CN110444257B - Cement free calcium soft measurement method based on unsupervised and supervised learning - Google Patents
Cement free calcium soft measurement method based on unsupervised and supervised learning Download PDFInfo
- Publication number
- CN110444257B CN110444257B CN201910717686.9A CN201910717686A CN110444257B CN 110444257 B CN110444257 B CN 110444257B CN 201910717686 A CN201910717686 A CN 201910717686A CN 110444257 B CN110444257 B CN 110444257B
- Authority
- CN
- China
- Prior art keywords
- layer
- coding
- input
- unsupervised
- learning
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
Landscapes
- Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Chemical & Material Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a cement free calcium soft measurement method based on unsupervised and supervised learning, wherein variables are selected as input variables of clinker fCaO soft measurement by analyzing a cement process, and a time sequence of each variable is input as a model; constructing a prediction model combining unsupervised learning and supervised learning by using the selected input variables; removing a decoding layer of sparse self-coding, stacking coding layers to form a deep network structure, initializing deep network parameters by using determined initial parameters, and performing supervised learning by adopting a BP reverse error correction algorithm; and (3) carrying out real-time prediction on the fCaO of the cement clinker by using a well-trained prediction model combining unsupervised learning and supervised learning. The model forward adopts a layer-by-layer greedy unsupervised learning mode to extract the high-level characteristics of the data; and further optimizing parameters by combining with supervised reverse fine adjustment, and realizing clinker fCaO real-time prediction by using the trained deep network.
Description
Technical Field
The invention relates to the field of industrial cement quality soft measurement monitoring, in particular to a cement free calcium soft measurement method based on unsupervised and supervised learning.
Background
The content of free calcium oxide (fCaO) of cement clinker is an important index for measuring the quality of the clinker in the novel dry-process cement production. The fCaO content of the clinker not only influences the stability and the strength of the clinker, but also is directly related to the energy consumption for cement firing. At present, the fCaO of the cement clinker is mainly sampled one hour manually and tested by laboratory tests, but the offline measurement result has obvious hysteresis phenomenon in the cement burning process, and meanwhile, the label samples are fewer, so that the real-time control and optimization of the cement burning process are difficult to realize. The cement clinker sintering process has the characteristics of large inertia, large time lag, multiple coupling, few label samples and the like, so that an accurate cement clinker fCaO prediction model is difficult to establish. In response to the above problems, some scholars have used different soft-sensing methods to study clinker fcoa predictions. And selecting five related variables burnt by the cement clinker in Zhao and Peng courses and the like to establish a multi-core LSSVM cement clinker fCaO prediction model. The Zhao-wave and others establish a soft measuring model of fCaO for cement clinker of convolutional neural network. In the methods, a supervised method is adopted to search the relation between the clinker fCaO and the auxiliary variable, but the labeled clinker fCaO has less data and insufficient learned characteristics. In consideration of the fact that clinker fCaO has few label samples, the overfitting phenomenon is easy to occur when a small amount of sample data trains the deep neural network. And fully utilizing the unlabeled data, fully extracting the characteristic information of the variable by utilizing a large amount of unlabeled data, and then carrying out fine adjustment by utilizing the sample data with the labels. The more effective soft measurement modeling method for the fCaO of the cement clinker is obtained.
Disclosure of Invention
Aiming at the existing problems, the invention provides a cement free calcium soft measurement method based on unsupervised and supervised learning, so that the influence of insufficient network training or overfitting caused by less label sample data is eliminated, and the characteristic information of the variable is further fully acquired.
In order to realize the purpose, the following technical scheme is adopted: the method comprises the following steps:
step 1: selecting 10 variables as input variables of clinker fCaO soft measurement by analyzing a cement process, and inputting a time sequence of each variable as a model; the input variable acquires a sample data set from a corresponding DCS, wherein the sample data set comprises a training data set with a label and without the label and a prediction data set, and the sample data of the input variable is subjected to unified maximum and minimum normalization processing; removing abnormal values by adopting a 3 sigma principle; the 10 variables comprise decomposing furnace coal feeding quantity feedback, high-temperature fan rotating speed feedback, decomposing furnace outlet temperature, pressure feedback under a two-chamber grate, EP fan rotating speed, feeding quantity 1 feedback, kiln head negative pressure, secondary air temperature feedback, kiln current and kiln tail temperature;
step 2: constructing a prediction model combining unsupervised learning and supervised learning by using the input variables selected in the step 1, and determining initial parameters, wherein the initial parameters comprise a network layer number, a hidden layer node number, a learning rate, a sparse penalty term coefficient and a sparsity coefficient; performing unsupervised forward training on each sparse self-coding by adopting a greedy training mode from bottom to top, namely training each layer of neural network layer by layer, updating only one self-coding parameter each time, and determining the initial weight w and bias b of each layer of the depth network;
and step 3: removing the decoding layer of the sparse self-coding, stacking the coding layers to form a deep network structure, initializing the deep network parameters by using the initial parameters determined in the step 2, and further optimizing the initial weight w and the bias b of the parameters by combining a supervision BP reverse error correction algorithm to finish the training of a supervised part;
and 4, step 4: and (4) carrying out real-time prediction on the fCaO of the cement clinker by using the prediction model combining unsupervised learning and supervised learning which is trained in the step (3).
The invention has the following beneficial effects:
1. aiming at the characteristics of large inertia, large time lag, multiple coupling and the like in the cement clinker sintering process, the deep structure is adopted to replace a single hidden layer structure, the capability of high-order potential interpretation factors of learning contents is improved, and the complex nonlinear characteristic of the cement industry is solved.
2. Aiming at the problem that the cement clinker fCaO label samples are few, a method combining unsupervised learning and supervised learning is provided; the sparse self-encoder adopts unsupervised layer-by-layer training to extract internal features of data, and combines the influence of different time delays of the data on the fCaO of the cement clinker in a period of supervised fine tuning learning, so that the complex process of deep network initialization training is solved, and the training speed of the network is improved.
3. Aiming at the high nonlinear characteristic of the cement clinker fCaO sintering process, the nonlinear characteristic of the asymmetric characteristic learning data of the encoder and the decoder is adopted, and the asymmetry can also solve the overfitting problem in the self-encoding training process.
4. The method can accurately predict the fCaO content of the cement clinker, and plays an important guiding role in improving the quality of the cement clinker and reducing the production energy consumption.
Drawings
FIG. 1 is a diagram of a soft measurement scheme for free calcium of cement clinker according to the present invention;
FIG. 2 is a schematic diagram of the unsupervised learning training of the present invention;
FIG. 3 is a diagram of supervised learning training of the present invention;
FIG. 4 is a graph comparing the accuracy of the present invention.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
the invention provides a cement free calcium soft measurement method based on unsupervised and supervised learning, which is characterized in that a specially designed depth self-encoder is adopted as a core, and a stack type (depth) asymmetric Sparse Complete self-encoder (SC-AE) with 3 hidden layers is designed by combining the characteristics of Sparse self-encoding and Complete self-encoding, as shown in figure 3, and the feature extraction capability of a single hidden layer is improved by adopting multiple hidden layers structurally. The structural diagram of the soft measurement is shown in fig. 1, firstly, selecting variables, then, performing abnormal value processing and normalization processing on a sample set, and selecting a training sample set and a prediction sample set, wherein the training sample set comprises a labeled data set and an unlabeled data set; reconstructing an input layer of each SC-AE to obtain initial parameters of the network; the decoding layer of each SC-AE is removed, and the coding layers are stacked to form a deep network structure. And finally, carrying out supervised global reverse fine adjustment by adopting a BP algorithm, correcting errors and optimizing weights and offsets, and completing the construction of a cement clinker fCaO prediction model, wherein the content comprises the following steps:
step 1: selecting input variables according to a cement process, performing data processing, and constructing an input layer of the model; the inclusion of temporal characteristics in the input layer enables the predictive model to learn the time delays in the data variables.
It is known from cement technology that cement raw materials are calcined in a rotary kiln at high temperature to form a sintering reaction, solid particle materials obtained by cooling are called cement clinker, and a solidified body contains a small amount of unreacted calcium oxide called free calcium oxide (fCaO). Too high a level of fCaO will reduce the stability of the cement, and too low a level will increase the energy consumption for cement firing, so that the fCaO needs to be controlled within a reasonable range. In the cement calcination process, all parameters of a burning zone play a crucial role in the content of the clinker fCaO, so that the parameters of a burning system are main factors for realizing soft measurement of the clinker fCaO. The heat source of the burning zone is the coal feeding amount of the decomposing furnace, the coal feeding amount of the kiln head and the secondary air temperature recycled into the kiln from the grate cooler, and the temperature of the burning zone influences the content of calcium oxide generated in the raw material decomposing process and the absorption condition of the generated calcium oxide by other compounds. The high-temperature fan and the EP fan enable huge air pressure difference to be generated in the kiln, so that the air passage of a cement burning system is guaranteed to be smooth, and the pressure in the kiln is kept stable. When the rotary kiln rotates, a kiln motor is required to provide power, the uniformity of chemical reaction of materials in the rotary kiln is ensured, and the higher the current of a kiln main machine is, the higher the viscosity of the materials in the kiln is, and the higher the temperature in the kiln is.
From the above analysis, 10 variables closely related to the clinker fcoa content were selected: the method comprises the following steps of decomposing furnace coal feeding amount feedback, high-temperature fan rotating speed feedback, decomposing furnace outlet temperature, pressure feedback under a two-chamber grate, EP fan rotating speed, feeding amount 1 feedback, kiln head negative pressure, secondary air temperature feedback, kiln current and kiln tail temperature.
Step 2: and (3) initially establishing a model, constructing a stacked sparse self-encoder, and initially determining an initial weight, bias, a sparse penalty term coefficient and a sparsity coefficient.
The stack SC-AE prediction model fuses the time sequence into the input layer, so that the input layer comprises time sequence data in a period of time, and the influence of time-varying time delay on the precision of the prediction model in the cement production process is eliminated. And (3) extracting data characteristics by adopting sparse self-encoding (SAE), and completing forward training by adopting greedy unsupervised learning layer by layer.
1. Preliminary model building
The unsupervised and supervised learning combined prediction model provided by the invention adopts a 5-layer network structure, and the number of neuron nodes of each layer is as shown in figure 3: 600,800,900,1200,1. The learning rate is selected to be 0.0001, the sparse penalty term coefficient is selected to be 3, the sparse coefficient is selected to be 0.001, the initial weight among the network structures adopts a random number among the truncated positive-Tai distribution [0,1], and the bias is initially 0.
2. Construction of stacked sparse autoencoder
The middle layer of the stacked SC-AE is formed by continuously stacking a plurality of sparse self-encodings (SAEs), namely removing the decoding layer of each SAE, and stacking the encoding layers to form a deep network structure. Individual SAE learn data characteristics using an unsupervised approach. The principle of the self-encoder is that the output of the decoder can reproduce the input by training an encoding and decoding mechanism by taking the characteristics of the human brain as reference. Input reconstruction at the output end is not easy and has no practical significance, but through designing a special structure, constraint is added during input reconstruction, a special cost function is used, only approximate reconstruction can be realized, a model is forced to reconstruct input data according to weight, and therefore useful characteristics of the data are learned in a hidden layer.
(1) Stacked SC-AE input layer
Let x be the input to the stacked SC-AE, which contains a time series of selected 10 variables, which can be expressed as:
x＝(x1,x2,...x10) (1)
each variable time series contains t sample points:
xi＝(xi(1),xi(2)...xi(t)) (2)
in the formula, xi(i ═ 1, 2.. 10) is the time series for the ith variable.
(2) SC-AE coding layer and decoding layer
SC-AE encoding/decoding process As shown in FIG. 2, if the l-th layer is an encoding layer, the layer inputs xl-1And output xl+1And the expression of layer l is formula (3) (4):
xl＝f(xl-1wl+bl) (3)
xl+1＝f(xl(wl)T+bl+1) (4)
in the formula, xl-1Features representing input of coding layerEigenvectors, xl+1Representing the coded layer output eigenvectors, wlRepresenting the weight between the coding layer and the input layer, (w)l)TWeight between coding layer and output layer, bl，bl+1Representing the corresponding bias of the output feature vector. Where f (.) is the activation function. The SC-AE model adopts a ReLU function as an activation function, and the expression of the ReLU function is shown as formula (5).
f(x)＝max(0,x) (5)
(3) Stacked SC-AE
And eliminating the decoding layer of each SC-AE, stacking the coding layers together to form a deep network, wherein the input of the prediction output layer is the coding layer of the third SAE and is expressed as:
y＝wh3+b (6)
wherein y represents the predicted output layer, w represents the weight between the output layer and the third hidden layer, h3Represents the coding layer of the third SAE, and b represents the corresponding offset of the output.
The design of the stack type sparse self-editor adds sparsity limitation, and the loss function of each sparse self-coding is expressed as:
wherein, the first term of the above formula is a least square loss function, the second term is a regularization term (preventing model overfitting), and the third term is a sparse term (only a small part of neurons in a hidden layer are in an activated state, and the rest of neurons are in a suppressed state); n is the number of hidden layer neurons, m is the number of input layer neurons, ρ is a sparsity parameter, usually a smaller value close to 0, λ is a regularization coefficient, where KL contrasts and divergences are expressed as:
the formula is that one rho is mean value and one is
Beta controls the weight of the sparsity penalty factor, which is the relative entropy between two bernoulli random variables of the mean. We adjust w, b by minimizing the loss function using a gradient descent method so that the reconstruction error is minimized, as shown in the following equation:
and step 3: unsupervised learning process
And fixing the neuron number of the hidden layer according to the loss function of the self-coding model, so that the network output value can reconstruct the input as much as possible, thereby training the self-coding model. And after the self-coding model is converged, removing the output layer, and taking the output value of the hidden layer as the characteristic of the original signal. Similarly, we use the features from the previous layer as the input of the self-coding model of the next layer to train the second layer network. By analogy, we can build a stacked sparse self-coding model, and use the network to predict the cement clinker fCaO as shown in FIG. 3.
And 4, step 4: supervised fine-tuning stacked sparse self-coding model
And constructing a neural network model with the same structure as the stacked sparse self-coding network. Setting initial values of a weight matrix and a bias item of each layer in the neural network model as an input-hidden layer weight matrix and a bias vector of an output layer of each layer in the stacked sparse self-coding model respectively, inputting a sample data set with a label, training the neural network model again by using a BP algorithm, and updating the weight matrix and the bias item again.
By error function
Wherein
t is the true value of the sample and y is the predicted value of the network.
For a single sample xiThe weight of the kth neuron of the output layer (L layer) and the jth neuron of the L-1 layer
The partial derivatives are:
the partial derivatives for the bias are:
for L-1 hidden layer:
similarly:
order to
Then:
similarly, the variance of the weights and offsets for layer l is:
and finally, updating formulas of the weight and the bias:
wl＝wl-η*Δwl (21)
bl＝bl-η*Δbl (22)
wherein
Is the output of the ith neuron of layer i; f (-) is the activation function and η is the learning rate.
It can be known from the above analysis that the updating method of the reverse fine tuning specifically includes that firstly, the variable quantity of the weight and the offset between the output layer (L layer) and the second last layer (L-1 layer) is obtained according to the formulas (12) and (13), then the updating of the weight and the offset is realized according to the formulas (21) and (22), and then all the weights and the offsets are sequentially updated layer by layer in reverse by using the formulas (21) and (22), so as to complete the parameter fine tuning which has supervision on the whole network.
And 5: predicted output of cement clinker fCaO.
And (4) predicting clinker fCaO by using the model which is trained in the steps 2, 3 and 4 and combines supervision and unsupervised, wherein the prediction result is shown in figure 4.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and not restrictive, and various changes and modifications may be made to the technical solution of the present invention by those skilled in the art without departing from the spirit of the present invention, which is defined by the claims.
Claims (2)
1. A cement free calcium soft measurement method based on unsupervised and supervised learning is characterized by comprising the following steps:
step 1: selecting 10 variables as input variables of clinker fCaO soft measurement by analyzing a cement process, taking a time sequence of each variable as model input, and performing abnormal value processing and normalization processing on each selected variable time sequence; the 10 variables comprise decomposing furnace coal feeding quantity feedback, high-temperature fan rotating speed feedback, decomposing furnace outlet temperature, pressure feedback under a two-chamber grate, EP fan rotating speed, feeding quantity 1 feedback, kiln head negative pressure, secondary air temperature feedback, kiln current and kiln tail temperature;
step 2: constructing a prediction model combining unsupervised learning and supervised learning by using the input variables selected in the step 1, constructing a stacked sparse autoencoder, extracting data characteristics by adopting a multi-hidden-layer network structure and adopting sparse autoencoding, and determining initial parameters, wherein the initial parameters comprise a network layer number, a hidden-layer node number, a learning rate, a sparse penalty item coefficient and a sparsity coefficient; performing unsupervised forward training on each sparse self-coding by adopting a greedy training mode from bottom to top, namely training each layer of neural network layer by layer, updating only one self-coding parameter each time, and determining the initial weight w and bias b of each layer of the depth network; the construction of the stacked sparse self-encoder specifically comprises the following steps:
(1) stacked SC-AE input layer
Let x be the input to the stacked SC-AE, which contains a time series of selected 10 variables, which can be expressed as:
x＝(x1,x2,...x10) (1)
each variable time series contains t sample points:
xi＝(xi(1),xi(2)...xi(t)) (2)
in the formula, xi(i ═ 1, 2.. 10) is a time series for the ith variable;
(2) SC-AE coding layer and decoding layer
If the l-th layer is a coding layer, then the layer inputs xl-1And output xl+1And the expression of layer l is formula (3) (4):
xl＝f(xl-1wl+bl) (3)
xl+1＝f(xl(wl)T+bl+1) (4)
in the formula, xl-1Feature vector, x, representing the input of the coding layerl+1Representing the coded layer output eigenvectors, wlRepresenting the weight between the coding layer and the input layer, (w)l)TWeight between coding layer and output layer, bl，bl+1Representing the corresponding offset of the output feature vector; wherein f (.) is an activation function; the SC-AE model adopts a ReLU function as an activation function, and the expression of the ReLU function is as shown in formula (5):
f(x)＝max(0,x) (5)；
(3) stacked SC-AE
And eliminating the decoding layer of each SC-AE, stacking the coding layers together to form a deep network, wherein the input of the prediction output layer is the coding layer of the third SAE and is expressed as:
y＝wh3+b (6)
wherein y represents the predicted output layer, w represents the weight between the output layer and the third hidden layer, h3Represents the coding layer of the third SAE, b represents the corresponding offset of the output;
the design of the stack type sparse self-editor adds sparsity limitation, and the loss function of each sparse self-coding is expressed as:
wherein the first term of the above equation is a least squares loss function, the second term is a regularization term, and the third term is a sparse term; n is the number of hidden layer neurons, m is the number of input layer neurons, ρ is a sparsity parameter, which is a smaller value close to 0, λ is a regularization coefficient, where KL contrasts and divergences are expressed as:
where the formula is one p as the mean and oneTo make
Beta controls the weight of the sparsity penalty factor as the relative entropy between two bernoulli random variables of the mean; the loss function is minimized using a gradient descent method to minimize the reconstruction error, thereby adjusting w, b as shown in the following equation:
and step 3: removing the decoding layer of the sparse self-coding, stacking the coding layers to form a deep network structure, learning data characteristics by adopting an unsupervised method for a single sparse self-coding, initializing deep network parameters by using the initial parameters determined in the step 2, and further optimizing the initial weight w and the bias b of the parameters by combining a supervised BP reverse error correction algorithm to finish the training of a supervised part;
and 4, step 4: and (4) carrying out real-time prediction on the fCaO of the cement clinker by using the prediction model combining unsupervised learning and supervised learning which is trained in the step (3).
2. The cement free calcium soft measurement method based on unsupervised and supervised learning as recited in claim 1, wherein: in step 1, the input variable collects a sample data set from a corresponding DCS, wherein the sample data set comprises a training data set and a prediction data set which are labeled and unlabeled, and the sample data of the input variable is subjected to unified maximum and minimum normalization processing; and (5) carrying out abnormal value elimination by adopting a 3 sigma principle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910717686.9A CN110444257B (en) | 2019-08-05 | 2019-08-05 | Cement free calcium soft measurement method based on unsupervised and supervised learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910717686.9A CN110444257B (en) | 2019-08-05 | 2019-08-05 | Cement free calcium soft measurement method based on unsupervised and supervised learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110444257A CN110444257A (en) | 2019-11-12 |
| CN110444257B true CN110444257B (en) | 2022-02-18 |
Family
ID=68433295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910717686.9A Active CN110444257B (en) | 2019-08-05 | 2019-08-05 | Cement free calcium soft measurement method based on unsupervised and supervised learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110444257B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111916156B (en) * | 2020-06-23 | 2023-10-27 | 宁波大学 | Tail gas sulfur-containing substance concentration real-time prediction method based on stack-type self-encoder |
| CN112536797A (en) * | 2020-12-01 | 2021-03-23 | 南京航空航天大学 | Comprehensive compensation method for position and attitude errors of industrial robot |
| CN112951349A (en) * | 2021-03-24 | 2021-06-11 | 湖州槐坎南方水泥有限公司 | Method for predicting strength of portland cement clinker based on DNN neural network |
| CN113177358B (en) * | 2021-04-30 | 2022-06-03 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
| CN113420815B (en) * | 2021-06-24 | 2024-04-30 | 江苏师范大学 | Nonlinear PLS intermittent process monitoring method of semi-supervision RSDAE |
| CN113593657B (en) * | 2021-07-26 | 2024-01-09 | 燕山大学 | Cement free calcium soft measurement system with quality target as guiding semi-supervised learning |
| CN117692346A (en) * | 2024-01-31 | 2024-03-12 | 浙商银行股份有限公司 | Message blocking prediction method and device based on spectrum regularization variation self-encoder |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106202946A (en) * | 2016-07-18 | 2016-12-07 | 燕山大学 | Clinker free calcium levels Forecasting Methodology based on degree of depth belief network model |
| CN106323636A (en) * | 2016-08-16 | 2017-01-11 | 重庆交通大学 | Adaptive extraction and diagnosis method for degree features of mechanical fault through stack-type sparse automatic coding depth neural network |
| CN109147878A (en) * | 2018-10-08 | 2019-01-04 | 燕山大学 | A kind of clinker free calcium flexible measurement method |
| CN109342703A (en) * | 2018-12-06 | 2019-02-15 | 燕山大学 | A kind of clinker free calcium levels measurement method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10219510B4 (en) * | 2002-04-30 | 2004-05-06 | imko Intelligente Micromodule Köhler GmbH | Procedure for determining the moisture of a substrate |
| CN104536396A (en) * | 2014-12-08 | 2015-04-22 | 沈阳工业大学 | Soft measurement modeling method used in cement raw material decomposing process in decomposing furnace |
-
2019
- 2019-08-05 CN CN201910717686.9A patent/CN110444257B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106202946A (en) * | 2016-07-18 | 2016-12-07 | 燕山大学 | Clinker free calcium levels Forecasting Methodology based on degree of depth belief network model |
| CN106323636A (en) * | 2016-08-16 | 2017-01-11 | 重庆交通大学 | Adaptive extraction and diagnosis method for degree features of mechanical fault through stack-type sparse automatic coding depth neural network |
| CN109147878A (en) * | 2018-10-08 | 2019-01-04 | 燕山大学 | A kind of clinker free calcium flexible measurement method |
| CN109342703A (en) * | 2018-12-06 | 2019-02-15 | 燕山大学 | A kind of clinker free calcium levels measurement method and system |
Non-Patent Citations (2)
| Title |
|---|
| 基于粒子群算法与连续型深度信念网络的水泥熟料游离氧化钙预测;刘彬;《计量学报》;20180531;第39卷(第3期);420-424 * |
| 基于连续型深度信念网络的水泥熟料游离钙含量软测量研究;高伟;《中国优秀硕士学位论文全文数据库（电子期刊）工程科技Ⅰ》;20180515;第2018年卷(第5期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110444257A (en) | 2019-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110444257B (en) | Cement free calcium soft measurement method based on unsupervised and supervised learning | |
| CN109147878B (en) | Soft measurement method for free calcium of cement clinker | |
| CN109685252B (en) | Building energy consumption prediction method based on cyclic neural network and multi-task learning model | |
| CN109816095B (en) | Network flow prediction method based on improved gated cyclic neural network | |
| CN106202946A (en) | Clinker free calcium levels Forecasting Methodology based on degree of depth belief network model | |
| CN110378044B (en) | Multi-time scale convolution neural network soft measurement method based on attention mechanism | |
| CN110210495B (en) | XGboost soft measurement modeling method based on parallel LSTM self-encoder dynamic feature extraction | |
| CN109342703B (en) | Method and system for measuring content of free calcium in cement clinker | |
| CN112734002B (en) | Service life prediction method based on data layer and model layer joint transfer learning | |
| CN110083125B (en) | Machine tool thermal error modeling method based on deep learning | |
| WO2022121944A1 (en) | Method and apparatus for intelligent forecasting of industrial process operation indicators, device, and storage medium | |
| Li et al. | Dynamic time features expanding and extracting method for prediction model of sintering process quality index | |
| CN109685283B (en) | Method for predicting submerged arc furnace working condition based on confidence rule base reasoning | |
| CN107868979A (en) | A kind of silicon single crystal diameter control method based on permanent casting speed control structure | |
| CN113780420B (en) | GRU-GCN-based method for predicting concentration of dissolved gas in transformer oil | |
| CN109143853B (en) | Self-adaptive control method for liquid level of fractionating tower in petroleum refining process | |
| CN111639111A (en) | Water transfer engineering-oriented multi-source monitoring data deep mining and intelligent analysis method | |
| CN115034465B (en) | Coke quality prediction method based on artificial intelligence | |
| CN111639820A (en) | Energy consumption parameter optimization method and system for cement kiln production | |
| CN112906976A (en) | Cement clinker free calcium sample data enhancement and prediction method based on R-WGAN | |
| Liu et al. | Model fusion and multiscale feature learning for fault diagnosis of industrial processes | |
| CN114721263A (en) | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm | |
| Ji et al. | Application of the improved the ELM algorithm for prediction of blast furnace gas utilization rate | |
| CN113627594A (en) | One-dimensional time sequence data amplification method based on WGAN | |
| CN117289652A (en) | Numerical control machine tool spindle thermal error modeling method based on multi-universe optimization |
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 |