CN-122020607-A - FDM process energy efficiency modeling and multi-objective optimization method based on BP neural network and NSGA-II

CN122020607ACN 122020607 ACN122020607 ACN 122020607ACN-122020607-A

Abstract

The invention provides an energy efficiency modeling and multi-objective optimization method for an FDM (frequency division multiplexing) process based on a BP neural network and NSGA-II, which relates to the technical field of fused deposition modeling energy efficiency prediction, and aims to analyze energy consumption characteristics in the FDM process by combining an actually measured processing power curve graph, construct an energy efficiency function and a printing efficiency function, design a five-factor three-level test by using a Box-Behnken method, study key factors influencing energy consumption by analysis of variance, respectively establish an energy consumption prediction model by a back propagation neural network, support vector regression and response surface regression fitting, determine an optimal prediction model by analysis and comparison, and comprehensively consider specific energy consumption, material deposition rate and surface roughness by combining actual processing working conditions based on the optimal prediction model, and solve the model by adopting an NSGA-II algorithm. The invention researches the energy consumption influence mechanism of the FDM process, realizes the relatively accurate prediction of energy consumption, and simultaneously provides theoretical support and optimization strategies for the optimization of the parameters of the FDM process.

Inventors

Jia shun
LI CHANGJUN
GU FU
Cao Quanyao
An Tongtong
WANG SHANG
LI ZONGSHU
ZHOU GUANGYAO
PENG JIANBIN

Assignees

山东科技大学

Dates

Publication Date: 20260512
Application Date: 20260121

Claims (7)

1. The FDM process energy efficiency modeling and multi-objective optimization method based on the BP neural network and NSGA-II is characterized by comprising the following steps: S1, analyzing energy efficiency characteristics of an FDM process; Determining energy efficiency characteristics of standby, preheating, printing and resetting stages by analyzing an actual measurement power curve of the FDM process; Constructing an energy efficiency function and a printing efficiency function of the FDM process, and calculating mass specific energy consumption and material deposition rate; s2, analyzing energy efficiency influence factors of the FDM process; Selecting an FDM 3D printer, printing consumables, nozzles, a power analyzer, a computer configured with test data analysis software, an electronic scale and a printing model to construct a test system; Designing five-factor three-level test by using Box-Behnken method, wherein five selected process parameters are layering thickness, printing speed, printing temperature, hot bed temperature and idle speed, and taking mass specific energy consumption and material deposition rate as test response; Carrying out a five-factor three-level test by using the constructed test system, obtaining test responses corresponding to five technological parameters, and analyzing the influence degree of the technological parameters on the mass specific energy consumption and the material deposition rate; s3, modeling the FDM process energy efficiency based on the BP neural network; based on experimental data, respectively establishing a prediction model of mass specific energy consumption and material deposition rate through a back propagation neural network BPNN and a support vector regression SVR, and determining an optimal FDM process energy efficiency prediction model through analyzing and comparing evaluation indexes of the FDM process energy efficiency prediction model established by the back propagation neural network BPNN, the support vector regression SVR and a response surface regression fitting RSM S4, an FDM energy efficiency optimization model and solving; Reserving layering thickness, printing speed, printing temperature, hot bed temperature and idle running speed as optimization variables according to the energy efficiency characteristics of each operation stage of the FDM process and the influence degree of process parameters on mass specific energy consumption and material deposition rate, and setting the constraint range of the optimization variables; Constructing a surface roughness prediction model based on the test data; an FDM energy efficiency optimization model aiming at minimizing mass specific energy consumption, maximizing material deposition rate and minimizing surface roughness is constructed, and the model is specifically as follows: ; Wherein LT is layering thickness, PS is printing speed, NT is printing temperature, BT is hot bed temperature, TS is idle speed, SEC is mass specific energy consumption, ra is surface roughness, MDR is material deposition rate; and solving the established FDM energy efficiency optimization model by adopting an NSGA-II algorithm to obtain a Pareto optimal solution set, and screening a global optimal solution by adopting a superior-inferior solution distance method TOPSIS.
2. The method for modeling and multi-objective optimization of energy efficiency of an FDM process based on a BP neural network and NSGA-II according to claim 1, wherein the energy efficiency function and the printing efficiency function are expressed as follows: ; ; Where SEC is the mass specific energy consumption, w is the total energy consumption of the printing process, m is the mass of the workpiece being printed, MDR is the material deposition rate, and t is the total time of the printing process.
3. The energy efficiency modeling and multi-objective optimization method based on the BP neural network and NSGA-II FDM process according to claim 2, wherein the FDM 3D printer is a huge shadow T10000 FDM 3D printer, the printing consumable is polylactic acid, the diameter of the printing consumable is 1.75mm, the diameter of the nozzle is 0.4mm, the power analyzer is a YOKOGAWA WT333E cross river power analyzer, the computer is provided with WTViewerFreePlus software, the measuring precision of the electronic scale is 0.01g, the printing model is of a hollow cube structure, the peripheral size is 20mm by 20mm, and the internal size is 17.6mm by 17.6mm.
4. The method for modeling and multi-objective optimization of energy efficiency of an FDM process based on a BP neural network and NSGA-II according to claim 3, wherein the analyzing the extent of the influence of the process parameters on the mass specific energy consumption and the material deposition rate comprises: Performing quadratic polynomial fitting on five process parameters and corresponding test responses by adopting response surface regression fitting RSM through Design-Expert 13 software, constructing a prediction model of mass specific energy consumption and material deposition rate, performing variance analysis, and evaluating the significance of main effects and interaction effects of the process parameters; According to the significance of the main effect and the interaction effect of each process parameter, the influence degree of the mass specific energy consumption is determined to be the thermal bed temperature, the layering thickness, the printing speed, the idle running speed, the printing temperature and the influence degree of the material deposition rate from large to small, and the influence degree of the mass specific energy consumption is determined to be the layering thickness, the printing speed, the thermal bed temperature, the idle running speed and the printing temperature from large to small.
5. The method for modeling and optimizing multiple objectives of FDM process energy efficiency based on a BP neural network and NSGA-II according to claim 4, wherein the determining an optimal FDM process energy efficiency prediction model by respectively establishing a prediction model of mass specific energy consumption and material deposition rate through a back propagation neural network BPNN and a support vector regression SVR and analyzing and comparing evaluation indexes of the FDM process energy efficiency prediction model established by the back propagation neural network BPNN, the support vector regression SVR and a response surface regression fitting RSM based on experimental data comprises: Based on PyCharm 2024.1 integrated development environment, constructing a BP neural network model by means of Keras API of a TensorFlow 2.12.0 deep learning framework, wherein the BP neural network model adopts a Sequential sequence model, input layer variables of the BP neural network model comprise layering thickness, printing speed, hot bed temperature, idle running speed and printing temperature, output layer variables of the BP neural network model comprise mass specific energy consumption and material deposition rate, the BP neural network model consists of three layers of hidden layers, the neuron numbers of each layer are 128, 64 and 32 in sequence, a ReLU activation function is adopted, a dataset is divided into a training set, a verification set and a testing set according to the proportion of 64%, 16% and 20%, and parameter settings of the BP neural network model are shown in a table 5; TABLE 5 BP neural network model parameter settings ; Establishing a prediction model of mass specific energy consumption and material deposition rate through support vector regression SVR, wherein the parameter setting of the support vector regression SVR is shown in a table 6; TABLE 6 SVR parameter settings for support vector regression ; And selecting error percentage PE, average absolute error MAE, mean square error MSE and decision coefficient R 2 as evaluation indexes, evaluating an FDM process energy efficiency prediction model established by a back propagation neural network BPNN, a support vector regression SVR and a response surface regression fitting RSM, and determining the FDM process energy efficiency prediction model established by the BP neural network as an optimal FDM process energy efficiency prediction model.
6. The method for modeling and multi-objective optimization of FDM process energy efficiency based on BP neural network and NSGA-II according to claim 5, wherein the constructing the surface roughness prediction model based on the experimental data comprises: By adopting a Box-Behnken test Design method, using layering thickness LT, printing speed PS, printing temperature NT and hot bed temperature BT as key technological parameters, fitting four key technological parameters and corresponding surface roughness by adopting response surface regression fitting RSM through Design-Expert 13 software, and establishing a second-order polynomial prediction model of the surface roughness, wherein the specific expression is as follows: Ra=102.21212+170.45832LT+0286418PS-0.998533NT-0.280567BT-0.243842LT*PS-0.114613LT*NT+0.447006LT*BT-0.000218PS*NT+0.000378PS*BT+0000330NT*BT-362.27834LT-0.002454PS+0.002426NT-+0.001286BT.
7. The method for modeling and optimizing energy efficiency and multiple targets of FDM process based on BP neural network and NSGA-II according to claim 6, wherein the method for screening global optimal solution by using top solution distance method comprises: And comprehensively sequencing the Pareto solution sets by adopting a superior-inferior solution distance method TOPSIS, selecting the production efficiency as a positive index, the cutting specific energy consumption and the roller abrasion depth as negative indexes, finding out the optimal and worst matrix vectors, calculating the distance between the Pareto non-dominant solution and the positive ideal solution or the negative ideal solution to obtain a comprehensive score, and determining the technological parameter combination with the highest comprehensive score as the global optimal solution.

Description

FDM process energy efficiency modeling and multi-objective optimization method based on BP neural network and NSGA-II Technical Field The invention relates to the technical field of fused deposition modeling energy efficiency prediction, in particular to an FDM process energy efficiency modeling and multi-objective optimization method based on a BP neural network and NSGA-II. Background Additive manufacturing (Additive Manufacturing, AM) is a method of manufacturing solid parts by layering build-up materials through a Computer Aided Design (CAD) model. The method has strong application potential in a plurality of high-end manufacturing fields such as aerospace, medical equipment, automobile manufacturing and the like. The fused deposition modeling (Fused Deposition Modeling, FDM) serves as an important branch of the additive manufacturing technology, and gradually becomes a hot spot direction of additive manufacturing research and application by virtue of the advantages of simplicity and convenience in operation, low cost, high material adaptability and the like. Compared with the traditional material reduction processing technology, the FDM can save 30% -50% and 20% -35% in processing time and manufacturing cost respectively, and production efficiency is remarkably improved. However, the energy efficiency problem of the FDM process is increasingly prominent behind its green manufacturing advantages, becoming an important factor limiting its large-scale popularization. FDM technology generally involves the heating and melting of thermoplastic materials, nozzle movement, and operation of a precision control system, which consumes a significant amount of energy, especially when the printing time is long or the workpiece volume is large, with a rapid rise in energy load, further exacerbating the overall energy efficiency problem. Therefore, the energy efficiency mechanism in the forming process of the FDM technology is researched, key factors influencing energy efficiency are analyzed, an energy efficiency prediction model is established, and the process parameter combination is further optimized, so that the method has important theoretical significance and engineering value for reducing energy consumption, improving energy utilization efficiency and prolonging equipment service life, and further sustainable development of the additive manufacturing technology is promoted. It should be noted, however, that the objectives of the existing research predictions are less inclusive of material deposition rates, and that the predictive performance comparisons between conventional polynomial regression models and machine learning algorithm models are less common. The existing optimization model is based on a traditional polynomial model, and less advanced modeling methods such as a neural network are adopted as modeling basis. Meanwhile, the optimization target comprehensively considers the less common of specific energy consumption, roughness and material deposition rate. Disclosure of Invention Aiming at the problems existing in the prior art, the invention provides an FDM process energy efficiency modeling and multi-objective optimization method based on a BP neural network and NSGA-II, which comprises the steps of firstly analyzing energy consumption characteristics in the process of FDM technology processing by combining an actually measured processing power curve graph, constructing an energy efficiency function and a printing efficiency function, secondly, designing a five-factor three-level test by using a Box-Behnken method in a response surface method, researching key factors influencing energy consumption by variance analysis and the like, thirdly, respectively establishing an energy consumption prediction model by using a Back Propagation Neural Network (BPNN), a Support Vector Regression (SVR) and a response surface regression fit (RSM), determining an optimal prediction model by analysis and comparison, and finally, constructing a multi-objective optimization model by combining actual processing working conditions and comprehensively considering specific energy consumption, material deposition rate and surface roughness, and solving the model by using a refined-strategy non-dominant genetic algorithm (NSGA-II). The research method researches the energy consumption influence mechanism of the FDM process, realizes relatively accurate prediction of energy consumption, and simultaneously provides theoretical support and optimization strategies for parameter optimization of the FDM process. An FDM process energy efficiency modeling and multi-objective optimization method based on a BP neural network and NSGA-II comprises the following steps: S1, analyzing the energy efficiency characteristics of the FDM process, and determining the energy efficiency characteristics of standby, preheating, printing and resetting stages by analyzing the actual measured power curve of the FDM process. The energy efficiency func