Method for predicting f-CaO of clinker in real time based on data of full-automatic free calcium oxide analyzer
1. The method for predicting f-CaO of clinker in real time based on data of a full-automatic free calcium oxide analyzer is characterized by comprising the following steps of:
acquiring field data, acquiring production data in a field control system DCS and detection data of a full-automatic free calcium oxide analyzer, and processing the data;
selecting input variables related to f-CaO of the clinker as auxiliary variables for predicting f-CaO of the clinker;
step three, constructing a clinker f-CaO prediction model according to and in combination with the selected variables;
and step four, in the same time sequence, acquiring f-CaO data of the clinker by using a full-automatic free calcium oxide detector, then performing contrast correction on the f-CaO data of the clinker obtained by prediction calculation, and adjusting f-CaO prediction model parameters of the clinker.
2. The method for predicting f-CaO of clinker in real time based on data of a fully automatic free calcium oxide analyzer according to claim 1, wherein the data processing comprises:
and (3) first-order filtering treatment: the first-order filtering expression is Y (n) ═ Coef x (n) + (1-Coef) × Y (n-1), where x (n) represents the current filtering input value, new sampling data signal; y (n) represents the output value of the current filtering; y (n-1) represents the last filtered output value; coef represents a filter coefficient, the value of the filter coefficient is between 0 and 1, the intensity of filtering is adjusted by adjusting the filter coefficient, and if the filter coefficient is closer to 1, the filtering is weaker, namely the filtering is closer to the original signal; if the filter coefficient is closer to 0, the stronger the filtering is;
abnormal value elimination: and directly eliminating the data with obvious abnormality in the field control system DCS.
3. The method for predicting f-CaO of clinker in real time based on data of full-automatic free calcium oxide analyzer as claimed in claim 1, wherein the input variables are sample data sets collected from a control system DCS, the sample data sets comprise a training data set with a label and a training data set without a label and a prediction data set, and the sample data are subjected to unified maximum and minimum normalization processing and abnormal value elimination processing.
4. The method for predicting f-CaO of clinker in real time based on data of a full-automatic free calcium oxide analyzer according to claim 1, wherein the third step is specifically as follows:
taking input variables as input, taking clinker f-CaO data of the full-automatic free calcium oxide detector for 1 h/time as output, carrying out normalization processing on all input and output variables, and constructing a clinker f-CaO prediction model based on a time sequence, wherein the model expression is as follows:
f-CaO=k0+k1x1+k2x2+k3x3+k4x4+
k5x5+k6x6+k7x7+k8x8
wherein k0 represents the intercept; k1-k8 are the contribution rates of all parameters, and particularly, k1 represents colorimetric high temperature; k2 represents the limestone saturation ratio KH; k3 represents the silicic acid ratio SM; k4 represents the aluminate ratio IM; k5 represents high temperature NOx; k6 denotes the smoke chamber temperature; k7 denotes the brightness of the flame; k8 represents the secondary air temperature.
5. The method for predicting f-CaO of clinker in real time based on data of a full-automatic free calcium oxide analyzer according to claim 1, wherein the fourth step is specifically: firstly, a primary pulse signal is sent out by a full-automatic free calcium detector every time sampling is carried out, a system records sampling time, then predicted clinker f-CaO data at the sampling moment is compared and corrected with the result of the full-automatic free calcium oxide detector, the confidence coefficient of the clinker f-CaO data detected by the full-automatic free calcium oxide detector is set, iterative adjustment is carried out on clinker f-CaO prediction model parameters by taking the confidence coefficient as a standard, and clinker f-CaO prediction automatic rolling optimization of the model is realized.
6. The method for predicting f-CaO of clinker in real time based on data of a fully automatic free calcium oxide analyzer according to claim 1, wherein the parameter configuration of the clinker f-CaO prediction model comprises setting a confidence factor, a test value safety check and a constraint correction coefficient;
wherein the setting confidence factor specifically comprises: the confidence factor ranges between 0 and 1, with values closer to 1 indicating more confidence in the most recent assay value, i.e.: f-CaO value of clinker detected by a full-automatic free calcium oxide detector;
the assay value safety check specifically comprises:
checking the maximum value: if the assay value is greater than the maximum value, not taking;
checking a minimum value: if the assay value is less than the minimum value, not taking;
checking the change rate: if the deviation of the test value is larger than the maximum deviation amount, not adopting the test value;
and fourthly, checking sampling time: if the sampling time is greater than the maximum sampling interval from the current time interval, not taking the assay value;
the constraint correction coefficients, namely: and (4) adding upper and lower limit constraints to the correction coefficient, and forcibly equaling the upper and lower limits after the correction coefficient is exceeded.
7. The method for predicting f-CaO of clinker in real time based on data of a fully automatic free calcium oxide analyzer according to claim 6, wherein the deviation of the assay value is calculated by the formula:
BiasDifference=Lab-PredUnbiased,
wherein Lab represents an assay value; prednbiased represents the predicted output of the soft meter at the same sampling instant.
8. The method for predicting f-CaO of clinker in real time based on data of a fully automatic free calcium oxide analyzer according to claim 6, wherein the calculation of the correction coefficient employs a relative value correction method or an absolute value correction method.
Background
The f-CaO content of the clinker is an important index for assessing the quality of the clinker, and represents the residual degree after the combination of calcium oxide, silicon oxide, aluminum oxide, iron oxide and the like in the calcination of the raw material, so the height of the f-CaO content represents the rationality of the ingredients and the complete degree of the calcination of the clinker, and further represents the stability quality of the clinker.
The clinker f-CaO is an important factor for reflecting the quality and influencing the stability of the cement, and is an important reference index in the production process. And controlling relevant technological parameters in the cement production process by a central control room operator according to the f-CaO content of the clinker. Therefore, the realization of the clinker f-CaO prediction real-time detection value has important significance for controlling the cement quality and guiding the cement production.
At present, clinker f-CaO detection modes are divided into laboratory offline detection and full-automatic free calcium oxide analyzer detection. The content of f-CaO in clinker is generally detected by a glycerol-ethanol method or a glycol-ethanol laboratory test method in China. The detection of the full-automatic free calcium oxide analyzer is mainly implemented by sampling through the full-automatic free calcium oxide analyzer, feeding samples through a shell and automatically detecting the f-CaO content of clinker. The sampling and testing method of the full-automatic free calcium oxide analyzer is that the clinker f-CaO content is obtained by sampling and testing automatically every 1 hour to the site, and the cement clinker calcining process has a certain time delay, and the result obtained by analysis has a certain hysteresis for guiding the control of a calcining system. The instantaneity is poor, and the production and control requirements of cement enterprises are often difficult to meet.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention provides a method for predicting clinker f-CaO in real time based on data of a full-automatic free calcium oxide analyzer, which combines an automatic control theory with an actual production mechanism, applies a computer technology, changes the clinker f-CaO which is difficult to monitor online in real time in the production process as an output variable, selects process data which is relatively easy to measure as an input variable, and infers and estimates the internal relation between the process data and the input variable through mathematical calculation, thereby achieving the purpose of replacing the real-time data of the clinker f-CaO and being difficult to measure, and the specific technical scheme is as follows:
the method for predicting f-CaO of clinker in real time based on data of a full-automatic free calcium oxide analyzer comprises the following steps:
acquiring field data, acquiring production data in a field control system DCS and detection data of a full-automatic free calcium oxide analyzer, and processing the data;
selecting input variables related to f-CaO of the clinker as auxiliary variables for predicting f-CaO of the clinker;
step three, constructing a clinker f-CaO prediction model according to and in combination with the selected variables;
and step four, in the same time sequence, acquiring f-CaO data of the clinker by using a full-automatic free calcium oxide detector, then performing contrast correction on the f-CaO data of the clinker obtained by prediction calculation, and adjusting f-CaO prediction model parameters of the clinker.
Further, the data processing includes:
and (3) first-order filtering treatment: the first-order filtering expression is Y (n) ═ Coef x (n) + (1-Coef) × Y (n-1), where x (n) represents the current filtering input value, new sampling data signal; y (n) represents the output value of the current filtering; y (n-1) represents the last filtered output value; coef represents a filter coefficient, the value of the filter coefficient is between 0 and 1, the intensity of filtering is adjusted by adjusting the filter coefficient, and if the filter coefficient is closer to 1, the filtering is weaker, namely the filtering is closer to the original signal; if the filter coefficient is closer to 0, the stronger the filtering is;
abnormal value elimination: and directly eliminating the data with obvious abnormality in the field control system DCS.
Further, the input variable is a sample data set acquired from the control system DCS, the sample data set comprises a training data set with a label and a training data set without a label and a prediction data set, and the sample data is subjected to unified maximum and minimum normalization processing and abnormal value elimination processing.
Further, the third step specifically includes:
taking input variables as input, taking clinker f-CaO data of the full-automatic free calcium oxide detector for 1 h/time as output, carrying out normalization processing on all input and output variables, and constructing a clinker f-CaO prediction model based on a time sequence, wherein the model expression is as follows:
f-CaO=k0+k1x1+k2x2+k3x3+k4x4+k5x5+k6x6+k7x7+k8x8
wherein k0 represents the intercept; k1-k8 are the contribution rates of all parameters, and particularly, k1 represents colorimetric high temperature; k2 represents the limestone saturation ratio KH; k3 represents the silicic acid ratio SM; k4 represents the aluminate ratio IM; k5 represents high temperature NOx; k6 denotes the smoke chamber temperature; k7 denotes the brightness of the flame; k8 represents the secondary air temperature.
Further, the fourth step specifically includes: firstly, a primary pulse signal is sent out by a full-automatic free calcium detector every time sampling is carried out, a system records sampling time, then predicted clinker f-CaO data at the sampling moment is compared and corrected with the result of the full-automatic free calcium oxide detector, the confidence coefficient of the clinker f-CaO data detected by the full-automatic free calcium oxide detector is set, iterative adjustment is carried out on clinker f-CaO prediction model parameters by taking the confidence coefficient as a standard, and clinker f-CaO prediction automatic rolling optimization of the model is realized.
Further, the parameter configuration of the clinker f-CaO prediction model comprises setting a confidence factor, a test value safety check and a constraint correction coefficient;
wherein the setting confidence factor specifically comprises: the confidence factor ranges between 0 and 1, with values closer to 1 indicating more confidence in the most recent assay value, i.e.: f-CaO value of clinker detected by a full-automatic free calcium oxide detector;
the assay value safety check specifically comprises:
checking the maximum value: if the assay value is greater than the maximum value, not taking;
checking a minimum value: if the assay value is less than the minimum value, not taking;
checking the change rate: if the deviation of the test value is larger than the maximum deviation amount, not adopting the test value;
and fourthly, checking sampling time: if the sampling time is greater than the maximum sampling interval from the current time interval, not taking the assay value;
the constraint correction coefficients, namely: and (4) adding upper and lower limit constraints to the correction coefficient, and forcibly equaling the upper and lower limits after the correction coefficient is exceeded.
Further, the calculation formula of the assay value deviation is as follows:
BiasDifference=Lab-PredUnbiased,
wherein Lab represents an assay value; prednbiased represents the predicted output of the soft meter at the same sampling instant.
Further, the calculation of the correction coefficient adopts a relative value correction method or an absolute value correction method.
The invention has the advantages that:
the method can update the clinker f-CaO real-time data according to the actual process variable change condition, can enable an operator and an expert controller system to obtain the clinker f-CaO real-time data information in time, and has the advantages of low cost, quick response, flexible use, simple maintenance and the like.
Drawings
FIG. 1 is a schematic diagram of the present invention;
FIG. 2 is a schematic diagram of configuration model parameters of the present invention;
FIG. 3 is a schematic diagram of the calculation of the correction coefficient according to the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and technical effects of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings.
As shown in figure 1, the method for predicting f-CaO of clinker in real time based on data of a full-automatic free calcium oxide analyzer specifically comprises the following steps:
step one, data acquisition and processing, specifically comprising:
collecting field data: the method comprises the steps of obtaining production data in a field control system DCS, detection data of a full-automatic free calcium oxide analyzer and sampling time of the full-automatic free calcium oxide analyzer, and acquiring process data values of cement production in real time.
And (3) carrying out data processing: aiming at the condition that the acquired field data has bad values, preprocessing methods such as filtering and noise reduction and standard are adopted to process the data acquisition abnormal values, and the method specifically comprises the following steps:
abnormal value elimination: directly removing acquired data with obvious abnormity caused by sudden change of certain parameters due to the influence of production field accidents or sensor faults in the process of acquiring data of a cement firing field;
and (3) first-order filtering treatment: and filtering the measurement signal to eliminate the influence of the noise of the signal on the control.
The filtering adopts first-order filtering, also called first-order inertia filtering or first-order low-pass filtering, and has good effect on filtering high-frequency part noise in signals, and the expression of the first-order filtering is as follows:
Y(n)=Coef*X(n)+(1-Coef)*Y(n-1),
wherein, X (n) represents the current filtering input value and a new sampling signal; y (n) represents the output value of the current filtering; y (n-1) represents the last filtered output value; coef represents a filter coefficient, the value of the Coef is between 0 and 1, a proper filter coefficient is selected according to actual needs, the intensity of filtering can be adjusted by adjusting the filter coefficient, and if the filter coefficient is closer to 1, the filtering is weaker, namely closer to the original signal; the closer the filter coefficient is to 0, the stronger the filtering.
Selecting variables to determine initial parameters, specifically:
the method comprises the steps of firstly, deeply researching a cement sintering process and a clinker f-CaO generation mechanism, analyzing the influence degree of each variable on the clinker f-CaO, and selecting characteristic input closely related to the clinker f-CaO as an auxiliary variable for predicting the clinker f-CaO.
The clinker f-CaO generation mechanism shows that the clinker f-CaO content mainly depends on the components of the raw meal and the calcination condition of the calcination system. The clinker f-CaO auxiliary variable set is selected preliminarily as follows: the limestone saturation ratio KH, the silicic acid rate SM, the aluminic acid rate IM, the current of a kiln main machine, the secondary air temperature, the smoke chamber NOx, the smoke chamber temperature, the pressure under a secondary grate, the outlet temperature of a decomposing furnace, the kiln rotating speed, the tertiary air temperature, the coal feeding amount of the decomposing furnace, the rotating speed of a high-temperature fan, the rotating speed of an EP fan, the feeding amount, the negative pressure of a kiln head, the colorimetric high temperature and the like.
The variables selected by the cement process are analyzed to be used as input variables of clinker f-CaO soft measurement, and the time sequence of each variable is used as model input; the input variables are sample data sets collected from a corresponding control system DCS, the sample data sets comprise labeled training data sets and unlabeled prediction data sets, and unified maximum and minimum normalization processing and abnormal value elimination processing are carried out on the sample data.
Step three, analyzing data, and establishing a clinker f-CaO prediction model, specifically comprising the following steps:
as shown in fig. 2, for characteristics such as multivariate coupling, nonlinear uncertain time lag and the like in the cement burning process, characteristic variables, i.e., input variables, of the prediction model are determined through correlation analysis of data; taking characteristic variables as input, taking clinker f-CaO data of the full-automatic free calcium oxide detector for 1 h/time as output, carrying out normalization processing on all input and output variables, and constructing a clinker f-CaO prediction model based on a time sequence, wherein the model expression is as follows:
f-CaO=k0+k1x1+k2x2+k3x3+k4x4+k5x5+k6x6+k7x7+k8x8
wherein k0 represents the intercept; k1-k8 are the contribution rates of all parameters, and particularly, k1 represents colorimetric high temperature; k2 represents the limestone saturation ratio KH; k3 represents the silicic acid ratio SM; k4 represents the aluminate ratio IM; k5 represents high temperature NOx; k6 denotes the smoke chamber temperature; k7 denotes the brightness of the flame; k8 represents the secondary air temperature.
Step four, predicting clinker f-CaO data for automatic correction, specifically:
in the same time sequence, a full-automatic free calcium oxide detector is adopted to obtain f-CaO data of the clinker, the f-CaO data of the clinker obtained by prediction and calculation is contrasted and corrected, and f-CaO prediction model parameters of the clinker are adjusted.
Specifically, firstly, the full-automatic free calcium detector sends out a pulse signal every time sampling, the system records the sampling time, then the predicted clinker f-CaO data at the sampling moment is compared and corrected with the result of the full-automatic free calcium oxide detector, a certain confidence coefficient is given to the clinker f-CaO data of the full-automatic free calcium oxide detector, and the clinker f-CaO prediction model parameters are iteratively adjusted by taking the value as a standard, so that the clinker f-CaO prediction automatic rolling optimization of the model is realized.
And configuring corresponding model parameters by combining the preliminary clinker f-CaO prediction model, wherein the method specifically comprises the following steps:
setting a confidence factor and safety check: because of the uncertainty in the analysis of the assay values, a confidence factor is introduced to compensate for the uncertainty. The confidence factor beliefFactor ranges from 0 to 1, with values closer to 1 indicating more confidence in the most recent assay value. In the actual calculation process, the correction coefficient BiasRatio/BiasAdd is further added with upper and lower limit constraints, and the upper and lower limits are forcibly equal after the upper and lower limits are exceeded.
And (3) carrying out correction safety check on f-CaO value of clinker detected by the full-automatic free calcium oxide detector:
checking the maximum value: if the assay value is greater than the maximum value, this value is rejected.
Checking a minimum value: if the assay value is less than the minimum value, this value is rejected.
Checking the change rate: if the assay value deviates by more than the maximum deviation, the value is rejected.
And fourthly, checking sampling time: if the sampling time is greater than the maximum sampling interval from the current time interval, the assay value may be rejected.
As shown in fig. 3, the assay value correction coefficient is calculated using a relative value correction method or an absolute value correction method.
The deviation of the predicted output value of the soft meter from the assay value is calculated prior to using the two methods:
lab-predunaided, wherein Lab represents assay value; prednbiased represents the predicted output of the soft meter at the same sampling instant.
Because assay analysis requires time, the input time of the assay value is typically later than the sampling time. Therefore, according to the sampling time acquired by the system, the test value is compared with the output value of the soft meter to obtain the test value deviation, and the final corrected predicted value is obtained according to the test value deviation.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way. Although the foregoing has described the practice of the present invention in detail, it will be apparent to those skilled in the art that modifications may be made to the practice of the invention as described in the foregoing examples, or that certain features may be substituted in the practice of the invention. All changes, equivalents and modifications which come within the spirit and scope of the invention are desired to be protected.
- 上一篇:石墨接头机器人自动装卡簧、装栓机
- 下一篇:一种用于橡胶制品的老化试验装置