Published Papers - HARADA Toshinobu

Decision Supporting System for Car Surface Forms : Reasoning Optimum Solutions from Imperfect Solutions Reasoned by Fuzzy Reasoning

Harada Toshinobu, Mori Norihiko, Sugiyama Kazuo

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  42 ( 3 ) 49 - 58   1995  [Refereed]

　View Summary

In this study, construction of decision supporting system for reasoning of car surface forms from concepts and required specifications was attempted. Reasoning engine of this system was adopted fuzzy reasoning. And, we concentrated on study of method for reasoning optimum solutions from imperfect solutions reasoned by fuzzy reasoning. Two systems were constructed using Hopfield neural network and genetic algorithms in the part of reasoning optimum solutions from imperfect solutions. And then, two systems were compared. As a result, it was confirmed that we acquired only one reasoning solution by the system using Hopfield neural network, instead of requiring no prior knowledge of surface forms' mismatching. Further to this, it was confirmed that we were able to acquire plural reasoning solutions by the system using genetic algorithms, instead of requiring some knowledges of surface forms' mismatching.

DOI

Development of Various Solution Using Identify Mapping Model : a Support System to Define Car Configuration (2)

Harada Toshinobu, Mori Norihiko

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  41 ( 1 ) 51 - 58   1994

　View Summary

A designer's creative work is to inquire into the shape and configuration for a given concept, this activity is aimed at solving the "inverse problem" talked about recently. There are a few studies of the solving of "inverse problems" in the field of design. However, these studies use fuzzy relational equations and genetic algorithms, in the former, the most suitable solution within expressed limits is found and in the latter we get a small number of solutions. Thus it is doubtful that we get all suitable solutions. Therefore, an inverse reasoning system with an identity mapping model, based on a naural network, was developed. The system is able to compress the information from a neural network model into three-dimensional data, enabling easy visualization. Accordingly, we can inquire into various solutions.

DOI

Algorithm of Creating the Curve by Controling it's Characteristics

Harada Toshinobu, Mori Norihiko, Sugiyama Kazuo

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  41 ( 4 ) 1 - 8   1994  [Refereed]

　View Summary

Initially, by means of example, the way in which the characteristics of a curve affect an observer's impression of the curve was studied. As a result it was possible to raise a hypothesis defining some relationship between the characteristics and the impressions formed. Secondly, an algorithm was developed to output the loci of a curve derived from a definition of desired impression. This algorithm was used to produce a number of curves and these were assessed as to whether they affected the desired impression. The success of this second investigation allowed the confirmation of the hypothesis. Subsequent investigations determined that it was possible by this algorithm to generate a variety of curves of specified natures by controlling these characteristics. Futher to this it was considered that by this algorithm curves could be generated with characteristics beyond those of general curve expressions (eg. parabolas, sine curves, etc.) and that these may be applied to the key lines in vehicle design.

DOI

Study of Quantitative Analysis of Curve's Character

Harada Toshinobu, Mori Norihiko, Sugiyama Kazuo

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  40 ( 6 ) 9 - 16   1994  [Refereed]

　View Summary

The aim of this study was to develop a quantitative analysis of a curve's character from the view point of a human observer. The research investigated how designers control a curve's character by the use of image words. As a result of this it was found that the character was controlled by changing the pattern of curvature and volume. From this a quantitative analysis was developed. Further to this, a number of drawing curve and curves used on existing car surfaces were analysed. As a result, it was verified that it is possible to correlate the character and numerical expression for those curves. It is thought that this quantitative analysis could form an index in the design of a car's surface and a key in researching the relation between a curve's character and image words.

DOI

Analysis of Process of Defining Car Configuration : Developing a Support System to Define Car Configuration

Harada Toshinobu, Mori Norihiko

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  1993 ( 96 ) 39 - 46   1993  [Refereed]

　View Summary

Recently, as a car stlye has reached maturity, the configuration (it means the layout and the proportion) of a car becomes to be defined early in the phase of the development of a car style by planners, designers, and engineers. Aims of this study is the analysis of factors effecting the cinfigration of a car. And, we construct a neual net work model -reasoning a cabin configuration from the configuration of a car body-, and a linear multiple regression model -reasoning a cabin configration from the interior configration items- and then, make simulations by means of these models. As a result, it becomes clear that the configrations of a car are brought partially by a linear thinking and by a non-linear thinking. And then, we get some guide line for designing the configuration of a car.

DOI

Inquiring into Image Using Genetics Algorithm

Harada Toshinobu, Mori Norihiko, Sugiyama Kazuo

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  40 ( 2 ) 41 - 48   1993  [Refereed]

　View Summary

So far, some studies of image characteristics and inference have been done. However, thoes methods obtain results convergently, not divergently. In this study we develop a system in which results are obtained divergently using 3 genetics algorithm that bases on the the theory of a life's evolution (selection and mutation), and compare these results with those obtained by a neural network using some sample concepts. Subsequently, we confirm the viability of the system using the genetics algorithm. Further, we develop a system connecting genetics algorithm and neural network in paralle, and as a result we confirm that this system gets results has advantages of both of methods, and can controls these advantages by a parameter.

DOI

A Research on Relation between Constituent Element and Preference of Car Interior Design(TECHNICAL SESSION CONTENTS,THE 37TH ANNUAL CONFERENCE OF THE JSSD)

Endo Makoto, Mori Norihiko, Ushida Msayuki, Harada Toshinobu

Bulletin of Japanese Society for the Science of Design ( Japanese Society for the Science of Design )  1990 ( 81 ) 71 - 71   1990

DOI