CN116384257B - Method for predicting assembly errors and optimizing tolerance of air separation integral cold box - Google Patents
Method for predicting assembly errors and optimizing tolerance of air separation integral cold box Download PDFInfo
- Publication number
- CN116384257B CN116384257B CN202310615208.3A CN202310615208A CN116384257B CN 116384257 B CN116384257 B CN 116384257B CN 202310615208 A CN202310615208 A CN 202310615208A CN 116384257 B CN116384257 B CN 116384257B
- Authority
- CN
- China
- Prior art keywords
- tolerance
- assembly
- model
- cold box
- flexible body
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000000926 separation method Methods 0.000 title claims description 23
- 238000005457 optimization Methods 0.000 claims abstract description 71
- 238000004458 analytical method Methods 0.000 claims abstract description 35
- 230000006870 function Effects 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000001816 cooling Methods 0.000 claims abstract description 24
- 238000013461 design Methods 0.000 claims abstract description 24
- 230000008569 process Effects 0.000 claims abstract description 24
- 238000010801 machine learning Methods 0.000 claims abstract description 22
- 238000004088 simulation Methods 0.000 claims abstract description 18
- 238000005070 sampling Methods 0.000 claims description 17
- 238000009826 distribution Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 13
- 239000002245 particle Substances 0.000 claims description 13
- 238000005259 measurement Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 abstract description 6
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 229910000831 Steel Inorganic materials 0.000 description 12
- 239000010959 steel Substances 0.000 description 12
- 238000004519 manufacturing process Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 10
- 239000003795 chemical substances by application Substances 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 238000012546 transfer Methods 0.000 description 3
- 238000003466 welding Methods 0.000 description 3
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000003394 haemopoietic effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 229910052757 nitrogen Inorganic materials 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000001739 rebound effect Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting assembly errors and optimizing tolerances of a space division self-contained cooling box, and belongs to the field of machine learning. The invention constructs the agent model capable of being applied to the assembly error prediction of the rigid flexible body of the cold box by training the machine learning model, thereby applying the machine learning agent model to tolerance optimization. The agent model is established based on sample data of three-dimensional tolerance analysis simulation, and model parameter optimization is performed by using an optimization algorithm. In addition, the agent model considers deformation conditions of the flexible body under the condition of external force and temperature change in the assembly process, and is coupled with tolerance transmission, so that the assembly error prediction is more comprehensive and accurate. Meanwhile, the invention establishes the cost-tolerance function of the cold box on the basis of the prediction model, and can further optimize the tolerance design. Therefore, the invention can improve the error prediction precision of the assembly of the rigid and flexible body of the cold box and the tolerance optimization design efficiency of parts, and reduce the tolerance design cost.
Description
Technical Field
The invention belongs to the field of machine learning, and particularly relates to a method for predicting assembly errors and optimizing tolerances of an air-conditioner box through a machine learning model.
Background
The air separation equipment takes air as raw material to prepare industrial gases such as oxygen, nitrogen and the like, is important core equipment of strategic industries such as energy, chemical refining integration and the like, and is a hematopoietic organ which depends on the existence of modern industry. Cryogenic cold boxes are the most critical main structure for cryogenic rectification in air separation plants. However, due to the large size and the numerous components, the problems of difficult installation, long period, more manpower and the like in the process of installing the real objects, difficult control of the assembly error of the large-sized air-separation integral cold box, difficult design of tolerance of parts and the like are caused. Meanwhile, the flexible body can not only influence the assembly precision due to the tolerance of parts in the assembly process, but also deform due to the change of stress and temperature, so that the assembly precision is further influenced. Therefore, how to optimize the tolerance of the space-filling cold box and meet the double requirements of assembly errors and cost is a technical problem to be solved in the prior art.
Disclosure of Invention
The invention aims to solve the defect that the tolerance is difficult to accurately control at low cost in the assembling process of the air-separation integral cold box, and provides a method for predicting the assembling error and optimizing the tolerance of the air-separation integral cold box. According to the method, a large-scale cold box rigid-flexible body assembly error prediction proxy model is established through machine learning, a cost-tolerance function is further established, and an optimization algorithm is used for optimizing the assembly tolerance of the cold box parts, so that the aim of reducing production cost is achieved.
The technical scheme for realizing the purpose of the invention is as follows:
a method for predicting assembly errors and optimizing tolerances of a space division self-contained cold box comprises the following steps:
s1, sampling all tolerance variables to be optimized in the assembly process of the air separation integral cold box parts in a respective tolerance zone range to form a sample set, wherein each sample consists of a group of tolerance values of all the tolerance variables to be optimized;
s2, adding tolerance information of all tolerance variables to be optimized to a rigid-flexible body assembly model of the cold box in a three-dimensional tolerance analysis system, and simultaneously importing a rigidity matrix and a temperature load file of a flexible body in the model; generating feature points on the flexible body uniformly, and associating each feature point with the nearest flexible body grid point; setting the assembly sequence of parts in the rigid and flexible body assembly model according to actual procedures;
s3, taking the tolerance value contained in each sample in the sample set, the assembly environment temperature of the cold box and the applied external force as inputs of a three-dimensional tolerance analysis system, and obtaining the assembly error value and tolerance contribution degree of the position of the appointed measuring point on the cold box through tolerance simulation analysis; screening and dimension reduction are carried out on all tolerance variables to be optimized according to the tolerance contribution degree, and key tolerance variables are obtained;
s4, based on tolerance simulation analysis data of each sample obtained in the S3, taking a tolerance value corresponding to a key tolerance variable, the assembly environment temperature of the cold box and the external force applied to the cold box as input of a machine learning model, taking the assembly error value obtained by the tolerance simulation analysis as a truth value label, and performing supervision training on the machine learning model to obtain a proxy model;
s5, optimizing all key tolerance variables by adopting an optimization algorithm according to a cost-tolerance function of the cold box and taking a measurement point error as a constraint condition and taking the minimum cost as an optimization target, wherein the assembly error value corresponding to each group of key tolerance variable feasible solutions in the optimization process is obtained by calculating a proxy model; and finally, taking the optimal solution obtained by optimization as a tolerance variable design value of the air-separation self-contained cold box.
Preferably, in the step S1, the sampling method uses latin hypercube sampling.
Preferably, the three-dimensional tolerance analysis system is implemented by 3DCS software.
Preferably, in the three-dimensional tolerance analysis system, a tolerance distribution type of tolerance information added by the rigid-flexible body assembly model is set to a constant type.
Preferably, in the step S2, the stiffness matrix and the temperature load file of the flexible body are generated by gridding the flexible body structure of the cold box and then performing finite element analysis.
Preferably, in the step S3, a minimum contribution threshold is preset before the screening and dimension reduction, and tolerance variables lower than the minimum contribution threshold among all the tolerance variables to be optimized are removed, and the remaining tolerance variables are used as key tolerance variables.
Preferably, in the S4, the machine learning model uses a neural network model or a support vector machine model.
Preferably, in the step S4, the machine learning model adopts a support vector regression model, the kernel function adopts a gaussian kernel function, and the parameter optimization algorithm during model training adopts a TPE algorithm.
Preferably, in S5, the optimization algorithm is a particle swarm optimization algorithm.
Preferably, in the step S5, the constraint condition of the optimization algorithm is that the upper limit of the measured point error does not exceed the maximum allowable error, and the optimization objective is that the sum of the tolerance costs of all the key tolerance variables calculated based on the cost-tolerance function is minimum.
Compared with the prior art, the invention has the remarkable advantages that:
according to the invention, a prediction model between the tolerance, stress, temperature and assembly errors of parts is established by using the agent model based on three-dimensional tolerance transfer and flexible body finite element rigidity temperature matrix simulation sample data, and the prediction model is further combined with a particle swarm optimization algorithm to optimally design the tolerance of the large-sized cold box parts. In addition, in the invention, the method of adding the finite element rigidity matrix enables the prediction of the assembly error to be more relevant to the actual situation. In order to further improve the accuracy of the proxy model assembly error prediction, the model is optimized by adopting a TPE super-parameter optimizing algorithm. Finally, on the basis of the optimized agent model, an extra-large integral cold box cost-tolerance function is further constructed, and the tolerance production cost is optimally designed by utilizing a particle swarm algorithm, so that the integral manufacturing quality of the large cold box is ensured, and the design and generation processing cost is reduced. The method is applied to the prediction of the assembly error and the tolerance optimization of the large-sized cold box, and can efficiently find the assembly error and the optimal tolerance design parameter from the complex high-dimensional parameters, so that the design accuracy is ensured for the design of the large-sized cold box, the design period is shortened, and the design and production cost is reduced.
Drawings
FIG. 1 is a schematic diagram of steps of a method for predicting assembly errors and optimizing tolerances of a space division self-contained cooling box according to the present invention.
FIG. 2 is a flowchart of an optimization method in an embodiment of the invention.
FIG. 3 is a drawing of a large cold box side skin steel sheet model in an embodiment of the invention.
FIG. 4 is a histogram of statistical distribution of the proxy model training set error in an embodiment of the present invention.
FIG. 5 is a histogram of statistical distribution of the error of the proxy model test set in an embodiment of the present invention.
Fig. 6 is a schematic diagram illustrating a change of cost in an iterative process of a particle swarm optimization algorithm according to an embodiment of the present invention.
The reference numerals in the drawings are: measuring point 1, cold box upper surface skin steel sheet 2, cold box lower surface skin steel sheet 3, locating bolt 4.
Detailed Description
For a clearer understanding of the present invention, a method of embodying the present invention will be explained in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following specific examples. It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other.
The air-separation self-contained cooling box is generally assembled by a rigid body and a flexible body, and is a rigid-flexible body assembling structure. However, the tolerance of the flexible body and the deformation of the flexible body caused by external force and temperature change affect the assembly precision, so that the tolerance of the assembled parts should not be excessively controlled. But if the tolerances are controlled too much, this will also result in excessive processing costs. The method for predicting the assembly error and optimizing the tolerance of the oversized air-separation self-contained cooling box can be used for predicting and optimizing the tolerance of any assembly body which is just flexibly matched in the oversized self-contained cooling box, for example, the method can be used for predicting the assembly error of a flexible body of a side surface skin steel plate of the oversized self-contained cooling box and optimizing the tolerance of a rigid positioning bolt and a part, and also can be used for correspondingly analyzing the welding assembly of a flexible part of a rectifying tower in the self-contained cooling box. Specific implementations of the method are described in detail below.
As shown in FIG. 1, as a preferred implementation manner of the present invention, a method for predicting assembly errors and optimizing tolerances of a space division self-contained cooling box is provided, which comprises the following steps S1-S5.
S1, sampling all tolerance variables to be optimized in the assembly process of the air separation integral cold box parts in a respective tolerance zone range to form a sample set, wherein each sample consists of a group of tolerance values of all the tolerance variables to be optimized.
It should be noted that the tolerance variable to be optimized needs to be selected according to the actual assembly condition of the cold box components, and is not limited to what specific tolerance variable combination is adopted. For example, in an embodiment of the present invention, for a rigid-flexible assembly formed by assembling a flexible body of a skin steel plate and a rigid positioning bolt in a cold box, all or part of tolerance variables in lower surface flatness, reference surface profile, lower surface general surface profile, lower surface circular hole dimensional tolerance, upper surface general surface profile, upper surface two circular hole dimensional tolerance and simplified bolt roundness can be adopted as tolerance variables to be optimized.
In addition, in practical application, when the respective tolerance zone ranges of the tolerance variables are determined, the allowable tolerance and deviation of the tolerance ratio variables can be further determined according to the preset tolerance information of the two-dimensional design drawing of the cold box rigid flexible body assembly model, so that the corresponding tolerance zone ranges are determined.
In addition, the sample set formed by sampling is used as input data of a subsequent tolerance analysis system, corresponding assembly errors under different tolerance variable values can be formed through tolerance analysis, and therefore a machine learning model can be trained through a large amount of data. The sampling method adopts Latin hypercube sampling, namely, if N samples are to be taken from the tolerance zone range of each tolerance variable, the tolerance zone range of each tolerance variable is divided into N identical subintervals, a value is randomly selected from each subinterval, the N values of each tolerance variable and the values of other tolerance variables are randomly combined, and each group formed by combination covers sampling data of all tolerance variable values, namely, one sample. Of course, other sampling methods besides Latin hypercube sampling may theoretically be employed, provided that the representativeness of the sampling is ensured.
S2, adding tolerance information of all tolerance variables to be optimized to a rigid flexible body assembly model of the cold box in a three-dimensional tolerance analysis system, and simultaneously importing a rigidity matrix and a Temperature Load (Temperature Load) file of a flexible body in the model; generating feature points on the flexible body uniformly, and associating each feature point with the nearest flexible body grid point; the assembly sequence of the parts in the rigid-flexible body assembly model is set according to the actual working procedure.
It should be noted that the three-dimensional tolerance analysis system in the present invention may be any software, platform or self-building program capable of implementing three-dimensional tolerance analysis. In one embodiment of the present invention, the three-dimensional tolerance analysis system may be implemented using 3DCS software, which may better implement the tolerance analysis functions required by the present invention.
In addition, in order to ensure that the analysis result in the three-dimensional tolerance analysis system can meet the requirement of the subsequent machine learning model on the training sample, the tolerance distribution type of the tolerance information added in the rigid-flexible body assembly model of the cold box is set to be a constant type. The constant tolerance type is defined as statistical distribution data that is set to a fixed value at the part-on-tolerance, as defined by the GD/T tolerance, rather than to meet the tolerance band requirements.
In addition, the rigid-flexible body assembly model of the air separation integral cooling box can be drawn in advance and then is led into a system, corresponding initial parameters including tolerance information (tolerance zone information such as tolerance and deviation of parts) and part assembly sequence are set in the system, and particularly, in order to realize the processing of the flexible body structure, a rigidity matrix and a temperature load file corresponding to the flexible body structure are also required to be led in. The rigidity matrix and the temperature load file of the flexible body can be generated by gridding the flexible body structure of the cold box and then analyzing the flexible body structure by finite elements. The gridding treatment can be realized by hypermesh software, and the finite element analysis can be realized by abaqus software.
S3, taking the tolerance value contained in each sample in the sample set obtained in the S1, the assembly environment temperature of the cold box and the applied external force as inputs of a three-dimensional tolerance analysis system, and obtaining the assembly error value and tolerance contribution degree of the designated measuring point position on the cold box through tolerance simulation analysis; and screening and dimension reduction is carried out on all tolerance variables to be optimized according to the tolerance contribution degree, so that key tolerance variables are obtained.
It should be noted that the position of the designated measuring point on the cold box is determined according to the error sensitive position corresponding to the assembly of the cold box, and can be determined according to the actual assembly process or quality inspection requirements.
It should be noted that, the assembly environment temperature and the external force of the air-separation self-contained cooling box need to be determined according to the working condition of the air-separation self-contained cooling box for actual assembly, and if no other additional external force exists in the air-separation self-contained cooling box, only the gravity of the air-separation self-contained cooling box can be considered.
It should be noted that, the purpose of screening and dimension reduction for all tolerance variables to be optimized is to reduce the number of critical tolerance variables to be optimized, and only the tolerance variables having a critical influence on the final assembly error value are reserved for optimization. The optimization variable selects tolerance contribution degree to rank the tolerance in front so as to achieve the purposes of simplifying the characteristics and highlighting the key points. The screening and dimension reduction process can be realized through a threshold method, a lowest contribution threshold value can be preset before screening and dimension reduction, tolerance variables lower than the lowest contribution threshold value in all tolerance variables to be optimized are removed, and the reserved tolerance variables are used as key tolerance variables. It is particularly noted, however, that the screening dimension reduction process is primarily directed to cases where there are more tolerance variables or where there are significantly non-critical tolerance variables, and if there are not more tolerance variables to be optimized, the corresponding minimum contribution threshold may be set to a lower value, with all tolerance variables being retained as much as possible or as whole. The tolerance contribution degree required by screening dimension reduction can be output by three-dimensional tolerance analysis software such as 3DCS and the like.
S4, based on the tolerance simulation analysis data of each sample obtained in the S3, taking a tolerance value corresponding to a key tolerance variable, the assembly environment temperature of the cold box and the external force as input of a machine learning model, taking the assembly error value obtained by the tolerance simulation analysis as a truth value label, and performing supervision training on the machine learning model to obtain the agent model.
It should be noted that, the machine learning model for training the proxy model in the present invention may be implemented by various learning models such as a neural network model or a support vector machine model. Model training belongs to the prior art, because the assembly error value of the designated measurement point position corresponding to each sample in the sample set is obtained through tolerance simulation analysis in the step S4, the assembly error value can provide a supervision signal for model training. Each training sample for training the machine learning model is a tolerance value corresponding to a key tolerance variable (other non-key variables are not required to be used as input) in each sample obtained by sampling in the step S1, the assembly environment temperature of the cold box and the external force applied to the cold box are used as sample inputs, and an assembly error value corresponding to the sample input obtained by tolerance simulation analysis is used as a truth value label (group trunk). After the machine learning model is trained until the prediction precision meets the requirement, the machine learning model can be used for carrying out the subsequent optimization process.
In one embodiment of the present invention, for the training objectives of the present invention, a machine learning model may employ a support vector regression (Support Vector Regression, SVR) model, which implements input and output fits by the following formula to construct a proxy model:
wherein x is the fit inputThe value of the sum of the values,for fitting the output values +.>Is Lagrangian multiplier +.>Representing a kernel function, b being the displacement.
The kernel function of the SVR model is a Gaussian kernel function, and the formula is as follows:
wherein:is a Gaussian kernel function, and I.I.I is the modulus of the vector, and is ++>Is the center of the kernel function.
The parameter optimization algorithm in SVR model training preferably adopts a TPE (Tree-structured Parzen Estimator) algorithm under a Bayesian optimization algorithm framework. During training, all training samples can be divided into a training set and a testing set, after iterative training is performed by using the training set, fitting effect evaluation is performed by using the testing set, the score of the testing set is a decision coefficient of a proxy model fitting testing set, the higher the score is, the better the fitting effect is, and the formula is as follows:
wherein: r is R 2 For the SVR model score, the sum of squares of u residuals, v total squared difference,for the actual viewing tag value, +.>Output value predicted for SVR model, +.>To represent the average of all the values of the visual label.
S5, optimizing all key tolerance variables by adopting an optimization algorithm according to a cost-tolerance function of the cold box and taking a measurement point error as a constraint condition and taking the minimum cost as an optimization target, wherein the assembly error value corresponding to each group of key tolerance variable feasible solutions in the optimization process is obtained by calculating a proxy model; and finally, taking the optimal solution obtained by optimization as a tolerance variable design value of the air-separation self-contained cold box.
It should be noted that, the cost-tolerance function of the above air-separation integral-package cold box can be selected or constructed according to practical situations, which belongs to the prior art, and a great deal of related literature exists for researching the cost-tolerance function, and a function form of the cold box with the corresponding type can be selected.
When the optimization algorithm is adopted to optimize all the key tolerance variables, the optimization algorithm can be realized by adopting a genetic algorithm, a Bayesian optimization algorithm and the like, and the function of the optimization algorithm is to find a set of feasible solutions meeting the optimization target in the whole solution space, and a set of key tolerance variable values which enable the cost to be minimum are found for the invention. In an embodiment of the present invention, the optimization algorithm may be a particle swarm algorithm, and the speed and position update formula of the particle swarm algorithm is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,for particle speed, +.>For particle position->Called inertial factor->And->Called acceleration constant, is generally taken as->，/>Representing interval [0,1 ]]Random number on->D-th dimension of the individual extremum of the i-th variable,>and d-th dimension representing the globally optimal solution.
The optimization objective of the optimization algorithm is to minimize the sum of the tolerance costs of all the key tolerance variables calculated based on the cost-tolerance function expressed by the formula:
in the method, in the process of the invention,tolerance for parts>For the manufacturing cost of the parts->For the total manufacturing cost of the product, min () represents the minimization of the solution.
In addition, in optimizing, besides minimizing cost, the optimization algorithm needs to meet that the measurement point error is within the allowable error range, so that the constraint condition can be set that the upper limit of the measurement point error does not exceed the maximum allowable error. The specific maximum allowable error value can be determined according to the actual process requirements.
It should be noted that, in the optimization, the proxy model allows the fitting error values corresponding to a set of key tolerance values to be obtained in a single time, and the upper limit of the measurement point errors in the constraint condition is obtained through statistics of the fitting error values under a large number of different fitting times. Therefore, for each set of feasible solutions of the key tolerance variables, wherein the value of each key tolerance variable corresponds to a tolerance zone, all the key tolerance variables can be input into the proxy model after tolerance values are selected according to the corresponding actual tolerance distribution types, an assembly error value is predicted, the prediction process can be repeated continuously to simulate the assembly process of a large number of different cold boxes in actual assembly, the distribution situation of the assembly error values obtained by a large number of different cold boxes under the set of feasible solutions is obtained, and the upper limit of the assembly error of the designated measuring point is counted. The actual tolerance distribution type of each tolerance traversal can be determined according to the processing technology, the assembly mode and the like of the parts, and meanwhile, the external force applied to the cold box and the temperature condition of the assembly environment can be determined according to the actual operation condition. In the actual optimization solving process, the actual distribution type of the tolerance and the actual working condition can be used as the reference, the value in the tolerance zone range of the part is used as the input, and the assembly error prediction is continuously carried out by transmitting the value into the optimized agent model, so that the accurate simulation of the assembly process is realized.
In order to better demonstrate the specific implementation and technical effects of the invention, the method for predicting the assembly error and optimizing the tolerance of the space-division self-contained cooling box shown in the steps S1 to S5 in the preferred implementation is applied to a specific example.
Examples
In this embodiment, the assembly error prediction and the tolerance optimization of the parts are performed for a flexible body and a rigid positioning bolt of a certain skin steel plate on the side surface of the oversized self-contained cooling box, and analysis is performed according to the method flowchart shown in fig. 2. The steps of the overall process flow are described in detail below.
Step 1, according to the preset tolerance information of a two-dimensional design drawing of a large-scale self-contained cold box side surface skin steel plate flexible body and a rigid positioning bolt assembly model, each tolerance is within the toleranceLatin hypercube sampling is performed in-band. The model involves a tolerance of 7, denoted T 1 ~T 7 Referring specifically to table 1, 3000 data were extracted for each tolerance band to form 3000 sets of assembly error prediction input data, each set of data containing 7 features, i.e., the data dimension was 3000 x 7.
And step 2, simplifying the large-scale self-contained cold box model according to the two-dimensional drawing, and then carrying out three-dimensional modeling. Then, the built three-dimensional model to be analyzed is imported into a three-dimensional tolerance analysis software 3DCS, tolerance information is added to the cold box rigid flexible body assembly model to be analyzed, and the tolerance distribution type is selected as a constant distribution type. The three-dimensional model structure of the air-split self-contained cold box is shown in fig. 3, and is integrally formed by connecting a cold box upper surface skin steel plate 2 and a cold box lower surface skin steel plate 3 through a plurality of positioning bolts 4, wherein the cold box upper surface skin steel plate 2 and the cold box lower surface skin steel plate 3 are regarded as flexible bodies, the positioning bolts 4 are regarded as rigid bodies, and a measuring point 1 of an assembly error concerned is positioned at the corner position of the upper surface of the whole cold box.
And 3, meshing the flexible structure in the assembly model by using hypermesh mesh processing software. Since the oversized self-contained cold box skin steel sheet analyzed in this example is a thin plate structure, it is divided into shell units. And then importing the generated grid file into abaqus finite element software to perform finite element analysis, and generating a rigidity matrix and a temperature load file.
And 4, importing the generated grid file, the rigidity matrix and the temperature load file into a cold box rigid-flexible body assembly model by utilizing a finite element advanced module FEA Compliant Modeler in the 3 DCS.
In the step 5, 3DCS software, a series of characteristic points are uniformly generated on the flexible body structure, the distance between the characteristic points and grid points formed by surrounding flexible body grid division (namely grid points corresponding to the rigidity matrix and the temperature load file) is calculated, the characteristic points are associated with the nearest grid points, and the purpose that the displacement and angle change generated by the stress and the temperature of the flexible body can be transmitted to the measuring points through the characteristic points is achieved. After the association, the connection condition between the characteristic points and the grid nodes is checked through a FEA Point Linking function, the difference between the characteristic points and the grid nodes is considered acceptable when the difference is smaller than 5mm, and otherwise, the characteristic points need to be continuously subdivided. After confirmation, the connection condition in this embodiment is qualified and valid.
And 6, adding assembly steps to the parts in the 3DCS software according to the actual assembly sequence, process and operation type of the rigid and flexible body of the large-scale cold box. And then clicking a Nominal Build assembly button to observe whether the assembly process achieves the expected assembly effect, and particularly observing whether the flexible body assembly is deformed under the influence of force and temperature. In the simulation assembly process, the flexible body deforms due to welding force, and after clamping constraint is released, a rebound effect is generated, so that the deformation condition in the welding process is met, and the process is reasonable. Further generating an assembly tolerance transfer cloud chart, observing whether tolerance transfer meets expectations and meets requirements, checking whether tolerance and assembly steps are added correctly, and combining the cloud chart to know that part tolerance really affects a measuring point and generates errors, so that the embodiment tolerance and the assembly steps are added correctly.
And 7, determining the distance between the upper left tip ends of the upper and lower thin plates as an assembly error measuring point, as shown by a measuring point 1 marked in fig. 3. And (3) carrying out batch processing by using a dcs_ doe _form module in the 3DCS, wherein variables comprise data obtained by sampling in the step (1), namely, tolerance values of 7 tolerance variables, assembly environment temperature of the cold box and external force condition of the cold box, and output data are measurement point assembly error values corresponding to a group of input data and ranking of each tolerance contribution degree. The present embodiment has fewer tolerance variables, only 7 tolerance variables and in order to optimize all tolerances, all tolerances are treated as critical tolerance variable inputs for model training, prediction and tolerance optimization in subsequent optimizations.
And 8, taking 7 tolerance variables, the assembly environment temperature of the cold box and the external force condition of the cold box as sample input of the proxy model, and taking the assembly error value of the corresponding measuring point as sample output of the proxy model to form a training sample. The data generated in step 7 was applied to 7:3, forming a training set and a testing set by proportion division, and training the SVR proxy model of the support vector machine. Wherein regularization parameter c=1, epsilon=0 in the insensitive loss function, model training set score of 0.911, test set score of 0.789, and average absolute error rate of 9.450%.
And 9, taking the score of the test set as an objective function, and optimizing the super-parameters of the agent model by adopting a TPE optimization algorithm to obtain an optimized agent model. Wherein regularization parameter c= 2.6493, epsilon= 0.05801 in the insensitive loss function, model training set score of 0.987, test set score of 0.861, average absolute error rate of 6.891%. Obtaining a trained surrogate model predictor (y p ) And the true value (y t ) The statistical histograms of (a) are shown in fig. 4 and 5, wherein fig. 4 is training set data and fig. 5 is test set data. The statistics data, the model score value and the error rate can be known, the prediction performance of the optimized agent model is improved, the assembly error of the rigid flexible body of the large-scale cold box can be well predicted, and the model is effective.
And step 10, determining that tolerance types of all parts are standard normal distribution according to the processing technology, the assembly mode and the like of the parts, wherein external force is only influenced by earth gravity, the temperature is 20 ℃ ambient temperature, and determining that the simulation assembly time of a single feasible solution is 3000 times. In 3000 simulations, the actual distribution type of the tolerance and the actual working condition are taken as references, namely, input data of all 3000 groups of tolerance values are required to be normally distributed in a corresponding tolerance zone, and the input data are transmitted into an optimized proxy model to predict assembly errors. The setting conditions of the assembly tolerance of each part are shown in the following table:
TABLE 1 Assembly tolerance set conditions
After 3000 assembly error predictions for one feasible solution, the error results are shown in the following table. The distance between the two points is not more than 1.5mm, and the upper limit error of the measuring point is predicted to be 1.83mm through the proxy model, namely, the condition that the cold box assembly is possibly unqualified under the tolerance condition is required, so that the tolerance of the cold box assembly needs to be optimally designed.
Table 2 prediction error statistics
And the possibility of unqualified assembly errors is obtained by the prediction model, so that the error requirement of the measurement point is used as an optimization problem constraint condition for optimization in the subsequent optimization. Further, in order to ensure that the processing cost of the oversized self-contained cooling box is the lowest in constraint requirements, a cost-tolerance mathematical model of the oversized cooling box rigid flexible body is constructed, wherein the functional relation between the tolerance and the cost is shown in the following table:
TABLE 3 cost-tolerance function
The tolerance cost function of this embodiment is obtained by combining tables 2 and 3:
wherein T is i For each tolerance zone code shown in table 2, the tolerance zone values before optimization are carried in, and the tolerance cost of the cold box rigid-flexible body model before optimization is obtained as follows:
and then, performing tolerance optimization design by using a particle swarm optimization algorithm, taking the minimum tolerance cost as an optimization target, taking the measurement point error requirement as a constraint condition and taking the tolerance zone range as a variable. Because the optimization problem with the constraint is solved, a penalty function is added to the particle swarm algorithm to achieve the constraint purpose. The penalty function is to add an obstacle function to the original solution, and a great value is given to the infeasible point or the point which is attempted to pass through the boundary and escape, namely, the constraint is converted into the solution of the unconstrained optimization problem. The mathematical expression for optimizing the model is therefore as follows:
wherein:representing the upper limit of the measurement point error for a set of feasible solutions T.
In this embodiment, the number of particles is set to 40, the inertia constant is 0.7, and the acceleration constant c 1 =c 2 =2, performing tolerance design optimization, and performing optimization iteration for 200 cost changes as shown in fig. 6 to obtain a final optimization result of the tolerance, and performing carry operation on the tolerance due to limitation of actual machining conditions to obtain a corrected tolerance as shown in the following table 4:
table 4 tolerance after optimization
To verify the optimization result, the optimized tolerance band was input to the proxy model, and 3000 assembly simulation predictions were performed as well, and the prediction error statistics were obtained as shown in table 5 below. As shown in Table 5, the upper limit error is 1.49mm, which is smaller than the required 1.5mm, so that the error variation is within the required range, the optimum design changes the unqualified assembly condition into qualified one, and the optimum design is effective.
Table 5 statistical results of prediction errors after optimization
Bringing the tolerance data after the optimal design into a cost function, and obtaining the cost after the optimal design as follows:
the cost before the optimization design is 12.7881, the cost after the optimization is 11.3553, the cost is reduced by 11.204%, and the tolerance cost is optimized. In summary, the embodiment of the invention shows that the method can accurately predict the assembly error of the rigid flexible body of the large-scale cold box, and achieves the purposes of optimizing tolerance and reducing production cost.
In summary, the method of connecting the finite element rigidity and the temperature matrix grid points by using the object feature points on the basis of considering three-dimensional tolerance transmission by adopting a support vector machine algorithm in the embodiment further considers the influence of stress and temperature change on assembly errors, and generates a proxy model for predicting the assembly errors of the rigid flexible body of the large cold box. Based on the proxy model, adopting TPE algorithm to optimize the proxy model. On the basis of the optimized agent model, an extra-large self-contained cold box cost-tolerance function is further constructed, and the tolerance production cost is optimized and designed by utilizing a particle swarm algorithm. The agent model is used for greatly improving the optimization design efficiency and reducing the optimization difficulty, and can also effectively reduce the production and processing cost, thereby having important significance for guaranteeing the whole manufacturing quality of the large-sized cold box.
The above embodiment is only a preferred embodiment of the present invention, but it is not intended to limit the present invention. Various changes and modifications may be made by one of ordinary skill in the pertinent art without departing from the spirit and scope of the present invention. Therefore, all the technical schemes obtained by adopting the equivalent substitution or equivalent transformation are within the protection scope of the invention.
Claims (8)
1. The method for predicting the assembly error and optimizing the tolerance of the air separation self-contained cooling box is characterized by comprising the following steps of:
s1, sampling all tolerance variables to be optimized in the assembly process of the air separation integral cold box parts in a respective tolerance zone range to form a sample set, wherein each sample consists of a group of tolerance values of all the tolerance variables to be optimized;
s2, adding tolerance information of all tolerance variables to be optimized to a rigid-flexible body assembly model of the cold box in a three-dimensional tolerance analysis system, and simultaneously importing a rigidity matrix and a temperature load file of a flexible body in the model; generating feature points on the flexible body uniformly, and associating each feature point with the nearest flexible body grid point; setting the assembly sequence of parts in the rigid and flexible body assembly model according to actual procedures;
s3, taking the tolerance value contained in each sample in the sample set, the assembly environment temperature of the cold box and the applied external force as inputs of a three-dimensional tolerance analysis system, and obtaining the assembly error value and tolerance contribution degree of the position of the appointed measuring point on the cold box through tolerance simulation analysis; screening and dimension reduction are carried out on all tolerance variables to be optimized according to the tolerance contribution degree, and key tolerance variables are obtained;
s4, based on tolerance simulation analysis data of each sample obtained in the S3, taking a tolerance value corresponding to a key tolerance variable, the assembly environment temperature of the cold box and the external force applied to the cold box as input of a machine learning model, taking the assembly error value obtained by the tolerance simulation analysis as a truth value label, and performing supervision training on the machine learning model to obtain a proxy model; the machine learning model adopts a support vector regression model, the kernel function adopts a Gaussian kernel function, and a parameter optimization algorithm during model training adopts a TPE algorithm;
s5, optimizing all key tolerance variables by adopting an optimization algorithm according to a cost-tolerance function of the cold box and taking a measurement point error as a constraint condition and taking the minimum cost as an optimization target, wherein the assembly error value corresponding to each group of key tolerance variable feasible solutions in the optimization process is obtained by calculating a proxy model; and finally, taking the optimal solution obtained by optimization as a tolerance variable design value of the air-separation self-contained cold box.
2. The method for predicting assembly errors and optimizing tolerances of a space-division self-contained cooling box according to claim 1, wherein in the step S1, the sampling method is Latin hypercube sampling.
3. The method for predicting assembly errors and optimizing tolerances of a space division self-contained cooling box according to claim 1, wherein the three-dimensional tolerance analysis system is implemented by 3DCS software.
4. The method for predicting assembly errors and optimizing tolerances of a space division self-contained cooling tank according to claim 1, wherein the tolerance distribution type of tolerance information added by a rigid-flexible body assembly model in the three-dimensional tolerance analysis system is set to be a constant type.
5. The method for predicting assembly errors and optimizing tolerances of a space-division self-contained cold box according to claim 1, wherein in the step S2, the stiffness matrix and the temperature load file of the flexible body are generated by gridding the flexible body structure of the cold box and then performing finite element analysis.
6. The method for predicting assembly errors and optimizing tolerances of air-separation self-contained cooling boxes according to claim 1, wherein in the step S3, a minimum contribution threshold is preset before screening and dimension reduction, tolerance variables lower than the minimum contribution threshold in all tolerance variables to be optimized are removed, and the reserved tolerance variables are used as key tolerance variables.
7. The method for predicting assembly errors and optimizing tolerances of a space-division self-contained cooling box according to claim 1, wherein in S5, the optimization algorithm is a particle swarm optimization algorithm.
8. The method for predicting assembly errors and optimizing tolerances of a space-division self-contained cooling tank according to claim 1, wherein in S5, the constraint condition of the optimization algorithm is that the upper limit of the error of the measuring point does not exceed the maximum allowable error, and the optimization target is that the sum of the tolerance costs of all key tolerance variables calculated based on the cost-tolerance function is minimum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310615208.3A CN116384257B (en) | 2023-05-29 | 2023-05-29 | Method for predicting assembly errors and optimizing tolerance of air separation integral cold box |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310615208.3A CN116384257B (en) | 2023-05-29 | 2023-05-29 | Method for predicting assembly errors and optimizing tolerance of air separation integral cold box |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116384257A CN116384257A (en) | 2023-07-04 |
| CN116384257B true CN116384257B (en) | 2023-09-29 |
Family
ID=86969732
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310615208.3A Active CN116384257B (en) | 2023-05-29 | 2023-05-29 | Method for predicting assembly errors and optimizing tolerance of air separation integral cold box |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116384257B (en) |
Citations (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201411277A (en) * | 2012-08-06 | 2014-03-16 | Asml Netherlands Bv | Methods for providing lithography features on a substrate by self-assembly of block copolymers |
| CN106202731A (en) * | 2016-07-12 | 2016-12-07 | 南京理工大学 | Bridge crane multi-flexibl e dynamics structural optimization method |
| WO2019048506A1 (en) * | 2017-09-08 | 2019-03-14 | Asml Netherlands B.V. | Training methods for machine learning assisted optical proximity error correction |
| CN110096762A (en) * | 2019-04-11 | 2019-08-06 | 天津大学 | A kind of prediction of lathe rigging error and control method |
| CN110161968A (en) * | 2019-06-14 | 2019-08-23 | 重庆邮电大学 | A kind of numerical control machining tool heat error prediction technique based on packaging type principle |
| CN110457783A (en) * | 2019-07-24 | 2019-11-15 | 武汉理工大学 | A kind of the rigging error analysis and tolerance optimization method of parallel connection lifting mechanism |
| WO2020000248A1 (en) * | 2018-06-27 | 2020-01-02 | 大连理工大学 | Space reconstruction based method for predicting key performance parameters of transition state acceleration process of aircraft engine |
| CN111784061A (en) * | 2020-07-07 | 2020-10-16 | 广东电网有限责任公司 | Training method, device and equipment for power grid engineering cost prediction model |
| CN111931404A (en) * | 2020-07-31 | 2020-11-13 | 长沙理工大学 | Bore complex surface contact collision response prediction method based on self-optimization CNN |
| CN111985122A (en) * | 2020-07-07 | 2020-11-24 | 北京机电工程研究所 | Part tolerance optimization design method |
| AU2020103882A4 (en) * | 2020-12-03 | 2021-02-11 | Zhimin Li | A Method for Allocating Assembly Tolerance of car body considering the Influence of Welding Deformation |
| KR102219026B1 (en) * | 2020-11-27 | 2021-02-23 | (주) 토즈 | Ship design error prediction system through model learning |
| WO2021055847A1 (en) * | 2019-09-18 | 2021-03-25 | 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 |
| CN113111454A (en) * | 2021-04-01 | 2021-07-13 | 浙江工业大学 | RV reducer dynamic transmission error optimization method based on Kriging model |
| CN113204851A (en) * | 2021-06-07 | 2021-08-03 | 中国核动力研究设计院 | Tolerance optimization method for all-welded assembly |
| CN113486470A (en) * | 2021-07-15 | 2021-10-08 | 西北工业大学 | Assembly body pose calculation method considering non-ideal surface contact state |
| CN113705094A (en) * | 2021-08-27 | 2021-11-26 | 江苏杰瑞信息科技有限公司 | Ship fuel oil pipeline fault prediction method based on PSO-GRU |
| CN114048557A (en) * | 2021-10-25 | 2022-02-15 | 安徽科技学院 | Tolerance optimization design method for key parts of multi-axis numerical control machine tool |
| CN114529040A (en) * | 2022-01-05 | 2022-05-24 | 南京理工大学 | On-line prediction method for assembly error of electromechanical product |
| CN115203865A (en) * | 2022-09-19 | 2022-10-18 | 南京理工大学 | Product assembly process mechanical performance online prediction method based on digital twinning |
| CN115759175A (en) * | 2022-11-28 | 2023-03-07 | 武汉大学 | Approximation method of detail level grid model based on approximate global optimization |
| CN116050007A (en) * | 2022-12-08 | 2023-05-02 | 浙江大学 | Complex equipment optimization design method based on hybrid self-adaptive sampling agent model |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8000945B2 (en) * | 2008-08-08 | 2011-08-16 | The United States Of America Represented By The Secretary Of The Navy | Method for determining heterogeneous bottom friction distributions using a numerical wave model |
| US11124317B2 (en) * | 2018-01-26 | 2021-09-21 | Dalian University Of Technology | Method for prediction of key performance parameters of aero-engine in transition condition |
| CN111475903B (en) * | 2019-01-07 | 2024-06-18 | 哈尔滨工业大学 | Multistage part dynamic characteristic step-by-step measuring and distributing method for large-scale high-speed rotary equipment based on multi-bias error synchronous compensation |
| US20220076159A1 (en) * | 2020-09-10 | 2022-03-10 | National Technology & Engineering Solutions Of Sandia, Llc | Entity modification of models |
-
2023
- 2023-05-29 CN CN202310615208.3A patent/CN116384257B/en active Active
Patent Citations (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201411277A (en) * | 2012-08-06 | 2014-03-16 | Asml Netherlands Bv | Methods for providing lithography features on a substrate by self-assembly of block copolymers |
| CN106202731A (en) * | 2016-07-12 | 2016-12-07 | 南京理工大学 | Bridge crane multi-flexibl e dynamics structural optimization method |
| WO2019048506A1 (en) * | 2017-09-08 | 2019-03-14 | Asml Netherlands B.V. | Training methods for machine learning assisted optical proximity error correction |
| WO2020000248A1 (en) * | 2018-06-27 | 2020-01-02 | 大连理工大学 | Space reconstruction based method for predicting key performance parameters of transition state acceleration process of aircraft engine |
| CN110096762A (en) * | 2019-04-11 | 2019-08-06 | 天津大学 | A kind of prediction of lathe rigging error and control method |
| CN110161968A (en) * | 2019-06-14 | 2019-08-23 | 重庆邮电大学 | A kind of numerical control machining tool heat error prediction technique based on packaging type principle |
| CN110457783A (en) * | 2019-07-24 | 2019-11-15 | 武汉理工大学 | A kind of the rigging error analysis and tolerance optimization method of parallel connection lifting mechanism |
| WO2021055847A1 (en) * | 2019-09-18 | 2021-03-25 | 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 |
| CN111784061A (en) * | 2020-07-07 | 2020-10-16 | 广东电网有限责任公司 | Training method, device and equipment for power grid engineering cost prediction model |
| CN111985122A (en) * | 2020-07-07 | 2020-11-24 | 北京机电工程研究所 | Part tolerance optimization design method |
| CN111931404A (en) * | 2020-07-31 | 2020-11-13 | 长沙理工大学 | Bore complex surface contact collision response prediction method based on self-optimization CNN |
| KR102219026B1 (en) * | 2020-11-27 | 2021-02-23 | (주) 토즈 | Ship design error prediction system through model learning |
| AU2020103882A4 (en) * | 2020-12-03 | 2021-02-11 | Zhimin Li | A Method for Allocating Assembly Tolerance of car body considering the Influence of Welding Deformation |
| CN113111454A (en) * | 2021-04-01 | 2021-07-13 | 浙江工业大学 | RV reducer dynamic transmission error optimization method based on Kriging model |
| CN113204851A (en) * | 2021-06-07 | 2021-08-03 | 中国核动力研究设计院 | Tolerance optimization method for all-welded assembly |
| CN113486470A (en) * | 2021-07-15 | 2021-10-08 | 西北工业大学 | Assembly body pose calculation method considering non-ideal surface contact state |
| CN113705094A (en) * | 2021-08-27 | 2021-11-26 | 江苏杰瑞信息科技有限公司 | Ship fuel oil pipeline fault prediction method based on PSO-GRU |
| CN114048557A (en) * | 2021-10-25 | 2022-02-15 | 安徽科技学院 | Tolerance optimization design method for key parts of multi-axis numerical control machine tool |
| CN114529040A (en) * | 2022-01-05 | 2022-05-24 | 南京理工大学 | On-line prediction method for assembly error of electromechanical product |
| CN115203865A (en) * | 2022-09-19 | 2022-10-18 | 南京理工大学 | Product assembly process mechanical performance online prediction method based on digital twinning |
| CN115759175A (en) * | 2022-11-28 | 2023-03-07 | 武汉大学 | Approximation method of detail level grid model based on approximate global optimization |
| CN116050007A (en) * | 2022-12-08 | 2023-05-02 | 浙江大学 | Complex equipment optimization design method based on hybrid self-adaptive sampling agent model |
Non-Patent Citations (2)
| Title |
|---|
| 一种粒子群模糊支持向量机的航天器参量预测方法;顾胜;魏蛟龙;皮德常;;宇航学报(11);正文全文 * |
| 支持向量机替代模型的遗传优化设计;向国齐;严志坚;黄大贵;;电子科技大学学报(03);正文全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116384257A (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hu et al. | Engineering design under uncertainty and health prognostics | |
| CN106951695A (en) | Plant equipment remaining life computational methods and system under multi-state | |
| Khodaygan | Meta-model based multi-objective optimisation method for computer-aided tolerance design of compliant assemblies | |
| CN113393211A (en) | Method and system for intelligently improving automatic production efficiency | |
| CN116561681A (en) | Digital twinning-based fusion type fault diagnosis method | |
| Al-Aomar | A robust simulation-based multicriteria optimization methodology | |
| Imen et al. | Generation of disassembly plans and quality assessment based on CAD data | |
| CN114692507A (en) | Counting data soft measurement modeling method based on stacking Poisson self-encoder network | |
| CN116384257B (en) | Method for predicting assembly errors and optimizing tolerance of air separation integral cold box | |
| CN111861002A (en) | Building cold and hot load prediction method based on data-driven Gaussian learning technology | |
| CN116360339A (en) | Numerical control machine tool reliability optimization method based on digital twin model of meta-action unit | |
| CN107480381B (en) | The method of response surface model is built based on simulated annealing and applies its system | |
| CN114329805A (en) | Connecting piece multidisciplinary collaborative design optimization method based on self-adaptive agent model | |
| CN115640720A (en) | Self-gravitation simulation method based on distance control grid encryption | |
| CN112990601B (en) | Worm wheel machining precision self-healing system and method based on data mining | |
| CN109858699B (en) | Water quality quantitative simulation method and device, electronic equipment and storage medium | |
| CN112488248A (en) | Method for constructing proxy model based on convolutional neural network | |
| Shen et al. | Positioning failure error identification of industrial robots based on particle swarm optimization and Kriging surrogate modeling | |
| Yu-zhao et al. | Price forecasting algorithm for coal and electricity based on PSO and RBF neural network | |
| Jones et al. | Bayes linear analysis of risks in sequential optimal design problems | |
| CN115081200B (en) | Acceleration factor and failure boundary domain analysis method for complex equipment | |
| Song et al. | Equipment spare parts demand forecasting model based on grey neural network | |
| CN113449258B (en) | Quality evaluation method and device for intelligent electric meter, terminal equipment and storage medium | |
| CN112560898B (en) | Load space-time prediction method based on deep learning | |
| Akinlana | New Developments in Statistical Optimal Designs for Physical and Computer Experiments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |