基于回归方法和神经网络的电弧增材制造单道成形形貌预测

目的 针对电弧增材制造技术实际应用中工艺参数选取困难和成形结果难预测的问题,确定高效、准确的电弧增材制造单道成形形貌预测的数学方法,以快速、方便地选取丝材电弧增材制造工艺参数并指导成形质量控制。方法 在单道单层丝材电弧增材制造实验的基础上,采用多种回归方法和神经网络方法分别建立焊接电流、电压和焊枪移动速度等多个工艺参数与增材层宽度、增材层高度及熔池深度等成形形貌参数之间的数学关系模型。结果 电弧增材制造单道成形形貌与焊接电流、电压和焊枪移动速度显著相关,且各参数间存在非线性交互作用;采用多元线性回归法可较准确地预测单道增材层宽度,但对于增材层高度和熔深的预测效果较差;神经网络可良好地处理各工艺参数间复杂的非线性关系,其对增材层宽度、增材层高度和熔深的预测平均误差率分别为4.17%、6.60%和7.01%,显著优于多元线性回归法。结论 采用神经网络法可以准确预测电弧增材制造单道成形的形貌参数,进而指导增材制造工艺参数的选取和成形质量的控制。     

基于BAS算法优化的电弧增材制造焊道尺寸预测

目的 为提高实际应用中电弧增材制造对工艺参数的选取效率及成形形貌的控制效果,建立高效且精准的成形尺寸预测模型,实现对焊道尺寸的合理预测。方法 在单层单道CMT电弧增材制造实验的基础上,建立基于天牛须搜索算法(Beetle Antennae Search,BAS)优化BP神经网络的焊道尺寸预测模型,利用BAS算法实现对BP神经网络初始权值和阈值的优化,可以实现预测不同工艺参数(焊接速度、送丝速度、干伸长)下焊道的成形尺寸(熔宽、余高)。利用试验验证BAS-BP预测模型的性能,与现有模型进行对比,结果 结果表明该模型具有较高精度的预测效果,能够有效映射工艺参数与焊道尺寸之间的非线性关系,印证了该模型具有良好的拟合和泛化能力,同时其对焊道熔宽和余高的预测误差分别不超过0.2、0.12 mm,预测平均误差率均不超过6%,相对于其他预测模型表现出较好的准确性和稳定性。结论 BAS-BP神经网络预测模型的输出误差较小,网络训练收敛速度加快,避免了过拟合及欠拟合的风险,有效提高了预测模型的泛化能力和预测精度,可以实现一定工艺参数范围内的焊道尺寸预测,为后续电弧增材的实时预测及控制参数应用提供了技术支持。     

基于神经网络的电弧增材制造铝合金力学性能预测

目的 预测不同工艺参数下电弧增材制造铝合金的力学性能。方法 通过实验建立了电弧增材制造6061铝合金及Ti C增强6061铝合金力学性能的数据集,并建立了一种以焊接电流、焊接速度、脉冲频率、TiC颗粒含量为输入,以屈服强度和抗拉强度为输出的神经网预测模型,对比了反向传播神经网络(BP)、粒子群算法优化BP神经网络(PSO-BP)、遗传算法优化BP神经网络(GA-BP)3种预测模型的精度。结果 与BP模型和PSO-BP模型相比,GA-BP预测模型具有更好的预测精度。其中,GA-BP模型预测6061铝合金屈服强度最佳结果的相关系数(R)为0.965,决定系数(R²)为0.93,平均绝对误差(Mean Absolute Error,MAE)为2.35,均方根误差(Root Mean Square Error,RMSE)为2.67;预测Ti C增强的6061铝合金抗拉强度最佳结果的R=1,R²高达0.99,MAE为0.46,RMSE为0.49,GA-BP具有良好的预测精度。结论 BP、PSO-BP、GA-BP 3种神经网络模型可以用来预测电弧增材制造...     

Utilization of neural networks for the recognition of variance shifts in correlated manufacturing process parameters

Traditional statistical process control (SPC) charting techniques were developed for use in discrete industries where independence exists between process parameters over time. Process parameters from many manufacturing industries are not independent, however, but they are serially correlated. Consequently, the power of traditional SPC charts was greatly weakened. The paper discusses the development of neural network models to identify successfully shifts in the variance of correlated process parameters. These neural network models can be used to monitor manufacturing process parameters and signal when process adjustments are needed.     

激光增材制造超高温氧化物共晶陶瓷研究进展

超高温氧化物共晶陶瓷具有优异的高温强度、高温蠕变性能、高温结构稳定性以及良好的高温抗氧化和抗腐蚀性能, 成为1400 ℃以上高温氧化环境下长期服役的新型候选超高温结构材料之一, 在新一代航空航天高端装备热结构部件中具有重要的应用前景。基于熔体生长技术, 以选择性激光熔化和激光定性能量沉积为代表的激光增材制造技术具有一步快速近净成形大尺寸、复杂形状构件的独特优势, 近年来已发展成为制备高性能氧化物共晶陶瓷最具潜力的前沿技术。本文从工作原理、成形特点、技术分类等方面概述了基于熔体生长的两种典型激光增材制造技术, 综述了激光增材制造技术在超高温氧化物共晶陶瓷制备领域的研究现状和特点优势, 重点介绍了选择性激光熔化和激光定向能量沉积超高温氧化物共晶陶瓷在激光成形工艺、凝固缺陷控制、凝固组织演化、力学性能等方面的研究进展。最后, 指出了实现氧化物共晶陶瓷激光增材制造工程化应用亟需突破的关键瓶颈, 并对该领域未来的重点发展方向进行了展望。     

Application of fuzzy neural network to laser bending process of sheet metal

Abstract

The optimisation and selection of process plans is very important for laser bending of sheet metal to achieve the anticipated bending deformation. In this paper, an adaptive fuzzy neural network has been proposed to predict the bending deformation. This network integrates the learning power of neural networks with fuzzy inference systems. During the establishing process of the energy density (composed of three process parameters: laser power, scanning velocity, and spot diameter), width, thickness of sheet, and scanning path curvature were taken as four input variables of the network. The gradient descent learning algorithm was applied to optimally adjust the weight coefficients of the neural network and the parameters of the fuzzy membership functions. Then, the trained network was used to predict the laser bending deformation. Good correlation was found between the predictive and experimental results.     

增材制造铜/钢双金属材料研究进展

铜/钢双金属材料具有力学强度高、物理化学性能优良等优势,在交通运输、电力能源和建筑工业等领域应用前景广阔。然而,传统熔铸工艺在制造铜/钢双金属材料时,容易在铜/钢界面处产生偏析现象,在一定程度上限制了铜/钢双金属材料的发展。与传统工艺相比,增材制造技术不仅能实现复杂加工零件的快速制造,而且在成形过程中较短的保温时间能缓和或消除异种金属材料界面产生的冶金缺陷,进而增强铜/钢双金属材料的力学性能。由于双金属材料是近年来的研究热点,有关增材制造铜/钢双金属材料的综述性文章较少,故综述了近年来激光、电子束及电弧增材制造技术制造铜/钢双金属材料的研究发展现状,分析了各技术的优缺点,并从制备方法、工艺参数及界面合金元素等角度,分析了影响材料界面组织性能变化的关键因素。发现在增材制造铜/钢双金属材料方面,目前激光增材制造技术主要应用于精度要求较高的小尺寸零部件,电子束增材制造技术适用于某些具有特殊性能的合金,如钛合金,而电弧增材制造技术适用于精度要求较低的大型复杂零部件。在铜/钢双金属材料增材制造过程中,界面处易形成显微组织分布不均匀、界面晶粒尺寸差异较大等现象,导致界面处产生应力集中,从而造成材料...     

Measurement and characterization of porosity in aluminium selective laser melting parts using X-ray CT

ABSTRACT

Selective laser melting (SLM) is an additive manufacturing technique which has the capability to produce complex metal parts with almost 100% density and good mechanical properties. Despite the potential benefits of SLM technology, there are technical challenges relating to the qualification and certification of the manufactured parts that limits its application in safety-critical industries, such as aerospace. Material porosity in SLM parts is detrimental for aerospace applications since it compromises structural integrity and could result in premature structural failure of parts. This paper describes the application of the non-destructive X-ray computed tomography (XCT) method to characterize the internal structure to enhance the understanding of the process parameters on material porosity and thus provide quality control of the SLM AlSi10Mg parts. An efficient and reliable XCT image processing procedure that involves image enhancement and ring artefact removal prior to image segmentation is presented. The obtained porosity level is compared with the conventional Archimedes method, showing good agreement. The characteristics of pores, such as shapes and sizes, are also discussed.     

Direct Ink Writing of Pickering Emulsions Generates Ultralight Conducting Polymer Foams with Hierarchical Structure and Multifunctionality

Porous materials with multiple hierarchy levels can be useful as lightweight engineering structures, biomedical implants, flexible functional devices, and thermal insulators. Numerous routes have integrated bottom-up and top-down approaches for the generation of engineering materials with lightweight nature, complex structures, and excellent mechanical properties. It nonetheless remains challenging to generate ultralight porous materials with hierarchical architectures and multi-functionality. Here, the combined strategy based on Pickering emulsions and additive manufacturing leads to the development of ultralight conducting polymer foams with hierarchical pores and multifunctional performance. Direct writing of the emulsified inks consisting of the nano-oxidant—hydrated vanadium pentoxide nanowires—generated free-standing scaffolds, which are stabilized by the interfacial organization of the nanowires into network structures. The following in situ oxidative polymerization transforms the nano-oxidant scaffolds into foams consisting of a typical conducting polymer—polyaniline. The lightweight polyaniline foams featured by hierarchical pores and high surface areas show excellent performances in the applications of supercapacitor electrodes, planar micro-supercapacitors, and gas sensors. This emerging technology demonstrates the great potential of a combination of additive manufacturing with complex fluids for the generation of functional solids with lightweight nature and adjustable structure-function relationships.     

Feasibility and robustness of transiently chaotic neural networks applied to the cell formation problem

Cell formation is a key issue in the design of cellular manufacturing systems. Effective grouping of parts and machines can improve considerably the performance of manufacturing cells. The transiently chaotic neural network (TCNN) is a recent methodology in intelligent computation that has the advantages of both the chaotic neural network and the Hopfield neural network. The present paper investigates the dynamics of the TCNN network and studies the feasibility and robustness of final solutions of TCNN when applied to the cell formation problem. The paper provides insight into the feasibility and robustness of TCNN for cell formation problems. It also discusses how to set the initial values of the TCNN parameters in the case of well-structured and ill-structured cell formation problems.     

Cost-tolerance analysis model based on a neural networks method

The purpose of tolerance design in product components is to produce a product with the least manufacturing cost possible, while meeting all functional requirements of the product. The product designer and process planner must fully understand the process accuracy and manufacturing cost of all kinds of manufacturing process to perform a good process plan job. Usually, the cost-tolerance model is constructed by a linear or non-linear regression analysis based on the data of the cost-tolerance experiment and to derive the correlation curve between the two. Though these correlation curves can show the relationship between manufacturing cost and tolerance, a fitting error is inevitable. In particular, there is considerable discrepancy in terms of the non-experimental data. A cost-tolerance analysis model based on a neural networks method is proposed. The cost-tolerance experimental data are used to set the training sets to establish a cost-tolerance network. Three representation modes of the cost-tolerance relationship are presented. First, the cost-tolerance relationship is derived from the grid points setting by the required tolerance accuracy. Second, a reasonable manufacturing cost of an unknown cost-tolerance experimental pair can be derived by the simulation of a cost-tolerance network. Third, an inference model based on a network's output is proposed to express the scope of the cost variation of various tolerances by means of a cost band. Comparison is also made with the high-order polynomial power function and exponential function cost-tolerance curves adopted by Yeo et al . Analytical results prove that the application of the cost-tolerance analysis model based on neural networks yields better performance in controlling the average fitting error than all conventional fitting models. The representation model using a cost band can identify precisely the possible cost variation range and reduce the chances of error in the tolerance design and cost estimation. It can thus provide important references for tolerance designers and process planners.     

Effect of support on printed properties in fused deposition modelling processes

ABSTRACT

Additive manufacturing still suffers from redundant support material usage when printing parts with overhanging features. All the supports will be removed after fabrication, resulting in wasted materials. There are many works conducted for reducing support waste by improving support strategies. However, using different support strategies may lead to different printed qualities. In this paper, the effect of support strategy on printed qualities is investigated in fused deposition modelling processes. Three different support strategies are adopted for manufacturing the same 3D part. The finished surface roughness and flexural properties are compared for evaluating different support strategies, as well as the material waste and printing time. The results show that different support strategies may result in different printed surface roughness and flexural properties. To achieve the balance between support consumption and properties of printed parts, it becomes necessary to understand the effect of supports on printed qualities for choosing a best support strategy.     

Convolutional and generative adversarial neural networks in manufacturing

Manufacturing is undergoing transformation driven by the developments in process technology, information technology, and data science. A future manufacturing enterprise will be highly digital. This will create opportunities for machine learning algorithms to generate predictive models across the enterprise in the spirit of the digital twin concept. Convolutional and generative adversarial neural networks have received some attention of the manufacturing research community. Representative research and applications of the two machine learning concepts in manufacturing are presented. Advantages and limitations of each neural network are discussed. The paper might be helpful in identifying research gaps, inspire machine learning research in new manufacturing domains, contribute to the development of successful neural network architectures, and getting deeper insights into the manufacturing data.     

Combining neural networks and genetic algorithms for optimising the parameter design of the inter-metal dielectric process

An inter-metal dielectric (IMD) is deposited between metal layers to provide isolation capability to a device and separate the different metal layers that are unnecessary in the conduction of electricity. Owing to the complicated input/response relationships of the IMD process, the void problem results in electric leakage and causes wafer scraping. In the current study, we combined neural networks, genetic algorithms (GAs) and the desirability function in order to optimise the parameter settings of the IMD process. Initially, a backpropagation (BP) neural network was developed to map the complex non-linear relationship between the process parameters and the corresponding responses. Moreover, the desirability function and GAs were employed to obtain the optimum operating parameters in respect to each response. The implementation of the proposed approach was carried out in a semiconductor manufacturing company in Taiwan, and the results illustrate the practicability of the proposed approach.     

A machine learning approach to yield management in semiconductor manufacturing

Yield improvement is one of the most important topics in semiconductor manufacturing. Traditional statistical methods are no longer feasible nor efficient, if possible, in analysing the vast amounts of data in a modern semiconductor manufacturing process. For instance, a typical wafer fabrication process has more than 1000 process parameters to record on a single wafer and one manufacturing plant may produce tens of thousands wafers a day. Traditional approaches have limits in extracting the full benefits of the data. Therefore, the manufacturing data is poorly exploited even in the most sophisticated processes. Now it is widely accepted that machine learning techniques can provide powerful tools for continuous quality improvement in a large and complex process such as semiconductor manufacturing. In this work, memory based reasoning (MBR) and neural network (NN) learning are combined for yield improvement and an integrated framework is proposed for a yield management system based on hybrid machine learning techniques. In this hybrid system of NN and MBR, the feature weight set which is calculated from the trained neural network plays the core role in connecting both learning strategies and the explanation on prediction can be given by obtaining and presenting the most similar examples from the case base. The proposed system has advantages in typical semiconductor manufacturing problems such as scalability to large datasets, high dimensions and adaptability to dynamic situations.     

基于GA–BP神经网络的汽车内饰板虚拟制造研究

目的 以某汽车内饰板为研究对象进行虚拟制造,以提前得到相对准确的工艺参数并减少成形缺陷的产生。方法 研究了工艺参数对产品拉延成形质量的影响,并确定了拉丁超立方抽样区间,在抽样区间内抽取60组样本数据,以最大减薄率为目标值,以前50组样本数据为测试集、后10组样本数据为预测集,使用基于GA–BP神经网络的遗传算法得到最优工艺参数,并将其代入有限元分析软件DYNAFORM中进行虚拟制造。结果 训练后GA–BP模型的预测值与期望值最大误差为0.299 7%,最大预测误差率为1.747 38%;遗传算法预测的最大减薄率为16.548%,虚拟制造得到的减薄率为16.167%,虚拟制造值与预测值的大小仅相差0.318%,仿真误差的误差率为2.36%。结论 虚拟制造结合先进算法的优化方法可以指导后续生产。     

3D printing vertically: Direct ink writing free-standing pillar arrays

Over the last three decades, a variety of additive manufacturing techniques have gradually gained maturity and will potentially play an important role in future manufacturing industries. Among them, direct ink writing has attracted significant attention from both material and tissue engineering areas, where the colloidal ink is extruded and dispensed according to a pre-designed path, usually in the X-Y plane with suitable increments in the Z direction. Undoubtedly, this way of disassembling geometries, simple or complex, can facilitate most of the printing process. However, for one extreme case, i.e. pillar arrays, the size resolution can deviate from both nozzle and design if the common way of slicing and additive manufacturing is used. Therefore, a different printing path is required – directly depositing pillars in a converse gravitational direction. This paper gives multiple examples of printing viscoelastic colloidal ceramic and metal inks uniaxially and periodically into free-standing and height-adjustable pillar arrays. It is expected to inspire the additive manufacturing community that more versatile degrees of freedom and complex printing paths, not confined within only complex shapes, can be achieved by ink-based 3D printing.     

Using neural networks for 3D measurement in stereo vision inspection systems

This paper presents a stereo vision inspection process which derives precise 3D measurements. Two artificial neural networks are used to facilitate the whole measurement process. At first, a simple camera calibration process is developed to derive the focal lengths and the relative information. A Hopfield neural network is used to solve the stereo matching problem, which has been constructed as an energy function. By means of a recursive process, the disparities of extracted feature points are obtained. In addition, a backpropagation neural network-based measurement error correction model for 3D measurement is proposed. It reduces the errors of 3D measurement associated with a part's orientation, position, magnitude and distance between the object and cameras. Four procedural processes are designed to implement this model. Our laboratory experiments demonstrate that the proposed measurement process has a satisfactory measurement result.     

Global optimization of a feature-based process sequence using GA and ANN techniques

Operation sequencing has been a key area of research and development for computer-aided process planning (CAPP). An optimal process sequence could largely increase the efficiency and decrease the cost of production. Genetic algorithms (GAs) are a technique for seeking to ‘breed’ good solutions to complex problems by survival of the fittest. Some attempts using GAs have been made on operation sequencing optimization, but few systems have intended to provide a globally optimized fitness function definition. In addition, most of the systems have a lack of adaptability or have an inability to learn. This paper presents an optimization strategy for process sequencing based on multi-objective fitness: minimum manufacturing cost, shortest manufacturing time and best satisfaction of manufacturing sequence rules. A hybrid approach is proposed to incorporate a genetic algorithm, neural network and analytical hierarchical process (AHP) for process sequencing. After a brief study of the current research, relevant issues of process planning are described. A globally optimized fitness function is then defined including the evaluation of manufacturing rules using AHP, calculation of cost and time and determination of relative weights using neural network techniques. The proposed GA-based process sequencing, the implementation and test results are discussed. Finally, conclusions and future work are summarized.     

MODELLING AND OPTIMIZING OF A MIG WELDING PROCESS—A CASE STUDY USING EXPERIMENTAL DESIGNS AND NEURAL NETWORKS

This paper describes an application of an integrated method using experimental designs and neural network technologies for modelling and optimizing a metal inert gas (MIG) welding process. To achieve optimization, the process parameters must be adjusted in such a way that the deviations from target are minimized while the robustness to noise and to process fluctuations are maximized. This new method consists of an experiment reference template for designing and collecting training data samples, and a parallel distributed computational adaptive neural network system to provide a powerful tool for data modelling and empirical investigations. The relevant data is established using experimental design methods and highlighted in the case study. An adaptive GaRBF neural network is used to approximate the stochastically non-linear dynamics of the welding process to optimize the basic welding parameters. The neural network is trained with welding experimental data, tested and compared in an actual welding environment in terms of its ability to determine weld quality. The results show that the proposed adaptive neural network is capable of mapping the complex relationships between the welding parameters and the corresponding output weld quality. The implementation for this case study was carried out using a ‘semi-automatic’ welding facility, to mass weld a 20″ × 0.438″ pin/box onto a 20″ × 0.5″ × 37′ pipe (tubular drilling products), in an actual workshop which makes oilfield equipment. The entire range of welding combinations that might be experienced during actual welding operations is included to study the weld quality. © 1997 by John Wiley & Sons, Ltd.