CN111103846A - Numerical control machine tool state prediction method based on time sequence - Google Patents
Numerical control machine tool state prediction method based on time sequence Download PDFInfo
- Publication number
- CN111103846A CN111103846A CN201911369633.9A CN201911369633A CN111103846A CN 111103846 A CN111103846 A CN 111103846A CN 201911369633 A CN201911369633 A CN 201911369633A CN 111103846 A CN111103846 A CN 111103846A
- Authority
- CN
- China
- Prior art keywords
- numerical control
- control machine
- working condition
- network
- machine tool
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 38
- 230000001133 acceleration Effects 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000005070 sampling Methods 0.000 claims description 7
- 238000012546 transfer Methods 0.000 claims description 5
- 238000013519 translation Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 4
- 238000011161 development Methods 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 6
- 238000012544 monitoring process Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000007621 cluster analysis Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
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
-
- 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
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/31—From computer integrated manufacturing till monitoring
- G05B2219/31434—Zone supervisor, collects error signals from, and diagnoses different zone
Abstract
The invention relates to a numerical control machine state prediction method based on a time sequence, which is characterized in that working condition parameters of a numerical control machine are obtained from a numerical control machine controller and a joint driving motor, wherein the working condition parameters comprise the position, the position error, the speed, the acceleration, the torque, the current, the temperature and the operation time of each shaft; performing similarity clustering on the working condition parameters by using a same-class comparison method, and extracting the working condition parameters in the same cluster as training data; establishing a time sequence-based NARMA-L2 model of each axis position error and working condition parameter based on an improved wavelet neural network, and training the NARMA-L2 model; and collecting working condition data in real time, inputting the data into a trained NARMA-L2 model, and predicting and judging the position error of each axis of the machine tool and the working condition parameters. The accurate prediction and identification of the motion errors of all joints of the numerical control machine can be realized, and the predictive maintenance and the cluster operation and maintenance of the numerical control machine under the time-varying working condition are realized.
Description
Technical Field
The invention belongs to the technical field of operation monitoring of numerical control machines, and particularly relates to a numerical control machine state prediction method based on a time sequence.
Background
Under the background that the wave of global industrial internet rises day by day, the numerical control machine tool is applied more and more widely in the production of flexible processing and the like. However, the numerical control machine tool lacks real-time monitoring in the working process, and is difficult to control the working state and the service life.
The operation condition of the numerical control machine tool is complex, and the health condition of the numerical control machine tool is difficult to evaluate. The existing numerical control machine tool monitoring method mostly needs to be additionally provided with an external sensor, so that the internal control of the numerical control machine tool is influenced, and a large amount of cost waste is caused. On the other hand, in the process of extracting data from the sensor, the traditional waveform signal feature extraction method needs to perform high-frequency sampling on the sensing signal, and has high requirements on signal processing and transmission technologies/equipment of the acquisition terminal, which invisibly increases the monitoring cost.
Disclosure of Invention
In view of the above, the present invention is intended to disclose a method for predicting the state of a numerically-controlled machine tool based on a time sequence, so as to realize accurate prediction and identification of the motion error of each joint of the numerically-controlled machine tool, and realize predictive maintenance and cluster operation and maintenance of the numerically-controlled machine tool under a time-varying working condition.
In order to solve the technical problems, the invention adopts the following technical scheme:
a numerical control machine state prediction method based on time series is characterized by comprising the following steps:
s1; obtaining working condition parameters of the numerical control machine tool from a numerical control machine tool controller and a joint driving motor, wherein the working condition parameters comprise the position, the position error, the speed, the acceleration, the torque, the current, the temperature and the operation time of each shaft;
s2: performing similarity clustering on the working condition parameters by using a same-class comparison method, and extracting the working condition parameters in the same cluster as training data;
s3: establishing a time sequence-based NARMA-L2 model of each axis position error and working condition parameter based on an improved wavelet neural network, and training the NARMA-L2 model;
s4: and collecting working condition data in real time, inputting the data into a trained NARMA-L2 model, and predicting and judging the position error of each axis of the machine tool and the working condition parameters.
In the above technical solution, step S1 acquires the working condition parameters from the numerical control machine tool in a TCP socket communication manner, the sampling frequency is not more than 1Hz, and the working condition parameters are stored as excel documents in the following format, and each excel document becomes a set of data.
In the above technical solution, step S2 uses a K-means dynamic clustering algorithm to cluster different operating condition parameters of each joint of the numerically-controlled machine tool to obtain operating condition parameters in the same cluster.
In the above technical solution, for the operating condition state where the position error of each axis needs to be predicted, in step S2, the K-means dynamic clustering algorithm is first used to calculate the comprehensive distance parameter of each operating condition parameter in the operating condition state, and then the distance between the operating condition parameter and the clustering center of each group of operating condition parameters is calculated, and the set group data is selected from the closest cluster to train the WNN-NARMAL2 model.
In the above technical solution, in step S3, the numerical control machine tool axis position error and working condition parameter NARMA-L2 model based on the time series is established as follows:
wherein k represents the time scale of the system; n represents the backtracking step number of the time scale, and n is more than or equal to 1;
y (k), y (k +1), y (k-1), …, y (k-n +1) represents system output, specifically, each axis position error of the numerical control machine tool under each time scale;
u (k), u (k-1), u (k-2), … u (k-n +1) represent system inputs, i.e., operating parameters on each time scale.
In the technical scheme, each axis position error and working condition parameter NARMA-L2 model of the numerical control machine tool based on the time sequence in the step S3 comprises an input layer, a wavelet hidden layer and an output layer; each node of the wavelet hidden layer is expanded according to time scale and contains an expansion coefficient ah1,ah2And a translation coefficient bh1,bh2Each wavelet transfer function of (a); the output of the network is a function y (k + 1); the model is divided into an upper part and a lower part according to the time scale development sequence, wherein the upper part is a pair function g [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), … and u (k-n +1)]u (k), the lower half being the function f [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), …, u (k-n +1)]Is calculated.
In the above technical solution, the training process of the NARMA-L2 model in step S3 is specifically as follows:
s31: determining each of the networksNumber of nodes of layer, connection weight coefficient W of initialized each nodeh1i,Wh2i,Wjh1,Wh2And the expansion coefficient a of the wavelet transfer functionh1,ah2And a translation coefficient bh1,bh2Are all [0,1]A random value within a range;
s32: inputting training data [ y (k), y (k-1), …, y (k-n +1), u (k-1), …, u (k-n +1)]And corresponding desired target output yT(k +1), wherein k ═ 1,2, ·, N denotes the time scale of the training data;
s33: output of the computing network:
s331: firstly, calculating the output of the wavelet hidden layer, and for the wavelet hidden layer output of the upper half part of the network:
subscript h in formula (2)1Representing h in the wavelet hidden layer of the upper half of the network1A node; for wavelet hidden layer output in the lower half of the network:
in the formulae (2) and (3), layerout _ uh1kAnd layerout _ lh2kRespectively representing wavelet hidden layer outputs of the upper half part and the lower half part; Ψ (x) is the Morlet mother wave represented as follows:
ψ(x)＝cos(1.75x)exp(-x2/2) (4)
s332: then, the output of the network output layer is calculated as follows for the upper half and the lower half, respectively:
in the formulae (5) and (6), outout _ ujkAnd outout _ lkOutput layer outputs representing the upper and lower halves respectively; (x) is a sigmoidal logarithmic function represented as:
f(x)＝1/(1+e-x) (7)
in the above equation, subscript j denotes the j-th node of the output layer of the upper half of the network; h1And H2The total number of nodes of the wavelet hidden layers of the upper half part and the lower half part is respectively;
s333: finally, the output of the whole network is calculated:
y(k+1)＝outout_uk·u(k)+outout_lk(8)
s34: calculating the fitness function of the whole network by taking the mean square error as an index:
in the formula (9), MSE represents a mean square error of the network; n represents the time scale of the training data;
adjusting the weight coefficient a along the direction of the negative gradient by adopting a back propagation BP methodh1，ah2,bh1，bh2,Wh1i,Wh2i,Wjh1And W andh2the back propagation BP method is performed in calculating a fitness function of the entire network. S35: if the fitness function error meets the precision control requirement, ending the training; otherwise, returning to step S33 to recalculate the output of the network until the fitness function error of step S35 meets the requirement of accuracy control.
In the technical scheme, when 50 groups of data are selected from the nearest clusters to train the WNN-NARMAL2 model, the precision of the working condition parameter of the corner error of training and prediction is between-0.1 and 0.06 um.
Compared with the prior art, the invention has the following beneficial effects:
if an external sensor is additionally arranged, the internal control of the numerical control machine tool is influenced, and a large amount of cost waste is caused, the numerical control machine tool controller and the joint driving motor are selected as key parts, working condition parameters can be obtained through the numerical control machine tool controller and the joint driving motor, working condition signals such as joint position, speed, acceleration, torque, current, temperature and the like are obtained from the numerical control machine tool controller and the joint driving motor through a TCP socket communication method, the sampling frequency is low (less than or equal to 1Hz), the external sensor is not required to be additionally arranged, the predictive maintenance and the cluster operation and maintenance management of the numerical control machine tool under the time-varying working condition are realized, the sampling is convenient, and the cost is low.
The method adopts a K-means dynamic clustering algorithm to perform clustering analysis on working condition parameters (speed, acceleration, torque, current, temperature and the like of each shaft) when each shaft executes actions, and selects 50 groups of training data from the closest clusters as training data for state prediction modeling of the numerical control machine, and the training data has low extraction cost, high speed, simplicity and feasibility.
The hidden neural network model based on the time sequence is established based on the wavelet, the model has high approximation precision, the training array has small scale, the weight with precision requirement can be quickly obtained, the network adaptability can be adjusted, and the hidden neural network model can be used for online real-time monitoring.
By combining with the actual working condition of the numerical control machine, a novel model based on Nonlinear Auto-Regressive Moving Average (NARMA-L2) and a wavelet neural network is adopted, and a time sequence prediction method is based, so that the prediction and identification of the states and errors of all joints of the numerical control machine are realized, and the defect of experience monitoring is overcome.
Drawings
FIG. 1 is a flow chart of a numerical control machine state prediction method based on time series according to the method of the present invention.
FIG. 2 is a NARMA-L2 model structure diagram of position errors and working condition parameters of each axis of the numerically-controlled machine tool based on time series.
FIG. 3 is a parameter cluster analysis diagram of the operation condition of the numerical control machine tool.
Fig. 4 is a timing diagram of the X-axis training/predicted position error and the actual position error of the cnc machine according to the present invention.
Fig. 5 is a timing diagram of the difference between the X-axis training/predicted position error and the actual position error of the cnc machine according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to fig. 1-4 and the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The numerical control machine state prediction method based on time series of the invention comprises four steps, as shown in figure 1,
firstly, selecting a numerical control machine controller and a joint driving motor as key components to acquire working condition parameters of the numerical control machine;
then, performing clustering analysis on working condition data and extracting training data;
then according to the collected working condition parameters, establishing a time-series-based NARMA-L2 model of position errors and working condition parameters of each axis of the numerical control machine tool, training the model and stabilizing a network structure;
and finally, actually acquiring working condition parameters for state prediction based on the time sequence.
The numerical control machine state prediction method based on the time sequence is realized by the following specific steps:
1) working condition parameter acquisition of numerical control machine tool
Working condition parameter signals such as the position, the position error, the speed, the acceleration, the torque, the current, the temperature and the operation time of each shaft are obtained from a numerical control machine tool controller and a joint driving motor in a TCP socket communication mode, the sampling frequency is less than or equal to 1Hz, the working condition parameter signals are stored as excel documents in the following format, and each excel document becomes a group of data.
As shown in table 1 below:
TABLE 1 sampling condition data sheet
2) State prediction method based on time series
Numerically controlled machines, as complex electromechanical equipment, are typically nonlinear systems. The NARMA-L2 model (nonlinear autoregressive Moving Average model) has high approximation accuracy and high convergence speed, and is widely applied to nonlinear system identification. The method comprises the following steps of establishing a time-series-based NARMA-L2 model of position errors and working condition parameters of each axis of the numerical control machine tool as follows:
wherein k represents the time scale of the system, y (k), y (k +1), y (k-1), …, y (k-n +1) represents the system output and is the position error of each axis of the numerical control machine tool under each time scale; u (k), u (k-1), u (k-2), … u (k-n +1) represent system inputs, i.e., operating parameters (e.g., joint velocity, acceleration, torque, current, temperature) on various time scales.
According to the basic structure of the wavelet neural network, combined with a NARMA-L2 model, the method provides an improved wavelet neural network for state prediction based on time series, and the structure of the improved wavelet neural network is shown in FIG. 2.
As shown in FIG. 2, the NARMA-L2 model for establishing the position error and working condition parameters of each axis of the numerical control machine tool based on the time sequence is divided into three layers, namely an input layer, a hidden layer (wavelet layer) and an output layer. The upper half of the model is an approximation to the function g [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), …, u (k-n +1) ] u (k), and the lower half is an approximation to the function f [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), …, u (k-n +1) ].
The training process of the NARMA-L2 model of the position errors and the working condition parameters of each axis of the numerical control machine tool based on the time sequence is specifically as follows:
(1) determining the number of nodes in each layer of the network at 0,1]Initializing connection weight coefficient W of each node in rangeh1i,Wh2i,Wjh1,Wh2And the expansion coefficient a of the wavelet transfer functionh1,ah2And a translation coefficient bh1,bh2Is a random value;
(2) inputting training data [ y (k), y (k-1), …, y (k-n +1), u (k-1), …, u (k-n +1)]And corresponding desired target output yT(k +1), wherein k ═ 1,2, ·, N denotes the time scale of the training data, and N denotes the time scale end point of the training data;
(3) the output of the computing network is as follows:
i) first, the output of the wavelet hidden layer is calculated. For wavelet hidden layer output in the upper half of the network:
subscript h in formula (2)1Representing h in the wavelet hidden layer of the upper half of the network1And (4) each node. For wavelet hidden layer output in the lower half of the network:
in the formulae (2) and (3), layerout _ uh1kAnd layerout _ lh2kRespectively representing the wavelet hidden layer output of the upper half part and the lower half part. Ψ (x) is the Morlet mother wave represented as follows:
ψ(x)＝cos(1.75x)exp(-x2/2) (4)
ii) subsequently, the output of the output layer is calculated. The following are shown for the upper and lower halves respectively:
in the formulae (5) and (6), outout _ ujkAnd outout _ lkRespectively representing the output of the output layers of the upper and lower half. (x) is a sigmoidal logarithmic function represented as:
f(x)＝1/(1+e-x) (7)
in the above equation, the subscript j denotes the j-th nodes of the output layer in the upper half of the network. H1And H2The total number of nodes of the wavelet hidden layers of the upper half part and the lower half part is respectively.
Iii) finally, the output of the whole network is calculated as follows:
y(k+1)＝outout_uk·u(k)+outout_lk(8)
(4) calculating the fitness function of the whole network by taking the mean square error as an index:
in equation (9), MSE represents the mean square error of the network. In order to reduce network error and improve function approximation capability, a weight coefficient a is requiredh1，ah2,bh1，bh2,Wh1i,Wh2i,Wjh1And W andh2and carrying out dynamic adjustment. The specific adjusting method adopts a Back-Propagation (BP) method to adjust the weight along the direction of the negative gradient.
(5) And if the fitness function error meets the requirement, ending the training. Otherwise, returning to the step (3).
3) Operating condition data cluster analysis and training data extraction
And (3) performing cluster analysis on different operating condition parameters (joint speed, acceleration, torque, current and temperature) of each joint of the numerical control machine tool by adopting a K-means dynamic cluster algorithm. The method collects 500 groups of working condition parameter data under different operating conditions to obtain 6 groups of clustering centers as shown in figure 3.
For the operating condition that the position error of each shaft needs to be predicted, firstly, a K-means dynamic clustering algorithm is adopted to calculate the comprehensive distance parameter of the operating condition, further, the distance between the operating condition and the 6 groups of clustering centers is calculated, and 50 groups of data are selected from the clusters closest to each other to train a WNN-NARMAL2 model. The training and validation effect is shown in fig. 4 and 5. Where fig. 4 is a timing chart of the X-axis training/predicted position error and the actual position error, and fig. 5 is a timing chart of the difference between the X-axis training/predicted position error and the actual position error.
As shown in figures 4 and 5, after the WNN-NARMAL2 neural network is trained in a time sequence of 100s by using 50 groups of data, the model can be used for predicting the position error of each axis of the numerical control machine, and the training and predicting results are between-0.1 and 0.06um, so that higher accuracy is achieved.
It should be noted that, according to the implementation requirement, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can be combined into new steps/components to achieve the purpose of the present invention.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (8)
1. A numerical control machine state prediction method based on time series is characterized by comprising the following steps:
s1; obtaining working condition parameters of the numerical control machine tool from a numerical control machine tool controller and a joint driving motor, wherein the working condition parameters comprise the position, the position error, the speed, the acceleration, the torque, the current, the temperature and the operation time of each shaft;
s2: performing similarity clustering on the working condition parameters by using a same-class comparison method, and extracting the working condition parameters in the same cluster as training data;
s3: establishing a time sequence-based NARMA-L2 model of each axis position error and working condition parameter based on an improved wavelet neural network, and training the NARMA-L2 model;
s4: and collecting working condition data in real time, inputting the data into a trained NARMA-L2 model, and predicting and judging the position error of each axis of the machine tool and the working condition parameters.
2. The numerical control machine tool state prediction method based on the time series as claimed in claim 1, characterized in that step S1 collects the working condition parameters from the numerical control machine tool in a TCP socket communication manner, the sampling frequency is less than or equal to 1Hz, and the working condition parameters are stored as excel documents in the following format, and each excel document becomes a set of data.
3. The numerical control machine state prediction method based on the time series as claimed in claim 1, characterized in that step S2 clusters different operation condition parameters of each joint of the numerical control machine by using a K-means dynamic clustering algorithm to obtain the condition parameters in the same cluster.
4. The method of claim 1, wherein in step S2, for the operating condition status requiring the prediction of the position error of each shaft, the K-means dynamic clustering algorithm is used to calculate the comprehensive distance parameter of each condition parameter under the operating condition status, and further calculate the distance to the cluster center of each set of condition parameters, and select the set of data from the closest cluster for training the WNN-NARMAL2 model.
5. The method of claim 1, wherein the time-series based numerical control machine tool state prediction method is characterized in that the time-series based numerical control machine tool axis position error and condition parameter NARMA-L2 model in step S3 is established as follows:
wherein k represents the time scale of the system; n represents the backtracking step number of the time scale, and n is more than or equal to 1;
y (k), y (k +1), y (k-1), …, y (k-n +1) represents system output, specifically, each axis position error of the numerical control machine tool under each time scale;
u (k), u (k-1), u (k-2), … u (k-n +1) represent system inputs, i.e., operating parameters on each time scale.
6. The numerical control machine tool state prediction method based on time series according to claim 5, characterized in that in step S3, the prediction method is based on time seriesThe NARMA-L2 model of each axis position error and working condition parameter of the numerical control machine tool comprises an input layer, a wavelet hidden layer and an output layer; each node of the wavelet hidden layer is expanded according to time scale and contains an expansion coefficient ah1,ah2And a translation coefficient bh1,bh2Each wavelet transfer function of (a); the output of the network is a function y (k + 1); the model is divided into an upper part and a lower part according to the time scale development sequence, wherein the upper part is a pair function g [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), … and u (k-n +1)]u (k), the lower half being the function f [ y (k), y (k-1), …, y (k-n +1), u (k-1), u (k-2), …, u (k-n +1)]Is calculated.
7. The method of claim 5, wherein the training of the NARMA-L2 model in step S3 is as follows:
s31: determining the number of nodes in each layer of the network, and initializing the connection weight coefficient W of each nodeh1i,Wh2i,Wjh1,Wh2And the expansion coefficient a of the wavelet transfer functionh1,ah2And a translation coefficient bh1,bh2Are all [0,1]A random value within a range;
s32: inputting training data [ y (k), y (k-1), …, y (k-n +1), u (k-1), …, u (k-n +1)]And corresponding desired target output yT(k +1), wherein k ═ 1,2, ·, N denotes the time scale of the training data;
s33: output of the computing network:
s331: firstly, calculating the output of the wavelet hidden layer, and for the wavelet hidden layer output of the upper half part of the network:
subscript h in formula (2)1Representing h in the wavelet hidden layer of the upper half of the network1A node; for wavelet hidden layer output in the lower half of the network:
in the formulae (2) and (3), layerout _ uh1kAnd layerout _ lh2kRespectively representing wavelet hidden layer outputs of the upper half part and the lower half part; Ψ (x) is the Morlet mother wave represented as follows:
ψ(x)＝cos(1.75x)exp(-x2/2) (4)
s332: then, the output of the network output layer is calculated as follows for the upper half and the lower half, respectively:
in the formulae (5) and (6), outout _ ujkAnd outout _ lkOutput layer outputs representing the upper and lower halves respectively; (x) is a sigmoidal logarithmic function represented as:
f(x)＝1/(1+e-x) (7)
in the above equation, subscript j denotes the j-th node of the output layer of the upper half of the network; h1And H2The total number of nodes of the wavelet hidden layers of the upper half part and the lower half part is respectively;
s333: finally, the output of the whole network is calculated:
y(k+1)＝outout_uk·u(k)+outout_lk(8)
s34: calculating the fitness function of the whole network by taking the mean square error as an index:
in the formula (9), MSE represents a mean square error of the network; n represents the time scale of the training data;
adjusting the weight coefficient a along the direction of the negative gradient by adopting a back propagation BP methodh1，ah2,bh1，bh2,Wh1i,Wh2i,Wjh1And W andh2the back propagation BP method is performed in calculating a fitness function of the entire network.
S35: if the fitness function error meets the precision control requirement, ending the training; otherwise, returning to step S33 to recalculate the output of the network until the fitness function error of step S35 meets the requirement of accuracy control.
8. The numerical control machine state prediction method based on time series according to claim 1, characterized in that when 50 groups of data are selected from the nearest clusters for training the WNN-NARMAL2 model, the precision of the operating condition parameter of the rotation angle error of training and prediction is between-0.1 and 0.06 um.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911369633.9A CN111103846A (en) | 2019-12-26 | 2019-12-26 | Numerical control machine tool state prediction method based on time sequence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911369633.9A CN111103846A (en) | 2019-12-26 | 2019-12-26 | Numerical control machine tool state prediction method based on time sequence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111103846A true CN111103846A (en) | 2020-05-05 |
Family
ID=70424727
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911369633.9A Pending CN111103846A (en) | 2019-12-26 | 2019-12-26 | Numerical control machine tool state prediction method based on time sequence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111103846A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114328131A (en) * | 2022-03-09 | 2022-04-12 | 深圳市佳贤通信设备有限公司 | Function time consumption monitoring method based on clock period record |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103878791A (en) * | 2014-04-12 | 2014-06-25 | 福州大学 | Industrial robot external-sensor-free external force detection method |
| CN205318211U (en) * | 2016-01-20 | 2016-06-15 | 湖南工学院 | Inertia match parameter formula digit control machine tool is optimized and real -time monitoring system |
| CN107678398A (en) * | 2016-09-22 | 2018-02-09 | 华中科技大学 | The method that breaking for Digit Control Machine Tool detects |
| CN108490880A (en) * | 2018-04-24 | 2018-09-04 | 湖北文理学院 | A kind of numerical control machine tool cutting cutting-tool wear state method of real-time |
-
2019
- 2019-12-26 CN CN201911369633.9A patent/CN111103846A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103878791A (en) * | 2014-04-12 | 2014-06-25 | 福州大学 | Industrial robot external-sensor-free external force detection method |
| CN205318211U (en) * | 2016-01-20 | 2016-06-15 | 湖南工学院 | Inertia match parameter formula digit control machine tool is optimized and real -time monitoring system |
| CN107678398A (en) * | 2016-09-22 | 2018-02-09 | 华中科技大学 | The method that breaking for Digit Control Machine Tool detects |
| CN108490880A (en) * | 2018-04-24 | 2018-09-04 | 湖北文理学院 | A kind of numerical control machine tool cutting cutting-tool wear state method of real-time |
Non-Patent Citations (1)
| Title |
|---|
| 金超: "《基于工况的数控加工热误差与切削振动预测方法研究》", 《中国博士学位论文全文数据库 (工程科技Ⅰ辑)》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114328131A (en) * | 2022-03-09 | 2022-04-12 | 深圳市佳贤通信设备有限公司 | Function time consumption monitoring method based on clock period record |
| CN114328131B (en) * | 2022-03-09 | 2022-05-27 | 深圳市佳贤通信设备有限公司 | Function time consumption monitoring method based on clock period record |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111590581B (en) | Positioning compensation method and device for robot | |
| Fu et al. | Integrated thermal error modeling of machine tool spindle using a chicken swarm optimization algorithm-based radial basic function neural network | |
| Zhang et al. | Machine tool thermal error modeling and prediction by grey neural network | |
| CN109472057B (en) | Product processing quality prediction device and method based on cross-process implicit parameter memory | |
| CN110909762B (en) | Robot posture recognition method and device based on multi-sensor fusion | |
| CN111581832A (en) | Improved gray Elman neural network hovercraft motion prediction method based on ARMA model correction | |
| CN112276944A (en) | Man-machine cooperation system control method based on intention recognition | |
| CN111752151A (en) | Adaptive tracking and compensating method and system for grinding and polishing industrial blade | |
| CN112416021B (en) | Learning-based path tracking prediction control method for rotor unmanned aerial vehicle | |
| CN111881574A (en) | Wind turbine generator key component reliability modeling method based on distribution function optimization | |
| CN111190429A (en) | Unmanned aerial vehicle active fault-tolerant control method based on reinforcement learning | |
| CN113910218B (en) | Robot calibration method and device based on kinematic and deep neural network fusion | |
| CN111103846A (en) | Numerical control machine tool state prediction method based on time sequence | |
| CN115128978A (en) | Internet of things environment big data detection and intelligent monitoring system | |
| CN112462608B (en) | Discrete sliding mode track and speed tracking control method for high-speed train | |
| CN108959787A (en) | Consider the thermal deformation prediction technique and system of the macro dual drive system of actual condition | |
| CN112149896A (en) | Attention mechanism-based mechanical equipment multi-working-condition fault prediction method | |
| CN114030008B (en) | Industrial robot practical training energy consumption measurement method based on data driving | |
| CN114399024B (en) | Oil gas concentration big data intelligent detection system | |
| Jiang et al. | Nonlinear time-series modeling of feed drive system based on motion states classification | |
| CN114012733A (en) | Mechanical arm control method for scribing PC (personal computer) component mold | |
| Peng et al. | Wind turbine blades icing failure prognosis based on balanced data and improved entropy | |
| CN111752144B (en) | Method for predicting cyclic ripple and intelligent control system | |
| Shi et al. | Online Monitoring Dynamic Characteristics in Thin-Walled Structure Milling: A Physics-Constrained Bayesian Updating Approach | |
| CN109656229A (en) | The construction method of robot end's performance prediction model based on GA-RBF network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200505 |
|
| RJ01 | Rejection of invention patent application after publication |