pdf (accessed 11.04.16). Ahmed, S.M., Sang, L.P., Torbica, Z.M., 2003. Use of quality function deployment in civil engineering capital project planning. J. Constr. Eng. Manage. 129 (4), 358-368. ASCE, 2013. 2013 report card for America's Infrastructure. Technical Report by the American Society of Civil Engineers. http://www. infrastructurereportcard.org/a/documents/2013-Report-Card.pdf (accessed 11.04.16). Bhuyan, P.K., Nayak, S.M., 2013. A review of level of service analysis of urban streets. Transport Rev. 33 (2), 219-238. http://dx.doi.org/ 10.1080/01441647.2013.779617. Bolar, A., Tesfamariam, S., Sadiq, R., 2012. Condition assessment for bridges: a hierarchical evidential reasoning (HER) framework. Struct. Infrastruct. Eng. 9 (7), 648-666. http://dx.doi.org/10.1080/ 15732479.2011.602979. Bolar, A., Tesfamariam, S., Sadiq, R., 2014. Management of civil infrastructure systems: A QFD-based approach. J. Infrastruct. Syst.. http://dx.doi.org/10.1061/(ASCE)IS.1943-555X.0000150. Bolar, A.A., 2014. A Quality Function Deployment (QFD) Approach for Bridge Maintenance Management. The University of British Columbia, Kelowna, Canada, Okanagan School of Engineering. Bunks, C., McCarthy, D., Al-Ani, T., 2000. Condition-based maintenance of machines using hidden Markov models. Mech. Syst. Signal Process. 14 (4), 597-612.

California Department of Transportation CALTRANS, 2005. 2005 Statewide Results - Maintenance Customer Survey. http://dot. ca.gov/hq/maint/external_survey/2005_survey/SurveyResults/Statewi-deResults.pdf (accessed 11.04.16).

Cariaga, I., El-Diraby, T., Osman, H., 2007. Integrating value analysis and quality function deployment for evaluating design alternatives. J. Constr. Eng. Manage. 133 (10), 761-777.

Cassidy, M.J., Penny, W.D., 2002. Bayesian nonstationary autoregressive models for biomedical signal analysis. IEEE Trans. Biomed. Eng. 49 (10), 1142-1152.

Carnevalli, J.A., Miguel, P.C., 2008. Review analysis and classification of the literature on QFD—types of research difficulties and benefits. Int. J. Prod Econ. 114 (2), 737-754.

Chan, L., Wu, M., 2002. Quality function deployment: A literature review. Eur. J. Operat. Res. 143 (3), 463-497.

Chong, Y.T., Chen, C.H., 2010. Customer needs as moving targets of product development: a review. Int. J. Adv. Manufact. Technol. 48 (14), 395-406.

Federation of Canadian Municipalities and National Research Council (FCM & NRC), 2003. Developing Levels of Service Federation of Canadian Municipalities and National Research Council https://www. fcm.ca/Documents/reports/Infraguide/Developing_Levels_of_Ser-vice_EN.pdf, accessed 11.04.16.

Felio, G., 2012. Canadian Infrastructure Report Card Vol 1 2012 Municipal Roads and Water Systems. Canadian Construction Association; Canadian Public Works Association; Canadian Society for Civil Engineering; Federation of Canadian Municipalities. http:// www.canadainfrastructure.ca/downloads/Canadian_Infrastructure_ Report_Card_EN.pdf (accessed 11.04.16).

Felio, G., Lounis, Z., 2009. Model framework for assessment of state, performance, and management of Canada's core public infrastructure. NRTSI - Services Committee & NRC - Assets Committee, Ottawa, ON.

Francisque, A., Rodriguez, M.J., Sadiq, R., Miranda, L.F., Proulx, F., 2011. Reconciling 'actual' risk with 'perceived' risk for distributed water quality: a QFD-based approach. J. Water Supply Res. Technol. AQUA 60 (6), 321-342.

Gargione, L.A., 1999. Using quality function deployment (QFD) in the design phase of an apartment construction project. In: Tommelein, I. D., Ballard, G. (Eds.), Proc. 7th Annu. Conf. of Int. Group for Lean Constr.. University of California, Berkeley, CA, pp. 357-368.

Hughes, J., 2009. Computing the observed information in the hidden Markov model using the EM algorithm. Stat. Probabil. Lett. 32 (1), 107-114.

Junhai, M.A., Chen, A., Jun, H.E., 2007. General framework for bridge life cycle design (Translated from Chinese). J. Tongji Univ. (Natural Science) 35 (8), 1003-1007.

Kamara, J.M., Anumba, C.J., Evbuomwan, N.F.O., 1999. Client requirements processing in construction: a new approach using QFD. J. Architect. Eng. 5 (1), 8-15.

Kasemir, B., Jaeger, J., Jaeger, C., Gardner, T., 2003. Public Participation in Sustainability Science: A Handbook. Cambridge University Press, West Nyack, NY, USA.

Kobayashi, K., Kaito, K., Lethanh, N., 2012. A statistical deterioration forecasting method using hidden Markov model for infrastructure management. Transport. Res. B 46 (4), 544-561.

Lair, J., Rissanen, T., Sarja, S., 2004. Methods for optimization and decision-making in lifetime management of structures LIFECON deliverable D2.3.Technical Research Centre of Finland (VTT), VTT Building and Transport, European Community, Fifth Framework Program: GROWTH. http://lifecon.vtt.fi/d23.pdf (accessed 11.04.16).

Lami, I.M., Vitti, E.L., 2011. A combination of Quality Function Development and Analytic Network Process to evaluate urban redevelopment projects: an application to the Belle de Mai - la Friche of Marseille France. J. Appl. Oper. Res. 3 (1), 2-12.

Lee, D.E., Arditi, D., 2006. Total quality performance of design/build firms using quality function deployment. J. Constr. Eng. Manage. 132 (1), 49-57.

Lethanh, N., Adey, B.T., 2013. Use of exponential hidden Markov models for modelling pavement deterioration. Int. J. Pave. Eng. 14 (7), 645654.

14 A.A. Bolar et al. linternational Journal of Sustainable Built Environment xxx (2017) xxx-xxx

893 Lethanh, N., Kaito, K., Kobayashi, K., 2015. Infrastructure deterioration

894 prediction with a poisson hidden markov model on time series data. J.

895 Infrastruct. Syst. 21 (3). http://dx.doi.org/10.1061/(ASCE)IS.1943-

896 555X.0000242, 04014051.

897 Local Government Association of Southern Australia (LGASA), 2006.

898 Guidelines for Infrastructure and Asset Management Plan Template.

899 Institute of Public Works Engineering Australia. https://www.lga.

900 sa.gov.au/webdata/resources/files/Infrastructure_and_Asset_Manage-

901 ment_Plan_Template_Guidelines.pdf (accessed 11.04.16).

902 Malekly, H., Mousavi, S.M., Hashemi, H., 2010. A fuzzy integrated

903 methodology for evaluating conceptual bridge design. Expert Syst.

904 Appl. 37 (9), 4910-4920.

905 Mallya, G., Tripathi, S., Kirshner, S., Govindaraju, R., 2013. Probabilistic

906 assessment of drought characteristics using hidden Markov Model. J.

907 Hydrol. Eng. 18 (7), 834-845.

908 Mckinsey Global Institute, 2013. Infrastructure productivity: how to save

909 $1 Trillion a year. http://www.mckinsey.com/~-/media/McKinsey/dot-

910 com/Insights%20and%20pubs/MGI/Research/Urbanization/Infras-

911 tructure%20productivity/MGI_Infrastructure_Full_report_Jan2013.

912 ashx.

913 Milan, M., Barros, J.W.D., Gava, J.L., 2003. Planning soil tillage using

914 Quality Function Deployment (QFD). Sci. Agricol. 60 (2), 217-221.

915 Miller, J.D., 2013. Infrastructure 2013: Global priorities, global insights.

916 Urban Land Institute, Ernst & Young, Washington, DC, USA, http://

917 www.ey.com/Publication/vwLUAssets/Infrastructure_2013/%

918 24FILE/Infrastructure_2013.pdf on accessed (10.02.14).

919 Mirza, S., 2007. Danger ahead the coming collapse of Canada's municipal

920 infrastructure. Federation of Canadian Municipalities, Ottawa, Ont.,

921 http://site.ebrary.com/id/10202348.

922 Rabiner, L.R., Juang, B.H., 1986. An introduction to hidden Markov

923 models. IEEE ASSP Mag., 4-16

924 Rabiner, L.R., 1989. A tutorial on hidden Markov models and selected

925 applications in speech recognition. Proc. IEEE 77 (2), 257-286.

926 Radharamanan, R., Juang, Jeng-Nan, Felix, C.J.A.K., 2008. Service

927 quality in a collective urban transportation system. 2008 IEEE Int.

928 Symp. Service-Oriented Syst. Eng., 227-233

929 Rammohan, R., Taha, M.R., 2005 Exploratory investigations for intel-

930 ligent damage prognosis using hidden Markov models. In: Systems,

931 Man and Cybernetics 2005 IEEE International Conference (2) 1524932 1529.

933 ReVelle, B.J., Moran, J.W., Cox, C.A., 1998. The QFD Handbook. John

934 Wiley & Sons, New York, NY, USA.

935 Sarja, A., 2004. Reliability based life cycle design and maintenance

936 planning LIFECON deliverable D2.1. Technical Research Centre of

937 Finland (VTT), VTT Building and Transport, European Community,

Fifth Framework Program: GROWTH. http://lifecon.vtt.fi/d21.pdf 938

(accessed 11.04.16). 939

Sepideh, S., Aaghaie, A., 2011. Introducing busy customer portfolio using 940

hidden Markov model. Iran. J. Manage. Stud. 4, 99-119. 941

Sharma, J.R., Rawani, A.M., Barahate, M., 2008. Quality function 942

deployment: a comprehensive literature review. Int. J. Data Anal. 943

Tech. Strategies 1 (1), 78-103. 944

Shen, L., Bai, L., 2006. A review on Gabor wavelets for face recognition. 945

Pattern Analysis and Applications 9 (2-3), 273-292. 946

Shieh, J., Wu, H., 2009. Applying a hidden Markov chain model in quality 947

function deployment to analyze dynamic customer requirements. J. 948

Qual. Quantity 43 (4), 635-644. 949

Soderqvist, M.K., Vesikari, E., 2003. LIFECON LMS generic technical 950

handbook for a predictive life cycle management system of concrete 951

structures LIFECON deliverable D1.1. Technical Research Centre of 952

Finland (VTT), VTT Building and Transport, European Community, 953

Fifth Framework Program: GROWTH. Retrieved on May 22, 2012 954

via: http://lifecon.vtt.fi/d11.pdf (accessed 11.04.16). 955

Soubaras, H., 2009. Symbolic and quantitative approaches to reasoning 956

with uncertainty. Volume 5590 of the series Lecture Notes in 957

Computer Science. pp. 863-874. 958

Rivenbark, W.C., Ballard, E.C., 2011. Using citizen surveys to influence 959

and document culture change in local government. Public Performance 960

Manage. Rev. 35 (3), 475-484. 961

Utne, I.B., 2009. Improving the environmental performance of the fishing 962

fleet by use of Quality Function Deployment (QFD). J. Cleaner 963

Product. 17 (8), 724-731. 964

Van Luu, T., Kim, S.Y., Truong, T.Q., Ogunlana, S.O., 2009. Quality 965

improvement of apartment projects using fuzzy-QFD approach: A 966

case study in Vietnam. KSCE J. Civil Eng. 13 (5), 305-315. 967

Xu, W., Yan, X., Wu, C., 2011. A recognition method for lane change 968

intention based on hidden Markov model. Proceedings of the First 969

International Conference on Transportation Information and Safety. 970

Yang, M., Khan, F.I., Sadiq, R., Amyotte, P., 2011. A rough set-based 971

quality function deployment (QFD) approach for environmental 972

performance evaluation: a case of offshore oil and gas operations. 973

t JOCP 19 (13), 1513-1526. 974

Yasamis-Speroni, F., Lee, D.E., Arditi, D., 2012. Evaluating the quality 975

performance of pavement contractors. J. Constr. Eng. Manage. 138 976

(10), 1114-1124. 977

Yu, J., 2012. Health condition monitoring of machines based on hidden 978

markov model and contribution analysis. IEEE Trans. Instrum. Meas. 979

61 (8), 2200-2211. 980

Zhang, Y., Wang, H.P., Zhang, C., 1999. Green QFD-II: a life cycle 981

approach for environmentally conscious manufacturing by integrating 982

LCA and LCC into QFD matrices. Int. J. Res. 37 (5), 1075-1091. 983