CN102393884B - Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network - Google Patents
Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network Download PDFInfo
- Publication number
- CN102393884B CN102393884B CN201110307393.7A CN201110307393A CN102393884B CN 102393884 B CN102393884 B CN 102393884B CN 201110307393 A CN201110307393 A CN 201110307393A CN 102393884 B CN102393884 B CN 102393884B
- Authority
- CN
- China
- Prior art keywords
- steel billet
- neural network
- temperature
- matrix
- heating
- 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
Abstract
The invention discloses a hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network. The existing prediction method is dependent on manual work, and thus, has the defects of low efficiency and poor reliability. The method disclosed by the invention comprises the following steps: selecting prediction model variables: reasonably selecting input/output variables of a prediction model by utilizing mechanism analysis and prior information; normalizing data to be input; establishing a BP neural network, and training and testing the neural network; and finally, denormalizing the data obtained by the neural network, thereby obtaining the predicted heating temperature. In the invention, the steel billet temperature is predicted according to the operation history data of the electromagnetic induction heater, and has higher prediction precision than the traditional engineering computation method.
Description
Technical field
The invention belongs to technical field of automation, be specifically related to a kind of Forecasting Methodology of the hot continuous rolling electromagnetic induction heating steel billet temperature based on neural network.
Background technology
At steel industry, traditional steel-making, continuous casting, steel rolling process are respective independently production links.The hot continuous rolling integral process that but modern production model progressively changes into " steel smelting-continuous casting-steel rolling ", the continuous hot-rolling automatic production line pattern that the height that Here it is generally praises highly at present is both at home and abroad intensive.The firing equipment (heating, soaking, holding furnace) adopted between continuous casting and tandem rolling is the key equipment being connected continuous casting and rolling production line.From technological angle, middle firing equipment needs to carry out concurrent heating with the function making strand homogeneous temperature reach tandem rolling temperature requirement to it in the process completing conveying continuous casting billet, solves the problem of board briquette field inequality.Heating furnace, as the buffer zone between conticaster and tandem mill two kinds of different process speed component, is connected the logistics between casting machine with milling train, cushions, the production of both coordinations.
Due to the plurality of advantages of electromagnetic induction heating technology, such as firing rate is fast, power density is controlled, pollution-free, be easy to control, oxidization burning loss is few etc., compared to traditional gas-fired furnace, without the need to naked light thermal source, there is the advantage of zero-emission, be more suitable for using in hot continuous rolling production line.Therefore, high-power medium frequency induction heater is progressively applied to hot continuous rolling production line, see Fig. 1.
Because electromagnetic induction heating is a complicated non-linear Large Time Delay Process, be difficult to set up accurate mechanism model, be difficult to obtain satisfied effect by the control method (as PID regulates) of routine, the general artificial experience that adopts is debugged and is controlled.
In view of neural network has powerful non-linear mapping capability and good fault-tolerance, the time of day data of system can well be approached.Therefore, on a large amount of induction heating service data bases accumulated, adopt neural net method to set up hot continuous rolling electromagnetic induction heating temperature prediction model, steel billet temperature is predicted, can be electromagnetic induction heating accurate temperature controlling and foundation is provided.
Summary of the invention
The present invention is directed to the deficiencies in the prior art, a kind of Forecasting Methodology of the hot continuous rolling electromagnetic induction heating steel billet temperature based on neural network is provided.The concrete steps of the method are:
Step (1) selects forecast model variable.
Nerual network technique is adopted to set up hot continuous rolling electromagnetic induction heating steel billet temperature forecast model, for ensureing the neural net model establishing validity based on data, avoid the blindness of pure black-box modeling, first utilize Analysis on Mechanism and prior imformation, the input/output variable of choose reasonable forecast model.
Select the steel billet temperature after electromagnetic induction heating to be the output variable of neural network model based on Analysis on Mechanism, select the principal element affecting steel billet temperature to be neural network model input variable: the steel billet temperature 1. before electromagnetic induction heating; 2. the voltage of induction heater; 3. the electric current of induction heater.Neural network model output variable is the steel billet temperature after electromagnetic induction heating.
In inductive coil, being induction heated process for steel billet, is depart from magnetic field to steel billet from steel billet enters coil magnetic field.First, ignoring in heat transfer, heat radiation situation, heats up and is mainly subject to voltage, the function of current of this process induction heater in a certain cross section of steel billet.Therefore, the voltage of induction heater, electric current are sequence variables.Secondly, consider heat transfer, heat radiation situation, the steel billet temperature before and after electromagnetic induction heating also gets sequence variables.
Step (2) data normalization process.
Input packet in training sample is containing three, and order of magnitude difference is comparatively large, and for ensureing each factor par, convergence speedup speed, is normalized data, is converted into the value of [0,1] interval range
.
Wherein
for the maximal value in input data,
for the minimum value in input data.
for input data,
for the value after the process of input data normalization.
Step (3) builds BP neural network framework.
The newff function called in MatlabR2009a Neural Network Toolbox sets up BP neural network, Net=newff (PR,
,
, BTF, BLF, PF); Net is BP neural network framework, and PR is the span determined by greatest member and least member in input matrix,
be
ithe neuronic number of layer,
be
ithe transport function of layer,
,
for the total number of plies of neural network, BTF is the training function of BP neural network, and BLF is weights and bias, and PF is network performance function.
Step (4) training BP neural network.Concrete grammar is:
A, initialization BP neural network, the value assignment utilizing random function to produce, to weights and bias, is then called init function and is carried out initialization BP neural network.
B, network training number of times and training objective error are set.
C, to arrange training data be input matrix P, and Offered target value is matrix T, and the train function called in MatlabR2009a Neural Network Toolbox carries out data training to BP neural network Net until convergence, Net=train (Net, P, T).
Step (5) test b P neural network.
The BP neural network trained is tested, historical data composition is used for electromagnetic induction heating temperature prediction network test matrix P_test, directly call the sim function in MatlabR2009a Neural Network Toolbox, D=sim (Net, P_test), emulate test matrix, wherein D is objective function.
The process of step (6) data renormalization.
To the steel billet temperature after the electromagnetic induction heating of test gained according to formula
carry out renormalization process, wherein
for the steel billet temperature after renormalization process,
for the steel billet temperature that emulation testing obtains,
for steel billet maximum temperature,
for steel billet minimum temperature.
The present invention will utilize the simulation calculation ability of computing machine, build error back propagation (BP) neural network, automatic acquisition knowledge from induction heater operation history data, step by step new knowledge is attached in its mapping function, thus realize approaching of nonlinear function, Accurate Prediction can be carried out to the steel billet temperature after inductor heating.
The beneficial effect of the inventive method:
1. BP neural network has the ability of approaching nonlinear mapping function, and therefore utilize electromagnetic induction heater operation history data to predict steel billet temperature, the precision of prediction obtained than Traditional project computing method is high.
2. the hot continuous rolling electromagnetic induction steel billet temperature prediction based on BP neural network avoids the finite element model solving tradition and be coupled with temperature field based on the electromagnetic field that induction heating, heat transfer, heat radiation theory are set up, also avoid this finite element model of computer solving to the rigors of the basic datas such as boundary condition simultaneously, only need utilize a large amount of electromagnetic induction heating history datas, practicality is stronger.
3. increase the length of BP neural metwork training data, the dimension of input matrix, improve the computing velocity of computing machine, can precision of prediction be improved.
Accompanying drawing explanation
Fig. 1 is hot continuous rolling induction heating schematic diagram.
Embodiment
For the steel billet induction heating of certain steel mill's hot continuous rolling production line, carry out the hot continuous rolling electromagnetic induction heating steel billet temperature forecast model modeling embodiment based on BP neural network.
Step (1) forecast model variables choice
Analysis on Mechanism.The output of model is the steel billet temperature after heating.According to law of conservation of energy, the energy that the energy before the energy=heating steel billet after heating steel billet+steel billet absorbs, the energy of wherein steel billet absorption is from induction heater.By known voltage, the electric current respectively recording induction heater of the PLC control system of electromagnetic induction heater, if steel billet is known through the time of induction heater, the voltage of induction heater, electric current, the product of time are exactly the energy that steel billet absorbs so in theory.That can determine neural network model based on above-mentioned Analysis on Mechanism is input as 3: the temperature before heating steel billet, the voltage and current of inductor.
Before this modeling, first pre-service is carried out to each sampled data, each sample (a steel billet induction heating process) data are integrated, wherein, temperature before heating steel billet and the temperature after heating are integrated into 100 values, according to the length of inductor and steel billet, voltage and current is integrated into 152 values.According to position correlation analysis, the steel billet temperature correspondence n-th+25 of n-th is to the electric current of the inductor of n+46 and magnitude of voltage, now get the electric current and voltage value of the temperature before the heating steel billet of n-th and correspondence thereof, as input matrix, temperature after the heating steel billet of n-th is as output matrix, so, each sample just can form 100 pairs of input and output matrixes.
Make A matrix be temperature before heating steel billet, B matrix is the voltage of induction heater, and C matrix is the electric current of induction heater, and D is the temperature after heating steel billet.Then input matrix: P=
, concrete arrangement mode is:
p=
, wherein
be the temperature before the heating steel billet in the n-th moment,
be the voltage of the induction heater in the n-th moment,
be the electric current of the induction heater in the n-th moment, like this, each row of matrix P form input matrix.
Corresponding output matrix: D=
Step (2) data normalization process
Utilize normalization formula
respectively input and output matrix is normalized.
For input matrix P, specific practice is normalized A, B, C tri-matrixes respectively.For A matrix:
A=
A matrix is normalized according to normalization formula, obtains
.Carry out similar process to B, C matrix successively again, the input matrix after normalized is:
=
.
For output matrix D=
,
for the temperature after heating steel billet,
for the minimum value of the temperature after heating steel billet,
for the maximal value of the temperature after heating steel billet,
for the temperature after heating steel billet after normalized.Input matrix after normalization is
.
Step (3) builds BP neural network
Build BP neural network framework, call the newff function in MatlabR2009a function library,
Net=newff（threshold,[5,1],’tansig’,’purelin’,trainlm）
Wherein Threshold is minimum value and the maximal value of defined matrix 45 input and output vectors of a 45*2; [5,1] represents that ground floor has 5 neurons, and the second layer has 1 neuron; Tansig is input layer transport function; Purelin is output layer transport function; Trainlm is the training function based on l-m algorithm.
Step (4) training BP neural network
A. initialization network
The net.initFcn initialization function deciding whole network.The parameter net.layer{i}.initFcn initialization function deciding each layer.Initwb function is according to initiation parameter (net.inputWeights{i, j}.initFcn) the initializes weights matrix of every one deck oneself and be biased, and initializes weights is set to rands usually, and concrete grammar is as follows:
net.layers{1}.initFcn = 'initwb';
net.inputWeights{1,1}.initFcn = 'rands';
net.layerWeights{2,1}.initFcn = 'rands';
net.biases{1,1}.initFcn = 'rands';
net.biases{2,1}.initFcn = 'rands';
net = init(net);
Net.IW{1,1} are the weight matrix of input layer to hidden layer
Net.LW{2,1} are the weight matrix between hidden layer and output layer; Net.b{1,1} are the threshold values vector of hidden layer, and net.b{2,1} are the threshold values of output contact;
B., network training number of times, training objective error are set and be used for show step number
net.trainParam.epochs=2000;
net.trainParam.goal=0.002;
net.trainParam.show=50;
Arranging network training number of times is 2000 steps, and training objective error is 0.0002, and display train epochs is 50 steps.
C. input matrix is utilized
be set to objective matrix
, by calling train function, net=train (net,
,
) carry out the temperature prediction network training after heating steel billet until convergence.
Step (5) network test
The historical data being used for testing is used for the matrix p_test of steel billet temperature network test according to the input matrix form composition in step (1), then is normalized according to step (2), the test matrix after normalization is
.Call the sim () function in Matlab tool box, the network trained is emulated.Calling program code is: D=sim (net,
); D matrix is sewage pumping station the water level of the frontal pool predicted value.
Step (6) renormalization process
To the temperature after the heating steel billet of test gained according to formula
carry out renormalization process, wherein
for steel billet temperature final after renormalization process,
for the steel billet temperature that emulation testing obtains,
for the maximal value of steel billet temperature,
for the minimum value of steel billet temperature.After renormalization, steel billet temperature is
, namely testing the steel billet temperature obtained is
.
Claims (1)
1., based on the hot continuous rolling electromagnetic induction heating temperature predicting method of BP neural network, it is characterized in that the method comprises the following steps:
Step 1. selects forecast model variable, specifically utilizes Analysis on Mechanism and prior imformation, selects the input/output variable of forecast model;
The detailed process of Analysis on Mechanism is: the output of model is the steel billet temperature after heating; According to law of conservation of energy, the energy that the energy before the energy=heating steel billet after heating steel billet+steel billet absorbs, the energy of wherein steel billet absorption is from induction heater; By known voltage, the electric current respectively recording induction heater of the PLC control system of electromagnetic induction heater, if steel billet is known through the time of induction heater, the voltage of induction heater, electric current, the product of time are exactly the energy that steel billet absorbs so in theory; Three are input as: the steel billet temperature before electromagnetic induction heating, the voltage of induction heater and the electric current of induction heater based on above-mentioned Analysis on Mechanism determination neural network model;
Before this modeling, first pre-service is carried out to each sampled data, each sample data is integrated, wherein, temperature after temperature before heating steel billet and heating is integrated into 100 values, according to the length of inductor and steel billet, voltage and current is integrated into 152 values; According to position correlation analysis, the steel billet temperature correspondence n-th+25 of n-th is to the electric current of the inductor of n+46 and magnitude of voltage, now get the electric current and voltage value of the temperature before the heating steel billet of n-th and correspondence thereof, as input matrix, temperature after the heating steel billet of n-th is as output matrix, so, each sample just forms 100 pairs of input and output matrixes;
Make A matrix be heating before steel billet temperature, B matrix is the voltage of induction heater, and C matrix is the electric current of induction heater, and D is the temperature after heating steel billet; Then input matrix: P=[A B C]
t, concrete arrangement mode is:
B
nbe the voltage of the induction heater in the n-th moment,
C
nbe the electric current of the induction heater in the n-th moment, like this, each row of matrix P form input matrix;
Corresponding output matrix: D=[d
1d
2d
100];
The process of step 2. data normalization, specifically:
Input packet in training sample is containing three, and order of magnitude difference is comparatively large, and for ensureing each factor par, convergence speedup speed, is normalized data, is converted into the value of [0,1] interval range
Wherein x
maxfor the maximal value in input data, x
minfor the minimum value in input data, x is input data,
for the value after the process of input data normalization;
Step 3. builds BP neural network framework, specifically:
The newff function called in MatlabR2009a Neural Network Toolbox sets up BP neural network, Net=newff (PR, [s
1, s
2..., s
i], { TF
1, TF
2..., TF
i, BTF, BLF, PF); Net is BP neural network framework, and PR is the span determined by greatest member and least member in input matrix, s
ibe i-th layer of neuronic number, TF
ibe the transport function of i-th layer, 1≤i≤N
1, N
1for the total number of plies of neural network, BTF is the training function of BP neural network, and BLF is weights and bias, and PF is network performance function;
Step 4. trains BP neural network, and concrete grammar is:
A, initialization BP neural network, the value assignment utilizing random function to produce, to weights and bias, is then called init function and is carried out initialization BP neural network;
B, network training number of times and training objective error are set;
C, to arrange training data be input matrix P, and Offered target value is matrix T, and the train function called in MatlabR2009a Neural Network Toolbox carries out data training to BP neural network Net until convergence, Net=train (Net, P, T);
Step 5. test b P neural network, specifically:
The BP neural network trained is tested, historical data composition is used for electromagnetic induction heating temperature prediction network test matrix P_test, directly call the sim function in MatlabR2009a Neural Network Toolbox, D=sim (Net, P_test), emulate test matrix, wherein D is objective function;
The process of step 6. data renormalization, specifically:
To the steel billet temperature after the electromagnetic induction heating of test gained according to formula
carry out renormalization process, wherein x ' is the steel billet temperature after renormalization process,
for the steel billet temperature that emulation testing obtains, x '
maxfor steel billet maximum temperature, x '
minfor steel billet minimum temperature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110307393.7A CN102393884B (en) | 2011-10-12 | 2011-10-12 | Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110307393.7A CN102393884B (en) | 2011-10-12 | 2011-10-12 | Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102393884A CN102393884A (en) | 2012-03-28 |
| CN102393884B true CN102393884B (en) | 2015-04-15 |
Family
ID=45861207
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110307393.7A Active CN102393884B (en) | 2011-10-12 | 2011-10-12 | Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102393884B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102662039A (en) * | 2012-04-17 | 2012-09-12 | 戴会超 | BP neutral network-based method for predicting dissolved oxygen saturation in water body |
| CN102719644B (en) * | 2012-06-29 | 2013-12-04 | 武汉大学 | Forecasting method of inner and outer wall temperature difference of 9% Cr martensitic steel thick wall pipeline in heat treatment |
| CN102779216B (en) * | 2012-07-30 | 2014-09-17 | 杭州电子科技大学 | Systematic identification method of electromagnetic induction heating process based on finite element model |
| CN102816917B (en) * | 2012-09-11 | 2014-12-17 | 河北沧海重工股份有限公司 | Position determining method for postweld heat treatment temperature equivalent points of inner walls of steel pipes with 9 percent of Cr |
| CN104007659B (en) * | 2014-05-28 | 2016-08-24 | 重庆科技学院 | BP neutral net implementation method in S7-300 series of PLC |
| CN104407642B (en) * | 2014-12-01 | 2016-09-07 | 杭州电子科技大学 | A kind of continuous casting billet sensing heating process temp. control method controlled based on iterative learning |
| CN107729694B (en) * | 2017-11-17 | 2020-09-25 | 电子科技大学 | Multi-parameter electromagnetic field modeling simulation method based on neural network |
| CN109033505A (en) * | 2018-06-06 | 2018-12-18 | 东北大学 | A kind of ultrafast cold temprature control method based on deep learning |
| WO2021229645A1 (en) * | 2020-05-11 | 2021-11-18 | 東芝三菱電機産業システム株式会社 | Inductive heating method and inductive heating system |
| CN113849020B (en) * | 2021-09-22 | 2022-04-19 | 北京科技大学 | Billet heating curve design method and device based on artificial intelligence algorithm |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1588346A (en) * | 2004-08-30 | 2005-03-02 | 邢台钢铁有限责任公司 | Method for predicting converter terminal point using artificial nurve network technology |
| CN1979496A (en) * | 2005-12-02 | 2007-06-13 | 中国科学院金属研究所 | Copper-alloy pipe-material casting-milling technology parameter designing and optimizing method |
| CN101850410A (en) * | 2010-06-22 | 2010-10-06 | 攀钢集团钢铁钒钛股份有限公司 | Continuous casting breakout prediction method based on neural network |
-
2011
- 2011-10-12 CN CN201110307393.7A patent/CN102393884B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1588346A (en) * | 2004-08-30 | 2005-03-02 | 邢台钢铁有限责任公司 | Method for predicting converter terminal point using artificial nurve network technology |
| CN1979496A (en) * | 2005-12-02 | 2007-06-13 | 中国科学院金属研究所 | Copper-alloy pipe-material casting-milling technology parameter designing and optimizing method |
| CN101850410A (en) * | 2010-06-22 | 2010-10-06 | 攀钢集团钢铁钒钛股份有限公司 | Continuous casting breakout prediction method based on neural network |
Non-Patent Citations (3)
| Title |
|---|
| 基于BP 神经网络的热轧带钢卷取温度预报;马丽坤等;《钢铁研究学报》;20061130;第18卷(第11期);第27-30页 * |
| 基于MATLAB神经网络工具箱的BP网络实现;罗成汉;《计算机仿真》;20040531;第21卷(第5期);第109-115页 * |
| 归一化问题;王红旗;《科学网》;20091106;第1-3页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102393884A (en) | 2012-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102393884B (en) | Hot continuous rolling electromagnetic induction heating temperature prediction method based on BP (back-propagation) neural network | |
| CN102937784A (en) | Artificial neural network based method for controlling online prediction of casting billet quality | |
| CN102682345B (en) | Based on the traffic flow forecasting method of two variable optimal learning rate Fast Learning neural network | |
| CN102129259B (en) | Neural network proportion integration (PI)-based intelligent temperature control system and method for sand dust environment test wind tunnel | |
| CN107909220A (en) | Electric heating load prediction method | |
| CN105095962A (en) | Method for predicting dynamic mechanical property of material based on BP artificial neural network | |
| CN105204333B (en) | A kind of energy consumption Forecasting Methodology for improving iron and steel enterprise's energy utilization rate | |
| CN104636823B (en) | A kind of wind power forecasting method | |
| CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
| CN104517162A (en) | On-line hardness forecasting method of continuous annealing product by means of integrated learning | |
| CN102880908B (en) | Method for calculating remanufactured part environmental loss based on back propagation (BP) neural network | |
| CN107609712A (en) | A kind of steer method and apparatus based on classification prediction | |
| CN106203683A (en) | A kind of modeling method of power customer electro-load forecast system | |
| CN106779165A (en) | Power system environment dispatching method based on urban air-quality Predicting Technique | |
| CN107704962A (en) | A kind of smelter steam flow interval prediction method based on imperfect time series data collection | |
| CN109725537A (en) | A kind of improved linear optimal semi-active control method | |
| CN109886471A (en) | Fired power generating unit load distribution method based on neural network and intelligent optimization algorithm | |
| CN109598381A (en) | A kind of Short-time Traffic Flow Forecasting Methods based on state frequency Memory Neural Networks | |
| CN103823504A (en) | Maximum power tracing control method based on least squares support vector machine | |
| CN102564644A (en) | Temperature online measuring method for plate blank in production process of heating furnace | |
| CN112084701A (en) | System transient temperature prediction method based on data driving | |
| CN102789542B (en) | Load distribution method for hydrothermal power system based on multi-target distribution estimation | |
| CN112016754A (en) | Power station boiler exhaust gas temperature advanced prediction system and method based on neural network | |
| CN105094053A (en) | Machine tool thermal error modeling method based on ant colony neural network | |
| CN107292029A (en) | A kind of determination method that sheet forming technological parameter is predicted based on forming defects |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20190904 Address after: 311305 No. 59 South Ring Road, Qingshan Lake Street, Linan District, Hangzhou City, Zhejiang Province Patentee after: Hangzhou Sida Electric Cooker Complete Plant Co., Ltd. Address before: Hangzhou City, Zhejiang province 310018 Xiasha Higher Education Park No. 2 street Patentee before: Hangzhou Electronic Science and Technology Univ |