37Copyright © 2010 Japan Society of Kansei Engineering.

All Rights Reserved.

Kansei Engineering International Journal Vol.10 No.1 pp.37-47 (2010)

Received 2009.10.05Accepted 2010.06.30

1. INTRODUCTION

During and after the period of rapid economic growth of Japan, much focus has been placed on quantity than on quality for bridge structure development. Later on, however, people began to value the quality of life or comfort, and it is almost 20 years since the society began to acknowledge landscape design as important, which was backed by the rising attention on bridge landscape since the late 1980’s. During that time, while Japan experienced the rapid social and economical changes and aging popu-lation, recognition of landscape design was significantly changed. Public works adopting new styles have been increasing in which involvement of local residents is encouraged, a shift from the conventional public works planned by government and municipal offices. Some local governments started to launch public works, in particular, civil structure construction project where local residents have participation even from the planning stage.

As mentioned above, based on the reflection on the past public works where functionality and economy were considered as the most important, civil structure works that are oriented toward stabilizing public involvement have been desired while respecting local characteristics and residents’ aesthetic feelings. For citizen-involved development, opportunities such as public discussion meetings and workshops are often provided to gather opinions of people concerned. Construction plans presented at such public meetings are usually created by government officials or private designers, but in reality they do not always reflect diverse residents’ needs and their changing sense of value. If discussions between concerned parties or consensus building processes fail, the development processes will be affected. In future

bridge plans and design, much consideration has to be given to increasing awareness of citizen involvement in civil structure works, diverse sense of value, full reflection of users’ needs (Kansei) , etc.

A lot of research has been conducted regarding bridge landscape. A large amount of data exists showing aware-ness differences among subjects, where ordinary users and designers evaluate civil structures including bridges from different point of views [1-3]. Research has been made abundantly also on topics such as landscape evalua-tion/support system and search via database in the construction and urban planning fields [4-6]. We also paid attention on Kansei engineering techniques [7] that have been already applied in fields such as product marketing. We have expanded the application of Kansei engineering to girder bridges or arch bridges, and analyzed the rela-tionship between aesthetic feelings that students and bridge engineers have and design component factors [8-10]. However, although Kansei database is necessary to reflect users’ awareness of landscape on design to obtain consensus, there has been only a little research regarding the use of the Kansei database. The purpose of this research is to change design elements required at consen-sus-building occasions and display accompanying evaluation difference while using the Kansei database.

For consensus building, residents’ needs or sense of value must be reflected on construction plans. However, the issue is that there are no tools to reflect them into design. In other words, government officials or engineers cannot present design element changes and accompanying evaluation difference on the spot at consensus-building occasions such as public discussion meetings and work-shops.

To cope with this issue, in this research, girder bridges are selected because they are the most common bridge in

ORIGINAL ARTICLE

Establishment of Kansei Database and Application to Design for Consensus Building

Keiichi YASUDA* and Wataru SHIRAKI*** NEWJEC Inc., 2-3-20 Honjoh Higashi, Kita-ku, Osaka-city, 531-0074, Japan** Kagawa University, 2217-20 Hayashi-cho, Takamatsu-city, 761-0396, Japan

Abstract: Reflecting the recent social background where the importance of bridge landscape design is recognized and the new business style of citizen-involved infrastructure development has started, there has been a growing need of design where aesthetic feeling of actual users is reflected. In this research, a focus has been placed on the Kansei engineering technique where users’ needs are reflected on product development. A questionnaire survey has been conducted for bridge engineers who are most intensively involved in design work and students as actual users. The result was analyzed by factor analysis and the Hayashi’s quantification methods (category I). A tool required at consensus-building occasions has been created to change design elements and display accompanying evaluation differ-ence while using the Kansei database.Keywords: Aesthetic landscape, Girder bridges, Kansei engineering, Kansei database, Aesthetic design

38

Kansei Engineering International Journal Vol.10 No.1

Japan. Aesthetic feelings expressed by bridge engineers and students are analyzed and evaluated based on the Kansei engineering techniques, and incorporated into Kansei database. Then, the system is created that displays evaluation right after design elements are partially changed. Further, the system is created such that one or more adjectives with weightings can be selected from design concept to output the sensitivity of design elements. Finally, attempts are made on the system displaying design elements that affect evaluation together with pictures of bridges.

2. LANDSCAPE EVALUATION FOR GIRDER BRIDGES

2.1 SEQUENCE BY KANSEI ENGINEERING TECHNIQUE

In this research, to study design method using the Kansei database analyzed by the Kansei engineering techniques, girder bridges selected are the most common in Japan but landscape design is considered less important than arch bridges and cable stayed bridges that have erected struc-ture. The process is shown in Figure 1. Specifically, using 43 adjectives to express images of the bridges, a question-naire survey is conducted for 40 university students (composed of 20 female students and 20 male students) and 15 bridge engineers, showing them 90 pictures of bridges for evaluation. The pictures used in this research are selected as follows with the agreement of 5 collage teachers and people who are actually involved in landscape design. First, about 300 bridges are selected from “Bridge

Annals [11]”, the yearbook featuring girder bridges. Then, the bridges are whittled down to 90 by excluding samples that have similar shape, view point, and background. The factor analysis is conducted for each category: all subjects, female students and male students, to examine major clas-sifications of aesthetic feelings people have toward the girder bridges. The girder bridges are divided into small portions and they are considered as design elements. The design elements that affect aesthetic feelings (adjectives describing image) are analyzed using the items/categories table in accordance with the quantification method (cate-gory I). As a result, the impact of design elements of bridge landscape on the adjectives has been clarified. In the end, Kansei database is created incorporating ques-tionnaire results, images of bridges, basic specifications of bridges, analysis results obtained in accordance with the quantification method (category I) such as category scores. Further, the Kansei evaluation and design system created based on this database is expected to directly assist designing of future girder bridges on a practical level.

2.2 QUESTIONNAIRE SURVEYA questionnaire survey is conducted for bridge engi-

neers who are mainly involved in design work and university students (male and female) who mainly use bridges. Totally 90 pictures of girder bridge are selected from a magazine titled “Bridge Annals” [11] and processed into A4-horizontal size. Regarding these 90 pictures, the subjects fill in the questionnaire sheet as shown in Figure 2 with 43 pairs of aesthetic words according to the 5-level scale (-2, -1, 0, +1, +2). The 43 pairs of aesthetic words are prepared by conducting preliminary research as commonly used words for bridge landscape design.

The adjectives describing image are extracted and selected as follows. First, as adjectives representing aesthetic feelings about bridge landscape, 150 adjectives are selected from magazines featuring bridges (for instance, documents 12, 13). Then, the selected adjectives are whittled down to 90 as primary aesthetic words by (a) excluding adjectives which are similar in meaning and (b) keeping adjectives that are definitely necessary for girder bridge design with the help of the comments of actual bridge engineers. Still, the 90 adjectives include ones that are difficult to evaluate or ones that will be evaluated equivalently. Therefore, a preliminary questionnaire survey is conducted in accordance with five evaluation levels to whittle down the 90 adjectives to 43. The selec-tion of the female students is based on the assumption that women will gain a greater voice through future public involvement.

The pictures used for evaluation have various layouts and view points. Quantitative values of partial correlation

43 adjectives

Five-rank evaluation of90 bridges by SD method

Grasp of semantic space(20 females, 20 males,etc.)

60 categories for 20 items

Partial correlation betweenkansei and design elements

Photographs and basicspecifications of bridges,results of analyses, etc.

Display of photographs andcategory scores of bridges

Extraction/selection ofimage-representing

adjectives

Item/category table

Factor analysis

Questionnairesurvey

Analysis by 1st-classmulti-dimensional

quantification theory

Construction of designsupport system

Construction ofkansei database

Figure 1: Procedure of Study

39

Establishment of Kansei Database and Application to Design for Consensus Building

coefficient obtained by the quantification methods (cate-gory I) are used to examine the influence of various layouts and view points on the evaluation.

2.3 FACTOR ANALYSIS RESULTFactor analysis is used to explain multiple variables by

several factors. In this research, the questionnaire result from the female students, the male students, and bridge engineers are respectively analyzed to recognize semantic space with aesthetic beauty of bridge expressed by the adjectives describing image. The factor loadings obtained by factor analysis are shown in Table 1. The loadings are listed in descending order by factor regarding the ques-tionnaire result of the 20 female students.

In the case of the female students, 5 factors are selected as a result of the factor analysis: “aesthetics/artistic nature”, “harmony”, “dynamism”, “affinity”, and “harmo-ny with locality”. The factor loadings of all the factor axes

1 Feminine Masculine2 Youthful Old3 Stable Unstable4 Natural Artificial5 Practical Impractical6 Straight Curved7 Modern Classical8 Urban Rural9 Stylish Unfashionable10 Asserting itself No11 Close, or friendly Distant12 Beautiful Ugly13 Merged into background No14 Trim, or neat squalid15 Warm Cold16 Impressive Unimpressive17 Balanced Imbalanced18 Feeling of material No19 Spacious Congested20 Stereoscopic Flat21 Not losing interest in Losing interest in22 Functional No23 Soft Hard24 Lovely Unlovely25 Weighty No26 Elegant Inelegant27 Robust Weak28 Magnificent Plain29 Harmony with locality No30 Japanese style No31 Refined Boorish32 Colorful Monotone33 Graceful Ungraceful34 Playful No35 Individualistic Common36 Distinctive style Plain37 Symbolic No38 Open Oppressive39 Artistic Plain40 Cozy Uncomfortable41 Harmonized Disharmonized42 Dynamic Lethargic43 Pleasing Displeasing

Figure 2: Questionnaire

from the 1st to the 5th are 0.4 or more, which indicates that the result is reliable. However, there are three adjec-tives that cannot be explained by the factor axes; straight, feeling of material, and functional. Explanation is still impossible even when factor axes are added and the result shows that one factor axis is expressed by one adjective. It can also be concluded that the subjects cannot easily picture these adjectives because the adjectives straight and functional are not categorized in the five factor axes in the case of the male students and bridge engineers. The table below Table 1 shows the specific values, contribution ratios, and accumulated contribution ratios that are obtained by the Varimax method in factor analysis.

Then, the factor axes are given names according to the impression of each factor shown in Table 1. The impres-sion is determined by the adjectives derived from each factor.

Table 2 shows the result when the subject named the factor axes. The factor axis #1 which represents compre-

Table 1: Result of Factor Analysis (Female students)

1st 2nd 3rd 4th 5th

Aes

thet

ics/

artis

tic n

atur

e

37 Symbolic 0.92475 0.04677 -0.01216 0.19948 0.1532335 Individualistic 0.90324 0.05225 -0.01308 0.32629 0.1195416 Impressive 0.89952 0.12319 -0.01595 0.21930 0.0151534 Playful 0.86826 0.05916 -0.08074 0.38971 0.1648310 Asserting itself 0.85566 -0.02059 0.29653 0.06567 -0.192249 Stylish 0.81750 0.45800 0.01866 0.24670 0.0833028 Magnificent 0.80887 0.13278 0.12231 0.16569 0.0220539 Artistic 0.80037 0.31986 -0.09290 0.34430 0.171717 Modern 0.73052 0.49774 0.11909 -0.03459 -0.241092 Youthful 0.72601 0.51779 0.03435 0.13491 -0.259448 Urban 0.70000 0.42307 0.18687 -0.08183 -0.3359031 Refined 0.68070 0.60066 -0.00714 0.15856 0.0266812 Beautiful 0.67366 0.62507 -0.03726 0.28437 0.0039220 Stereoscopic 0.64986 0.03485 0.22415 -0.03418 -0.3468033 Graceful 0.63926 0.56737 -0.00433 0.38611 0.0998824 Lovely 0.63566 0.47705 -0.01984 0.49098 0.1675936 Distinctive style 0.59722 0.15943 0.51310 0.09028 0.1043243 Pleasing 0.59482 0.70925 0.07525 0.17340 0.1299232 Colorful 0.53413 -0.03620 -0.05675 0.42634 -0.06260

Har

mon

y

14 Trim, or neat -0.13213 0.82541 -0.19471 -0.13879 0.0334021 Not losing interest in 0.11332 0.80809 0.03479 -0.09491 0.0424341 Harmonized 0.17612 0.71300 0.19013 0.26389 0.2376126 Elegant 0.44814 0.70820 -0.16294 0.28539 0.1531240 Cozy 0.45538 0.68628 0.09536 0.27962 0.0821317 Balanced 0.09299 0.64153 0.47264 0.03739 0.1329138 Open 0.40198 0.54794 -0.18519 0.27955 0.1673413 Merged into background -0.31834 0.52869 0.02168 0.08909 0.57485

Dyn

amis

m27 Robust 0.09358 -0.11196 0.89707 -0.14683 -0.168893 Stable 0.08986 0.07528 0.85076 -0.05554 0.0214625 Weighty 0.03500 -0.29941 0.83556 -0.26182 -0.2219242 Dynamic 0.41537 -0.18092 0.62153 -0.19791 -0.3837119 Spacious 0.16943 0.31952 0.60803 0.37517 0.203695 Practical -0.30450 0.24651 0.58731 -0.38525 -0.16893

Affi

nity

1 Feminine 0.31181 0.16812 -0.35480 0.74333 0.1141315 Warm 0.26885 0.15713 0.00882 0.69585 0.4331623 Soft 0.20433 0.35786 -0.34171 0.67250 0.2775511 Close, or friendly 0.30666 0.42175 0.14685 0.54824 0.40888

Loc

ality 29 Harmony with locality 0.11220 0.19574 -0.14716 0.09018 0.83761

30 Japanese style -0.02161 0.06997 -0.11126 0.03198 0.812954 Natural -0.49870 0.21935 -0.08430 0.35813 0.539246 Straight -0.23949 0.02999 0.17952 -0.48990 0.1519018 Feeling of material 0.46957 0.11316 0.32286 -0.19683 -0.1641722 Functional -0.05772 0.27621 0.44823 -0.33681 -0.42664

Eigenvalue 12.52945 7.53072 4.83297 4.38653 3.65938Contribution ratio(%) 29.13826 17.51329 11.23947 10.20123 8.51018Accumulated contribution ratio(%)

29.13826 46.65155 57.89103 68.09226 76.60244

40

Kansei Engineering International Journal Vol.10 No.1

hensive beauty shows variations among the subjects. The factors such as “beautiful”, “stylish”, “pleasing”, “grace-ful” and “refined” that are derived from “aesthetics” are common, but while the female and male students add an “artistic nature” axis expressed by “artistic”, “modern”, “magnificent”, “playful”, “symbolic” and “individualis-tic”, the bridge engineers add an “harmony” axis expressed by “harmonized”, “not losing interest in”, “merged into background”, “balanced”, and “open”. The students consider aesthetics and artistic nature similarly as compre-hensive beauty, whereas the bridge engineers consider aesthetics including harmony as comprehensive beauty because they plan and design amiable bridges on the daily basis so that the bridges become harmonious with the surrounding environment in a balanced manner as well as merged into the background.

The factor axis #2 expressed by the female and male students is “harmony”, while that expressed by the bridge engineers is “artistic nature”. Regarding the factor “dyna-mism”, there is a characteristic difference where the female and male students express it by the adjectives “robust”, “stable”, “weighty”, and “dynamic”, whereas the bridge engineers do not regard “dynamic” as dyna-mism, but regard it as included in artistic nature. The all subjects express the axis “locality” by the same adjectives “harmony with locality” and “Japanese style”.

2.4 FACTOR ANALYSIS RESULT BY QUANTIFICATION THEORY (CATEGORY I)

For analysis by the quantification theory (category I), items/categories table are created at first. Items refer to the matters regarding design elements that affect girder bridge scene such as girder and bridge railing color, scene, back-ground color, substructure, and substructure section. Categories refer to further classification of each design element, such as circular type, rectangular type or oval type in the case of substructure, or mountains, flats or rivers in the case of scene. Therefore, 20 items that seem to affect girder bridge scene and their categories are selected and the items are further classified into 60 catego-ries in total.

When people use bridges, some evaluate bridge land-scape similarly but others don’t. A partial correlation coefficient obtained by the quantification theory (category I) represents the degree of impact given on each item, and the larger it is, the more important its element is. To clari-fy the relationship between the created items/categories table and adjectives describing image, average evaluation value is used as input data for each picture and analyzed using the quantification theory (category I) regarding all the 43 adjective pairs for students and engineers. Figure 3 shows the analysis result by the quantification theory

(category I) regarding “beautiful” evaluated by a bridge engineer. The score is represented in histogram on the right side in Figure 3.

A multiple correlation coefficient of each adjective pair represents the reliability regarding the relationship between items/categories and adjectives. Table 3 shows the multiple correlation coefficient of the 43 adjective

Table 2: Designations of Factor Axes

Factor Axes

Female Students Male Students Bridge Engineers

1stAesthetics/

artistic natureAesthetics/

artistic natureAesthetic harmony

2nd Harmony Harmony Artistic nature

3rd Dynamism Dynamism Sense of relief

4th Affinity Locality Dynamism

5th Locality Sense of relief Locality

Table 3: multiple correlation coefficient

adjective No.

Female Students Male Students Bridge engineers

1 0.8148387 0.8519795 0.75026772 0.7895273 0.8305257 0.69986563 0.8787718 0.7994932 0.75603454 0.8301973 0.8323803 0.76861665 0.8095092 0.704629 0.74270816 0.8622028 0.8622518 0.87467047 0.8166912 0.776294 0.7202298 0.8752098 0.8289541 0.76623729 0.7364622 0.7676869 0.683146310 0.7414058 0.7791501 0.750537511 0.6957061 0.7924634 0.693677712 0.7268043 0.7384113 0.707126613 0.7702685 0.7740488 0.802790114 0.8366792 0.8457741 0.72215315 0.7522777 0.7841308 0.756989416 0.6708791 0.6988939 0.647818717 0.7295816 0.7512527 0.740954618 0.7609128 0.7430286 0.839221319 0.7533106 0.7418989 0.709548720 0.8421459 0.8413107 0.802301521 0.7837147 0.7563568 0.654242322 0.755992 0.6685006 0.750329123 0.7722325 0.8215611 0.769169524 0.7426253 0.783294 0.68400325 0.860082 0.878931 0.772846526 0.7858425 0.8074253 0.69529327 0.8458136 0.8325015 0.748568728 0.7602305 0.7327293 0.805305329 0.740013 0.7267234 0.713330530 0.6674385 0.8049419 0.623401131 0.7336285 0.7249115 0.66269232 0.8657165 0.8316611 0.768064433 0.7516882 0.7520237 0.750445134 0.7350024 0.7365523 0.762023835 0.6949731 0.7384491 0.748241236 0.6809605 0.7173119 0.783263437 0.6520694 0.7051879 0.748965538 0.7955365 0.7999528 0.697373239 0.7275389 0.7221718 0.797505140 0.7579134 0.7856287 0.700973141 0.7500517 0.7909434 0.687264942 0.819884 0.7952803 0.797161943 0.7419851 0.7351143 0.7091157

41

Establishment of Kansei Database and Application to Design for Consensus Building

pairs. Some adjective pairs show a coefficient of less than 0.7, but by and large it is 0.7 or more. In conclusion, the reliability of the items/categories is ensured.

3. CREATION OF KANSEI DATABASE AND APPLICATION TO DESIGN

3.1 KANSEI DATABASEThe Kansei database is the database that designers

easily access on PC, and use the Kansei evaluation system

where adjectives, pictures, factor analysis results, scores obtained in accordance with the Hayashi’s quantification method (category I) are computed based on the question-naire results obtained as described above. As for the conventional scene design support system using neutral network, expert system etc., there have been issues such as that it does not successfully handle scene constituents and evaluation changes when the constituents are partially changed, and thus it is only by trial-and-error approach that engineers can determine how and what design

Beauty‑Not beauty

multiple correlation coefficient = 0.707126553

uniform 0.0323 0.0252 0.0059transform -0.0193straight 0.1805 0.1798 0.0320curved -0.1478red 0.2103 0.3261 -0.0736blue 0.0402ivory -0.0110brown -0.0375gray 0.2525green -0.0208gray 0.3749 0.3468 -0.2061brown 0.1407white 0.0903green -0.0470overhanging 0.2566 0.4146 -0.0294columnar 0.1285rectangle 0.0135inverted trapeziod -0.2861circular 0.3568 0.4238 0.2298block‑triangular -0.1940oval 0.02701 0.2448 0.2994 -0.16732 -0.01263 0.03924 0.13205 more -0.0470wall 0.2821 0.3087 0.1981longitudinal sash 0.0051transversal sash -0.1106exist 0.3110 0.2178 -0.1307nil 0.0871exist 0.0362 0.0299 0.0086nil -0.0213exist 0.0836 0.0725 -0.0604nil 0.0121close 0.4237 0.3249 -0.1336middle distant 0.1913lateral 0.2078 0.3121 -0.2878oblique 0.0243upper 0.3450 0.3290 0.1711horizontal -0.1579lower 0.0358mountain 0.3846 0.3306 0.0143flatland 0.1552river -0.1754white 0.1825 0.1863 -0.0365green 0.1468blue -0.0394brown 0.0166green 0.4136 0.5016 -0.0938black‑brown -0.0300blue 0.4078gray‑white -0.0118large 0.0310 0.0305 0.0236middle -0.0069samll 0.0008exist 0.1014 0.0926 -0.0772nil 0.0154exist 0.2872 0.2139 -0.1355nil 0.0784

17.Colors of lower background

20.Presence or absence of obstacles

18.Clearance

15.Scenery

16.Color of upper background

19.Parallel bridge

13.Angle of incidence

14.Height of viewpoint

11.Inspection passage & otheraccessory structures

12.Sight distance

9.Drain pipes

10.Lighting posts

7.Number of piers

8.Type of balustrades

5.Shapes of substructures

6.Sectional forms of substructures

4.Color of balustrades

1.Plan view

2.Shape of main girders

3.Color of Girders

Item Category  Score Range Partial correlation

coefficient

‑0.5 0.0 0.5

Figure 3: The result of quantification theory (category I) for bridge engineeer

42

Kansei Engineering International Journal Vol.10 No.1

elements should be changed to obtain a higher evaluation score. Figure 4 shows the configuration of the Kansei database proposed in this research. In the database, ques-tionnaire results of the female and male students and bridge engineers (average and variance), SD profile of each bridge, and category score of each adjective, basic data of girder bridge, images, etc. are registered. Design-ers use the database to search for aesthetic words (adjectives describing images) regarding the bridge they start planning and designing. As a result, images that suit the aesthetic words, SD (semantic differential) profiles, several most highly ranked and poorly ranked bridges according to category score and the questionnaire result are displayed on the screen. Engineers can proceed design work based on this search result. In the Kansei database shown in Figure 4, engineers and ordering companies are also included, but they are not applicable in this research.

3.2 APPLICATION TO DESIGN3.2.1 APPLICATION AT DETAILED DESIGN STAGE

This section explains how to apply the created Kansei database to landscape design operations.

At detailed design stage, trial-and-error approach must be taken while determining what changes of design element and how much affect evaluation. Evaluation simulation according to category change is useful for such

occasion. As shown in Figure 5, by selecting applicable category by items, the score of adjective describing image in need can be obtained immediately. Changes frequently occur while proceeding detailed designing, and often re-evaluation is necessary when category is partially changed whether it occurs at plan designing stage or maintenance and control stage. Using this simulation, however, one can immediately know whether the change has a good effect or not.

Figure 7 shows the calculation model for the input/output data.

Figure 5 shows the example where female student is selected and category is selected for each item. Score for each adjective from no.1 to 43 is displayed as shown in Figure 6. In this way, evaluation can be conducted imme-diately when design element is partially changed.

3.2.2 APPLICATION AT PLAN DESIGNING STAGEApplication of the Kansei database at planning design

structure(bridge , load,

ri ver, dum, airport, park,

et c.)

input of user images

(adjectives)

Int ernet

Query

Int ernet Int ernet

Questionnaireoutput

kansei database

resul ts(quanti fi cation, category score,

SD profiles , bridge bas ic

dat a, etc.)

student s, bridge

engineers, designers ,

orderers, et c.

Figure 4: Configuration of Kansei Database

Figure 5: Image-1of System (Picture of Inquiry)

Figure 6: Image-1of System (output)

43

Establishment of Kansei Database and Application to Design for Consensus Building

stage is also possible. When the design concept is deter-mined at the initial stage of landscape design, it needs to be converted to specific adjectives describing image.

For instance, when the concept is converted to 5 adjec-tives describing image; “beautiful”, “merged into the background”, “balanced”, “never getting tired of looking at” and “harmonious”, the applicable boxes are checked as shown in Figure 8, and expected users are selected. At this point, weighting can also be considered for each adjective. All boxes can be checked too.

After input is completed, category score is calculated based on the selected multiple adjectives which describe

image and their weightings, and several bridges that are most highly ranked and poorly ranked in the order of total category score, respectively, are displayed in Figure 9 and 10. Engineers can refer to this output example and imme-diately grasp design features of highly ranked bridges and poorly ranked bridges.

Figure 11 shows the calculation model for the input/output data.

3.2.3 VerificationNew data was input in the established Kansei database

E=s · c (1)

where,E: total score of each adjectives: category score of each subject calculated by quanti-

fication theory 1 (1 ~ 60)c: selected category (1 or 0)

scorecategory

1 2 3 4 5 · 56 57 58 59 60

adjective

12345·

3940414243

×

selected category

category1 2 3 4 5 · 56 57 58 59 60

1 0 1 0 1 1 1 0 1 0

Figure 7: Calculation model for input/output data

Figure 8: Image-2 of System (Picture of Inquiry)

Figure 9: Total high rank category Scores of Bridges

Figure 10: Total low-rank category Scores of Bridges

44

Kansei Engineering International Journal Vol.10 No.1

to evaluate the system in terms of matching ratio or repro-duction ratio of the data retrieved.

First, a questionnaire survey was conducted for 24 female students regarding two bridges (Photos 1, 2), focusing on how evaluations change according to category change. The questionnaire result was compared with the result when the categories for the two bridges were input in the system (see Table 4). Figure 12 shows the system output when the categories for Photo 2 were input.

The questionnaire and system results show substantially equal output, which means evaluation result can be obtained simply by inputting categories for bridge in the system without questionnaire surveys. Girder color and clearance are different in Photo 1 and 2, which is repre-sented well in the figures for “massive”. Photo 2 is merged into the background better, as the questionnaire result also shows. Category can be changed only by clicking the

necessary button and it was verified that results could be obtained instantly.

Next, the result was verified using the past actual design regarding display of bridges that conform to a design concept. The verification was conducted by inputting one or more adjectives expanded from the design concept. The actual bridge over the river flowing through the residential area was selected. Based on the actual landscape concept of “balance in landscape (harmony between uniqueness and symbol)”, adjectives impressive, colorful, symbolic, unique, and beautiful were selected and input in the system. Figure 13 shows the input screen and Figures 14 and 15 show the output result. Figure 14 lists the output in

I=s · w · b (2)

where,I: total score of each adjectives: category score of each subject calculated by quanti-

fication theory 1 (1 ~ 60)w: selected concept weightingb: bridge basic data classified by items/categories (1 or 0)

scorecategory

weight1 2 3 4 5 · 56 57 58 59 60

×adjective

123 145·

3940 1414243 1

×bridge item/

categorycategory

1 2 3 4 5 · 56 57 58 59 60

bridge

1 1 0 1 0 1 1 1 0 1 0

Σ(score*w eight*br

idge item/cat egory)

2 1 0 1 0 1 1 1 0 1 03 0 1 1 0 1 1 0 1 1 04 1 0 1 0 1 1 1 0 1 05 1 0 1 0 0 0 1 0 1 06 1 0 1 0 1 1 1 0 1 0·

85 1 0 1 0 1 1 1 0 1 086 0 1 0 1 1 1 1 0 1 087 1 0 1 0 0 1 0 1 1 088 1 0 1 0 0 1 1 0 1 089 1 0 1 0 0 0 1 0 1 090 1 0 1 0 1 1 1 0 1 0

Figure 11: Calculation model for input/output data

Table 4 : Comparison between questionnaire result and system output (female students)

Photo 1 Photo 2Question

naire result

System output result

Question naire result

System output result

Stylish 0.32 0.2517 0.45 0.4303Beautiful 0.25 0.3081 0.22 0.4735Merged into the back -0.55 -0.3611 0.87 0.5052Massive -1.02 -0.6348 0.02 -0.1238

橋梁 11

Photo 1

橋梁 17

Photo 2

45

Establishment of Kansei Database and Application to Design for Consensus Building

ascending order, whereas Figure 15 lists it in descending order. When comparison is made between Figure 14 and Figure 15 in terms of impressive, colorful, symbolic, and unique, it is obvious that bridges with a higher total score conform to the design concept better. As the figure shows, similar bridge images are displayed when inputting adjec-tives related to the design concept, and designers can perform design work with reference to these images. It is effective in design work to understand factors that will give negative impact on evaluation. In other words, designers can make effective use of the opportunity when they see bad examples. Comparison among bridges with both lower and higher total scores makes it easier to

Figure 12: System input/output resultunderstand design elements that affect evaluation result.

It seems that we need a mechanism where items/catego-ries or adjectives are assumed automatically from pictures to establish the database system. Currently, items/catego-ries are assumed manually while engineers look at pictures. Automatic assumption has not been achieved yet because category classification often requires judgment of engineers. For example, it is extremely difficult only from the pictures to tell the shape of main girders, substructures, and angles of incidence. If a mechanism is achieved where categorization is carried out automatically from the pictures, establishment of the database will be accelerated. We will continue research on this point.

Figure 13: Input screen

Figure 14: Output screen (in ascending order of score)

46

Kansei Engineering International Journal Vol.10 No.1

4. CONCLUSION

In this research, Kansei that differs between students who use bridges and bridge engineers has been studied focusing on girder bridges to obtain more helpful Kansei database. Also, a specific method of applying the Kansei database to actual design work has been studied, and summarized as below.

1) As a result of factor analysis, the 6 factors are derived for female students, male students, and bridge engineers, respectively. While the female and male students add an “artistry” axis expressed by “sense of beauty/artistry” to the aesthetic factor, the bridge engineers add an “affinity” axis expressed by “aesthetic affinity” to the aesthetic factor. Another difference between the students and engi-neers is that the female and male students consider “affinity” as the factor axes #3, but the bridge engi-neers consider it as the factor axes #1 and axes #2.

2) A tool required at consensus-building occasions has been created to change design elements and display accompanying evaluation difference while using the Kansei database.

Aesthetic feeling of users have been considered more and more important as infrastructure development through public involvement has become mainstream these days. Design reflecting users’ Kansei 100% is not mandatory,

Figure 15: Output screen (in descending order of score)

but accountability to users should be prioritized in the future to respond to social demand such as information disclosure and transparency. The Kansei engineering techniques using database is thought to be appropriate for realizing such design where Kansei of users and engineers is associated with design elements quantitatively.

The verification conducted for the students and bridge engineers will be oriented toward designers, ordering companies and housewives, and the target bridge for analysis will be expanded to arch bridges. It is expected that bridges that afford satisfaction to users is possible by further improving and systematizing the Kansei database, so that anyone can access and use it easily and effectively.

REFERENCES1. Y. Mori, N. Nishimura, H. Sato, S. Tanaka ; A Study

on resident’s evaluation for the landscaping and aesthetic design of expressway, Proc. of JSCE, Jour-nal of Infrastructure planning and Management, No.524, IV-29, pp.23-35 (1995).

2. Y. Shono, Y. Inoue, M. Nakazono, K. Nakagawa ; Relation between design method and psychological evaluation for the structures in highway, Proc. of JSCE, Journal of Construction Management and Engineering, No.528, VI-29, pp.103-113 (1995).

3. Y. Shono, Y. Inoue, M. Nakazono, K. Nakagawa ; Dependence on evaluator’s occupation in case of evaluation for the design of highway structures, Proc. of JSCE, Journal of Construction Management and Engineering, No.553, VI-33, pp.93-102 (1996).

4. N. Uchiyama and T. Shibata ; Analysis and model on harmonization between urban landscape and build-ings - Method of multi-modal Kansei retrieval systems- , Journal of Architecture Planning, Architec-tural Institute of Japan, Vol.73, No.623, pp.241-248 (2008).

5. T.Nanmatsu, K. Tadamura, S. Ikaruga, Y. Tabuchi : Development of a system for supporting landscape assessment, Journal of Information Proceeding Soci-ety of Japan, Vol.45, No.6, pp.1663-1671 (2004).

6. N. Murooka, Y. Horita, K. Honda, T. Murai : Consid-eration of the image retrieval system using KANSEI information as the reference key, Technical report of Information Proceeding Society of Japan, 2004-CVIM-142, pp.1-8 (2004).

7. M.Nagamachi ; Kansei Engineering, Kaibundo, 1989.8. K.Yasuda, W.Shiraki, M.Dogaki, K.Kawazu,

M.Adachi ; A study on assessment and design of aesthetics of landscape with girder bridge by Kansei Engineering approach, JSCE, Journal of Structural engineering, Vol.45A, pp.543-551 (1999).

9. W.Shiraki, H.Noda, M. Nagamachi, Y.Matsubara,

47

Establishment of Kansei Database and Application to Design for Consensus Building

M.Adachi ; Construction of Kansei database for arched bridges and its application to aesthetic assess-ment, JSCE, Journal of Structural Engineering, Vol.45A, pp.553-560 (1999).

10. K.Yasuda, W.Shiraki, M.Adachi, Y.Mikumo, M.Dogaki ; A study on assessment and design of aesthetics with girder bridge using Kansei engineer-ing method, Proc. of JSCE, Journal of Construction Management and Engineering, No.665, VI-48, pp.103-116 (2000).

11. Japan Society of Bridge Construction ; Bridge Annals, No.1987 to 1993.

12. Japan Society of Civil engineers : Manual for Aesthetic Design of Bridge, 1982.6.

13. O. Shinohara: Hashi no keikan dezain wo kangaeru, Giho-do, 1994.6.

Keiichi YASUDAreceived the B.E. degree in Civil Engineering in

1983 from Tottori University. Afterward, he had

joined NEWJEC Inc. and worked as a bridge plan-

ning, analysis and design engineer. Then, he

received the Ph.D. degree in 1999 from Tottori

University. His research interest includes a relation-

ship of Kansei Engineering with bridge design and estimation,

infrastructure management, soft computing. He is a member of JSKE,

JSCE, JCI, JCCA and so on.

Wataru SHIRAKIreceived the M.E. degree in Engineering from

Kansai University in 1974 and received Doctor’s

degree in Civil Eng. from Nagoya University in

1980. Since 1998, he has been a professor in the

Faculty of Engineering, Kagawa University. He was

a guest professor at Innsbruck University from 1987

to 1988. His research interest includes “Aesthetical Assessment and

Design of Structures using Kansei Engineering, “Applications of Soft-

Computing to Civil Engineering”, “Environmental Engineering”, “Social

Systems Engineering”,” Information and Communication Engineering”

and “Reliability-based and Performance-based Design of Structural

Systems and Large Scale Network Systems”. He is a member of JSKE,

JSCE, JGS, JSSC,JES and so on.