CN110928186B - Automatic covering agent adding control method for tundish based on machine vision - Google Patents
Automatic covering agent adding control method for tundish based on machine vision Download PDFInfo
- Publication number
- CN110928186B CN110928186B CN201911173230.7A CN201911173230A CN110928186B CN 110928186 B CN110928186 B CN 110928186B CN 201911173230 A CN201911173230 A CN 201911173230A CN 110928186 B CN110928186 B CN 110928186B
- Authority
- CN
- China
- Prior art keywords
- covering agent
- tundish
- image
- self
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B22—CASTING; POWDER METALLURGY
- B22D—CASTING OF METALS; CASTING OF OTHER SUBSTANCES BY THE SAME PROCESSES OR DEVICES
- B22D11/00—Continuous casting of metals, i.e. casting in indefinite lengths
- B22D11/16—Controlling or regulating processes or operations
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Mechanical Engineering (AREA)
- Image Analysis (AREA)
- Feedback Control In General (AREA)
Abstract
The invention relates to a machine vision-based automatic covering agent adding control method for a tundish, which can fully mine field image information data and other influence factors, can gradually accumulate a charging port image database in the using process, increases a self-learning correction function due to more considered influence factors, has stronger model robustness, achieves a temperature hit rate of 95 percent, reduces the waste of a covering agent by 30 percent, can effectively add the covering agent, and improves the production benefit of enterprises.
Description
Technical Field
The invention relates to the field of covering agent adding for a tundish of a continuous casting platform in the steel industry, in particular to a control method for automatically adding a covering agent for the tundish based on machine vision.
Background
The refined clean molten steel with qualified components and temperature can contact with air in the transfer process from a steel ladle to a tundish and then injected into a crystallizer, so that the molten steel is subjected to secondary oxidation, the quality of continuous casting molten steel is seriously affected, and therefore, a covering agent is added into the tundish to achieve the effect of preserving heat and preventing secondary oxidation. At present, the mode of adding the covering agent to the tundish on the platform of the continuous casting machine is mostly a manual mode, the state of the hole is observed at any moment by people, and whether the covering agent is added or not is determined by manual experience. The manual addition of the covering agent is greatly influenced by the factors of an operator, the stability of the addition is difficult to ensure, the labor intensity of workers is high, and the working environment is severe.
Nowadays, with the increase of automation level, more and more steel enterprises have started to be intelligently transformed. The core technology of automatic covering agent adding of the tundish is feed opening image feature extraction and feature detection, and machine vision is combined with an automatic control system to obviously improve the field working efficiency and stability. In the current generation of the rapid development of artificial intelligence, the application of artificial intelligence and machine vision to the production of the steel industry is an inevitable trend, and steel enterprises also urgently need a system with high automation degree for automatically adding a covering agent to a tundish to ensure the process of adding the covering agent with high efficiency and high precision, and the core of the system lies in that the image characteristic extraction of a feeding port is combined with other influence factors to judge the amount of the covering agent to be added.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a method for realizing control of a covering agent by combining image characteristic information of a feed inlet with other influence factors. The method is characterized in that a deep convolution neural network is added into the system, the characteristic of an image at a feeding port is extracted to obtain image characteristic information, and the image characteristic information and other influence factors are comprehensively calculated to judge whether a covering agent needs to be added.
The specific technical scheme is that the control method for automatically adding the covering agent to the tundish based on machine vision comprises the following steps:
Wherein, a Relu activation function is connected behind each convolution layer, the output layer uses a Softmax function, and uses a cross entropy function as a loss function;
step 4, outputting the test set to the optimized deep convolution neural network model, outputting the opening grade of the tundish feeding port, and increasing the self-learning coefficient omega t The opening grade of the charging opening and a self-learning coefficient omega are adjusted t Carrying out variable proportion weighting with other factors influencing the combustion of the covering agent to obtain the amount of the covering agent required to be added at the feeding port of the tundish, sending a signal to a PLC (programmable logic controller), and controlling the addition by the PLCThe material feeder quantitatively adds the covering agent, and the self-learning coefficient omega t And correcting and calculating the parameters by acquiring field physical parameters and feedback values.
Further, other factors influencing the burning of the covering agent in the step 4 include, but are not limited to, the carbon content C of the molten steel, which is the carbon content of the molten steel flowing into the tundish from the packaging process, a time difference T, which indicates the time from the last addition of the covering agent to the last addition of the covering agent, and a temperature R, which indicates the temperature before the addition of the covering agent, and are obtained by a temperature sensor.
Furthermore, the specific implementation manner of obtaining the required covering agent adding amount of the tundish feeding port in the step 4 is as follows,
setting the output data of the deep convolution neural network model as P k ，P d ＝P k * And 10, comprehensively weighting the covering agent with other influence factor values on site in a variable proportion manner to obtain the covering agent addition quantity Q, wherein the unit of Q is kg, and the method comprises the following steps:
wherein, P d Data in a fractional system for the neural network model output data, gamma,
σ, μ is the scaling weight, ω t Is a self-learning coefficient; c, T and R are respectively values of corresponding scores, time difference values and temperature values of corresponding steel grades after unified measurement, namely values normalized to be between 0 and 100.
Further, in step 4, the coefficient ω is learned by self t Obtained by the following formula,
wherein, ω is t Self-learning coefficient, omega, representing the addition of the covering agent at the t-th time t-1 Denotes the self-addition of the covering agent at the t-1 th timeLearning coefficient, the initial value of the self-learning coefficient is 1,T t-1 Temperature after addition of the covering agent for the T-1 th time, T tar Is the target temperature.
Further, a specific implementation manner of the preprocessing in step 1 is as follows, firstly, gray processing is performed on the acquired image, and a gray image can be obtained by performing weighted average on RGB three components according to the following formula:
f(i,j)＝0.30R(i,j)+0.59G(i,j)+0.11B(i,j)
in the formula: r-image red channel, G-image green channel, B-image blue channel, M < i, j < N, wherein M, N are image horizontal and vertical resolution;
then the image after the gradation processing is subjected to threshold processing,
setting a global threshold T, and dividing the data of the image into two parts by using T: pixel groups larger than T and pixel groups smaller than T; and setting the pixel value of the pixel group larger than T as white or black, and setting the pixel value of the pixel group smaller than T as black or white, thereby obtaining the binary gray image at the feeding port of the tundish.
The invention has the beneficial effects that: the method can fully mine field image information data and other influence factors, gradually accumulate the charging opening image database in the using process, increase the self-learning correction function due to more considered influence factors, has stronger model robustness, achieves the temperature hit rate of 95 percent, reduces the covering agent waste by 30 percent, can effectively add the covering agent, and improves the production benefit of enterprises.
Drawings
Fig. 1 is a machine vision equipment layout;
FIG. 2 is a process flow diagram of the present invention;
FIG. 3 is a comparison graph of various activation functions;
fig. 4 is an activation function application diagram.
Detailed Description
The technical scheme of the invention is further explained by combining the drawings and the embodiment.
As shown in fig. 2, the invention provides a control method for automatically adding a covering agent to a tundish based on machine vision, which comprises the following steps:
step 1: as shown in fig. 1, an industrial camera is installed right above a tundish charging hole on a continuous casting platform, image acquisition is carried out on the tundish charging hole every 1s, the acquired image is preprocessed to construct a training set and a test set of a deep convolution neural network, and the opening grade of the charging hole is obtained through manual marking;
the specific implementation manner of the preprocessing in the step 1 is as follows, firstly, the gray level processing is carried out on the collected image, and the gray level image can be obtained by carrying out weighted average on RGB three components according to the following formula:
f(i,j)＝0.30R(i,j)+0.59G(i,j)+0.11B(i,j)
in the formula: r-image red channel, G-image green channel, B-image blue channel, M < i, j < N, wherein M, N are image horizontal and vertical resolution sizes;
then the image after the gradation processing is subjected to threshold processing,
setting a global threshold T, and dividing the data of the image into two parts by using T: pixel groups larger than T and pixel groups smaller than T; and setting the pixel value of the pixel group larger than T as white or black, and setting the pixel value of the pixel group smaller than T as black or white, thereby obtaining the binary gray image at the feeding port of the tundish.
The training set and the test set for constructing the deep convolutional neural network in the step 1 are as follows:
the input layer of the neural network model is a preprocessed acquired image of a tundish material port, and the output layer is a charging port opening grade P k ，(0<k<K) In that respect Finally, the network output is a score system, namely, the network output is marked as 10 when the opening grade is 1, and is marked as 20 \8230whenthe opening grade is 2, and finally, the obtained data set is divided into a test set and a training set according to the proportion of 2;
and 2, step: establishing a deep convolutional neural network model, connecting a Relu activation function behind each convolutional layer, using a softmax function as an output layer, and using a cross entropy function as a loss function in the model;
determining the network structure as convolutional layer N = 1 3, a pooling layer N 2 =3, adopt Max pooling mode, all connect layer N 3 Number of output layer units is N =2 4 =10, i.e. P k ，(0≤k<K) After the regression of softmax, the opening grade of the material opening is divided into 10 grades;
in deep learning, signals are calculated by linear superposition before being transmitted from one neuron to the next layer of neurons, and the signals need to pass through a nonlinear activation function to continue to be transmitted downwards, and the process is circulated;
further, as shown in fig. 3, a comparison graph of each activation function is shown, and a Relu activation function is used in step 2, and the specific form is as follows:
in the formula, y i For each layer of neuron data, w i As weights between neurons, b i Is an offset matched with the weight; where figure 4 is an activation function application diagram.
Further, the cross entropy as the loss function in step 2 is specifically as follows:
the cross entropy characterizes the distance between two probability distributions, given two probability distributions p and q, the cross entropy of p is represented by q as:
H(p,q)＝-∑p(x)logq(x)
in the formula, p represents a correct answer, q represents a predicted value, and the smaller the cross entropy value is, the closer the two probability distributions are;
before cross entropy is carried out, softmax regression needs to be carried out on the output values of the neural network, namely, the network output values are changed into probability distribution from real numbers, and the form is as follows:
the output of the primary neural network is P 1 ，…，P k ,(0<k<K) The output after the softmax regression process is:
the predicted value after softmax regression is [ y 0 ，y 1 ，…，y k ]。
And step 3: training and optimizing the deep convolutional neural network model by using a training set, reducing model overfitting by using an L2 regularization and Dropout algorithm, and training and adjusting model weights by using an Adam algorithm to obtain an optimized deep convolutional neural network model;
in step 3, the deep convolutional neural network model is trained and optimized by using the training set as follows:
training model parameters w, b by using a training set to minimize a model cross entropy loss function, wherein w represents weights between layers, and b represents trainable offset between layers;
the cross entropy loss function is calculated as follows:
training the concentrated output layer to be an opening grade P k ，(0<k<K) K is 10, estimating a total of 10 classes.
When the opening grade is 0 min, namely 0 grade, the opening grade is expressed as [1, 0., 0];
the opening grade is 10 minutes, namely, the opening grade is represented as [0, 1., 0] when the opening grade is 1 grade;
...
the opening is represented as [0, 0., 0,1] when the opening grade is 100 minutes, namely 10 grades;
the cross entropy loss function calculation process is as follows:
H 0 ((1,0,…,0,0),(y 0 ，y 1 ，…，y K ))＝-(1×logy 0 +0×logy 1 +…+0×logy K )
H 1 ((0,1,…,0,0),(y 0 ，y 1 ，…，y K ))＝-(0×logy 0 +1×logy 1 +…+0×logy K )
...
H K ((0,0,…,0,1),(y 0 ，y 1 ，…，y K ))＝-(0×logy 0 +0×logy 1 +…+1×logy K )
during training, the input layer image is assumed to correspond to a label opening grade of 1 grade, namely [1,0, \ 8230;, 0]The model predicted value is [ y 0 ，y 1 ，…，y K ]Then the loss function is of the form:
H 1 ((0,0,…,0,1),(y 0 ，y 1 ，…，y K ))＝-(0×logy 0 +1×logy 1 +…+0×logy K )
at the moment, the neural network model trains model weight parameters by minimizing cross entropy function loss through an Adam algorithm;
further, adam in step 3 above is a first-order optimization algorithm that can replace the conventional stochastic gradient descent process, and it can iteratively update the neural network weights based on the training data, and the form is as follows:
V dw ＝β 1 V dw +(1-β 1 )dw
V db ＝β 1 V db +(1-β 1 )db
S dw ＝β 2 S dw +(1-β 2 )dw 2 ，S db ＝β 2 S db +(1-β 2 )db 2
in the formula (I), the compound is shown in the specification,
w weight, dw weight gradient;
b-bias, db-bias gradient;
V dw -an exponential moving average of the weight gradient, initialized to 0 at training;
V db -exponentially moving average of the bias gradient, initialized to 0 at training;
S dw -an exponential moving average of the weight gradient squared, initialized to 0 at training;
S db -an exponential moving average of the squared bias gradient, initialized to 0 at training;
α — the learning rate, also known as the step factor, controls the update rate of the weights; debugging is needed;
β 1 -exponential decay rate of first order moment estimate, moving average of dw, common default value 0.9;
β 2 -exponential decay Rate of second moment estimate, calculating dw 2 And db 2 The usual default value of the moving weighted average of (2) is 0.999;
e — this parameter is a very small number, which is to prevent division by zero in the implementation (e.g., 10E-8);
further, the L2 regularization and Dropout algorithm is of the form:
the L2 regularization means that an L2 norm penalty term is added to a weight parameter w
In combination with the Adam algorithm, the parameter optimization formula is as follows:
wherein, alpha is a learning rate, and a factor E smaller than 1 is multiplied by the weight parameter w relative to a normal gradient optimization formula in the Adam algorithm, so that w is continuously reduced;
the deep neural network using the Dropout algorithm is in the form of the training process:
in the formula, the Bernoulli function is to generate a probability r vector, that is, a vector of 0 and 1 is randomly generated, and p is a random probability value; when the code level is realized in forward transmission, for a certain layer of network units, multiplying the network units by 0 according to a certain probability, and temporarily discarding the network units, so that the appearance of one neuron is independent of other neurons;
for the ith neuron data of layer l +1,
to be the weight between the ith neurons of layer l +1,
is the ith offset matched with the weight for the l +1 th layer;
preferably, upper
Step 4, outputting the test set to the optimized deep convolutional neural network model, outputting the opening grade of a tundish feeding port, and increasing a self-learning coefficient omega t The opening grade of the charging opening and the self-learning coefficient omega are set t Carrying out variable proportion weighting with other factors influencing the combustion of the covering agent to obtain the amount of the covering agent required to be added to the charging hole of the tundish, sending a signal to a PLC (programmable logic controller), and quantitatively adding the covering agent by controlling a charging device by the PLC, wherein the self-learning coefficient omega is t And correcting and calculating the parameters by acquiring field physical parameters and feedback values.
Further, the specific implementation manner of obtaining the required covering agent adding amount of the tundish feeding port in the step 4 is as follows,
setting the output data of the deep convolutional neural network model as P k ，P d ＝P k * And 10, comprehensively weighting the change proportion with other influence factor values on site to obtain the covering agent addition Q (kg), which is concretely as follows:
wherein, P d Data in a score system for the neural network model output data, gamma,
σ, μ is the scaling weight, ω t For the self-learning coefficient, the gamma,
σ, μ is set to 0.4,0.2,0.1,0.3 according to field work experience; c, T and R are respectively the carbon content of the molten steel, the time difference value and the temperature value after the uniform measurement, namely the values are standardized to be between 0 and 100.
Further, the self-learning coefficient ω in step 4 t Obtained by the following formula,
wherein, ω is t Self-learning coefficient, omega, representing the addition of the covering agent at the t-th time t-1 The self-learning coefficient of the covering agent added at the t-1 th time is shown, and the initial value of the self-learning coefficient is 1,T t-1 Temperature after addition of the covering agent for the T-1 th time, T tar Is the target temperature. The main function of adding the covering agent is heat preservation, so that the temperature of the molten steel is stabilized at a set value, and the self-learning coefficient is self-adapted by using the temperature as a weighing factor.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (3)
1. A control method for automatically adding a covering agent to a tundish based on machine vision is characterized by comprising the following steps:
step 1, preprocessing an acquired tundish charging hole image on a continuous casting platform to construct a training set and a test set of a deep convolution neural network model, and manually marking to obtain the opening grade of the charging hole;
step 2, establishing a deep convolutional neural network model with a network structure of convolutional layer N 1 A pooling layer N 2 A full connection layer N 3 N, the number of output layer units 4 The number of output layer units is also equal to the opening grade P of the feed inlet k ；
Wherein, a Relu activation function is connected behind each convolution layer, the output layer uses a Softmax function, and uses a cross entropy function as a loss function;
step 3, training and optimizing the deep convolution neural network model by using a training set, reducing overfitting of the network model by using L2 regularization and Dropout algorithm in the training process, and adjusting the weight of the model by using Adam algorithm to obtain an optimized deep convolution neural network model;
step 4, inputting the test set into the optimized deep convolutional neural network model, outputting the opening grade of a tundish feeding port, and increasing a self-learning coefficient omega t The opening grade of the charging opening and the self-learning coefficient omega are set t Carrying out variable proportion weighting with other factors influencing the combustion of the covering agent to obtain the amount of the covering agent required to be added to the feeding port of the tundish, sending a signal to a PLC (programmable logic controller), controlling a feeder to carry out quantitative addition on the covering agent by the PLC, and carrying out self-learning coefficient omega t The method comprises the steps of correcting and calculating by acquiring field physical parameters and feedback values;
in the step 4, other factors influencing the combustion of the covering agent comprise the carbon content C of the molten steel, a time difference T and a temperature value R, wherein the carbon content C of the molten steel is the carbon content of the molten steel which flows into the tundish in a packaging mode, the time difference T represents the time from the last time of adding the covering agent to the last time of adding the covering agent, and the temperature R represents the temperature value before adding the covering agent and is obtained through a temperature sensor;
the specific implementation manner of obtaining the amount of the covering agent required to be added to the feeding port of the tundish in the step 4 is as follows,
setting the output data of the deep convolutional neural network model as P k ，P d ＝P k * And 10, comprehensively weighting the covering agent with other influence factor values on site according to the transformation ratio, and obtaining the adding amount Q of the covering agent, wherein the unit of Q is kg, and the concrete steps are as follows:
wherein, P d Data in a fractional system for the neural network model output data, gamma,
σ, μ is the scaling weight, ω t Is a self-learning coefficient; c, T and R are respectively the values of corresponding molten steel carbon content, time difference value and temperature value after unified measurement, namelyNormalized to a value between 0 and 100.
2. The machine vision-based automatic covering agent adding control method for the tundish of claim 1, wherein the method comprises the following steps: self-learning coefficient omega in step 4 t Obtained by the following formula,
wherein, ω is t Self-learning coefficient, omega, representing the addition of the covering agent at the t-th time t-1 Shows the self-learning coefficient of the covering agent added at the t-1 st time, and the initial value of the self-learning coefficient is 1,T t-1 Temperature after addition of the covering agent for the T-1 th time, T tar Is the target temperature.
3. The machine vision-based automatic covering agent adding control method for the tundish of claim 1, wherein the method comprises the following steps: the specific implementation manner of the preprocessing in the step 1 is as follows, firstly, the gray level processing is carried out on the collected image, and the gray level image can be obtained by carrying out weighted average on RGB three components according to the following formula:
f(i,j)＝0.30R(i,j)+0.59G(i,j)+0.11B(i,j)
in the formula: r-image red channel, G-image green channel, B-image blue channel, M < i, j < N, wherein M, N are image horizontal and vertical resolution;
then the image after the gradation processing is subjected to threshold processing,
setting a global threshold T ', and dividing the image data into two parts by T': a pixel group larger than T 'and a pixel group smaller than T'; and setting the pixel value of the pixel group larger than T 'as white or black, and setting the pixel value of the pixel group smaller than T' as black or white, thereby obtaining the binary gray image at the feeding port of the tundish.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911173230.7A CN110928186B (en) | 2019-11-26 | 2019-11-26 | Automatic covering agent adding control method for tundish based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911173230.7A CN110928186B (en) | 2019-11-26 | 2019-11-26 | Automatic covering agent adding control method for tundish based on machine vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110928186A CN110928186A (en) | 2020-03-27 |
| CN110928186B true CN110928186B (en) | 2022-10-14 |
Family
ID=69851935
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911173230.7A Active CN110928186B (en) | 2019-11-26 | 2019-11-26 | Automatic covering agent adding control method for tundish based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110928186B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106180619A (en) * | 2016-08-12 | 2016-12-07 | 湖南千盟物联信息技术有限公司 | A kind of system approach of casting process Based Intelligent Control |
| CN106552909A (en) * | 2015-09-28 | 2017-04-05 | 宝山钢铁股份有限公司 | For the robot coverture added automatically system and its method of continuous casting work |
| CN109215009A (en) * | 2017-06-29 | 2019-01-15 | 上海金艺检测技术有限公司 | Continuous casting billet surface image defect inspection method based on depth convolutional neural networks |
| CN110340322A (en) * | 2019-08-22 | 2019-10-18 | 联峰钢铁（张家港）有限公司 | A kind of method and apparatus of continuous casting automatic casting |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2360357A (en) * | 2000-03-17 | 2001-09-19 | Alex Davidkhanian | Slag detector for molten steel transfer operations |
| CA3025870A1 (en) * | 2016-07-28 | 2018-02-01 | Refractory Intellectual Property Gmbh & Co. Kg | System with a spraying nozzle unit and method for spraying an inorganic mass |
-
2019
- 2019-11-26 CN CN201911173230.7A patent/CN110928186B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106552909A (en) * | 2015-09-28 | 2017-04-05 | 宝山钢铁股份有限公司 | For the robot coverture added automatically system and its method of continuous casting work |
| CN106180619A (en) * | 2016-08-12 | 2016-12-07 | 湖南千盟物联信息技术有限公司 | A kind of system approach of casting process Based Intelligent Control |
| CN109215009A (en) * | 2017-06-29 | 2019-01-15 | 上海金艺检测技术有限公司 | Continuous casting billet surface image defect inspection method based on depth convolutional neural networks |
| CN110340322A (en) * | 2019-08-22 | 2019-10-18 | 联峰钢铁（张家港）有限公司 | A kind of method and apparatus of continuous casting automatic casting |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110928186A (en) | 2020-03-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110781944B (en) | Automatic molten iron slag-off control method based on deep learning | |
| US11294338B2 (en) | Use of comprehensive artificial intelligence in primary industry plants | |
| CN110438284A (en) | A kind of converter intelligence tapping set and control method | |
| CN106119458B (en) | Converter steelmaking process cost control method and system based on BP neural network | |
| CN111650903B (en) | Intelligent control system for bottom argon blowing of steel ladle based on visual identification | |
| CN111142060B (en) | Adaptive threshold adjustment diagnosis method based on improved BP neural network | |
| CN108251591A (en) | Utilize the top bottom blowing converter producing process control method of LSTM systems | |
| CN106096724B (en) | Converter steelmaking process cost control method and system based on ELM neural network | |
| CN110097929A (en) | A kind of blast furnace molten iron silicon content on-line prediction method | |
| CN108891421B (en) | Method for constructing driving strategy | |
| CN108265157A (en) | Intelligent arc furnace steelmaking system | |
| CN110928186B (en) | Automatic covering agent adding control method for tundish based on machine vision | |
| CN116852665A (en) | Injection molding process parameter intelligent adjusting method based on mixed model | |
| CN110288591B (en) | Zinc flotation working condition discrimination method based on improved self-adaptive multi-population genetic algorithm | |
| CN116427950A (en) | Adaptive intelligent shield construction fine construction auxiliary decision-making system and method | |
| CN112132700A (en) | Formula decision method for regulating and controlling water quality | |
| CN107292387A (en) | A kind of method that honey quality is recognized based on BP | |
| CN210765379U (en) | Device for intelligent tapping of converter | |
| CN115994692B (en) | Intelligent river and lake management platform based on 5G and big data | |
| CN117213583A (en) | Method for measuring coal level of raw coal bunker of thermal power generation coal conveying system | |
| CN115584375B (en) | Automatic tapping method and system for converter based on image recognition | |
| CN116753536A (en) | Boiler pressure control method based on pressure prediction model | |
| CN116197904A (en) | Multi-degree-of-freedom 3D printing correction method based on deep learning | |
| CN109325594A (en) | A kind of power communication scene O&M aided diagnosis method improving CBR | |
| CN115130846A (en) | Method for analyzing maximum influence factors of lower part heat system of blast furnace based on tuyere monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |