CN116523819A - Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network - Google Patents

Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network Download PDF

Info

Publication number
CN116523819A
CN116523819A CN202211515853.XA CN202211515853A CN116523819A CN 116523819 A CN116523819 A CN 116523819A CN 202211515853 A CN202211515853 A CN 202211515853A CN 116523819 A CN116523819 A CN 116523819A
Authority
CN
China
Prior art keywords
convolution
layer
error
parameter
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.)
Pending
Application number
CN202211515853.XA
Other languages
Chinese (zh)
Inventor
李帷韬
童倩倩
孙伟
李奇越
李俊辰
顾嘉钦
黄鑫兴
张心茹
刘威
吕顺辰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hefei University of Technology
Original Assignee
Hefei University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hefei University of Technology filed Critical Hefei University of Technology
Priority to CN202211515853.XA priority Critical patent/CN116523819A/en
Publication of CN116523819A publication Critical patent/CN116523819A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Quality & Reliability (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a magnesia furnace abnormal condition identification method and device based on a random configuration convolution network, wherein the method comprises the following steps: acquiring a magnesium furnace image training data set; setting parameters of a convolution network and initializing; configuring an L-th convolution parameter of a Layer convolution of a convolution network, judging whether the parameter meets a configuration condition, and if so, constructing a convolution parameter matrix; generating a feature map by using the convolution parameter matrix; pooling the feature images to generate a downsampled feature image; updating input data by utilizing a downsampling feature map, calculating the error of the predicted output O and the real output T of the random configuration convolution network, if the error is larger than an expected error, continuing to configure the parameters of the convolution network, and if the error is smaller than or equal to the expected error, completing the construction of the convolution network, and detecting the abnormal working condition of the magnesium furnace to be detected; the invention has the advantages that: the method has high accuracy in identifying the abnormal working condition of the magnesium furnace, and is convenient for on-site deployment.

Description

Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network
Technical Field
The invention relates to the field of intelligent control of industrial magnesium smelting, in particular to a method and a device for identifying abnormal working conditions of a magnesium furnace based on a random configuration convolution network.
Background
At present, the method for producing the fused magnesia by the fused magnesia enterprises in China mainly utilizes a three-phase alternating current fused magnesia furnace (called as fused magnesia furnace for short) to heat and smelt powdery raw materials mainly comprising magnesite by electric arc. In the smelting process of the electric smelting magnesium furnace, the consumed active power is mainly influenced by smelting current, and the smelting power of the electric smelting magnesium furnace can be changed by adjusting the smelting current.
Because the fault data of the electric smelting magnesium furnace cannot be directly measured due to the temperature of the molten liquid of the ultrahigh temperature electric smelting magnesium furnace, most of the existing under-burning abnormal working conditions are diagnosed by regularly observing the characteristics of the furnace shell by field workers, and a small amount of researches adopt various mature algorithms for automatic monitoring. For example, chinese patent publication No. CN115033724a discloses a method for identifying abnormal conditions of an electric melting magnesium furnace based on a multi-mode modulation algorithm, which comprises: collecting historical data of the magnesium melting furnace, and carrying out data screening; carrying out normalization processing on the data samples; constructing a multi-mode modulation neural network MNN and initializing network parameters; training a multi-modal modulation neural network MNN; and identifying abnormal working conditions of the electric smelting magnesium furnace based on multi-mode understanding to obtain an identification result. The method and the device can realize the diagnosis of the abnormal working condition of the electric smelting magnesium furnace, evaluate the health state of the electric smelting magnesium furnace and provide maintenance advice of the electric smelting magnesium furnace, ensure the safe, efficient and stable operation of the factory, and further improve the accuracy of the identification of the abnormal working condition and the stability of the factory production. However, the patent application adopts multi-mode fault monitoring, relies on a large amount of historical process measurement data, is affected by performance attenuation, faults, external interference and the like of the on-site detection instrument and the system, so that the measurement data is inaccurate, the accuracy of subsequent modeling is affected, the model is easy to fall into local optimum, and the built model has low accuracy in identifying abnormal working conditions of the magnesia furnace; and the text characteristic sequence of the abnormal fault of the electric smelting magnesium furnace required by the multi-mode network is not easy to obtain, so that the electric smelting magnesium furnace is difficult to deploy on site.
Disclosure of Invention
The invention aims to solve the technical problems that the identification method for the abnormal working condition of the magnesium furnace in the prior art is low in accuracy, the text characteristic sequence of the abnormal fault of the electric smelting magnesium furnace is not easy to obtain and is difficult to deploy on site.
The invention solves the technical problems by the following technical means: a magnesium furnace abnormal condition identification method based on a random configuration convolution network comprises the following steps:
step a: acquiring a magnesium furnace image training data set;
step b: setting parameters of a convolution network and initializing;
step c: configuring an L-th convolution parameter of a Layer convolution of a convolution network, judging whether the parameter meets a configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met;
step d: generating a feature map by using the convolution parameter matrix;
step e: pooling the feature images to generate a downsampled feature image;
step f: and updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, if the error is larger than the expected error, continuing to configure the parameters of the convolution network, and if the error is smaller than or equal to the expected error, completing the construction of the convolution network, and detecting the abnormal working condition of the magnesium furnace to be detected.
The beneficial effects are that: the method configures the convolution parameters of the convolution network, the next operation is performed only when the configuration conditions are met, the reliability of the model parameters is guaranteed to a certain extent, the errors of the prediction output O and the real output T of the convolution network are calculated, the parameters of the convolution network are continuously configured under the condition that the errors exceed expected errors, the constructed convolution network prediction errors are avoided to be overlarge, the accuracy of the output result of the convolution network is improved, and therefore the accuracy of identifying abnormal working conditions of the magnesium furnace is high.
Further, the step a includes:
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples, x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
Further, the step b includes:
setting parameters of convolution network, and maximizing the number Layer of convolution layers max Maximum convolution number L per layer max Expected error ε, convolution kernel size k;
initializing a convolution network, wherein the convolution times L=0, the number of convolution layers is layer=0, and the input I=X and the error are input
Still further, the step c includes:
step c01: cutting the input data I into input data I' with the size of k multiplied by k;
step c02: randomly selecting the weight W by using a formula (1) L,Layer Bias b L,Layer ；
Wherein λ is a configuration fixed parameter, rand (a, b) is a random function of generating a matrix of a×b with uniformly distributed random samples over [0,1 ], C in The number of channels for the input data I;
step c03: calculating h using equation (2) L,Layer ，
h L,Layer ＝g(W L,Layer ·I′+b L,Layer ) (2)
Wherein g (·) is an activation function, · is a matrix multiplication;
step c04: calculating constraint ζ using equation (3) L,Layer If xi L,Layer If the parameter is less than 0, the parameter does not meet the configuration condition, let r=r+τ and return to step c01 to reselect the convolution kernel parameter, otherwise, continuing to step c05;
wherein e T Transpose of error e, h L,Layer T Is h L,Layer Is transposed of r is constraint parameter, 0<r<1, tau is the updating step length of the configuration fixed parameter;
step c05: constructing a weighting parameter matrix W of a Layer Layer Layer ＝[W 1,Layer ,W 2,Layer ,...,W L,Layer ]Bias parameter matrix b Layer ＝[b 1,Layer ,...,b L,Layer ]The weight parameter matrix and the bias parameter matrix form a convolution parameter matrix.
Still further, the step d includes:
performing convolution operation on the training data set of the magnesium furnace image by using a convolution parameter matrix and adopting a formula (4) to generate a feature map F L,Layer ；
Wherein I is i,k For the kth channel matrix of the ith input data, i=0, 1, N, k=1, 2, C in Is a cross-correlation operator.
Still further, the step e includes:
the characteristic diagram F is compared with the formula (5) L,Layer Performing pooling operation to generate a downsampled feature map F' L,Layer ；
Wherein F is i L,Layer I=1..n is the feature map of the i-th input data, h=0, 1..h-1, H is the feature map F L,Layer W=0, 1.., W-1, W is the number of rows of the feature map F L,Layer Is a column number of columns.
Further, in the step f, updating the input data using the downsampled feature map includes:
step f01: updating input data I to I g ，I g ＝[F′ 1,Layer ,...,F′ L,Layer ]Let l=l+1;
step f02: calculating an output layer parameter beta of the convolution network by using a formula (6);
wherein,,is a pseudo-inverse operator;
step f03: calculating a predicted output O of the convolutional network by using a formula (7);
O＝I g ·β (7)
step f04: updating the error e using equation (8);
wherein t is i To output the ith component of T, o i To predict the ith component of output O.
Further, in the step f, calculating the error between the predicted output O and the real output T of the randomly configured convolutional network, if the error is greater than the expected error, continuing to configure the convolutional network parameters, if the error is less than or equal to the expected error, completing the construction of the convolutional network, including:
step f05: judging whether the error e is larger than the expected error epsilon, if e is larger than epsilon, continuing the step f06, otherwise, completing the construction of the convolution network, and jumping to the step f08;
step f06: judging whether the convolution times L is less than or equal to the maximum convolution times L max If L < = L max Step c01 is skipped to continue to configure the convolution parameters, otherwise, layer=layer+1, l=1;
step f07: judging whether the number of the convolution layers is less than or equal to the maximum number of the convolution layers max If Layer < = Layer max Step c01 is skipped to continue to configure the convolution parameters, otherwise, the convolution network is constructed, and step f08 is skipped;
step f08: the constructed convolution network is used for detecting abnormal working conditions of the magnesium furnace to be detected.
The invention also provides a device for identifying abnormal working conditions of the magnesium furnace based on the random configuration convolution network, which comprises the following steps:
the data set acquisition module is used for acquiring a magnesium furnace image training data set;
the network initialization module is used for setting parameters of the convolutional network and initializing;
the parameter configuration module is used for configuring the L-th convolution parameter of the Layer convolution of the convolution network, judging whether the parameter meets the configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met;
the feature map generating module is used for generating a feature map by utilizing the convolution parameter matrix;
the pooling operation module is used for pooling the feature images to generate downsampled feature images;
the error judging module is used for updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, continuing to configure the parameters of the convolution network if the error is larger than the expected error, and completing the construction of the convolution network if the error is smaller than or equal to the expected error, and detecting the abnormal working condition of the magnesium furnace to be detected.
Further, the data set acquisition module is further configured to:
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples, x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
Further, the network initialization module is further configured to:
setting parameters of convolution network, and maximizing the number Layer of convolution layers max Maximum convolution number L per layer max Expected error ε, convolution kernel size k;
initializing a convolution network, wherein the convolution times L=0, the number of convolution layers is layer=0, and the input I=X and the error are input
Still further, the parameter configuration module is further configured to:
step c01: cutting the input data I into input data I' with the size of k multiplied by k;
step c02: randomly selecting the weight W by using a formula (1) L,Layer Bias b L,Layer ；
Wherein λ is a configuration fixed parameter, rand (a, b) is a random function of generating a matrix of a×b with uniformly distributed random samples over [0,1 ], C in The number of channels for the input data I;
step c03: calculating h using equation (2) L,Layer ，
h L,Layer ＝g(W L,Layer ·I′+b L,Layer ) (10)
Wherein g (·) is an activation function, · is a matrix multiplication;
step c04: calculating constraint ζ using equation (3) L,Layer If xi L,Layer If the parameter is less than 0, the parameter does not meet the configuration condition, let r=r+τ and return to step c01 to reselect the convolution kernel parameter, otherwise, continuing to step c05;
wherein e T Transpose of error e, h L,Layer T is h L,Layer Is transposed of r is constraint parameter, 0<r<1, tau is the updating step length of the configuration fixed parameter;
step c05: constructing a weighting parameter matrix W of a Layer Layer Layer ＝[W 1,Layer ,W 2,Layer ,...,W L,Layer ]Bias parameter matrix b Layer ＝[b 1,Layer ,...,b L,Layer ]The weight parameter matrix and the bias parameter matrix form a convolution parameter matrix.
Still further, the feature map generating module is further configured to:
performing convolution operation on the training data set of the magnesium furnace image by using a convolution parameter matrix and adopting a formula (4) to generate a feature map F L,Layer ；
Wherein I is i,k For the kth channel matrix of the ith input data, i=0, 1, N, k=1, 2, C in Is a cross-correlation operator.
Still further, the pooling operation module is further configured to:
the characteristic diagram F is compared with the formula (5) L,Layer Performing pooling operation to generate a downsampled feature map F' L,Layer ；
Wherein F is i L,Layer I=1..n is the feature map of the i-th input data, h=0, 1..h-1, H is the feature map F L,Layer W=0, 1.., W-1, W is the number of rows of the feature map F L,Layer Is a column number of columns.
Further, the error determination module updates the input data by using the downsampled feature map, including:
step f01: updating input data I to I g ，I g ＝[F′ 1,Layer ,...,F′ L,Layer ]Let l=l+1;
step f02: calculating an output layer parameter beta of the convolution network by using a formula (6);
wherein,,is a pseudo-inverse operator;
step f03: calculating a predicted output O of the convolutional network by using a formula (7);
O＝I g ·β (15)
step f04: updating the error e using equation (8);
wherein t is i To output the ith component of T, o i To predict the ith component of output O.
Further, the error judging module calculates the error between the predicted output O and the real output T of the randomly configured convolutional network, if the error is greater than the expected error, the convolutional network parameter is continuously configured, if the error is less than or equal to the expected error, the convolutional network construction is completed, including:
step f05: judging whether the error e is larger than the expected error epsilon, if e is larger than epsilon, continuing the step f06, otherwise, completing the construction of the convolution network, and jumping to the step f08;
step f06: judging whether the convolution times L is less than or equal to the maximum convolution times L max If L < = L max Step c01 is skipped to continue to configure the convolution parameters, otherwise, layer=layer+1, l=1;
step f07: judging whether the number of the convolution layers is less than or equal to the maximum number of the convolution layers max If Layer < = Layer max Step c01 is skipped to continue to configure the convolution parameters, otherwise, the convolution network is constructed, and step f08 is skipped;
step f08: the constructed convolution network is used for detecting abnormal working conditions of the magnesium furnace to be detected.
The invention has the advantages that:
(1) The method configures the convolution parameters of the convolution network, the next operation is performed only when the configuration conditions are met, the reliability of the model parameters is guaranteed to a certain extent, the errors of the prediction output O and the real output T of the convolution network are calculated, the parameters of the convolution network are continuously configured under the condition that the errors exceed expected errors, the constructed convolution network prediction errors are avoided to be overlarge, the accuracy of the output result of the convolution network is improved, and therefore the accuracy of identifying abnormal working conditions of the magnesium furnace is high.
(2) Because the magnesium furnace image size is larger, the parameters of the traditional fully-connected network are more easy to be over-fitted, the characteristic characterization in the magnesium furnace image is extracted through the convolution operation formula (4), and the characteristic map is compressed by using the maximum pooling operation formula (5), so that the parameter number in the network is reduced, the operation speed is increased, and the model is prevented from being over-fitted.
Drawings
Fig. 1 is a flowchart of a method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to embodiment 1 of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described in the following in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1, the method for identifying abnormal working conditions of the magnesium furnace based on a randomly configured convolution network comprises the following steps:
s1: acquiring a magnesium furnace image training data set;
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
S2: setting parameters of a random configuration convolution network and initializing; the specific process of the steps is as follows:
setting parameters of random configuration convolutional network, and maximizing the number Layer of convolutional layers max Maximum convolution number L per layer max Expected error ε, convolution kernel size k;
initializing a random configuration convolution network, wherein the convolution times L=0, the number of convolution layers is layer=0, and I=X is input to the system, so that errors are generated
S3: configuring an L-th convolution parameter of a Layer convolution of a convolution network, judging whether the parameter meets a configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met; the specific process of S3 is as follows:
s31: cutting the input data I into input data I' with the size of k multiplied by k;
s32: randomly selecting the weight W by using a formula (1) L,Layer Bias b L,Layer ；
Wherein λ is a configuration fixed parameter, rand (a, b) is a random function of generating a matrix of a×b with uniformly distributed random samples over [0,1 ], C in The number of channels for the input data I;
s33: calculating h using equation (2) L,Layer ，
h L,Layer ＝g(W L,Layer ·I′+b L,Layer ) (18)
Wherein g (·) is an activation function, · is a matrix multiplication;
s34: calculating constraint ζ using equation (3) L,Layer And judge xi L,Layer If is greater than 0, if ζ L,Layer If the parameter is less than 0, the parameter does not meet the configuration condition, let r=r+τ and return to step S31 to reselect the convolution kernel parameter, otherwise, continue to step S35;
wherein e T Transpose of error e, h L,Layer T Is h L,Layer Is transposed of r is constraint parameter, 0<r<1, tau is the updating step length of the configuration fixed parameter;
s35: constructing a weighting parameter matrix W of a Layer Layer Layer ＝[W 1,Layer ,W 2,Layer ,...,W L,Layer ]Bias parameter matrix b Layer ＝[b 1,Layer ,...,b L,Layer ]The weight parameter matrix and the bias parameter matrix form a convolution parameter matrix.
S4: generating a feature map by using the convolution parameter matrix; the specific process of the step is as follows:
carrying out convolution operation on the training data set of the magnesium furnace image by using a convolution parameter matrix and adopting a formula (4) to generateCharacterization map F L,Layer ；
Wherein I is i,k For the kth channel matrix of the ith input data, i=0, 1, N, k=1, 2, C in Is a cross-correlation operator.
S5: pooling the feature images to generate a downsampled feature image; the specific process of the step is as follows:
the characteristic diagram F is compared with the formula (5) L,Layer Performing pooling operation to generate a downsampled feature map F' L,Layer ；
Wherein F is i L,Layer I=1..n is the feature map of the i-th input data, h=0, 1..h-1, H is the feature map F L,Layer W=0, 1.., W-1, W is the number of rows of the feature map F L,Layer Is a column number of columns.
S6: and updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, if the error is larger than the expected error, continuing to configure the parameters of the convolution network, and if the error is smaller than or equal to the expected error, completing the construction of the convolution network, and detecting the abnormal working condition of the magnesium furnace to be detected. The specific process of the step is as follows:
s61: updating input data I to I g ，I g ＝[F′ 1,Layer ,...,F′ L,Layer ]Let l=l+1;
s62: calculating an output layer parameter beta of the convolution network by using a formula (6);
wherein,,is a pseudo-inverse operator;
s63: calculating a predicted output O of the convolutional network by using a formula (7);
O＝I g ·β (23)
step S64: updating the error e using equation (8);
wherein t is i To output the ith component of T, o i To predict the ith component of output O.
Step S65: judging whether the error e is larger than the expected error epsilon, if e is larger than epsilon, continuing to step S66, otherwise, completing the construction of the convolution network, and jumping to step S68;
step S66: judging whether the convolution times L is less than or equal to the maximum convolution times L max If L < = L max Step S31 is skipped to continue configuration of the convolution parameters, otherwise, let layer=layer+1, l=1;
step S67: judging whether the number of the convolution layers is less than or equal to the maximum number of the convolution layers max If Layer < = Layer max Step S31 is skipped to continue configuration of convolution parameters, otherwise, the convolution network is constructed, and step S68 is skipped;
step S68: the constructed convolution network is used for detecting abnormal working conditions of the magnesium furnace to be detected.
By the technical scheme, the convolution parameters of the convolution network are configured, the next operation is performed after the configuration conditions are met, the reliability of the model parameters is guaranteed to a certain extent, the errors of the predicted output O and the real output T of the convolution network are calculated, the parameters of the convolution network are continuously configured under the condition that the errors exceed expected errors, the prediction errors of the constructed convolution network are prevented from being overlarge, the accuracy of the output result of the convolution network is improved, and therefore the accuracy of identifying abnormal working conditions of the magnesium furnace is high.
Example 2
Based on embodiment 1, embodiment 2 of the present invention further provides a device for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network, where the device includes:
the data set acquisition module is used for acquiring a magnesium furnace image training data set;
the network initialization module is used for setting parameters of the convolutional network and initializing;
the parameter configuration module is used for configuring the L-th convolution parameter of the Layer convolution of the convolution network, judging whether the parameter meets the configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met;
the feature map generating module is used for generating a feature map by utilizing the convolution parameter matrix;
the pooling operation module is used for pooling the feature images to generate downsampled feature images;
the error judging module is used for updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, continuing to configure the parameters of the convolution network if the error is larger than the expected error, and completing the construction of the convolution network if the error is smaller than or equal to the expected error, and detecting the abnormal working condition of the magnesium furnace to be detected.
Specifically, the data set acquisition module is further configured to:
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples, x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
Specifically, the network initialization module is further configured to:
setting parameters of convolution network, and maximizing the number Layer of convolution layers max Maximum convolution number L per layer max Expected error ε, convolution kernel size k;
initializing a convolution network, wherein the convolution times L=0, the number of convolution layers is layer=0, and the input I=X and the error are input
More specifically, the parameter configuration module is further configured to:
step c01: cutting the input data I into input data I' with the size of k multiplied by k;
step c02: randomly selecting the weight W by using a formula (1) L,Layer Bias b L,Layer ；
Wherein λ is a configuration fixed parameter, rand (a, b) is a random function of generating a matrix of a×b with uniformly distributed random samples over [0,1 ], C in The number of channels for the input data I;
step c03: calculating h using equation (2) L,Layer ，
h L,Layer ＝g(W L,Layer ·I′+b L,Layer ) (26)
Wherein g (·) is an activation function, · is a matrix multiplication;
step c04: calculating constraint ζ using equation (3) L,Layer If xi L,Layer If the parameter is less than 0, the parameter does not meet the configuration condition, let r=r+τ and return to step c01 to reselect the convolution kernel parameter, otherwise, continuing to step c05;
wherein e T Transpose of error e, h L,Layer T Is h L,Layer Is transposed of r is constraint parameter, 0<r<1, tau is the updating step length of the configuration fixed parameter;
step c05: constructing a weighting parameter matrix W of a Layer Layer Layer ＝[W 1,Layer ,W 2,Layer ,...,W L,Layer ]Bias parameter matrix b Layer ＝[b 1,Layer ,...,b L,Layer ]The weight parameter matrix and the bias parameter matrix form a convolution parameter matrix.
More specifically, the feature map generating module is further configured to:
performing convolution operation on the training data set of the magnesium furnace image by using a convolution parameter matrix and adopting a formula (4) to generate a feature map F L,Layer ；
Wherein I is i,k For the kth channel matrix of the ith input data, i=0, 1, N, k=1, 2, C in Is a cross-correlation operator.
More specifically, the pooling operation module is further configured to:
the characteristic diagram F is compared with the formula (5) L,Layer Performing pooling operation to generate a downsampled feature map F' L,Layer ；
Wherein F is i L,Layer I=1..n is the feature map of the i-th input data, h=0, 1..h-1, H is the feature map F L,Layer W=0, 1.., W-1, W is the number of rows of the feature map F L,Layer Is a column number of columns.
More specifically, the updating the input data by using the downsampling feature map in the error determination module includes:
step f01: updating input data I to I g ，I g ＝[F′ 1,Layer ,...,F′ L,Layer ]Let l=l+1;
step f02: calculating an output layer parameter beta of the convolution network by using a formula (6);
wherein,,is a pseudo-inverse operator; />
Step f03: calculating a predicted output O of the convolutional network by using a formula (7);
O＝I g ·β (31)
step f04: updating the error e using equation (8);
wherein t is i To output the ith component of T, o i To predict the ith component of output O.
More specifically, the error judging module calculates the error between the predicted output O and the real output T of the randomly configured convolutional network, if the error is greater than the expected error, the convolutional network parameter is continuously configured, if the error is less than or equal to the expected error, the convolutional network construction is completed, including:
step f05: judging whether the error e is larger than the expected error epsilon, if e is larger than epsilon, continuing the step f06, otherwise, completing the construction of the convolution network, and jumping to the step f08;
step f06: judging whether the convolution times L is less than or equal to the maximum convolution times L max If L < = L max Step c01 is skipped to continue to configure the convolution parameters, otherwise, layer=layer+1, l=1;
step f07: judging whether the number of the convolution layers is less than or equal to the maximum number of the convolution layers max If Layer < = Layer max Step c01 is skipped to continue to configure the convolution parameters, otherwise, the convolution network is constructed, and step f08 is skipped;
step f08: the constructed convolution network is used for detecting abnormal working conditions of the magnesium furnace to be detected.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. The method for identifying the abnormal working condition of the magnesium furnace based on the random configuration convolution network is characterized by comprising the following steps:
step a: acquiring a magnesium furnace image training data set;
step b: setting parameters of a convolution network and initializing;
step c: configuring an L-th convolution parameter of a Layer convolution of a convolution network, judging whether the parameter meets a configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met;
step d: generating a feature map by using the convolution parameter matrix;
step e: pooling the feature images to generate a downsampled feature image;
step f: and updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, if the error is larger than the expected error, continuing to configure the parameters of the convolution network, and if the error is smaller than or equal to the expected error, completing the construction of the convolution network, and detecting the abnormal working condition of the magnesium furnace to be detected.
2. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 1, wherein the step a comprises:
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples, x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
3. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 1, wherein the step b comprises:
setting parameters of convolution network, and maximizing the number Layer of convolution layers max Maximum convolution number L per layer max Expected error ε, convolution kernel size k;
initializing a convolution network, wherein the convolution times L=0, the number of convolution layers is layer=0, and the input I=X and the error are input
4. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 3, wherein the step c comprises:
step c01: cutting the input data I into input data I' with the size of k multiplied by k;
step c02: randomly selecting the weight W by using a formula (1) L,Layer Bias b L,Layer ；
Wherein λ is a configuration fixed parameter, rand (a, b) is a random function of generating a matrix of a×b with uniformly distributed random samples over [0,1 ], C in The number of channels for the input data I;
step c03: calculating h using equation (2) L,Layer ，
h L,Layer ＝g(W L,Layer ·I′+b L,Layer ) (34)
Wherein g (·) is an activation function, · is a matrix multiplication;
step c04: calculating constraint ζ using equation (3) L,Layer If xi L,Layer If the parameter is less than 0, the parameter does not meet the configuration condition, let r=r+τ and return to step c01 to reselect the convolution kernel parameter, otherwise, continuing to step c05;
wherein e T Transpose of error e, h L,Layer T Is h L,Layer Is transposed of r is constraint parameter, 0<r<1, tau is the updating step length of the configuration fixed parameter;
step c05: constructing a weighting parameter matrix W of a Layer Layer Layer ＝[W 1,Layer ,W 2,Layer ,...,W L,Layer ]Bias parameter matrix b Layer ＝[b 1,Layer ,...,b L,Layer ]The weight parameter matrix and the bias parameter matrix form a convolution parameter matrix.
5. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 4, wherein the step d comprises:
performing convolution operation on the training data set of the magnesium furnace image by using a convolution parameter matrix and adopting a formula (4) to generate a feature map F L,Layer ；
Wherein I is i,k For the kth channel matrix of the ith input data, i=0, 1, N, k=1, 2, C in Is a cross-correlation operator.
6. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 5, wherein the step e comprises:
the characteristic diagram F is compared with the formula (5) L,Layer Performing pooling operation to generate a downsampled feature map F' L,Layer ；
Wherein F is i L,Layer I=1..n is the feature map of the i-th input data, h=0, 1..h-1, H is the feature map F L,Layer W=0, 1.., W-1, W is the number of rows of the feature map F L,Layer Is a column number of columns.
7. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 6, wherein the updating the input data in the step f by using the downsampling feature map comprises:
step f01: updating input data I to I g ，I g ＝[F′ 1，Layer ,...,F′ L,Layer ]Let l=l+1;
step f02: calculating an output layer parameter beta of the convolution network by using a formula (6);
wherein,,is a pseudo-inverse operator;
step f03: calculating a predicted output O of the convolutional network by using a formula (7);
O＝I g ·β (39)
step f04: updating the error e using equation (8);
wherein t is i To output the ith component of T, o i To predict the ith component of output O.
8. The method for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 7, wherein in the step f, an error between a predicted output O and a true output T of the randomly configured convolutional network is calculated, if the error is greater than an expected error, parameters of the convolutional network are continuously configured, and if the error is less than or equal to the expected error, the construction of the convolutional network is completed, including:
step f05: judging whether the error e is larger than the expected error epsilon, if e is larger than epsilon, continuing the step f06, otherwise, completing the construction of the convolution network, and jumping to the step f08;
step f06: judging whether the convolution times L is less than or equal to the maximum convolution times L max If L < = L max Step c01 is skipped to continue to configure the convolution parameters, otherwise, layer=layer+1, l=1;
step f07: judging whether the number of the convolution layers is less than or equal to the maximum number of the convolution layers max If Layer < = Layer max Step c01 is skipped to continue to configure the convolution parameters, otherwise, the convolution network is constructed, and step f08 is skipped;
step f08: the constructed convolution network is used for detecting abnormal working conditions of the magnesium furnace to be detected.
9. The utility model provides a magnesium stove abnormal condition recognition device based on random configuration convolutional network which characterized in that, the device includes:
the data set acquisition module is used for acquiring a magnesium furnace image training data set;
the network initialization module is used for setting parameters of the convolutional network and initializing;
the parameter configuration module is used for configuring the L-th convolution parameter of the Layer convolution of the convolution network, judging whether the parameter meets the configuration condition, if so, constructing a convolution parameter matrix, and if not, reconfiguring the parameter until the configuration condition is met;
the feature map generating module is used for generating a feature map by utilizing the convolution parameter matrix;
the pooling operation module is used for pooling the feature images to generate downsampled feature images;
the error judging module is used for updating input data by utilizing the downsampling characteristic diagram, calculating the error of the predicted output O and the real output T of the random configuration convolution network, continuing to configure the parameters of the convolution network if the error is larger than the expected error, and completing the construction of the convolution network if the error is smaller than or equal to the expected error, and detecting the abnormal working condition of the magnesium furnace to be detected.
10. The device for identifying abnormal conditions of a magnesium furnace based on a randomly configured convolutional network according to claim 9, wherein the data set acquisition module is further configured to:
the training data set input is x= { X 1 ,x 2 ，...，x N Output is t= { T } 1 ,t 2 ，...，t N N is the number of samples, x i For the ith input sample, t i For the i-th output sample, where i=1, 2,..n.
CN202211515853.XA 2022-11-30 2022-11-30 Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network Pending CN116523819A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211515853.XA CN116523819A (en) 2022-11-30 2022-11-30 Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211515853.XA CN116523819A (en) 2022-11-30 2022-11-30 Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network

Publications (1)

Publication Number Publication Date
CN116523819A true CN116523819A (en) 2023-08-01

Family

ID=87392714

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211515853.XA Pending CN116523819A (en) 2022-11-30 2022-11-30 Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network

Country Status (1)

Country Link
CN (1) CN116523819A (en)

Similar Documents

Publication Publication Date Title
CN110220602B (en) Switch cabinet overheating fault identification method
CN104390657B (en) A kind of Generator Unit Operating Parameters measurement sensor fault diagnosis method and system
CN110559966B (en) Multi-mode combined control method and system for reaction kettle
EP2677383B1 (en) Process monitoring system and method
CN116111727B (en) Comprehensive distribution box abnormity monitoring method based on dynamic temperature threshold
WO2019234810A1 (en) Learning device, inference device, method, and program
CN104281179B (en) Controlling system and data collecting/displaying method
CN107423433A (en) A kind of data sampling rate control method and device
CN110245460B (en) Intermittent process fault monitoring method based on multi-stage OICA
KR20200039964A (en) Method and Apparatus for Operating Optimal of Equipment based on Machine Learning Model
CN116424096B (en) New energy automobile battery acquisition assembly method and system for dynamic resource optimization configuration
Omitaomu et al. Online support vector regression with varying parameters for time-dependent data
CN113048807A (en) Air cooling unit backpressure abnormity detection method
JP7182059B2 (en) Measurement value prediction module, measurement value prediction program, and measurement value prediction method
CN115640860A (en) Electromechanical equipment remote maintenance method and system for industrial cloud service
CN107222169A (en) A kind of photovoltaic array based on outlier detection is monitored and fault detection method in real time
Lutska et al. Forecasting the efficiency of the control system of the technological object on the basis of neural networks
CN116523819A (en) Method and device for identifying abnormal working conditions of magnesium furnace based on random configuration convolution network
CN111665882A (en) Temperature control method and system
EP3252557B1 (en) Method for detection of diagnosing control valve stiction
CN107766880A (en) The flexible measurement method of Fermentation by Photosynthetic Bacteria process key parameter based on BA LSSVM
KR102110319B1 (en) System for generating learning data
CN116505556A (en) Wind farm power control system and method based on primary frequency modulation
CN115700324A (en) Tile temperature early warning method and device and electronic equipment
CN116070734A (en) Transformer fault prediction method, device, equipment and storage medium

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