Method for predicting quality or manufacturing condition of cement

Patent metadata

Field	Value
Publication number	US-9679244-B2
Application number	US-201314403753-A
Country	US
Kind code	B2
Filing date	Feb 22, 2013
Priority date	May 30, 2012
Publication date	Jun 13, 2017
Grant date	Jun 13, 2017

Abstract

Provided is a method capable of predicting the quality of cement in a short time period and with high accuracy. The method of predicting the quality or manufacturing conditions of cement through use of a neural network including an input layer and an output layer includes: performing learning of the neural network for a sufficiently large number of times of learning such that σ L <σ M is obtained, using learning data and monitor data; then repeating the learning of the neural network until σ L ≧σ M is obtained while the number of times of learning is decreased; inputting specific observation data to the input layer of the neural network in which a judgment value for analysis degree obtained from the neural network after the learning is less than a preset value; and outputting an estimated value of specific evaluation data from the output layer of the neural network.

First claim

The invention claimed is:

1. A method of predicting quality or manufacturing conditions of
cement through use of a neural network including an input layer and an output layer,
the input layer being used for inputting an actually measured value of observation data in cement manufacturing, and the output layer being used for outputting an estimated value of evaluation data related to evaluation of the quality or the manufacturing conditions of the cement,
the observation data and the evaluation data being used in one of the following combinations:
(i) a combination in which the observation data comprises one or more kinds of data selected from data on a clinker raw material, data on burning conditions, data on grinding conditions, and data on clinker, and the evaluation data comprises data on a clinker raw material, data on burning conditions, data on grinding conditions, data on clinker, or data on cement; and
(ii) a combination in which the observation data comprises one or more kinds of data selected from data on a clinker raw material, data on burning conditions, data on grinding conditions, data on clinker, and data on cement, and the evaluation data comprises data on physical properties of a cement-containing hydraulic composite,
the method comprising the steps of:
(A) performing initial setting of a number of times of learning;
(B) performing learning of the neural network for the set number of times of learning through use of a plurality of learning data each comprising a combination of an actually measured value of the observation data and an actually measured value of the evaluation data;
(C) calculating a mean square error (σ L ) between an estimated value of the evaluation data obtained by inputting an actually measured value of the observation data of the plurality of learning data to the input layer of the neural network in which learning has been performed in the latest step (B) and an actually measured value of the evaluation data of the plurality of learning data, and a mean square error (σ M ) between an estimated value of the evaluation data obtained by inputting an actually measured value of the observation data in monitor data, which comprise a combination of an actually measured value of the observation data and an actually measured value of the evaluation data and which is used for confirming reliability of a learning result of the neural network, to the input layer of the neural network in which learning has been performed in the latest step (B) and an actually measured value of the evaluation data in the monitor data, performing a step (D) when the calculated σ L and am satisfy a relationship of σ L ≧σ M , and performing a step (E) when the calculated σ L and σ M satisfy relationship of σ L <σ M ;
(D) increasing the set number of times of learning to reset the increased set number of times of learning as a new number of times of learning, and performing the steps (B) and (C) again;
(E) resetting a number of times of learning obtained by reducing the number of times of learning for which the latest learning of the neural network has been performed as a new number of times of learning;
(F) performing the learning of the neural network for the set number of times of learning through use of the plurality of learning data used in the step (B);
(G) calculating a mean square error (σ L ) between an estimated value of the evaluation data obtained by inputting an actually measured value of the observation data of the plurality of learning data to the input layer of the neural network in which learning has been performed in the latest step (F) and an actually measured value of the evaluation data of the plurality of learning data and a mean square error (σ M ) between an estimated value of the evaluation data obtained by inputting an actually measured value of the observation data in the monitor data to the input layer of the neural network in which learning has been performed in the latest step (F) and an actually measured value of the evaluation data in the monitor data, performing a step (I) when the calculated σ L and σ M satisfy a relationship of σ L ≧σ M , and performing a step (H) when the calculated σ L and σ M satisfy a relationship of σ L <σ M ;
(H) performing the steps (E) to (G) again when the number of times of learning of the neural network in the step (F) performed most recently is more than a preset numerical value, and performing a step (K) when the number of times of learning of the neural network in the step (F) performed most recently is equal to or less than the preset numerical value;
(I) calculating a judgment value for analysis degree by the following equation (1), and when the analysis degree determination value is less than a preset value, inputting an actually measured value of the observation data in the cement manufacturing to the input layer and outputting an estimated value of the evaluation data related to the evaluation of the quality or the manufacturing conditions of the cement from the output layer, and when the analysis degree determination value is equal to or more than the preset value, performing the step (K); and
(K) initializing learning conditions, and performing the steps (A) to (K) again:

Analysis
⁢

⁢
degree
⁢

⁢
determination
⁢

⁢
value
⁢

⁢

(
%
)

=

Mean
⁢

⁢
square
⁢

⁢
error
⁢

⁢

(

σ
L

)

of
⁢

⁢
learning
⁢

⁢

Assignees

• Taiheiyo Cement Corp

Inventors

• Ohno Maiko
• Kurokawa Daisuke
• Hirao Hiroshi

Classifications

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• C04B7/361 Primary

  Condition or time responsive control in hydraulic cement manufacturing processes (controlling or regulating in general G05; F27B7/42 takes precedence) · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title