CN107608207B - Aluminum profile shakeout detection method based on machine learning - Google Patents
Aluminum profile shakeout detection method based on machine learning Download PDFInfo
- Publication number
- CN107608207B CN107608207B CN201710666194.2A CN201710666194A CN107608207B CN 107608207 B CN107608207 B CN 107608207B CN 201710666194 A CN201710666194 A CN 201710666194A CN 107608207 B CN107608207 B CN 107608207B
- Authority
- CN
- China
- Prior art keywords
- aluminum profile
- shakeout
- diameter
- method based
- machine learning
- 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
Landscapes
- Length Measuring Devices With Unspecified Measuring Means (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an aluminum profile shakeout detection method based on machine learning, which adopts a neural network to perform pre-prediction processing and combines a Q-learning algorithm to perform strategy processing of specific conditions, wherein a pre-processing part of the neural network can predict the amount of primary shakeout, so as to ensure that the effect required by the diameter of a black spot is close to the effect required by the diameter of the black spot under the condition of operation as little as possible, and on the basis, the wear-resisting quality grade is introduced as part of the Q-learning algorithm for input, and different shakeout amounts are regulated and controlled according to the real-time diameter of the black spot and the resistance value of a resistor, so that the detection efficiency and the accuracy are maximized. The aluminum profile shakeout detection method based on machine learning can be widely applied to the industrial field.
Description
Technical Field
The invention relates to the field of industry, in particular to an aluminum profile shakeout detection method based on machine learning.
Background
In the production process of the aluminum profile, quality detection is particularly important, wherein the wear resistance detection of the aluminum profile is important work of the quality detection. At present, a method adopted for detecting the wear resistance of the aluminum profile in the industry is a shakeout test method, the shakeout test is mainly completed manually by people, however, the shakeout test process is particularly long, people need to observe the change of the surface of the aluminum profile while adding the sand continuously, inaccurate data and labor consumption caused by fatigue are easy to occur, and the method can be used for indicating that the requirement is difficult to meet by only relying on the labor. Carrying out automatic transformation and being favorable to reducing the human cost, can raise the efficiency simultaneously.
In the whole process of the shakeout test, when sand impacts the surface of the aluminum profile, a black point (a mark for abrasion of a coating) can slowly appear on the surface of the aluminum profile, when the diameter of the black point is larger than or equal to the empirical diameter, the aluminum profile is worn, whether the aluminum profile is worn or not is confirmed by measuring the resistance at the position of the black point, when the resistance value of the resistance is smaller than the empirical value, the coating is worn, and the amount of the total sand used at the moment is used for measuring the abrasion resistance of the aluminum profile.
Due to the characteristics of the test method, the diameter of the black spot and the resistance value of the resistor need to be monitored in real time during the test process. The common method is to process the image of the diameter of the black spot by the image processing technology, and the resistance value of the resistor can be obtained by measuring the resistance module.
The common control method comprises the steps of obtaining a fixed sand drop amount through experience, measuring the diameter of a black spot through fixing a certain amount of sand drop, performing cyclic operation until the diameter of the black spot is larger than or equal to an empirical value, changing the fixed sand drop amount at the moment, performing resistance test, performing cyclic operation until the resistance is smaller than the empirical resistance, and stopping operation.
And a prediction function is deduced through a large amount of experimental data by more advanced intelligent modification, so that the relation between the fixed sand falling amount and the time axis is obtained, and the efficiency is improved.
The fixed shakeout amount is adopted for testing, the efficiency is very low, if the value of the shakeout amount is set to be smaller, the required time is overlong, and if the shakeout amount is set to be overlarge, the operation is easy to occur, so that the data is inaccurate.
The relationship between the sand falling amount and the time axis is simulated through data, so that the efficiency problem can be improved, but the sand falling test method is to measure the quality of different aluminum profiles, so that the relationship between the sand falling amount and the time axis of the aluminum profiles with different quality and wear resistance is different, and the relationship is difficult to accurately simulate.
Disclosure of Invention
In order to solve the technical problems, the invention aims to: provided is an aluminum profile shakeout detection method based on machine learning.
The technical scheme adopted by the invention is as follows: an aluminum profile shakeout detection method based on machine learning comprises the following steps:
A. training a BP neural network which takes the thickness and the process conditions of the aluminum profile as input and the initial shakeout quantity under the standard quality grade of the aluminum profile as output according to the experimental data of shakeout detection;
B. calculating according to the initial sand falling amount and the diameter of the black points on the surface of the aluminum profile to obtain the surface quality grade of the aluminum profile;
C. the method comprises the following steps of (1) taking the quality grade of the surface of the aluminum profile, the diameter of black spots on the surface of the aluminum profile and the resistance which are detected in real time as input, and calculating and controlling the shakeout time by adopting a Q-learning algorithm;
D. and when the detection resistance value of the black points on the surface of the aluminum profile reaches the empirical resistance value, outputting the total shakeout amount as the detection result of the wear resistance of the aluminum profile.
Further, in the Q-learning algorithm in step C, a discrete state space D ═ D1, D2, D3,.. gtang, dn ], R ═ R1, R2, R3,. gtang, rn ] and an action space T ═ T1, T2, T3,. gtang, tn ] are obtained according to the fuzzy control principle and the historical data,
the state function is f (d, r, degree), f is the function related to D, R and degree, d, r and degree are the respective currently detected diameter, resistance and quality level,
the reward function Q (, t) ═ Q (, t) + α (r + γ Q (', t') -Q (, t)),
alpha is the update step, gamma is the learning rate, T is the parameter in the time matrix T,
policy function pi () ═ argmaxQ (, t'), the policy function is the state space of the greedy algorithm.
And further, calculating according to the initial sand falling amount and the diameter of the black points on the surface of the aluminum profile in the step B and referring to an industry standard to obtain a surface quality grade value of the aluminum profile.
The invention has the beneficial effects that: the invention adopts the neural network to carry out pre-prediction processing and combines the Q-learning algorithm to carry out strategy processing of specific conditions, the pre-processing part of the neural network can predict the amount of primary sand falling, the effect required by the diameter of the black spot is ensured to be close to under the condition of operation as less as possible, the wear-resistant quality grade is introduced as part of the input of the Q-learning algorithm on the basis, the adjustment and control of different sand falling amounts are carried out according to the real-time diameter of the black spot and the resistance value of the resistor, and the detection efficiency and the accuracy are maximized.
Drawings
FIG. 1 is a flow chart of the steps of the method of the present invention.
Detailed Description
The following further describes embodiments of the present invention with reference to the accompanying drawings:
referring to fig. 1, an aluminum profile shakeout detection method based on machine learning is characterized by comprising the following steps:
A. training a BP neural network which takes the thickness and the process conditions of the aluminum profile as input and the initial shakeout quantity under the standard quality grade of the aluminum profile as output according to the experimental data of shakeout detection;
B. calculating according to the initial sand falling amount and the diameter of the black points on the surface of the aluminum profile to obtain the surface quality grade of the aluminum profile;
C. the method comprises the following steps of (1) taking the quality grade of the surface of the aluminum profile, the diameter of black spots on the surface of the aluminum profile and the resistance which are detected in real time as input, and calculating and controlling the shakeout time by adopting a Q-learning algorithm;
D. and when the detection resistance value of the black points on the surface of the aluminum profile reaches the empirical resistance value, outputting the total shakeout amount as the detection result of the wear resistance of the aluminum profile.
Further preferably, in the Q-learning algorithm in step C, a discrete state space D ═ D1, D2, D3,. times, dn ], R ═ R1, R2, R3,. times, rn ], and an action space T ═ T1, T2, T3,. times, tn ] are obtained according to a fuzzy control principle and historical data,
the state function is f (d, r, degree), f is the function related to D, R and degree, d, r and degree are the respective currently detected diameter, resistance and quality level,
the reward function Q (, t) ═ Q (, t) + α (r + γ Q (', t') -Q (, t)),
alpha is the update step, gamma is the learning rate, T is the parameter in the time matrix T,
policy function pi () ═ argmaxQ (, t'), the policy function is the state space of the greedy algorithm.
Further as a preferred embodiment, in the step B, the surface quality grade value of the aluminum profile is calculated according to the initial shakeout amount and the diameter of the black spot on the surface of the aluminum profile and the industry standard.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (2)
1. An aluminum profile shakeout detection method based on machine learning is characterized by comprising the following steps:
A. training a BP neural network which takes the thickness and the process conditions of the aluminum profile as input and the initial shakeout quantity under the standard quality grade of the aluminum profile as output according to the experimental data of shakeout detection;
B. calculating according to the initial sand falling amount and the diameter of the black points on the surface of the aluminum profile to obtain the surface quality grade of the aluminum profile;
C. the method comprises the following steps of (1) taking the quality grade of the surface of the aluminum profile, the diameter of black spots on the surface of the aluminum profile and the resistance which are detected in real time as input, and calculating and controlling the shakeout time by adopting a Q-learning algorithm;
D. and when the detection resistance value of the black points on the surface of the aluminum profile reaches the empirical resistance value, outputting the total shakeout amount as the detection result of the wear resistance of the aluminum profile.
2. The aluminum profile shakeout detection method based on machine learning according to claim 1, characterized in that: and in the step B, calculating according to the initial sand falling amount and the diameter of the black points on the surface of the aluminum profile and referring to an industry standard to obtain a surface quality grade numerical value of the aluminum profile.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710540690 | 2017-07-05 | ||
| CN2017105406903 | 2017-07-05 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107608207A CN107608207A (en) | 2018-01-19 |
| CN107608207B true CN107608207B (en) | 2020-10-27 |
Family
ID=61064736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710666194.2A Active CN107608207B (en) | 2017-07-05 | 2017-08-07 | Aluminum profile shakeout detection method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107608207B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110553942B (en) * | 2019-09-23 | 2024-06-25 | 中节能太阳能科技（镇江）有限公司 | Automatic change wear-resisting testing arrangement of circulation shakeout |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202403992U (en) * | 2011-12-29 | 2012-08-29 | 咸阳天誉建材检验有限公司 | Shakeout abrasion resistance tester |
| CN204116209U (en) * | 2014-10-24 | 2015-01-21 | 中国石油天然气集团公司 | A kind of coating high-temp abrasion resistance tester |
| JP2015151411A (en) * | 2014-02-10 | 2015-08-24 | 住友ベークライト株式会社 | Window for machine tool |
| CN104897498A (en) * | 2015-06-19 | 2015-09-09 | 芜湖精塑实业有限公司 | Method for testing wear resistance of surface coating of profile via falling sand tester |
| CN106878403A (en) * | 2017-01-25 | 2017-06-20 | 东南大学 | Based on the nearest heuristic service combining method explored |
-
2017
- 2017-08-07 CN CN201710666194.2A patent/CN107608207B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202403992U (en) * | 2011-12-29 | 2012-08-29 | 咸阳天誉建材检验有限公司 | Shakeout abrasion resistance tester |
| JP2015151411A (en) * | 2014-02-10 | 2015-08-24 | 住友ベークライト株式会社 | Window for machine tool |
| CN204116209U (en) * | 2014-10-24 | 2015-01-21 | 中国石油天然气集团公司 | A kind of coating high-temp abrasion resistance tester |
| CN104897498A (en) * | 2015-06-19 | 2015-09-09 | 芜湖精塑实业有限公司 | Method for testing wear resistance of surface coating of profile via falling sand tester |
| CN106878403A (en) * | 2017-01-25 | 2017-06-20 | 东南大学 | Based on the nearest heuristic service combining method explored |
Non-Patent Citations (2)
| Title |
|---|
| 建筑铝合金型材耐磨性测试方法;纪红 等;《理化检验·物理分册》;20101231;第180-182页 * |
| 氧化膜标准试样50次落砂试验规律性研究;陈超 等;《2016中国铝加工产业技术创新交流大会》;20161231;第517-524页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107608207A (en) | 2018-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11574192B2 (en) | Controller training based on historical data | |
| KR101889668B1 (en) | Rolling simulation device | |
| CN113909996B (en) | High-end equipment machining state monitoring method and system based on digital twinning | |
| CN101863088A (en) | Method for forecasting Mooney viscosity in rubber mixing process | |
| CN107608207B (en) | Aluminum profile shakeout detection method based on machine learning | |
| KR20200035550A (en) | Estimation method of transmission temperature of molten steel using artificial neural network technique | |
| CN117469603B (en) | Multi-water-plant water supply system pressure optimal control method based on big data learning | |
| CN109696829B (en) | Method and device for controlling a process in a system, in particular a grinding process in a grinding device | |
| CN104550801A (en) | Control method and control device for continuous bloom casting light reduction | |
| CN1609736B (en) | Controlling device, system utilizing the same controller and controlling method | |
| JP7264697B2 (en) | Plant operation support system and plant operation support method | |
| JP4517048B2 (en) | Pouring trainer system | |
| JP6098553B2 (en) | Rejuvenated phosphorus amount prediction device, recovered phosphorus amount prediction method, and converter dephosphorization control method | |
| US20220307944A1 (en) | Apparatus, method, and computer-readable medium | |
| CN103135444A (en) | Steel production energy consumption immune prediction control model | |
| CN103713619A (en) | Test method for ore grinding process operation control method | |
| JP7263020B2 (en) | Treated water quality estimation device, treated water quality estimation method and program | |
| CN112596378A (en) | Coating thickness control method and training method and device of coating thickness control model | |
| JP4834988B2 (en) | Continuous process control method and continuous process control system | |
| CN109902867B (en) | Water level prediction and measurement control method | |
| CN111250545B (en) | Control system and method for reducing swing steel coil over-thickness rate | |
| CN111985146A (en) | Method and device for determining measure potential well | |
| JP2019184575A (en) | Measurement operation parameter adjustment device, machine learning device, and system | |
| KR102683969B1 (en) | Plating control method and system using coating weight prediction model | |
| JPH06199904A (en) | Method for changing continuous operation condition of continuous polymerization process |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Li Qidong Inventor after: Xiao Pan Inventor after: Huang Guancheng Inventor after: Lin Jianfa Inventor before: Li Qidong Inventor before: Xiao Pan Inventor before: Cai Nian Inventor before: Chen Xindu Inventor before: Wang Han |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |