Prediction method of iron slag content in zinc pot based on VAR multivariate time series
Through the prediction method of iron slag content of zinc pot based on VAR multi-time series, the problem of iron slag deposition in hot-dip galvanizing production is solved, and the accurate prediction and control of iron slag content in zinc pot is achieved, and the galvanizing quality and use of zinc pot are improved. life.
Patent Information
- Application Number
- CN202210448457.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-26
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2042-04-26
AI Technical Summary
During the hot-dip galvanizing production process, the deposition of iron slag in the zinc pot leads to a decrease in the service life of the zinc pot, the galvanizing quality decreases, and the suspended iron slag is difficult to clean, affecting the galvanizing quality.
The zinc pot iron slag content prediction method based on VAR multivariate time series is adopted. By selecting process parameters and constructing a vector autoregression model, the changes in the iron slag content are predicted, and the production parameters are adjusted through impulse response analysis to control the iron slag content.
It realizes accurate prediction and effective control of the iron slag content in the zinc pot, improves the galvanizing quality, and extends the service life of the zinc pot.
Smart Images

Figure CN114898817B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of hot-dip galvanizing, and in particular to a method for predicting the iron slag content in a zinc pot based on a VAR multivariate time series. Background Art
[0002] In the production process of hot-dip galvanizing, zinc slag is inevitably generated. Due to its large mass fraction, iron slag will be deposited at the bottom of the zinc pot, reducing the service life of the zinc pot; it will also continue to accumulate on the sinking roller and the stabilizing roller, forming indentations on the surface of the strip, reducing the quality of strip galvanizing; in addition, iron slag will also be suspended in the zinc liquid, making it difficult to clean the zinc slag. Analyzing the changing trend of iron slag content and influencing factors is of great significance to inhibiting the generation of zinc slag and improving the quality of galvanizing.
[0003] The vector autoregression (VAR) method is a multivariate data analysis method. The model directly considers the relationship between the variables in the time series, and uses multiple equations to construct the model by taking each endogenous variable in the system as a function of the lagged value of all endogenous variables in the system. In each equation of the model, the endogenous variable regresses the lagged value of all endogenous variables in the model, and then estimates the dynamic relationship of all endogenous variables and makes predictions. The VAR model can not only make predictions, but also analyze the impact between variables in the system, and has obvious advantages in analyzing systems composed of multiple variables. Summary of the invention
[0004] In order to solve the problems of the prior art, the present invention provides a method for predicting the iron slag content of a zinc pot based on VAR multivariate time series.
[0005] The present invention provides a method for predicting the iron slag content of a zinc pot based on a VAR multivariate time series, comprising the following steps:
[0006] Step 1: Select the process parameters produced in a zinc pot, including collection time, zinc bath temperature, total aluminum content, strip speed, and iron slag content;
[0007] Step 2: Construct a sample data set to obtain stable production data of the zinc pot area over a continuous period of time;
[0008] Step 3: Establish a prediction model for zinc pot iron slag content based on a vector autoregression model;
[0009] Step 3-1: Perform a stationarity test on each variable in the sample set of step 2;
[0010] Step 3-2: Select the optimal order of the vector autoregression model;
[0011] Based on the sample number T, the maximum lag order is established to be no more than Then, the optimal lag order is determined by LR test, AIC criterion, FPE criterion, SC criterion and HQ criterion respectively, and the order selected as the optimal lag order by most criteria is used as the autoregressive coefficient p value to build the VAR (p) model;
[0012] Step 3-3: Establish a vector autoregression model based on the optimal order, i.e., a prediction model for the iron slag content in the zinc pot;
[0013] Assuming that the optimal order is p, the sample set is T, the iron slag content, total aluminum content, zinc bath temperature, and strip speed are endogenous variables, the vector autoregression model is constructed as follows:
[0014] Y t =β1Y t-1 +…+β p Y t-p +C+u t ,t=1,2,…,T
[0015] Among them, Y t Y is a vector composed of the t-th iron slag content, total aluminum content, zinc bath temperature, and strip speed; t-p Yes t p-order lag; C is a 4-dimensional constant column vector, u t represents the random error term, β1,…,β p Represents the coefficient matrix to be estimated;
[0016] Step 3-4: Establish a vector autoregression model based on the optimal order and verify whether the model is stable. If it is unstable, re-build the model according to step 3;
[0017] Step 4: Analyze the impact of influencing factors on the iron slag content through pulse response, and predict the future changes in iron slag content based on the model.
[0018] Further improvement, the sample data set constructed in step 2 is to obtain stable production data of the zinc pot area for 3-10 consecutive days from a factory's strip hot-dip galvanizing production workshop, and the fluctuation of zinc bath composition is relatively stable within this time range. The step 2 is specifically to aggregate the data set obtained in step 1 to the minute level, including the iron slag content, strip speed, total aluminum content, and zinc bath temperature every 5 minutes, and obtain a single sample by sampling the data in step 1 every 5 minutes or taking the average every 5 minutes.
[0019] Further improvement, the step 3-1 is specifically to perform a stationarity test on the iron slag content, total aluminum content, zinc bath temperature, and strip speed in the sample set, and after all four variables pass the stationarity test, proceed to the subsequent steps. If they are not stationary, perform a differential calculation on the variables until the stationarity test passes. The stationarity test uses the ADF method to perform a unit root test.
[0020] As a further improvement, the step 3-4 is specifically to calculate the characteristic roots of the VAR (p) equation established in step 3-3. If all the characteristic roots are within the unit circle, it means that the constructed VAR (p) model is stable, and this model is used for subsequent analysis.
[0021] The beneficial effects of the present invention are:
[0022] 1. The prediction results are accurate and reliable: The present invention incorporates the influence of strip speed, total aluminum content and zinc bath temperature into the prediction model, and accurately and effectively predicts the generation trend of iron slag in the zinc bath.
[0023] 2. The process parameters in the production process can be adjusted according to the analysis results of the pulse response function to achieve effective control of the iron slag content in the zinc bath, which is of great significance for guiding production. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0025] Figure 1 It is a schematic flow chart of the method for predicting zinc pot iron slag in the present invention;
[0026] Figure 2 It is the impulse response function analysis result of the zinc slag prediction model in the embodiment of the present invention. DETAILED DESCRIPTION
[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0028] An accurate and effective prediction method for the iron slag content of zinc pots, the flow chart is as follows Figure 1 As shown, including:
[0029] Step 1: Select the process parameters and other data produced in a zinc pot, including acquisition time, zinc bath temperature, total aluminum content, strip speed, iron slag content, etc.;
[0030] The production data of seven consecutive days were selected from the hot-dip galvanizing workshop of a factory, including production time, strip speed, zinc bath temperature, total aluminum content, and iron slag content;
[0031] Step 2: Build a sample dataset;
[0032] The data in step 1 are sampled every five minutes to obtain a sample data set. The sample set contains 1808 samples, each of which is time, iron slag content, zinc bath temperature, strip speed, and total aluminum content;
[0033] Step 3: Establish a prediction model for zinc pot iron slag content based on a vector autoregression model;
[0034] Step 3-1: Perform a stationarity test on each variable in the sample set of step 2;
[0035] The ADF method is used to perform stationarity tests on the iron slag content, total aluminum content, zinc bath temperature, and strip speed series in the sample set. The test results are shown in the table. At the 5% significance level, the four series are stationary within the 95% confidence interval. Therefore, the four selected variables have passed the stationarity test and can be used as endogenous variables in the multivariate time series model to construct a model.
[0036] Step 3-2: Select the optimal order of the vector autoregression model;
[0037] According to the sample number of 1808, it is determined that the maximum lag order does not exceed 12, and then the optimal lag order is determined by LR test, AIC criterion, FPE criterion, SC criterion and HQ criterion respectively. The results are as follows:
[0038] Table 1 Test values of various selection criteria for lag order of VAR model
[0039]
[0040] Note: * indicates the optimal lag order selected under this criterion, and NA indicates no value.
[0041] It can be seen from the table results that the optimal lag order of the VAR model finally determined is 3.
[0042] Step 3-3: Establish a vector autoregression model based on the optimal order, that is, a prediction model for the iron slag content in the zinc pot.
[0043] Assuming that the optimal order is p, the sample set is T, the iron slag content, total aluminum content, zinc bath temperature, and strip speed are endogenous variables, the vector autoregression model is constructed as follows:
[0044] Y t =β1Y t-1 +…+β p Y t-p +C+u t ,t=1,2,...,T
[0045] Among them, Yt Y is a vector composed of the t-th iron slag content, total aluminum content, zinc bath temperature, and strip speed; t-p Yes t p-order lag; C is a 4-dimensional constant column vector, u t represents the random error term, β1,…,β p Represents the coefficient matrix to be estimated.
[0046] In this case, the bottom slag content, zinc bath temperature, strip running speed, and total aluminum content are used as the four endogenous variables of the vector autoregression model. The vector autoregression model is constructed with a lag order of 3. The model results are shown in the formula:
[0047]
[0048] Among them, DROSS is the bottom slag content, TBATH is the zinc bath temperature, V is the strip running speed, and TAL is the total aluminum content.
[0049] Step 3-4: Establish a vector autoregression model based on the optimal order and verify whether the model is stable. If it is unstable, re-build the model according to step 3.
[0050] The characteristic roots of the VAR (3) equation established in the example are calculated and the results are as follows:
[0051] Table 2 Characteristic root table of vector autoregression model
[0052]
[0053] All characteristic roots calculated by the characteristic equation are within the unit circle, indicating that the constructed VAR model is stable, that is, the system composed of bottom slag content, zinc bath temperature, strip speed, and total aluminum content is stable. The equation for bottom slag content is as follows:
[0054] DROSS t =1.954*DROSS t-1 +(8.083e-05)*TBATH t-1 -(2.363e-07)*V t-1 +0.0876*TAL t-1
[0055] -0.98*DROSS t-2 +(1.524e-04)*TBATH t-2 +(3.523e-06)*V t-2 -0.17*TAL t-2
[0056] +0.021*DROSS t-3-(7.119e-05)*TBATH t-3 -(1.158e-06)*V t-3 +0.083*TAL t-3
[0057] -(2.6166e-04)
[0058] Step 4: Analyze the impact of influencing factors on the iron slag content through pulse response, and predict the future changes in iron slag content based on the model.
[0059] Impulse response function analysis:
[0060] For the established bottom slag content prediction model, the VAR (3) model, a pulse of unit standard deviation is applied to each explanatory variable in the model, and the cumulative change amplitude and impact duration of the bottom slag content in the zinc bath after each explanatory variable emits a unit pulse are analyzed over time after being impacted by its pulse.
[0061] The pulse response function analysis results of zinc bath temperature, strip running speed and total aluminum content to bottom slag content are shown in the figure, where V is the strip running speed, TBATH is the zinc bath temperature, and TAL is the total aluminum content.
[0062] Figure 2 The impulse response function analysis results of the zinc slag prediction model in the embodiment of the present invention are shown in FIG. According to the impulse response function analysis results in the figure, the following conclusions can be drawn:
[0063] 1) After applying a pulse shock of one unit standard deviation to the strip running speed, zinc bath temperature and total aluminum content, their effects on the bottom dross content in the zinc bath tend to stabilize at about 45 minutes, that is, a change in a related process parameter at a certain moment will continue to affect the bottom dross content in the zinc bath for the next 45 minutes, and then gradually dissipate.
[0064] 2) Under the production condition that the average running speed of the strip is 2m / s, the running speed of the strip is negatively correlated with the amount of bottom dross in the zinc bath. In the actual production process, briefly increasing the running speed of the strip can reduce the bottom dross content in the zinc bath to a certain extent in the short term.
[0065] 3) Under the production condition that the average zinc bath temperature is 460°C, there is a weak negative correlation between the zinc bath temperature and the amount of bottom slag in the zinc bath. In the actual production process, if it is necessary to effectively reduce the bottom slag content in the zinc bath by increasing the zinc bath temperature, it is necessary to make a significant change to the zinc bath temperature.
[0066] 4) Under the production conditions where the average total aluminum content in the zinc bath is 0.135%, the total aluminum content is strongly positively correlated with the bottom slag content in the zinc bath. In the actual production process, the input system of zinc ingots with different aluminum contents when adding zinc to the zinc liquid needs to be carefully managed and studied.
[0067] Prediction accuracy analysis:
[0068] The production data of 10 steel coils of the steel plant were selected, and the model was used to predict the bottom slag content in the next 5 minutes, 15 minutes, 30 minutes and 60 minutes. The prediction results of each period are shown in the following table:
[0069] Table 3 Comparison of dynamic prediction values and actual values for the next 5 minutes
[0070]
[0071] Table 4 Comparison of model prediction values and true values for the next 15 minutes
[0072]
[0073] Table 5 Comparison of model prediction values and true values for the next 30 minutes
[0074]
[0075] Table 6 Comparison of model prediction values and true values for the next 60 minutes
[0076]
[0077] After the model is dynamically predicted, the mean absolute error (MAE) and mean absolute percentage error (MAPE) are used to evaluate the accuracy of the model prediction in each time period. The calculation formulas of the two evaluation indicators are as follows:
[0078]
[0079] The results of the model prediction evaluation indicators for each time period are shown in Table 7.
[0080] Table 7 Evaluation index table of model prediction results in each time period
[0081]
[0082] It can be seen from Table 7 that when the dynamic prediction time does not exceed 30 minutes, the average absolute percentage error of the model prediction is less than 5%, and the model accuracy is very high; when the prediction time reaches 60 minutes, the average absolute percentage error of the model prediction reaches about 18%, and the prediction accuracy is poor. Considering that the average fluctuation range of the bottom slag content in the zinc bath in actual production is about 65%, it can be considered that the model can effectively and accurately predict the changes in the bottom slag content in the next 30 minutes.
[0083] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, the above is only a preferred embodiment of the present invention. Since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited to this. Any technician familiar with the technical field is within the technical scope disclosed by the present invention. For ordinary technicians in the technical field, changes or replacements that can be easily thought of should be covered within the protection scope of the present invention without departing from the principle of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for predicting the iron slag content in a zinc pot based on VAR multivariate time series, characterized in that The following steps are involved: Step 1: Select the process parameters produced in a zinc pot; Step 2: Construct a sample data set to obtain stable production data of the zinc pot area over a continuous period of time; Step 3: Establish a prediction model for zinc pot iron slag content based on a vector autoregression model; Step 3-1: Perform a stationarity test on each variable in the sample set of step 2; Step 3-2: Select the optimal order of the vector autoregression model; Based on the sample number T, the maximum lag order is established to be no more than Then, the optimal lag order is determined by LR test, AIC criterion, FPE criterion, SC criterion and HQ criterion respectively, and the order selected as the optimal lag order by most criteria is used as the autoregressive coefficient p value to build the VAR (p) model; Step 3-3: Establish a vector autoregression model based on the optimal order, i.e., a prediction model for the iron slag content in the zinc pot; Assuming that the optimal order is p, the sample set is T, the iron slag content, total aluminum content, zinc bath temperature, and strip speed are endogenous variables, the vector autoregression model is constructed as follows: WATER t Gβ1Y t-1 +...+β p WATER t-p +C+u t ,TJ1,2,...,T Among them, Y t Y is a vector composed of the t-th iron slag content, total aluminum content, zinc bath temperature, and strip speed; t-p Yes t p-order lag; C is a 4-dimensional constant column vector, u t represents the random error term, β1,…,β p Represents the coefficient matrix to be estimated; Step 3-4: Establish a vector autoregression model based on the optimal order to verify whether the model is stable. If it is unstable, re-build the model according to step 3; Step 4: Analyze the impact of influencing factors on the iron slag content through pulse response, and predict the future changes in iron slag content based on the model.
2. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 1, characterized in that: The process parameters of the zinc pot production in step 1 include collection time, zinc bath temperature, total aluminum content, strip speed, and iron slag content.
3. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 1, characterized in that: The sample data set constructed in step 2 is obtained from a factory's strip hot-dip galvanizing production workshop to obtain stable production data of the zinc pot area for 3-10 consecutive days. The fluctuation of zinc bath composition is relatively stable within this time range.
4. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 3 is characterized in that: Specifically, the step 2 is to aggregate the data set obtained in step 1 to the minute level, including the iron slag content, strip speed, total aluminum content, and zinc bath temperature every 5 minutes, and obtain a single sample by sampling the data in step 1 every 5 minutes or taking the average every 5 minutes.
5. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 1, characterized in that: The step 3-1 is specifically to perform a stationarity test on the iron slag content, total aluminum content, zinc bath temperature, and strip speed in the sample set respectively. After all four variables pass the stationarity test, proceed to the subsequent steps. If they are not stationary, perform a differential calculation on the variables until the stationarity test passes.
6. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 1 or 5, characterized in that: The stationarity test adopts the ADF method to perform a unit root test.
7. The method for predicting the iron slag content in a zinc pot based on VAR multivariate time series according to claim 1, characterized in that: The step 3-4 is specifically to calculate the characteristic roots of the VAR (p) equation established in step 3-3. If all the characteristic roots are within the unit circle, it means that the constructed VAR (p) model is stable, and this model is used for subsequent analysis.
