FIELD: fire safety.
SUBSTANCE: invention relates to fire and industrial safety (development of methods and ways for investigation of explosive and fire hazard properties of substances and materials) and can be used to determine group of explosive mixture for selection of type of explosion-proof electrical equipment, in development of measures to ensure fire-explosion safety of technological processes in various industries. Essence: the main elements of the proposed method are molecular descriptors and artificial neural networks (ANN) "back-propagation" taking into account use of the method of improving gradient descent, in which the moment m is introduced, when the effect of the gradient on the change of weight varies with time. Technical task of the invention is determination of fire hazardous parameters, in particular self-ignition temperature, anthraquinone and dyes based on it. Set task is achieved by the fact that in rapid method for prediction of fire-hazardous properties of anthraquinone and dyes based on it, in particular self-ignition temperature, which includes formation of a database of molecular descriptors consisting of basic physical and chemical properties of substances under consideration, novel is that the applied back propagation network employs a method of improved gradient descent. Such method is introduction of moment m, when effect of gradient on change of weights varies with time. An additional advantage of moment introduction is the ability of the algorithm to overcome small local minima. Besides, applying "back-propagation" with the improved gradient descent the extremely effective method of finding the gradient of the error function is obtained.
EFFECT: possibility of determining fire-hazardous properties, in particular self-ignition temperature, and high accuracy of analysis.
1 cl, 7 tbl, 1 dwg, 5 ex
Title | Year | Author | Number |
---|---|---|---|
EXPRESS-METHOD FOR PREDICTING FIRE-HAZARDOUS PROPERTIES OF LIMITING ALDEHYDES WITH USE OF MOLECULAR DESCRIPTORS AND ARTIFICIAL NEURAL NETWORKS | 2018 |
|
RU2672268C1 |
EXPRESS-METHOD FOR PREDICTING FIRE-HAZARDOUS PROPERTIES OF LIMITING KETONS WITH THE USE OF MOLECULAR DESCRIPTORS AND ARTIFICIAL NEURAL NETWORKS | 2017 |
|
RU2662716C1 |
EXPRESS-METHOD FOR FORECASTING FIRE-PROOF PROPERTIES OF COMPOUNDS OF OIL AND PROPIONIC ACIDS WITH USE OF MOLECULAR DESCRIPTERS AND ARTIFICIAL NEURAL NETWORKS | 2016 |
|
RU2621669C1 |
NEURAL NETWORK TRANSFER OF THE FACIAL EXPRESSION AND POSITION OF THE HEAD USING HIDDEN POSITION DESCRIPTORS | 2020 |
|
RU2755396C1 |
DIAGNOSTIC TECHNIQUE FOR ACUTE CORONARY SYNDROME | 2020 |
|
RU2733077C1 |
METHOD FOR VISUALIZING A 3D PORTRAIT OF A PERSON WITH ALTERED LIGHTING AND A COMPUTING DEVICE FOR IT | 2021 |
|
RU2757563C1 |
METHOD AND A SYSTEM FOR PREDICTING TIME SERIES VALUES USING AN ARTIFICIAL NEURAL NETWORK | 2019 |
|
RU2744041C1 |
NEURAL DOT GRAPHIC | 2019 |
|
RU2729166C1 |
RAPID TWO-LAYER NEURAL NETWORK SYNTHESIS OF REALISTIC IMAGES OF A NEURAL AVATAR BASED ON A SINGLE IMAGE | 2020 |
|
RU2764144C1 |
METHOD AND SYSTEM FOR OPTIMIZING LABORATORY ANALYSIS OF ROCK SAMPLES | 2019 |
|
RU2725506C1 |
Authors
Dates
2019-06-24—Published
2018-05-30—Filed