CN109359673A - A kind of intelligence manufacture failure prediction method and device based on on-line study - Google Patents
A kind of intelligence manufacture failure prediction method and device based on on-line study Download PDFInfo
- Publication number
- CN109359673A CN109359673A CN201811113244.5A CN201811113244A CN109359673A CN 109359673 A CN109359673 A CN 109359673A CN 201811113244 A CN201811113244 A CN 201811113244A CN 109359673 A CN109359673 A CN 109359673A
- Authority
- CN
- China
- Prior art keywords
- data
- node
- module
- vector
- index
- 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
Links
Classifications
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Machine Translation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to machine learning techniques fields, more particularly to a kind of intelligence manufacture failure prediction method and device based on on-line study, by collecting and extracting historical data, using self-organizing map neural network, construct the model that can assess critical component health status trend, and then obtain the health characteristics index of critical component, using the health characteristics index as the magnitude of quantized key unit status, by acquiring the data of critical component generation in real time as input data, obtain failure predication result, approach is extracted and detected the present invention provides a kind of efficient fault signature.
Description
Technical field
The present invention relates to machine learning techniques fields, and in particular to a kind of intelligence manufacture failure predication based on on-line study
Method and device.
Background technique
In machine learning field, on-line study (Online-learning) refers to every time through a trained case-based learning mould
The learning method of type, the purpose of on-line study are the marks of correctly predicted trained example.The most important feature of on-line study
It is that, when primary prediction is completed, correct result is just obtained, this result can be used directly to correction model.
In on-line learning algorithm, we it is not assumed that training data from some probability distribution or random process.When
When training example comes, we classify to it using classifier, and requirement of the on-line Algorithm to data is looser, therefore it
It is more practical algorithm；Meanwhile it is also more practical training algorithm.
And predict that manufacturing quality is one of key measure of quality management, accurately predict the feature learning with manufacturing process
Closely related, therefore, how using big data providing a kind of efficient fault detection approach in intelligence manufacture and becoming is worth solution
Certainly the problem of.
Summary of the invention
The present invention provides a kind of intelligence manufacture failure prediction method and device based on on-line study, can be in intelligence manufacture
In a kind of efficient fault signature be provided extract and detection approach.
A kind of intelligence manufacture failure prediction method based on on-line study provided by the invention, comprising the following steps:
Step S1, the historical data generated to critical component during intelligence manufacture is collected；
Step S2, using the historical data as sample set { xi, carry out predicted characteristics extraction；
Step S3, using self-organizing map neural network, the mould that can assess critical component health status trend is constructed
Type；
Step S4, the health characteristics index for obtaining critical component, using the health characteristics index as quantized key component
The magnitude of state；
Step S5, the data that acquisition critical component generates in real time obtain failure predication using the data as input data
As a result.
Further, the step S2 is specifically included:
Step S21, each data vector x is extractediK neighbor data point；
Step S22, it calculates and x is reconstructed by neighbor data pointiOptimal weights Wij, that is, solve the minimum two of a belt restraining
Multiply, solution formula is as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Step S23, it is based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstructed error φ (Y) is minimum,
Calculation formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
Further, the step S3 is specifically included:
Step S31, data vector x in random alignment mappingiWeight vector Wij；
Step S32, from xiIt is middle to choose any input vector, each node is traversed in the map；
Step S33, the distance between the weight vector of input vector and mapping node is obtained using Euclidean distance formula；
Step S34, the smallest node of selected distance, the node are best match unit；
Step S35, the node for updating best match unit, using the node in the best match unit neighborhood as input
The calculation formula of vector, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit refer to
Number, α (s) are the learning coefficients of a monotone decreasing, and θ (u, v, s) is provided between neuron u and neuron v in the case where step-length is s
The neighbouring function of distance；
Step S36, increase step index s and jump to step S32, until Feature Mapping tends towards stability.
A kind of intelligence manufacture fault prediction device based on on-line study provided by the invention, comprising:
Collection module, the historical data for generating to critical component during intelligence manufacture are collected；
First extraction module, for using the historical data as sample set { xi, carry out predicted characteristics extraction；
Module is constructed, for using self-organizing map neural network, critical component health status can be assessed by, which constructing, becomes
The model of gesture；
Quantization modules are closed for obtaining the health characteristics index of critical component using the health characteristics index as quantization
The magnitude of key member state；
First prediction module, the data for acquiring critical component generation in real time are obtained using the data as input data
Obtain failure predication result.
Further, first extraction module specifically includes:
Second extraction module, for extracting each data vector xiK neighbor data point；
First computing module reconstructs x by neighbor data point for calculatingiOptimal weights Wij, that is, solve a band about
The least square of beam, solution formula are as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Second computing module, for for being based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstruct
Error φ (Y) is minimum, and calculation formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
Further, the building module specifically includes:
Module is arranged, for data vector x in random alignment mappingiWeight vector Wij；
Module is chosen, is used for from xiIt is middle to choose any input vector, each node is traversed in the map；
Third computing module, for being obtained between input vector and the weight vector of mapping node using Euclidean distance formula
Distance；
Second chooses module, is used for the smallest node of selected distance, the node is best match unit；
Update module makees the node in the best match unit neighborhood for updating the node of best match unit
Calculation formula for input vector, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit refer to
Number, α (s) are the learning coefficients of a monotone decreasing, and θ (u, v, s) is provided between neuron u and neuron v in the case where step-length is s
The neighbouring function of distance；
Stable module, for increasing step index s and executing selection module, until Feature Mapping tends towards stability.
The beneficial effects of the present invention are: the present invention disclose a kind of intelligence manufacture failure prediction method based on on-line study and
Device, by collecting historical data, using self-organizing map neural network, critical component health status can be assessed by, which constructing, becomes
The model of gesture, so that the data that acquisition critical component generates in real time obtain failure predication as a result, the present invention mentions as input data
A kind of efficient fault signature has been supplied to extract and detect approach.
Detailed description of the invention
The invention will be further described with example with reference to the accompanying drawing.
Fig. 1 is a kind of flow chart of the intelligence manufacture failure prediction method based on on-line study of the embodiment of the present invention；
Fig. 2 is a kind of process of the intelligence manufacture failure prediction method step S2 based on on-line study of the embodiment of the present invention
Figure；
Fig. 3 is the process in a kind of intelligence manufacture failure prediction method step S3 based on on-line study of the embodiment of the present invention
Figure.
Specific embodiment
With reference to Fig. 1~3, a kind of intelligence manufacture failure prediction method based on on-line study provided in an embodiment of the present invention,
The following steps are included:
Step S1, the historical data generated to critical component during intelligence manufacture is collected；
Step S2, using the historical data as sample set { xi, carry out predicted characteristics extraction；
Step S3, using self-organizing map neural network, the mould that can assess critical component health status trend is constructed
Type；
Step S4, the health characteristics index for obtaining critical component, using the health characteristics index as quantized key component
The magnitude of state；
Step S5, the data that acquisition critical component generates in real time obtain failure predication using the data as input data
As a result.
Further, the step S2 is specifically included:
Step S21, each data vector x is extractediK neighbor data point；
Step S22, it calculates and x is reconstructed by neighbor data pointiOptimal weights Wij, that is, solve the minimum two of a belt restraining
Multiply, solution formula is as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Step S23, it is based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstructed error φ (Y) is minimum,
Calculation formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
Further, the step S3 is specifically included:
Step S31, data vector x in random alignment mappingiWeight vector Wij；
Step S32, from xiIt is middle to choose any input vector, each node is traversed in the map；
Step S33, the distance between the weight vector of input vector and mapping node is obtained using Euclidean distance formula；
Step S34, the smallest node of selected distance, the node are best match unit；
Step S35, the node for updating best match unit, using the node in the best match unit neighborhood as input
The calculation formula of vector, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit refer to
Number, α (s) are the learning coefficients of a monotone decreasing, and θ (u, v, s) is provided between neuron u and neuron v in the case where step-length is s
The neighbouring function of distance；
Step S36, increase step index s and jump to step S32, until Feature Mapping tends towards stability.
A kind of intelligence manufacture fault prediction device based on on-line study provided by the invention, comprising:
Collection module, the historical data for generating to critical component during intelligence manufacture are collected；
First extraction module, for using the historical data as sample set { xi, carry out predicted characteristics extraction；
Module is constructed, for using self-organizing map neural network, critical component health status can be assessed by, which constructing, becomes
The model of gesture；
Quantization modules are closed for obtaining the health characteristics index of critical component using the health characteristics index as quantization
The magnitude of key member state；
First prediction module, the data for acquiring critical component generation in real time are obtained using the data as input data
Obtain failure predication result.
Further, first extraction module specifically includes:
Second extraction module, for extracting each data vector xiK neighbor data point；
First computing module reconstructs x by neighbor data point for calculatingiOptimal weights Wij, that is, solve a band about
The least square of beam, solution formula are as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Second computing module, for for being based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstruct
Error φ (Y) is minimum, and calculation formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
Further, the building module specifically includes:
Module is arranged, for data vector x in random alignment mappingiWeight vector Wij；
Module is chosen, is used for from xiIt is middle to choose any input vector, each node is traversed in the map；
Third computing module, for being obtained between input vector and the weight vector of mapping node using Euclidean distance formula
Distance；
Second chooses module, is used for the smallest node of selected distance, the node is best match unit；
Update module makees the node in the best match unit neighborhood for updating the node of best match unit
Calculation formula for input vector, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit refer to
Number, α (s) are the learning coefficients of a monotone decreasing, and θ (u, v, s) is provided between neuron u and neuron v in the case where step-length is s
The neighbouring function of distance；
Stable module, for increasing step index s and executing selection module, until Feature Mapping tends towards stability.
The above, only presently preferred embodiments of the present invention, the invention is not limited to above embodiment, as long as
It reaches technical effect of the invention with identical means, all should belong to protection scope of the present invention.
Claims (6)
1. a kind of intelligence manufacture failure prediction method based on on-line study, which comprises the following steps:
Step S1, the historical data generated to critical component during intelligence manufacture is collected；
Step S2, using the historical data as sample set { xi, carry out predicted characteristics extraction；
Step S3, using self-organizing map neural network, the model that can assess critical component health status trend is constructed；
Step S4, the health characteristics index for obtaining critical component, using the health characteristics index as quantized key unit status
Magnitude；
Step S5, the data that acquisition critical component generates in real time obtain failure predication knot using the data as input data
Fruit.
2. a kind of intelligence manufacture failure prediction method based on on-line study according to claim 1, which is characterized in that institute
Step S2 is stated to specifically include:
Step S21, each data vector x is extractediK neighbor data point；
Step S22, it calculates and x is reconstructed by neighbor data pointiOptimal weights Wij, that is, the least square of a belt restraining is solved,
Solution formula is as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Step S23, it is based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstructed error φ (Y) is minimum, calculate
Formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
3. a kind of intelligence manufacture failure prediction method based on on-line study according to claim 1, which is characterized in that institute
Step S3 is stated to specifically include:
Step S31, data vector x in random alignment mappingiWeight vector Wij；
Step S32, from xiIt is middle to choose any input vector, each node is traversed in the map；
Step S33, the distance between the weight vector of input vector and mapping node is obtained using Euclidean distance formula；
Step S34, the smallest node of selected distance, the node are best match unit；
Step S35, update best match unit node, using the node in the best match unit neighborhood as input to
The calculation formula of amount, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit index, α
(s) be a monotone decreasing learning coefficient, θ (u, v, s) be step-length be s under provide distance between neuron u and neuron v
Neighbouring function；
Step S36, increase step index s and jump to step S32, until Feature Mapping tends towards stability.
4. a kind of intelligence manufacture fault prediction device based on on-line study characterized by comprising
Collection module, the historical data for generating to critical component during intelligence manufacture are collected；
First extraction module, for using the historical data as sample set { xi, carry out predicted characteristics extraction；
Module is constructed, for using self-organizing map neural network, critical component health status trend can be assessed by constructing
Model；
Quantization modules, for obtaining the health characteristics index of critical component, using the health characteristics index as quantized key portion
The magnitude of part state；
First prediction module, using the data as input data, obtains event for acquiring the data of critical component generation in real time
Hinder prediction result.
5. a kind of intelligence manufacture fault prediction device based on on-line study according to claim 4, which is characterized in that institute
The first extraction module is stated to specifically include:
Second extraction module, for extracting each data vector xiK neighbor data point；
First computing module reconstructs x by neighbor data point for calculatingiOptimal weights Wij, that is, solve a belt restraining
Least square, solution formula are as follows:
It enables reconstructed error ε (W) minimum, obtains optimal weights Wij；
Second computing module, for for being based on WijCalculate optimal reconstruct low-dimensional data vector yiEven following reconstructed error
φ (Y) is minimum, and calculation formula is as follows:
Y is calculatedi, constitute low-dimensional characteristic data set { yi}。
6. a kind of intelligence manufacture fault prediction device based on on-line study according to claim 4, which is characterized in that institute
Building module is stated to specifically include:
Module is arranged, for data vector x in random alignment mappingiWeight vector Wij；
Module is chosen, is used for from xiIt is middle to choose any input vector, each node is traversed in the map；
Third computing module, for using Euclidean distance formula obtain between input vector and the weight vector of mapping node away from
From；
Second chooses module, is used for the smallest node of selected distance, the node is best match unit；
Update module, for updating the node of best match unit, using the node in the best match unit neighborhood as defeated
The calculation formula of incoming vector, the node for updating best match unit is as follows:
WV(s+1)=WV(s)+θ (u, v, s) α (s) (D (t)-WV(s))
Wherein s is step index, and i is the index of training sample, xiIt is input vector, u is xiBest match unit index, α
(s) be a monotone decreasing learning coefficient, θ (u, v, s) be step-length be s under provide distance between neuron u and neuron v
Neighbouring function；
Stable module, for increasing step index s and executing selection module, until Feature Mapping tends towards stability.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811113244.5A CN109359673A (en) | 2018-09-25 | 2018-09-25 | A kind of intelligence manufacture failure prediction method and device based on on-line study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811113244.5A CN109359673A (en) | 2018-09-25 | 2018-09-25 | A kind of intelligence manufacture failure prediction method and device based on on-line study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109359673A true CN109359673A (en) | 2019-02-19 |
Family
ID=65351342
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811113244.5A Pending CN109359673A (en) | 2018-09-25 | 2018-09-25 | A kind of intelligence manufacture failure prediction method and device based on on-line study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109359673A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651573A (en) * | 2020-12-31 | 2021-04-13 | 上海竞动科技有限公司 | Risk prediction method and device based on deep learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104537415A (en) * | 2014-12-02 | 2015-04-22 | 北京化工大学 | Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM |
| CN107941537A (en) * | 2017-10-25 | 2018-04-20 | 南京航空航天大学 | A kind of mechanical equipment health state evaluation method |
-
2018
- 2018-09-25 CN CN201811113244.5A patent/CN109359673A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104537415A (en) * | 2014-12-02 | 2015-04-22 | 北京化工大学 | Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM |
| CN107941537A (en) * | 2017-10-25 | 2018-04-20 | 南京航空航天大学 | A kind of mechanical equipment health state evaluation method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651573A (en) * | 2020-12-31 | 2021-04-13 | 上海竞动科技有限公司 | Risk prediction method and device based on deep learning |
| CN112651573B (en) * | 2020-12-31 | 2021-07-06 | 上海竞动科技有限公司 | Risk prediction method and device based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104318261B (en) | A kind of sparse representation face identification method representing recovery based on figure embedding low-rank sparse | |
| US11418029B2 (en) | Method for recognizing contingencies in a power supply network | |
| CN109784692B (en) | Rapid safety constraint economic dispatching method based on deep learning | |
| US11487266B2 (en) | Method for recognizing contingencies in a power supply network | |
| US11860608B2 (en) | Industrial equipment operation, maintenance and optimization method and system based on complex network model | |
| CN109409024B (en) | Photovoltaic module voltage and current characteristic modeling method based on one-dimensional depth residual error network | |
| CN108490115B (en) | Air quality abnormity detection method based on distributed online principal component analysis | |
| CN112383052A (en) | Power grid fault repairing method and device based on power internet of things | |
| Pumir et al. | Smoothed analysis of the low-rank approach for smooth semidefinite programs | |
| Tasfi et al. | Deep neural networks with confidence sampling for electrical anomaly detection | |
| CN116029221B (en) | Power equipment fault diagnosis method, device, equipment and medium | |
| CN117272196A (en) | Industrial time sequence data anomaly detection method based on time-space diagram attention network | |
| Omitaomu et al. | Online support vector regression with varying parameters for time-dependent data | |
| CN110334478A (en) | Machinery equipment abnormality detection model building method, detection method and model | |
| CN114298134A (en) | Wind power prediction method and device and electronic equipment | |
| Gu et al. | Identification of concurrent control chart patterns with singular spectrum analysis and learning vector quantization | |
| CN113723541B (en) | Slope displacement prediction method based on hybrid intelligent algorithm | |
| CN109359673A (en) | A kind of intelligence manufacture failure prediction method and device based on on-line study | |
| Liu et al. | State of health estimation of lithium-ion batteries based on multi-feature extraction and temporal convolutional network | |
| CN117349595A (en) | Industrial scene-oriented extremely few sample fault diagnosis method | |
| Ren et al. | AdaM: Adaptive-maximum imputation for neighborhood-based collaborative filtering | |
| Yu et al. | A hybrid learning-based model for on-line monitoring and diagnosis of out-of-control signals in multivariate manufacturing processes | |
| Peng et al. | Cross-band spectrum prediction algorithm based on transfer learning and meta learning | |
| EP3816805A1 (en) | Analysis of anomalies in a facility | |
| CN110739030B (en) | Soft measurement method for small sample in ethylene production process |
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 |