CN106682685B - Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning - Google Patents
Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning Download PDFInfo
- Publication number
- CN106682685B CN106682685B CN201611106593.5A CN201611106593A CN106682685B CN 106682685 B CN106682685 B CN 106682685B CN 201611106593 A CN201611106593 A CN 201611106593A CN 106682685 B CN106682685 B CN 106682685B
- Authority
- CN
- China
- Prior art keywords
- data
- microwave heating
- layer
- temperature
- parameters
- 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
Images
Classifications
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Control Of High-Frequency Heating Circuits (AREA)
Abstract
The invention discloses a local temperature change anomaly detection method based on deep learning of microwave heating temperature field distribution characteristics, which is characterized in that multidimensional big data distributed in the temperature field in the microwave heating process are learned by adopting a method of combining a convolutional neural network and automatic coding, the deep logical relationship among the data is found out, the characteristics with more expressive ability than input are obtained by learning the structure of the data, and then anomaly detection is carried out by adopting an Isolation Forest algorithm. The invention can reliably detect the phenomenon that the dielectric coefficient and the thermal conductivity of the heated medium change along with the rise of the temperature due to the coupling of the complex time-varying electromagnetic field and the temperature field in the microwave heating process, so that the medium is locally overheated and even thermally out of control. And then timely processing is carried out, and safety accidents are avoided.
Description
Technical Field
The invention relates to a microwave heating control technology.
Background
Microwave heating essentially utilizes the energy characteristics of microwaves. Microwave energy directly penetrates through the medium, and molecules in the medium absorb the microwave energy to make the molecular thermal motion aggravated, so that the temperature rises, and the heating purpose is achieved. Compared with other traditional heating modes, the heating device has the advantages of high efficiency, no pollution, high heating speed, small heat loss and the like. Microwave heating is a new cleaning heating mode and has no doubt great application value.
In the microwave heating process, however, a coupling condition of a complex time-varying electromagnetic field and a temperature field is involved, the dielectric coefficient and the thermal conductivity of the heated medium change with the rise of temperature, and the change of the variable factors can cause the phenomenon of local overheating and even thermal runaway of the medium. This situation can lead to safety accidents if not handled in a timely manner.
Disclosure of Invention
The invention aims to solve the problem of local overheating (thermal runaway) in the microwave heating process.
The technical scheme adopted for achieving the purpose of the invention is that the local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning is characterized by comprising the following steps:
obtaining a temperature anomaly detection model of the microwave heating device through the steps 1-4:
1) arranging m acquisition nodes in a microwave heating temperature field of a microwave heating device, and acquiring temperature data of each acquisition node at n0 moments to form a multi-dimensional data sample set Cq;
2) The multidimensional data sample set C obtained in the step 1 is processedqCarrying out normalization processing to obtain data
3) Depth feature extraction:
3-1) selecting a training sample S in the multi-dimensional data of the normalized microwave distribution fieldqAs an input variable, a convolution kernel size L is setkAnd step length AkObtaining a data set H formed by the characteristic mapping valuesk。
3-2) setting the size of the pooling layer, and processing the data set H by adopting a maximum pooling methodkObtaining a data set Ik,
3-3) data set IkThe automatic encoder is brought in, parameters of each layer of the network are reversely corrected by utilizing a back propagation algorithm through the minimum reconstruction error, and finally the characteristics are learnedRepeating the step 3), continuously learning new sample data, and finally updating to obtain a characteristic data set through layer-by-layer iteration of the network
during operation, whether the temperature of the microwave heating device is abnormal is obtained through the following steps:
A) recording the temperature data of each acquisition node in m acquisition nodes of the microwave heating device at n0 moments by adopting the same method as the step 1), and storing the temperature data into a matrix E;
B) performing normalization processing on the matrix E by adopting the same method as the step 2) to obtain a matrix F:
C) selecting the matrix F as an input variable, and setting the length L of the convolutional layer sliding windowKAnd step length AKSequentially extracting parameters of each line from input variables, traversing the parameters of each line by using a sliding window, and cutting the parameters into a plurality of data fragmentsWherein u isRounding up to the obtained value, K is 1, 2 … … K, K is the number of convolutional network layers.
D) By data fragmentsAnd a linear filterPerforming convolution, adding an offset pbkAs input of an activation function, wherein the activation function adopts sigmoid, tanh, or relu, and a data set PH formed by feature mapping values obtained by the activation functionk:
E) Setting a pooling window with a GK size and a maximum pooling data set PHkObtaining a data set PIk:
F) To the data set PIkCarrying in an automatic encoder, reversely correcting parameters of each layer of the network layer by using a back propagation algorithm through minimum reconstruction errors, and training to obtain characteristics
Learning new data training samples by adopting the same method as the step 3), updating the input variables in the step 3-1), obtaining new samples for training each layer parameter of the network according to the previous training samples, updating k by k +1, and coiling the length L of the sliding window of the layerk+1And step length Ak+1The length of the pooling window is unchanged. Repeating the process of 3), and finally learning the characteristics
LKAnd AKRespectively setting the length and the step length of the sliding window of the convolutional layer when the steps D) to E) are carried out for the Kth time; gKSetting the size of the pooling window for the Kth time of performing the steps D) to E);
G) feature matrixAs the temperature abnormality detection model obtained in step 4):the output is abnormal or normal.
It is worth to be noted that the invention adopts a multilayer convolution network structure, extracts deep-level feature expressions from an original multi-dimensional data signal, and then transmits the deep-level feature expressions to an Isolation Forest anomaly detection model to detect anomaly features. The method is high-efficiency, can extract better feature expression from the original data acquired in the microwave heating process, brings the extracted features into an anomaly detection system for detection, and can reliably detect whether the temperature change of the heated medium distribution field is abnormal or not.
Drawings
FIG. 1. Process flow of the invention
FIG. 2 is a schematic diagram of convolutional neural network feature extraction
FIG. 3 is a diagram of an automatic coding feature map
FIG. 4 shows the detection of anomalies (red dots for anomalies) in 700 power microwave data using the method of the present invention
FIG. 5 illustrates the detection of anomalies (red dots for anomalies) in 800 power microwave data using the method of the present invention.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
An anomaly detection method of multi-dimensional big data information of temperature field distribution in a microwave heating process based on deep learning,
building a background analysis system and a detection device;
the background analysis system heats the data sample C in the original microwave heating processaDuring analysis, obtaining a temperature anomaly detection model of the microwave heating device through the steps 1-4:
the method comprises the following steps:
1) obtaining a raw data sample CqQ is the data sample number, q is 1, 2 … …,
wherein: microwave data for a node, sample CqThe method comprises the steps that m different nodes are provided, data of each node are collected at time t, and t is 1, 2 … … n 0;
3) Depth feature extraction:
3-1) selecting data samples S in the data setqAs an input variable, the size L of the convolution kernel is setkAnd step length AkSequentially extracting parameters of each line from input variables, traversing the parameters of each line by using a sliding window, and cutting the parameters into a plurality of data fragmentsWherein i isRounding up to obtain a value, wherein K is 1 and 2 … … K, and K is the number of layers of the convolutional network;
3-2) with data fragmentsAnd a linear filterPerforming convolution and adding an offset bkAs the input variable of the activation function, obtaining a data set H formed by feature mapping values through the activation functionkThe activation function can adopt sigmoid, tanh, relu and other functions.
3-3) setting a pooling window with the size of GK, and processing the data set H by adopting a maximum pooling methodkObtaining a data set Ik:
3-4) obtaining a characteristic data set IkSubstituting into an automatic encoder, correcting parameters of each layer of the network layer by layer in a reverse direction by using a back propagation algorithm through minimum reconstruction errors, and training to obtain characteristics
Learning a new data training sample by adopting the same method as the step 3), updating the input variable in the step 3-1) by using the new sample, training the new sample according to the network layer parameters obtained by the previous training sample, updating k by using k +1, keeping the length Lk +1 and the step length Ak +1 of the convolutional layer sliding window unchanged, and keeping the length of the pooling window unchanged. Repeating the process of 3), and finally learning the characteristics
LKAnd AKRespectively setting the size and the step length of a convolution kernel when the Kth time is subjected to the steps 3-2) to 3-3); gKSetting the size of a pooling window when the step 3-2) to the step 3-3) is carried out for the Kth time;
4) with data setsObtaining an abnormal score S (X) for inputting Isolation Forest, and obtaining a temperature abnormal detection model:
the detection device comprises a data collection part and a data analysis part; during working, whether the temperature of the microwave heating device is abnormal is obtained through the steps A to G:
A) the data collection part collects data of the heating nodes, stores the data into a matrix E and transmits the data to the data analysis part;
wherein:
the data of one node is obtained, in a sample E, m different nodes are provided, the parameter of each node is collected at pt time, and pt is 1 and 2 … … pn 0;
B) normalizing the data sample E by the same method as the step 2) to obtain the data sample E
C) Selecting the matrix F as an input variable, and setting the length L of the convolutional layer sliding windowKAnd step length AKSequentially extracting parameters of each line from input variables, traversing the parameters of each line by using a sliding window, and cutting the parameters into a plurality of data fragmentsWherein u isRounding up to the obtained value, K is 1, 2 … … K, K is the number of convolutional network layers.
D) By data fragmentsAnd a linear filterPerforming convolution, adding an offset pbkAs the input of the activation function, wherein the activation function may adopt sigmoid, tanh, relu, etc., and the data set PH formed by the feature mapping values obtained by the activation functionk:
E) Setting a pooling window with a GK size and a maximum pooling data set PHkObtaining a data set PIk:
F) To the data set PIkCarrying in an automatic encoder, reversely correcting parameters of each layer of the network layer by using a back propagation algorithm through minimum reconstruction errors, and training to obtain characteristics
Learning new data training samples by adopting the same method as the step 3), updating the input variables in the step 3-1), training new samples according to network layer parameters obtained by the previous training sample, and updating by using k +1k, the length Lk +1 and the step Ak +1 of the convolutional layer sliding window are unchanged, and the length of the pooling window is also unchanged. Repeating the process of 3), and finally learning the characteristics
LKAnd AKRespectively setting the length and the step length of the sliding window of the convolutional layer when the steps D) to E) are carried out for the Kth time; gKSetting the size of the pooling window for the Kth time of performing the steps D) to E);
Claims (1)
1. A local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning is characterized in that:
obtaining a temperature abnormity detection model of the microwave heating device through steps 1) to 4):
1) in the microwave heating temperature field of the microwave heating device, m collection nodes are arranged, and n is0At each moment, acquiring the temperature data of each acquisition node to form a multi-dimensional data sample set Cq;
2) A multi-dimensional data sample set C obtained in the step 1) is usedqCarrying out normalization processing to obtain data
3) Depth feature extraction:
3-1) selecting a training sample S in the multi-dimensional data of the normalized microwave distribution fieldqAs an input variable, a convolution kernel size L is setKAnd step length AkSequentially extracting parameters of each line from input variables, traversing the parameters of each line by using a sliding window, and cutting the parameters into a plurality of data fragmentsWherein i isRounding up to obtain a value, wherein K is 1 and 2.. K, and K is the number of layers of the convolutional network; by data fragmentsAnd a linear filter Wi kPerforming convolution and adding an offset bkAs input of the activation function, a data set H formed by feature mapping values is obtained through the activation functionk;
3-2) setting the size of the pooling layer, and processing the data set H by adopting a maximum pooling methodKObtaining a data set Ik,
3-3) data set IkThe automatic encoder is brought in, parameters of each layer of the network are reversely corrected by utilizing a back propagation algorithm through the minimum reconstruction error, and finally the characteristics are learnedRepeating the step 3), continuously learning new sample data, and finally updating to obtain characteristics through layer-by-layer iteration of the network
during working, whether the temperature of the microwave heating device is abnormal is obtained through the steps A to G:
A) recording m collection nodes of the microwave heating device by the same method as the step 1), wherein each collection node is positioned at n0Temperature data of each moment are stored in a matrix E;
B) normalizing the matrix E by adopting the same method as the step 2) to obtain a matrix F;
C) selecting the matrix F as an input variable, and setting the length L of the convolutional layer sliding windowKAnd step length AkSequentially extracting parameters of each line from input variables, traversing the parameters of each line by using a sliding window, and cutting the parameters into a plurality of data fragmentsWherein u isRounding up to obtain a value, wherein K is 1 and 2.. K, and K is the number of layers of the convolutional network; pn-N complex0The total number of the acquisition time points of the matrix E is;
D) by data fragmentsAnd a linear filterPerforming convolution, adding an offset pbkAs input of an activation function, wherein the number of activations is sigmoid, tanh, or relu, and a data set PH formed by feature mapping values obtained by the activation functionk:
E) Is provided withIs sized as GKPooling window, max pooling data set PHkObtaining a data set PIk:
F) To the data set PIkCarrying in an automatic encoder, reversely correcting parameters of each layer of the network layer by using a back propagation algorithm through minimum reconstruction errors, and training to obtain characteristics
Learning new data training samples by adopting the same method as the step 3), updating the input variables in the step 3-1), obtaining new samples for training each layer parameter of the network according to the previous training samples, updating k by k +1, and coiling the length L of the sliding window of the layerk+1And step length Ak+1The length of the pooling window is not changed; repeating the process of 3), and finally learning the characteristics
LKAnd AKRespectively setting the length and the step length of the sliding window of the convolutional layer when the steps D) to E) are carried out for the Kth time; gKSetting the size of the pooling window for the Kth time of performing the steps D) to E);
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611106593.5A CN106682685B (en) | 2016-12-06 | 2016-12-06 | Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611106593.5A CN106682685B (en) | 2016-12-06 | 2016-12-06 | Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106682685A CN106682685A (en) | 2017-05-17 |
CN106682685B true CN106682685B (en) | 2020-05-01 |
Family
ID=58866338
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611106593.5A Active CN106682685B (en) | 2016-12-06 | 2016-12-06 | Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106682685B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108777873A (en) * | 2018-06-04 | 2018-11-09 | 江南大学 | The wireless sensor network abnormal deviation data examination method of forest is isolated based on weighted blend |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109521725A (en) * | 2017-09-20 | 2019-03-26 | 西门子公司 | The method, apparatus and equipment and machine readable media of detection abnormal data |
CN108182450B (en) * | 2017-12-25 | 2021-03-30 | 电子科技大学 | Airborne ground penetrating radar target identification method based on deep convolutional network |
CN108921440B (en) * | 2018-07-11 | 2022-08-05 | 平安科技(深圳)有限公司 | Pollutant abnormity monitoring method, system, computer equipment and storage medium |
CN109655483B (en) * | 2018-12-14 | 2021-06-15 | 四川大学 | Material microstructure defect detection method based on deep learning algorithm |
CN110766056B (en) * | 2019-09-27 | 2022-05-06 | 中山大学 | Abnormal image detection method integrating image generation and multi-label classification |
CN111428886B (en) * | 2020-04-10 | 2023-08-04 | 青岛聚好联科技有限公司 | Method and device for adaptively updating deep learning model of fault diagnosis |
CN113177290B (en) * | 2021-03-25 | 2023-09-26 | 中国人民解放军军事科学院国防科技创新研究院 | Satellite component temperature field prediction method based on depth agent model normalization |
CN113253125B (en) * | 2021-05-19 | 2023-02-17 | 北方工业大学 | Information fusion-based lithium iron phosphate battery thermal runaway monitoring method and system |
CN114547970B (en) * | 2022-01-25 | 2024-02-20 | 中国长江三峡集团有限公司 | Intelligent diagnosis method for abnormality of top cover drainage system of hydropower plant |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104755892A (en) * | 2012-10-26 | 2015-07-01 | 富士通株式会社 | Temperature measuring system and abnormality detecting method |
CN105164508A (en) * | 2013-02-18 | 2015-12-16 | 赛拉诺斯股份有限公司 | Systems and methods for collecting and transmitting assay results |
CN105447512A (en) * | 2015-11-13 | 2016-03-30 | 中国科学院自动化研究所 | Coarse-fine optical surface defect detection method and coarse-fine optical surface defect detection device |
CN105678267A (en) * | 2016-01-08 | 2016-06-15 | 浙江宇视科技有限公司 | Scene recognition method and device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070010906A1 (en) * | 2005-07-11 | 2007-01-11 | Tokyo Electron Limited | Apparatus and system for monitoring a substrate processing, program for monitoring the processing and storage medium storing same |
-
2016
- 2016-12-06 CN CN201611106593.5A patent/CN106682685B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104755892A (en) * | 2012-10-26 | 2015-07-01 | 富士通株式会社 | Temperature measuring system and abnormality detecting method |
CN105164508A (en) * | 2013-02-18 | 2015-12-16 | 赛拉诺斯股份有限公司 | Systems and methods for collecting and transmitting assay results |
CN105447512A (en) * | 2015-11-13 | 2016-03-30 | 中国科学院自动化研究所 | Coarse-fine optical surface defect detection method and coarse-fine optical surface defect detection device |
CN105678267A (en) * | 2016-01-08 | 2016-06-15 | 浙江宇视科技有限公司 | Scene recognition method and device |
Non-Patent Citations (2)
Title |
---|
Monitoring and Extracting Abnormalities in Land Surface Temperature Images for Automatic Identification of Forest Fires;Narasimha Prasad.etc;《2013 European Modelling Symposium》;20140331;第215-219页 * |
基于图像分析的微波加热温度场监测系统研究与实现;崔丽艳;《中国优秀硕士学位论文全文数据库》;20160615;I138-1404 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108777873A (en) * | 2018-06-04 | 2018-11-09 | 江南大学 | The wireless sensor network abnormal deviation data examination method of forest is isolated based on weighted blend |
Also Published As
Publication number | Publication date |
---|---|
CN106682685A (en) | 2017-05-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106682685B (en) | Local temperature change abnormity detection method based on microwave heating temperature field distribution characteristic deep learning | |
CN106682814B (en) | Wind turbine generator fault intelligent diagnosis method based on fault knowledge base | |
CN106600140B (en) | Gas pipeline fault prediction early warning system and method based on improved support vector machine | |
Yin et al. | Wasserstein Generative Adversarial Network and Convolutional Neural Network (WG‐CNN) for Bearing Fault Diagnosis | |
CN108398268B (en) | Bearing performance degradation evaluation method | |
CN104318305B (en) | Inverter low-frequency noise fault diagnosis method based on wavelets and neural network | |
CN109460621A (en) | Based on the gas turbine blades fault detection method for improving neural network | |
CN113132399A (en) | Industrial control system intrusion detection method based on time convolution network and transfer learning | |
CN114363195B (en) | Network flow prediction and early warning method for time and frequency spectrum residual convolution network | |
CN115037543B (en) | Abnormal network flow detection method based on bidirectional time convolution neural network | |
CN111259953B (en) | Equipment defect time prediction method based on capacitive equipment defect data | |
CN109597315A (en) | A kind of mechanical equipment health degenerate state discrimination method, equipment and system | |
CN112231975A (en) | Data modeling method and system based on reliability analysis of railway power supply equipment | |
CN116929815A (en) | Equipment working state monitoring system and method based on Internet of things | |
CN111259949A (en) | Fault identification model construction method, model and identification method for aircraft environmental control system | |
CN104677629A (en) | Fault detection method for vehicle transmission | |
KR102471427B1 (en) | Predictive diagnosis system for mechanical and electrical failures through motor noise and thermal distribution analysis | |
Huang et al. | Automatic classification of magnetic tiles internal defects based on acoustic resonance analysis | |
CN116738367A (en) | Rolling bearing fault diagnosis method based on multi-scale map transducer | |
CN115146675B (en) | Rotary machine migration diagnosis method under variable working condition of depth multi-feature dynamic countermeasure | |
CN103399490B (en) | A kind of carbon fibre precursor wet method coagulation bath temperature control technique that study is controlled based on immunological memory | |
CN105843891A (en) | Incremental online characteristic extraction and analysis method and system | |
CN108459585A (en) | Power station fan method for diagnosing faults based on sparse locally embedding depth convolutional network | |
Ali et al. | Analysis on propeller design for medium-sized drone (DJI phantom 3) | |
Li et al. | Tool wear monitoring using machine learning |
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 |