CN108803486B - Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network - Google Patents
Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network Download PDFInfo
- Publication number
- CN108803486B CN108803486B CN201810936815.9A CN201810936815A CN108803486B CN 108803486 B CN108803486 B CN 108803486B CN 201810936815 A CN201810936815 A CN 201810936815A CN 108803486 B CN108803486 B CN 108803486B
- Authority
- CN
- China
- Prior art keywords
- thermal error
- deep learning
- machine tool
- numerical control
- control machine
- 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
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
- G05B19/404—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form characterised by control arrangements for compensation, e.g. for backlash, overshoot, tool offset, tool wear, temperature, machine construction errors, load, inertia
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Manufacturing & Machinery (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Automatic Control Of Machine Tools (AREA)
- Numerical Control (AREA)
Abstract
The invention discloses a numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network, which comprises the following steps: A. collecting sample data: selecting a heat source measuring point on a numerical control machine tool, detecting a temperature value of the heat source measuring point and a spindle thermal error value corresponding to a time point as sample data; B. establishing a deep learning thermal error prediction model based on a parallel deep belief network; C. training a deep learning thermal error prediction model by using the collected sample data; D. detecting the temperature value of a heat source measuring point of the numerical control machine tool in real time, inputting the trained deep learning thermal error prediction model, and predicting the thermal error value in real time; E. and taking the predicted thermal error value as the compensation translation amount of the origin of the coordinate system of the numerical control machine tool, and realizing real-time thermal error compensation through the deviation of the origin of the coordinate system. The method has the advantages of accurately representing the complex mapping relation between the monitoring temperature signal and the thermal error under the condition of big data, being beneficial to improving the prediction and compensation precision of the thermal error and the like.
Description
Technical Field
The invention relates to the technical field of precision control in the numerical control machine tool industry, in particular to a numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network.
Background
According to statistics, in the precision machining process, machining errors caused by thermal deformation of a process system account for 40% -70% of the total machining errors, wherein the thermal deformation errors of the numerical control machine tool account for a large proportion and even account for more than 50% of the machining errors of the whole workpiece. The reasonable and effective thermal error control is an important guarantee for improving the processing precision of the numerical control machine tool. Error compensation is one of the most commonly used and effective methods. The thermal error compensation is premised on establishing a mapping relation between the thermal error and the temperature of the machine tool as accurately as possible, so that the thermal error is forecasted by the temperature value of the machine tool in the real-time compensation process.
Since the thermal error itself has the comprehensive characteristics of quasi-static time variation, nonlinearity, attenuation delay and coupling, it is difficult to establish an accurate thermal error mathematical model by using theoretical analysis. The currently common thermal error modeling method is an experimental modeling method, namely, performing correlation analysis on thermal error data and a machine tool temperature value according to a statistical theory and performing fitting modeling by using a least square principle. In recent years, shallow neural network theory (BP network, RBF network), gray system theory, and the like have also been applied to thermal error modeling. However, the thermal error mathematical model established by the traditional method has two defects besides the problems of compensation precision and robustness: firstly, a large number of signal processing technologies need to be mastered to extract signal characteristics by combining with abundant engineering practice experience; and secondly, the complex mapping relation between the monitoring signal and the thermal error is difficult to represent under the condition of big data by using a shallow model.
Compared with the traditional method, the deep learning method is driven by data, can automatically extract characteristics (knowledge) from the data, and has remarkable advantages for analyzing unstructured, variable and cross-domain big data.
Disclosure of Invention
Aiming at the defects of the prior art, the technical problems to be solved by the invention are as follows: how to provide a numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network, which can automatically extract deep features of numerical control machine tool temperature data, accurately represent a complex mapping relation between a monitoring temperature signal and a thermal error under the condition of big data and is beneficial to improving the thermal error prediction and compensation precision, instantaneity and adaptability.
In order to solve the technical problems, the invention adopts the following technical scheme:
a numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network is characterized by comprising the following steps:
A. collecting sample data: selecting a heat source measuring point on a numerical control machine tool, detecting a temperature value of the heat source measuring point and a spindle thermal error value corresponding to a time point as sample data;
B. establishing a deep learning thermal error prediction model based on a parallel deep belief network;
C. training a deep learning thermal error prediction model by adopting the sample data acquired in the step A;
D. detecting the temperature value of a heat source measuring point of the numerical control machine tool in real time, inputting the trained deep learning thermal error prediction model, and predicting the thermal error value in real time;
E. and taking the predicted thermal error value as the compensation translation amount of the origin of the coordinate system of the numerical control machine tool, and realizing real-time thermal error compensation through the deviation of the origin of the coordinate system.
Further, the method is characterized in that the normalization processing is carried out on the input temperature sample data to an interval [0, 1], and the de-normalization processing is carried out on the thermal error value output by the network prediction.
Furthermore, the deep learning thermal error prediction model is mainly formed by connecting three deep belief networks with the same network structure in parallel, and is respectively DBN1 for predicting the thermal error value of the main shaft in the X-axis direction, DBN2 for predicting the thermal error value of the main shaft in the Y-axis direction and DBN3 for predicting the thermal error value of the main shaft in the Z-axis direction; the DBN1, DBN2 and DBN3 each include 1 visual input layer, 1 output layer and 3 restricted Boltzmann machine RBM hidden layers, RBM1, RBM2 and RBM3, respectively; the DBN1, the DBN2 and the DBN3 respectively have different weight parameters and share one RBM1 layer.
Further, the number of neurons of the visual input layer is consistent with the number of the heat source measuring points selected in the step A; the number of neurons in the output layer is 1, and the output value is a thermal error value of the main axis in the X-axis direction, the Y-axis direction or the Z-axis direction.
Further, the limiting number of neurons in the RBM hidden layer of the boltzmann machine is determined by the following steps: first, the initial number of neurons is set
In the formula, r is the number of heat source measuring points selected in the step A, then the number is gradually increased by taking the step length as s, the root mean square error is predicted by the deep learning thermal error prediction model, and the number of the nerve cells corresponding to the minimum root mean square error predicted by the deep learning thermal error prediction model is determined as the number of the nerve cells of the limitation Boltzmann machine RBM hidden layer.
Further, a logarithmic divergence unsupervised learning method is adopted, a deep belief network DBN1 in a model is pre-trained to obtain a DBN1 network initial weight, the trained DBN1 parameters are assigned to the DBN2 and the DBN3 correspondingly, and initial weight sharing is achieved
Furthermore, the temperature data at each moment and the thermal error at the corresponding moment form label sample data, and the label sample data is used for respectively fine-tuning and generating the optimal weight of the 3 deep belief networks by adopting a BP algorithm.
Further, in the step a, at least 15 heat source measurement points are selected, and at least 100 groups of temperature values and corresponding heat error values are collected as sample data.
Furthermore, the heat source measuring points are mainly distributed at the main shaft of the numerical control machine tool, the screw rod and nut pairs of the feeding shafts, the machine body, the cooling liquid and the working chamber.
Further, the thermal error values include 3 thermal error values of the main shaft in the X-axis direction, the Y-axis direction and the Z-axis direction, which are respectively ex、ey、ez。
In summary, the invention has the following advantages:
1. the method is based on the parallel deep belief network, and solves the problem that the complex mapping relation between the monitoring temperature signal and the thermal error is difficult to represent under the condition that a shallow model is used in the traditional method. Compared with the traditional thermal error prediction system, the prediction accuracy is greatly improved, and meanwhile, the real-time performance and the adaptability of thermal error compensation are obviously improved.
2. The method is based on the deep learning principle, can automatically extract deep features of the temperature data of the numerical control machine, does not need intervention of human, can get rid of dependence on a large number of signal processing technologies and diagnosis experiences, completes self-adaptive extraction of the features and intelligent prediction and compensation of a thermal error state, and has good robustness and strong real-time performance.
3. The invention improves the deep learning capability of the thermal error prediction network and solves the problem that the traditional method needs to master a large number of signal processing technologies and combines rich engineering practice experience to extract the signal characteristics.
Drawings
Fig. 1 is a flow chart of a numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network.
FIG. 2 is a schematic view of thermal error measurement of a main shaft of a gantry machining center.
FIG. 3 is a structure of a deep learning thermal error prediction model of a numerical control machine.
FIG. 4 is a graph of X-axis thermal error compensation.
FIG. 5 is a Y-axis thermal error compensation curve.
FIG. 6 is a Z-axis thermal error compensation curve.
Detailed Description
The present invention is described in further detail below in connection with thermal error prediction and compensation in a gantry machining center.
In specific implementation, as shown in fig. 1, the following steps are specifically adopted:
firstly, collecting a key point temperature value of a numerical control machine tool and a spindle thermal error value of corresponding time as sample data;
(1) according to the structure, working condition and heat source distribution of the gantry machining center, heat source key points are arranged at or near the larger heating part of the machine tool, and 18 temperature sensors are adopted to detect temperature data. The temperature key points are numbered from T1 to T18, arranged as follows:
1) left and right upright post parts
The screw rod is mounted on a lower bearing seat: t1, T2; a screw nut: t3, T4; guide rail: t5, T6.
2) Cross beam part
Bearing frame about the lead screw: t7, T8; a lead screw nut T9; a guide rail T10.
3) Part of the ram
The upper surface of the main shaft box: t11; the left side of the main shaft box: t12; the right side of the main shaft box: t13; a main shaft flange: t14.
4) Bed and table position
A lathe bed: t15; a workbench: t16.
5) Cooling liquid part: the coolant inlet line is at T17.
6) Ambient temperature: the workplace temperature T18.
(2) Measuring 3 thermal error values of a main shaft of the gantry machining center in the radial direction and the axial direction: e.g. of the typex、ey、ezThe measurement is shown in fig. 2.
(3) The simulation gantry machining center is used for simulating a continuous cycle machining state, main shaft rotation, feed shaft movement and cooling liquid circulation, but actual cutting is not carried out. Data collection is carried out in the afternoon, the gantry machining center is preheated for 0.5 hour before collection, collection is stopped at 1 hour in the noon, the gantry machining center does not stop, data collection is continued in the afternoon, the numerical value of each temperature sensor and the numerical value of the spindle displacement sensor are recorded every 5 minutes, and 150 groups of temperature and thermal error data are collected as sample data.
(4) And normalizing 150 groups of sample data to an interval [0, 1] according to the following formula for training and verifying the deep learning prediction model:
in the formula ai' denotes a value normalized for each sample data, aiFor each sample data original value, amaxAnd aminRespectively representing the minimum value and the maximum value of each type of sample data.
Secondly, establishing a deep learning thermal error prediction model based on a parallel deep belief network;
(1) construction of numerical control machine tool deep learning thermal error prediction model structure
The numerical control machine tool deep learning thermal error prediction model is formed by connecting 3 deep belief networks (DBN1, DBN2 and DBN3) in parallel. The 3 deep belief networks have the same network structure and different weight parameters and share 1 restricted Boltzmann machine, namely RBM1 layer.
DBN1, DBN2, DBN3 predict thermal error values e of the spindle in the radial X, y and Z directions, respectivelyx、ey、ez. The numerical control machine tool deep learning thermal error prediction model is shown in a figure 3.
(2) Deep belief network structure determination
1) Each deep belief network contains 1 visual input layer, 3 restricted boltzmann machine RBM hidden layers, and 1 output layer.
2) The number of neurons in the visual input layer is the same as the number of the set key heat source points, and the number of neurons in the visual input layer is 18. The number of neurons in the output layer is 1, the output of the neurons is a thermal error value of the main shaft in the X-axis direction, the Y-axis direction or the Z-axis direction, and specifically, the output of the DBN1 is a thermal error value of the main shaft in the X-axis direction; the output of the DBN2 is the thermal error value for the spindle in the Y-axis direction; the output of the DBN3 is the thermal error value of the spindle in the Z-axis direction;
3) the number of neurons of each RBM hidden layer has a large influence on the prediction accuracy and generalization capability of the model. The loss of temperature characteristic information can be caused by too few neurons, so that the characteristic extraction is incomplete, and the prediction precision is low. The prediction accuracy is improved by increasing the number of neurons, but the generalization capability of the model is poor due to the excessive number of neurons. The method for determining the neuron number of the RBM hidden layer comprises the following steps:
firstly, according to the number r of heat source key points and the quantity L of collected sample data, the initial number of neurons of each RBM hidden layer is set to be 18 and 150
The meaning of the formula is that an integer part of r/2 is taken, the prediction accuracy of the model is evaluated by taking root mean square error RMSE as an index, and the root mean square error RMSE is defined as:
wherein, ytFor the actual value of the thermal error,
the thermal error predicted value of the model is shown, n is the total number of the thermal errors, and t is the serial number of the thermal errors. In this example, p0At an initial value of 9, the root mean square error was calculated to be 0.102.
② keeping the neuron number of the hidden layer of the RBM2 and RBM3 constant, and gradually increasing the neuron number p of the hidden layer of the RBM1 by the step length s of 21In this embodiment, when p is1When the error is 21, the root mean square error of the model prediction reaches the minimum value of 0.025, and p is calculated1The number of neurons identified as the hidden layer of RBM1 was determined 21.
In the same way, when the RBM2 is determined, the number of the RBM1 neurons is fixed at 21, the number of the RBM3 neurons is fixed at 9, and finally the number of the RBM2 neurons is determined to be 21. When the RBM3 is determined, the neuron number of the RBM1 is fixed at 21, the neuron number of the RBM2 is fixed at 21, and finally the neuron number of the RBM3 is determined to be 19.
Determining the number of neurons in the hidden layers of RBM2 and RBM3 as p2＝21，p3＝19。
Training a deep learning thermal error prediction model by using the collected sample data;
(1) and (3) pre-training 1 of the deep belief networks DBN1 in the model by adopting a logarithmic divergence unsupervised learning method to obtain a network initial weight, wherein the deep belief networks DBN2 and DBN3 share the initial weight. The deep belief network DBN log-divergence unsupervised learning method comprises the following steps:
1) randomly initializing each layer of neurons of the DBN to 0 or 1, connecting weights w of neurons between different layersijSet to any value in the range of (0, 1). The RBM1 is first pre-trained.
2) Using a temperature sample data to generate a visual layer neuron viCalculating hidden layer neuron hjThen connect the weight wijHas a positive gradient of Hij＝vi×hj。
3) By hidden layer neurons hjThe visual layer neuron obtained by reverse calculation is vi' if the inverse gradient of the connection weight is Hij′＝vi′×hj。
4) Updating the weight: w is aij＝wij+ε×(Hij-Hij') where ε is the learning rate, typically set to 0<ε<1。
5) The 18 temperature data collected at each time are one sample data, and 150 sample data are collected at 150 collection times. And circularly training the RBM1 network by using 150 collected temperature sample data, and continuously iterating until convergence, namely Hij-HijDelta is less than or equal to delta, and delta is a convergence threshold value which is less than 0.01.
In this embodiment, when the RBM1 is pre-trained, the learning rate ∈ is set to 0.1, the iteration convergence threshold is set to 0.005, and pre-training is completed after 990 iterations. In the same way, pre-training of RBM2 and RBM3 was completed after 860 and 1100 iterations, respectively. And correspondingly assigning the trained DBN1 parameters to the DBN2 and the DBN3 to realize initial weight sharing.
(2) And the temperature data at each moment and the thermal errors at the corresponding moment form label sample data. And respectively fine-tuning and generating the optimal weight of the 3 deep belief networks by using label sample data and adopting a BP algorithm.
Reversely correcting the initial weight obtained by pre-training by using the label data through a BP method; then, forward testing newly obtained weight parameters of each layer; this is repeated until the root mean square error of the model prediction converges.
The root mean square error E convergence discriminant is:
where t is the training number and λ is a minimum threshold, typically set to less than 0.01.
In this embodiment, the convergence threshold λ of the root mean square error E is set to 0.005, the fine-tuning learning rate epsilon is set to 0.1, and the 3 deep belief networks respectively complete weight fine tuning after 35, 46, and 39 iterations, and at this time, the root mean square errors predicted by the 3 deep belief networks are respectively 0.01, 0.015, and 0.012.
After the pre-training and the tuning, the construction of the prediction model is completed, and all model parameters are successfully obtained. A series of temperature data detected in real time may then be input to a prediction model for predicting a corresponding spindle thermal error value.
And fourthly, detecting the key point temperature value of the numerical control machine tool in real time, and inputting the key point temperature value into a deep learning prediction network to predict a thermal error value.
And measuring 30 groups of key heat source point temperature data of the gantry machining center in real time, inputting the data into the deep learning prediction network to predict the spindle thermal error, and obtaining a very high-precision thermal error prediction result as shown in figures 4-6. The thermal error root mean square error of the model prediction is 0.018, the performance indexes of the real-time data prediction and the training sample data prediction are basically consistent, and the deep learning thermal error prediction model has very good generalization capability.
And fifthly, taking the predicted thermal error value as the compensation translation amount of the origin of the coordinate system of the machine tool, and realizing real-time thermal error compensation through the offset of the origin of the coordinate system.
The thermal errors of the main shaft of the gantry machining center in X, Y, Z3 directions are respectively ex、ey、ezUsing the compensation translation amount as the compensation translation amount of the machine tool coordinate system origin, the machine tool coordinate system origin is respectively translated in X, Y, Z shaft 3 directions by-ex、-ey、-ezNamely, the thermal error of the machine tool can be accurately compensated, the maximum residual errors in the X, Y, Z directions after compensation are respectively 2.5 μm, 3 μm and 2.5 μm, and the residual errors after compensation are shown in fig. 4 to 6.
The above description is only exemplary of the present invention and should not be taken as limiting, and any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A numerical control machine tool thermal error prediction and compensation method based on a parallel deep learning network is characterized by comprising the following steps:
A. collecting sample data: selecting a heat source measuring point on a numerical control machine tool, detecting a temperature value of the heat source measuring point and a spindle thermal error value corresponding to a time point as sample data;
B. establishing a deep learning thermal error prediction model based on a parallel deep belief network;
C. training a deep learning thermal error prediction model by adopting the sample data acquired in the step A;
D. detecting the temperature value of a heat source measuring point of the numerical control machine tool in real time, inputting the trained deep learning thermal error prediction model, and predicting the thermal error value in real time;
E. the predicted thermal error value is used as the compensation translation amount of the origin of the coordinate system of the numerical control machine tool, and the real-time compensation of the thermal error is realized through the deviation of the origin of the coordinate system;
the deep learning thermal error prediction model is mainly formed by connecting three deep belief networks with the same network structure in parallel, and is respectively DBN1 for predicting the thermal error value of the main shaft in the X-axis direction, DBN2 for predicting the thermal error value of the main shaft in the Y-axis direction and DBN3 for predicting the thermal error value of the main shaft in the Z-axis direction; the DBN1, DBN2 and DBN3 each include 1 visual input layer, 1 output layer and 3 restricted Boltzmann machine RBM hidden layers, RBM1, RBM2 and RBM3, respectively; the DBN1, the DBN2 and the DBN3 respectively have different weight parameters and share one RBM1 layer.
2. The method as claimed in claim 1, wherein in the step a, the inputted temperature sample data is normalized to the interval [0, 1], and in the steps C and D, the outputted thermal error value is denormalized.
3. The method for predicting and compensating the thermal error of the numerical control machine tool based on the parallel deep learning network according to claim 1, wherein the number of the neurons of the visual input layer is consistent with the number of the heat source measuring points selected in the step A; the number of neurons in the output layer is 1, and the output value is a thermal error value of the main axis in the X-axis direction, the Y-axis direction or the Z-axis direction.
4. The parallel deep learning network-based numerical control machine tool thermal error prediction and compensation method according to claim 3, wherein the limiting number of neurons in the RBM hidden layer of the Boltzmann machine is determined by the following steps: first, the initial number of neurons is set
In the formula, r is the number of heat source measuring points selected in the step A, then the heat source measuring points are gradually increased by taking the step length as s, meanwhile, the root mean square error is predicted by the deep learning heat error prediction model, and the deep learning heat is usedAnd determining the number of the neurons corresponding to the minimum root mean square error predicted by the error prediction model as the number of the neurons of the RBM hidden layer of the limit Boltzmann machine.
5. The method for predicting and compensating the thermal error of the numerical control machine tool based on the parallel deep learning network as claimed in claim 4, wherein a logarithmic divergence unsupervised learning method is adopted, a deep belief network DBN1 in a model is pre-trained to obtain an initial weight of a DBN1 network, and a trained DBN1 parameter is assigned to the DBN2 and the DBN3 correspondingly to realize initial weight sharing.
6. The method for predicting and compensating the thermal error of the numerical control machine tool based on the parallel deep learning network as claimed in claim 5, wherein the temperature data at each moment and the thermal error at the corresponding moment form label sample data, and the label sample data is used to generate the optimal weight of the 3 deep belief networks by respectively fine tuning through a BP algorithm.
7. The method as claimed in claim 1, wherein in the step a, at least 15 heat source measurement points are selected, and at least 100 sets of temperature values and corresponding thermal error values are collected as sample data.
8. The method for predicting and compensating the thermal error of the numerical control machine tool based on the parallel deep learning network as claimed in claim 1, wherein the heat source measuring points are mainly distributed at a main shaft of the numerical control machine tool, each feed shaft screw nut pair, the machine body, the cooling liquid and the working chamber.
9. The method as claimed in claim 1, wherein the thermal error values include 3 thermal error values of the main axis in the X-axis direction, the Y-axis direction and the Z-axis direction, which are respectively ex、ey、ez。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810936815.9A CN108803486B (en) | 2018-08-16 | 2018-08-16 | Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810936815.9A CN108803486B (en) | 2018-08-16 | 2018-08-16 | Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108803486A CN108803486A (en) | 2018-11-13 |
| CN108803486B true CN108803486B (en) | 2021-02-02 |
Family
ID=64080245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810936815.9A Active CN108803486B (en) | 2018-08-16 | 2018-08-16 | Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108803486B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111273607B (en) * | 2018-12-04 | 2022-12-13 | 沈阳高精数控智能技术股份有限公司 | Spark-based numerical control machine tool running state monitoring method |
| CN109800537B (en) * | 2019-02-20 | 2022-11-18 | 大连理工大学 | Machine tool thermal error model reliability calculation method based on deep neural network and Monte Carlo method |
| CN109765846B (en) * | 2019-03-11 | 2022-03-18 | 哈尔滨理工大学 | Method for selecting thermal error modeling temperature sensitive points of double-turntable five-axis numerical control machine tool |
| 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 |
| CN110083125B (en) * | 2019-05-19 | 2021-05-07 | 重庆理工大学 | Machine tool thermal error modeling method based on deep learning |
| CN110568817A (en) * | 2019-09-13 | 2019-12-13 | 深圳市烨嘉为技术有限公司 | machine tool motion temperature difference compensation method based on big data analysis and prejudgment |
| CN111240268B (en) * | 2020-01-14 | 2021-03-16 | 重庆大学 | Axle system thermal error modeling method and thermal error compensation system based on SLSTM neural network |
| CN114609967A (en) * | 2020-12-04 | 2022-06-10 | 迈鑫机械工业股份有限公司 | Real-time space precision compensation intelligent module of numerical control machine |
| CN113021078B (en) * | 2021-03-01 | 2022-10-14 | 珠海艾诚精密模具有限公司 | Method for detecting dynamic elongation of NC spindle |
| CN113341878B (en) * | 2021-06-23 | 2023-04-18 | 重庆理工大学 | Thermal error measuring method of five-axis numerical control machine tool |
| CN114265365B (en) * | 2021-12-21 | 2024-05-28 | 重庆大学 | Online measurement-based dynamic modeling and compensation method for thermal error of gear grinding machine |
| CN116224905B (en) * | 2023-03-02 | 2023-10-20 | 广东工业大学 | Machine tool thermal error prediction method and system based on joint distribution self-adaption |
| CN117226599B (en) * | 2023-11-10 | 2024-01-30 | 上海诺倬力机电科技有限公司 | Numerical control machine tool thermal error prediction method, device, equipment and medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104764527A (en) * | 2015-03-31 | 2015-07-08 | 上海大学 | Thermal error measuring device for workpiece machining |
| CN104898562A (en) * | 2015-04-27 | 2015-09-09 | 攀枝花学院 | Modeling method of thermal error compensation of numerically-controlled machine tool |
| CN105607575A (en) * | 2016-01-27 | 2016-05-25 | 电子科技大学 | FA-LSSVM based numerical control machine tool main shaft heat drift modeling method |
| CN106372337A (en) * | 2016-09-05 | 2017-02-01 | 华中科技大学 | Thermal deformation prediction method of preheating stage of numerical control machine tool |
| 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 |
| JP6679086B2 (en) * | 2015-11-11 | 2020-04-15 | 国立研究開発法人情報通信研究機構 | Learning device, prediction device, learning method, prediction method, and program |
-
2018
- 2018-08-16 CN CN201810936815.9A patent/CN108803486B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104764527A (en) * | 2015-03-31 | 2015-07-08 | 上海大学 | Thermal error measuring device for workpiece machining |
| CN104898562A (en) * | 2015-04-27 | 2015-09-09 | 攀枝花学院 | Modeling method of thermal error compensation of numerically-controlled machine tool |
| JP6679086B2 (en) * | 2015-11-11 | 2020-04-15 | 国立研究開発法人情報通信研究機構 | Learning device, prediction device, learning method, prediction method, and program |
| CN105607575A (en) * | 2016-01-27 | 2016-05-25 | 电子科技大学 | FA-LSSVM based numerical control machine tool main shaft heat drift modeling method |
| CN106372337A (en) * | 2016-09-05 | 2017-02-01 | 华中科技大学 | Thermal deformation prediction method of preheating stage of numerical control machine tool |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108803486A (en) | 2018-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108803486B (en) | Numerical control machine tool thermal error prediction and compensation method based on parallel deep learning network | |
| CN111177911B (en) | Part surface roughness online prediction method based on SDAE-DBN algorithm | |
| CN111079342B (en) | TBM tunneling performance prediction method based on online rock mass grade classification | |
| WO2020041956A1 (en) | Bayes- and fault tree-based reliability evaluation method for computer numerical control machine tool | |
| CN107492098B (en) | It is a kind of based on PCA and CNN high-temperature forging surface defect in position detecting method | |
| WO2020168584A1 (en) | Method for calculating, based on deep neural network and monte carlo method, reliability of machine tool thermal error model | |
| CN113051831A (en) | Machine tool thermal error self-learning prediction model modeling method and machine tool thermal error control method based on digital twins | |
| CN110561191B (en) | Numerical control machine tool cutter abrasion data processing method based on PCA and self-encoder | |
| CN114619292B (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
| CN111881574B (en) | Wind turbine generator set key component reliability modeling method based on distribution function optimization | |
| CN113188794B (en) | Gearbox fault diagnosis method and device based on improved PSO-BP neural network | |
| CN114881289B (en) | Digital twinning-based fusion-casting charge forming process parameter optimization method and system | |
| CN102889988B (en) | Precision prediction method of ball screw pair | |
| CN111126255A (en) | Numerical control machine tool cutter wear value prediction method based on deep learning regression algorithm | |
| CN116578832B (en) | Device for predicting surface roughness of machined part | |
| CN116757057A (en) | Air quality prediction method based on PSO-GA-LSTM model | |
| CN108090515A (en) | A kind of environmental rating appraisal procedure based on data fusion | |
| CN116088453A (en) | Production quality prediction model training method and device and production quality monitoring method | |
| CN116975645A (en) | Industrial process soft measurement modeling method based on VAE-MRCNN | |
| CN117709203A (en) | Comprehensive error modeling method and system for hydrostatic lead screw feeding system | |
| CN110705841A (en) | Chemical production safety assessment method based on improved fuzzy analytic hierarchy process | |
| Ren et al. | Vibration monitoring with dependencies attention for gear hobbing quality evaluation | |
| CN114547963A (en) | Tire modeling method and medium based on data driving | |
| CN103983332A (en) | Method for error compensation of sensor based on HGSA-BP algorithm | |
| CN116540632A (en) | CNN-GRU (computer numerical control-grid-GRU) combined neural network-based machine tool feeding system thermal error prediction method |
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 |