CN106407527A - Wearing capacity prediction method based on Bayesian network - Google Patents
Wearing capacity prediction method based on Bayesian network Download PDFInfo
- Publication number
- CN106407527A CN106407527A CN201610803213.7A CN201610803213A CN106407527A CN 106407527 A CN106407527 A CN 106407527A CN 201610803213 A CN201610803213 A CN 201610803213A CN 106407527 A CN106407527 A CN 106407527A
- Authority
- CN
- China
- Prior art keywords
- bayesian network
- probability
- parameter
- prediction method
- method based
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Length Measuring Devices With Unspecified Measuring Means (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
Disclosed is a wearing capacity prediction method based on a Bayesian network. The wearing capacity prediction method is characterized by comprising the following steps of 1, obtaining parameter data under different working conditions through experiments; 2, performing data analysis and learning, carrying out analysis and modeling on the parameter data obtained in the step 1, including structural learning and parameter learning, wherein the structural learning is for completing the establishment of the Bayesian network while the parameter learning is for obtaining the Bayesian network; and updating the parameter of each node in the Bayesian network according to a parameter data set; and 3, performing reasoning and prediction on the Bayesian network, according to the Bayesian network obtained in the step 2, by inputting one group (relative speed, load, temperature and hardness) of data, the wearing capacity of the fifth parameter of the nodes corresponding to the group of the data, and the corresponding probability corresponding to the wearing capacity can be deduced in the Bayesian network obtained in the step 2. The patterned Bayesian network based on probability reasoning has high advantage in solving the problem caused by uncertainty and relevancy.
Description
Technical field
The invention belongs to mechanical field, more particularly to it is based on computer intelligence algorithm and predicts workpiece under different operating modes
Wear extent Forecasting Methodology.
Background technology
With regard to abrasion, Chinese scholars have all done substantial amounts of research it is proposed that various theory.Wear form mainly has abrasive material
Abrasion, adhesive wear, erosive wear, Surface fatigue wear, corrosive wear etc..Prediction to wear extent, currently used does
Method is usually test method(s), calculating method and Statistics Method.
Factor influential on wear extent is a lot, has relative velocity, load, temperature, the factor such as hardness.Even if but in phase
With operating mode, same material sample is tested, the data obtaining wear extent is all not necessarily the same.Thus existing Forecasting Methodology
All can not accurately realize to wear extent under different operating modes and to should corresponding probability under wear extent prediction.
Content of the invention
The present invention is to solve the above problems it is proposed that a kind of Abrasion prediction method based on Bayesian network, the method
The Bayesian network using is a kind of probability net, and it is the graphical network based on probability inference.Bayesian network is for solution
The problem that never definitiveness and relatedness cause has very big advantage.
A kind of Abrasion prediction method based on Bayesian network, using the relative velocity under different operating modes, load, temperature
Degree, hardness parameter are predicted it is characterised in that comprising the following steps to the wear extent of workpiece and corresponding probability：
Step one, obtains the supplemental characteristic under different operating modes by experiment；
Step 2, data analysiss and study, the supplemental characteristic that step one is obtained is analyzed modeling, and this step includes tying
Structure study and parameter learning,
Structure learning is used for completing the structure of Bayesian network,
Parameter learning is used in the Bayesian network obtaining, and the set according to supplemental characteristic is every in Bayesian network to update
The parameter of individual node；
Step 3, Bayesian Network Inference is predicted, the Bayesian network being obtained according to step 2, one group of input (speed relatively
Degree, load, temperature, hardness) data can derive to the node that should organize data in the Bayesian network in step 2
The 5th parameter wear extent and to should corresponding probability under wear extent.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：Wherein,
Structure learning is realized by structure learning algorithm.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：Wherein,
Structure learning algorithm includes K2 algorithm, K3 algorithm.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：Wherein,
The parameter of node includes predictor and predicts the outcome, and predictor includes relative velocity, load, temperature and hardness, prediction
Result includes wear extent and to should corresponding probability under wear extent.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：Wherein,
Described relative velocity is rotating speed V, and rotating speed V is divided into 4 parts：V1 represents 0<V<V m/s, V2 represent V m/s<V<F m/s, V3 generation
Table F m/s<V<T m/s, V4 represent T m/s<V<fm/s；Described load is load F, and load F is divided into 4 parts：F1 generation table 0<F
<E N, F2 represent e N<F<G N, F3 represent g N<F<HN, F4 represent hN<F<i N；Described temperature is experimental temperature T, will test
Temperature T is divided into 3 parts：T1 represents 0<T<J DEG C, T2 represents j DEG C<T<K DEG C, T3 represents k DEG C<T<L DEG C, wherein, V, F, T, f, e,
According to corresponding described rotating speed V, corresponding described load F, corresponding described experimental temperature T is carried out the numerical value of g, h, i, j, k, l
Equilibrium divides.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：Wherein,
Described wear extent is abrasion percentage ratio f, and abrasion percentage ratio f is divided into 3 parts：F1 represents 0<f<M%, f2 represent m%<f<N%,
F3 represents n%<f<P%, wherein, the numerical value of m, n, p carries out equalizing division according to corresponding described abrasion percentage ratio f, and 0<m
<n<p<100.
The Abrasion prediction method based on Bayesian network that the present invention provides, can also have the feature that：To step
Supplemental characteristic in rapid one carry out statistical analysiss obtain different relative velocities, different loads, wear extent under different temperatures right
Answer probability：Under V1F1T1 operating mode, probability f1 is q%, and the probability of f2 is r%, and the probability of f3 is 0；Under V2F1T1 operating mode,
Probability f1 is s%, and the probability of f2 is t%, and the probability of f3 is 0.
Invention effect and effect
According to the Abrasion prediction method based on Bayesian network provided by the present invention, used due to this Forecasting Methodology
Bayesian network is a kind of probability net, and it is the graphical network-Bayesian network based on probability inference, never true for solution
The problem that qualitative and relatedness causes has very big advantage.
Brief description
Fig. 1 is that the probabilistic relation of each node in the Abrasion prediction method based on Bayesian network of the present invention is illustrated
Figure.
Specific embodiment
In order that technological means, creation characteristic, reached purpose and effect that the present invention realizes are easy to understand, below tie
Close embodiment the present invention is specifically addressed based on the principle steps using effect of the Abrasion prediction method of Bayesian network.
Embodiment
Based on the Abrasion prediction method of Bayesian network, using the relative velocity under different operating modes, load, temperature, hard
Degree parameter is predicted it is characterised in that comprising the following steps to the wear extent of workpiece and corresponding probability：
Step one, obtains the supplemental characteristic under different operating modes by experiment.These supplemental characteristics include relative velocity, carry
Lotus, temperature, hardness and wear extent and to the corresponding probability that should occur under wear extent.
Step 2, data analysiss and study, the supplemental characteristic that step one is obtained is analyzed modeling, and this step includes tying
Structure study and parameter learning.
Structure learning is used for completing the structure of Bayesian network, and Structure learning is that the structure of network can pass through association area
The analysis of expert also can pass through some structure learning algorithms such as K2 algorithm, and K3 algorithm etc. is determined.
Parameter learning is used in the Bayesian network obtaining, and the set according to supplemental characteristic is every in Bayesian network to update
The parameter of individual node, wherein, the parameter of node includes predictor and predictor, predictor include relative velocity, load,
Temperature and hardness, predictor includes wear extent and to should corresponding probability under wear extent.
Then update conditional probability under its father node for each node according to data, such as case hardness is a section
Point, its father node be temperature, can obtain at a certain temperature according to data, the value that case hardness is likely to occur and they go out
Existing probability.
Step 3, Bayesian Network Inference is predicted, the Bayesian network being obtained according to step 2, one group of input (speed relatively
Degree, load, temperature, hardness) data can derive to the node that should organize data in the Bayesian network in step 2
The 5th parameter wear extent and to should corresponding probability under wear extent.
Wherein, relative velocity is rotating speed V, rotating speed V is 4m/s, is then divided into 1,2,3,4 four 4 parts of sections：
V1 represents 0<V<1m/s,
V2 represents 1m/s<V<2m/s,
V3 represents 2m/s<V<3m/s,
V4 represents 3m/s<V<4m/s；
Load is load F, and load F is that 40N is then divided into 10,20,30,40 four 4 parts of sections：
F1 generation table 0<F<10N,
F2 represents 10N<F<20N,
F3 represents 20N<F<30N,
F4 represents 30N<F<40N；
Temperature is experimental temperature T, experimental temperature T is 30 DEG C and is then divided into 10,20,30 three 3 parts of sections：
T1 represents 0<T<10 DEG C,
T2 represents 10 DEG C<T<20 DEG C,
T3 represents 20 DEG C<T<30℃.
Wear extent is abrasion percentage ratio f, abrasion percentage ratio f is 30% and is divided into 10%, 20%, 30% 3 section
3 parts：
F1 represents 0<f<10%,
F2 represents 10%<f<20%,
F3 represents 20%<f<30%.
Supplemental characteristic in step one is carried out statistical analysiss obtain different relative velocities, different loads, under different temperatures
Wear extent corresponding probability：
Under V1F1T1 operating mode, occur f1 probability be 80%, f2 probability be 20%, f3 probability be 0；
Under V2F1T1 operating mode, occur f1 probability be 70%, f2 probability be 30%, f3 probability be 0.
Fig. 1 is that the probabilistic relation of each node in the Abrasion prediction method based on Bayesian network of the present invention is illustrated
Figure.
As shown in figure 1, having obtained the probabilistic relation between the structure of Bayesian network and each node in previous step, then
In known speed, after the state of load and temperature, we can progressively release surface temperature, wear extent possible values and they go out
Existing corresponding probability.
When input corresponding (relative velocity, load, temperature, hardness) node when it is possible to accordingly obtain right
Corresponding probability under wear extent fn answered and this wear extent fn.
The effect of embodiment and beneficial effect
The Abrasion prediction method based on Bayesian network being provided according to the present embodiment, because this Forecasting Methodology uses
Bayesian network be a kind of probability net, it is the graphical network based on probability inference.Bayesian network is for solution never
The problem that definitiveness and relatedness cause has very big advantage.
Claims (7)
1. a kind of Abrasion prediction method based on Bayesian network, using the relative velocity under different operating modes, load, temperature,
Hardness parameter is predicted it is characterised in that comprising the following steps to the wear extent of workpiece and corresponding probability：
Step one, obtains the supplemental characteristic under different operating modes by experiment；
Step 2, data analysiss and study, the described supplemental characteristic that step one is obtained is analyzed modeling, and this step includes tying
Structure study and parameter learning,
Structure learning is used for completing the structure of described Bayesian network,
Parameter learning is used in the described Bayesian network obtaining, and the set according to described supplemental characteristic is updating described Bayes
The parameter of each node in network；
Step 3, Bayesian Network Inference is predicted, the Bayesian network being obtained according to step 2, and (relative velocity carries for one group of input
Lotus, temperature, hardness) data can derive to the described section that should organize data in the described Bayesian network in step 2
5th parameter wear extent of point and to should corresponding probability under wear extent.
2. the Abrasion prediction method based on Bayesian network according to claim 1 it is characterised in that：
Wherein, described Structure learning is realized by structure learning algorithm.
3. the Abrasion prediction method based on Bayesian network according to claim 2 it is characterised in that：
Wherein, described structure learning algorithm includes K2 algorithm, K3 algorithm.
4. the Abrasion prediction method based on Bayesian network according to claim 1 it is characterised in that：
Wherein, the parameter of described node includes predictor and predicts the outcome,
Described predictor includes described relative velocity, load, temperature and described hardness,
Described predict the outcome including described wear extent and to should described corresponding probability under wear extent.
5. the Abrasion prediction method based on Bayesian network according to claim 4 it is characterised in that：
Wherein, described relative velocity is rotating speed V, and rotating speed V is divided into 4 parts：
V1 represents 0<V<Am/s,
V2 represents am/s<V<B m/s,
V3 represents b m/s<V<C m/s,
V4 represents c m/s<V<d m/s；
Described load is load F, and load F is divided into 4 parts：
F1 generation table 0<F<E N,
F2 represents e N<F<G N,
F3 represents g N<F<HN,
F4 represents h N<F<i N；
Described temperature is experimental temperature T, and experimental temperature T is divided into 3 parts：
T1 represents 0<T<J DEG C,
T2 represents j DEG C<T<K DEG C,
T3 represents k DEG C<T<L DEG C,
Wherein, the numerical value of a, b, c, d, e, g, h, i, j, k, l, according to corresponding described rotating speed V, corresponding described load F, corresponds to
Described experimental temperature T carry out equalize divide.
6. the Abrasion prediction method based on Bayesian network according to claim 5 it is characterised in that：
Wherein, described wear extent is abrasion percentage ratio f, and abrasion percentage ratio f is divided into 3 parts：
F1 represents 0<f<M%,
F2 represents m%<f<N%,
F3 represents n%<f<P%,
Wherein, the numerical value of m, n, p carries out equalizing division according to corresponding described abrasion percentage ratio f, and 0<m<n<p<100.
7. the Abrasion prediction method based on Bayesian network according to claim 6 it is characterised in that：
Supplemental characteristic in step one is carried out with statistical analysiss and obtains different relative velocities, different loads, the mill under different temperatures
The corresponding probability of damage amount：
Under V1F1T1 operating mode, probability f1 is q%, and the probability of f2 is r%, and the probability of f3 is 0；
Under V2F1T1 operating mode, probability f1 is s%, and the probability of f2 is t%, and the probability of f3 is 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610803213.7A CN106407527A (en) | 2016-09-06 | 2016-09-06 | Wearing capacity prediction method based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610803213.7A CN106407527A (en) | 2016-09-06 | 2016-09-06 | Wearing capacity prediction method based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106407527A true CN106407527A (en) | 2017-02-15 |
Family
ID=57999737
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610803213.7A Pending CN106407527A (en) | 2016-09-06 | 2016-09-06 | Wearing capacity prediction method based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106407527A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102176217A (en) * | 2010-12-20 | 2011-09-07 | 西安瑞特快速制造工程研究有限公司 | Method for estimating reliability of numerical control machine tool cutting tool based on logistic model |
| CN102689230A (en) * | 2012-05-09 | 2012-09-26 | 天津大学 | Tool wear condition monitoring method based on conditional random field model |
| CN105205221A (en) * | 2015-08-26 | 2015-12-30 | 电子科技大学 | Precision reliability analysis method for heavy numerical control machine tool |
-
2016
- 2016-09-06 CN CN201610803213.7A patent/CN106407527A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102176217A (en) * | 2010-12-20 | 2011-09-07 | 西安瑞特快速制造工程研究有限公司 | Method for estimating reliability of numerical control machine tool cutting tool based on logistic model |
| CN102689230A (en) * | 2012-05-09 | 2012-09-26 | 天津大学 | Tool wear condition monitoring method based on conditional random field model |
| CN105205221A (en) * | 2015-08-26 | 2015-12-30 | 电子科技大学 | Precision reliability analysis method for heavy numerical control machine tool |
Non-Patent Citations (2)
| Title |
|---|
| 袁成清 等: "基于贝叶斯网络的摩擦学系统状态辨识的知识获取", 《润滑与密封》 * |
| 陈洪涛: "基干多参量信息融合的刀具磨损状态识别及预测技术研究", 《中国博士学位论文全文数据库 工程科技Ⅰ辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tang et al. | Data analytics and optimization for smart industry | |
| Zhao et al. | EnLSTM-WPEO: Short-term traffic flow prediction by ensemble LSTM, NNCT weight integration, and population extremal optimization | |
| Shahsavari-Pour et al. | A novel hybrid meta-heuristic algorithm for solving multi objective flexible job shop scheduling | |
| Lu et al. | Bayesian optimization with reference models: A case study in MPC for HVAC central plants | |
| Zhao et al. | An improved particle swarm optimization with decline disturbance index (DDPSO) for multi-objective job-shop scheduling problem | |
| Cui et al. | An improved Lagrangian relaxation approach to scheduling steelmaking-continuous casting process | |
| Jia et al. | An ACO algorithm for makespan minimization in parallel batch machines with non-identical job sizes and incompatible job families | |
| CN106022521A (en) | Hadoop framework-based short-term load prediction method for distributed BP neural network | |
| Ahmad et al. | The best candidates method for solving optimization problems | |
| Erten et al. | Stochastic optimization methods | |
| Huang et al. | Energy consumption forecasting for the nonferrous metallurgy industry using hybrid support vector regression with an adaptive state transition algorithm | |
| Chen et al. | Solving chemical dynamic optimization problems with ranking-based differential evolution algorithms | |
| Lee et al. | Multistage MR-CART: Multiresponse optimization in a multistage process using a classification and regression tree method | |
| Ren et al. | Immune genetic algorithm for multi-objective flexible job-shop scheduling problem | |
| Wu et al. | A deep learning approach for solving linear programming problems | |
| Fernández-González et al. | Mixed discrete-continuous heuristic generative planning based on flow tubes | |
| Lotfi et al. | Fuzzy goal programming to optimization the multi-objective problem | |
| Al Adwan et al. | Predicting compressive strength of concrete using histogram-based gradient boosting approach for rapid design of mixtures | |
| Zaheer et al. | A differential evolution approach for solving integer programming problems | |
| CN106407527A (en) | Wearing capacity prediction method based on Bayesian network | |
| Jia et al. | A novel backtracking search with grey wolf algorithm for optimization | |
| CN115097737B (en) | Multi-level regulation and control method capable of being re-entered into manufacturing system | |
| Xu et al. | Fuzzy-like multiple objective multistage decision making | |
| Schmidt et al. | Reviving Simulated Annealing: Lifting its Degeneracies for Real-Time Job Scheduling. | |
| Liao et al. | Recurrent reinforcement learning for predictive overall equipment effectiveness |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170215 |
|
| RJ01 | Rejection of invention patent application after publication |