Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Application of Neuro-Fuzzy System to Evaluate Sustainability in Highway Design

658 views

Published on

http://www.ijmer.com/papers/Vol2_Issue6/BG2641534158.pdf

  • Be the first to comment

  • Be the first to like this

Application of Neuro-Fuzzy System to Evaluate Sustainability in Highway Design

  1. 1. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645 Application of Neuro-Fuzzy System to Evaluate Sustainability in Highway Design Thoedtida Thipparat1, Thongpoon Thaseepetch2 1, 2 (Facility of engineering and architecture, Rajamangala University of Technology, Bangkok, ThailandABSTRACT: In this paper, adaptive neuro-fuzzy inference Considering sustainability early in planning and design willsystems (ANFIS), using Sugeno ANFIS is used to evaluate result in high potential reductions in operations’ sustainablethe construction sustainability of highway design projects in impacts could be generated [8-10].Thailand. Input variables include factors associated with As a result, the planning and design stage is ageometrics and alignments, earthworks, pavement, critical stage for the study of ecology, the systematicdrainage, retaining walls, slope protection, landscape and analysis of resource and energy flows within theecology, transportation facilities, transportation anthroposphere, and the realm of man-made or managedmaintenance, bridges, sound insulation, tunnels, electrical resources. The wide variety of materials used inand mechanical work, and lighting. Finally, on the basis of construction work, as well as fuel and electricity used forour analysis, our results hold an important lesson for construction machinery and recycling plants has adecision makers who may clearly picture the rules and use significant environmental impact. The risk score is anthe fuzzy neural network approach to promote effective overall aggregation of construction risks that are usuallyenvironmental protection to meet the highway district’s assessed against different criteria such as safety,goals. functionality, sustainability and environment and characterized by risk ratings such as High, Medium, Low orKeywords: Sustainability, Highway design, Neuro-fuzzy None. The aggregation process involves a large number ofsystem, Uncertainty, Evaluation subjective judgments of construction experts, but there is no explicit functional relationship between risk score and risk I. INTRODUCTION ratings. Public awareness of global environmental The AASHTO (American Association of Stateproblems such as global warming and ozone depletion has Highway and Transportation Officials) conducted aincreased, to become one of the biggest issues for the new compendium of practices, procedures, and policies formillennium, since the concept of sustainable development integrating environmental stewardship into highwayhas attracted public attention as the greatest concern of the construction and maintenance activities for Departments oflate twentieth century. Sustainable development in the Transportation (DOT) in many states [11]. It is one of theconstruction industry is now an important issue [1]. standards and improvements in environmental processes,However, it is lagging behind other sectors because, by its practices, and significant environmental items.very nature, construction involves manipulation and use of Currently, several studies have been done onlarge quantities of natural and man-made materials. Also, development of the forecasting models. However, a modelthe construction and operations of infrastructure are large used to predict the sustainability of highway design is rarelyusers of energy. A construction project generates a large concerned. The research focus is to investigate the factorsamount of pollutants, including noise, air, solid waste and affecting sustainability in highway design in Thailand.water [2-3]. There are suggestions that the environmental The ability of hybrid ANN models in predictingeffects of the construction project can be studied under the country risk rating was determined [12]. The hybrid ANNfollowing headings: (i) energy consumption; (ii) dust and models were compared with traditional statistical techniquesgas emission; (iii) noise distribution; (iv) waste generation; such as discriminate analysis (DA), logit model, probit(v) water discharge; (vi) use of water resources; (vii) model and ordinary neural networks. Their results indicatedunnecessary building consumption; (viii) pollution by that hybrid neural networks outperformed all the otherbuilding materials; (ix) land use; and (x) use of natural models. In order to determine the constructionresources (Ball, 2002).Environmental impacts of buildings sustainability, mathematical models need to be developed toover their entire life cycle process have been recognized as predict the risk scores of construction project. Adaptivea serious problem [4]. Nevertheless, the construction neuro-fuzzy inference systems (ANFIS), using Sugenoindustry has not been showing very much concern about the ANFIS and logistic regression analysis are two alternativeenvironmental issues [5]. approaches for modelling construction sustainability data. The sustainable concepts can be effectively This study compares the modelling mechanisms andintegrated into construction project in the design stage [6]. performances in modelling construction sustainability data.One of the most importance stage of the life cycle in most This study firstly identified the performancegreen building guidelines and evaluation methods is the measurement indicator for construction sustainabilitydesign stage [7]. It is important to focus on planning and assessment associated with a construction project. Then thedesign services provided by engineering consultants in the impact of regulatory compliance, auditing activities andinitial stages of the life cycle of infrastructure. The resources consumption on construction sustainability weresustainable planning and design lead to the good investigated. The study framework is a simple linear modelmanagement of the energy and materials required of the relationship between the independent and dependentconstructing highway, and waste produced by construction. variables. Sustainability items as the independent variables www.ijmer.com 4153 | Page
  2. 2. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645consists of 14 groups, namely geometrics and alignments, 5.2 Vegetationearthworks, pavement, drainage, retaining walls, slopeprotection, landscape and ecology, transportation facilities, 5.3 Grinding stones or softtransportation maintenance, bridges, sound insulation, reinforcingtunnels, electrical and mechanical work, and lighting. Thereare 60 input variables. . These input variables used in the 6.1 Vegetationmeasurement of the sustainability are proposed by [13]. 6. SlopeTable 1 shows sustainable items for highway design. protection 6.2 Reinforced slopes 6.3 Waste reuseTable 1 Sustainable items for highway design (Tsai andChang, 2012) 7. Landscape and 7.1 Avoidance of natural Activities Sustainable items preservation sites ecology 1. Geometrics and 1.1 Reduction in volume or 7.2 Embankments or cutting alignments weight replaced by bridges or tunnels 1.2 Mild curves 7.3 Native trees 1.3 Mild slopes 7.4 Tree transplanting 2. Earthworks 2.1 Earthwork balance 7.5 Vegetation 2.2 Minimum excavation and 7.6 Topsoil recycling fills 7.7 Culverts for wildlife 2.3 Topsoil recycling crossings 7.8 Ecological ponds 2.4 Waste reuse 7.9 Habitat connectivity 3. Pavement 3.1 Reduction in volume or weight 7.10 Biological porous environment 3.2 Permeable materials 7.11 Reduction in 3.3 Recycled materials landscaping facilities 3.4 Noise reduction materials 7.12 High bridges 3.5 Fiber materials Table 1 Sustainable items for highway design (Tsai and Chang, 2012) (con’t)Table 1 Sustainable items for highway design (Tsai and Activities Sustainable itemsChang, 2012) (con’t) 8.1 Reduction in volume or Activities Sustainable items weight 8. Transportation 4.1 Runoff reduction facilities 8.2 Multi-function poles 4. Drainage 4.2 Vegetated or gravel 9. Transportation ditches maintenance 9.1 Reduction in path changes 4.3 Rainwater catchments 10. Bridges 10.1 Reduction in volume or 4.4 Infiltration trenches or weight catch basins 10.2 Long-span bridges 4.5 Sediment ponds 10.3 Pre-casting techniques 4.6 Regional materials 10.4 Temporary bridges for 5.1 Reduction in volume or construction 5. Retaining walls weight 10.5 Hollow railings www.ijmer.com 4154 | Page
  3. 3. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645 10.6 Reinforced materials The ANFIS approach adopts Gaussian functions (or other membership functions) for fuzzy sets, linear 10.7 High strength concrete functions for the rule outputs, and Sugeno’s inference mechanism [14]. The parameters of the network are the 10.8 Self-compacting mean and standard deviation of the membership functions concrete (antecedent parameters) and the coefficients of the output linear functions as well (consequent parameters). The 10.9 Lightweight concrete ANFIS learning algorithm is then used to obtain these parameters. This learning algorithm is a hybrid algorithm 10.10 Steel consisting of the gradient descent and the least-squares 11. Sound 11.1 Reduction in volume or estimate. Using this hybrid algorithm, the rule parameters insulation weight are recursively updated until an acceptable level of error is reached. Each iteration includes two passes, forward and 11.2 Landscaping backward. In the forward pass, the antecedent parameters are fixed and the consequent parameters are obtained using 12.1 Reduction in volume or the linear least-squares estimation. In the backward pass, the weight consequent parameters are fixed and the error signals 12. Tunnels propagate backward as well as the antecedent parameters 12.2 Vegetation are updated by the gradient descent method. Based on the original ANFIS study [15]; the learning mechanisms should 12.3 Reduction in ventilation not be applied to determine membership functions in the facilities Sugeno ANFIS, since they convey linguistic and subjective descriptions of possibly ill-defined concepts. Hence, the 12.4 Waste reuse choice of membership function should depend on the specific types of data. 12.5 Fiber materials 13.1 Reduction in III. TRAINING ANFIS FOR 13. Electrical and transportation controlling SUSTAINABLILITY EVALUATION mechanical work facilities For proving the applicability of the model and illustration, the proposed model was applied in a highway 14. Lighting 14.1 Reduction in lighting district in Thailand. The first step to apply the model was to equipment construct the decision team. The decision team included head and staffs of highway district, construction consultants, 14.2 Renewable energy contractors with capital of more than hundred million baht, contractors with capital of less than hundred million baht, and sub-contractors. For training the ANFIS, a 14.3 Shading board questionnaire was designed including 14 activities (60 variables). Fig 2 presents an input-output model for earthworks. The decision team was asked to give a score to II. NEURO-FUZZY MODEL them, based on their knowledge about the highway projects The neuro-fuzzy system attempts to model the in a case study highway district. Then, they rated theuncertainty in the factor assessments, accounting for their highway projects with respect to each criterion in the rangequalitative nature. A combination of classic stochastic of [0, 10] with triangular or trapezoidal membershipsimulations and fuzzy logic operations on the ANN inputs functions describing the linguistic variables such as theas a supplement to artificial neural network is employed. impact of reduction in volume or weight on sustainability,Artificial Neural Networks (ANN) has the capability of self- for example: "low", "medium" or "high" (Fig 3). The systemlearning, while fuzzy logic inference system (FLIS) is was built in the Matlab Fuzzy Toolbox and Simulinkcapable of dealing with fuzzy language information and environment. A Matlab programme was generated andsimulating judgment and decision making of the human compiled. The pre-processed input/output matrix whichbrain. It is currently the research focus to combine ANN contained all the necessary representative features, was usedwith FLIS to produce fuzzy network system. ANFIS is an to train the fuzzy inference system. Fig 4 show membershipexample of such a readily available system, which uses functions describing the linguistic variables such as theANN to accomplish fuzzification, fuzzy inference and operation performance, for example: "low", "medium" ordefuzzification of a fuzzy system. ANFIS utilizes ANN’s "high" after training the network. Fig 5 shows the structurelearning mechanisms to draw rules from input and output of the ANFIS; a Sugeno fuzzy inference system was used indata pairs. The system possesses not only the function of this investigation. From 60 attributes generated by theadaptive learning but also the function of fuzzy information proposed approach, the decision makers derived 60describing and processing, and judgment and decision completed questions. We used 50 of them for training themaking. ANFIS is different from ANN in that ANN uses the ANFIS and the rest (15) for checking and validation of theconnection weights to describe a system while ANFIS uses model. For rule generation, we used subtractive clusteringfuzzy language rules from fuzzy inference to describe a where the range of influence, squash factor, acceptancesystem. ratio, and rejection ratio were set at 0.5, 1.25, 0.5 and 0.15, www.ijmer.com 4155 | Page
  4. 4. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645respectively during the process of subtractive clustering.Rules (clusters) of the trained FIS for earthworks arepresented in Fig 6. Because by using subtractive clustering,input space was categorized into 18 clusters. Each input hasGaussian curve built-in membership functions. Fig 7 showsthe surface of ANFIS after training. The training error of theANFIS was ranged from 0.32 to 0.75 after 50 epochs aspresent in Fig 8. The rate of highway projects with respectto criteria and the output of ANFIS (sustainability level ofhighway design) have been shown in Table 2. Figure 5 Network of sustainability evaluation by the ANFIS Figure 2 Input-output model for earthworks Figure 6 Rules (clusters) of the trained FIS Figure 7 Trained main ANFIS surface: construction sustainability Figure 3 Membership functions of the operation performance (before Training) Figure 8 Training error of the ANFIS Figure 4 Membership functions of the operation performance (after Training) www.ijmer.com 4156 | Page
  5. 5. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645Table 2. Inputs (rates of a case study) and output of ANFIS IV. CONCLUSION A major objective of this paper was to present an application of ANFIS, and its performances in modeling a set of construction sustainability data. Considering data obtained from experts based on eighty scenarios, the sustainability level is calculated. Scenarios No. 1 to 50 were used as training cases for both methods. Scenarios No. 51 to 80 were used as test cases to compare performance of the Logistic regression and ANFIS methods. The five scenarios were used to compare the performance of the proposed approach by comparing with the values given by experts in scenarios being tested. The average and standard deviation of the differences between the estimated and the possibilityTable 2. Inputs (rates of a case study) and output of ANFIS obtained from expert produced by ANFIS are calculated to(con’t) be 9.91 and 7.89 respectively. The performance of the models developed by ANFIS is summarized in Table 2. Since it is an important lesson for decision makers aiming to promote effective environmental protection to meet the highway district’s goals, the findings present a model of how sustainability for highway design was predicted regarding factors associated with 14 activities. Results present that the ANFIS model enhances sustainability forecasting by using fuzzy rules, which are easy for communication. The ANFIS model can explain the training procedure of outcome and how to simulate the rules for prediction. www.ijmer.com 4157 | Page
  6. 6. International Journal of Modern Engineering Research (IJMER) www.ijmer.com Vol. 2, Issue. 5, Sep.-Oct. 2012 pp-4153-4158 ISSN: 2249-6645 REFERENCES [9] C. Cerdan, C. Gazulla, M. Raugei, E. Martinez, P.[1] I. Holton, J. Glass, A.D.F. Price, Managing for Fullana-i-Palmer, Proposal for new quantitative eco- sustainability: findings from four company case design indicators: a first case study, Journal of studies in the UK precast concrete industry, Journal of Cleaner Production 17, 2009,1638-1643. Cleaner Production, 18, 2010 152-160. [10] J.H. Spangenberg, A. Fuad-Luke, K. Blincoe, Design[2] B. Polster, B. Peuportier, I.B. Sommereux, P.D. for sustainability (Dfs): the interface of sustainable Pedregal, C. Gobin, Durand, E.: Evaluation of the production and consumption. Journal of Cleaner Environmental Quality of Buildings towards A More Production 18, 2010, 1485-1493 Environmentally Conscious Design. Solar Energy. 57, [11] Transportation Research Board (TRB), Environmental 1996, 219–30. stewardship practices, procedures, and policies for[3] Y.C.Tse and C. Raymond, The Implementation of high way construction and maintenance. NCHRP EMS in Construction Firms: Case Study in Hong Report 25e25 (04), (National Cooperative Highway Kong. Journal of Environmental Assessment Policy Research Program, prepared by Venner Consulting and Management, 3, 2001, 177–94 and Parsons Brinckerhoff, 2004).[4] R. Morledge, and F. Jackson, Reducing Environmental [12] J. Yim, and H. Mitchell, Comparison of country risk Pollution Caused by Construction Plant. Environ models: hybrid neural networks, logit models, mental Management and Health.12, 2001, 191–206 discriminant analysis and cluster techniques. Expert[5] R.B. Clements, Complete guide to ISO 14000 Systems with Applications. 28(1), 2005, 137–148. (Prentice Hall: Englewood Cliffs, NJ, 1996). [13] C.Y. Tsai, A.S. Chang, Framework for developing[6] National Research Council (NRC) Improving construction sustainability items: the example of Engineering Design: Designing for Competitive highway design. Journal of Cleaner Production, 20, Advantage. National Academy Press (Washington, 2012, 127-136 DC, 1999) [14] T. Takagi, and M. Sugeno, Fuzzy Identification of[7] K.R. Bunz, G.P. Henze, D.K.Tiller, Survey of Systems and its Application to Modeling and Control. sustainable building design practices in north IEEE Transactions on Systems Man and Cybernetics, America, Europe, and Asia, Journal of Architectural 5, 1985, 116–132 Engineering, ASCE 12 (1), 2006, 33-62. [15] J.S.R. Jang, ANFIS: Adaptive- Network based Fuzzy[8] B.C. McLellan, G.D. Corder, D. Giurco, S. Green, Inference System, IEEE Trans. IEEE Transactions on Incorporating sustainable development in the design of Systems Man and Cybernetics, 23, 1993, 665-685. mineral processing operations review and analysis of current approaches, Journal of Cleaner Production 17, 2009, 1414-1425 www.ijmer.com 4158 | Page

×