Part 1. Problem
Suppose we have an order for designing a new variant of a known product. For example, a toothbrush. This is a design problem and it implies a great variety of possible solutions. We usually have some requirements and restrictions. They decrease the number of possible solutions, but, and after that we have a lot of equal variants and their combinations. What will the properties of a bristle be? What will the surface of a handle be? What will the shape of a head be? What kind of a unit between the head and handle will be? And so on... How to find and select an optimal solution?

A determining solution direction is normally made by the method of extrapolation. The past evolution of a product is examined and a trend is continued. But such method gives us a very narrow field of view. It looks like a strict route for the visitors in a forest reserve instead of free walking in any direction in an ordinary park. We think it is very useful for a designer to have a big map for his searching. In this situation we can see all the possibilities at the same time. 
 

Research work of a designer can be compared to a work of an exploration geologist. If a geologist has a special map of his region of exploration, he works more effectively. Such map may include some parameters. This is an informational fund, which is structured and united in one image at the same time. By the aggregate of the information a geologist can determine the most probable place of occurrence of the minerals.

A similar map can help a designer too. A designer must keep a lot of information about a product: all its history, tens of the new and known possible solutions. When a map contains complex information about the possible variants of a product, a designer can free his head for more useful things: analyzing and selecting solutions. Informational fund in the structural and visual form helps him.
System presentation of information permits to analyze and to compare all the possibilities. A designer does not forget anything. All information is kept in one image. By a design map of the projected product a designer could select more perspective solutions. 
 

How to create such a map? How to enumerate all the possible variants and how to structure them?

Part 2. Basic principles
What do we need for creating a map of possible solutions? First is to select an objective classification criterion. Second is to determine the degree of generalization of information on a map. Third is to determine the information unit.

Classification Criterion
For technical world an objective criterion is laws and trends of engineering system evolution. They are thoroughly elaborated in TRIZ. Every step of every trend is transformation of an object. The known regularities may be used for building a “tree of transformations”. The evolution lines will correspond to guides of the tree. The steps on the evolution lines will correspond to key points of the tree. 
 

Degree of Generalization
Selecting solution occurs at the stage of the conceptual projecting. This is a stage, when a designer needs to select the general idea. Our map can give a recommendation, some advice, but not the precise prescription. Excessive specification will increase the number of solutions that differ little from each other. But a high degree of abstraction will give too indefinite recommendations. We think that it is the level of principal distinction between transformations that will correspond to the level of conceptual generalization.  

An Information Unit
What will the information unit be in a transformation tree?
In TRIZ and in traditional designing, an engineering system is determined through the function it performs. The main useful function of an engineering system being designed may be divided into a series of subfunctions (elementary functions). An elementary function is unable of being divided into any more functions that have distinction of kind. The examples of such functions are: “to cut an object”, “to move an object”, “to cool an object”, and so on.
To describe all possible ways of performing an elementary function, the latter should be presented in the form of an elementary action in the operational zone. To perform the action, we need: a tool (to perform the action) and an article (that is acted upon). In addition, environment may be taken.

We think that the information unit for a tree of transformations is a single, simple transformation of an element-participant of the operational zone. For example: “sand flow cleans a surface”, “two smaller sand flows clean a surface”, “many smallest sand flows clean a surface”.

Part 3. A basic transformation tree
A short definition of the main idea of the method:

Determining the ways of performing the functions of a designed system on the basis of engineering system evolution trends.

Movie “tree” demonstrates an experiment with building a fragment of the basic, universal tree of evolution for a monolithic solid 3D (three dimensional) object. In principle, any of trends may become the “trunk” – the main axis – of a tree of transformations. But it is known from experience that it is more convenient to use the trend “Segmentation”: (a solid object, an object divided into two parts, an object divided into many parts, powder, paste or gel, emulsion, or suspension, liquid, foam, gas, plasma, field, vacuum).
Each step is a starting point for the evolution of other trends. There is one interesting feature of that tree. With the evolution of the trend, the number of possible branches decreases. A solid object may evolve along the following lines: mono-by-poly, geometrical evolution, segmentation of volume, dynamization, activization, introduction of additional substance. At the “gas” step of the segmentation line, they are the following lines: mono-by-poly, dynamization, activization, introduction of additional substance. In the “field” state, they are: mono-by-poly, dynamization. Vacuum development is extremely limited (dynamization of the state “vacuum – non-vacuum”). 

A new structure is built from each point, because each new transformation may be accompanied by a number of transformations in different directions. As a result, a multidimensional structure arises. It is similar to the multidimensional morphological box. But a transformation tree differs from a morphological box in that the variants are arranged along the axes not randomly but in accordance with the evolution trends.

For a specific design problem, a tree of transformations will have its own, unique form. But any specific tree is built on the basis of a universal, basic tree for a given class of objects. The availability of a universal sample facilitates the work. Let us refer again to the analogy with geographical maps, for instance, with a route map of city transport. 
 

There is some background in the form of a city map. There are bright lines of bus routes drawn against this background. If we need to schematically represent the routes of trams and underground, we will use the same background, but bright lines will be drawn along other streets. On the basic tree of transformations, you will select a 'reference point' – the actual state of an object – and will line out existing and probable “routes” of its changes.

Part 4. Program structure
We have developed the prototype of a program based on the above-described methods. The program has the next destinations: to acquaint users with possible methods of performing a necessary function, to enable them to combine the methods and to select the optimal solutions using the parametric filter.
The heart of the program is the basic tree of transformations. The database composed of images of a single transformation of an object and text descriptions of the transformation. Database is connected with the tree. The working field is the “scene” where the “assembly” of function-performing elements occurs. 
 

Prior to starting the practical work with the program, it is necessary to analyze the system being designed, to identify its main function, analyze it into elementary functions, and determine a tool and article for each function. TRIZ has developed detailed methodical recommendations necessary for performing these operations.
1. From the list of elementary functions, the user must select the function, the methods for performing which he wishes to determine.

2. Then he must select graphic analogs of the tool, article and, if he wishes, of the environment. The specific name of the object must be transformed into “descriptive” by identifying the main distinctive features of the object.
The selected objects are visualized in the operational field – the “scene” -- thus visualizing the function performance.

3. To narrow the search field, it is recommended to apply a “parametric filter”: to select from the offered list a parameter to be improved under the problem conditions.
A unique menu is created for the user. This menu reflects the properties preset by the user. This menu will list only those trends and separate steps of the trends, which improve the selected parameter. The menu consists of three parts – a separate part for the tool, article and environment. The menu can be textual or more visual – pictographic. If the user does not apply the filter, he will obtain a full menu with all transformation alternatives.

4. Selecting a specific transformation step from the menu changes a respective element of the scene. Transformations may be summed up (except mutually exclusive cases) and combined. Placing them in a special storage may save the obtained versions of function performance -(concepts) – and then the user may work only with selected versions.

5. Elementary functions may be combined to form functional chains.

Conclusion
The availability of the multi-dimensional tree of transformations allows determining all the basically differing methods of function performance. This increases the effectiveness of concept generation by a designer and makes possible objective selection of a concept, provides basis for making forecast of system evolution, and enables the designer to by-pass patents and to create patent umbrellas.
The computer program based on the above-described methods will allow formalizing and automating the process of building a tree of transformations for a system selected by a user.

An earlier version of this paper was presented at the European TRIZ Association meeting, “TRIZ Future 2002,” 6-8 November 2002.