Optimization of formulations and design of experiments
Links between instrumental analysis and sensory analysis
This set of operations allows to process instrumental data (production criteria, physical & chemical analysis, ...) to meet various objectives such as:
. The product study by the mean of their instrumental analysis
. The operation control management by linking data production and sensory non-conformities
. The monitoring of the productions with the identification of key factors changes over time
. The development of new products with links between the product formulation and their sensory descriptions.
The system proposes the creation of data entry grids or data import through Excel©, then different and various calculations are suggested as:
. Description of data with different distribution stats with mean, standard deviation, confidence interval,
. analysis of the data structure with the analysis of variance
. multivariate analysis with Principal Component Analysis of averaged data, data or raw data at different dates or different classifications.
Tastel+ is also capable to involve instrumental data or formulation with corresponding sensory descriptive data, the joint processing of data will then identify links between these two groups of data. It will be possible by introducing a new sensory profile to know the conditions of its development.
The idea is to know what would be the optimal formulation of components to enable the development of a product which is known also its sensory characteristics.
A very simple example with 4 cookies to improve illustrates these links:
1. Import formulation data
2. Links between the formulation data and corresponding sensory profiles
3. Modeling an ideal product on the basis of a desired target sensory (from, eg consumer tests or merger with a competitor)
Importing formulation data in Tastel:
Links between the formulation data and corresponding sensory profiles:
The system will then model a new product from correlations between the two types of data so that the user can then look at the elements of this formulation to develop this new product matching these sensory expectations.
Other links can also exist between sensory description and production or instrumental analysis, easily highlighted by Tastel+ :
Mix Plan Designs
Mix plan designs meet the needs of optimization techniques for formulating the mix in relation to a given answer: stability, viscosity, texture, preference ... or at a price optimization by a different combination of less expensive components.
Protocols are proposed by the system according to the type of mixture (I, II, III, IV) and the number of elements in the mix with their levels in order to issue optimum formulation graphs.
The components entering in mixt with their range of variation can determine automatically the appropriate protocol, but also the type of mix (Type I, II, III or IV). Thus, it is considered an example with a mixture of type I with 3 components.
Protocols may be proposed for the mixture of 2 to 4 elements.
Following the data entry of the experimental data, Tastel+ will test the validity of models and display results using response surfaces depending on the model: linear, quadratic, other ... Complete results are then published stating all information before the models tested, the validation of these tests, then the model coefficients in case of accepted model.
Optimization of factors
Optimizations of factors are dealing with critical factor formulation for a 'goal' reply: target stability, sensory description, or yield, to identify what those factors are important, and therefore how to work the answer the most efficiently.
This module can easily return to the factors influencing an experimental result in order to propose a protocol of experiments, and thus to identify laws causing the process to be optimized.
Then, the suggested protocols are automatically available in order to prepare experiments: a model three complete factor model is displayed below.
Once the requested data entry done, the system will be able to test the difference between the experimental result and the calculated result of the modeling to test the validity of the model under different laws possible (linear or quadratic).
These treatments are then used to represent the curves of iso-response plans for the different considered factors, additional results are also published in order to return among other information, the model coefficients when this one is accepted.