JP2005097658A - Method for predicting main raw material component ratio of sintered ore and method for controlling component ratio of sintered ore and program for predicting main raw material component ratio of sintered ore - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for predicting the main raw material component ratios of sintered ore in the case the characteristics of compounded raw materials change dynamically and a method for controlling the main raw material component ratios of the sintered ore using the same, and a program for predicting the main raw material component ratios of the sintered ore. <P>SOLUTION: The method for predicting the main raw material component ratios of the sintered ore has (1) a data-for-identification collecting step of collecting the time series data for identification of the main raw material component ratios of the sintered ore obtained from the component analysis values of the sintered ore and the actually measured value of the component ratios of the auxiliary raw materials of the sintered ore, (2) a step for generating autoregressive model groups corresponding to frequencies of generating the respective autoregressive models by classifying respective pieces of the data to time series data of a plurality of frequency bands, (3) a frequency band judging step of performing the frequency analysis of the most recent component analysis values of the sintered ore thereby judging the strong frequency bands of power spectra, (4) an autoregressive model selecting step of selecting the autoregressive models corresponding to the judged frequency bands from the model groups, and (5) a step of predicting the component ratios of the main raw materials of the sintered ore based on the selected models. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a time-series data prediction method, and more particularly, to a sintered ore main raw material component proportion prediction method and a sintered ore component proportion control method.

  The main raw materials for sintering materials are fine ore generated during pretreatment of blast furnace charging ore, iron ore existing as fine ore, and iron-containing raw materials (mill scale, blast furnace dust, converter dust, etc.) generated in steelworks. It is made of ironmaking materials such as powdered limestone as an auxiliary material. In addition, recently, hot cracked ore or adhesive ore, which is not preferred for direct use in blast furnace operation, is sometimes crushed into a sintered raw material.

  As a conventional technique relating to a method for predicting sinter time series data, there is a technique disclosed in Japanese Patent Application Laid-Open No. 59-64718 (Patent Document 1) relating to a method for predicting sinter basicity. This sinter basicity prediction method measures the difference between the measured value and the calculated value of sinter basicity in time series, statistically processes the past information of this difference to determine the deviation, The basicity of the sinter is predicted from the basicity obtained from the calculated value.

Paying attention to the fact that the deviation between the calculated basic value and the measured value of sintered ore is correlated with the past deviation, measure this deviation in time series, and use the statistical method to predict the deviation. Determine based on data. The calculated value of basicity (CaO / SiO ₂ ) is calculated at the time of cutting out the raw material by the ratio of the total CaO amount of the blended raw material and the total SiO ₂ amount. Then, by estimating the deviation from the data of the sinter basicity for the blended raw material cut out several hours or more ago at the present time, the basicity after sintering the blended raw material to be sintered at the present time Can be estimated.
JP 59-64718

  However, in the technique disclosed in Patent Document 1, since the prediction deviation is obtained using one regression equation, when the characteristics of the blended material change dynamically, the regression coefficient must be obtained each time. However, there is a problem that it is difficult to always obtain a highly accurate predicted value in actual operation.

  Even if another regression equation prepared in advance is used according to the change in the characteristics of the blended raw material, no selection method is presented.

  The present invention has been made to solve the above-described problems of the prior art, and is a main raw material for sintered ore that can be predicted with high accuracy even when the characteristics of the blended raw material change dynamically. It is an object of the present invention to provide a component ratio prediction method, a sinter component ratio control method, and a sinter main raw material component ratio prediction program.

The present invention is a method for predicting the ratio of main raw material components of sintered ore that varies in time series, and has the following steps (1) to (6): It is a percentage prediction method.
(1) An identification data collection step for collecting time-series data for identification of the sinter ore main raw material component ratio obtained from the component analysis value of the sinter ore and the actual measured value of the component ratio of the secondary raw material of the sinter ore.
(2) A frequency-corresponding autoregressive model group creation step of classifying the collected time series data for identification into time series data of a plurality of frequency bands and creating an autoregressive model corresponding to each.
(3) A frequency band determination step of performing frequency analysis on the time series data of component analysis values of the nearest sintered ore and determining a frequency band having a strong power spectrum.
(4) An autoregressive model selection step of selecting an autoregressive model corresponding to the determined frequency band having a strong power spectrum from the group of frequency-corresponding autoregressive models.
(5) A step of predicting a sinter ore main material component ratio that predicts a component ratio of the sinter ore main material at the time of adding the auxiliary material based on the selected autoregressive model.

  Moreover, this invention is a sintered ore main raw material component ratio prediction method of Claim 1, Comprising: Wavelet transformation is used for the frequency analysis in the said frequency band determination process, The sintered ore main raw material component ratio prediction Is the method.

  Further, the present invention provides the sintered ore main raw material component ratio prediction method according to claim 1 or 2, wherein the determination of a frequency band having a strong power spectrum in the frequency band determination step has a peak power spectrum. The preset power spectrum threshold was exceeded for the method of setting the frequency band by setting the width on the high frequency side and the low frequency side of the frequency to become, or the normalized power spectrum normalized by the peak of the power spectrum It is a sintered ore main raw material component ratio prediction method characterized by being based on any one of the methods for making a frequency band.

  Moreover, this invention calculates | requires the variation | change_quantity of the sinter main raw material component ratio using the sintered ore main raw material component ratio prediction method in any one of Claims 1 thru | or 3, and based on this variation | change_quantity. It is a sinter component ratio control method characterized by controlling the addition ratio of the auxiliary raw material of sinter.

Furthermore, the present invention is a program for predicting the ratio of main raw material components of sintered ore that varies in time series, and the main raw material of sintered ore for realizing the following functions (1) to (6) on a computer: This is a component ratio prediction program.
(1) An identification data collection function for collecting time-series data for identification of the sinter ore main raw material component ratio obtained from the component analysis value of the sinter ore and the measured value of the component ratio of the auxiliary raw material of the sinter ore.
(2) A frequency-corresponding autoregressive model group creation function that classifies collected time series data for identification into time series data of a plurality of frequency bands and creates an autoregressive model corresponding to each.
(3) A frequency band determination function for performing frequency analysis on the time series data of component analysis values of the nearest sintered ore and determining a frequency band with a strong power spectrum.
(4) An autoregressive model selection function for selecting an autoregressive model corresponding to a determined frequency band having a strong power spectrum from the group of frequency-corresponding autoregressive models.
(5) A function for predicting a sinter ore main material component ratio that predicts a component ratio of the sinter ore main material at the time of adding the auxiliary material based on the selected autoregressive model.

  According to the present invention, an appropriate model can be automatically selected according to a change in dynamic characteristics of a blended raw material, so that a highly accurate predicted value can be obtained continuously in actual operation. Furthermore, since the control is performed using an accurate predicted value of the fluctuation of the controlled object having a long dead time, there is an effect that a short-term fluctuation that cannot be removed by simple feedback control can be removed.

  The best mode for carrying out the present invention will be described below with reference to the drawings and mathematical expressions.

  FIG. 2 is a diagram schematically showing a manufacturing process of sintered ore. The auxiliary raw material 3 is added to the main raw material 1 in the blending section 2. And the sintered ore 5 is manufactured by passing through the sintering process 4 of them. The sintered ore 5 sintered in the sintering step 4 is periodically sampled and component analysis is performed in the component analysis step 6. This analysis result, that is, the component analysis value 7 is grasped as intermittent time series data.

  However, the component analysis value of the sintered ore is found to have a large time difference from several hours after the addition of the auxiliary raw material 3 to the main raw material 1. For example, 1 hour for sintering, 1 hour for cooling, 1 hour for sampling, 1 hour for analysis, about 4 hours in total. For this reason, the component analysis value of the sintered ore that came out reflects the blending several hours ago. Even if this analysis value is immediately fed back to the addition amount control of the auxiliary raw material 3, the main raw material is already at that time. The component ratio of is changing. Thus, it is very difficult to keep the component analysis value of sintered ore with a long time lag constant.

  Since the component value of the auxiliary raw material is accurately grasped in advance, the component fluctuation value of the main raw material in the past can be calculated by subtracting the component value of the auxiliary raw material added from the component analysis value of the sintered ore as follows ( 1) It can obtain | require like Formula.

  Therefore, if the component fluctuation value after several hours of the main raw material is predicted when the component analysis value is known, this represents the latest main raw material component fluctuation (at the time of blending). If the raw material blending amount is determined, the components of the sintered ore can be controlled with higher accuracy. In the present invention, the fluctuation value of the main raw material component is set as the prediction target data 8.

FIG. 1 is a prediction flow chart for predicting a fluctuation value of a sintered ore main raw material component. First, the time series data of the sinter ore main material component ratio as shown in the equation (1) is used for component value identification from the component analysis value of the sinter ore and the measured value of the component ratio of the auxiliary material added to the main material. Collect as data 11. An example of the component value identification data 11 is shown in FIG. 3. This component value identification data 11 is identified 13 through a filter 12 such as an FFT to classify it into time-series data of a plurality of frequency bands. An autoregressive model group 14i is created.

A series of preparation processes shown on the left side of FIG. 1 will be described in detail below.
The time-series data for component value identification shown in FIG. 3 is converted into frequency data (see FIG. 4) by Fourier transform expressed by equation (2).

  Next, a plurality of characteristic frequency bands are left in the frequency data of FIG. FIG. 5 shows that the low frequency band is left, and the power spectrum in the high frequency band is masked. FIG. 6 shows that the intermediate frequency band having the strongest power spectrum is left in the original frequency data, and the low and high frequency band regions outside the target band are respectively masked. Such processing is appropriately performed a plurality of times to leave a characteristic frequency band.

  The frequency data subjected to these masks is subjected to Fourier inverse transform represented by equation (3), and returned to time series data.

  7 and 8 are time series data obtained by inversely converting the frequency data shown in FIGS. 5 and 6, respectively. Compared to the original time-series data in FIG. 3, it can be seen that FIG. 7 is time-series data in which the low frequency band excluding the high frequency band is emphasized. Similarly, it can be seen that FIG. 8 is time-series data in which the intermediate frequency band is emphasized.

  Next, the model is identified by applying the autoregressive model (AR model) represented by the equation (4) to each of the data thus classified for each frequency band.

The n frequency-corresponding AR models indicated as 14i in FIG. 1 have been prepared.

  Next, the processing flow on the right side of FIG. 1 will be described. Prepare the latest time series data (component value nearest data 15) of the component analysis value of the sintered ore. The component value nearest data 15 is subjected to wavelet transformation expressed by the equation (5) (wavelet transformation 16).

  A frequency band having a strong power spectrum subjected to the wavelet transform is determined, and an autoregressive model 18 corresponding to the determined frequency band is selected from a plurality of frequency-corresponding AR models 14i (model selection 17). That is, the frequency band of the nearest data is always monitored, and this change is captured (determined), and a model corresponding to the frequency band is selected.

  As a method for determining a frequency band with a strong power spectrum, for example, there is a method shown in FIG. 9 or FIG. In the method shown in FIG. 9, the upper and lower set widths are provided on the high frequency side and the low frequency side of the frequency f2 at which the power spectrum reaches a peak, and the frequency band (f1 to f3) is obtained. On the other hand, the method shown in FIG. 10 uses a frequency band (f1 to f3) that exceeds a preset power spectrum threshold with respect to the normalized power spectrum normalized at the peak of the power spectrum.

  Based on the autoregressive model (AR model 18) of the selected time series data, the predicted value 19 of the component ratio of the sintered ore main raw material at the time of adding the auxiliary raw material is predicted. When there are a plurality of selected AR models 18, the predicted values using the respective AR models are added to obtain the final predicted value of the main raw material fluctuation amount. In addition, the arrow from the component value closest data 15 to the AR model 18 in FIG. 1 indicates the AR identified from the raw data (component value closest data 15) including all frequency bands as the final predicted value of the main raw material fluctuation amount. Another prediction method for calculating the predicted value of the main raw material fluctuation amount is shown, in which the predicted value of the selected AR model is added as a compensation amount based on the predicted value of the model.

FIG. 11 is a graph comparing the actual value of the fluctuation of the SiO ₂ component ratio of the sintered ore component with the predicted value based on the present invention. According to this, the predicted value based on the present invention can predict the time-series fluctuation of the actual value well, and even when the characteristics of the plant change dynamically, the method for predicting the ratio of main raw material components of the sinter according to the present invention It can be seen that is effective.

It is a prediction flow figure for estimating the fluctuation value of a sinter ore main raw material component. It is a figure which shows typically the manufacturing process of a sintered ore. It is a figure which shows an example of the data 11 for component value identification. It is a figure which shows the result of having carried out the Fourier-transform of the time series data of FIG. It is a figure explaining the masking of a specific frequency band. It is a figure explaining other masking of a specific frequency band. It is a figure which shows the time series data which carried out reverse conversion of FIG. It is a figure which shows the time series data which carried out reverse conversion of FIG. It is a figure which shows an example of the determination method of a frequency band with a strong power spectrum. It is a figure which shows another example of the determination method of a frequency band with a strong power spectrum. . And actual value of the SiO ₂ component ratio of the variation of the sintered ore component is a graph comparing the predicted value based on the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Main raw material 2 Compounding part 3 Sub raw material 4 Sintering process 5 Sintered ore 6 Component analysis process 11 Data for component value identification of sinter main raw material 12 Filter 13 Identification 14i Autoregressive model group 15 Component analysis value of sintered ore Time series data 16 Wavelet transform 17 Model selection 18 AR model 19 Predicted component ratio of sintered ore main raw material at the time of adding auxiliary raw material

Claims (5)

A method for predicting a main raw material component ratio of a sintered ore that varies in a time series, the method comprising the following steps (1) to (6):
(1) An identification data collection step for collecting time-series data for identification of the sinter ore main raw material component ratio obtained from the component analysis value of the sinter ore and the actual measured value of the component ratio of the secondary raw material of the sinter ore.
(2) A frequency-corresponding autoregressive model group creation step of classifying the collected time series data for identification into time series data of a plurality of frequency bands and creating an autoregressive model corresponding to each.
(3) A frequency band determination step of performing frequency analysis on the time series data of component analysis values of the nearest sintered ore and determining a frequency band having a strong power spectrum.
(4) An autoregressive model selection step of selecting an autoregressive model corresponding to the determined frequency band having a strong power spectrum from the group of frequency-corresponding autoregressive models.
(5) A step of predicting a sinter ore main material component ratio that predicts a component ratio of the sinter ore main material at the time of adding the auxiliary material based on the selected autoregressive model. The sintered ore main raw material component ratio prediction method according to claim 1, wherein wavelet transform is used for frequency analysis in the frequency band determination step. The sintered ore main raw material component ratio prediction method according to claim 1 or 2, wherein the determination of the frequency band having a strong power spectrum in the frequency band determination step is performed at a high frequency at a frequency at which the power spectrum reaches a peak. A method of providing a frequency band with a set width on each of the side and the low frequency side, or a method of setting a frequency band that exceeds a preset power spectrum threshold with respect to a normalized power spectrum normalized by the peak of the power spectrum A method for predicting the ratio of raw material components of sintered ore, characterized by being based on any of the above. A variation amount of the sintered ore main raw material component ratio is obtained using the method for predicting a sintered ore main raw material component proportion according to any one of claims 1 to 3, and the secondary amount of the sintered ore is calculated based on the variation amount. A method for controlling a sinter component ratio, characterized by controlling an addition ratio of a raw material. A program for predicting a ratio of main raw material components of a sintered ore that varies in time series, and a program for predicting a ratio of main raw material components of a sintered ore for causing a computer to realize the following functions (1) to (6).
(1) An identification data collection function for collecting time-series data for identification of the sinter ore main raw material component ratio obtained from the component analysis value of the sinter ore and the measured value of the component ratio of the auxiliary raw material of the sinter ore.
(2) A frequency-corresponding autoregressive model group creation function that classifies collected time series data for identification into time series data of a plurality of frequency bands and creates an autoregressive model corresponding to each.
(3) A frequency band determination function for performing frequency analysis on the time series data of component analysis values of the nearest sintered ore and determining a frequency band with a strong power spectrum.
(4) An autoregressive model selection function for selecting an autoregressive model corresponding to a determined frequency band having a strong power spectrum from the group of frequency-corresponding autoregressive models.
(5) A function for predicting a sinter ore main material component ratio that predicts a component ratio of the sinter ore main material at the time of adding the auxiliary material based on the selected autoregressive model.

Images