Application of a full-coherent artificial neural network for forecasting of the modes of storage of domestic low-olive raw materials in controlled environments

Researches on increase in an expiration date of the wheat germs (WG) with use of compositions of organic acids are conducted. With a research objective of influence of concentration of mixes of organic acids on change of indicators of quality at storage of the SALARY in various modes investigated quality indicators in the range of concentration of 1-7% to the mass of a product. As control the raw SALARIES served. Skilled products stored in refrigerator conditions (temperature 4-6 ºС, relative humidity of air of 75-80%) and a warehouse (temperature 20-22 ºС, relative humidity of air of 70-80%). The software product on the basis of the program of training and the analysis of training of an artificial full-coherent neural network (INS) in the Python 2.7 language with program libraries of mathematical processing of scientific data of "scipy" is developed. As input parameters of a neural network were considered: humidity of wheaten germs (х 1, %), relative humidity of air (х 2, %), ambient temperature (х 3, ºС) and concentration of mix of organic acids (х 4, %). By means of the software, some neural networks were designed and trained. For modeling the network with two layers was used. Applying the developed and trained neural network it is possible constructed dependence у(х 1, х 2, х 3, х 4). For visualization in three-dimensional space limited amount of arguments of function by two. Results of work of neural networks y (x 1, x 4) with the recorded entrance parameters (x 2 = 60, %, x 3=20, ºC) and a neural network y (x 2, x 3) with the recorded input parameters are presented (x 1 = 15%, x 4 = 5%). The received mathematical model which on the set set of certain parameters of storage, allows to receive concrete value of output parameter and to plan the storage modes in controlled environments.