CN109739181A - A kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network - Google Patents
A kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network Download PDFInfo
- Publication number
- CN109739181A CN109739181A CN201910060083.6A CN201910060083A CN109739181A CN 109739181 A CN109739181 A CN 109739181A CN 201910060083 A CN201910060083 A CN 201910060083A CN 109739181 A CN109739181 A CN 109739181A
- Authority
- CN
- China
- Prior art keywords
- neural network
- image
- machine tool
- main shaft
- detection neural
- 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.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
A kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network, it belongs to numerically-controlled machine tool machining accuracy technical field, this method key step: step 1: using thermal infrared imager acquisition main shaft of numerical control machine tool heating and cooling image, after converting array for image, initial pictures array is subtracted, obtained image array is then converted to image；Step 2: carrying out calibration frame pretreatment to the image after conversion, data set is made；Step 3: selection detection neural network model, the precision of prediction of training pattern；Step 4: input test collection, the precision of prediction of testing model；Step 5: carrying out finite element verifying and experimental verification.The present invention is able to achieve the Thermal Error Robust Modeling under complex working condition, can accurately predict Spindle thermal error, is conducive to improve numerically-controlled machine tool overall processing precision.
Description
Technical field
The invention belongs to numerically-controlled machine tool machining accuracy technical field, in particular to a kind of numerical control based on detection neural network
Thermal Error Model for Spindle of Machine Tool detection method.
Background technique
For Modern Manufacturing Technology because of the fast development of science and technology, the demands such as high-precision, high efficiency, high quality are more urgent,
And this requirement for numerically-controlled machine tool machining accuracy is also higher and higher, the requirement for improving numerically-controlled machine tool machining accuracy is also more compeled
It cuts.
Through a large number of studies show that, machine tool thermal error be influence numerically-controlled machine tool machining accuracy one of principal element, because heat accidentally
Workpiece machining error caused by difference about occupies the 40%-70% of total mismachining tolerance.Therefore, precise numerical control machine is further increased
Machining accuracy, it is necessary to machine tool thermal error is compensated.And main shaft is the core component and main heating source of numerically-controlled machine tool,
Therefore establishing a kind of effective Spindle thermal error modeling detection method is the key that current numerical control machine heat error compensation research.
Although having the largely research in terms of main shaft of numerical control machine tool thermal error modeling method at present, current items error
Modeling technique is difficult to be suitable for various process conditions and the application of numerically-controlled machine tool processing as the solution of general character.And
The reason is that, on the one hand, there are the practical problems of industrial application, further aspect is that current heat error compensation modeling technique can not
It realizes robustness modeling, and not can guarantee data acquisition accuracy.
Summary of the invention
The present invention is to overcome the shortcomings of the existing technology, provide a kind of main shaft of numerical control machine tool Thermal Error based on detection neural network
Model detection method.This method is passed using thermal infrared imager acquisition main shaft of numerical control machine tool heating and cooling image, arrangement displacement
Sensor acquires axially and radially Thermal Error, after pre-processing to image data, using Thermal Error as label, makes image data
Collection, input YOLO v3 detect neural network model, obtain main shaft main heating source position and accurate Spindle thermal error.
Step 1: converting number for image using thermal infrared imager acquisition main shaft of numerical control machine tool heating and cooling image
After group, initial pictures array is subtracted, obtained image array is then converted to image；
Step 2: calibration frame pretreatment is carried out to the image after conversion, to demarcate position data and Thermal Error as label,
And data set is made, the image data set made is divided into training set and forecast set two parts；
Step 3: selection detection neural network model, modifies total losses function, adjusts network structure, start to be trained,
Average error rate reaches 5% hereinafter, deconditioning, preservation model；
Step 4: input test collection, the precision of prediction of testing model, if average error rate is not up to 5% hereinafter, right again
Model is trained, until average error rate reaches 5% or less；
Step 5: carrying out finite element verifying and experimental verification；Simulation analysis is carried out using ANSYS software, will be emulated
To main shaft temperature field picture input to obtain the predicted value of main shaft thermal deformation as detection neural network model, and obtained with emulation
Deformation field data be compared, if the deviation of the two in certain deviation range, i.e., verifying model accuracy meet the requirements,
Otherwise, third step is returned to, ginseng and test verifying are adjusted in the training carried out again；
The experimental data on different lathes is taken to carry out model prediction again, if precision reaches allowed band, method can
Otherwise row returns to second step, ginseng and test verifying are adjusted in the training carried out again.
Further, third step selection YOLO v3 detects neural network model, will be by the Classification Loss in total losses function
Function is changed to L1Loss function, and last classification layer is changed to full articulamentum, sets neuronal quantity as 1, the detection mind
It is full convolutional neural networks through network YOLO v3 detection main body, finally connects two layers of full articulamentum；
Two types of convolution block and residual block are utilized in YOLO v3 detection neural network model calculates, convolution block is pressed
It is executed according to convolutional layer, batch normalization layer and activation primitive layer sequence；
Convolutional layer: each pixel of image is numbered first, xd,i+m,j+nD layer the i-th+m row jth of expression image+
N column pixel；ωd,m,nD layer m row the n-th column weight for indicating filter, uses wbIndicate the bias term of filter；To characteristic pattern
Each element be numbered, use ai,jIndicate the i-th row jth column element of characteristic pattern；D is depth；F is the size of filter, is used
F indicates activation primitive, W2It is the width of characteristic pattern after convolution；W1It is the width of image before convolution；F is the width of filter；P is
The quantity of zero padding, S are stride, H2It is the height of characteristic pattern after convolution, H1It is the width of image before convolution, uses following equation
Calculate convolution；
W2=(W1-F+2P)/S+1
H2=(H1-F+2P)/S+1
Characteristic pattern is obtained in convolutional layer；
Batch normalization layer: the characteristic pattern that convolution obtains is normalized make characteristic pattern matrix numerical value magnitude range [-
1,1]；
Activation primitive layer: LeakyRelu function activates operational formula as follows:
F (x)=max (0, x)+negative_slope × min (0, x)；
Wherein, negative_slope is a small non-zero number；
After calculating, start to be trained model tune ginseng.
The beneficial effect of the present invention compared with prior art is: detection neural network, which has, carries out great amount of images data information
The ability analyzed in real time, and the inherent law of objective for implementation can be excavated from complicated mass image data, Thermal Error can be overcome
Difficulty present in detection technique is modeled, the Thermal Error Robust Modeling detection under complex working condition is realized, can accurately predict main shaft
Thermal Error, and it is subsequent can to test and detection neural network be further improved, realize numerically-controlled machine tool global error inspection
It surveys, improves the overall processing precision of numerically-controlled machine tool.
Detailed description of the invention
Fig. 1 is that the present invention is based on the main shaft of numerical control machine tool thermal error modeling detection process figures of detection neural network；
Fig. 2 is the YOLO v3 detection neural network architecture diagram for adjusting structure；
Fig. 3 is convolution block frame composition in Fig. 2；
Fig. 4 is residual block architecture diagram in Fig. 2；
Fig. 5 is main shaft axial error output quantity schematic diagram；
Fig. 6 is main shaft diameter to error output quantity schematic diagram；
Fig. 7 is the dotted line chart of prediction output valve and the real output value comparison of Spindle thermal error in embodiment；
Fig. 8 is the position registration figure in embodiment between predicted position and true calibration frame position.
Specific embodiment
The present invention is further described with embodiment with reference to the accompanying drawing:
Referring to shown in Fig. 1-Fig. 6, a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network,
Step 1: converting number for image using thermal infrared imager acquisition main shaft of numerical control machine tool heating and cooling image
After group, initial pictures array is subtracted, obtained image array is then converted to image；The step, which carries out processing to image, ensure that absolutely
Temperature rise to property is conducive to the Accurate Prediction of later period Thermal Error without being influenced by environment temperature；It is also beneficial to later period extension pair
Experiment improves, and thermal imaging system arrangement shooting is integrally carried out to lathe, while corresponding every orientation error is label, can be carried out
The heat source of lathe entirety detects and orientation error prediction；
Step 2: calibration frame pretreatment is carried out to the image after conversion, to demarcate position data and Thermal Error as label,
And data set is made, the image data set made is divided into training set and forecast set two parts；
Step 3: selection detection neural network model, modifies total losses function, adjusts network structure, start to be trained,
Average error rate reaches 5% hereinafter, deconditioning, preservation model；The step takes detection neural network model, can accurately mark
The step for determining heat source position, and can accurately predict Thermal Error, and saving screenshot in image procossing early period；
Average error rate calculates:
Wherein: yiBatch true value is trained for i-th,For i-th training batch predicted value, n is i-th training batch
Sample size；N is training total number of samples amount.
Third step selects YOLO v3 to detect neural network model, and the Classification Loss function in total losses function is changed to L1
Loss function, and last classification layer is changed to full articulamentum, sets neuronal quantity as 1, the detection neural network YOLO
It is full convolutional neural networks that v3, which detects main body, finally connects two layers of full articulamentum；Total losses function is the prior art, or quoted from text
Chapter Joseph Redmon, Ali Farhadi.YOLOv3:An Incremental Improvement.arXiv
PreprintarXiv:1804.02767,2018, it can obtain.L1 loss function:Wherein: yiFor i-th
Training batch true value,For i-th training batch predicted value, n is i-th training batch sample size.
Illustrated to detect how neural network carries out calculating training with Fig. 2:
As shown in Figure 2, Figure 3 and Figure 4, detection neural network YOLO v3 detection main body is full convolutional neural networks, is most followed by
Upper two layers of full articulamentum.
As shown in Figure 3 and Figure 4, convolution block and residual block two is utilized in YOLO v3 detection neural network model calculates
A type, convolution block are executed according to convolutional layer, batch normalization layer and activation primitive layer sequence.
Convolutional layer: in order to clearly describe convolutional calculation process, being first numbered each pixel of image,
xd,i+m,j+nIndicate d layer the i-th+m row jth+n column pixel of image, ωd,m,nIndicate that the d layer m row n-th of filter arranges power
Weight, uses wbIndicate the bias term of filter；Each element of characteristic pattern is numbered, a is usedi,jIndicate the i-th row the of characteristic pattern
J column element；D is depth；F is the size (width or height, the two are identical) of filter, indicates activation primitive, W with f2It is convolution
The width of characteristic pattern afterwards；W1It is the width of image before convolution；F is the width of filter；P is the quantity of zero padding (in original graph
0) as several circles of surrounding benefit, S is stride, H2It is the height of characteristic pattern after convolution, H1It is the width of image before convolution；
Convolution is calculated using following equation；
W2=(W1-F+2P)/S+1
H2=(H1-F+2P)/S+1
Characteristic pattern is obtained in convolutional layer；
Batch normalization layer: the characteristic pattern that convolution obtains is normalized, characteristic pattern matrix numerical value magnitude range is made
[-1,1].Convenient for subsequent activation calculating, prevent gradient from disappearing.
Activation primitive layer: LeakyRelu function activates operational formula as follows:
F (x)=max (0, x)+negative_slope × min (0, x)；Wherein x is input feature vector figure tensor size.
Wherein, negative_slope is a small non-zero number；
The convolution kernel size 3 × 3/2 being related to refers to that stride is increased to 2 by 1, makes characteristic pattern using convolution stride
Size reduces, and realizes the function in maximum pond, extracts significantly more efficient image information.
After calculating, start to be trained model tune ginseng, full articulamentum can use gradient descent method, by anti-
Optimal weight parameter and bias term parameter are obtained to propagation algorithm.The principle of convolutional layer training with full articulamentum be it is the same,
Convolutional layer training, to the partial derivative of each weight, is then updated according to gradient descent method using chain type derived function loss function
Weight obtains optimal weight parameter and bias term parameter by back-propagation algorithm, adjusts weight parameter and bias term
After parameter, YOLO v3 detection neural network model precision is made to reach expected precision.Convolutional calculation in convolutional neural networks swashs
Work function calculating, Chi Hua, back-propagation algorithm are the prior art, or quoted from books Ian Goodfellow, Yoshua
Bengio, Aaron Courville. deep learning People's Telecon Publishing House .2017.07.01.Can obtain calculation formula and
Related derivation process.
Residual block relevant calculation is the prior art, or quoted from paper K.He, X.Zhang, S.Ren, and J.Sun.Deep
Residual learning for image recognition.arXiv preprint arXiv:1512.03385,2015.
It can obtain relevant calculation formula and derivation process.
Position detection relevant calculation is the prior art, or quoted from paper Joseph Redmon, Ali
Farhadi.YOLOv3:An Incremental Improvement.arXiv preprint arXiv:1804.02767,
2018.It can obtain relevant calculation formula and derivation process.
Step 4: input test collection, the precision of prediction of testing model, if average error rate 5% or more, again to model
It is trained, until average error rate reaches 5% or less；Here average error rate calculates described in third step as above；
Step 5: carrying out finite element verifying and experimental verification；Simulation analysis is carried out using ANSYS software, will be emulated
To main shaft temperature field picture input to obtain the predicted value of main shaft thermal deformation as detection neural network model, and obtained with emulation
Deformation field data be compared, if the deviation of the two in certain deviation range, i.e., verifying model accuracy meet the requirements,
Otherwise, third step is returned to, ginseng and test verifying are adjusted in the training carried out again；
The experimental data on different lathes is taken to carry out model prediction again, if precision reaches allowed band, method can
Otherwise row returns to second step, ginseng and test verifying are adjusted in the training carried out again.
Embodiment based on above scheme is as follows:
The implementation case is operated on a numerically-controlled machine tool (VMC650m type lathe), is illustrated according to Fig. 1-Fig. 6, will
Acquisition image is dimensioned to 416 × 416, in data scaling, by collected Thermal Error (position sensor acquisition it is axial with
And radial Thermal Error) and fore bearing heat source calibration frame position data be set as label, pretreatment and data are carried out to data
After expansion, data set is made, training set and forecast set are divided into.YOLO v3 detection neural network is set, and by the last layer
Classification layer is changed to full articulamentum, and original Classification Loss function is changed to mean square deviation loss function in total losses function, then defeated
Enter data set to be trained and predict, obtains the position between accurate Thermal Error and predicted position and true calibration frame
Error, as a result as shown in Figure 7 and Figure 8.
The result of Fig. 7 is the comparison diagram of prediction and the actual value output of Spindle thermal error: wherein the curve with ■ represents
The actual measured value of Thermal Error, band ● curve indicate based on detection neural network model model predication value, band ▲ curve
It indicates prediction residual, is the error of actual measured value and model predication value.Fig. 8 result be predicted position and true calibration frame it
Between location error comparison diagram, Fig. 7 shows, prediction residual illustrates the master based on detection neural network prediction within ± 3 microns
Axis thermal deformation is identical with actual measured value.Spindle thermal error can be accurately predicted to demonstrate this method, be main shaft heat
The accurate compensation of error provides theoretical guarantee, improves the overall processing precision of numerically-controlled machine tool.Fig. 8 shows the increasing with frequency of training
Add, the position registration between predicted position and true calibration frame increases namely the two location error is gradually reduced, prediction
Accuracy rate is gradually increased.
The present invention is disclosed as above with preferable case study on implementation, and however, it is not intended to limit the invention, any to be familiar with this profession
Technical staff, without departing from the scope of the present invention, according to the technical essence of the invention to the above case study on implementation institute
Any simple modification, equivalent change and modification done still belong to technical solution of the present invention range.
Claims (6)
1. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network, it is characterised in that:
Step 1: using thermal infrared imager acquisition main shaft of numerical control machine tool heating and cooling image, after converting array for image,
Initial pictures array is subtracted, obtained image array is then converted to image；
Step 2: carrying out calibration frame pretreatment to the image after conversion, to demarcate position data and Thermal Error as label, and make
Make data set, the image data set made is divided into training set and forecast set two parts；
Step 3: selection detection neural network model, modifies total losses function, adjusts network structure, start to be trained, it is average
Error rate reaches 5% hereinafter, deconditioning, preservation model；
Step 4: input test collection, the precision of prediction of testing model, if average error rate is not up to 5% hereinafter, again to model
It is trained, until average error rate reaches 5% or less；
Step 5: carrying out finite element verifying and experimental verification；Simulation analysis is carried out using ANSYS software, emulation is obtained
Main shaft temperature field picture inputs to obtain the predicted value of main shaft thermal deformation, and the change obtained with emulation as detection neural network model
Shape field data is compared, if the deviation of the two in certain deviation range, i.e. verifying model accuracy is met the requirements, no
Then, third step is returned to, ginseng and test verifying are adjusted in the training carried out again；
The experimental data on different lathes is taken to carry out model prediction again, if precision reaches allowed band, method is feasible, no
Then, second step is returned to, ginseng and test verifying are adjusted in the training carried out again.
2. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network according to claim 1,
It is characterized by: third step selection YOLO v3 detects neural network model, the Classification Loss function in total losses function is changed to
L1 loss function, and last classification layer is changed to full articulamentum, sets neuronal quantity as 1, the detection neural network
It is full convolutional neural networks that YOLO v3, which detects main body, finally connects two layers of full articulamentum；
Two types of convolution block and residual block are utilized in YOLO v3 detection neural network model calculates, convolution block is according to volume
Lamination, batch normalization layer and activation primitive layer sequence execute；
Convolutional layer: each pixel of image is numbered first, xd,i+m,j+nIndicate d layer the i-th+m row jth+n column of image
Pixel；ωd,m,nD layer m row the n-th column weight for indicating filter, uses wbIndicate the bias term of filter；To the every of characteristic pattern
A element is numbered, and uses ai,jIndicate the i-th row jth column element of characteristic pattern；D is depth；F is the size of filter, with f table
Show activation primitive, W2It is the width of characteristic pattern after convolution；W1It is the width of image before convolution；F is the width of filter；P is zero padding
The quantity filled, S are stride, H2It is the height of characteristic pattern after convolution, H1It is the width of image before convolution, is calculated using following equation
Convolution；
W2=(W1-F+2P)/S+1
H2=(H1-F+2P)/S+1
Characteristic pattern is obtained in convolutional layer；
Batch normalization layer: the characteristic pattern that convolution obtains is normalized；
Activation primitive layer: LeakyRelu function activates operational formula as follows:
F (x)=max (0, x)+negative_slope × min (0, x)；
Wherein, negative_slope is a small non-zero number；
After calculating, start to be trained model tune ginseng.
3. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network according to claim 2,
It is characterized by: the convolution kernel size 3x3/2 being related to refers to be walked using convolution in YOLO v3 detection neural network model
Stride is increased to 2 by 1 by width, reduces characteristic pattern size, is realized maximum pond, is extracted effective image information.
4. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network according to claim 3,
It is characterized by: batch normalization layer: making characteristic image prime matrix value magnitude range [- 1,1].
5. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network according to claim 4,
It is characterized by: full articulamentum can use gradient descent method in YOLO v3 detection neural network model, by reversely passing
It broadcasts algorithm and obtains optimal weight parameter and bias term parameter.
6. a kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network according to claim 5,
It is characterized by: convolutional layer training utilizes chain type derived function loss function to the partial derivative of each weight, then according to gradient
Descending method updates weight, obtains optimal weight parameter and bias term parameter by back-propagation algorithm, adjusts weight ginseng
After several and bias term parameter, YOLO v3 detection neural network model precision is made to reach expected precision.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910060083.6A CN109739181B (en) | 2019-01-22 | 2019-01-22 | Machine tool spindle thermal error modeling detection method based on detection neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910060083.6A CN109739181B (en) | 2019-01-22 | 2019-01-22 | Machine tool spindle thermal error modeling detection method based on detection neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109739181A true CN109739181A (en) | 2019-05-10 |
| CN109739181B CN109739181B (en) | 2021-03-16 |
Family
ID=66365627
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910060083.6A Active CN109739181B (en) | 2019-01-22 | 2019-01-22 | Machine tool spindle thermal error modeling detection method based on detection neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109739181B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058569A (en) * | 2019-05-19 | 2019-07-26 | 重庆理工大学 | A kind of numerical control machining tool heat error modeling method based on Optimization of Fuzzy neural network |
| CN111240268A (en) * | 2020-01-14 | 2020-06-05 | 重庆大学 | Axle system thermal error modeling method and thermal error compensation system based on SLSTM neural network |
| CN111985149A (en) * | 2020-06-05 | 2020-11-24 | 宁波大学 | Convolutional network-based five-axis machine tool rotating shaft thermal error modeling method |
| CN112200788A (en) * | 2020-10-16 | 2021-01-08 | 清华大学 | High-temperature deformation measuring device and method |
| CN112475904A (en) * | 2020-11-12 | 2021-03-12 | 安徽江机重型数控机床股份有限公司 | Numerical control milling and boring machine machining precision prediction method based on thermal analysis |
| CN114310483A (en) * | 2021-12-13 | 2022-04-12 | 华中科技大学 | Numerical control machining size error prediction method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201209382A (en) * | 2010-08-25 | 2012-03-01 | Nat Univ Chung Cheng | An error compensation apparatus for the built-in motor spindle |
| CN104597842A (en) * | 2015-02-02 | 2015-05-06 | 武汉理工大学 | BP neutral network heavy machine tool thermal error modeling method optimized through genetic algorithm |
| CN108363870A (en) * | 2018-02-11 | 2018-08-03 | 宁波大学 | A kind of main shaft of numerical control machine tool thermal error modeling method based on deep learning |
| CN108415369A (en) * | 2018-05-28 | 2018-08-17 | 河北工业大学 | A kind of main shaft of numerical control machine tool Thermal Error intelligent perception system and cognitive method |
-
2019
- 2019-01-22 CN CN201910060083.6A patent/CN109739181B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201209382A (en) * | 2010-08-25 | 2012-03-01 | Nat Univ Chung Cheng | An error compensation apparatus for the built-in motor spindle |
| CN104597842A (en) * | 2015-02-02 | 2015-05-06 | 武汉理工大学 | BP neutral network heavy machine tool thermal error modeling method optimized through genetic algorithm |
| CN108363870A (en) * | 2018-02-11 | 2018-08-03 | 宁波大学 | A kind of main shaft of numerical control machine tool thermal error modeling method based on deep learning |
| CN108415369A (en) * | 2018-05-28 | 2018-08-17 | 河北工业大学 | A kind of main shaft of numerical control machine tool Thermal Error intelligent perception system and cognitive method |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058569A (en) * | 2019-05-19 | 2019-07-26 | 重庆理工大学 | A kind of numerical control machining tool heat error modeling method based on Optimization of Fuzzy neural network |
| CN110058569B (en) * | 2019-05-19 | 2021-05-11 | 重庆理工大学 | Numerical control machine tool thermal error modeling method based on optimized fuzzy neural network |
| CN111240268A (en) * | 2020-01-14 | 2020-06-05 | 重庆大学 | Axle system thermal error modeling method and thermal error compensation system based on SLSTM neural network |
| CN111985149A (en) * | 2020-06-05 | 2020-11-24 | 宁波大学 | Convolutional network-based five-axis machine tool rotating shaft thermal error modeling method |
| CN112200788A (en) * | 2020-10-16 | 2021-01-08 | 清华大学 | High-temperature deformation measuring device and method |
| CN112475904A (en) * | 2020-11-12 | 2021-03-12 | 安徽江机重型数控机床股份有限公司 | Numerical control milling and boring machine machining precision prediction method based on thermal analysis |
| CN114310483A (en) * | 2021-12-13 | 2022-04-12 | 华中科技大学 | Numerical control machining size error prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109739181B (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109739181A (en) | A kind of main shaft of numerical control machine tool thermal error modeling detection method based on detection neural network | |
| Chen et al. | DenseLightNet: A light-weight vehicle detection network for autonomous driving | |
| CN109948165B (en) | Fine granularity emotion polarity prediction method based on mixed attention network | |
| CN108363870B (en) | Numerical control machine tool spindle thermal error modeling method based on deep learning | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN106855853A (en) | Entity relation extraction system based on deep neural network | |
| CN110825829B (en) | Method for realizing autonomous navigation of robot based on natural language and semantic map | |
| CN110245754A (en) | A kind of knowledge distillating method based on position sensing figure | |
| CN116402821A (en) | Aircraft skin gluing quality defect detection method based on neural network | |
| CN115797808A (en) | Unmanned aerial vehicle inspection defect image identification method, system, device and medium | |
| CN109709895A (en) | A kind of main shaft of numerical control machine tool thermal error modeling training method based on image threshold setting | |
| CN116228754A (en) | Surface defect detection method based on deep learning and global difference information | |
| CN108628956B (en) | The method and system of design innovative | |
| Zhang et al. | Thermal error modeling for heavy duty CNC machine tool based on convolution neural network | |
| Jia et al. | Automated detection of lunar craters using deep learning | |
| CN114170154A (en) | Remote sensing VHR image change detection method based on Transformer | |
| CN110503148A (en) | A kind of point cloud object identifying method with scale invariability | |
| Qin et al. | LSDDN: a lightweight end-to-end network for surface defect detection | |
| Wang et al. | Kalman filter tracking of sequence spot centroid ablated by femtosecond laser | |
| Wang et al. | Artificial neural network-based thermal error modelling in ball screw | |
| Zheng et al. | Robot Target Location Based on the Difference in Monocular Vision Projection | |
| CN106469393A (en) | A kind of Chinese film international communication merit rating method and system | |
| Lee et al. | A Study on Real-Time Defect Detection System Using CNN Algorithm During Scaffold 3D Printing | |
| Enea et al. | Damage detection in composites by AI and high-order modelling surface-strain-displacement analysis | |
| Tang et al. | An assembly system based on industrial robot with binocular stereo vision |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20190510 Assignee: Ningbo Science and Technology Innovation Association Assignor: Ningbo University Contract record no.: X2023980033633 Denomination of invention: A Thermal Error Modeling and Detection Method for Machine Tool Spindles Based on Detection Neural Network Granted publication date: 20210316 License type: Common License Record date: 20230317 |