CN113240096A - Casting cylinder cover micro-structure prediction method based on rough set and neural network - Google Patents

Casting cylinder cover micro-structure prediction method based on rough set and neural network Download PDF

Info

Publication number
CN113240096A
CN113240096A CN202110635459.9A CN202110635459A CN113240096A CN 113240096 A CN113240096 A CN 113240096A CN 202110635459 A CN202110635459 A CN 202110635459A CN 113240096 A CN113240096 A CN 113240096A
Authority
CN
China
Prior art keywords
casting
neural network
data
cylinder cover
attribute
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110635459.9A
Other languages
Chinese (zh)
Other versions
CN113240096B (en
Inventor
黄渭清
李冬伟
刘金祥
李媛
任培荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN202110635459.9A priority Critical patent/CN113240096B/en
Publication of CN113240096A publication Critical patent/CN113240096A/en
Application granted granted Critical
Publication of CN113240096B publication Critical patent/CN113240096B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/22—Indexing; Data structures therefor; Storage structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Human Resources & Organizations (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a casting cylinder cover micro-structure prediction method based on a rough set and a neural network, and belongs to the related field of casting aluminum alloy cylinder covers. The invention aims to solve the problems that the final reasonable process parameters can be determined only by repeated trial and error in the existing casting process, and the casting process parameters can be optimized in the product design stage in order to reduce the design and production cost, so that the aims of reducing the metal consumable rate and the casting rejection rate are fulfilled, and the invention provides the method for predicting the microstructure of the cast aluminum alloy cylinder cover based on the rough set and the BP neural network; the method utilizes a rough set theory to reduce various index attributes of the casting and heat treatment process which affect the microstructure appearance of the material, and the data reduction function selects the index with larger weight on one hand, reduces the input dimension of a neural network on the other hand, enhances the learning efficiency of the neural network, and finally achieves the purpose of reducing the metal consumable rate and the casting rejection rate on the premise of improving the accuracy and efficiency of the microstructure appearance prediction.

Description

Casting cylinder cover micro-structure prediction method based on rough set and neural network
Technical Field
The invention relates to a casting cylinder cover micro-structure prediction method based on a rough set and a neural network, and belongs to the related field of casting aluminum alloy cylinder covers.
Background
In the field of casting aluminum alloy cylinder heads, the casting and heat treatment process parameters of the cylinder heads have influence on the microstructure morphology of the cylinder heads. If the casting and heat treatment process parameters are reasonably set, compact microstructures can be obtained, so that the metal consumption rate and the casting rejection rate are greatly reduced. The casting processing and heat treatment process parameters of the cast aluminum alloy cylinder cover are many, and the mapping relation between the processing technology of the cast aluminum alloy cylinder cover and the microstructure morphology of the cast aluminum alloy cylinder cover is difficult to reasonably describe through a simple analytical expression. In the actual production process, the cylinder cover mainly adopts an empirical production mode, usually, the technological parameters are determined by the production experience, and then the trial-made sample is produced. Repeated trial and error are needed in the casting process to determine the final reasonable process flow, the rejection rate is increased in the process, and the economic benefit is greatly reduced. If the microstructure morphology can be directly predicted through casting and heat treatment process parameters, great help is provided for the processing and production process of the cast aluminum alloy cylinder cover. How to rapidly predict the microstructure morphology of the cast aluminum alloy cylinder cover and establish the mapping relation between effective casting and heat treatment process parameters and the microstructure morphology is one of the problems to be solved in the current production process of the cast aluminum alloy cylinder cover.
At present, a numerical simulation technology is a mature and effective means for metallic property prediction, and a learner simulates the casting process of the aluminum alloy cylinder cover by using numerical simulation software to obtain the microstructure morphology, but the simulation time is long, the simulation times are many, an intelligent learning function is not provided, and the requirements for efficient and rapid prediction cannot be met.
The prediction method based on the rough set and the neural network is used in many fields, such as price prediction of real estate, application in transformer fault diagnosis, judgment and analysis of atmospheric pollution, prediction of earthquake and the like. After the rough set is extracted, outstanding results are obtained in the fields of artificial intelligence, mode recognition, fault diagnosis and the like. The research combines the respective advantages of the rough set and the neural network technology, establishes an intelligent prediction method under the condition of not changing the classification quality of the sample, and predicts the microstructure morphology of the cast aluminum alloy cylinder cover by using casting process parameters and heat treatment parameters. Compared with the prior prediction methods such as an empirical formula method, a finite element simulation analysis method and the like, the prediction method is more accurate and faster, more importantly, a large amount of time and economic cost are saved, and the optimization design process of casting the aluminum alloy cylinder cover product is promoted.
Disclosure of Invention
The invention aims to solve the problems that the final reasonable process parameters can be determined only by repeatedly trial and error in the existing casting process, and provides a method for predicting the microstructure of a cast aluminum alloy cylinder cover based on a rough set and a BP neural network in order to reduce the design and production cost, optimize the casting process parameters and achieve the purposes of reducing the metal consumable rate and the casting rejection rate; the method utilizes a rough set theory to reduce various index attributes of the casting and heat treatment process which affect the microstructure appearance of the material, and the data reduction function selects the index with larger weight on one hand, reduces the input dimension of a neural network on the other hand, enhances the learning efficiency of the neural network, and finally achieves the purpose of reducing the metal consumable rate and the casting rejection rate on the premise of improving the accuracy and efficiency of the microstructure appearance prediction.
The purpose of the invention is realized by the following technical scheme.
A cast cylinder cover micro-structure prediction method based on a rough set and a neural network comprises the following steps:
acquiring a casting and heat treatment process parameter database of a cast aluminum alloy cylinder cover, a micro-structure morphology database of a cylinder cover material and position parameters of data acquisition points; the casting process parameters comprise: solidification rate, casting temperature, casting mold temperature, mold filling time, mold filling pressure, mold filling speed, crusting time, pressurizing pressure, pressurizing speed, pressure maintaining time, pressure relief time and the like; the heat treatment process parameters comprise: solution treatment temperature, heat preservation time, cooling speed, aging temperature, aging time and the like; the microtexture topographic data of the cylinder head material comprises: the grain size, the secondary dendrite arm spacing, the two-dimensional porosity, the metallographic pore area, the maximum Feret size of the metallographic pore, the area mean value of eutectic silicon, the length-width ratio mean value of eutectic silicon, the maximum Feret size mean value of eutectic silicon, the roundness mean value of eutectic silicon particles and the like; the coordinate parameters of the cylinder cover data acquisition points comprise: x-coordinate value, y-coordinate value, z-coordinate value. Wherein the location parameter of the data acquisition point is a necessary parameter.
Step two, carrying out normalization processing on all the data acquired in the step one;
the data normalization processing formula is as follows:
Figure BDA0003103843180000021
wherein X represents test data before normalization, X*Denotes the normalized test data, XmaxAs the maximum value of the test data, XminIs the minimum of the experimental data.
Step three, discretizing the test data subjected to the normalization processing in the step two, then coding the data set, and obtaining a discretized decision table after coding;
step four, calculating importance of the condition attributes in the decision table obtained in the step three, and then carrying out data reduction according to the importance to obtain a reduction subset of the test data;
s41, defining the casting and heat treatment process parameters, the micro-structure topography data and the position parameter decision table as a quadruple:
DT＝(U，C∪D，V，f)
wherein: u: a domain of discourse;
c, U.D: c is a condition attribute set, and D is a decision attribute set;
v: v is the value range of the attribute;
f: an information function of the decision table.
S42 gives a decision table DT ═ (U, C ═ D, V, f), there is a set of attributes
Figure BDA0003103843180000022
There is a e C-B,
defining:
Figure BDA0003103843180000023
the importance of the condition attribute a to the condition attribute set B relative to the decision attribute set D.
S43, using genetic algorithm to reduce the attribute of the decision table to obtain the attribute reduced subset. Where the location parameters do not participate in the reduction.
And fifthly, obtaining a reduction subset through rough set attribute reduction, using the reduction subset as an input layer of the BP neural network, and selecting a training function and a transfer function to establish a BP neural network prediction model. The training function of the BP neural network comprises: trangd, trainlm, traingdm, trangda, traingdx, trainrp, etc.; the transfer function includes: log-sigmoid type, tan-sigmoid type, linear purelin, etc.
And step six, verifying the effectiveness of the BP neural network prediction model, and judging the effectiveness of the BP neural network model according to an acceptable relative error range.
And seventhly, after training the training data for multiple times, when the prediction error of the neural network prediction model reaches an acceptable range, storing the BP neural network, and then adopting the BP neural network model to perform subsequent micro-tissue morphology prediction.
Advantageous effects
The method can predict the microstructure state of the material at the casting process design stage of the product, not only can effectively reduce the design and production cost, but also can achieve the aim of reducing the metal consumable rate and the casting rejection rate, and plays a guiding role in the process production of the casting cylinder cover;
the invention utilizes rough set theory to reduce data, extracts core index as input, can reduce the complexity of neural network structure, accelerate the network training speed and predict the accuracy;
3, the method is widely applicable to prediction of the microstructure appearance of various casting cylinder covers, particularly prediction of the microstructure of an aluminum alloy cylinder cover casting material, and prediction and defect avoidance are provided for material promotion and optimization;
compared with the traditional data calculation, the BP neural network can reversely propagate the error, correct the weight and deviation of each layer of unit in time, and has the advantage of higher accuracy.
Drawings
FIG. 1 is a flow chart of a method for predicting the microstructure of a cast aluminum alloy cylinder head according to an embodiment of the present invention.
FIG. 2 is a diagram of a BP neural network prediction model.
FIG. 3 conditional attribute importance calculation.
FIG. 4 BP neural network predictions of porosity and secondary dendrite arm spacing are compared to experimental results. Wherein, the graph (a) is a porosity graph; graph (b) is the secondary dendrite arm spacing graph.
Detailed Description
The embodiments of the present invention will be described in further detail with reference to the drawings and examples.
A casting cylinder cover micro-structure prediction method based on a rough set and a neural network is disclosed, wherein the prediction process is shown in figure 1, and comprises the following steps:
step one, acquiring a micro-structure information database of a cast aluminum alloy cylinder cover material;
the casting processing technological parameters and the heat treatment parameters of the selected cast aluminum alloy cylinder cover are provided by a casting factory, and the micro-structure parameters of the cylinder cover are obtained by measuring a laboratory scanning microscope and an electron microscope. Respectively taking test samples from different position points of the low-pressure cast aluminum alloy cylinder cover, and taking position coordinate values x, y and z; the selected casting and heat treatment process parameters comprise: solidification rate, heat transfer coefficient, aging temperature and aging time; the fracture of the test sample piece is subjected to microstructural observation by utilizing a scanning electron microscope and an optical microscope, and the measured microstructural structure parameters of the sample piece at each position comprise grain structure size, pore structure size and eutectic silicon structure size, wherein the grain structure comprises grain size and secondary dendrite arm spacing, the pore structure comprises two-dimensional porosity, metallographic pore area, eutectic silicon structure comprises eutectic particle area mean value, eutectic particle length-width ratio mean value and eutectic particle maximum Feret size mean value. In this embodiment, the secondary dendrite arm spacing and porosity in the microstructure morphology are selected as the prediction objects of the BP neural network. The prediction of other micro-tissue morphology parameters can be implemented according to the prediction method of the embodiment.
Step two, carrying out normalization processing on the data acquired in the step one;
the obtained data base has different reference of each parameter, which can affect the prediction result, so the data set is normalized.
The data normalization processing formula is as follows:
Figure BDA0003103843180000042
wherein X represents test data before normalization, X*Denotes the normalized test data, XmaxAs the maximum value of the test data, XminIs the minimum of the experimental data. The data after the data set normalization processing of the casting process and the micro-structure of the cast aluminum alloy cylinder cover are shown in the table 1.
TABLE 1 micro-organization structure database after normalization processing
Figure BDA0003103843180000041
Figure BDA0003103843180000051
Carrying out discretization on the test data after the normalization processing, coding and establishing a decision table to obtain a discretized decision table;
s31 sets the casting process parameters and heat treatment process parameters as conditional attributes C ═ { a1, a2, …, a7}, a1 represents x-axis coordinates with the casting cylinder head nozzle as the origin, a2 represents y-axis coordinates with the casting cylinder head nozzle as the origin, a3 represents z-axis coordinates with the casting cylinder head nozzle as the origin, a4 represents the heat transfer coefficient, a5 represents the solidification rate, a6 represents the aging temperature, and a7 represents the aging time for the data set in table 1. Considering the limitations of the test conditions, the heat transfer coefficient, aging temperature and aging time in this example were all set to the same conditions, and the implementation of the prediction method was not affected. Because the casting process and the heat treatment process of the cast aluminum alloy cylinder cover both influence the size of the secondary dendrite arm spacing, and the porosity is only influenced by the casting process, the Secondary Dendrite Arm Spacing (SDAS) in the micro-texture structure parameters is set as a decision attribute D ═ SDAS, and a decision table is constructed;
s32 performs attribute reduction on the decision table, which may include the following steps:
processing abnormal data points, and removing coarse data;
discretizing the normalized data set;
establishing a discretized decision table;
the decision table after discretization is shown in table 2.
TABLE 2 decision table of micro-organization structure by discretization
Figure BDA0003103843180000052
Fourthly, attribute reduction is carried out by adopting a rough set-based method, and the importance of the condition attribute is calculated to obtain a reduction subset of the test data;
s41, defining the casting process parameters, the heat treatment parameters and the micro-structure parameter decision table as a quadruple:
DT＝(U，C∪D，V，f)
wherein:
u: a domain of discourse;
c, U.D: c is a condition attribute set, and D is a decision attribute set;
v: v is the value range of the attribute;
f: an information function representing a decision table.
S42, calculating the importance of the condition attribute, wherein the importance of the condition attribute is the quantitative representation of each influence factor of the SDAS.
Given a decision table DT ═ (U, C ═ D, V, f), there is a set of attributes
Figure BDA0003103843180000061
The existence of a epsilon C-B defines:
Figure BDA0003103843180000062
the importance of the condition attribute a to the condition attribute set B relative to the decision attribute D is shown in fig. 3, in which the importance of a2 is the largest and is 0.822, and the importance of a5 is the smallest and is 0.222.
S43, using genetic algorithm to reduce attribute of decision table, taking out redundant attribute, obtaining reduced RED of conditional attribute C relative to decision attribute DC(D) { { a1}, { a2, a5}, { a3, a5} }. Where the location parameters do not participate in the reduction.
Step five, obtaining a reduction subset of a cast aluminum alloy cylinder cover micro-structure decision table through rough set attribute reduction, using the reduction subset as an input layer of a BP neural network, and establishing a BP neural network prediction model;
s51, adopting a three-layer BP neural network, selecting condition attributes as the number of input nodes N and the number of output nodes M as the number of decision attributes according to the reduction subset result;
s52, taking the condition attribute as the input of the neural network, inputting four parameters of position parameters x, y, z and solidification rate, and taking two micro-tissue structure parameters of porosity and decision attribute secondary dendrite arm spacing as the output target of the neural network;
s53, training the neural network model by adopting an LM algorithm, training all samples in an abbreviated subset, and obtaining a predicted value meeting the error requirement;
provided with a hidden layerThe number of nodes is q, and the weight of the input layer and the hidden layer is omegamiThe threshold is bmThe weight of the hidden layer and the output layer is omegaijThe threshold is bi，f1Selecting log-Sigmoid, f for transfer function of hidden layer2Selecting a linear transfer function purelin for an output layer, wherein i is 1, 2 and …; selecting a train function from the trainlm;
transfer function log-Sigmoid:
Figure BDA0003103843180000063
the output of the input layer is equal to the input signal of the entire network signal:
Figure BDA0003103843180000064
where x (n) is the input signal, n is 1, 2, …,
Figure BDA0003103843180000065
n, m is 1, 2, …;
the inputs to the hidden layer node are:
Figure BDA0003103843180000071
wherein
Figure BDA0003103843180000072
An input that is a hidden layer;
the output of the hidden layer node is:
Figure BDA0003103843180000073
wherein
Figure BDA0003103843180000074
Is the output of the hidden layer;
the inputs to the output layer nodes are:
Figure BDA0003103843180000075
wherein
Figure BDA0003103843180000076
Is an input to the output layer;
the output of the output layer node is:
Figure BDA0003103843180000077
wherein
Figure BDA0003103843180000078
Is the output of the output layer.
The method for correcting the weight and the threshold value between layers is as follows:
weight change of the hidden layer:
Δωmi＝ηδmixi
where η is the learning rate.
Weight change of the output layer:
Figure BDA0003103843180000079
there is an output error E when the implemented output does not coincide with the desired output, as follows:
Figure BDA00031038431800000710
wherein d isj(n) is the desired output of the network. And adjusting the weight value and the threshold value to reduce the value of the output error E to a desired range. The method for determining the number q of hidden layer nodes is as follows:
empirical formula using hidden layer node number
Figure BDA00031038431800000711
N and M are the numbers of neurons of the input layer and the output layer respectively, 4 and 2 are taken, and a is a constant between [0 and 10 ]. And rounding the calculated q value, so that the number of the neurons of the hidden layer is 6. The BP neural network prediction model diagram is shown in FIG. 2, 4 nodes of coordinate point x coordinate, y coordinate value, z coordinate value and coagulation rate are used as an input layer of the BP neural network, 2 nodes of SDAS and porosity are used as an output layer of the BP neural network, and a hidden layer is 6 neurons.
And step six, verifying the effectiveness of the neural network prediction model, and taking a part of test data as a test set to verify the correctness of the trained BP neural network. After multiple times of training, the relative error of the predicted value reaches within 15 percent, and the predicted value is within an acceptable range, and the neural network training is determined to be effective.
And seventhly, when the prediction error of the neural network prediction model reaches an acceptable range, storing the BP neural network, and then adopting the BP neural network model to perform subsequent micro-tissue morphology prediction.
According to the weight and threshold calculation formula and the output error calculation formula in step S5, the value of the output error E is reduced to a desired range by adjusting the weight and threshold. The training process of the BP neural network is to continuously correct the weight and the threshold value, so that the output error E is as small as possible, and the set precision standard is reached.
As can be seen from the comparison of the BP neural network prediction results of porosity and secondary dendrite arm spacing with the test results in FIG. 4, the predicted value is close to the test value, the prediction error of porosity is 22.75% at most, 0.4% at least, and the average error is 10.85%; the maximum prediction error of the secondary dendrite arm spacing is 4.25%, the minimum prediction error is 0%, and the average error is 2.53%.
According to the method, a rough set theory is utilized to preprocess test data of the cast aluminum alloy cylinder cover, discretization and attribute reduction are carried out on a decision table of the test data of the cast aluminum alloy cylinder cover, and a reduction sample space is obtained after irrelevant attributes are removed, so that input variables are simplified; and taking the reduced sample space as the input and output of the BP neural network to obtain a network training model, and then carrying out BP neural network model training. After repeated learning and training, the training is stopped after the minimum error is obtained. And finally obtaining the prediction result of the micro-tissue structure.
The above detailed description is intended to illustrate the objects, aspects and advantages of the present invention, and it should be understood that the above detailed description is only exemplary of the present invention and is not intended to limit the scope of the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (1)

1. A casting cylinder cover micro-structure prediction method based on a rough set and a neural network is characterized in that: the method comprises the following steps:
acquiring a casting and heat treatment process parameter database of a cast aluminum alloy cylinder cover, a micro-structure morphology database of a cylinder cover material and position parameters of data acquisition points;
the casting process parameters comprise: the solidification rate, the casting temperature, the casting mold temperature, the mold filling time, the mold filling pressure, the mold filling speed, the crust forming time, the pressurization pressure, the pressurization speed, the pressure maintaining time and the pressure relief time;
the heat treatment process parameters comprise: solution treatment temperature, heat preservation time, cooling speed, aging temperature and aging time;
the microtexture topographic data of the cylinder head material comprises: the method comprises the following steps of (1) grain size, secondary dendrite arm spacing, two-dimensional porosity, metallographic pore area, maximum Feret size of metallographic pore, area mean value of eutectic silicon, length-width ratio mean value of eutectic silicon, maximum Feret size mean value of eutectic silicon and roundness mean value of eutectic silicon particles;
step two, carrying out normalization processing on all the data acquired in the step one;
the data normalization processing formula is as follows:
Figure FDA0003103843170000011
wherein X represents experimental data before normalization, X*Denotes the normalized experimental data, XmaxMaximum value of experimental data, XminIs the minimum value of experimental data;
step three, discretizing the test data subjected to the normalization processing in the step two, then coding the data set, and obtaining a discretized decision table after coding;
step four, calculating importance of the condition attributes in the decision table obtained in the step three, and then carrying out data reduction according to the importance to obtain a reduction subset of the test data;
s41, defining the casting and heat treatment process parameters, the micro-structure topography data and the position parameter decision table as a quadruple:
DT＝(U，C∪D，V，f)
wherein: u: a domain of discourse;
c, U.D: c is a condition attribute set, and D is a decision attribute set;
v: v is the value range of the attribute;
f: an information function of the decision table;
s42, giving a decision table DT ═ (U, C ═ D, V, f), there is a set of attributes
Figure FDA0003103843170000012
The existence of a epsilon C-B defines:
Figure FDA0003103843170000013
the importance of the condition attribute a to the condition attribute set B relative to the decision attribute set D;
s43, performing attribute reduction on the decision table by using a genetic algorithm to obtain an attribute reduction subset; wherein the location parameters do not participate in the reduction;
step five, obtaining a reduction subset through rough set attribute reduction, using the reduction subset as an input layer of the BP neural network, and selecting a training function and a transfer function to establish a BP neural network prediction model;
sixthly, verifying the effectiveness of the BP neural network prediction model;
and seventhly, if the neural network prediction model is verified to be effective, adopting the BP neural network model to predict the microstructure morphology.
CN202110635459.9A 2021-06-07 2021-06-07 Casting cylinder cover micro-structure prediction method based on rough set and neural network Active CN113240096B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110635459.9A CN113240096B (en) 2021-06-07 2021-06-07 Casting cylinder cover micro-structure prediction method based on rough set and neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110635459.9A CN113240096B (en) 2021-06-07 2021-06-07 Casting cylinder cover micro-structure prediction method based on rough set and neural network

Publications (2)

Publication Number Publication Date
CN113240096A true CN113240096A (en) 2021-08-10
CN113240096B CN113240096B (en) 2022-03-11

Family

ID=77137285

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110635459.9A Active CN113240096B (en) 2021-06-07 2021-06-07 Casting cylinder cover micro-structure prediction method based on rough set and neural network

Country Status (1)

Country Link
CN (1) CN113240096B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111832610A (en) * 2020-06-01 2020-10-27 成都飞机工业（集团）有限责任公司 3D printing organization prediction method, system, medium and terminal equipment

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2010149208A (en) * 2008-12-24 2010-07-08 Toyota Motor Corp Method of treating surface of porous flat plate body, porous flat plate, and fuel cell
CN108319223A (en) * 2018-02-06 2018-07-24 合肥工业大学 A kind of thread turning process parameter optimizing method of Oriented Green manufacture
CN109472397A (en) * 2018-10-19 2019-03-15 东华大学 Polymerization technique parameter adjusting method based on viscosity change
CN109934421A (en) * 2019-04-24 2019-06-25 中南大学 A kind of blast furnace molten iron silicon content prediction and compensation method towards the fluctuation working of a furnace
CN111177976A (en) * 2019-12-25 2020-05-19 广东省焊接技术研究所（广东省中乌研究院） Arc welding seam forming accurate prediction method based on deep learning
CN111652402A (en) * 2019-03-04 2020-09-11 湖南师范大学 Optical fiber preform deposition process intelligent optimization method based on big data analysis

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2010149208A (en) * 2008-12-24 2010-07-08 Toyota Motor Corp Method of treating surface of porous flat plate body, porous flat plate, and fuel cell
CN108319223A (en) * 2018-02-06 2018-07-24 合肥工业大学 A kind of thread turning process parameter optimizing method of Oriented Green manufacture
CN109472397A (en) * 2018-10-19 2019-03-15 东华大学 Polymerization technique parameter adjusting method based on viscosity change
CN111652402A (en) * 2019-03-04 2020-09-11 湖南师范大学 Optical fiber preform deposition process intelligent optimization method based on big data analysis
CN109934421A (en) * 2019-04-24 2019-06-25 中南大学 A kind of blast furnace molten iron silicon content prediction and compensation method towards the fluctuation working of a furnace
CN111177976A (en) * 2019-12-25 2020-05-19 广东省焊接技术研究所（广东省中乌研究院） Arc welding seam forming accurate prediction method based on deep learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李海娜: "基于PSO-BP神经网络的铝带坯晶粒度软测量建模及优化", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111832610A (en) * 2020-06-01 2020-10-27 成都飞机工业（集团）有限责任公司 3D printing organization prediction method, system, medium and terminal equipment

Also Published As

Publication number Publication date
CN113240096B (en) 2022-03-11

Similar Documents

Publication Publication Date Title
JP4681426B2 (en) Apparatus and method for analyzing relation between operation and quality in manufacturing process, computer program, and computer-readable recording medium
CN112435363B (en) Cutter wear state real-time monitoring method
CN107273924B (en) Multi-data fusion power plant fault diagnosis method based on fuzzy clustering analysis
CN112838946B (en) Method for constructing intelligent sensing and early warning model based on communication network faults
CN114678080B (en) Converter end point phosphorus content prediction model, construction method and phosphorus content prediction method
CN111179592B (en) Urban traffic prediction method and system based on spatio-temporal data flow fusion analysis
CN108805142A (en) A kind of crime high-risk personnel analysis method and system
CN113240096B (en) Casting cylinder cover micro-structure prediction method based on rough set and neural network
CN108256307A (en) A kind of mixing enhancing intelligent cognition method of intelligent business Sojourn house car
CN113283153A (en) Method for predicting service life of ship vibration equipment under variable working conditions
Wu et al. Hot-rolled steel strip surface inspection based on transfer learning model
CN111753751A (en) Fan fault intelligent diagnosis method for improving firework algorithm
CN115019510A (en) Traffic data restoration method based on dynamic self-adaptive generation countermeasure network
JP4653547B2 (en) Apparatus and method for analyzing relation between operation and quality in manufacturing process, computer program, and computer-readable recording medium
CN111723909A (en) Optimization method and system of fuzzy neural network model
CN111651929A (en) Multi-objective optimization method based on fusion of Dynaform and intelligent algorithm
Zhang Application of artificial intelligence recognition technology in digital image processing
Li et al. Prior mask R-CNN based on graph cuts loss and size input for precipitation measurement
CN115147341A (en) Fabric surface defect classification depth network method based on biological vision inspiration
CN113240095B (en) Casting cylinder cover mechanical property prediction method based on rough set and neural network
CN111258996B (en) Product quality multi-source depth fusion forecasting method for industrial big data
CN112967267B (en) Laser directional energy deposition sputtering counting method of full convolution neural network
CN114357878A (en) Method for predicting water surface load of typical section of amphibious aircraft
Yao et al. Multi-step-ahead tool state monitoring using clustering feature-based recurrent fuzzy neural networks
CN114117684A (en) Selective laser melting forming 316L stainless steel abrasion prediction method based on machine learning

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