Product-cost modelling approach for the development of a decision support system for optimal roofing material selection

Expert Systems with Applications 39 (2012) 6857–6871

Contents lists available at SciVerse ScienceDirect

Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa

Product-cost modelling approach for the development of a decision support system for optimal roofing material selection Sazzadur Rahman a,⇑,1, Henry Odeyinka a, Srinath Perera b, Yaxin Bi c a

School of Built Environment, University of Ulster, Newtownabbey, BT37 0QB, United Kingdom School of Built & Natural Environment, Northumbria University, Newcastle, United Kingdom c School of Computing and Mathematics, University of Ulster, Newtownabbey, BT37 0QB, United Kingdom b

a r t i c l e

i n f o

Keywords: TOPSIS Knowledge-based system Decision support system Roofing material selection Energy efficient building Product-cost modelling

a b s t r a c t Selection of optimal roofing materials is very important but it is a complex and onerous task as varieties of materials are available for housing roof construction. In order to select suitable materials, an extensive range of criteria would need to be considered. This paper presents the framework and the development of a knowledge-based decision support system for material selection implemented in roofing material selection domain, called ‘Knowledge-based Decision Support system for roofing Material Selection and cost estimating’ (KDSMS). It was developed to facilitate the selection of optimal materials for different roof sub elements. The system consists of a database and knowledge base that is equipped with an inference engine. The former is used to store different types of roofing materials with assigned attribute values. The later is used to hold qualitative and quantitative knowledge which were collected from domain experts and other technical literatures such as building regulations, price guide book and product catalogues. The proposed system employs the TOPSIS (Technique of ranking Preferences by Similarity to the Ideal Solution) multiple criteria decision making method to solve materials selection and optimisation problem. This study utilised the available roofing materials in the UK housing market in developing the system reported. The main contribution of the developed system is that it provides a tool for the architects, quantity surveyors or self house builder to select optimal materials from a wide array of possibilities for different roof sub elements and also to estimate the conceptual cost for the roof element in the early stage of building design. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Different types of materials and technologies are available for building design and construction while new materials and advanced technologies are continuously being introduced into the market (Wong & Li, 2008). The selection of materials is a complex procedure and it is difficult to match materials based on design requirements (Ashby, Brechet, Cebon, & Salvo, 2004). Materials are generally selected from the existing catalogues of materials and traditionally experts apply trial and error methods or use experiences to choose new materials or materials having better performance (Shanian & Savadogo, 2006). It is acknowledged that the selection of appropriate materials may reduce the energy consumption and maintenance cost of buildings (Papadopoulos & Giama, 2007). As buildings are responsible for significant impact on the environment, eco-friendly materials are becoming popular for housing construction (Hymers, 2006). ⇑ Corresponding author. Tel.: +44 7902977669. 1

E-mail address: [email protected] (S. Rahman). Now based at Robert Gordon University, Aberdeen, United Kingdom.

0957-4174/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2012.01.010

Moreover, there is an increasing demand for sustainable and energy-efficient construction (UNEP, 2001) and the use of environmental friendly materials (Chan & Tong, 2007; Roaf, Fuentes, & Thomas, 2007). However, there is a lack of public awareness about sustainable and energy-efficient construction and these issues are unfamiliar to many architects, engineers, and contractors (UNEP, 2001). Evidences from literature suggest that the building owners and clients tend to emphasise the initial cost rather than operating cost (Wilson et al., 1998). Karolides (2006) and Woolley (2006) emphasised that the amount of energy needed can be reduced by using high performance and extra insulation, which is the easiest and least expensive way to solve energy problem. Architects or cost engineers need to consider several factors in order to select optimum materials to meet clients’ requirements. In order to solve this problem of material selection in a way that meets design and clients’ requirements and results in sustainable construction, it is required to analyse and synthesis a multitude of criteria (Perera, Odeyinka, & Bi, 2009a, 2009b). Different approaches regarding materials selection have been devised for different purposes. For instance, knowledge-based or expert systems have been developed to select materials for different purposes.

6858

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

Bullinger, Warschat, and Fischer (1991) proposed a knowledgebased system to select optimal materials for construction with fibre-reinforced composite materials. Soronis (1992) proposed a method for the selection of roofing materials where several factors have been taken into consideration to assess durability only. Chen, Sun, and Hwang (1995) developed an intelligent system for composite material selection in structural design. Mahmoud, Aref, and Al-Hammad (1996) developed a method for selection of finishing materials that covered floors, walls and ceilings. Mohamed and Celik (1998) proposed a knowledge-based method regarding materials selection and cost estimating for a residential building where users could choose their preferred one from a list of materials without evaluation and synthesis of multiple design criteria and client requirements. Instead of expert or knowledge-based systems, Perera and Fernando (2002) proposed a cost modelling system for roofing material selection where several factors are identified and considered in the selection process. Chan and Tong (2007) acknowledged the fact that the decision to select appropriate material is not simply a consideration of cost and materials properties but also there is a need to consider environmental impacts. It is identified that the selection of material is a key issue for the environment (Chan & Tong, 2007) and the choice of material is the optimal way to achieve the energy efficient construction of a building (Krope & Goricanec, 2009). In view of the foregoing, the design team needs to consider several factors in order to select the more suitable materials to meet clients’ requirements. In order to solve this problem of material selection in a way that meets the requirement of the design team and those of the construction clients and results in sustainable construction and cost effective solutions, it is required to simultaneously analyse and synthesise multitudes of criteria in order to achieve an optimum solution. It is identified that few decision support systems have been devised for roofing materials selection but the proposed systems do not have the facility to select the appropriate materials by evaluating them with respect to the multitudes of criteria to be considered in order to meet the clients’ expectations. Some systems attempt to solve the problem of materials selection by adopting rule-based knowledge representation in terms of IFTHEN rules. However, it is difficult to rank the most suitable materials using conditional expressions. This clearly indicates a research gap with respect to selecting the optimum roofing materials by analysing and synthesising a multitude of design and client’s requirements that are both cost effective and sustainable. In order to fill this gap, it is necessary to develop a system that has the capability of simultaneously evaluating multiple criteria in the optimisation of materials selection for roof design. Hence, this research aims to bridge the current knowledge gap by developing a knowledge-based decision support system, called Knowledge-based Decision support System for roofing Material Selection and cost estimating (KDSMS). Its aim is to optimise the selection of roofing materials and model the associated cost for the roof element at an early stage of building design. This system adopts the Technique Of ranking Preferences by Similarity to the Ideal Solution (TOPSIS) method to solve Multi Criteria Decision Making (MCDM) problems. The advantage of this method is its efficiency and simplicity to use and the ability to rank the materials indisputably (Shanian & Savadogo, 2006). Architects, Cost Engineers, Quantity Surveyors and self builders are the potential users of this system. It has the potential of assisting them in selecting optimal materials from the list of alternatives based on the level of importance of the criteria set by them. In addition, the system estimates the cost of the optimal materials selected to determine the budget. This system also can be used to educate the users about new materials by providing relevant information.

2. Methodology The essence of this research is to develop a method for evaluating multitudes of criteria in order to select optimal roofing materials and also to estimate the associated cost. Extensive literature review was carried out as an initial step to identify the multitude of criteria to consider in roofing material selection. This was followed by structured questionnaire survey and interviews of domain experts in order to elicit the relevant knowledge for building the decision support system. Upon developing the system, relevant data were also gathered to test and evaluate the system. The data set included elemental costs, total roof costs and the level of importance attached to selection criteria of material. The data set were collected from case study projects obtained from housing developers in order to evaluate the system. Fig. 1 illustrates the research methodology used in developing the system. The key stages of the research methodology are explained in detail in the following subsections. 2.1. Knowledge gathering A forum of domain experts consisting of four industry specialists (Architects and Quantity Surveyors) and academics were used as the main source for knowledge elicitation. Relevant knowledge was elicited both at the pre and post development phases of the system. The elicited knowledge can be divided into qualitative and quantitative categories. Qualitative knowledge includes material selection criteria, material selection process, material selection regulations for thermal requirements and cost adjustment techniques. The qualitative knowledge were compiled into rules and built into the knowledge base. Material selection criteria for different roof sub elements were identified through extensive literature review. Fifteen criteria were initially identified (Rahman et al., 2009a, Perera, Odeyinka, & Bi, 2009b). Material selection criteria, cost estimating and adjustment processes were identified through a questionnaire survey of the domain experts. This was followed by further interviews. The expert forum was also involved in validating the inclusion of the criteria and also in determining the subjective or objective type from the identified material evaluation criteria. Ten quantitative criteria were decided by the expert forum for the selection process. Out of the ten material selection criteria identified in Table 1, five were used for roof structure, eight criteria were used for roof coverings, seven criteria were used for roof insulation, five criteria were used for roof drainage, six criteria were used for roof lights and two were used for roof features. These ten roofing material selection criteria represent the attributes or properties of materials. Thermal regulation guidelines were obtained from Technical Booklet F (2008), Conservation of fuel and power, by Building Regulations of Northern Ireland. The qualitative knowledge acquired from domain experts was conceptualised and then analysed using Inferential Modelling Technique and by applying top-down and bottom-up techniques (Chen et al., 1995 cited in Zhou, Huang, & Chan, 2004). Using these techniques, the main tasks were decomposed into several subtasks. Then the subtasks were further decomposed so that each of them can be handled easily. 2 shows the process decomposition diagram of the KDSMS system consisting of three main tasks, namely, Select Material, Estimate Cost and Maintain System; and some other associated subtasks. Top-down technique was applied to the Select Material and Maintain System tasks to decompose it into subtasks to define material selection tasks and maintain materials for different roof sub elements. Conversely, bottom-up technique was applied to Estimate Cost task where cost of different roof sub element was estimated separately to obtain the total cost.

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

6859

Literature Search

Cost Modelling

Multi Criteria Decision Making (MCDM)

Materials Selection and Roofing Materials

Literature Review Knowledge Gathering

Identification of Criteria for Roofing Material Selection

MCDM and Cost Estimate Methods Selection

KDSMS Framework Expert Forum with Domain Experts Material Information Gathering

Knowledge Gathering from Experts

Academicians Practitioners

Structuring the Information

Knowledge Analysis

Knowledge Base Creation

Data Modelling KDSMS System Development

Process Modelling

Prototype Development

Prototype Implementation

KDSMS System Evaluation

Data Validation through Case Study

Fig. 1. The research process.

Table 1 Criteria for materials selection for different roof sub elements. Criteria of materials selection

Values or units

Roof sub-element where used

Life span

Years

Initial cost

Materials and labour cost per unit

Sustainability Ease of installation Freedom from maintenance Maintenance cost Life cycle cost

BREEAM rating- A+, A, B, C, D and E Labour hours Replacement factor (life expectancy of roof divided by lifespan of material) Percentage of initial cost Maintenance cost multiplied by replacement factor and added by initial cost U value (W/m2K) Kilogram/m2 Millimetre

Roof structure, roof coverings, roof insulation, roof drainage, roof lights and roof features Roof structure, roof coverings, roof insulation, roof drainage, roof lights and roof features Roof structure, roof coverings, roof insulation and roof drainage Roof structure, roof coverings, roof insulation and roof lights Roof structure, roof coverings, roof insulation, roof drainage, roof lights and roof features Roof coverings Roof coverings and roof drainage

Thermal performance Weight Thickness

The quantitative knowledge is the values of roofing material properties obtained from the ten selection criteria as shown in

Roof insulation and roof lights Roof coverings Roof insulation

Table 1. The internet is a vast source of information and many manufacturers publish their product description on their web sites.

6860

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

Manufacturers are the primary sources to provide information about their materials. The values of roofing material properties were collected from manufacturers of widely used roofing materials such as Marley Eternit (Eternit, 2010), Monier Redland (Redland, 2010), Lagan Building Solutions (LBS, 2010), Northstone (Northstone, 2010), Kingspan Insulation (Kingspan, 2010), Knauf Insulation (Knauf, 2010) and Velux (Velux, 2010) along with other materials suppliers. Manufacturers’ product information was gathered as hard copies from their catalogues and also from their websites. Labour hours and cost data were collected from Spon’s Price Guide Book (Langdon, 2010) where the cost data includes material and labour cost. Building Research Establishment (BRE) publishes sustainability information and they are the primary sources for gathering of sustainability information. The sustainable rating was considered according to the green guide of Building Research Establishment Environmental Assessment Method (BREEAM) where A+, A, B, C, D and E ratings were used (Anderson, Shiers, & Steele, 2009). Different types of roofing materials were gathered throughout the development of the system and the values of material properties on the ten criteria were stored into the spreadsheet files. 2.2. KDSMS system development The system knowledge base was created using relevant information derived from literature review and interviews with the domain experts. Its essence was to provide solutions to roofing materials selection problem and cost estimating. The knowledge base contains the facility for implementing the TOPSIS method for roofing material selection, building regulations for roof, cost estimating and cost adjustment processes. The inference engine was implemented by SQL statements in Oracle PL/SQL environment to manipulate knowledge in the knowledge base to reach the solution. Entity–Relationship (E–R) model was adopted for data modelling. E–R modelling was carried out in two phases – logical data modelling and physical data modelling. The logical data modelling was carried out based on the domain objects and relations decomposition method adopted from the knowledge analysis. The logical data model shows all entities, attributes and the relationships between entities associated with the proposed KDSMS system. The physical data modelling was carried out based on the logical data model. The database was created at this stage where the logical model was more refined, some names of entities and

attributes are shortened, foreign key are assigned against the primary key. The logical data model was transformed into Oracle Database 10 g and a database script file was generated. It is then used to create physical storage in the computer for the database. The entity and attribute in the logical data model were transformed to table and column respectively in the database. An example of a database table is shown in Table 2 which is part of the physical data model. It shows the table name and associated attribute names, data type and primary key defined for an attribute. Process modelling was carried out based on the processes identified during the knowledge analysis. The main system, KDSMS, was decomposed into subsystems or sub processes that can be called a child diagram. The process modelling of this system was carried out based on the process decomposition diagram as in Fig. 2. Context diagram was created first as this is the first level of process modelling. The main system was then decomposed into sub processes next level that can be called a child diagram. The main tasks are represented as parent processes and subtasks are represented as child processes where each child process may have more than two processes. Fig. 3 illustrates an example of a process modelling representation of the KDSMS system where parent processes, data stores and associated information flow are presented. It shows the parent processes Select Material, Estimate Cost and Maintain System. Fig. 3 shows that a user inputs roof information and specification, roofing materials, roof regulations, cost index and materials cost data. The Maintain System process receives the above inputs from the user and stores roofing materials information into Roofing Materials data store, cost data into Approximate Cost data store, roof regulations into Roof Regulation data store, cost index into Cost Index data store and roof project information and specification into Roof Project data store. The user inputs roof information and specification and material selection criteria weights and the Select Material process receives the roof information and specification and criteria weights from the user; stores roof information into data store Roof Project. It obtains the roofing materials information from Roofing Materials data store. It evaluates the materials and stores into Roof Design data store. It then provides the optimal materials selected by the user to the Estimate Cost process. The Estimate Cost process receives the data provided by the Select Material process. It then obtains cost data from Approximate Cost data store, roof area from Roof Project data store. It receives the inputs of desired roof lights area and roof features quantity from the user and obtains cost data from Approximate Cost data store. After estimating the cost for the materials selected, it obtains location

Table 2 Database table. CREATE Table project (project_id varchar2(3), project_name varchar2(30), CONSTRAINT project_pk PRIMARY KEY (project_id));

KDSMS

Select Material

Estimate Cost

Manage Requirements

Compute Total Cost

Evaluate Materials

Adjust Total Cost Fig. 2. KDSMS process decomposition.

Maintain System

Maintain Materials

Maintain Roof Regulation

6861

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

D2

D1 Roofing Materials

Roofing Materials Obtained

Approximate Cost

D5 Roof Project

D3 Roof Regulation

Regulation Commitment

Roofing Materials Commitment Cost Commitment

Roof Information & Specification Roof Project Created

Specification Materials & Cost Roofing Materials Provided

Cost Obtained Roof Information & Specification

Roof Regulations Provided

Roof Area Obtainded 1

Designers/ Self Builders

3

2 Select Materials

Optimal Materials Selected

Estimate Cost

Roof Information & Specification

Maintain System

Criteria Weights Materials Evaluated

Optimal Materials and Cost Commitment

Roof Lights & Features Quantity

D7

Designers/ Self Builders

Location & Tender Price Index Provided Index Commitment

D6 Roof Design Materials Selected and Cost Model

System Maintainance Users

Roof Design Cost

D4

Cost Index

Fig. 3. Process modelling representation of KDSMS main processes.

and tender price index data from Cost Index data store and it then adjusts cost. After adjusting the cost, it stores and optimal materials selected and associated cost into the Roof Design Cost data store and it then provides the materials selected and cost model to the user. After the data and process modelling, the prototype is developed in Oracle Developer Suite 10 g where user interfaces are developed by Oracle Forms and application coding is implemented by Procedural Language/Structured Query Language (PL/SQL).

Step 2: Multiply every computed element of step 1 by the criteria weights to construct the weighted normalised decision matrix, V by the following formula:

V ¼ ½v ij mn ;

where

v ij ¼ rij  wj

ð4Þ

Step 3: Determine the ideal and negative-ideal solution sets from step 2.

Ideal set;Aþj ¼ fthe maximum value under benefit criteria or minimum value under cost

2.3. Materials selection and TOPSIS

criteria for each column of the matrix;

TOPSIS is a multi criteria decision making technique which is based on the idea that the chosen alternative should have the shortest distance from the ideal solution and farthest from the negative ideal solution (Hwang & Yoon, 1981). This method was adopted in this research as the decision support system for the selection of optimal materials. In the TOPSIS method, the criteria weights,wj(j = 1, 2, . . . , number of criteria, n) and values of attributes at criteria, xij(i = 1, 2, . . . , number of alternatives, m; j = 1, 2, . . . , number of criteria, n) build the weighted matrix, W and decision matrix, D respectively; can be expressed as follows (Perera, Odeyinka, & Bi, 2008; Shanian & Savadogo, 2006; Yong, 2006):

W ¼ ½w1 ; w2 ; . . . ; wn  2

x11 6 6  6 D¼6 6  6 4  xm1

3 x1n 7  7 7     7 7 7     5    xmn

ð5Þ

Negative-ideal set Aj ¼ fthe minimum value under benefit criteria or maximum value under cost criteria for each column of the matrix; V; in step 2g ð6Þ Step 4: Calculate the distance of each alternative from the ideal and negative ideal solution sets obtained in step 3.

Sþi

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX u m ¼ t ðv ij  Aþj Þ2

ð7Þ

j¼1

ð1Þ Si

     

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX u m ¼ t ðv ij  Aj Þ2

ð8Þ

j¼1

ð2Þ

Step 5: Calculate the relative closeness to the ideal solution obtained in step 4 to rank the alternatives:

Ci ¼

Step 1: Calculate every element of the decision matrix to obtain normalised matrix, R:

xij rij ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pm 2 i¼1 xij

V; in step 2g

ð3Þ

Sþi

Si þ Si

ð9Þ

where 0 6 Ci 6 1 Step 6: Rank the alternatives obtained in step 5 according to descending order of Ci. TOPSIS is implemented by SQL statements in Oracle PL/SQL environment. Weighted matrix, W and decision matrix, D as per

6862

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

Eqs. (1) and (2) are built separately for each roof sub element. Fig. 4 illustrates an example of the TOPSIS implementation in Oracle PL/ SQL, where it ranks roof coverings materials with respect to lifespan, sustainability and initial cost criteria and provides materials name, TOPSIS score and rank based on the TOPSIS score. The result of a lower step is used as source in an upper step for further calculation. For example, the result of s1 is used for further calculation in s2. The decision matrix, as per Eq. (3), is formed by selecting the material name and values of lifespan, sustainability and initial cost properties from roof coverings materials database table and Eq. (3) is calculated inside s1. The numerical values of lifespan, sustainability and initial cost criteria weights are stored into v_lifespan, v_sustainability and v_initialcost to form weighted matrix as per Eq. (1); and these variables are used in s2 to solve Eqs. (4)– (6) are calculated in s3. Oracle SQL functions MAX () OVER () and MIN () OVER () functions are used to obtain the maximum and minimum values of each column for all materials. The results of both s2 and s3 are used to solve Eqs. (7) and (8) in s4. The TOPSIS score is calculated for all materials in TOPSIS_COVERINGS_VIEW and Eq. (9) is calculated from the result of s4. The materials are then ranked based on the TOPSIS score calculated from TOPSIS_COVERINGS_VIEW and DENSE_RANK () OVER () functions are used to rank the materials. Fig. 5 illustrates the materials ranking mechanism of the system where TOPSIS calculation is part of the Select Material process shown in Fig. 3. When the user provides criteria weighting for each sub element, the system retrieves material data of each sub element based on the roof information provided such as roof span, roof type, and angle of pitch. It also checks against the knowledge provided in the knowledge base such as thermal requirement for roof insulation and roof lights. It applies the rules stored in the knowledge base to retrieve data from database against the roof information and knowledge provided. It then performs TOPSIS calculation to rank the materials. TOPSIS calculation needs decision matrix of material selection criteria weights and performance of alternatives; hence the system forms separate logical decision matrix of each sub element by retrieving properties of materials data from database and

using corresponding criteria weight provided by the user. As TOPSIS ranks the materials based on the criteria weights indicated by the user, the materials ranking can be changed for different order of importance. 2.4. Cost estimating The initial cost of materials supplied in the database comprised of material, labour cost, overheads and profit including VAT; and the cost data is entered in the database and updated using the tender price index (TPI) 580 of greater London area for the first quarter in 2009. The system estimates cost for the optimal materials selected for each roof sub element; and hence the total roofing cost can be different for different materials selected. The cost is estimated for the typical pitched roof specification of a residential building in the UK and the specification is comprised of roof structure, roof coverings, roof insulation, rainwater goods, plasterboard and skim. The cost of roof lights and roof features are added in the total roof cost as extra cost. Total roof cost is calculated as follows:

RC ¼ RSMC þ RLC þ RFC

ð10Þ

Where RC is total roof cost, RSMC is the cost of roof specification comprised of the structure, coverings, insulation and drainage; RLC is roof lights cost, RFC is roof features cost. RSMC in Eq. (10) is calculated as:

RSMC ¼ RA  RSMCm2

ð11Þ 2

2

Where RA is roof area; RSMCm is per m cost to be calculated for new roof specification. RLC in Eq. (10) is

RLC ¼ RLA  RLCm2

ð12Þ 2

2

Where RLA is area covered by roof lights; RLCm is per m cost to be calculated for selected roof lights. RFC in Eq. (10) is calculated by following formula:

RFC ¼ RFQ  RFC u

ð13Þ

Where RFQ is roof features’ quantity; RFCu is unit cost of roof features. The Price Guide Books, such as SPON’s by Langdon (2010), provides elemental cost per unit and approximate cost range per square metre for the general roof specification. The price book does not provide either the exact cost per square metre of a roof specification for specific materials or exact cost per square metre of specific roof lights. Thus it is difficult to estimate exact cost per square metre for a roof specification or roof lights based on the elemental cost of new materials. Moreover, it may be possible to estimate cost based on the past projects if the same specification and materials are used; but it is difficult to estimate cost from the past projects if different types of materials are chosen. In order to resolve this problem, the following formulae were used to calculate per m2 cost for a new roof specification and roof lights. The cost per m2 of a new roof specification, RSMCm 2 in Eq. (11) is calculated as:

RSMCm2 ¼ RSMCm2min "

( )# RSMCm2max  RSMCm2min þ ðRSMC new  RSMC minÞ  RSMC max  RSMC min ð14Þ

RSMCm2min

Fig. 4. SQL code of TOPSIS implementation.

2

Where is the predefined minimum cost per m of a combined roof structure, coverings, insulation and drainage; RSMCm2max is the predefined maximum cost per m2 of a combined roof structure, coverings, insulation and drainage;

6863

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

User Input for weights of criteria for each sub element

Accepts the weights of criteria as linguistic values and convert to numeric values

Thermal requirement data

Retrieves material data from the database based on the roof information and thermal requirment

Roof information

Material database

Ranks materials by TOPSIS calculation

Provides ranking materials to the user Fig. 5. Materials ranking mechanism using TOPSIS.

RSMCmin is the predefined combined minimum unit cost of a combined roof structure, coverings, insulation and drainage; RSMCmax is the predefined combined maximum unit cost of a combined roof structure, coverings, insulation and drainage; RSMCnew is the combined unit cost of a selected roof structure, coverings, insulation and drainage. RSMCnew is calculated by aggregating the unit cost of roof structure, coverings, insulation and drainage materials selected. Eq. (14) can be explained by the following example: The minimum cost per square metre of a roof specification, RSMCm2min , consisted of roof structure, roof coverings, insulation, drainage, plasterboard, and skim is £143.02; The maximum cost per square metre, RSMCm2max , of the same specification is £220; The minimum combined unit cost, RSMCmin, consisted of a roof structure, roof coverings, insulation, and drainage is £239.05; The maximum combined unit cost unit cost, RSMCmax, consisted of a roof structure, roof coverings, insulation, and drainage is £310.66; The cost per unit of the selected new materials of roof structure, roof coverings, roof insulation, and drainage are £126.13, £24.18, £16.22 and £13.67 respectively; and then the combined per unit cost of the selected materials of roof structure, roof coverings, roof insulation, and drainage, RSMCnew, is £180.2. If the above values are substituted in Eq. (14), the RSMCm2 can be calculated as follows:

   220  143:02 RSMCm2 ¼ 143:02 þ ð180:2  239:05Þ  310:66  239:05 ¼ £79:76 If the roof area, RA, is 5168 m2, the eqn. (12) can be calculated as follows:

RSMC ¼ RA  RSMCm2 ¼ 5168  79:76 ¼ £412; 183:5 The cost per square metre of a new roof lights, RLCm2 in eqn. (13) is calculated as follows:

RLCm2 ¼ RLCm2min "

( )# RLCm2max  RLCm2min þ ðRLC new  RLC minÞ  RLC max  RLC min

ð15Þ

Where RLCm2min is the predefined minimum cost per m2 of a roof light; RLCm2max is the predefined maximum cost per m2 of a roof light; RLCmin is the predefined cost per m2 of a roof light; RLCmax is the predefined maximum cost per m2 of a roof light; RLCnew is the unit cost of a selected roof light. The total estimated cost is adjusted according to the tender price index (TPI) of a new area for the same or different time as follows:

adjusted cost ¼ estimated cost 

new TPI  new location index base TPI  base location index

ð16Þ

If the new TPI and new location index are 333 and .72 respectively and base TPI and base location index are 1 respectively, and ‘estimated cost’ is RSMC, Eq. (16) can be calculated as follows:

adjusted cost ¼ 412; 183:5 

333  :72 ¼ £170; 388 580  1

Fig. 6 illustrates the cost estimating mechanism. The system retrieves the approximate cost of a roof specification and roof lights cost from roof specification cost database table and predefined roof lights cost database table. The system calculates m2 cost for roof specification and roof lights by using Eq. (14) and (15) respectively. The system calculates total roof specification cost by multiplying new cost per m2 of roof specification by the roof area. The system accepts input of roof lights area and roof features quantity from the user and then calculates new roof lights cost by multiplying new cost per m2 of roof lights and roof area and new roof features cost by multiplying unit cost of roof features and roof features quantity. The system retrieves the tender price index and area index from the cost index database table and then adjusts new total roof specification cost, roof lights cost and roof features cost. Finally the system calculates the total roof cost by using Eq. (10) and then pro-

6864

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

Provides ranked materials to the user

User chooses a material of each sub element

Accepts selected materials

Predefined Roof Lights Cost Database Table

Predefined Roof Lights Cost

Retrieves approximate cost of roof specification and roof lights

Roof Specification Cost

Roof Specification Cost Database table

Calculates m2 cost of new roof specification and roof lights

User provides roof lights area and roof features quantity

Calculates total roof specification cost, roof lights cost and roof features cost

Roof area

Adjusts total roof specification cost, roof lights cost and roof features cost

Tender price index and area index

Cost Index Database Table

Adds cost of roof specification , roof lights and roof features

Provides materials selected and its cost to the user Fig. 6. Cost estimating mechanism of the KDSMS system.

vides the selected materials and the associated total cost to the user.Fig. 6 illustrates the cost estimating mechanism. The system retrieves the approximate cost of a roof specification and roof lights cost from roof specification cost database table and predefined roof lights cost database table. The system calculates m2 cost for roof specification and roof lights by using Eqs. (14) and (15) respectively. The system calculates total roof specification cost by multiplying new cost per m2 of roof specification by the roof area. The system accepts input of roof lights area and roof features quantity from the user and then calculates new roof lights cost by multiplying new cost per m2 of roof lights and roof area and new roof features cost by multiplying unit cost of roof features and roof features quantity. The system retrieves the tender price index and area index from the cost index database table and then adjusts new total roof specification cost, roof lights cost and roof features cost. Finally the system calculates the total roof cost by using Eq. (10) and then provides the selected materials and the associated total cost to the user.

3. KDSMS model This section explains the architecture of the KDSMS system (Fig. 4). It contains four components: (1) database management system of the roof element database with user interface facility; (2) knowledge base with the integration of material selection and cost estimating rules; (3) inference engine with user interface facility; and (4) user interface with input and output facilities. This system is currently operated in stand alone computer, but it is capable of being operated in web based environment. The components are explained in the following section (Fig. 7). 3.1. Roof element database The properties of materials of roof structure, roof coverings, roof insulation, roof drainage, roof lights and roof features related to the respective selection criteria (see Table 1) are stored in the corresponding data tables in the database. The cost index, roof insula-

6865

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871 Table 3 Sustainability rating and eco points.

User User Interface (Oracle Form 10g)

Inference Engine (SQL statements, TOPSIS) 1

Knowledge Base - Materials selection - Cost estimating - Cost adjustment

Roof element Database (Oracle Database 10g)

Fig. 7. KDSMS architecture.

tion regulations, roof information such as span and angle of pitch are stored in the database table. The structure of the database for KDSMS system is illustrated in Fig. 8. The user can enter, update, delete and view the data in the tables of database management system directly through the user interface. Data entry, delete, update or view functions are available for the database tables of thermal regulations, tender price and location index, roof project, roofing materials and approximate per square metre cost. Individual roof sub element has its own data table and values of materials properties are maintained through the user interface. The users can enter materials with the values of the properties, update the values, view or delete the materials through the user interface. During the data entry, the values of sustainability attribute are converted from BREEAM rating into numeric values and stored into the database. Table 3 shows the BREEAM rating for sustainability and the numeric values are converted for other roofing materials and insulation materials separately based on the eco points (Anderson et al., 2009).

3.2. Knowledge base Knowledge base was created using relevant information derived from literature review and interviews with the domain experts. It contains the facility for implementing the TOPSIS method for material selection, building regulations for roof, cost estimating and cost adjustment processes. The knowledge base

Rating1

Eco points for other roofing materials1

Eco points for insulation materials1

A+ A B C D E

0.51 0.70 1.05 1.30 1.62 1.87

0.04 0.08 0.12 0.16 0.20 0.25

(Anderson et al., 2009)

uses the IF-THEN syntax where the IF part represents condition and THEN part represents action. The IF-THEN rules were used to represent knowledge in the knowledge base. The rules were categorised into two types- selection rule and estimating rule where selection rule was used for materials selection purpose and estimating rule was used for cost estimating and cost adjustment purposes. An example of an estimating rule in the knowledge base for cost adjustment is shown is Table 4. SQL is a standard language to manipulate database. Thus, SELECT SQL statement is directly included in the rules. The advantage is that the SQL statement rules not only query the database but also can insert, update and delete records within the database. An example of SQL statement of a selection rule for insulation material selection is shown in Table 5. The DECODE is an Oracle SQL function which compares an expression to each search value one by one. If expression is equal to a search value, then Oracle Database returns the corresponding result. If no match is found, then Oracle returns default value. In the above example of Table 1, if insulation_level is rafter, it selects the insulation materials which u_value is smaller than or equal to 0.20; if insulation_level is ceiling, it selects the insulation materials which u_value is smaller than or equal to 0.16; and if insulation_level is not defined, it selects the insulation materials which u_value is smaller than or equal to 0.20. The rule in Table 5 is very

Table 4 Rule for cost adjustment. IF new_location_index=:v_new_location and new_tender_price_index=:v_new_tender_price_index THEN adjusted_ total_roof_cost:¼:v_total_roof_cost x ((:v_new_location_index: v_new_tender_price_index)580);

KDSMS System

Thermal Regulation for Roof Insulation and Roof Lights

Tender Price & Location Index

Data Entry

Data Update

Data Delete

Data View

´Roof Project

Roofing Materials

Approximate Cost / m2

Materials Selected & Cost Model

Roof Structure, Coverings, Insulation and Drainage Cost

Roof Lights Cost

Roof Structure

Roof Coverings

Roof Insulation

Fig. 8. Database structure of KDSMS.

Roof Drainage

Roof Lights

Roof Features

6866

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

Table 5 Rule for insulation material selection.

Table 6 Inference engine for roof coverings materials selection.

SELECT material_name FROM roof_insulation_material WHERE u_value <= DECODE (insulation_level,’rafter’,0.20,’ceiling’,0.16,0.20);

SELECT material_name, score, DENSE_RANK () OVER (ORDER BY score DESCENDING) Rank, lifespan, initial_cost, coverings_size from TOPSIS_COVERINGS_VIEW

difficult to represent by IF-THEN syntax, thus SQL statement is used to represent the rules. TOPSIS calculations are also implemented by SQL statement as rules which are the part of selection rule. 3.3. Inference engine The inference engine performs the reasoning process. By adopting the forward chaining mechanism, it provides the solution by linking the rules in the knowledge base with the facts provided in the database. The SQL statements are implemented in stored procedures and input screens; and the stored procedures are linked with input and output screens. The input screens accept the input from the user; then the related inference engine of each

roof sub element searches the data from the database and match with the rules. The inference engine is mainly used to rank materials by evaluating them and to estimate and adjust cost. An example of the inference engine is illustrated in Table 6. TOPSIS is a part of inference engine, which performs the material selection task as a decision making technique. The inference engine uses two special Oracle SQL functions, DENSE_RANK () and OVER (), to rank the materials based on the score in descending order calculated by TOPSIS where TOPSIS calculations are implemented by SQL statements in TOPSIS_COVERINGS_VIEW, which is an Oracle Database View. It produces the result by matching the values of roof specification parameters, rules in the knowledge base and data of material properties in the tables of the data base and the result is shown through a user interface.

Fig. 9. Data input screen for roof coverings materials.

Fig. 10. Data entry screen of create roof project.

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

6867

3.4. User interface

4. Illustrative example

The user interface is the medium through which the user interacts with the processes. It accepts input from users, supplies it to the inference engine activates the processes to produce the output to the users. Menu-driven user interface is developed for this system where it uses pull-down and cascading menu options. It is JAVA applet based form which is developed by using Oracle Forms 10 g. It is linked with all input and output screens of the KDSMS system. The system has different types of input and output screens for different purposes. An example of data entry, update, delete and view functions for roof coverings materials is shown in Fig. 9, where the users can interact with the database through this interface. It is noted that the importance level of criteria is depended on the requirements of users, hence the users can input any weights attached with the criteria entry screen, When the users input the criteria weights as linguistic values, the system converts these values into numeric values.

The selection of optimal materials for roof structure, roof coverings, roof insulation, roof drainage, roof lights and roof features; and the cost estimate of these selected materials are illustrated here. In the beginning of the material selection, the system asks the user to provide the roof information, as shown in Fig. 10. After selecting the Select Material option, the system redirects the user to an option to weight criteria for different roof sub elements in separate tab page. When a tab option of a roof sub element is selected, the system prompts the user to weight criteria for that sub element (Fig. 11). Since criteria weights vary from user to user or project to project, the system prompts the user for the weights attached to the criteria in such a way that these can then be changed by the user as required. Linguistic variables are used for weighting criteria and user can choose importance level by using radio buttons. The user can choose all sub elements or individual sub element and if the user does not wish to consider any of the roof sub element, it

Fig. 11. Interface for roof structure selection criteria.

Fig. 12. Interface of selected material for roof structure.

6868

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

can be ignored. Fig. 11 shows an example of criteria weighting of roof structure sub element. The system executes its knowledge base after weighting the criteria of different sub elements and selecting Select material option. After retrieving roof structure materials according to spans and angle of pitch, roof coverings materials according to angle of pitch,

roof insulation materials according to thermal regulations and other roof sub elements materials for pitched roof, the system executes TOPSIS method to rank materials for different sub elements by using corresponding criteria weights and the values of materials properties retrieved from database. Then the system prompts the recommendations of optimal materials and ranking, and then asks

Fig. 13. Interface of selected material for roof coverings.

Fig. 14. Interface of selected material for roof insulation.

Fig. 15. Interface of selected material for roof drainage.

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

the user to choose a specific material from the list of individual sub element. The user can use arrow key button or mouse to choose a material and when any material of a sub element is highlighted by mouse or arrow key button, it is stored into different box in the lower place of the screen. Figs. 12–16 show the lists of optimal materials and ranking along with the values of materials properties of roof structure, roof coverings, roof insulation, roof drainage and roof lights respectively where the system facilitates the user to choose the desired material of each roof sub element. After choosing desired optimal material of each sub element, the system estimates the cost based on these optimal materials selected if the Estimate Cost option is selected. The system then estimates the cost of these optimal materials and adjusts the cost according to quarter 1 of year 2009 for Northern Ireland and prompts the result to the user, as shown in Fig. 17; where the system provides optimal materials selected and associated total cost. The system also provides the facility for the user to update the required roof lights area and when the roof area is provided, the total cost is changed.

6869

5. Conclusion This paper presents the KDSMS system, a Knowledge-based Decision Support system for the selection of optimal Materials for building design. The system uses product cost modelling techniques and the MCDM technique of TOPSIS for optimal materials selection and has been implemented as a prototype system for optimal roofing material selection and cost modelling. The KDSMS system uses an architecture that integrates the knowledge base with the Oracle database system. This system resolves MCDM problem by identifying a multitude of criteria involved in roofing materials selection and evaluating them in the selection procedure. The system enables new material information to be added and the database can be updated easily. It also enables the updating of prices for both time factors and location using the tender price and location indices obtained from the BCIS. It also provides sustainable rating for materials so that it can facilitate the effective selection of sustainable and innovative building materials thereby facilitating the reduction of carbon footprint. In addition, the sys-

Fig. 16. Interface of selected material for roof lights.

Fig. 17. Roofing materials selected and cost model.

6870

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871

tem provides an approximate cost estimate for roof and its sub elements based on the optimal materials selected. Optimal materials are always preferred not only for environmental reasons but also for cost effectiveness and ease of maintenance. More sustainable materials contribute to the sustainable construction and help the environment by reducing the carbon footprint. This requires the need to simultaneously consider a multitude of criteria in selecting optimal materials with higher sustainability. Moreover, new innovative materials are frequently introduced to the market, but may not be used due to lack of information and experience of the designers or clients (e.g. self-build). Therefore, building designers or inexperienced users such as self builders are often faced with the problem of information overload and pressures on innovative and sustainable design. This system employs the use of a knowledge-based system to overcome this problem. The materials along with the values of properties are stored in a database; which allows quick and efficient retrieval of appropriate product details and the values of properties to evaluate performance when required. Several systems have attempted to solve this problem but none have successfully utilised the multi criteria decision making (MCDM) techniques in roofing materials selection within the housing design domain. This system fills this gap and proposes a knowledge-based model as the decision making support tool to provide optimal product/material selection and estimate of approximate cost from the early conceptual stage of design. The developed system is not without its limitations. It is limited to pitched roofs of housing projects and their related materials. As such, it did not include flat roofs and associated materials for flat roof construction. In addition, the study did not include qualitative and subjective criteria such as weather resistance, sound resistance, strength and stability, fire resistance and security, and associated building regulations, which may influence material selection. The reason for this exclusion in this present research is to avoid the use of subjective criteria in the present prototype that was developed. It is also noted that a building may have several roofs with different angle of pitch and roof spans. However, there is a limitation in this study in that the KDSMS system can handle a single roof with a roof span and angle of pitch and this limitation hinders the selection of materials and estimating of cost for multiple roofs of a building. However, this system has several advantages. Firstly, with the aid of this decision support system, the architects, quantity surveyors and home owners are made more aware in a user friendly manner about the multitudes of selection criteria to be considered in the selection of roofing materials. Secondly, the knowledge of material selection method through the use of multiple criteria decision making would assist the Cost engineers, Quantity Surveyors and Architects to evaluate and select optimal materials from the vast array of possibilities to meet specific requirements. Thirdly, the cost modelling facility offered by the system would assist the Quantity Surveyors and Architects to estimate roofing cost with new materials from an early stage of building design. This can save enormous time and a significant cost of roof construction. Moreover, they are able to overcome the information overload which might prevent them from the selection of suitable materials. Finally, the system enables the evaluation and choice of optimal materials for the construction of sustainable and energy-efficient buildings which can contribute to the reduction of carbon foot print. The optimal materials will require less energy, thus the end-users can save a significant energy cost. The approach adopted in this research is generic to all other building elements; as such further research needs to be carried out to cover flat roof and other elements of a building in order to facilitate the effective use of innovative and sustainable building materials and technologies. This research can be effectively ex-

panded to other building types such as educational buildings, retail, industrial buildings, health, commercial, hospital and sport centre. It is suggested that further research would be necessary to consider the inclusion of qualitative and other subjective criteria and associated building regulations which influence material selection. In addition, further research also needs to be conducted to estimate the cost for multi roofs of a building, and to obtain relevant data on life cycle assessment for a roof such as energy cost, cleaning cost among other running cost to evaluate materials more efficiently. The research can further be expanded and implemented worldwide with modifications to account for specific regional location factors and regulations. Moreover, the current system and database can be expanded and implemented as a commercial package and transformed to fully web enabled system. Acknowledgements This project was supported by the Built Environment Research Institute, University of Ulster, Northern Ireland, as part of a Ph.D. programme of research. We would like to acknowledge the support of the Director of the Research Institute, Professor Stanley McGreal for proof reading some part of the work and also for providing industrial contacts to evaluate the developed system. We would also like to acknowledge the inputs of those who participated in the expert forum. References Anderson, J., Shiers, D., & Steele, K. (2009). The green guide to specification: An environmental profiling system for building materials and components. (4th ed.). UK: Building Research Establishment. Ashby, M. F., Brechet, Y. J. M., Cebon, D., & Salvo, L. (2004). Selection strategies for materials and process. Materials and Design, 25(1), 51–67. Bullinger, H. J., Warschat, J., & Fischer, D. (1991). Knowledge-based system for material selection for design with new materials. Knowledge-Based Systems, 4(2), 95–102. Chan, J. W. K., & Tong, T. K. L. (2007). Multi-criteria material selections and end-oflife product strategy: Grey relational analysis approach. Materials and Design, 28(5), 1539–1546. Chen, J. L., Sun, S. H., & Hwang, W. C. (1995). An intelligent data base system for composite material selection in structural design. Engineering Fracture Mechanics, 50(5/6), 935–946. Marley Eternit. Roofing. (2010). http://www.marleyeternit.co.uk/ Accessed 17.05.2010. Hwang, C. L., & Yoon, K. (1981). Multiple attribute decision making- methods and applications: A state-of-the-art survey. Berlin: Springer-Verlag. Hymers, P. (2006). Converting to an Eco-friendly Home: The Complete Handbook. London: New Holland. Karolides, A. (2006). Green building approaches. In RSMeans (Eds.), Green Building (2nd ed.). USA: Reed Construction Data Inc. (pp. 1–25). Kingspan. (2010). Products. http://www.insulation.kingspan.com/uk/literature.htm Accessed 17.05.2010. Knauf (2010). Roof Insulation. http://www.knaufinsulation.co.uk/Default.aspx? page=2006 Accessed 17.05.2010. Krope, J., & Goricanec, D. (2009). Energy efficiency and thermal envelope. In D. Mumovic & M. Santamouris (Eds.), A handbook of sustainable building design and engineering. London: Earthscan. Langdon, D. (2010). Spon’s architects’ and builders’ price book (135th ed.). UK: Spon Press. LBS. (2010). Roofing. [online] Available from http://www.lbsproducts.com/ Accessed 17.05.2010. Mahmoud, M. A. A., Aref, M., & Al-Hammad (1996). An expert system for evaluation and selection of floor finishing materials. Expert Systems with Applications, 10(2), 281–303. Mohamed, A., & Celik, T. (1998). An integrated knowledge-based system for alternative design and materials selection and cost estimating. Expert Systems with Applications, 14(3), 329–339. Northstone. (2010). Northstone Materials. http://www.northstone-ni.com/aboutus/products-and-services/ Accessed 17.05.2010. Papadopoulos, A. M., & Giama, E. (2007). Environmental performance evaluation of thermal insulation materials and its impact on the building. Building and Environment, 42(5), 2178–2187. Part F. (2008). Conservation of fuel and power. Building Regulations (Northern Ireland) Amendments to Technical Booklets F1 (2006) and F2 (2006). Perera, R. S., & Fernando, U. L. A. S. B. (2002). Cost modelling for roofing material selection. Built Environment: Srilanka, 3(1), 11–24.

S. Rahman et al. / Expert Systems with Applications 39 (2012) 6857–6871 Rahman S., Perera, S., Odeyinka, H., & Bi, Y. (2008). A conceptual knowledge-based cost model for optimizing the selection of material and technology for building design. In 24th Annual ARCOM conference proceedings (pp. 217–225). Rahman S., Perera, S., Odeyinka, H., & Bi, Y. (2009a). A knowledge-based decision support system for roofing materials selection and cost estimating: A conceptual framework and data modelling. In 25th Annual ARCOM conference proceedings (pp. 1081–1190). Rahman S., Perera, S., Odeyinka, H., & Bi, Y. (2009b). A knowledge-based decision support system for roofing materials selection. In RICS COBRA research conference proceedings (pp. 1753–1762). Redland. (2010). Product catalogue. http://www.monier.co.uk/products/productcatalogue.html Accessed 17.05.2010]. Roaf, S., Fuentes, M., & Thomas, S. (2007). Ecohouse: A design guide. Oxford: Architectural Press. Shanian, A., & Savadogo, O. (2006). TOPSIS multiple-criteria decision support analysis for material selection of metallic bipolar plates for polymer electrolyte fuel cell. Journal of Power Sources, 159(2), 1095–1104. Soronis, G. (1992). An approach to the selection of roofing materials for durability. Construction and Building Materials, 6(1), 9–14.

6871

UNEP. (2001). Energy and Cities: Sustainable Building and Construction. IETC Side Event at United Nations Environment Programme Governing Council, 6 February 2001, Nairobi, Kenya. http://www.unep.or.jp/ietc/focus/Energy Cities1.asp Accessed 17.05.2010. Velux. (2010). Roofs windows. http://www.velux.co.uk/Private/Products/ velux_Roof_Windows/default.aspx Accessed 17.05.2010. Wilson, A., Uncapher, J. L., McManigal, L., Lovins, L. H., Cureton, M., & Browning, W. D. (1998). Green development: Integrating ecology and real estate. New York: John Wiley & Sons Inc.. Wong, J. K. W., & Li, H. (2008). Application of the analytic hierarchy process (AHP) in multi-criteria analysis of the selection of intelligent building systems. Building and Environment, 43(1), 108–125. Woolley, T. (2006). Natural building: A guide to materials and techniques. Marlborough: The Crowood Press. Yong, D. (2006). Plant location selection based on fuzzy TOPSIS. Advanced Manufacturing Technology, 28(78), 839–844. Zhou, Q., Huang, G. H., & Chan, C. W. (2004). Development of an intelligent decision support system for air pollution control at coal-fired power plants. Expert Systems with Applications, 26(3), 335–356.