CN115098824B - BP neural network-based ultrasonic sensor sensitivity compensation curve construction method - Google Patents
BP neural network-based ultrasonic sensor sensitivity compensation curve construction method Download PDFInfo
- Publication number
- CN115098824B CN115098824B CN202210755513.8A CN202210755513A CN115098824B CN 115098824 B CN115098824 B CN 115098824B CN 202210755513 A CN202210755513 A CN 202210755513A CN 115098824 B CN115098824 B CN 115098824B
- Authority
- CN
- China
- Prior art keywords
- neural network
- data
- training
- error
- actual
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 44
- 230000035945 sensitivity Effects 0.000 title claims abstract description 27
- 238000010276 construction Methods 0.000 title claims description 5
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000002604 ultrasonography Methods 0.000 abstract 2
- 238000004088 simulation Methods 0.000 abstract 1
- 238000001514 detection method Methods 0.000 description 13
- 229910000831 Steel Inorganic materials 0.000 description 4
- 239000002131 composite material Substances 0.000 description 4
- 229920001940 conductive polymer Polymers 0.000 description 4
- 239000010959 steel Substances 0.000 description 4
- 230000006378 damage Effects 0.000 description 3
- 230000007774 longterm Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001125 extrusion Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000001151 other effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Databases & Information Systems (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
By collecting ultrasound sensor data in different scenarios and recording sensitivity information of the ultrasound sensor. The method comprises the steps of using ultrasonic sensors of the same type with different sensitivities in a laboratory environment to obtain ultrasonic sensing data with different sensitivities, correcting errors of simulation data according to actual data, detecting test data by using a trained BP neural network, configuring parameters according to an actual application scene by using the trained BP neural network, and inputting acquired data into the BP neural network for training to obtain an ultrasonic sensor sensitivity compensation curve under the scene.
Description
Technical Field
The invention belongs to the field of strain sensor post-processing, and particularly relates to a method for generating hysteresis phenomenon of a strain sensor based on a flexible conductive polymer composite material based on BP (back propagation) neural network compensation, which is simple in calculation and accurate in compensation.
Background
Railway operation safety is a national strategic problem. The improvement of the railway operation speed brings higher requirements to the safety and the reliability, and particularly the railway maintenance detection work is important; in order to ensure the safety and smoothness of railway transportation and improve the operation efficiency, the integrity rate of equipment must be maintained for the whole national economy service, and the state of key equipment, especially the supporting body-steel rail of a train, is monitored in real time; during long-term use of the steel rail in the railway, peeling, spotting, cracking and breaking occur due to fatigue, and other steel rail injuries affecting the performance of the steel rail are caused; in a high-speed railway system, the long-term collision, extrusion and other effects generated by a high-speed train are more prominent, the probability of occurrence of cracks and the speed of crack propagation are both improved, if the cracks are not detected in time and safety measures are taken, the cracks are easily propagated under the external force of the follow-up continuous action, so that serious accidents such as rail breakage, even derailment and the like of the train are caused; therefore, the detection of the rail damage is a key technology for grasping the safety state of the rail, and is an indispensable condition for guaranteeing the safe operation of the high-speed rail.
The inventor finds that in the current-stage railway rail flaw detection work, most of the railway rail flaw detection work is to collect data on a line through an ultrasonic flaw detection vehicle, then directly display the collected damage information on a terminal in a B scanning mode, judge whether a rail has a flaw through flaw detection personnel, send a recheck to the field flaw detection personnel if a suspicious flaw exists, and confirm the flaw after recheck and then carry out subsequent repair treatment; the working mode is very dependent on flaw detection experience and interpretation capability of operators, the detection rate of a damaged rail is greatly related to the professional level of equipment operators, and the process of finding the damaged position from flaw detection data consumes a lot of time, which forms an important contradiction with the rapid growth of railway lines in China at present.
An obstacle detection system for a vehicle generally employs a sensor such as an ultrasonic sensor mounted on the vehicle body for detecting surrounding obstacles using ultrasonic waves. Generally, an ultrasonic sensor emits sound waves around a vehicle and senses obstacles through the sound waves reflected by the obstacles. To this end, the sensor determines the time interval at which the reflected sound wave is received and uses the known speed of sound to identify the distance between the obstacle and the vehicle.
It should be appreciated that in dry air at 20deg.C (68deg.F), the speed of sound is 343m/s. However, weather conditions affect the behavior of sound waves, and the speed of sound varies with pressure, temperature, and humidity. Ultrasonic sensors are generally used in a variety of vehicles to provide parking assistance, collision detection, automatic parking, or any other type of obstacle avoidance capability. Thus, suitable obstacle detection is required to avoid damaging the vehicle.
Currently, the speed of sound is compensated for changes in atmospheric temperature by employing an ambient temperature sensor mounted on the vehicle. However, this can be expensive. In addition, there is currently no solution to compensate the ultrasonic sensor for atmospheric pressure changes.
Disclosure of Invention
An ultrasonic sensor sensitivity compensation curve construction method based on BP neural network comprises the following steps:
The method comprises the steps of collecting ultrasonic sensing data under different scenes, recording sensitivity information of an ultrasonic sensor, using the ultrasonic sensors with the same type and different sensitivities in a laboratory environment to obtain the ultrasonic sensing data under different sensitivities, and carrying out error correction on analog data according to actual data;
using a BP neural network with double-layer input, positioning the number of hidden layer units as a set constant N, using tansing for a hidden layer activation function, using a ReLU for an activation function, using a clustering model based on K-means to acquire a sensitivity clustering center K of ultrasonic sensing according to actual acquired data, using ultrasonic data with sensitivity K 1 meeting the requirements of I K-K 1|<ts1 as test data, wherein ts 1 is a set first judgment threshold value, and using the rest data as training data;
Training the training data by using a defined BP neural network, calculating the error of the BP neural network according to the range of 2N weights between an input layer and a hidden layer and the range of N weights between the hidden layer and an output layer and combining a threshold ts 2 set by the output layer, and when the error does not meet the error standard, modifying the weights and the threshold until the error meets the error standard;
Detecting test data by using a trained BP neural network, firstly carrying out normalization operation on the test data to obtain a maximum value K max and a minimum value K min in the test data, and carrying out standard mapping on any one original test data K 2 to obtain corrected data
Setting the maximum iteration number and learning rate of the BP neural network to obtain test data, obtaining an analysis value through the BP neural network, comparing the analysis value with an actual value to obtain an error value, and calculating the consistency of a prediction curve { Kp i } and an actual curve { Ks i }
Wherein N sp is the number of points of the predicted curve and the actual curve
When g ps is larger than a third judgment threshold ts 3, judging that the BP neural network can be used after training, otherwise, judging that the BP neural network can not be used after training, extracting a point that the error of an actual value and a predicted value exceeds a fourth judgment threshold ts 4, and retraining until the BP neural network can be used after training;
And carrying out parameter configuration according to the trained BP neural network and the actual application scene, inputting acquired data into the BP neural network for training to obtain an ultrasonic sensor sensitivity compensation curve in the scene.
Compared with the prior art, the invention has the advantages that:
Compared with the prior art, the invention provides a strain sensor hysteresis compensation method based on a flexible conductive polymer composite material, which can compensate hysteresis generated in use of a sensor by combining a small amount of measured data with simple steps.
Compared with a function correction method and a piecewise linear interpolation method, the strain sensor based on the flexible conductive polymer composite material is compensated by adopting the BP neural network compensation sensor, so that the accuracy is remarkably improved, and the data processing and the application are facilitated.
The 3-layer BP neural network is adopted to realize the compensation of the strain sensor based on the flexible conductive polymer composite material, and the method is simple to realize and high in accuracy, and greatly improves the accuracy of the sensor in the use process.
Detailed Description
The invention is further illustrated by the following examples. The examples of the present invention are intended to better understand the present invention to those skilled in the art, and are not intended to limit the present invention in any way.
An ultrasonic sensor sensitivity compensation curve construction method based on BP neural network comprises the following steps:
The method comprises the steps of collecting ultrasonic sensing data under different scenes, recording sensitivity information of an ultrasonic sensor, using the ultrasonic sensors with the same type and different sensitivities in a laboratory environment to obtain the ultrasonic sensing data under different sensitivities, and carrying out error correction on analog data according to actual data;
using a BP neural network with double-layer input, positioning the number of hidden layer units as a set constant N, using tansing for a hidden layer activation function, using a ReLU for an activation function, using a clustering model based on K-means to acquire a sensitivity clustering center K of ultrasonic sensing according to actual acquired data, using ultrasonic data with sensitivity K 1 meeting the requirements of I K-K 1|<ts1 as test data, wherein ts 1 is a set first judgment threshold value, and using the rest data as training data;
Training the training data by using a defined BP neural network, calculating the error of the BP neural network according to the range of 2N weights between an input layer and a hidden layer and the range of N weights between the hidden layer and an output layer and combining a threshold ts 2 set by the output layer, and when the error does not meet the error standard, modifying the weights and the threshold until the error meets the error standard;
Detecting test data by using a trained BP neural network, firstly carrying out normalization operation on the test data to obtain a maximum value K max and a minimum value K min in the test data, and carrying out standard mapping on any one original test data K 2 to obtain corrected data
Setting the maximum iteration number and learning rate of the BP neural network to obtain test data, obtaining an analysis value through the BP neural network, comparing the analysis value with an actual value to obtain an error value, and calculating the consistency of a prediction curve { Kp i } and an actual curve { Ks i }
Wherein N sp is the number of points of the predicted curve and the actual curve
When g ps is larger than a third judgment threshold ts 3, judging that the BP neural network can be used after training, otherwise, judging that the BP neural network can not be used after training, extracting a point that the error of an actual value and a predicted value exceeds a fourth judgment threshold ts 4, and retraining until the BP neural network can be used after training;
And carrying out parameter configuration according to the trained BP neural network and the actual application scene, inputting acquired data into the BP neural network for training to obtain an ultrasonic sensor sensitivity compensation curve in the scene.
Claims (1)
1. The ultrasonic sensor sensitivity compensation curve construction method based on the BP neural network is characterized by comprising the following steps of:
The method comprises the steps of collecting ultrasonic sensing data under different scenes, recording sensitivity information of an ultrasonic sensor, using the ultrasonic sensors with the same type and different sensitivities in a laboratory environment to obtain the ultrasonic sensing data under different sensitivities, and carrying out error correction on analog data according to actual data;
using a BP neural network with double-layer input, positioning the number of hidden layer units as a set constant N, using tansing for a hidden layer activation function, using a ReLU for an activation function, using a clustering model based on K-means to acquire a sensitivity clustering center K of ultrasonic sensing according to actual acquired data, using ultrasonic data with sensitivity K 1 meeting the requirements of I K-K 1|<ts1 as test data, wherein ts 1 is a set first judgment threshold value, and using the rest data as training data;
Training the training data by using a defined BP neural network, calculating the error of the BP neural network according to the range of 2N weights between an input layer and a hidden layer and the range of N weights between the hidden layer and an output layer and combining a threshold ts 2 set by the output layer, and when the error does not meet the error standard, modifying the weights and the threshold until the error meets the error standard;
Detecting test data by using a trained BP neural network, firstly carrying out normalization operation on the test data to obtain a maximum value K max and a minimum value K min in the test data, and carrying out standard mapping on any one original test data K 2 to obtain corrected data
Setting the maximum iteration number and learning rate of the BP neural network to obtain test data, obtaining an analysis value through the BP neural network, comparing the analysis value with an actual value to obtain an error value, and calculating the consistency of a prediction curve { Kp i } and an actual curve { Ks i }
Wherein N sp is the number of points of the predicted curve and the actual curve
When g ps is larger than a third judgment threshold ts 3, judging that the BP neural network can be used after training, otherwise, judging that the BP neural network can not be used after training, extracting a point that the error of an actual value and a predicted value exceeds a fourth judgment threshold ts 4, and retraining until the BP neural network can be used after training;
And carrying out parameter configuration according to the trained BP neural network and the actual application scene, inputting acquired data into the BP neural network for training to obtain an ultrasonic sensor sensitivity compensation curve in the scene.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210755513.8A CN115098824B (en) | 2022-06-28 | 2022-06-28 | BP neural network-based ultrasonic sensor sensitivity compensation curve construction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210755513.8A CN115098824B (en) | 2022-06-28 | 2022-06-28 | BP neural network-based ultrasonic sensor sensitivity compensation curve construction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115098824A CN115098824A (en) | 2022-09-23 |
CN115098824B true CN115098824B (en) | 2024-04-19 |
Family
ID=83295339
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210755513.8A Active CN115098824B (en) | 2022-06-28 | 2022-06-28 | BP neural network-based ultrasonic sensor sensitivity compensation curve construction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115098824B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106679880A (en) * | 2016-12-21 | 2017-05-17 | 华南理工大学 | Pressure sensor temperature compensating method based on FOA-optimized SOM-RBF |
CN111145042A (en) * | 2019-12-31 | 2020-05-12 | 国网北京市电力公司 | Power distribution network voltage abnormity diagnosis method adopting full-connection neural network |
CN112287788A (en) * | 2020-10-20 | 2021-01-29 | 杭州电子科技大学 | Pedestrian detection method based on improved YOLOv3 and improved NMS |
CN112528568A (en) * | 2020-12-26 | 2021-03-19 | 浙江天行健智能科技有限公司 | Road feel simulation method based on K-Means and BP neural network |
WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
JP2021149893A (en) * | 2020-03-24 | 2021-09-27 | 株式会社東芝 | Neural network analysis apparatus, neural network analysis method, and program |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11231812B2 (en) * | 2020-03-19 | 2022-01-25 | Sensel, Inc. | System and method for calibrating a touch sensor |
-
2022
- 2022-06-28 CN CN202210755513.8A patent/CN115098824B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106679880A (en) * | 2016-12-21 | 2017-05-17 | 华南理工大学 | Pressure sensor temperature compensating method based on FOA-optimized SOM-RBF |
WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
CN111145042A (en) * | 2019-12-31 | 2020-05-12 | 国网北京市电力公司 | Power distribution network voltage abnormity diagnosis method adopting full-connection neural network |
JP2021149893A (en) * | 2020-03-24 | 2021-09-27 | 株式会社東芝 | Neural network analysis apparatus, neural network analysis method, and program |
CN112287788A (en) * | 2020-10-20 | 2021-01-29 | 杭州电子科技大学 | Pedestrian detection method based on improved YOLOv3 and improved NMS |
CN112528568A (en) * | 2020-12-26 | 2021-03-19 | 浙江天行健智能科技有限公司 | Road feel simulation method based on K-Means and BP neural network |
Also Published As
Publication number | Publication date |
---|---|
CN115098824A (en) | 2022-09-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101561379B (en) | Tap-scanning method for detecting structural damages | |
CN109142547B (en) | Acoustic online nondestructive testing method based on convolutional neural network | |
Al-Jumaili et al. | Characterisation of fatigue damage in composites using an Acoustic Emission Parameter Correction Technique | |
CN108804740B (en) | Long-distance pipeline pressure monitoring method based on integrated improved ICA-KRR algorithm | |
CN204463471U (en) | A kind of ship lockage ship's speed detects prior-warning device | |
CN113878214B (en) | Welding quality real-time detection method and system based on LSTM and residual distribution | |
CN101718396A (en) | Method and device for detecting leakage of fluid conveying pipeline based on wavelet and mode identification | |
CN106442635A (en) | Method for recognizing structure layer inside tree on basis of radar waves | |
CN108876771B (en) | Undercut welding defect detection method | |
WO2022179247A9 (en) | Damage determination method based on full cable system cable force measurement and error adaptive analysis | |
CN110568082A (en) | cable wire breakage distinguishing method based on acoustic emission signals | |
CN113657217A (en) | Concrete state recognition model based on improved BP neural network | |
CN115098824B (en) | BP neural network-based ultrasonic sensor sensitivity compensation curve construction method | |
CN115656335A (en) | Method for quickly detecting and identifying internal defects of bearing ring | |
CN111855793A (en) | Seamless rail internal temperature stress early diagnosis method based on surface magnetic memory signal | |
CN111855825B (en) | Rail head nuclear injury identification method and system based on BP neural network | |
CN116953087A (en) | Intelligent detection method for bridge segment assembly construction quality | |
CN113687192B (en) | Method for collecting and positioning discharge signal of power transmission line | |
CN115223044A (en) | End-to-end three-dimensional ground penetrating radar target identification method and system based on deep learning | |
CN113109666A (en) | Track circuit fault diagnosis method based on deep convolutional neural network | |
CN113138048A (en) | Nondestructive live-line detection method for cable joint interface pressure based on stress ultrasound | |
CN116626170B (en) | Fan blade damage two-step positioning method based on deep learning and sound emission | |
CN114925452B (en) | Online guided wave-hidden Markov model crack evaluation method based on dynamic state | |
JP3388180B2 (en) | Target tracking device and target tracking method | |
CN114487103B (en) | Damage detection analysis method based on old part acoustic emission signal chaos characteristic value |
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 |