JP2007527456A - Method and apparatus for producing perfume and fragrance compositions - Google Patents

Abstract

  The present invention relates to a method for generating a fragrance and / or fragrance composition using a composition database comprising a recipe vector and an attribute vector associated with a group of base compositions. A target attribute vector is defined in accordance with the method of the present invention, and an operator is determined that achieves the conversion of the target attribute vector into the attribute vector of the recipe vector in at least one environment. Also, the target recipe vector is converted into a target attribute factor using the operator, and is determined under the condition that the defined substance components are mixed in proportion to each other according to the target recipe vector.

Description

  The present invention relates not only to an apparatus for carrying out the method, but also to a method for producing a perfume and / or fragrance composition.

  In the field of perfume and fragrance production, there is a constant need for new or modified fragrances and fragrances that must meet the most diverse demands, respectively. A general problem in this connection is described, for example, in German Offenlegungsschrift 10144816 A1. The search for suitable fragrances and fragrances is particularly difficult due to the fact that the perception of fragrances and fragrances, respectively, and the relationship between the odor and the chemical structure of the flavoring substance is not well known. Furthermore, it has been found that minor changes in the structural composition of known odor and flavoring substances can cause strong changes in their olfactory and scent characteristics. The relationship between commonly used fragrances and fragrance compositions consisting of a mixture of odors and flavoring substances is even more complex.

  EP 1 271 140 A1 describes a method for determining sensory attributes of foods in a dispersed liquid phase. A sample of food is thereby put into an environment that matches the conditions in the mouth during food intake to the extent possible. As a result, sensory attribute prediction is performed based on the spatial distribution of the liquid phase under these conditions.

  However, it is not possible to make a general prediction of the sensory attributes of fragrances and fragrance compositions. In addition, EP-A-1 271 140 A1 discloses known sensations for the production of new creations that are as close as possible to the predetermined sensation profiles, especially those proposed in the field of fragrance and fragrance production. It does not address the problem of handling attributes in view of their practical application.

  Some time ago, various evaluation instruments were developed to make the complex impressions and relationships used as objective as possible to quantify sensory characteristics. Furthermore, the quantitative description analysis (QDA) introduced around 1974 was developed, for example, to the so-called QFP method (“Quantitative Flavor Profiling”. XP008033289 (CRStampanoni, quantitative fragrance profiling, fragrance recognition). An effective tool: Quantitative flavor profiling. An effective tool in flavor perception. Food Marketing and Technology, 1993, Sensory Dep., Glvaudan-Roure Flavor Ltd., Juvendorf, Switzerland ( Dubendorf, Switzerland), Vol. 7, No. 1, February 1993, pages 4-8. Such a method can be applied to a fragrance and / or fragrance composition depending on the subject being used. Allows independent assessment and explanation, however, the methodology can be applied to new fragrances and / or fragrance compositions (from now on and Converting the development and operational well-defined concept for the production of called "composition"), the has not been described heretofore.

  The object of the present invention is to specify a method for producing perfume and / or fragrance compositions so that the above disadvantages and limitations are avoided. This problem is solved by the method defined in claim 1 and by the device defined in claim 9 according to the invention.

In accordance with the method of the present invention, we start with a composition database consisting of a number of recipe vectors and attribute vectors associated with a group of base compositions. Each one of the base compositions can be manufactured by mixing predetermined substance components, the proportion of the individual substance components being represented by the corresponding recipe vector. In addition, evaluation results regarding the characteristics of the selected sensory attributes of the base composition are available and are expressed in the form of attribute vectors associated with each one of the base compositions. The method according to the invention is as follows:
a) specifying a target attribute vector;
b) determining an operator that achieves a conversion from the recipe vector to the attribute vector in at least the environment of the target attribute vector;
c) establishing a target recipe vector, provided that it is converted to a target attribute vector by using the operator;
d) mixing the designated substance components in proportions according to the target recipe vector;
Consists of.

  In this context, the solvents, carrier substances, etc. required depending on the application can also be regarded as “substance components” in addition to various perfumes and / or fragrance substances.

  A numerical ordered series that specifies the proportion of material components used to create a perfume and / or fragrance composition is considered a “recipe vector”. The m-th component (in algebraic sense) of the recipe vector of the n-th composition indicates what ratio the substance component “m” represents in the composition “n”. .

  Similar to the recipe vector, the “attribute vector” can be expressed as an ordered series of numerical values that specify numerical evaluation results of individual sensory attributes.

  The method according to the invention uses a composition database comprising information about a group of base compositions, whereby it is represented on the one hand—in the form of the recipe vector—the composition, and on the other hand—the form of the attribute vector. -Sensory attributes are stored for each base composition.

  Perfume and / or fragrance compositions can be generated with desired attribute characteristics based on this composition database. The desired and intended sensory attributes are thereby represented as “target attribute vectors” and are expressed in the form of an ordered series of numbers representing the desired numerical values of the individual sensory attributes. . As a result, the “target recipe vector” to be considered as a mixture support for the desired composition is determined. For this purpose, an attribute vector that is already stored in the composition database and is closest to the modified attribute vector so that the associated recipe vector is known is at least referenced. Based on the known relationship between these attribute vectors and the recipe vector, an operator is established to achieve the conversion from the recipe vector to the attribute vector at least in the environment of the modified attribute vector. From now on, the target recipe vector will be pursued, and therefore it must have features that can be transformed into the desired target attribute vector by using the previously established operator. Finally, the additional composition is created by mixing the specified substance components in proportions according to the target recipe vector.

  In this context, we are working on a method for the production of new but not yet known compositions, especially for the production of compositions that can be used. However, the method is based on a composition database that contains information about many base compositions, ie one attribute vector of the base composition is often desired when specifying the target attribute vector. “Production” can also be regarded as “modification” of a previously characterized composition since it is used and modified in the direction in which it is used. The terms “production” and “modification” are used hereinafter as equivalents in the context of this.

  An additional object of the invention is to specify an apparatus for carrying out the method according to the invention. This object is achieved by an apparatus as defined in claim 9. The device according to the invention features not only a data processing device but also a mixing device controlled thereby, the data processing device comprising means for determining an imaging operator, means for inputting an attribute vector, Input, storage and retrieval of at least one composition database as well as means for the calculation of modified recipe vectors and means for the transfer of control signals shaped according to the recipe vectors to the mixing device Means are also provided.

The mixing device comprises the following components:
a) a plurality of storage containers that can be filled with individual substance components;
b) a plurality of receptacles;
c) a controllable supply device for providing a specified amount of individual substance components from the corresponding storage container to the receptacle to create a perfume and / or fragrance composition;
Is featured.

  Advantageous embodiments of the invention are defined in the dependent claims.

A preferred method for creating a composition database is defined in claim 2 and includes the following procedural steps:
a) creating a group of base compositions by mixing the substance components in proportions according to each recipe vector associated with the base composition;
b) quantitatively evaluating each base composition with respect to the selected sensory attribute and the creation of the associated attribute vector;
c) creating the composition database by storing the recipe and attribute vectors such that the vectors associated with each base composition are usable with respect to each other and with respect to the base composition;
Is provided.

  Due to the fact that the vectors associated with each composition are usable with respect to each other and with respect to the composition, a highly complex relationship between the sensory attributes of the composition and the perfume and / or fragrance composition Can be measured systematically and used for additional analysis.

  As is generally known, the assessment of sensory attributes for fragrances and / or fragrance compositions is a significant variation based on the factors that the sensory assessment by an individual often influences the subject. It becomes difficult due to the fact of showing sex. It is therefore preferable to use a qualifiable procedure. This can be performed by a quantitative description analysis according to claim 3.

  Advantageously, only a selected portion of the attribute resulting from the quantitative evaluation is used to create the attribute vector. These attributes are advantageously selected by factor analysis as claimed in claim 4.

  The setting of the operator to achieve the conversion from recipe vector to attribute vector can be difficult due to the fact that multiple very different recipes can lead to compositions of very similar sensory attributes. Thus, the method of claim 5 is advantageously used, and the bijection relationship between the attribute vector and the recipe vector of the composition database is based on local conditions in the environment of the modified attribute vector. Used by multiple regression networks and / or neural networks and / or expert systems to set the desired operator.

  When preparing a group of fragrances and fragrance compositions, the sensory sensation can be achieved by selectively adding substance components based on several substance components that serve as a basis, such as defining a wonderful fragrance and flavor note. It is appropriate to build a series of compositions with slightly different attributes. Basically, different procedures can be applied to this. Moreover, it can be advantageously executed by the statistical test plan described in claim 6. An apparatus designed for this purpose is defined in claim 9. The means thereby provided for the generation of recipe vectors is not only storage, ie means for storing the generated recipe vectors within the composition database, but also data entry, data processing, and data It can be created in particular in a programmable recipe module that also comprises means for output.

  In order to explore the composition of the desired sensory attribute, a corresponding attribute vector is advantageously specified, which can be considered in particular a modification of the attribute vector already available in the composition vector. The modification can in particular consist of the fact that certain attributes must be strengthened or weakened. Advantageously, such modification is performed interactively based on appropriate visualization and input means, such as via a monitor or computer mouse. Although there is a desire to replicate a given fragrance and / or fragrance composition of an unknown composition, the modified attribute vector can be obtained by, for example, attribute analysis by quantitative description analysis according to claim 3, It can be determined according to claim 7 by the fact that it can be carried out for the composition.

  The most diverse types of graphic representations are possible for visualizing and processing attribute vectors. The attribute vector is advantageously expressed in the form of a polar diagram, so-called “cobweb”, in particular as claimed in claim 8. Associated with each attribute is a beam section originating from each other and whose length is a specified measurement of the attribute in question. Advantageously, the beams are arranged at equidistant angles.

  Embodiments of the invention are described in more detail below with reference to the drawings.

  The device represented in FIG. 1 for the production of a fragrance and / or fragrance composition comprises a data processing device 2 and a mixing device 4 controlled thereby. The mixing device 4 comprises a plurality of receptacles 8, 6a, 8b, 8c, etc. for fragrance and / or fragrance composition, such as reagent glass, as well as a plurality of storage containers 6, 6a, which can be filled with individual substance components. Etc.-for example small glass bottles-are also featured.

  The controllable supply device 10 is supplied by a pressure gas source, not shown, from which supply pipes 14, 14 a, etc. diverge and lead to the individual storage containers 6, 6 a etc. Is provided. Furthermore, each storage container 6, 6a, etc. is provided with a rising duct 18, 18a, etc. that extends to the bottom of the storage container with which it is originally associated. The individual rising ducts 18, 18 a, etc. are bundled at their ends 20, 20 a, etc. located away from the storage container towards the filling arm 22 belonging to the filling station 24. In addition, each rising duct 18, 18a, etc. includes a controllable valve 16, 16a, etc. The supply pipes 14, 14a, etc. of these storage containers 6, 6a, etc. and the rising ducts 18, 18a, etc. pass through the container lids 26, 26a, which generate a pressure seal. The corresponding valve 16 in the rising duct 18 is opened to move a certain amount of material component contained in the storage container 6 to the receptacle 8. Due to the overpressure in the storage container 6, the material component is pressurized in the riser 18 through the valve 16 and all the way to the filling arm 22.

  Furthermore, the mixing device 4 includes a turntable 28 that is originally arranged horizontally, and is provided with a plurality of openings 30 that are distributed in the entire area in the periphery. As is apparent from FIG. 1, the individual receptacles 8, 8 a, 8 b and 8 c are hung on the turntable 28 in the individual openings 30. For this purpose, the receptacles 8, 8a, 8b, 8c etc. are characterized by an upper collar that exceeds the width of the opening 30, which is used for suspended storage of the receptacles 8, 8a, 8b, 8c, etc. A support surface is created. The turntable 28 can rotate on its central axis A, and in addition can be moved to an upper position by a lower position in a direction along the axis displayed here. The cover disk 32, which is essentially coaxially located on the turntable 28, creates an upper cover for receptacles 8, 8a, 8b, 8c, etc. depending on the turntable 28. However, the cover disk 32 is shown in FIG. 1 with some lift compared to the turntable 28 for clarity.

  In addition, the filling station 24 includes a metering device 34 by which the amount of different substance components sent to the receptacle 8 arranged at the top thereof can be accurately determined. As can be seen from FIG. 1, the filling arm 22, the receptacle 8 to be filled, and the metering device 34 are essentially installed vertically one above the other. The gap 36 in the cover disk 32 leaves a free passage for the receptacle 8 installed on the weighing device 34.

  After the receptacle 8 has been successfully filled, the cover disk 32 and the filling arm 22 are first raised until the receptacle 8 is released. In addition, the turntable 28 is lifted until the receptacle 8 hangs away from it. Thereafter, the turntable 28 is rotated around the central axis A to the extent that the next receptacle 8a to be filled is placed on the weighing device 24.

  Thereafter, the turntable 28 is lowered, so that the receptacle 8a comes upright on the weighing device 24. Finally, the cover disk 32 is pulled down to the top of the receptacles 8, 8 b, 8 c, etc. depending on the turntable 28. Furthermore, the filling arm 22 is pulled down again to the extent that the end portions 20, 20a, etc., such as the rising ducts 18, 18a, etc. are located inside or substantially inside the mouthpiece of the receptacle 8a that can be filled.

  The aforementioned movements of the turntable 28, the filling arm 22 and the cover disk 32 are influenced by a traction group 38, represented by a simplified schematic, controlled by the data processing device 2.

  It is appropriate to provide various interchangeable turntables. The first turntable may be provided, for example, for 64 smaller containers that can be filled up to 20 g, and the second turntable may be provided for 28 larger containers that can be filled up to 100 g. it can.

  The device is advantageously composed of a plurality, for example four filling stations, and a corresponding number of metering devices, filling arms, valve groups, etc., so that filling of a large number of receptacles can be accelerated accordingly. .

  The data processing device 2 features a known functional group such as a computer, a storage medium, a monitor and a computer mouse used for data input, storage, processing and retrieval.

  As will be clarified later on the basis of the relevant procedural steps, the data processing device 2 comprises means for determining an imaging operator, means for inputting an attribute vector, calculation of a modified recipe vector. As well as means for transferring to the mixing device 4 a control signal shaped according to the recipe vector, in particular means for inputting, storing and retrieving at least one composition database Prepare.

  The recipe and sensory data of the fragrance and / or fragrance composition—hereinafter referred to as the “composition” —are managed using the apparatus described above. The user can intentionally go through a database available according to sensory characteristics and have a recipe for the displayed composition. The recipes can be modified utilizing a modification procedure when they do not have the required sensory characteristics. Multiple new recipes are created thereby that must be sufficient for the desired sensory characteristics. Base compositions that exist in the past and their sensory evaluation form the basis of the modification procedure. The investigation and analysis of these data as well as the operating mode of the reforming procedure are further described below.

  In practical use, groups of compositions are considered as so-called “modules”, thereby demonstrating the importance of each module being associated with the desired perfume or fragrance direction. The development of such a module is depicted in FIG. For example, to create a mango fragrance module, the technical literature and database are first reviewed for ingredients already used in mango fragrance or naturally present in the fruit. After successfully selecting, for example, 10 to 20 of such substance components, a group of compositions in which several substance components are mixed with each other is created each time.

  Depending on the field of application, ordinary solvents, carrier materials and the like are also mixed therein in addition to the fragrance components. The substance components are administered using a device as described herein in a reagent glass containing 8 to 15 grams, thereby allowing up to 64 compositions to be produced simultaneously. Each composition produced in this way is characterized by a recipe vector that indicates the proportions of the substance components that are used to create the composition.

For example, creating ₁₄ compositions K ₁ -K ₁₄ from 33 substance components S ₁ -S ₃₃ is expressed in the following algebraic notation:
(K) = R (S) (1)
Thereby, the composition vector (K) is a column vector of size 4, the substance component (S) is a column vector of size 33, and the recipe matrix R is a matrix of 14 rows and 33 columns. The matrix element R _ij represents the ratio of the substance component S _{i in} the composition K _j . The i-th row of the recipe matrix R is also interpreted as a row vector (R _i ) ^T representing the composition of the composition K _i and why (R _i ) ^T is also designated as the recipe vector of the composition K _i. (The row vector is represented as an index T to be “transported” in the notification used here).

  The section of the recipe matrix with percentages in weight percent is hereinafter displayed in Table 1.

Table 1: Recipe matrix extract

  The compositions thus prepared are subsequently evaluated for their sensory attributes. Not only the purification of the composition deemed appropriate, but also the initial disposal of the composition deemed inappropriate can be advantageously carried out at an early stage. For this, the compositions are smelled with respect to their aroma characteristics. A clearly recognizable mango fragrance recipe is purified, produced and tested again. Thereafter, excellent mango aroma is tasted with sugar aqueous solution. A good scented fragrance may not be appropriate and may need to be eliminated.

Quantitative evaluation of sensory attributes is performed on the “best” mango fragrance in the manner described above, which is therefore advantageously realized in a specialized sensor technology laboratory. Associated attribute vectors are created for each composition and correspond to the evaluation results of the individual sensory attributes. Similar to the recipe vector (R _i ) ^T introduced above, the attribute vector of the composition K _i can be created as a row vector (A _m ) ^T , and the attribute vector of the group of compositions is the attribute It can be expressed as a matrix A, so that the matrix element A _im represents a property of the attribute m in the composition K _i . The attribute matrix extracts with characteristic specifications in the norm scale from 0 to 15 are represented in Table 2 below.

Table 2: Extraction of attribute matrix

Factor analysis (also called “principal component analysis (PCA)” in English) is advantageously performed in the sense of data reduction. Attributes are thereby examined for similarities in their structure within the attribute matrix. For each extracted factor, the one attribute that has the highest factor loading on the one hand and that makes the most sense for the fragrance usage on the other hand is selected. Only the selected attributes can be changed intentionally later in the framework of the reforming process. Therefore, attributes that have a positive impact on the fragrances and fragrances that are being considered respectively are selected. For example, it may not be desirable to modify the mango fragrance present later to a stronger moldiness or woodiness. About 10 sensory attributes are actually selected for each module. Therefore, the number of dimensions of the attribute vector (A _m ) ^T is reduced, and the number of vector components is reduced from 50 to 10, for example.

  In addition to the selected attribute vector, the used recipe vector is also stored in the composition database so that the vectors associated with each composition can be retrieved with respect to each other and the composition.

  Individual points are determined in a multi-dimensional “attribute space” by the data collection described above, so that each composition of the module corresponds to a point in the attribute space. This is schematically represented in FIG. 3 for a composition at a level that is fixed by two attributes. The sensory attributes of the prepared composition with respect to their individual attributes should be characterized by the maximum variance possible to intentionally modify the attributes.

The attribute recipe matrix M is created for the relevant modules in view of the modification of the existing composition. The matrix element M _mj indicates how strongly the feature of the attribute k is specified by the proportion of the substance component S _j . These relationships are established, for example, by multiple linear regression and / or neural networks and / or by expert systems. Advantageously, the results obtained by various methods are compared with each other and adjusted if necessary. This process is represented schematically in FIG. In addition, the principle for determining the matrix element M is depicted in FIG. The attribute “flowery in fragrance” must be applied against each of the 14 components of the module against the proportion of the material component “butyl acetate”. The trend line established through linear regression shows that the peach aroma becomes stronger with increasing proportion of butyl acetate.

  The purpose of the procedure described above is to derive material components that have a positive or negative leverage effect on the attribute of interest. A positive leverage effect means that increasing the proportion of the substance component associated with an aroma recipe will enhance the sensory perception of the attribute of interest. On the other hand, a negative leverage effect means that the proportion of the material components involved must be lowered to enhance the sensory perception of the attribute of interest.

The attribute recipe matrix M can be viewed as a matrix representation of operators that achieve conversion from a recipe vector to an attribute vector in at least one local region of the attribute space. This can be formally represented by the following equation:
(A) = M (R) (2)
Here, (A) and (R) are an attribute vector and a recipe vector in the column representation. Equation (2) generates an associated attribute vector (A _i ) for the already characterized composition K _{i with} an associated recipe vector (R _i ). The required locality of the transformation means that equation (2) also applies to a slightly modified composition K _i ′ as compared to the composition K _i as a good approximation. In other words, the composition which is to the application of calculation rules for (R _i) slightly recipe is modified vector as compared with _{(R i) '(2)} , modified as good approximation This means that a modified modified attribute vector corresponding to the attribute vector (A _i ) ′ of the object K _i ′ is generated. When a new composition with specified attribute characteristics is desired, in principle it is necessary to solve the problem which is the reverse of the procedure just described. It is a pre-specified desired attribute vector (A _soll ), and an associated recipe vector (R _soll ) that can be considered as a manufacturing specification for the desired composition. Advantageously, the above-characterized compositions whose attribute properties are as close as possible to those desired are used for this search. In other words, modification of composition _{K ist} already characterized the recipe base vector _{R ist} an attribute vector _{A ist} is performed.

The reforming process can be realized as a so-called “trial and error” procedure in the simplest case. That is, for example, a plurality of new recipe vectors can be generated by utilizing a random generator in the environment of ( _Rist ), and the related attribute vectors can be calculated by using equation (2). This will have to be repeated until an attribute vector is found that is sufficiently close to the desired attribute vector (A _soll ). However, it may be more useful for the modification procedure to use a statistical test design. Possible procedures are further clarified below.

  Typically, 4 to 10 related substance components are determined for each attribute. They are divided into categories 1 to 3 according to their positive and negative leverage effects, so category 1 corresponds to the maximum leverage effect and category 3 corresponds to the minimum leverage effect. Multiple substance components can be assigned to category 3, while only one positive substance component and one negative substance component are assigned to categories 1 and 2 each time.

  Table 3 shows an extract of the data evaluation results from a module for apple aroma, for example. The attributes “taste is fresh” and “fragrance is like a flower” and the material components derived for this purpose are listed, for which the material components are listed by code.

  If the substance component C335 is added to an apple recipe, for example, or the substance component C8 is reduced therein, the recipe should be more flowery than in the basic recipe.

Table 3: Extracts of module data evaluation for apple aroma

  As described above, a predetermined recipe can be selected and later modified to intentionally enhance the characteristics of specific attributes. Thereby, for example, up to 14 new recipes are generated according to a specially developed test plan. The test plan addresses substance components originating from data evaluations that are assigned or reduced to basic recipes. The test planning process is depicted in Table 4. The concentration of the substance component with a positive leverage effect is increased in the first four recipes, and the concentration of the substance component with a negative leverage effect is reduced in recipes 5-8. Only the substance components contained in the basic recipe can be reduced.

  In other words, recipe number 6 is not generated if the basic prescription does not include a category 2 negative leveraged substance component. On the other hand, material components with a positive leverage effect are always added. Recipes 9 through 14 are created by a combination of the first eight actions.

Table 4: Statistical test plan

  Accordingly, the test plan indicates which material components must be increased or reduced. However, the absolute measure of these changes must be calculated separately for each material component and is realized as follows.

First, the difference between the maximum usage rate R _max and the minimum usage rate R _min is determined for each substance component across all recipes in the module under consideration.

The result is divided by the difference between the attribute maximum feature A _max and the minimum feature A _min through all recipes of the module under consideration.
_{X 1 = (R max -R min} ) / (A max -A min) (3)

The ratio R _ist of the material components used in the basic recipe, is divided by the sensory evaluation of the attribute of the selected recipe by the additional step.
_{X 2 = R ist / A ist} (4)

Since _Rist is then equal to zero, this calculation is not performed when the substance component is not available in the basic recipe.

The average value of both values calculated above is determined in the following steps.
X = (X ₁ + X ₂ ) / 2 (5)

This calculation is not performed when the second calculation is omitted. In this case, it is x = _{x 1.}

The last step is the result of Equation (5) consists of multiplying the difference between the attribute properties A _soll and the sensory evaluation A _ist is the desired the attribute of the selected recipe.
ΔR = X (A _soll -A _ist ) (6)

  The result ΔR is here the proportion of the substance component to be considered that is added to the basic recipe or reduced in the basic recipe. If ΔR is negative and the amount of ΔR exceeds the proportion Rist of the material component in the basic recipe, this material component is removed from the recipe.

  The calculations are designed for attribute enhancement, but can be used to weaken attributes as well.

  As a visual example of individual composition attribute vectors, they are depicted in FIG. 6 in the form of a polar chart (also called “cobweb”). Associated with each attribute is a beam section that originates from the mutual center and is a measure of the feature of the attribute whose length is related. The beam is thereby placed at an equal distance and diagonally. The attribute “citrus” has the weakest impression, and the attributes “vanilla” and “fresh” have the strongest impression of the compositions shown. Such a representation is well suited for interactive modification of the composition. The polar chart displayed on the monitor can be modified, for example, intentionally by a computer mouse, thus giving a modified attribute vector to the desired new composition.

1 is a schematic view of an apparatus for producing a fragrance and / or fragrance composition. It is the schematic of the procedure at the time of developing a module. FIG. 6 is a schematic diagram of six compositions in a plane spanning two attributes a and b. FIG. 2 is a schematic diagram of determining a relationship between a proportion of substance components and an attribute characteristic of a composition. It is a graphic diagram of the relationship between the specification of “peach aroma”, the ratio of the substance component “butyl acetate”, and the associated regression line. It is a figure of the attribute vector of the composition as a polar chart.

Claims (9)

To produce a fragrance and / or fragrance composition characterized by using a composition database composed of recipe vectors and attribute vectors associated with a group of base compositions that can be produced by mixing predetermined substance components Wherein each recipe vector specifies the proportion of the substance component required to create the associated base composition, whereby each attribute vector is selected of the associated base composition. Specify evaluation results for sensory attributes,
a) specifying a target attribute vector;
b) determining an operator that achieves a conversion from the recipe vector to the attribute vector at least in the environment of the target attribute vector;
c) establishing a target recipe vector on the condition that by using said operator it is converted to said target attribute vector;
d) mixing the predetermined substance components at a rate according to the target recipe vector;
A method for producing a fragrance and / or fragrance composition comprising: The composition database is
a) creating the group of base compositions by mixing the substance components in proportions according to a recipe vector associated with each base composition;
b) quantitatively evaluating each one of the base compositions with respect to creating the selected sensory attribute and the associated attribute vector;
c) creating the composition database by storing the recipe vector and attribute vector such that the vectors associated with each base composition can be retrieved with respect to each other and with respect to the base composition;
The method of claim 1 characterized by the fact that it is created by:   3. A method according to claim 1 or 2, characterized by the fact that the evaluation of the sensory attributes is based on quantitative description analysis.   4. A method according to any one of claims 1 to 3, characterized by the fact that the attributes used to create the attribute vector are selected by factor analysis.   5. A method according to any one of claims 1 to 4, characterized by the fact that the operator is established by multiple regression and / or neural networks and / or expert systems.   6. A method according to any one of the preceding claims, characterized by the fact that the target recipe vector is determined by a statistical test plan.   7. A method according to any one of the preceding claims, characterized by the fact that the target attribute vector is determined through the attribute evaluation of a given composition.   8. A method according to any one of the preceding claims, characterized by the fact that the attribute vector is represented in the form of a polar diagram. Not only the data processing device (2) but also the mixing device (4) controlled by it is provided, so that the data processing device (2) can provide means for inputting attribute vectors and the calculation of target recipe vectors. Means for input, storage and retrieval of at least one composition database as well as means for transferring control signals shaped according to the recipe vector to the mixing device (4) To that end, the mixing device (4) has the following components:
a) a plurality of storage containers (6, 6a) that can be filled with individual substance components;
b) a plurality of receptacles (8, 8a, 8b, 8c);
c) Controllable to bring a predetermined amount of individual substance components from the corresponding storage container (6, 6a) to the receptacle (8, 8a, 8b, 8c) for the production of a perfume and / or fragrance composition A supply device (10);
9. An apparatus according to any one of claims 1 to 8 for carrying out a method characterized by.

Images