CN116502959B - Product manufacturing quality prediction method based on meta learning - Google Patents
Product manufacturing quality prediction method based on meta learning Download PDFInfo
- Publication number
- CN116502959B CN116502959B CN202310738959.4A CN202310738959A CN116502959B CN 116502959 B CN116502959 B CN 116502959B CN 202310738959 A CN202310738959 A CN 202310738959A CN 116502959 B CN116502959 B CN 116502959B
- Authority
- CN
- China
- Prior art keywords
- sample
- representation
- quality
- representing
- sample data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 98
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 49
- 230000008569 process Effects 0.000 claims abstract description 42
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 238000012795 verification Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 17
- 230000007246 mechanism Effects 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 11
- 230000015654 memory Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000000153 supplemental effect Effects 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012512 characterization method Methods 0.000 claims description 3
- 230000007787 long-term memory Effects 0.000 claims description 3
- 238000005065 mining Methods 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 3
- 238000004886 process control Methods 0.000 abstract description 2
- 239000000047 product Substances 0.000 description 15
- 230000002159 abnormal effect Effects 0.000 description 5
- 230000008878 coupling Effects 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 4
- 238000005859 coupling reaction Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 206010027175 memory impairment Diseases 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Educational Administration (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Primary Health Care (AREA)
- Manufacturing & Machinery (AREA)
- General Factory Administration (AREA)
Abstract
The application relates to the technical field of product manufacturing quality prediction, solves the technical problem that small sample size data cannot support quality prediction of complex industrial structural parts with multiple production procedures, and particularly relates to a product manufacturing quality prediction method based on meta-learning, which comprises the following steps: s1, acquiring sample data of each manufacturing procedure in the process of manufacturing a complex structural member, wherein the sample data comprises process parameters, related sensor data and quality indexes; s2, preprocessing sample data to obtain a plurality of sample data sets, dividing the sample data sets into a training set, a verification set and a test set, and dividing each sample data set into a support set and a query set. The method can realize the prediction of the processing state of the complex structural member in the aerospace field, further realize the process control, obviously improve the processing efficiency and quality of the key structural member of the aero-engine and ensure the safety of the processing process.
Description
Technical Field
The application relates to the technical field of product manufacturing quality prediction, in particular to a product manufacturing quality prediction method based on meta learning.
Background
In recent years, the manufacturing industry at home and abroad gradually develops to organically combine the traditional processing technology with digital twin and artificial intelligence, so as to intelligently regulate and control the processing process through a virtual twin system. In the aerospace field, complex structural parts exist, the manufacturing process of the complex structural parts is a nonlinear multi-field coupling machining process, and along with material removal and surface creation, various complex physical processes are highly coupled, so that the shape precision and the surface state of the target structural parts are comprehensively influenced.
For the problem of data-driven processing state prediction, conventional machine learning algorithms such as support vector machines and the like have limitations of generalizing performance in the face of complex industrial scenarios of multiple process coupling. The deep neural network has good prediction performance, such as a recurrent neural network, a long-short-term memory network and the like, can capture time dynamic behaviors in data, but is dependent on data quantity, the prediction performance of small sample quantity data is general, data samples cannot be fully utilized, and the accuracy of the small sample quantity prediction under a complex technological process is difficult to realize.
Disclosure of Invention
Aiming at the defects of the prior art, the application provides a product manufacturing quality prediction method based on meta-learning, which solves the technical problem that small sample size data cannot support quality prediction of complex industrial structural parts with multiple production procedures.
In order to solve the technical problems, the application provides the following technical scheme: a product manufacturing quality prediction method based on meta learning, the prediction method comprising the steps of:
s1, acquiring sample data of each manufacturing procedure in the process of manufacturing a complex structural member, wherein the sample data comprises process parameters, related sensor data and quality indexes;
s2, preprocessing sample data to obtain a plurality of sample data sets, dividing the sample data sets into a training set, a verification set and a test set, and dividing each sample data set into a support set and a query set;
s3, mapping sample features of the sample data set corresponding to different procedures to the same dimension, and performing time sequence coding on the features according to the sequence of each procedure to serve as supplementary features;
s4, mining expected characterization reflecting sample commonalities related to measurement indexes through sample data in the support set to serve as newly generated supplementary features;
s5, constructing a decoder, and predicting the quality index of the sample in the query set by combining the input and the supplementary features of the support set to obtain a quality prediction result;
s6, respectively constructing a loss function for the newly generated supplementary features and the quality prediction result, and updating parameters through a back propagation algorithm to obtain a new digital twin model;
s7, inputting a sample of the complex structural member manufacturing process into a new digital twin model to finish the manufacturing quality prediction of the industrial manufacturing process.
Further, in step S2, the specific process includes the following steps:
s201, performing outlier processing and normalization processing on sample data of technological parameters of each manufacturing procedure and quality indexes of complex structural members to obtain a sample data set;
s202, dividing a sample data set into a training set, a verification set and a test set, taking out the first K samples in each data set on the basis of the training set, the verification set and the test set as supporting sets, and taking the rest samples as query sets.
Further, in step S3, the specific process includes the following steps:
s301, mapping sample characteristics of different dimensions corresponding to different working procedures to the same dimension through a fully connected neural network;
further, mapping sample features of different dimensions to the same dimension is accomplished using the following formula:
；
wherein ,sample characteristics are shown,/->Representing the feature representation mapped to the same dimension;
s302, constructing a two-way long-short-term memory network BiLSTM according to the sequence of each procedure, then encoding sample characteristics, and splicing the two-way hidden states of the two-way long-term memory network BiLSTM as output to serve as time sequence characteristics of each procedure by taking the mapped sample characteristics of each procedure in a sample as input;
s303, splicing the sample features with the dimensions mapped to the same dimension with the time sequence features to serve as initial feature codes of the samples.
Further, in step S4, the specific process includes the following steps:
s401, splicing initial feature codes and quality indexes of samples to obtain new sample representations;
s402, inputting a new sample representation into a multi-head attention mechanism, and outputting an updated representation as each sample, wherein the updated representation comprises correlations among the samples;
s403, the updated representation of all the support samples is averaged and scaled through a layer of fully connected network to be used as newly generated supplementary features.
Further, the newly generated supplemental features are accomplished using the following formula:
；
wherein ,representing new sample representations after splicing the support set feature representation with the quality index, K representing the number of support set samples,/for the support set feature representation>Representation->Dimension of->Reflecting the importance of the jth sample to the ith sample,/>Is an updated representation of sample i.
；
wherein ,for the temporary representation of the supplementary features, K represents the number of support set samples, +.>For supplementing features, a->Representing a weight matrix, +.>Representing the bias term.
Further, in step S5, the specific process includes the steps of:
s501, combining the sample representation of the support set with newly generated supplementary features to serve as an input feature representation of the support set, and taking the feature representation of the query set after encoding as the input of the query set;
s502, calculating an update representation of the support samples as part of the input of the decoder;
s503, calculating correlation between the support set and the sample of the query set through an attention mechanism, wherein the importance of each support sample is regarded as a weight parameter generated in a decoder;
s504, multiplying the generation parameters of the support samples with the characteristic representation of the query sample to add the bias, and obtaining the quality prediction result of the query sample.
Further, in step S6, the specific process includes the steps of:
s601, respectively constructing a multiple loss function for the newly generated supplementary features, the prediction result and the generation parameters of the decoder;
s602, weighting and summing the multiple loss functions to be used as a total loss function of the digital twin model;
and S603, carrying out iterative updating on parameters of the digital twin model through a back propagation algorithm to obtain a new digital twin model.
By means of the technical scheme, the application provides a product manufacturing quality prediction method based on meta learning, which has at least the following beneficial effects:
1. according to the application, the time sequence characteristics corresponding to each process in the product process flow can be mined through the bidirectional long-short-term memory network, and the influence relationship of mutual coupling among different processes is reserved in the characteristics;
based on the idea of meta-learning, the priori knowledge reflecting sample commonality corresponding to each quality index in the data is mined through a multi-head attention mechanism to be used as the supplement of the features, and limited data can be fully utilized, so that the prediction has better interpretability; the decoding prediction process is based on an attention mechanism, and the prediction accuracy of the small sample under the two supports can be effectively improved by combining the relation between the known result sample and the unknown result sample.
2. The method can realize the prediction of the processing state of the complex structural member in the aerospace field, further realize the process control, obviously improve the processing efficiency and quality of the key structural member of the aero-engine and ensure the safety of the processing process.
3. The application adopts the idea of meta learning, and combines the attention mechanism to mine the priori knowledge in the sample examples, fully utilizes the data samples, and is beneficial to improving the accuracy of small sample quantity prediction in the complex process.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:
FIG. 1 is a flow chart of a method for predicting manufacturing quality of a product according to the present application;
FIG. 2 is a flow chart of the critical fabrication of a complex structure of the present application;
FIG. 3 is a schematic diagram of a network architecture of the new digital twin model of the present application;
FIG. 4 is a graph of the predictive test results of the novel digital twin model of the present application.
Detailed Description
In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description. Therefore, the realization process of how to apply the technical means to solve the technical problems and achieve the technical effects can be fully understood and implemented.
Those of ordinary skill in the art will appreciate that all or a portion of the steps in a method of implementing an embodiment described above may be implemented by a program to instruct related hardware, and thus, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
Referring to fig. 1-4, a specific implementation manner of the present embodiment is shown, in which the present embodiment adopts the idea of meta-learning, and the attention mechanism is combined to mine priori knowledge in the sample instance, so that the data sample is fully utilized, and the accuracy of small sample size prediction in the complex process is improved.
Referring to fig. 1 and 2, taking a key manufacturing process of a complex structural member shown in fig. 2 as an example, a specific embodiment of the product manufacturing quality prediction method includes the following steps:
s1, acquiring sample data of each manufacturing procedure in the process of manufacturing a complex structural member, wherein the sample data comprises process parameters, related sensor data and quality indexes;
the sample data of each manufacturing procedure is used for determining the technological parameters of each procedure and the quality indexes obtained by measurement according to the specific technological process of industrial products in a manufacturing workshop, wherein the number of the technological parameters and the quality indexes is multiple, and each manufacturing procedure corresponds to at least one sample data, so that the technological parameters, the related sensor data and the quality indexes in each procedure are determined according to the sample data.
S2, preprocessing sample data to obtain a plurality of sample data sets, dividing the sample data sets into a training set, a verification set and a test set, and dividing each sample data set into a support set and a query set;
as a preferred mode of the present embodiment, the method adopted for realizing step S2 is as follows:
s201, performing outlier processing and normalization processing on sample data of technological parameters of each manufacturing procedure and quality indexes of complex structural members to obtain a sample data set;
abnormal value processing adopts abnormal value processing of data processing in the prior art, the abnormal value processing is realized by adopting a deletion, deletion value treatment, average value correction or capping method, wherein the abnormal value processing is realized by adopting a 3 sigma principle to establish that the interference or noise of singular data is difficult to meet the normal distribution on the basis of equal-precision repeated measurement of the normal distribution, so that abnormal value detection is completed. The normalization process adopts a normalization process method known in the prior art, and is not described in detail here.
S202, dividing a sample data set into a training set, a verification set and a test set, taking the first K samples in each data set as a supporting set and taking the rest samples as a query set, wherein the value of K is usually 5-20% of the number of the samples of the whole data set, and the ratio can provide enough training data to train the digital twin model, and meanwhile, excessive fitting of the digital twin model caused by using too much data is avoided.
In the step, each manufacturing procedure corresponds to at least one sample data, the sample data corresponds to at least one sample data set after pretreatment, and the digital twin model can be fully trained by selecting more sample data sets, and meanwhile, the accurate prediction of the manufacturing quality of the digital twin model for the manufacturing process of the complex structural member is improved.
S3, mapping sample features of the sample data set corresponding to different procedures to the same dimension, and performing time sequence coding on the features according to the sequence of each procedure to serve as supplementary features;
as a preferred mode of the present embodiment, the method adopted for realizing step S3 is as follows:
s301, mapping sample characteristics of different dimensions corresponding to different working procedures to the same dimension through a fully connected neural network;
in step S301, mapping sample features of different dimensions to the same dimension is accomplished using the following formula:
；
wherein ,representing sample characteristics->Representation of the feature after mapping to the same dimension, < +.>Representing a weight matrix, each row of which corresponds to an embedded vector,>representing the bias term.
S302, constructing a two-way long-short-term memory network BiLSTM according to the sequence of each procedure, then encoding sample characteristics, and splicing the two-way hidden states of the two-way long-term memory network BiLSTM as output to serve as time sequence characteristics of each procedure by taking the mapped sample characteristics of each procedure in a sample as input;
in step S302, the extraction of the time-series features is completed using the following formula:
；
wherein The hidden state, which is expressed as the output of the bi-directional long-short term memory network BiLSTM, is characterized as a time series.
The specific calculation method of the two-way long-short-term memory network BiLSTM comprises the following steps:
for a forward LSTM there are:
；
wherein Indicating forgetfulness door, < >>Representing the input door,/->Representing the output door, & lt + & gt>Representing the feature vector of the input +.>Indicating the temporary status of the cell,/->Indicates the hidden state of the last moment, +.>Representing a weight matrix, +.>Representing the bias term.
；
wherein Indicates the cell state at time t,/->The hidden state at time t is indicated.
；
wherein and />Respectively representing hidden states of the forward and reverse outputs of the two-way long-short-term memory network BiLSTM, and obtaining output +.>。
S303, splicing the sample features with the dimensions mapped to the same dimension with the time sequence features to serve as initial feature codes of the samples.
S4, mining expected characterization reflecting sample commonalities related to measurement indexes through sample data in the support set to serve as newly generated supplementary features;
as a preferred mode of the present embodiment, the method adopted for realizing step S4 is as follows:
s401, splicing initial feature codes and quality indexes of samples to obtain new sample representations;
s402, inputting a new sample representation into a multi-head attention mechanism, and outputting an updated representation as each sample, wherein the updated representation comprises correlations among the samples;
s403, the updated representation of all the support samples is averaged and scaled through a layer of fully connected network to be used as newly generated supplementary features.
In step S4, the newly generated supplemental features are completed using the following formula:
；
wherein ,representing new sample representations after splicing the support set feature representation with the quality index, K representing the number of support set samples,/for the support set feature representation>Representation->Dimension of->Reflecting the importance of the jth sample to the ith sample,/>For an updated representation of sample i, +.>Are weight matrices.
；
wherein ,for the temporary representation of the supplementary features, K represents the number of support set samples, +.>For supplementing features, a->Representing a weight matrix, +.>Representing the bias term.
S5, constructing a decoder, and predicting the quality index of the sample in the query set by combining the input and the supplementary features of the support set to obtain a quality prediction result;
as a preferred mode of the present embodiment, the method adopted for realizing step S5 is as follows:
s501, combining the sample representation of the support set with newly generated supplementary features to serve as an input feature representation of the support set, and taking the feature representation of the query set after encoding as the input of the query set;
s502, repeating the process in the step S402, and calculating an update representation of the support samples as part of the input of the decoder;
s503, calculating correlation between the support set and the sample of the query set through an attention mechanism, wherein the importance of each support sample is regarded as a weight parameter generated in a decoder;
the weight parameter generation is expressed as:
；
wherein , and />Representing the feature representation of the query set and the support set encoded, respectively, < >>Representation->Is a dimension of (2); />Sample for support set->The updated representation obtained after repeating the procedure of step S402, is->Representing a weight matrix, +.>Representing temporary generation parameters that are not projected via the fully connected network, K represents the number of support set samples.
；
wherein ,is +.>，/>Front middle>Item->Representing the generation weight, last item +.>For bias item->An activation function representing a fully connected network, typically a softmax function,/for example>Representing a weight matrix, +.>Representing the bias term.
S504, multiplying the generation parameters of the support samples with the characteristic representation of the query sample to add the bias, and obtaining the quality prediction result of the query sample.
The quality prediction result is calculated according to the following formula:
；
s6, respectively constructing a loss function for the newly generated supplementary features and the quality prediction result, and updating parameters through a back propagation algorithm to obtain a new digital twin model;
as a preferred mode of the present embodiment, the method adopted for realizing step S6 is as follows:
s601, respectively constructing a multiple loss function for the newly generated supplementary features, the prediction result and the generation parameters of the decoder;
multiple loss function、/> and />The method comprises the following steps of:
；
wherein ,for supporting the sample size of the set corresponding to all quality indicators, < >>Representation divide sample->The remaining sample sets, except ∈ ->For optimizing a priori feature generation such that the quality index characterizes a similar sample with a feature representation closer,/i>、Representing the newly generated supplemental features.
；
wherein ,for balancing weights in the decoder, preventing the overfitting phenomenon caused by the existence of excessive partial weights,is +.>，/>Front middle>Item->Representing the generation weight, last item +.>Is a bias term.
；
Wherein, the mean square error is taken as the loss of the product quality index prediction，/>Indicating the procedure->Is>Prediction of individual quality indicators, < >>Representing the true value +_>Represents the number of procedures in the current training batch, +.>Indicating procedure->The number of training samples included.
S602, weighting and summing the multiple loss functions to be used as a total loss function of the digital twin model;
；
wherein the weighted sum of the losses of each part is used as a multi-element loss function of the whole digital twin model iteration，Representing the weight parameters, respectively.
And S603, carrying out iterative updating on parameters of the digital twin model through a back propagation algorithm to obtain a new digital twin model.
S7, inputting a sample of the complex structural member manufacturing process into a new digital twin model to finish the manufacturing quality prediction of the industrial manufacturing process.
The product manufacturing quality prediction digital twin model based on meta learning provided by the embodiment can dig out time sequence characteristics corresponding to each process in the product process flow through a two-way long-short-term memory network, and the influence relationship of mutual coupling among different processes is reserved in the characteristics;
based on the idea of meta-learning, the priori knowledge reflecting sample commonality corresponding to each quality index in the data is mined through a multi-head attention mechanism to be used as the supplement of the features, and limited data can be fully utilized, so that the prediction has better interpretability; the decoding prediction process is based on an attention mechanism, and the prediction accuracy of the small sample under the two supports can be effectively improved by combining the relation between the known result sample and the unknown result sample.
Specifically, please refer toFig. 3 is a schematic diagram of a network structure of a new digital twin model, in fig. 3,representing support set sample features, +.>Representing sample characteristics of a query set,/->Sample tags representing support sets, i.e. quality index,/->Representing learned supplemental features->The prediction result of the quality index is shown, in this embodiment, the new digital twin model is verified and quantitatively analyzed by using the sample of the product manufacturing process shown in fig. 2, the average absolute error of the prediction is 0.36, the root mean square error is 0.48, and the prediction result is shown in fig. 4. Compared with other traditional algorithms, the method has better performance and lower error, can effectively solve the problem of product manufacturing quality prediction under the support of small sample size, and has the results shown in table 1.
The prediction method provided by the embodiment can be used for predicting the processing state of the complex structural member in the aerospace field, further realizing process regulation, remarkably improving the processing efficiency and quality of the key structural member of the aero-engine and ensuring the safety of the processing process.
And meanwhile, the idea of meta learning is adopted, priori knowledge in sample examples is mined by combining an attention mechanism, data samples are fully utilized, and the accuracy of small sample quantity prediction in a complex process is improved.
The foregoing embodiments have been presented in a detail description of the application, and are presented herein with a particular application to the understanding of the principles and embodiments of the application, the foregoing embodiments being merely intended to facilitate an understanding of the method of the application and its core concepts; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.
Claims (5)
1. A product manufacturing quality prediction method based on meta learning, characterized in that the prediction method comprises the following steps:
s1, acquiring sample data of each manufacturing procedure in the process of manufacturing a complex structural member, wherein the sample data comprises process parameters, related sensor data and quality indexes;
s2, preprocessing sample data to obtain a plurality of sample data sets, dividing the sample data sets into a training set, a verification set and a test set, and dividing each sample data set into a support set and a query set;
s3, mapping sample features of the sample data set corresponding to different procedures to the same dimension, and performing time sequence coding on the features according to the sequence of each procedure to serve as supplementary features; in step S3, the specific process includes the following steps:
s301, mapping sample characteristics of different dimensions corresponding to different working procedures to the same dimension through a fully connected neural network;
s302, constructing a two-way long-short-term memory network BiLSTM according to the sequence of each procedure, then encoding sample characteristics, and splicing the two-way hidden states of the two-way long-term memory network BiLSTM as output to serve as time sequence characteristics of each procedure by taking the mapped sample characteristics of each procedure in a sample as input;
s303, splicing the sample features with the dimensions mapped to the same dimension with the time sequence features to serve as initial feature codes of the samples;
s4, mining expected characterization reflecting sample commonalities related to measurement indexes through sample data in the support set to serve as newly generated supplementary features; in step S4, the specific process includes the following steps:
s401, splicing initial feature codes and quality indexes of samples to obtain new sample representations;
s402, inputting a new sample representation into a multi-head attention mechanism, and outputting an updated representation as each sample, wherein the updated representation comprises correlations among the samples;
s403, taking average and scaling of updated representations of all support samples through a layer of fully connected network to serve as newly generated supplementary features;
s5, constructing a decoder, and predicting the quality index of the sample in the query set by combining the input and the supplementary features of the support set to obtain a quality prediction result; in step S5, the specific process includes the following steps:
s501, combining the sample representation of the support set with newly generated supplementary features to serve as an input feature representation of the support set, and taking the feature representation of the query set after encoding as the input of the query set;
s502, calculating an update representation of the support samples as part of the input of the decoder;
s503, calculating correlation between the support set and the sample of the query set through an attention mechanism, wherein the importance of each support sample is regarded as a weight parameter generated in a decoder;
s504, multiplying the generation parameters of the support samples with the characteristic representation of the query sample to add a bias, so as to obtain a quality prediction result of the query sample;
s6, respectively constructing a loss function for the newly generated supplementary features and the quality prediction result, and updating parameters through a back propagation algorithm to obtain a new digital twin model; in step S6, the specific process includes the following steps:
s601, respectively constructing a multiple loss function for the newly generated supplementary features, the prediction result and the generation parameters of the decoder;
s602, weighting and summing the multiple loss functions to be used as a total loss function of the digital twin model;
s603, carrying out iterative updating on parameters of the digital twin model through a back propagation algorithm to obtain a new digital twin model;
s7, inputting a sample of the complex structural member manufacturing process into a new digital twin model to finish the manufacturing quality prediction of the industrial manufacturing process.
2. The product manufacturing quality prediction method according to claim 1, wherein in step S2, the specific process includes the steps of:
s201, performing outlier processing and normalization processing on sample data of technological parameters of each manufacturing procedure and quality indexes of complex structural members to obtain a sample data set;
s202, dividing a sample data set into a training set, a verification set and a test set, taking out the first K samples in each data set on the basis of the training set, the verification set and the test set as supporting sets, and taking the rest samples as query sets.
3. The method according to claim 1, wherein in step S301, mapping sample features of different dimensions to the same dimension is accomplished using the following formula:
；
wherein ,representing sample characteristics->Representation of the feature after mapping to the same dimension, < +.>Representing a weight matrix, each row of which corresponds to an embedded vector,>representing the bias term.
4. The method of claim 1, wherein in step S4, the newly generated supplemental features are accomplished using the following formula:
；
wherein ,representing new sample representations after splicing the support set feature representation with the quality index, K representing the number of support set samples,/for the support set feature representation>Representation->Dimension of->Reflecting the importance of the jth sample to the ith sample,/>For an updated representation of sample i, +.>Are weight matrixes;
；
wherein ,for the temporary representation of the supplementary features, K represents the number of support set samples, +.>For supplementing features, a->Representing a weight matrix, +.>Representing the bias term.
5. The product manufacturing quality prediction method according to claim 1, wherein in step S61, a multiple loss function is used、/> and />The method comprises the following steps of:
；
wherein ,for supporting the sample size of the set corresponding to all quality indicators, < >>Representation divide sample->The remaining sample sets, except ∈ ->For optimizing a priori feature generation such that the quality index characterizes a similar sample with a feature representation closer,/i>、/>Representing the newly generated supplemental features;
；
wherein ,for balancing weights in a decoder, preventing the existence of partial weights that are too large resulting in an overfitting phenomenon, < >>Is +.>，/>Front middle>Item->Representing the generation weight, last item +.>Is a bias term;
；
wherein, the mean square error is taken as the loss of the product quality index prediction，/>Indicating the procedure->Is>Prediction of individual quality indicators, < >>Representing the true value +_>Represents the number of procedures in the current training batch, +.>Indicating procedure->The number of training samples included.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310738959.4A CN116502959B (en) | 2023-06-21 | 2023-06-21 | Product manufacturing quality prediction method based on meta learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310738959.4A CN116502959B (en) | 2023-06-21 | 2023-06-21 | Product manufacturing quality prediction method based on meta learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116502959A CN116502959A (en) | 2023-07-28 |
| CN116502959B true CN116502959B (en) | 2023-09-08 |
Family
ID=87316787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310738959.4A Active CN116502959B (en) | 2023-06-21 | 2023-06-21 | Product manufacturing quality prediction method based on meta learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116502959B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116894180B (en) * | 2023-09-11 | 2023-11-24 | 南京航空航天大学 | Product manufacturing quality prediction method based on different composition attention network |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110728656A (en) * | 2019-09-06 | 2020-01-24 | 西安电子科技大学 | Meta-learning-based no-reference image quality data processing method and intelligent terminal |
| CN111241673A (en) * | 2020-01-07 | 2020-06-05 | 北京航空航天大学 | Health state prediction method for industrial equipment in noisy environment |
| CN112015902A (en) * | 2020-09-14 | 2020-12-01 | 中国人民解放军国防科技大学 | Least-order text classification method under metric-based meta-learning framework |
| CN112114791A (en) * | 2020-09-08 | 2020-12-22 | 南京航空航天大学 | Code self-adaptive generation method based on meta-learning |
| CN112949902A (en) * | 2021-01-25 | 2021-06-11 | 西北工业大学 | Runoff prediction method based on LSTM multi-state vector sequence-to-sequence model |
| CN114239949A (en) * | 2021-12-10 | 2022-03-25 | 中信银行股份有限公司 | Website access amount prediction method and system based on two-stage attention mechanism |
| JP2022067858A (en) * | 2020-10-21 | 2022-05-09 | セコム株式会社 | Learned model and data processor |
| CN114861522A (en) * | 2022-04-11 | 2022-08-05 | 深圳贝尔信息科技有限公司 | Precision manufacturing quality monitoring method and device based on artificial intelligence meta-learning technology |
| CN114972232A (en) * | 2022-05-17 | 2022-08-30 | 西安电子科技大学 | No-reference image quality evaluation method based on incremental meta-learning |
| CN115187569A (en) * | 2022-07-26 | 2022-10-14 | 北京芯联心科技发展有限公司 | Image quality evaluation model training method, device, equipment and medium |
| CN115297496A (en) * | 2022-09-28 | 2022-11-04 | 南昌航空大学 | Link quality prediction method combining Bi-LSTM and time mode attention |
| CN115908256A (en) * | 2022-10-19 | 2023-04-04 | 青岛大学 | Yarn quality prediction method based on meta-learning method |
| CN115906399A (en) * | 2022-10-11 | 2023-04-04 | 电子科技大学 | Improved method for predicting key process quality of product under small sample data |
-
2023
- 2023-06-21 CN CN202310738959.4A patent/CN116502959B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110728656A (en) * | 2019-09-06 | 2020-01-24 | 西安电子科技大学 | Meta-learning-based no-reference image quality data processing method and intelligent terminal |
| CN111241673A (en) * | 2020-01-07 | 2020-06-05 | 北京航空航天大学 | Health state prediction method for industrial equipment in noisy environment |
| CN112114791A (en) * | 2020-09-08 | 2020-12-22 | 南京航空航天大学 | Code self-adaptive generation method based on meta-learning |
| CN112015902A (en) * | 2020-09-14 | 2020-12-01 | 中国人民解放军国防科技大学 | Least-order text classification method under metric-based meta-learning framework |
| JP2022067858A (en) * | 2020-10-21 | 2022-05-09 | セコム株式会社 | Learned model and data processor |
| CN112949902A (en) * | 2021-01-25 | 2021-06-11 | 西北工业大学 | Runoff prediction method based on LSTM multi-state vector sequence-to-sequence model |
| CN114239949A (en) * | 2021-12-10 | 2022-03-25 | 中信银行股份有限公司 | Website access amount prediction method and system based on two-stage attention mechanism |
| CN114861522A (en) * | 2022-04-11 | 2022-08-05 | 深圳贝尔信息科技有限公司 | Precision manufacturing quality monitoring method and device based on artificial intelligence meta-learning technology |
| CN114972232A (en) * | 2022-05-17 | 2022-08-30 | 西安电子科技大学 | No-reference image quality evaluation method based on incremental meta-learning |
| CN115187569A (en) * | 2022-07-26 | 2022-10-14 | 北京芯联心科技发展有限公司 | Image quality evaluation model training method, device, equipment and medium |
| CN115297496A (en) * | 2022-09-28 | 2022-11-04 | 南昌航空大学 | Link quality prediction method combining Bi-LSTM and time mode attention |
| CN115906399A (en) * | 2022-10-11 | 2023-04-04 | 电子科技大学 | Improved method for predicting key process quality of product under small sample data |
| CN115908256A (en) * | 2022-10-19 | 2023-04-04 | 青岛大学 | Yarn quality prediction method based on meta-learning method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116502959A (en) | 2023-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110083125B (en) | Machine tool thermal error modeling method based on deep learning | |
| CN116502959B (en) | Product manufacturing quality prediction method based on meta learning | |
| Huang et al. | Exploiting fractional accumulation and background value optimization in multivariate interval grey prediction model and its application | |
| CN111832101A (en) | Construction method of cement strength prediction model and cement strength prediction method | |
| CN112733417A (en) | Abnormal load data detection and correction method and system based on model optimization | |
| CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
| CN116894180B (en) | Product manufacturing quality prediction method based on different composition attention network | |
| CN112508286A (en) | Short-term load prediction method based on Kmeans-BilSTM-DMD model | |
| CN116107279A (en) | Flow industrial energy consumption multi-objective optimization method based on attention depth neural network | |
| CN114066069A (en) | Combined weight byproduct gas generation amount prediction method | |
| CN111192158A (en) | Transformer substation daily load curve similarity matching method based on deep learning | |
| He et al. | A decomposition-based multi-objective particle swarm optimization algorithm with a local search strategy for key quality characteristic identification in production processes | |
| CN117458480A (en) | Photovoltaic power generation power short-term prediction method and system based on improved LOF | |
| CN112381591A (en) | Sales prediction optimization method based on LSTM deep learning model | |
| CN116562454A (en) | Manufacturing cost prediction method applied to BIM long-short-time attention mechanism network | |
| CN115660038A (en) | Multi-stage integrated short-term load prediction based on error factors and improved MOEA/D-SAS | |
| CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
| CN115061444A (en) | Real-time optimization method for technological parameters integrating probability network and reinforcement learning | |
| CN115908909A (en) | Evolutionary neural architecture searching method and system based on Bayes convolutional neural network | |
| CN115630582A (en) | Multi-sliding-window model fused soft rock tunnel surrounding rock deformation prediction method and equipment | |
| Chi et al. | Calculation method of probability integration method parameters based on MIV-GP-BP model | |
| CN112348275A (en) | Regional ecological environment change prediction method based on online incremental learning | |
| Yi et al. | Fast Sensitivity-Analysis-Based Online Self-Organizing Broad Learning System | |
| CN114841472B (en) | GWO optimization Elman power load prediction method based on DNA hairpin variation | |
| CN117975174B (en) | Three-dimensional digital core reconstruction method based on improvement VQGAN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |