CN113176761A - Machine learning-based multi-feature thin plate part quality prediction and process parameter optimization - Google Patents

Machine learning-based multi-feature thin plate part quality prediction and process parameter optimization Download PDF

Info

Publication number
CN113176761A
CN113176761A CN202110469230.2A CN202110469230A CN113176761A CN 113176761 A CN113176761 A CN 113176761A CN 202110469230 A CN202110469230 A CN 202110469230A CN 113176761 A CN113176761 A CN 113176761A
Authority
CN
China
Prior art keywords
data
quality
processing
machining
population
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
CN202110469230.2A
Other languages
Chinese (zh)
Other versions
CN113176761B (en
Inventor
王佩
常建涛
孔宪光
卜凡辉
高晓旭
刘德坤
张安集
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.)
Xidian University
Original Assignee
Xidian University
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 Xidian University filed Critical Xidian University
Priority to CN202110469230.2A priority Critical patent/CN113176761B/en
Publication of CN113176761A publication Critical patent/CN113176761A/en
Application granted granted Critical
Publication of CN113176761B publication Critical patent/CN113176761B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
    • G05B19/408—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 data handling or data format, e.g. reading, buffering or conversion of data
    • G05B19/4086—Coordinate conversions; Other special calculations
    • 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
    • G05B2219/35—Nc in input of data, input till input file format
    • G05B2219/35356—Data handling
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Manufacturing & Machinery (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Numerical Control (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a multi-quality target prediction and technological parameter optimization recommendation method for numerical control machining of a multi-machining-characteristic thin plate part, which solves the problems of multi-quality target prediction and technological parameter optimization of part machining characteristics and comprises the following implementation steps: collecting and sorting a data set; preprocessing the sorted data; sorting the feature importance of the data; dividing a correlation analysis data set into a training set and a test set; constructing a multi-target prediction model by using a machine learning algorithm; and (5) constructing a technological parameter optimization model by using a genetic algorithm to complete the technological parameter optimization of the machined part. The invention uses machine learning method to predict the processing characteristics with multi-quality targets, and uses improved genetic algorithm to optimize the technological parameters, so as to obtain the optimal technological parameters. The whole scheme is rigorous and complete, high in prediction precision and good in parameter optimization effect, is used for adaptive recommendation of process parameters and cutter parameters, and effectively improves the machining quality.

Description

Machine learning-based multi-feature thin plate part quality prediction and process parameter optimization
The technical field is as follows:
the invention belongs to the technical field of numerical control machining automatic control, relates to quality prediction and parameter optimization of machined parts, in particular to quality prediction and parameter optimization of a flat plate slot antenna, and particularly relates to a multi-quality target prediction and technological parameter optimization recommendation method for numerical control machining of a multi-machining-characteristic thin plate part, which is applied to adaptive control analysis of technological parameters and cutter parameters of the part in the numerical control machining process.
Background art:
numerical control milling is an important basic technology in intelligent manufacturing, and is widely applied to industries such as automobile manufacturing, ship engineering, aircraft manufacturing and the like, the traditional technological parameter selection usually depends on manual experience, the numerical control machine realizes the automation of the milling process, and along with the continuous emergence of new products, the structure of the product is also increasingly complex and the size of the product is also greatly changed in order to meet new requirements, so that the problem that the new product is difficult to process in the traditional numerical control processing is caused, in the existing numerical control machine, the technological parameters of processed parts cannot be self-adaptively adjusted according to the new product and the requirements, and the technological parameters set by depending on experience cannot be generally suitable for each processing; however, at present when big data is manufactured and developed vigorously, data value mining is more and more important for numerical control machining, and adaptive recommended process parameters not only can improve product quality, but also have important significance for comprehensive benefits of the whole machining system, so that research on a process parameter optimization method becomes a hotspot for research in the field of numerical control machining in recent years.
Zhang Zida et al, in the published paper "optimization of milling process parameters of N87 alloy steel under minimal lubrication conditions" (combined machine tool and automatic processing technology, 6 months 2020), performed three-level orthogonal tests of three factors, oil mass, gas flow and target distance, using a certain turbine blade material N87 as a research object, analyzed cutting effect and influence law under different parameter combinations, and obtained the best parameter combination under a certain working condition. The method adopts an orthogonal test, and obtains the optimal parameter combination through the rule and significance of the influence of three parameters of oil mass, air flow and target distance of micro-lubrication on the abrasion and surface roughness of the cutter under the same working condition. The method adopts the micro-lubrication after parameter optimization to compare with dry cutting, normal temperature cold air and pouring type cooling, and verifies the improvement effect of the parameter optimization on the performance of the micro-lubrication, but the method still has the defects that an orthogonal experiment mode is adopted, a plurality of groups of parameter values are given to carry out an experiment, a group of parameters with the best experiment effect are found, in the actual processing process, the variation range of the parameters is very large, more groups of experiments are needed by using the method, the cost is increased, the global search of a plurality of parameter values in the given range cannot be realized, and the obtained parameter values cannot reach the optimal state.
The article "optimization of numerical control milling processing parameters of injection mold" (equipment manufacturing technology, 9 months 2020) published by Mawei Dong proposes to optimize and research the milling process parameters of a mold forming part, establishes a multi-objective optimization model taking the rotation speed of a main shaft, the cutting depth, the feed amount of each tooth and the milling width as design variables and taking the maximum production efficiency, the minimum production cost and the surface roughness as sub-objectives, gives corresponding weights according to the importance degree of each optimized sub-objective, adopts a linear combination method to form a uniform objective function, adopts a particle swarm method for optimization and calculation, and finally obtains the optimized parameter combination. The algorithm establishes a targeted mathematical model, and adopts a reasonable optimization algorithm to realize the optimal comprehensive benefit on the premise of considering both the production cost and the production time, but when the method constructs a multi-target function, each target is expressed by using an empirical formula, and then the linear weighting is carried out to combine the targets into the multi-target function.
For the existing numerical control machining-oriented technological parameter optimization method, most of the machining special evidence objective functions are constructed by using the traditional empirical formula, and in the complex machining environment and multi-parameter optimization, the empirical formula is more complicated to express, and the precision is low; most of the methods for optimizing the process parameters are methods using orthogonal experiments, but the methods using the orthogonal experiments often need more groups of experiments for the case of large parameter variation range in actual processing, which not only increases the cost, but also increases the precision.
The invention content is as follows:
aiming at the defects of the prior art, the invention provides a multi-quality target prediction and process parameter optimization recommendation method for numerical control machining of a multi-machining-characteristic thin plate part, which has high prediction precision and process parameter optimization reliability.
The invention relates to a multi-quality target prediction and technological parameter optimization recommendation method for multi-machining characteristic sheet part numerical control machining, which takes a typical plate crack antenna of a machined part as a machining object and is characterized by comprising the following steps of:
step (1) collecting and arranging a processing technology and a working medium adding data set: collecting process parameter data of a planar slot antenna of a machined part by using a numerical control machine tool, collecting machining quality data of the planar slot antenna by using a three-coordinate measuring instrument, then extracting the data of three machining characteristics, namely a rectangle, a slot and a round hole of the planar slot antenna, corresponding to the rotating speed, the cutting depth and the feeding of the process parameters one by one, and sorting the data into a planar slot antenna machining quality prediction and process parameter optimization data set classified according to the machining characteristics;
step (2) data preprocessing: detecting abnormal values contained in three processing characteristic data, namely a rectangle, a crack and a round hole of the flat plate crack antenna processing quality prediction and process parameter optimization data set by using a box separation method, and replacing the abnormal values with the mean value of corresponding processing characteristics to obtain a preprocessed data set;
and (3) sorting the importance of the multidimensional features of the data: aiming at the preprocessed data set, respectively using a recursive feature elimination algorithm, a random forest algorithm and an XGboost algorithm to establish a regression model of the process parameters and the processing features of the flat plate slot antenna, calculating the importance of the rotating speed, the cutting depth and the feeding of the process parameters relative to the processing features, sequencing the importance of the process parameters, and obtaining the sequencing result of the importance of the process parameters of the processed parts by taking the intersection of the sequencing results of the algorithms;
and (4) carrying out correlation analysis on data characteristics and deleting redundant characteristics: performing relevance analysis on the rotating speed, the cutting depth and the feeding of the process parameters by using a relevance analysis method aiming at the preprocessed data set to obtain the relevant coefficients among the process parameters, judging the relevant coefficients among the process parameters with higher rank by combining the process parameter importance ranking results of the processed parts, deleting one of the two process parameters with the highest relevant coefficients, and reserving the other one of the two process parameters to respectively obtain the relevance analysis data sets of the processed characteristic rectangles, the cracks and the round holes of the processed parts;
step (5) respectively carrying out training set and test set division on the processing characteristic rectangle, crack and round hole correlation analysis data set: respectively dividing the respective correlation analysis data sets of the processing characteristic rectangles, the cracks and the round holes of the processed parts according to the proportion of 7:3, and respectively taking 70% of the respective data sets of the processing characteristic rectangles, the cracks and the round holes as training sets corresponding to the processing characteristics and taking the rest 30% as test sets corresponding to the processing characteristics;
and (6) constructing a multi-quality target prediction model by using a machine learning algorithm: aiming at the respective correlation analysis data sets of the processing characteristic rectangle, the crack and the round hole of the processing part, respectively training and learning the respective training sets of the processing characteristic rectangle, the crack and the round hole by using an XGboost algorithm, respectively constructing respective quality prediction models of the processing characteristic rectangle, the crack and the round hole, respectively using the process parameter rotating speed, the cutting depth and the feeding of the processing part in the prediction models as a group of input characteristics, respectively using the respective quality data of the processing characteristic rectangle, the crack and the round hole as output, adjusting the respective prediction models of the processing characteristic rectangle, the crack and the round hole one by one according to a 5-fold cross verification result, and finally evaluating the prediction models by using the respective test sets of the processing characteristic rectangle, the crack and the round hole to complete the construction of the multi-quality target prediction models of the rectangle, the crack and the round hole;
and (7) constructing a multi-process parameter synchronous optimization recommendation model based on the processing characteristic multi-quality target prediction model weighted improved genetic algorithm, and completing multi-process parameter synchronous optimization recommendation of the processed parts: the method comprises the steps of weighting and summing a multi-quality target prediction model of a rectangle, a crack and a round hole to serve as a part quality target function, using process parameter rotating speed, cutting depth and feeding as input variables, using a genetic algorithm to construct a multi-process parameter synchronous optimization recommendation model taking part quality as a target, optimizing the rotating speed, cutting depth and feeding of process parameters through the model, obtaining an optimal process parameter combination for recommendation, and completing multi-process parameter synchronous optimization recommendation of a machined part.
The method solves the technical problems of multi-quality target prediction and multi-process parameter synchronous optimization recommendation of the machining characteristics of the planar slot antenna in numerical control machining.
Compared with the prior art, the invention has the technical advantages that:
the multi-quality target prediction model has high precision: excavating parameters which have important influence on the processing characteristics through the multi-dimensional characteristic importance degree sequencing of the data and the correlation analysis of the data; the method has the advantages that the process parameter rotating speed, the cutting depth and the feeding are used as input, the XGboost algorithm is used for respectively establishing quality prediction models of the processing characteristic rectangle, the crack and the round hole, the condition that the complex and complicated empirical formula is used for expressing the relation between the process parameter and the processing characteristic is avoided, the prediction model is high in precision, and an accurate objective function is provided for a process parameter optimization model; the prediction model is used for predicting the processing quality of the rectangle, the crack and the round hole, so that early warning can be timely made on data with abnormal quality, and the processing quality is effectively improved;
the reliability of the process parameter optimization model is high: the method comprises the steps of weighting and summing a high-precision multi-quality target prediction model of a processing characteristic rectangle, a crack and a circular hole to serve as a part quality target function, using the rotating speed, the cutting depth and the feeding of process parameters as design variables, using the part quality target function as a target value, establishing a multi-process parameter synchronous optimization recommendation model based on a genetic algorithm of weighted improvement of the multi-quality target prediction model of the processing characteristic, avoiding cost rise caused by using more groups of orthogonal tests, recommending optimal process parameter values in an actual numerical control processing process, using the weighted summation of the high-precision multi-quality target prediction model as a target function to improve the genetic algorithm to enable the reliability of process parameter optimization to be higher, and improving the processing quality of the process parameter optimization model and the self-adaptive capacity of a cutter and equipment.
Description of the drawings:
FIG. 1 is a general flow diagram of the present invention;
FIG. 2 is a flow chart of the preparation and construction of a multi-quality target predictive model of the processing features of the present invention;
FIG. 3 is a flow chart of the improved genetic algorithm based multi-process parameter optimization model construction of the present invention;
FIG. 4 is a graph of the true value versus the predicted value of the rectangular error in the present invention;
FIG. 5 is a graph of the true value versus the predicted value of the round hole error in the present invention;
FIG. 6 is a graph of the true and predicted values of fracture error in the present invention.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawings.
Example 1
In the existing numerical control machining, the quality of a workpiece machined by a numerical control machine tool is difficult to predict, and the setting of the technological parameters of the numerical control machine tool is often dependent on the traditional experience. The flat plate slot antenna is a typical thin plate part, is used as key equipment for communication, broadcasting, radar, guidance, early warning, radar and missile resisting antennas, has excellent electrical properties such as high radiation efficiency, high power, light structure, small thickness, light weight and the like, is particularly suitable for airborne equipment requiring light weight and small size, has extremely high processing precision requirement and complex processing process, only depends on the traditional experience to guide the processing process, can generate abnormal phenomena of low processing precision and unstable processing process, seriously influences the processing quality and causes material waste; based on the problems of difficulty in processing quality prediction and inaccurate process parameter setting in the existing flat plate slot antenna processing process, the invention provides a multi-quality target prediction and process parameter optimization recommendation method for multi-processing characteristic sheet part numerical control processing through research and experiments.
The invention relates to a multi-quality target prediction and process parameter optimization recommendation method for numerical control machining of a sheet part with multiple machining characteristics, which takes a machined part flat plate slot antenna as a machining object, wherein the flat plate slot antenna consists of thin-wall cavities with different slotting modes, the flat plate slot antenna is mainly formed by connecting multiple layers of a coupling layer, a feed layer, a radiation layer and the like, and the structural design of different layers has certain difference in order to meet the requirements of actual electrical performance and mechanical performance, but three machining characteristics, specifically a rectangle, a slot and a round hole, are generally arranged on a rectangular flat plate-shaped workpiece to serve as a high-frequency element sample of the flat plate slot antenna, the following specific flat plate slot antenna is taken as an example for explanation, the machined workpiece is a rectangular machined workpiece, and the 3 machining characteristics, namely the rectangle, the slot and the round hole, on the flat plate, Crack and round hole, the direction of following the flat board of rectangle is long on the dull and stereotyped surface of flat board in proper order respectively independent be equipped with the closed groove of 4 rectangles forms, the crack of 1 rectangle form and the through-hole of 4 the same circular forms, wherein the centre of rectangular channel is link up, the length direction of crack is parallel with the length direction of flat board, the through-hole of 4 circular forms is put in a rectangular channel, when numerical control machine tool adds, mill processing according to the order of rectangle, crack and round hole, see figure 1, including following step:
step (1) collecting and arranging a processing technology and a working medium adding data set: the technical parameter data of the flat plate slot antenna of the machined part is collected through a numerical control machine tool, no matter the flat plate slot antenna is formed into different layers, the structure in each layer has no three machining characteristics of rectangle, slot and round hole, and the technical parameters also comprise rotating speed, cutting depth and feeding. The processing quality data of the flat plate slot antenna is collected through a three-coordinate measuring instrument, then the data of three processing characteristics, namely the rectangle, the slot and the round hole of the flat plate slot antenna, are extracted and are respectively corresponding to the process parameter rotating speed, the cutting depth and the feeding one by one, namely the processing quality data of the processing characteristic rectangle is corresponding to the process parameter rotating speed, the cutting depth and the feeding, the slot data is corresponding to the process parameter rotating speed, the cutting depth and the feeding, the data of the round hole is corresponding to the process parameter rotating speed, the cutting depth and the feeding, and all the data are arranged into a flat plate slot antenna processing quality prediction and process parameter optimization data set classified according to the processing characteristics. The processing quality data comprises data of processing characteristic rectangles, cracks and round holes, the processing quality data of the processing characteristic rectangles refer to processing errors of the length and the width of the rectangular closed groove, the data of the cracks refer to processing errors of the width of the cracks, and the data of the round holes refer to processing errors of the diameter of the round holes.
And (2) replacing an abnormal value with the mean value to perform data preprocessing: abnormal values contained in three processing characteristic data, namely rectangle, crack and round hole, of the flat plate crack antenna processing quality prediction and process parameter optimization data set are detected by using a box separation method, and the detected abnormal values are replaced by mean values of corresponding processing characteristics to obtain a preprocessed data set.
And (3) sorting the importance of the multidimensional features of the data: aiming at the data set preprocessed in the step (2), respectively using a recursive feature elimination algorithm, a random forest algorithm and an XGboost algorithm to establish regression models of the process parameters and the processing features of the flat plate slot antenna, obtaining the regression models corresponding to the three algorithms, respectively using the respective regression models established by the three algorithms to calculate the importance of the rotation speed, the cutting depth and the feeding of the process parameters relative to the rectangle, the slot and the round hole of the processing features, sequencing the importance of the process parameters, obtaining the sequencing results of the importance of the process parameters of each processing feature under different algorithms, taking intersection of the sequencing results of the importance of the same processing feature under different algorithms, obtaining the sequencing results of the importance of each processing feature, and taking intersection of the sequencing results of the importance of all the processing features to obtain the sequencing results of the importance of the process parameters of the processed parts.
And (4) carrying out correlation analysis on data characteristics and deleting redundant characteristics: and (3) aiming at the preprocessed data set obtained in the step (2), performing correlation analysis on the rotating speed, the cutting depth and the feeding of the process parameters of the flat plate slot antenna by using a correlation analysis method, calculating to obtain correlation coefficients among the process parameters, judging the correlation coefficients among the first three process parameters in the importance ranking by combining the correlation coefficients of the process parameters with the process parameter importance ranking result of the processed part, judging whether linear correlation exists between the two process parameters with the highest correlation coefficients, and keeping one of the two linearly correlated characteristics, or keeping the two process parameters at the same time, and finally obtaining respective correlation analysis data sets of the processed characteristic rectangle, the slot and the round hole of the processed part respectively.
And (5) respectively carrying out training set and test set division on the correlation analysis data set: the method comprises the steps of dividing a training set and a test set of correlation analysis data sets of a processing characteristic rectangle, a crack and a round hole of the flat plate crack antenna according to the proportion of 7:3, wherein the correlation analysis data sets of the processing characteristic rectangle, the crack and the round hole of a processing part have 70% of the respective training sets and 30% of the respective test sets, namely 70% of the respective data sets of the processing characteristic rectangle, the crack and the round hole are used as the training sets corresponding to processing characteristics, and the rest 30% of the respective data sets of the processing characteristic rectangle, the crack and the round hole are used as the test sets corresponding to the processing characteristics. The division ratio of the training set and the test set needs to be adjusted according to the size of actual data volume, the ratio of the training set to the test set can be set to be 7:3 or 8:2 generally, the ratio of the training set to the test set can be set to be 8:2 when the data volume is small, and due to the fact that the data volume of the machining characteristics of the flat plate slot antenna is moderate, the 7:3 division ratio is suitable, and the fitting capability of the prediction model is improved.
And (6) constructing a multi-quality target prediction model by using a machine learning algorithm: aiming at the respective correlation analysis data sets of the processing characteristic rectangles, the cracks and the round holes of the processing parts, an XGboost algorithm is used for respectively training the respective training sets of the processing characteristic rectangles, the cracks and the round holes, respective quality prediction models of the processing characteristic rectangles, the cracks and the round holes are respectively constructed, the technological parameters of the processing parts, such as rotating speed, cutting depth and feeding, in the prediction models are used as a group of prediction models for inputting, the respective quality data of the processing characteristic rectangles, the cracks and the round holes are used as output, the max _ depth parameter in the XGboost algorithm needs to be adjusted in the construction process of the prediction models to enable the model accuracy to be optimal, the respective prediction models of the processing characteristic rectangles, the cracks and the round holes are adjusted and optimized one by one according to 5-fold cross verification results, the respective test sets of the processing characteristic rectangles, the cracks and the round holes are used for evaluating the accuracy of the prediction models, and the rectangles, the cracks and the round holes are finished, And constructing a multi-quality target prediction model of the crack and the round hole. The multi-quality target in the invention refers to the processing quality of a rectangle, a crack and a round hole, and the establishment of the multi-quality target prediction model is to establish a respective quality prediction model for each processing characteristic by using an XGboost algorithm.
And (7) constructing a multi-process parameter synchronous optimization recommendation model based on the processing characteristic multi-quality target prediction model weighted improved genetic algorithm, and completing multi-process parameter synchronous optimization recommendation of the processed parts: the method comprises the steps of weighting and summing a multi-quality target prediction model of a rectangle, a crack and a round hole to serve as a part quality target function, using process parameter rotating speed, cutting depth and feeding as input variables, using a genetic algorithm to construct a multi-process parameter synchronous optimization recommendation model taking part quality as a target, optimizing the rotating speed, cutting depth and feeding of process parameters through the model to obtain an optimal process parameter combination, and completing multi-process parameter synchronous optimization recommendation of a machined part.
In the existing research method, the selection of process parameters and cutter parameters in the numerical control machining process cannot realize self-adaptive control analysis. The invention provides a multi-quality target prediction and process parameter optimization recommendation method oriented to multi-processing characteristic sheet part numerical control processing in order to change the current situation that the process parameter optimization is carried out by using the traditional empirical formula and the orthogonal experiment method in the prior art but the optimal result is not obtained actually in the process of the numerical control processing, relieves the problems of low processing quality prediction precision and inaccurate process parameter optimization, and effectively improves the prediction precision of the processing quality and the reliability of the process parameter optimization. Selecting the characteristics of data to obtain a correlation analysis data set of each processing characteristic, dividing the correlation analysis data set of each processing characteristic into 70% of training sets and 30% of testing sets corresponding to the processing characteristics, training each processing characteristic training set by adopting a machine learning algorithm XGboost, constructing quality prediction models of processing characteristic rectangles, cracks and round holes respectively, adjusting and optimizing each prediction model, evaluating the precision of each optimized prediction model by using each processing characteristic testing set to obtain a multi-quality target prediction model of the processing characteristic rectangles, cracks and round holes with higher precision, weighting and summing the multi-quality target prediction models of the processing characteristic rectangles, cracks and round holes to obtain a part quality objective function, constructing a multi-process parameter synchronous optimization recommendation model taking the part quality as the target by adopting a genetic algorithm to obtain a process parameter optimization result with high reliability, and completing the synchronous optimization recommendation of the multiple process parameters of the machined part.
Example 2
The method for predicting the multi-quality target and optimizing and recommending the process parameters for the numerical control machining of the multi-machining-characteristic thin plate part is the same as that in the embodiment 1, the machining process and the working medium adding data set are collected and sorted in the step (1), and the process for collecting and sorting the machining process and the working medium adding data set comprises the following steps:
(1.1) acquiring a data set: the cutter parameter information of the numerical control machine tool is an end mill, the diameter is 1, and the clamping length is 17; the processing characteristics are as follows: rectangles, slits and circular holes; the technological parameters are as follows: the numerical control machine tool conducts milling according to the sequence of the rectangle, the crack and the round hole, technological parameter data of the flat plate crack antenna are collected through an MDC module of the numerical control machine tool, machining quality data of the machining characteristic rectangle, the crack and the round hole of the flat plate crack antenna are collected through a three-coordinate measuring instrument, the machining quality data of the machining characteristic rectangle refer to machining errors of the length and the width of a rectangular closed groove, the data of the crack refer to machining errors of the width of the crack, and the data of the round hole refer to machining errors of the diameter of the round hole.
(1.2) sorting the data sets: the method comprises the steps of extracting data of three machining characteristics of a rectangle, a crack and a round hole of the flat plate crack antenna, wherein the data corresponds to process parameters of rotating speed, cutting depth and feeding one by one, enabling the rectangle to correspond to the process parameters of rotating speed, cutting depth and feeding, enabling the crack to correspond to the process parameters of rotating speed, cutting depth and feeding, enabling the round hole to correspond to the process parameters of rotating speed, cutting depth and feeding, and finally forming a flat plate crack antenna machining quality prediction and process parameter optimization data set classified according to the machining characteristics.
In the data acquisition and arrangement, aiming at three processing characteristics of a rectangle, a crack and a round hole on the surface of a flat plate crack antenna of a processing workpiece, a strategy of classifying a processing technology and a working medium adding data set according to the processing characteristics is adopted, the data of the rectangle, the crack and the round hole of the processing characteristics are corresponding to the process parameters of rotating speed, cutting depth and feeding one by one, the importance of the processing characteristics to the whole flat plate crack antenna of the processing workpiece is tightly grasped, a prediction model is respectively constructed for each processing characteristic, the construction complexity of the prediction model is reduced, and optimization preparation is also made for subsequent process parameters.
Example 3
The multi-quality target prediction and process parameter optimization recommendation method for multi-machining-feature sheet part numerical control machining is the same as that in embodiment 1-2, the mean value is used for replacing the abnormal value to perform data preprocessing in step (2), and the data preprocessing process comprises the following steps:
(2.1) statistics calculation for data set: and carrying out statistic calculation on the processing process data of the rectangle, the crack and the round hole of the flat plate crack antenna, wherein the statistic calculation comprises a lower quartile, a median, an upper quartile and a quartile interval.
(2.2) abnormal value judgment of data set: and (3) judging abnormal values of the processing characteristic rectangle, crack and round hole data of the processed part according to the following formula:
value>QU+1.5IQR or value<QL-1.5IQR
here, value indicates an abnormal value in each machining feature data, the abnormal value is data located outside a section between QU +1.5IQR and QL-1.5IQR, QU indicates an upper quartile of each machining feature data, QL indicates a lower quartile of each machining feature data, and IQR indicates a quartile pitch of each machining feature data, that is, IQR is QU-QL.
(2.3) outlier replacement of data set: and replacing the detected abnormal value with the average value of the data of the corresponding processing characteristic rectangle, the crack and the round hole to obtain a preprocessed data set.
The invention uses the box-dividing method to detect the mean value, the three-sigma principle can also be used for detecting the abnormal value, the three-sigma principle is more accurate when the data volume is large and requires that the data distribution is close to the normal distribution, the data set capacity of the flat plate slot antenna is smaller, the box-dividing method is more suitable to select, when the abnormal value is replaced, the mean value is used for replacing the abnormal value, the data set capacity can be ensured to be unchanged, and enough training data is provided for the prediction model.
Example 4
The method for predicting multiple quality targets and optimizing and recommending process parameters for numerical control machining of thin plate parts with multiple machining characteristics is the same as that in the embodiment 1-3, the correlation analysis and the redundant characteristic deletion of the data characteristics are described in the step (4), and the correlation analysis process of the data characteristics comprises the following steps:
(4.1) calculation of correlation coefficient of data characteristics: the relationship between two consecutive variables was analyzed using the pearson correlation coefficient, which was calculated as follows:
Figure BDA0003044715050000101
wherein xi，yiThe ith sample representing the continuity variable x sample set and y sample set,
Figure BDA0003044715050000102
is the mean of a sample set of continuous variables x,
Figure BDA0003044715050000103
is the mean value of a sample set of continuous variable y, r represents the correlation coefficient of the variable x and the variable y, the range of the correlation coefficient r is-1 ≦ r ≦ 1, r>0 is a positive correlation, r<0 is negative correlationThe linear correlation does not exist if r is less than or equal to 0.3, the R is not Y>And 0.8 is highly linear correlation, and a correlation coefficient matrix is constructed by calculating the Pearson correlation coefficient between every two of the process parameters of the rotating speed, the cutting depth and the feeding of the processed part based on the formula.
(4.2) judging the relevance of the data features: judging the correlation coefficient among the first three process parameters in the importance ranking result, judging whether the two process parameters with the highest correlation coefficient have linear correlation, reserving the parameters with the correlation coefficient smaller than 0.3, when the correlation coefficient of the two parameters is smaller than 0.3, the two parameters do not have linear correlation, and reserving all the process parameters according to the fact that the maximum correlation coefficient among all the obtained process parameters does not exceed 0.12.
The method uses the Pearson coefficient to carry out correlation analysis, which is very important for carrying out correlation analysis on the characteristics of the data, if two linear correlation characteristics exist, a method of only reserving one characteristic can be adopted, the problem of overfitting caused by leading the two linear correlation characteristics into a prediction model is avoided, the data is more in line with the data condition which is well processed by a machine learning algorithm, meanwhile, unnecessary calculation processing is avoided, and the training efficiency and precision of the prediction model are improved.
Example 5
The multi-quality target prediction and process parameter optimization recommendation method for multi-machining characteristic sheet part numerical control machining is the same as that in the embodiment 1-4, and the step (6) of constructing the multi-quality target prediction model by using a machine learning algorithm comprises the following steps:
(6.1) selection of a multi-quality target prediction model: the multi-quality target refers to the processing quality of a rectangle, a crack and a round hole, and the multi-quality target prediction model can use a machine to learn a random forest algorithm and an XGboost algorithm in the multi-quality target prediction model; the random forest is an integrated learning algorithm and consists of a plurality of weak learners, strong dependence does not exist among the weak learners, and the training results are voted by simultaneously training the weak learners, so that the performance of the random forest is generally superior to that of a single learner; the XGboost is also an integrated learning algorithm, but each weak learner has a strong dependency relationship, different weights are distributed to each weak learner according to the learning effect of each weak learner, then a plurality of learners are weighted and summed to construct a strong learner for training, and the strong learner can obtain better effect than a single learner; the method comprises the steps of using training sets of machining feature rectangles, cracks and round holes of machined parts to respectively construct prediction models by means of a random forest algorithm and an XGboost algorithm, then respectively inputting test sets of all machining features into the random forest prediction model and the XGboost prediction model, comparing mean square errors and fitting errors output by the two prediction models, and obtaining the XGboost prediction model with higher precision, so that the XGboost algorithm is used for constructing the prediction models.
(6.2) optimizing a multi-quality target prediction model: respectively introducing training sets of processing characteristic rectangles, cracks and round holes of a processed part into an XGboost algorithm for training, setting a traversing list of max _ depth parameters of the XGboost algorithm as [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], traversing each parameter in the list of max _ depth parameters by using a for-loop, constructing an XGboost model corresponding to each max _ depth parameter, evaluating the precision of each quality prediction model by using a mean square error and a fitting error obtained by 5-fold cross verification, comparing the precision of each model, selecting the max _ depth parameter corresponding to the model with the highest precision, realizing the search of the optimal max _ depth parameter, and finishing the tuning of a multi-quality target prediction model.
(6.3) evaluation of the multi-quality target prediction model: respectively leading training sets of processing characteristic rectangles, cracks and round holes of the processed parts into XGboost quality prediction models under respective optimal max _ depth parameters, using mean square error and fitting error to evaluate the model performance to obtain the training error of the models, leading test sets into XGboost prediction models corresponding to the optimal max _ depth parameters, using mean square error and fitting error to evaluate the model performance to obtain the test error of the models, using the test error as the performance evaluation of the multi-quality target prediction models, and finally completing the construction of the multi-quality target prediction models.
The invention establishes a quality prediction model for each processing characteristic, namely respectively constructs a rectangular processing quality prediction model, a crack processing quality prediction model and a round hole processing quality prediction model, the three machining quality prediction models adopt a machine learning algorithm, the XGboost algorithm is often used as a recommendation algorithm of the prediction model in machine learning and is widely applied, the XGboost can deeply mine the relation between the process parameters and the machining characteristics, the value of the machining workpiece data is effectively utilized, the accuracy of the prediction model reaches a higher level, the machining characteristic prediction model with high accuracy can avoid using a complex formula to express the relation between the process parameters and the machining characteristics, the designed process parameters can be imported into the prediction model to predict the data of the machining characteristics of the parts to be machined, and early warning of future abnormal data is realized.
Example 6
The multi-quality target prediction and process parameter optimization recommendation method for multi-machining-feature sheet part numerical control machining is the same as that in the embodiment 1-5, the multi-process parameter synchronous optimization recommendation model is constructed by the processing-feature-based multi-quality target prediction model weighted improved genetic algorithm in the step (7), and multi-process parameter synchronous optimization recommendation of the machined part is completed, wherein the construction of the multi-process parameter synchronous optimization recommendation model comprises the following steps:
(7.1) determining an objective function F (x) to be optimized: multiplying the multi-target prediction models of the machining characteristic rectangles, the cracks and the round holes of the machined part by the same weight coefficient respectively to obtain weighted multi-target prediction models of the rectangles, the cracks and the round holes, and then taking the weighted multi-target prediction models of the rectangles, the cracks and the round holes as a part quality objective function to be optimized, wherein the formula of the part quality objective function is as follows:
F(x1,x2,x3)＝w·F1(x1，x2,x3)+w·F2(x1,x2,x3)+w·F3(x1,x2,x3)
wherein x1,x2,x3Respectively, the process parameters of speed, depth of cut and feed, F1 (x)1,x2,x3)、F2(x1,x2,x3) And F3 (x)1，x2，x3) Prediction models for rectangle, round hole and crack respectively, w is weight coefficient, F (x)1，x2，x3) Is a part quality objective function.
(7.2) determining design variables and ranges of the genetic algorithm: the design variables of the genetic algorithm are the rotating speed, the cutting depth and the feeding of the process parameters to be optimized, the rotating speed of the process parameters to be optimized ranges from [30000(rpm) to 40000(rpm) ], the cutting depth ranges from [0(mm) to 0.3(mm) ], and the feeding range from [2800(mm) to 4000(mm) ].
(7.3) determining parameter settings of the genetic algorithm: determining to use Gray codes for genetic algorithm coding; determining the number of population individuals to be 20; determining the maximum genetic algebra to be 200; determining to adopt random sampling selection as a selection mode; determining to adopt two-point intersection as an intersection mode; determining a fitness calculation mode to be fitness distribution calculation based on grade division; and determining to adopt a binary mutation operator as a mutation mode to complete the construction of the multi-process parameter synchronous optimization recommendation model.
(7.3) initializing population: the technological parameters of rotating speed, cutting depth and feeding are used as a group of technological parameter combinations, a plurality of groups of technological parameters are randomly generated in the range of technological parameters to be optimized, the technological parameters of rotating speed, cutting depth and feeding of one group are a group individual, the plurality of groups of technological parameter combinations are equivalent to a group, and the initialization of the group is completed.
(7.4) calculating the fitness of the initial population: giving a fitness threshold condition f1, calculating the fitness of the initial population individuals according to a part quality objective function of the initial population individuals and a fitness distribution calculation method based on grade division, wherein the fitness calculation of the population individuals can be represented by the following formula:
f＝ranking(F(x1,x2,x3))
wherein f represents the fitness of the population individual, and ranking () represents a fitness distribution calculation method based on grade division;
obtaining the fitness f of the initial population individuals according to a fitness calculation formula of the population individuals0And compared with the set fitness threshold condition f1 if f0>f1, the initial population is the optimal population, otherwise, f0<f1, executing the step (7.5) and carrying out evolution on the initial population.
(7.5) evolving population: for a population to be evolved, a fitness limit value f2 which allows evolution is given, the fitness limit value which allows evolution is smaller than a fitness threshold condition, namely f2< f1, the fitness f of an individual of the population to be evolved is compared with the fitness limit value f2 which allows evolution, the population individual of f > f2 is selected to be used for generating a next generation population, the next generation population is population filial generation, the population individual of which the fitness is larger than f2 has a character which enables a part quality objective function value to reach a high-quality objective function value, then population filial generation individuals are recombined by using crossover and variation operators to obtain an evolved population individual, and the quality objective function corresponding to the evolved population individual is better compared with the previous generation so that the population filial generation inherits the good character of the previous generation population to complete the population evolution.
(7.6) calculating the fitness of the population after evolution: using a given fitness threshold condition f1, using a determined maximum genetic algebra 200, calculating the fitness of the evolved (current population) population individuals according to a fitness calculation formula of the population individuals, judging whether the fitness f of the evolved population individuals is greater than the fitness threshold condition f1, if f is greater than f1, determining that the evolved population is an optimal population, namely a process parameter combination which enables the quality of the parts to be optimal, and otherwise, re-evolving the evolved population, and returning to the step (7.5) to execute the next round of evolution; or directly judging whether the evolution algebra reaches the maximum genetic algebra, if the evolution algebra reaches the maximum genetic algebra, obtaining a process parameter combination which enables the part quality to reach the optimum, and completing the optimization of the process parameters, otherwise, the evolved population needs to be re-evolved, and the next round of evolution is executed by returning to the execution step (7.5) until the individual fitness value of the evolved population is greater than the fitness threshold condition or the evolution algebra reaches the maximum genetic algebra, and finally realizing the optimization of the process parameter rotating speed, the cutting depth and the feeding, obtaining the optimum process parameter combination, and completing the synchronous optimization recommendation of the multiple process parameters of the processed parts.
The invention uses a genetic algorithm based on processing characteristic multi-quality target prediction model weighted improvement to synchronously optimize and recommend the multi-process parameters of the parts to obtain a highly reliable parameter optimization result, the genetic algorithm is one of evolutionary algorithms, an optimal solution is found by simulating the mechanism of selection and inheritance in nature, and the genetic algorithm has three basic genetic operators: the method has the advantages that the problems that general numerical solution is easy to fall into local tiny traps are solved through selection, intersection and variation, global optimal solution search is achieved, the genetic algorithm is used for optimizing the process parameters, the search capability of the genetic algorithm on the optimal solution is also utilized, the population is continuously evolved through the action of a genetic operator, the optimal process parameters are obtained, in addition, the quality objective function obtained through weighted summation of the multi-quality target prediction model with high-precision processing characteristics is more accurate, and the reliability of process parameter optimization is improved.
The invention relates to a multi-quality target prediction and technological parameter optimization recommendation method for multi-machining-characteristic sheet part numerical control machining, which solves the problems of multi-quality target prediction and multi-technological parameter synchronous optimization recommendation of part machining characteristics, and comprises the following implementation steps: collecting and sorting a data set; preprocessing the sorted data; sorting the feature importance of the data; dividing a correlation analysis data set into a training set and a test set; constructing a multi-target prediction model by using a machine learning algorithm; and (5) constructing a technological parameter optimization model by using a genetic algorithm to complete the technological parameter optimization of the machined part. The invention uses machine learning method to predict the processing characteristics with multi-quality targets, and uses improved genetic algorithm to optimize the technological parameters, so as to obtain the optimal technological parameters. The whole scheme is rigorous and complete, high in prediction precision and good in parameter optimization effect, is used for adaptive recommendation of process parameters and cutter parameters, and effectively improves the machining quality.
The invention will be further illustrated by the following detailed example
Example 7
The method for predicting the multi-quality target and optimizing and recommending the process parameters for the numerical control machining of the multi-machining-characteristic sheet part is the same as the embodiment 1-6, and referring to fig. 1, the implementation steps of the invention are as follows:
step one, collecting and arranging a processing technology and a working medium adding data set: the cutter parameter information of the numerical control machine tool is an end mill, the diameter is 1, and the clamping length is 17; the processing characteristics are as follows: rectangles, slits and circular holes; the technological parameters are as follows: the method comprises the steps of collecting technological parameter data of a flat plate slot antenna through a numerical control machine tool, collecting machining quality data of machining characteristics of the flat plate slot antenna through a three-coordinate measuring instrument, then extracting the data of three machining characteristics of a rectangle, a slit and a round hole of the flat plate slot antenna to correspond to the technological parameter rotating speed, the cutting depth and the feeding one by one, enabling the rectangle to correspond to the technological parameter rotating speed, the cutting depth and the feeding, enabling the slit to correspond to the technological parameter rotating speed, the cutting depth and the feeding, enabling the round hole to correspond to the technological parameter rotating speed, the cutting depth and the feeding, and finally arranging the data into a flat plate slot antenna machining quality prediction and technological parameter optimization data set classified according to the machining characteristics. The processing quality data of the processing characteristic rectangle refers to the processing errors of the length and the width of the rectangular closed groove, the processing quality data of the crack refers to the processing errors of the width of the crack, and the processing quality data of the round hole refers to the processing errors of the diameter of the round hole.
Step two: preparation work of the multi-quality target prediction model and a model construction process are as follows: preprocessing a processing technology data set of a processing characteristic rectangle, a crack and a round hole of the flat plate crack antenna to obtain a preprocessed data set; carrying out feature importance ranking on the preprocessed data set to obtain an importance ranking result of the features; performing correlation analysis on the preprocessed data set, and then performing feature selection by combining feature importance ranking results to obtain respective correlation analysis data sets of the processed feature rectangle, the crack and the round hole; dividing the data sets according to the ratio of 7:3 by the respective correlation analysis data sets of the processing feature rectangles, the cracks and the round holes, so that each processing feature has a respective training set of 70% and a respective testing set of 30%; the method comprises the steps of introducing a training set of each machining feature into an XGboost algorithm for training, respectively constructing quality prediction models of a machining feature rectangle, a crack and a round hole, traversing a max _ depth parameter list in the XGboost algorithm, evaluating the precision of the prediction models of each machining feature by using a mean square error and a fitting error obtained by 5-fold cross validation, respectively finding a max _ depth parameter value which enables the precision of the prediction models of each machining feature to reach the highest, and then taking the prediction model corresponding to the parameter as an optimal prediction model to complete the construction of a multi-quality target prediction model of the machining feature rectangle, the crack and the round hole.
Step three: adopting a genetic algorithm of weighted improvement of a multi-quality target prediction model of processing characteristics to construct a multi-process parameter synchronous optimization recommendation model: the method comprises the steps of weighting and summing a multi-target prediction model of a rectangle, a crack and a round hole to serve as a part quality target function, using process parameter rotating speed, cutting depth and feeding as input variables, constructing a multi-process parameter synchronous optimization recommendation model taking part quality as a target based on a genetic algorithm, optimizing the rotating speed, cutting depth and feeding of the process parameters through the model to obtain the combination of the optimal process parameters, and completing multi-quality target prediction and process parameter optimization recommendation for multi-processing characteristic sheet part numerical control processing.
Referring to fig. 2, the step two implementation can be divided into the following steps:
(2a) data preprocessing, replacing detected outliers with mean values of data:
first, statistics calculation for the dataset: and carrying out statistic calculation on the processing process data of the characteristic rectangle, the crack and the round hole of the planar slot antenna, wherein the processing process data comprises a lower quartile, a median, an upper quartile and a quartile interval.
Secondly, judging abnormal values of the data set: and (3) judging abnormal values of the processing characteristic rectangle, the crack and the round hole data according to the following formula:
value>QU+1.5IQR or value<QL-1.5IQR
here, value indicates an abnormal value in each machining feature data, the abnormal value is data located outside a section between QU +1.5IQR and QL-1.5IQR, QU indicates an upper quartile of each machining feature data, QL indicates a lower quartile of each machining feature data, and IQR indicates a quartile pitch of each machining feature data, that is, IQR is QU-QL.
Thirdly, abnormal value replacement of the data set: and replacing the detected abnormal value with the average value of the data of the corresponding processing characteristic rectangle, the crack and the round hole to obtain a preprocessed data set.
(2b) Multidimensional feature importance ranking of data: aiming at a preprocessed data set, respectively using a recursive feature elimination algorithm, a random forest algorithm and an XGboost algorithm to establish regression models of technological parameters and processing features of the flat plate slot antenna, obtaining the regression models corresponding to the three algorithms, respectively using the regression models established by the three algorithms to calculate the importance of the technological parameters of rotating speed, cutting depth and feeding relative to processing feature rectangles, slots and round holes, sequencing the importance of the technological parameters, obtaining the sequencing results of the importance of the technological parameters of each processing feature under different algorithms, taking an intersection from the sequencing results of the importance of the same processing feature under different algorithms, obtaining the sequencing result of the importance of each processing feature, and then taking an intersection from the sequencing results of the importance of all processing features to obtain the final feature importance sequencing result.
(2c) Correlation analysis of data features:
firstly, calculating a correlation coefficient of data characteristics: and constructing a correlation coefficient matrix by calculating correlation coefficients between every two of the process parameters of rotating speed, cutting depth and feeding, and obtaining that the maximum correlation coefficient between the process parameters is not more than 0.12.
And secondly, judging the correlation of the data characteristics: and judging the correlation coefficient among the first three process parameters of the feature importance degree sorting result, reserving the parameters of which the correlation coefficient is less than 0.2 (nonlinear correlation among the parameters), and reserving all the process parameters according to the correlation coefficient matrix, wherein the maximum correlation coefficient among all the process parameters is not more than 0.12.
(2d) A multi-quality target prediction model of the processing characteristics is constructed by using a machine learning algorithm:
step one, optimizing a multi-target prediction model: respectively introducing training sets of machining feature rectangles, cracks and round holes into an XGboost prediction model for training, respectively constructing quality prediction models of the rectangles, the cracks and the round holes, needing to adjust max _ depth parameters of an XGboost algorithm to finish tuning of the quality prediction models of all machining features, setting traversing lists of the max _ depth parameters of all machining feature XGboost prediction models as [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], traversing each parameter in a max _ depth parameter list by using a for-loop, constructing XGboost models corresponding to each max _ depth parameter, evaluating the precision of each model by using a mean square error and a fitting error obtained by 5-fold cross verification, comparing the precision of each model, selecting the max _ depth parameter corresponding to the model with the highest precision, and finding the optimal max _ depth parameter, and taking the XGboost model corresponding to the optimal max _ depth parameter as the optimized prediction model to realize the tuning of the multi-quality target prediction model.
And secondly, evaluating a multi-quality target prediction model: respectively leading training sets of the processing characteristic rectangle, the crack and the round hole into an XGboost prediction model under the optimal max _ depth parameter, evaluating the model performance by using a mean square error and a fitting error to obtain a training error of the model, leading a test set into the XGboost prediction model under the optimal max _ depth parameter, evaluating the model performance by using an output mean square error and an output fitting error, and completing the construction of the multi-quality target prediction model of the processing characteristic rectangle, the crack and the round hole.
Referring to fig. 3, the third step can be divided into the following steps:
(3a) determining an objective function to be optimized: and respectively setting the same weight coefficients of the quality prediction models of the processing characteristic rectangle, the crack and the round hole, and then weighting and summing the multi-quality target prediction models of the processing characteristic rectangle, the crack and the round hole to be used as a part quality target function.
(3b) Determining design variables and parameters of a genetic algorithm, and constructing a process parameter optimization model: the design variables of the genetic algorithm are process parameter rotating speed, cutting depth and feeding, the range of the process parameter rotating speed is 30000(rpm) and 40000(rpm), the range of the cutting depth is 0(mm) and 0.3(mm), and the range of the feeding is 2800(mm) and 4000 (mm); and determining a coding method, a selection mode, a maximum genetic algebra, a fitness calculation mode, a cross mode and a variation mode of the genetic algorithm, and completing the construction of a multi-process-parameter optimization recommendation model.
(3c) Initializing a population: the technological parameters of rotating speed, cutting depth and feeding are used as a group of technological parameter combinations, a plurality of groups of technological parameters are randomly generated within the range of technological parameters to be optimized, the technological parameters of rotating speed, cutting depth and feeding of one group are a group individual, and the initialization of the group is completed by the combination of the plurality of groups of technological parameters.
(3d) Calculating the fitness of the initial population: and (3) giving a fitness threshold condition, calculating the fitness of the initial population individuals according to the part quality objective function and the fitness calculation mode of the initial population individuals, comparing the calculated fitness of the population individuals with the set fitness threshold condition, if the population fitness is greater than the set fitness threshold condition, the initial population is the optimal population, otherwise, the fitness is less than the set fitness threshold condition, and executing the step (3e) to evolve the initial population.
(3e) Carrying out evolution operation on the population, and searching the optimal solution by adopting operators of selection, intersection and variation: the method comprises the steps of giving a population evolution fitness limit value, selecting population individuals larger than the population evolution fitness limit value to generate a next generation, eliminating the population individuals smaller than the fitness limit value, using a next generation population as population filial generations, recombining the population filial generations by using a crossover and mutation operator, enabling the population individuals with high fitness to have the character that a quality objective function reaches the optimum, enabling the population filial generations to inherit the excellent character of the previous generation population through evolution operation, and enabling the population after evolution to gradually advance towards the direction of optimum solution.
(3f) And (3) evaluating the fitness of the evolved population: judging whether the population fitness after evolution is greater than a fitness threshold condition or whether the current algebra is greater than the maximum genetic algebra, if the population fitness is greater than the fitness threshold condition or the current evolutionary algebra is greater than the maximum genetic algebra, searching successfully, otherwise, carrying out a new round of evolution, returning to execute the step (3e) to carry out evolution operation on the population until the population fitness after evolution is greater than the fitness threshold condition or the current algebra is greater than the maximum genetic algebra, obtaining an optimal solution of a process parameter optimization problem, namely obtaining an optimal combination of the process parameter rotating speed, the cutting depth and the feeding, and enabling the quality of parts to be optimal.
The invention uses the machine learning method for multi-quality target prediction of the processing characteristics of the processing workpiece, and combines with the process parameter optimization, the multi-target prediction model of the processing characteristics has higher precision, the weighted summation of the multi-quality target prediction model is used as a target function to improve the genetic algorithm, so that the reliability of the process parameter optimization is higher, and in the actual processing, if the unreasonable parameter setting occurs, the prediction model can carry out early warning and timely processing on the future data, thereby ensuring the stability of the processing process.
The technical effects of the present invention will be further explained by experiments, data and visualization results thereof
Example 8
The method for predicting the multi-quality target and optimizing and recommending the process parameters for the numerical control machining of the multi-machining-characteristic thin plate part is the same as the embodiment 1-7, and the technical effects of the method are as follows:
simulation conditions and contents:
1) simulation conditions are as follows: the modeling experiment is completed on a Spyder (anaconda) running platform under an Intel (R) core (TN) i3-4170CPU @3.70GHz and a Windows 10 (multiplied by 64) system;
2) simulation content: based on the data of the processing characteristics of the flat plate slot antenna, the construction of a multi-quality target prediction model of a processing characteristic rectangle, a slot and a round hole is completed, and the performance and the predicted visual effect of each processing characteristic quality prediction model are output; aiming at the processing characteristic rectangles 1,2,3,4, cracks and round holes on the flat plate crack antenna, an optimization model of the process parameters is constructed by using a genetic algorithm, optimization of the process parameter rotating speed, the cutting depth and the feeding is completed, the optimal combination of the process parameter rotating speed, the cutting depth and the feeding is obtained, the processing quality of the processing characteristics corresponding to the optimized process parameters is compared with the processing quality of the processing characteristics corresponding to the process parameters before optimization, and the improvement effect of the optimized processing quality is shown.
Experimental results and analysis:
1) predicting the performance index and the visualization effect of the model:
the performance indexes of the prediction model obtained by simulation in the invention are shown in table 1, the table 1 is a performance index table of the XGboost prediction model obtained by aiming at the processing characteristics in the invention,
TABLE 1 performance index of XGboost prediction model
Machining features MSE (mean square error) R2 (error of fit)
Round hole 2.315e(-6) 0.579
Rectangle 4.38e(-7) 0.991
Crack (crack) 3.732e(-7) 0.995
From the model performance indexes in the table, the minimum mean square error of the processing characteristic cracks reaches 0.995, the closer the fitting error is to 1, the better the prediction effect is, the closer the fitting error is to 1, the fitting errors of the rectangle and the cracks are to 1, the mean square error of the round holes is lower than that of the rectangle and the cracks, the fitting error is lower, but the reasonable range is reached, and the prediction model precision of the three processing characteristics is higher overall.
The prediction curve of the invention can also be visually displayed, the curve of the true value and the predicted value of the rectangular error is shown in fig. 4, the curve of the true value and the predicted value of the round hole error is shown in fig. 5, the curve of the true value and the predicted value of the crack error is shown in fig. 6, the abscissa of fig. 4, fig. 5 and fig. 6 is the number of the processed workpiece, the ordinate is the processing error corresponding to the processing characteristic, the types of the prediction curves in the graph are all broken lines, and the types of the true data curves of the processed parts are all solid lines. The processing error fitting effect of the rectangular prediction curve in fig. 4 and the crack prediction curve in fig. 6 is the best, the prediction curves are almost close to the real curves, the fitting effect of the processing error prediction curve of the circular holes in fig. 5 is poorer than that of the other two processing characteristics, the prediction curves and the real curves have larger deviation at individual points, but the prediction curves of the circular holes are close to the variation trend of the real curves in the whole view, the mean square error and the fitting error of the whole are in reasonable ranges, and the prediction curves of the rectangular prediction curves, the cracks and the circular holes are close to the real curves to a greater extent and the prediction effects are better.
2) The process parameter optimization effect is good:
comparing the optimization effects of the process parameters, after the optimization of the process parameters with the part quality as the target is completed, comparing the processing quality after the optimization of the process parameters with the processing quality before the optimization in order to clear the optimization effect of the process parameters, and selecting the optimized parameters: [ rotational speed: 30000(rpm), cut depth: 0.15(mm), feed: 3400(mm), calculating the machining quality of each machining feature according to the optimized parameters, and referring to Table 2 for comparison of the machining quality before and after optimization,
TABLE 2 comparison of processing quality before and after optimization
Machining features Before optimization After the invention is optimized Improved effect after optimization
Rectangle 1 Long error (mm) 0.013 0.002 84.6％
Rectangle 1 width error (mm) 0.010 0.009 10％
Rectangle 2 Long error (mm) 0.006 0.002 66.7％
Rectangle 2 width error (mm) 0.011 0.003 72.7％
Rectangle
3 Long error (mm) 0.063 0.002 96.9％
Rectangle
3 width error (mm) 0.012 0.004 66.7％
Rectangle 4 long error (mm) 0.029 0.003 89.7％
Rectangle 4 width error (mm) 0.011 0.001 91％
Crack width error (mm) 0.011 0.008 27.3％
Round hole diameter error (mm) 0.018 0.013 27.8％
The data in the comprehensive table show that the processing quality of the rectangle is obviously improved after the technological parameters are optimized, the processing quality of the rectangle 3 with the length is improved by 96.9%, the processing quality of the rectangle 3 with the width is improved by 66.7%, the processing quality of the crack with the width is improved by 27.3%, and the processing quality of the circular hole with the diameter is improved by 27.8% aiming at the processing characteristics of the flat plate crack antenna. After the optimization of the technological parameters, the best lifting effect is the processing quality with the length of the rectangle 3, the lifting effect reaches 96.9 percent, the lowest lifting effect is the width of the rectangle 1 and also reaches 10 percent, the lifting effects of the processing quality with the length and the width of the other rectangles reach more than 60 percent, and the processing quality of all the processing characteristics is obviously improved.
In summary, the invention discloses a multi-quality target prediction and technological parameter optimization recommendation method for numerical control machining of a multi-machining-characteristic thin plate part, which solves the problems of multi-quality target prediction and technological parameter optimization of the machining characteristics of the part, and the implementation steps comprise: collecting and sorting a data set; preprocessing the sorted data; sorting the feature importance of the data; dividing a correlation analysis data set into a training set and a test set; constructing a multi-target prediction model by using a machine learning algorithm; and (5) constructing a technological parameter optimization model by using a genetic algorithm to complete the technological parameter optimization of the machined part. The invention uses machine learning method to predict the processing characteristics with multi-quality targets, and uses improved genetic algorithm to optimize the technological parameters, so as to obtain the optimal technological parameters. The whole scheme is rigorous and complete, high in prediction precision and good in parameter optimization effect, is used for adaptive recommendation of process parameters and cutter parameters, and effectively improves the machining quality.

Claims (5)

1. A multi-quality target prediction and technological parameter optimization recommendation method for multi-machining characteristic sheet part numerical control machining is provided, a typical plate crack antenna of a machined part is taken as a machining object, and the method is characterized by comprising the following steps:
step (1) collecting and arranging a processing technology and a working medium adding data set: collecting process parameter data of a planar slot antenna of a machined part by using a numerical control machine tool, collecting machining quality data of the planar slot antenna by using a three-coordinate measuring instrument, then extracting the data of three machining characteristics, namely a rectangle, a slot and a round hole of the planar slot antenna, corresponding to the rotating speed, the cutting depth and the feeding of the process parameters one by one, and sorting the data into a planar slot antenna machining quality prediction and process parameter optimization data set classified according to the machining characteristics;
step (2) data preprocessing: detecting abnormal values contained in three processing characteristic data, namely a rectangle, a crack and a round hole of the flat plate crack antenna processing quality prediction and process parameter optimization data set by using a box separation method, and replacing the abnormal values with the mean value of corresponding processing characteristics to obtain a preprocessed data set;
and (3) sorting the importance of the multidimensional features of the data: aiming at the preprocessed data set, respectively using a recursive feature elimination algorithm, a random forest algorithm and an XGboost algorithm to establish a regression model of the process parameters and the processing features of the flat plate slot antenna, calculating the importance of the rotating speed, the cutting depth and the feeding of the process parameters relative to the processing features, sequencing the importance of the process parameters, and obtaining the sequencing result of the importance of the process parameters of the processed parts by taking the intersection of the sequencing results of the algorithms;
and (4) carrying out correlation analysis on data characteristics and deleting redundant characteristics: performing relevance analysis on the rotating speed, the cutting depth and the feeding of the process parameters by using a relevance analysis method aiming at the preprocessed data set to obtain the relevant coefficients among the process parameters, judging the relevant coefficients among the process parameters with higher rank by combining the process parameter importance ranking results of the processed parts, deleting one of the two process parameters with the highest relevant coefficients, and reserving the other one of the two process parameters to respectively obtain the relevance analysis data sets of the processed characteristic rectangles, the cracks and the round holes of the processed parts;
step (5) respectively carrying out training set and test set division on the processing characteristic rectangle, crack and round hole correlation analysis data set: respectively dividing the respective correlation analysis data sets of the processing characteristic rectangles, the cracks and the round holes of the processed parts according to the ratio of 7:3, and respectively taking 70% of the respective data sets of the processing characteristic rectangles, the cracks and the round holes as training sets corresponding to the processing characteristics and the rest 30% as test sets corresponding to the processing characteristics;
and (6) constructing a multi-quality target prediction model by using a machine learning algorithm: aiming at the respective correlation analysis data sets of the processing characteristic rectangle, the crack and the round hole of the processing part, respectively training and learning the respective training sets of the processing characteristic rectangle, the crack and the round hole by using an XGboost algorithm, respectively constructing respective quality prediction models of the processing characteristic rectangle, the crack and the round hole, respectively using the process parameter rotating speed, the cutting depth and the feeding of the processing part in the prediction models as a group of input characteristics, respectively using the respective quality data of the processing characteristic rectangle, the crack and the round hole as output, adjusting the respective prediction models of the processing characteristic rectangle, the crack and the round hole one by one according to a 5-fold cross verification result, and finally evaluating the prediction models by using the respective test sets of the processing characteristic rectangle, the crack and the round hole to complete the construction of the multi-quality target prediction models of the rectangle, the crack and the round hole;
and (7) constructing a multi-process parameter synchronous optimization recommendation model based on the processing characteristic multi-quality target prediction model weighted improved genetic algorithm, and completing multi-process parameter synchronous optimization recommendation of the processed parts: the method comprises the steps of weighting and summing a multi-quality target prediction model of a rectangle, a crack and a round hole to serve as a part quality target function, using process parameter rotating speed, cutting depth and feeding as input variables, using a genetic algorithm to construct a multi-process parameter synchronous optimization recommendation model taking part quality as a target, optimizing the rotating speed, cutting depth and feeding of process parameters through the model, obtaining an optimal process parameter combination for recommendation, and completing multi-process parameter synchronous optimization recommendation of a machined part.
2. The multi-quality target prediction and process parameter optimization recommendation method for multi-machining-feature sheet part numerical control machining according to claim 1, wherein the data preprocessing of the step (2) comprises the following steps:
(2.1) statistics calculation for data set: calculating the statistical quantity of processing process data of the processing characteristic rectangle, the crack and the round hole of the processed part, wherein the processing process data comprises a lower quartile, a median, an upper quartile and a quartile interval;
(2.2) abnormal value judgment of data set: and (3) judging abnormal values of the processing characteristic rectangle, crack and round hole data of the processed part according to the following formula:
value>QU+1.5IQR or value<QL-1.5IQR
wherein value represents an abnormal value in each machining feature data, the abnormal value is data outside a section between QU +1.5IQR and QL-1.5IQR, QU represents an upper quartile of each machining feature data, QL represents a lower quartile of each machining feature data, IQR represents a quartile pitch of each machining feature data, that is, IQR is QU-QL;
(2.3) outlier replacement of data set: and replacing the detected abnormal value with the average value of the data of the corresponding processing characteristic rectangle, the crack and the round hole to obtain a preprocessed data set.
3. The multi-quality target prediction and process parameter optimization recommendation method for multi-machining feature sheet part numerical control machining according to claim 1, wherein the correlation analysis and redundant feature deletion of the data features in the step (4) comprise the following steps:
(4.1) calculation of correlation coefficient of data characteristics: constructing a correlation coefficient matrix by calculating correlation coefficients between every two of process parameters of rotating speed, cutting depth and feeding of the processed parts, and obtaining that the correlation degree between the process parameters is not more than 0.12 to the maximum extent;
(4.2) judging the relevance of the data features: judging the correlation coefficient among the process parameters of the machined parts with higher feature importance degree sequence, deleting one of the two parameters with the correlation coefficient larger than 0.3, reserving the other parameter, reserving the two parameters with the correlation coefficient smaller than 0.3 at the same time, and reserving all the process parameters according to the maximum correlation coefficient among the obtained process parameters which is not more than 0.12.
4. The multi-quality target prediction and process parameter optimization recommendation method for multi-machining-feature sheet part numerical control machining according to claim 1, wherein the step (6) of constructing the multi-quality target prediction model by using a machine learning algorithm comprises the following steps:
(6.1) tuning of the multi-quality target prediction model: respectively introducing training sets of processing characteristic rectangles, cracks and round holes of a processed part into an XGboost algorithm for training, setting a max _ depth parameter traversal list of the XGboost algorithm as [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], traversing each parameter in the max _ depth parameter list by using a for-loop, constructing an XGboost model corresponding to each max _ depth parameter, evaluating the precision of each model by using a mean square error and a fitting error obtained by 5-fold cross verification, circularly comparing the mean square error and the fitting error of every two models to obtain the max _ depth parameter corresponding to the model with the highest precision, realizing the search of the optimal max _ depth parameter, and finishing the optimization of a multi-quality target prediction model;
(6.2) evaluation of the multi-quality target prediction model: respectively leading training sets of processing characteristic rectangles, cracks and round holes of the processed parts into XGboost prediction models under respective optimal max _ depth parameters, using mean square error and fitting error to evaluate model performance to obtain training errors of the models, leading test sets into XGboost prediction models corresponding to the optimal max _ depth parameters, using mean square error and fitting error to evaluate model performance to obtain test errors of the models, using the test errors as evaluation of multi-target prediction models, and finally completing construction of multi-quality target prediction models.
5. The multi-quality target prediction and process parameter optimization recommendation method for multi-machining-feature sheet part numerical control machining according to claim 1, wherein the multi-process parameter synchronous optimization recommendation model is constructed based on the genetic algorithm for weighted improvement of the machining-feature multi-quality target prediction model in the step (7) to complete multi-process parameter synchronous optimization recommendation of the machined part, and the construction of the multi-process parameter synchronous optimization recommendation model comprises the following steps:
(7.1) determining an objective function to be optimized: respectively multiplying the multi-quality target prediction models of the processing characteristic rectangles, the cracks and the round holes of the processed parts by the same weight coefficient to obtain weighted multi-quality target prediction models of the rectangles, the cracks and the round holes, and then summing the weighted multi-quality target prediction models of the rectangles, the cracks and the round holes to serve as a part quality target function to be optimized;
(7.2) determining design variables and parameters of a genetic algorithm, and constructing a process parameter optimization model: the design variables of the genetic algorithm are process parameter rotating speed, cutting depth and feeding, the range of the process parameter rotating speed is 30000(rpm) and 40000(rpm), the range of the cutting depth is 0(mm) and 0.3(mm), and the range of the feeding is 2800(mm) and 4000 (mm); determining a coding method, a selection mode, a maximum genetic algebra, a fitness calculation mode, a cross mode and a variation mode of a genetic algorithm, and completing the construction of a multi-process-parameter synchronous optimization recommendation model;
(7.3) initializing population: the method comprises the following steps of combining process parameters of rotating speed, cutting depth and feeding as a group, randomly generating a plurality of groups of process parameters within the range of the process parameters to be optimized, wherein the rotating speed, cutting depth and feeding process parameters of one group are a group individual, and the initialization of the group is completed by combining the plurality of groups of process parameters;
(7.4) calculating the fitness of the initial population: giving a fitness threshold condition, calculating the fitness of the initial population individuals according to a part quality objective function and a fitness calculation mode of the initial population individuals, comparing the fitness of the population individuals with the set threshold condition, if the fitness is greater than the threshold condition, the initial population is the optimal population, otherwise, executing the step (7.5) and carrying out evolution on the initial population;
(7.5) evolving population: for a population to be evolved, a fitness limit value allowed to be evolved is given, the fitness of individuals of the population to be evolved is compared with the fitness limit value allowed to be evolved, the population individuals larger than the fitness limit value allowed to be evolved are used for generating a next generation population, the next generation population is population filial generation, then cross and variation operators are used for recombining population filial generation individuals to obtain population individuals after evolution, and a quality objective function corresponding to the population individuals after evolution is better than that of the previous generation;
(7.6) calculating the fitness of the population after evolution: calculating the fitness of the evolved population individuals according to a part quality target function of the evolved population individuals by using a given fitness threshold condition and a determined maximum genetic algebra, judging whether the fitness of the evolved population individuals is greater than the fitness threshold condition, if so, determining that the evolved population is an optimal population, combining the process parameters of rotating speed, cutting depth and feeding as a group to obtain a process parameter combination which enables the part quality to reach the optimal quality, and otherwise, returning to the step (7.5) to execute the next round of evolution; or directly judging whether the evolution algebra reaches the maximum genetic algebra, if so, obtaining a process parameter combination which enables the part quality to reach the optimum, and finishing the optimization of the process parameters, otherwise, returning to the execution step (7.5) to execute the next round of evolution until the population individuals after the evolution are larger than the fitness threshold condition or the evolution algebra reaches the maximum genetic algebra, finally realizing the optimization of the rotating speed, the cutting depth and the feeding of the process parameters, obtaining the combination of the optimum process parameters, and finishing the synchronous optimization recommendation of the process parameters of the processed parts.
CN202110469230.2A 2021-04-28 2021-04-28 Quality prediction and technological parameter optimization method for multi-processing characteristic sheet part Active CN113176761B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110469230.2A CN113176761B (en) 2021-04-28 2021-04-28 Quality prediction and technological parameter optimization method for multi-processing characteristic sheet part

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110469230.2A CN113176761B (en) 2021-04-28 2021-04-28 Quality prediction and technological parameter optimization method for multi-processing characteristic sheet part

Publications (2)

Publication Number Publication Date
CN113176761A true CN113176761A (en) 2021-07-27
CN113176761B CN113176761B (en) 2022-09-06

Family

ID=76925207

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110469230.2A Active CN113176761B (en) 2021-04-28 2021-04-28 Quality prediction and technological parameter optimization method for multi-processing characteristic sheet part

Country Status (1)

Country Link
CN (1) CN113176761B (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113779895A (en) * 2021-11-10 2021-12-10 深圳市裕展精密科技有限公司 Data model establishing device and method, and machining quality predicting device and method
CN114139789A (en) * 2021-11-23 2022-03-04 成都飞机工业（集团）有限责任公司 Airplane cost prediction method based on machine learning
CN114417739A (en) * 2022-03-29 2022-04-29 希望知舟技术(深圳)有限公司 Method and device for recommending process parameters under abnormal working conditions
CN114789364A (en) * 2022-04-18 2022-07-26 清华大学 Multi-index drilling quality control method, device and equipment
CN114926075A (en) * 2022-06-02 2022-08-19 西安电子科技大学 Mechanical part production scheduling method based on man-hour prediction
CN115048422A (en) * 2022-05-16 2022-09-13 上海控软网络科技有限公司 Process recommendation method, device, equipment and storage medium
CN115509196A (en) * 2022-10-27 2022-12-23 深圳市卓讯达科技发展有限公司 Manufacturing process optimization method and device based on machine learning
CN116151470A (en) * 2023-03-06 2023-05-23 联宝(合肥)电子科技有限公司 Product quality prediction method, device, equipment and storage medium
CN116300475A (en) * 2023-05-11 2023-06-23 苏州先准电子科技有限公司 Metal rolling control method and system
CN116628970A (en) * 2023-05-18 2023-08-22 浙江大学 Method for optimizing spinning forming technological parameters of aerospace thin-wall part based on data mining
WO2023184281A1 (en) * 2022-03-30 2023-10-05 京东方科技集团股份有限公司 Inspection parameter analysis method and apparatus
CN117312802A (en) * 2023-11-29 2023-12-29 安徽大学 SMT production line key process parameter optimization method based on man-machine fusion and storage medium
CN117495070A (en) * 2023-12-27 2024-02-02 中控技术股份有限公司 Technological parameter recommendation method and system for industrial assembly line
CN117744281A (en) * 2024-02-19 2024-03-22 北京航空航天大学 Bearing product manufacturing quality improvement analysis method
CN117952675A (en) * 2024-03-27 2024-04-30 山东鼎信数字科技有限公司 Method, system, device and storage medium for dynamic issuing of rights and interests products

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107480352A (en) * 2017-07-28 2017-12-15 东北大学 A kind of reliability optimization method of Milling Process technological parameter
CN107976956A (en) * 2017-11-22 2018-05-01 沈阳机床股份有限公司 The multi objective cutting data generating algorithm and cutting parameter optimization method of numerically-controlled machine tool
US10061300B1 (en) * 2017-09-29 2018-08-28 Xometry, Inc. Methods and apparatus for machine learning predictions and multi-objective optimization of manufacturing processes
CN108846521A (en) * 2018-06-22 2018-11-20 西安电子科技大学 Shield-tunneling construction unfavorable geology type prediction method based on Xgboost
CN109814513A (en) * 2019-03-20 2019-05-28 杭州辛孚能源科技有限公司 A kind of catalytic cracking unit optimization method based on data model
CN109828532A (en) * 2019-01-29 2019-05-31 贵州大学 A kind of Prediction of Surface Roughness method and process parameter optimizing method based on GA-GBRT
CN110516818A (en) * 2019-05-13 2019-11-29 南京江行联加智能科技有限公司 A kind of high dimensional data prediction technique based on integrated study technology
CN111563301A (en) * 2020-05-14 2020-08-21 北京工业大学 Thin-wall part milling parameter optimization method
CN111651929A (en) * 2019-12-24 2020-09-11 广东海洋大学 Multi-objective optimization method based on fusion of Dynaform and intelligent algorithm
CN111931307A (en) * 2020-08-11 2020-11-13 西北工业大学 Part quality prediction and evaluation optimization method and system
CN112380760A (en) * 2020-10-13 2021-02-19 重庆大学 Multi-algorithm fusion based multi-target process parameter intelligent optimization method
CN112379589A (en) * 2020-10-13 2021-02-19 重庆大学 Worm wheel machining shape controllable process
US20210110313A1 (en) * 2019-09-18 2021-04-15 Hartford Steam Boiler Inspection And Insurance Company Computer-based systems, computing components and computing objects configured to implement dynamic outlier bias reduction in machine learning models

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107480352A (en) * 2017-07-28 2017-12-15 东北大学 A kind of reliability optimization method of Milling Process technological parameter
US10061300B1 (en) * 2017-09-29 2018-08-28 Xometry, Inc. Methods and apparatus for machine learning predictions and multi-objective optimization of manufacturing processes
CN107976956A (en) * 2017-11-22 2018-05-01 沈阳机床股份有限公司 The multi objective cutting data generating algorithm and cutting parameter optimization method of numerically-controlled machine tool
CN108846521A (en) * 2018-06-22 2018-11-20 西安电子科技大学 Shield-tunneling construction unfavorable geology type prediction method based on Xgboost
CN109828532A (en) * 2019-01-29 2019-05-31 贵州大学 A kind of Prediction of Surface Roughness method and process parameter optimizing method based on GA-GBRT
CN109814513A (en) * 2019-03-20 2019-05-28 杭州辛孚能源科技有限公司 A kind of catalytic cracking unit optimization method based on data model
CN110516818A (en) * 2019-05-13 2019-11-29 南京江行联加智能科技有限公司 A kind of high dimensional data prediction technique based on integrated study technology
US20210110313A1 (en) * 2019-09-18 2021-04-15 Hartford Steam Boiler Inspection And Insurance Company Computer-based systems, computing components and computing objects configured to implement dynamic outlier bias reduction in machine learning models
CN111651929A (en) * 2019-12-24 2020-09-11 广东海洋大学 Multi-objective optimization method based on fusion of Dynaform and intelligent algorithm
CN111563301A (en) * 2020-05-14 2020-08-21 北京工业大学 Thin-wall part milling parameter optimization method
CN111931307A (en) * 2020-08-11 2020-11-13 西北工业大学 Part quality prediction and evaluation optimization method and system
CN112380760A (en) * 2020-10-13 2021-02-19 重庆大学 Multi-algorithm fusion based multi-target process parameter intelligent optimization method
CN112379589A (en) * 2020-10-13 2021-02-19 重庆大学 Worm wheel machining shape controllable process

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
唐超兰等: "6061铝合金铣削工艺参数多目标优化", 《广东工业大学学报》 *
蔡安江等: "基于BP神经网络的数控加工铣削参数优化", 《模具工业》 *

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113779895A (en) * 2021-11-10 2021-12-10 深圳市裕展精密科技有限公司 Data model establishing device and method, and machining quality predicting device and method
CN114139789A (en) * 2021-11-23 2022-03-04 成都飞机工业（集团）有限责任公司 Airplane cost prediction method based on machine learning
CN114417739A (en) * 2022-03-29 2022-04-29 希望知舟技术(深圳)有限公司 Method and device for recommending process parameters under abnormal working conditions
CN114417739B (en) * 2022-03-29 2022-07-26 希望知舟技术(深圳)有限公司 Method and device for recommending process parameters under abnormal working conditions
WO2023184281A1 (en) * 2022-03-30 2023-10-05 京东方科技集团股份有限公司 Inspection parameter analysis method and apparatus
CN114789364A (en) * 2022-04-18 2022-07-26 清华大学 Multi-index drilling quality control method, device and equipment
CN114789364B (en) * 2022-04-18 2023-11-24 清华大学 Multi-index drilling quality control method and device and equipment thereof
CN115048422A (en) * 2022-05-16 2022-09-13 上海控软网络科技有限公司 Process recommendation method, device, equipment and storage medium
CN114926075B (en) * 2022-06-02 2024-06-07 西安电子科技大学 Machine part production scheduling method based on man-hour prediction
CN114926075A (en) * 2022-06-02 2022-08-19 西安电子科技大学 Mechanical part production scheduling method based on man-hour prediction
CN115509196A (en) * 2022-10-27 2022-12-23 深圳市卓讯达科技发展有限公司 Manufacturing process optimization method and device based on machine learning
CN116151470A (en) * 2023-03-06 2023-05-23 联宝(合肥)电子科技有限公司 Product quality prediction method, device, equipment and storage medium
CN116300475A (en) * 2023-05-11 2023-06-23 苏州先准电子科技有限公司 Metal rolling control method and system
CN116300475B (en) * 2023-05-11 2023-08-22 苏州先准电子科技有限公司 Metal rolling control method and system
CN116628970A (en) * 2023-05-18 2023-08-22 浙江大学 Method for optimizing spinning forming technological parameters of aerospace thin-wall part based on data mining
CN117312802B (en) * 2023-11-29 2024-03-22 安徽大学 SMT production line key process parameter optimization method based on man-machine fusion and storage medium
CN117312802A (en) * 2023-11-29 2023-12-29 安徽大学 SMT production line key process parameter optimization method based on man-machine fusion and storage medium
CN117495070A (en) * 2023-12-27 2024-02-02 中控技术股份有限公司 Technological parameter recommendation method and system for industrial assembly line
CN117744281A (en) * 2024-02-19 2024-03-22 北京航空航天大学 Bearing product manufacturing quality improvement analysis method
CN117744281B (en) * 2024-02-19 2024-04-26 北京航空航天大学 Bearing product manufacturing quality improvement analysis method
CN117952675A (en) * 2024-03-27 2024-04-30 山东鼎信数字科技有限公司 Method, system, device and storage medium for dynamic issuing of rights and interests products

Also Published As

Publication number Publication date
CN113176761B (en) 2022-09-06

Similar Documents

Publication Publication Date Title
CN113176761B (en) Quality prediction and technological parameter optimization method for multi-processing characteristic sheet part
CN102567812B (en) Method for predicting control index by using processing parameter in tobacco processing procedure
CN109977464B (en) Prediction method of piston cutting deformation based on BP neural network
Majumder et al. Application of MCDM based hybrid optimization tool during turning of ASTM A588
CN113051683B (en) Numerical control machine tool life prediction method, system, equipment and storage medium
CN112364560B (en) Intelligent prediction method for working hours of mine rock drilling equipment
CN113591234B (en) Method for analyzing and checking parameters of self-punching riveting process simulation model based on machine learning
Hamzaçebi et al. Taguchi method as a robust design tool
CN107391890B (en) Prediction and optimal control method for oil casing threaded joint machining chatter defect
CN113469241A (en) Product quality control method based on process network model and machine learning algorithm
CN116520772A (en) Sheet part numerical control machining quality prediction method based on multitasking transfer learning
CN114818201A (en) Engine cylinder head milling surface quality prediction method based on mechanism and data driving
CN109885900A (en) A kind of Cutting parameters optimization method based on gray analysis
CN113868912B (en) Parameter identification and correction method for J-C constitutive model of titanium alloy
CN114254915A (en) Method for deciding and optimizing qualified state of full-flow processing quality of shaft parts
CN102621953A (en) Automatic online quality monitoring and prediction model updating method for rubber hardness
CN112861433A (en) Product low-carbon design method based on multi-level integrated framework
CN115018452A (en) Construction and comprehensive evaluation method for project progress management index system
CN113656970A (en) Aircraft engine product process lifting method, system, equipment and medium
Ibraheem et al. Prediction of cutting forces by using machine parameters in end milling process
Fan et al. Research on tool wear modeling of superalloy based on evolutionary cluster analysis
Roy et al. Optimization of process parameters using Taguchi coupled genetic algorithm: machining in CNC lathe
Xu et al. Cutting-state identification of machine tools based on improved Dempster-Shafer evidence theory
CN111291518A (en) Crusher cutter abrasion degree judging method based on BP neural network
CN117592223B (en) Intelligent design method of hole machining tool for aerospace materials

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

Inventor after: Wang Pei

Inventor after: Chang Jiantao

Inventor after: Kong Xianguang

Inventor after: Bu Fanhui

Inventor after: Gao Xiaoxu

Inventor after: Liu Dekun

Inventor after: Zhang Anji

Inventor after: Su Xin

Inventor before: Wang Pei

Inventor before: Chang Jiantao

Inventor before: Kong Xianguang

Inventor before: Bu Fanhui

Inventor before: Gao Xiaoxu

Inventor before: Liu Dekun

Inventor before: Zhang Anji

CB03 Change of inventor or designer information
GR01 Patent grant
GR01 Patent grant