CN114372181A - Intelligent planning method for equipment production based on multi-mode data - Google Patents

Intelligent planning method for equipment production based on multi-mode data Download PDF

Info

Publication number
CN114372181A
CN114372181A CN202111609391.3A CN202111609391A CN114372181A CN 114372181 A CN114372181 A CN 114372181A CN 202111609391 A CN202111609391 A CN 202111609391A CN 114372181 A CN114372181 A CN 114372181A
Authority
CN
China
Prior art keywords
data
formula
neural network
deep neural
manufacturing process
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
CN202111609391.3A
Other languages
Chinese (zh)
Inventor
梁泽锋
余志文
杨楷翔
孟献兵
陈俊龙
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN202111609391.3A priority Critical patent/CN114372181A/en
Publication of CN114372181A publication Critical patent/CN114372181A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/906—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Databases & Information Systems (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Manufacturing & Machinery (AREA)
  • Probability & Statistics with Applications (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses an intelligent planning method for equipment production based on multi-modal data, which comprises the following steps: 1) acquiring data in the manufacturing process to obtain multi-modal data; 2) standardizing the multi-modal data; 3) building a training set for a single manufacturing step; 4) constructing a category-action mapping relation for the state of each manufacturing step; 5) constructing and training a deep neural network; 6) classifying the data collected in each manufacturing process by using the trained deep neural network; 7) selecting corresponding actions according to the classification result; 8) and (4) repeatedly classifying and selecting actions for each manufacturing step to finish action planning. The invention fully utilizes the multi-modal data and the deep neural network collected in the manufacturing process, not only can identify the current state of each step, but also can carry out action planning according to the current state, thereby realizing the intelligent action planning in the manufacturing process and reducing manual intervention. Furthermore, the interpretability and accuracy of the action plan can be increased.

Description

Intelligent planning method for equipment production based on multi-mode data
Technical Field
The invention relates to the technical field of computer artificial intelligence, in particular to an intelligent planning method for equipment production based on multi-mode data.
Background
At present, with the development of the 4.0 th era of industry, artificial intelligence begins to permeate the traditional manufacturing industry, and the manufacturing industry begins to change to the direction of intellectualization and informatization. In the process of intelligent manufacturing, a large amount of multi-modal data with different types and various sources can be generated. How to plan the manufacturing process by using the multi-modal data generated in the manufacturing process, reducing manual intervention, improving the interpretability and accuracy of production planning becomes a technical difficulty of advancing the manufacturing process to intellectualization, and is also a key development direction of the intelligent manufacturing research field at the present stage.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide an intelligent planning method for equipment production based on multi-modal data, which can effectively utilize the multi-modal data to plan equipment production, fully utilize the multi-modal data and a deep neural network collected in the manufacturing process, identify the current state of each step, plan actions according to the current state, realize intelligent planning of the actions in the manufacturing process, reduce manual intervention and increase the interpretability and accuracy of the action planning.
In order to achieve the purpose, the technical scheme provided by the invention is as follows: an intelligent planning method for equipment production based on multi-modal data comprises the following steps:
1) carrying out data acquisition on the manufacturing process by using different types of sensors to obtain multi-mode data, and storing the multi-mode data in a server;
2) extracting multi-modal data from the server, and dividing the data into two types of image data and numerical data for preprocessing to obtain preprocessed multi-modal data; aiming at image data, acquiring a pixel value matrix of the image data and carrying out standardization processing on the pixel value matrix; directly carrying out data standardization processing on numerical data;
3) for the preprocessed multi-modal data, extracting multi-modal data related to the manufacturing process t; then, manually labeling the data according to expert knowledge to obtain a label of each piece of data, wherein the extracted multi-modal data and the corresponding labels form a training set;
4) constructing a category-manufacturing action mapping relation, so that each category corresponds to one manufacturing action in the training set obtained in the step 3);
5) inputting the training set of the step 3) into a deep neural network for training; the method comprises the following steps that a ReLU function is selected as an activation function of a deep neural network, a loss function is a cross entropy loss function, and category probability output is carried out by adopting a softmax function; obtaining a trained deep neural network by optimizing a loss function;
6) during the production of actual equipment, different types of sensors are used for data acquisition to form a multi-modal data stream, and the multi-modal data stream is predicted by using a trained deep neural network to obtain a classification result;
7) selecting the manufacturing action corresponding to the classification result in the step 6) according to the category-manufacturing action mapping relation constructed in the step 4), and finishing the action planning of the actual manufacturing process t;
8) and (4) repeating the steps 3) to 7) for each actually operated manufacturing process to form the intelligent plan of the action strategy of the equipment production manufacturing process under the multi-mode data.
Further, in step 2), the data normalization process is: after the numerical data stored in the server are read, the data are respectively standardized according to the category, and the standardization formula of the numerical data is as follows:
Figure BDA0003434735500000021
in the formula, XaIs raw numerical data，
Figure BDA0003434735500000022
Is the average of the numerical data,
Figure BDA0003434735500000023
is the standard deviation of the numerical data,
Figure BDA0003434735500000024
is XaNormalized numerical data;
aiming at image data, firstly, a pixel value matrix obtained by a sensor is spread into a one-dimensional vector to obtain an image vector XbThen normalized according to the following formula:
Figure BDA0003434735500000031
in the formula (I), the compound is shown in the specification,
Figure BDA0003434735500000032
is the average of the image vectors and is,
Figure BDA0003434735500000033
is the standard deviation of the image vector and,
Figure BDA0003434735500000034
is XbImage data after standardized processing;
the normalized image and the numerical data are two one-dimensional vectors which are spliced to obtain normalized data X*：
Figure BDA0003434735500000035
Further, in step 3), data of the manufacturing process t is collected M times, and the collected multi-modal data is represented as a data set Xt：
Figure BDA0003434735500000036
In the formula (I), the compound is shown in the specification,
Figure BDA0003434735500000037
normalized data collected at mth time; labeling the collected multi-modal data to obtain a data set XtThe label of (2):
Yt＝{Y1t,Y2t,...,YMt}
in the formula, YMtIs data
Figure BDA0003434735500000038
The corresponding label, determined by the expert; data set XtWith corresponding label YtAnd forming a training set.
Further, in step 4), a category-manufacturing action mapping relationship is constructed for the manufacturing process t, specifically as follows:
assuming that there are S categories for manufacturing process t, the set of categories CtExpressed as:
Ct＝{c1t,c2t,...,cSt}
in the formula, cStIs the S-th category of manufacturing process t; finally, a manufacturing action set A is constructedt：
At＝{a1t,a2t,...,aSt}
In the formula, aStIs a class cStCorresponding manufacturing action, thereby constructing a mapping relation between the category and the manufacturing action.
Further, in step 5), the training set of step 3) is input into the deep neural network HtTraining is carried out, a ReLU function is selected as an activation function of the network, a loss function is a cross entropy loss function, and probability calculation is carried out by adopting a softmax function; suppose a deep neural network has L layers, hkAnd hk+1The hidden layer networks corresponding to the k layer and the k +1 layer respectively can be the hidden layer network of the k +1 layer by using the following formulahk+1To show that:
hk+1＝δ(Wkhk)
in the formula, WkWeighting parameters corresponding to a k-th hidden layer network; δ (·) is an activation function, here a ReLU function, introduced non-linearity for deep neural networks, which is formulated as:
Figure BDA0003434735500000041
then, the output Z of the last layer of neural network is calculated:
Z＝WLhL-1
in the formula, WLIs the weight parameter of the last layer of hidden layer network, hL-1Is a penultimate hidden layer network; calculating the probability p that the input data belongs to the class i using the softmax functioni1, 2., S, the calculation formula is:
Figure BDA0003434735500000042
in the formula, ziIs the ith element of the output Z, ZjIs the jth element of output Z, j 1, 2.., S; get the probability piLater, the cross entropy Loss function Loss is used as the deep neural network HtThe specific formula of the loss function of (2) is:
Figure BDA0003434735500000043
in the formula, prIs the probability output of class r, yrIs a label for category r, S is the number of categories; and finally, optimizing the deep neural network by an RMSprop optimization method to obtain the trained deep neural network.
Further, in step 6), the deep neural network trained in step 5) is applied to the multi-modal data stream R of the manufacturing process in actual operationtCarrying out prediction; recording the well-trained deep neural network as
Figure BDA0003434735500000051
The classification result
Figure BDA0003434735500000052
Is shown as
Figure BDA0003434735500000053
Further, in step 7), the classification result obtained in step 6) is used
Figure BDA0003434735500000054
Corresponding to the category-manufacturing action mapping relation constructed in the step 4), and selecting the corresponding manufacturing action.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the method can effectively reduce the difficulty of manually designing the complex manufacturing process and realize the intelligent planning of the action strategy of the manufacturing process.
2. In the method, multi-modal data of each manufacturing step is collected, and a deep neural network is trained for each step according to the collected data. And classifying the multi-modal data which actually runs according to the classification result. Different manufacturing action strategies are planned for different classification results.
3. The method of the invention designs a deep neural network for each manufacturing process, and can effectively adjust the action according to the data characteristics of different steps; while the status of each step in the manufacturing process can be automatically identified.
4. The method designs different actions according to the current state of each manufacturing process, so that the action adjustment in the manufacturing process has higher interpretability. After the deep neural network is constructed, the action strategy planning method can automatically adjust the action of each step according to the multi-modal data in the manufacturing process, thereby effectively reducing manual intervention and improving the automation level.
Drawings
FIG. 1 is a schematic logic flow diagram of the method of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
As shown in fig. 1, the method for intelligently planning the production of equipment based on multi-modal data provided in this embodiment has the following specific conditions:
1) and carrying out data acquisition on the manufacturing process by using different types of sensors to obtain multi-mode data, and storing the multi-mode data in a server.
2) Extracting multi-modal data from the server, and dividing the data into two types of image data and numerical data for preprocessing to obtain preprocessed multi-modal data; aiming at image data, acquiring a pixel value matrix of the image data and carrying out standardization processing on the pixel value matrix; and directly carrying out data standardization processing on the numerical data.
The data normalization process is: after the numerical data stored in the server are read, the data are respectively standardized according to the category, and the standardization formula of the numerical data is as follows:
Figure BDA0003434735500000061
in the formula, XaIs the original data of a numerical type and,
Figure BDA0003434735500000062
is the average of the numerical data,
Figure BDA0003434735500000063
is the standard deviation of the numerical data,
Figure BDA0003434735500000064
is XaNormalized numerical data;
to is directed atImage data is obtained by firstly generating a one-dimensional vector from a pixel value matrix obtained by a sensor to obtain an image vector XbThen normalized according to the following formula:
Figure BDA0003434735500000065
in the formula (I), the compound is shown in the specification,
Figure BDA0003434735500000066
is the average of the image vectors and is,
Figure BDA0003434735500000067
is the standard deviation of the image vector and,
Figure BDA0003434735500000068
is XbImage data after standardized processing;
the normalized image and the numerical data are two one-dimensional vectors which are spliced to obtain normalized data X*：
Figure BDA0003434735500000071
3) For the preprocessed multi-modal data, extracting multi-modal data related to the manufacturing process t; then, according to expert knowledge, manually labeling the data to obtain a label of each piece of data, wherein the extracted multi-modal data and the corresponding label form a training set, and the method specifically comprises the following steps:
collecting data M times for the manufacturing process t, the collected multi-modal data being represented as a data set Xt：
Figure BDA0003434735500000072
In the formula (I), the compound is shown in the specification,
Figure BDA0003434735500000073
is the M-th collectionNormalized data of (a); labeling the collected multi-modal data to obtain a data set XtThe label of (2):
Yt＝{Y1t,Y2t,...,YMt}
in the formula, YMtIs data
Figure BDA0003434735500000074
The corresponding label, determined by the expert; data set XtWith corresponding label YtAnd forming a training set.
4) Constructing a category-manufacturing action mapping relation, so that each category corresponds to one manufacturing action in the training set obtained in the step 3); the method comprises the following steps of constructing a category-manufacturing action mapping relation aiming at a manufacturing process t, wherein the category-manufacturing action mapping relation specifically comprises the following steps:
assuming that there are S categories for manufacturing process t, the set of categories CtCan be expressed as:
Ct＝{c1t,c2t,...,cSt}
in the formula, cStIs the S-th category of manufacturing process t; finally, a manufacturing action set A is constructedt：
At＝{a1t,a2t,...,aSt}
In the formula, aStIs a class cStCorresponding manufacturing action, thereby constructing a mapping relation between the category and the manufacturing action. The correspondence of the category to the manufacturing action is determined by an expert.
5) Inputting the training set of the step 3) into a deep neural network HtAnd training, wherein the activation function of the network selects a ReLU function, the loss function is a cross entropy loss function, and probability calculation is performed by adopting a softmax function. Suppose a deep neural network has L layers, hkAnd hk+1The hidden layer networks corresponding to the k-th layer and the k + 1-th layer respectively, then for the hidden layer network of the k + 1-th layer, the hidden layer network h of the k + 1-th layer can be formed by the following formulak+1To show that:
hk+1＝δ(Wkhk)
in the formula, WkWeighting parameters corresponding to a k-th hidden layer network; δ (·) is an activation function, here a ReLU function, introduced non-linearity for deep neural networks, which is formulated as:
Figure BDA0003434735500000081
then, the output Z of the last layer of neural network is calculated:
Z＝WLhL-1
in the formula, WLIs the weight parameter of the last layer of hidden layer network, hL-1Is a penultimate hidden layer network. Calculating the probability p that the input data belongs to the class i using the softmax functioni1, 2., S, the calculation formula is:
Figure BDA0003434735500000082
in the formula, ziIs the ith element of the output Z, ZjIs the jth element of output Z, j 1, 2.., S; get the probability piLater, the cross entropy Loss function Loss is used as the deep neural network HtThe specific formula of the loss function of (2) is:
Figure BDA0003434735500000083
in the formula, prIs the probability output of class r, yrIs a label for category r, S is the number of categories; and finally, optimizing the deep neural network by an RMSprop optimization method to obtain the trained deep neural network.
6) Utilizing the deep neural network trained in the step 5) to perform multi-modal data flow R of the manufacturing process in actual operationtAnd (6) performing prediction. Recording the well-trained deep neural network as
Figure BDA0003434735500000091
The classification result
Figure BDA0003434735500000092
Can be expressed as:
Figure BDA0003434735500000093
7) classifying the results obtained in the step 6)
Figure BDA0003434735500000094
Corresponding to the category-manufacturing action mapping relation constructed in the step 4), and selecting the corresponding manufacturing action. Suppose that
Figure BDA0003434735500000095
In the category-manufacturing operation, the manufacturing operation is corresponded
Figure BDA0003434735500000096
The actions of the manufacturing process t are planned as
Figure BDA0003434735500000097
8) Repeating the actions from the step 3) to the step 7) for each step in the manufacturing process to form an intelligent planning scheme of the manufacturing process actions under multi-modal data
Figure BDA0003434735500000098
Wherein
Figure BDA0003434735500000099
Is the manufacturing operation of manufacturing process d (d ═ 1, 2.., T), where T is the number of steps in the manufacturing process.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims (7)

1. An intelligent planning method for equipment production based on multi-modal data is characterized by comprising the following steps:
1) carrying out data acquisition on the manufacturing process by using different types of sensors to obtain multi-mode data, and storing the multi-mode data in a server;
2) extracting multi-modal data from the server, and dividing the data into two types of image data and numerical data for preprocessing to obtain preprocessed multi-modal data; aiming at image data, acquiring a pixel value matrix of the image data and carrying out standardization processing on the pixel value matrix; directly carrying out data standardization processing on numerical data;
3) for the preprocessed multi-modal data, extracting multi-modal data related to the manufacturing process t; then, manually labeling the data according to expert knowledge to obtain a label of each piece of data, wherein the extracted multi-modal data and the corresponding labels form a training set;
4) constructing a category-manufacturing action mapping relation, so that each category corresponds to one manufacturing action in the training set obtained in the step 3);
5) inputting the training set of the step 3) into a deep neural network for training; the method comprises the following steps that a ReLU function is selected as an activation function of a deep neural network, a loss function is a cross entropy loss function, and category probability output is carried out by adopting a softmax function; obtaining a trained deep neural network by optimizing a loss function;
6) during the production of actual equipment, different types of sensors are used for data acquisition to form a multi-modal data stream, and the multi-modal data stream is predicted by using a trained deep neural network to obtain a classification result;
7) selecting the manufacturing action corresponding to the classification result in the step 6) according to the category-manufacturing action mapping relation constructed in the step 4), and finishing the action planning of the actual manufacturing process t;
8) and (4) repeating the steps 3) to 7) for each actually operated manufacturing process to form the intelligent plan of the action strategy of the equipment production manufacturing process under the multi-mode data.
2. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 2), the data normalization process is: after the numerical data stored in the server are read, classifying the data, and respectively standardizing the data according to categories, wherein the standardized formula of the numerical data is as follows:
Figure FDA0003434735490000021
in the formula, XaIs the original data of a numerical type and,
Figure FDA0003434735490000022
is the average of the numerical data,
Figure FDA0003434735490000023
is the standard deviation of the numerical data,
Figure FDA0003434735490000024
is XaNormalized numerical data;
aiming at image data, firstly, a pixel value matrix obtained by a sensor is spread into a one-dimensional vector to obtain an image vector XbThen normalized according to the following formula:
Figure FDA0003434735490000025
in the formula (I), the compound is shown in the specification,
Figure FDA0003434735490000026
is the average of the image vectors and is,
Figure FDA0003434735490000027
is the standard deviation of the image vector and,
Figure FDA0003434735490000028
is XbImage data after standardized processing;
the normalized image and the numerical data are two one-dimensional vectors which are spliced to obtain normalized data X*：
Figure FDA0003434735490000029
3. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 3), data of the manufacturing process t is collected M times, and the collected multi-modal data is represented as a data set Xt：
Figure FDA00034347354900000210
In the formula (I), the compound is shown in the specification,
Figure FDA00034347354900000211
normalized data collected at mth time; labeling the collected multi-modal data to obtain a data set XtThe label of (2):
Yt＝{Y1t,Y2t,...,YMt}
in the formula, YMtIs data
Figure FDA0003434735490000031
The corresponding label, determined by the expert; data set XtWith corresponding label YtAnd forming a training set.
4. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 4), a category-manufacturing action mapping relationship is constructed for the manufacturing process t, specifically as follows:
assuming that there are S categories for manufacturing process t, the set of categories CtExpressed as:
Ct＝{c1t,c2t,...,cSt}
in the formula, cStIs the S-th category of manufacturing process t; finally, a manufacturing action set A is constructedt：
At＝{a1t,a2t,...,aSt}
In the formula, aStIs a class cStCorresponding manufacturing action, thereby constructing a mapping relation between the category and the manufacturing action.
5. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 5), the training set of step 3) is input into a deep neural network HtTraining is carried out, a ReLU function is selected as an activation function of the network, a loss function is a cross entropy loss function, and probability calculation is carried out by adopting a softmax function; suppose a deep neural network has L layers, hkAnd hk+1The hidden layer networks corresponding to the k layer and the k +1 layer respectively can be the hidden layer network h of the k +1 layer by using the following formulak+1To show that:
hk+1＝δ(Wkhk)
in the formula, WkWeighting parameters corresponding to a k-th hidden layer network; δ (·) is an activation function, here a ReLU function, introduced non-linearity for deep neural networks, which is formulated as:
Figure FDA0003434735490000032
then, the output Z of the last layer of neural network is calculated
Z＝WLhL-1
In the formula, WLIs the last hidden layerWeight parameter of the network, hL-1Is a penultimate hidden layer network; calculating the probability p that the input data belongs to the class i using the softmax functioni1, 2., S, the calculation formula is:
Figure FDA0003434735490000041
in the formula, ziIs the ith element of the output Z, ZjIs the jth element of output Z, j 1, 2.., S; get the probability piLater, the cross entropy Loss function Loss is used as the deep neural network HtThe specific formula of the loss function of (2) is:
Figure FDA0003434735490000042
in the formula, prIs the probability output of class r, yrIs a label for category r, S is the number of categories; and finally, optimizing the deep neural network by an RMSprop optimization method to obtain the trained deep neural network.
6. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 6), the deep neural network trained in step 5) is applied to the multi-modal data stream R of the manufacturing process in actual operationtCarrying out prediction; recording the well-trained deep neural network as
Figure FDA0003434735490000043
The classification result
Figure FDA0003434735490000044
Is shown as
Figure FDA0003434735490000045
7. The intelligent planning method for equipment production based on multi-modal data as claimed in claim 1, wherein: in step 7), the classification result obtained in step 6) is used
Figure FDA0003434735490000046
Corresponding to the category-manufacturing action mapping relation constructed in the step 4), and selecting the corresponding manufacturing action.
CN202111609391.3A 2021-12-27 2021-12-27 Intelligent planning method for equipment production based on multi-mode data Pending CN114372181A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111609391.3A CN114372181A (en) 2021-12-27 2021-12-27 Intelligent planning method for equipment production based on multi-mode data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111609391.3A CN114372181A (en) 2021-12-27 2021-12-27 Intelligent planning method for equipment production based on multi-mode data

Publications (1)

Publication Number Publication Date
CN114372181A true CN114372181A (en) 2022-04-19

Family

ID=81142263

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111609391.3A Pending CN114372181A (en) 2021-12-27 2021-12-27 Intelligent planning method for equipment production based on multi-mode data

Country Status (1)

Country Link
CN (1) CN114372181A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117112857A (en) * 2023-10-23 2023-11-24 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Machining path recommending method suitable for industrial intelligent manufacturing

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160093048A1 (en) * 2014-09-25 2016-03-31 Siemens Healthcare Gmbh Deep similarity learning for multimodal medical images
CN112784919A (en) * 2021-02-03 2021-05-11 华南理工大学 Intelligent manufacturing multi-mode data oriented classification method
CN113112086A (en) * 2021-04-22 2021-07-13 北京邮电大学 Intelligent production system based on edge calculation and identification analysis

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160093048A1 (en) * 2014-09-25 2016-03-31 Siemens Healthcare Gmbh Deep similarity learning for multimodal medical images
CN112784919A (en) * 2021-02-03 2021-05-11 华南理工大学 Intelligent manufacturing multi-mode data oriented classification method
CN113112086A (en) * 2021-04-22 2021-07-13 北京邮电大学 Intelligent production system based on edge calculation and identification analysis

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘志洪, 顾宁: "制造业设计文档的模糊分类", 计算机辅助设计与图形学学报, no. 11, 20 November 2004 (2004-11-20) *
敬明旻;: "基于深度神经网络的多模态特征自适应聚类方法", 计算机应用与软件, no. 10, 12 October 2020 (2020-10-12) *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117112857A (en) * 2023-10-23 2023-11-24 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Machining path recommending method suitable for industrial intelligent manufacturing
CN117112857B (en) * 2023-10-23 2024-01-05 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Machining path recommending method suitable for industrial intelligent manufacturing

Similar Documents

Publication Publication Date Title
CN110580496A (en) Deep migration learning system and method based on entropy minimization
CN112987664B (en) Flow shop scheduling method based on deep reinforcement learning
CN109472057B (en) Product processing quality prediction device and method based on cross-process implicit parameter memory
CN110163429B (en) Short-term load prediction method based on similarity day optimization screening
CN113128113A (en) Poor information building load prediction method based on deep learning and transfer learning
CN113988449A (en) Wind power prediction method based on Transformer model
CN112308298B (en) Multi-scenario performance index prediction method and system for semiconductor production line
CN115204035A (en) Generator set operation parameter prediction method and device based on multi-scale time sequence data fusion model and storage medium
CN115424177A (en) Twin network target tracking method based on incremental learning
CN114372181A (en) Intelligent planning method for equipment production based on multi-mode data
Chen Estimating job cycle time in a wafer fabrication factory: A novel and effective approach based on post-classification
CN111753995A (en) Local interpretable method based on gradient lifting tree
CN108573275B (en) Construction method of online classification micro-service
CN115512214A (en) Indoor visual navigation method based on causal attention
CN114330089B (en) Rare earth element content change prediction method and system
CN115217534A (en) Method and system for monitoring service quality state of steam turbine
CN114943277A (en) Method for identifying operation condition and predicting operation performance of comprehensive energy system
CN114139677A (en) Unequal interval time sequence data prediction method based on improved GRU neural network
CN113159082A (en) Incremental learning target detection network model construction and weight updating method
CN115169617B (en) Mold maintenance prediction model training method, mold maintenance prediction method and system
CN111459927B (en) CNN-LSTM developer project recommendation method
CN113988181B (en) Target classification method based on adaptive feedforward neural network
CN117010459B (en) Method for automatically generating neural network based on modularization and serialization
Nguyen et al. Improve object detection performance with efficient task-alignment module
CN112101631B (en) Product construction period prediction method based on recurrent neural network

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