Artificial neural networks applied to spot welding process modeling

Artificial neural networks have been studied for applicability for modeling of spot welding process. Some basic concepts relating to neural networks are explained as well as how they can be used to model welding quality parameters in terms of the welding process parameter. The performance of the neural networks for modeling is presented and evaluated using actual welding data. It is concluded that neural network modeling is a good means of estimating spot welding quality on-line.     

Prediction of mechanical property of E4303 electrode using artificial neural network

Based on the method of artificial neural network, a new approach has been devised to predict the mechanical property of E4303 electrode. The outlined predication model for determining the mechanical propert) of electrode was built upon the production data. The research leverages a back propagation algorithm as the neural network‘ s learning rule. The result indicates that there are positive correlations between the predicted results and the practical production dota. Hence, using the neural network, predication of electrode property can be realized. For the first time, this research prorides a more scientific method for designing electrode.     

Thermo-mechanical and diffusion modelling in the process of ceramic–metal friction welding

In this paper a modelling of friction welding of elastic materials with elastic–plastic metals is presented. This model has been practically verified in the process of friction welding of corundum ceramic of 97.5% Al₂O₃ content and aluminium alloy 6061-T6 as well as in the same ceramic and electrolytic copper of 99.9% Cu content. Mechanical strength of the acquired welded joints was around 30 MPa.A simulation of the process was performed by means of the finite element method using two FEM systems, namely ADINA-T and ADINA. The simulations made it possible to observe the temperature distribution and thermo-mechanical fields that take place during the process. The obtained results show that the temperature, pressure and the deformation distribution near the contact surface are non-homogeneous. It causes not even conditions to create the bond and internal stresses generation. The agreement between the numerical geometry prediction and the experimental data proves the validity of the proposed model.The performed calculations and preliminary studies on the influence of the diffusion phenomena on the welding process showed that the diffusion depth is approximately 4 μm and the calculated diffusion coefficient of Al into Al₂O₃ is 1.8 × 10^?13 m²/s.Numerical simulation of the friction welding process allows better understanding of the whole process, final products shape prediction and can be helpful during design of the process using other materials.     

Application of BP neural network to semi-solid apparent viscosity simulation

1　INTRODUCTIONSincethethixotropic propertiesofvigorouslystirredtin leadslurrieswerediscoveredbySpenceretal[1] atMIT ,investigationandapplicationofthese mi solidtechnologyhavebeendevelopedquickly .Thestudyontheapparentviscosityisnotonlyoftheoreti calbutalsoofutilizationsignificance .Manyinvesti gatorshavecarriedoutagreatdealofexperimentalresearchandputforwardmanymathematicalmodelsofthesemi solidapparentviscosity .Someofthesemodels ,whicharerelatedto popularexponentialfunctionandaclassica…     

An approach to measure the pose of RHJD4-1 arc welding robot for calibration

A measurement setup used for robot calibration was designed to meet the requirement of off-line programming technique. The robot end-effector pose (position and orientation) can be calculated indirecdy by using this setup. The setup has been applied to RHJD4-1 arc welding robot. The experimental results show the method of pose measuring using the measurement setup is simple and reliable to finish pose measuring.for robot calibration. In addition, the setup can measure the position repeatability of robot.     

Optimisation of friction stir welding process parameters to weld cast aluminium alloy A413 – an experimental approach

Abstract

A413 is a high strength eutectic aluminium silicon cast alloy used in the food, chemical, marine, electrical and automotive industries. Fusion welding of these cast alloys can lead to problems such as porosity, microfissuring and hot cracking, etc. However, friction stir welding can be used to weld these cast alloys effectively, without defects. In this investigation, an attempt was made to optimise the friction stir welding process parameters for joining the cast aluminium alloy A413. Joints were made using four levels each of tool rotation speed, welding speed and axial force. The quality of the weld zone was analysed using macrostructure and microstructure analysis. Tensile strength of the joints were evaluated and correlated with the weld zone hardness and microstructure. The joint fabricated using a tool rotation speed of 900 rev min^?1, a welding speed of 75 mm min^?1 and an axial force of 3 kN showed the best tensile strength.     

On-line quality inspection in laser blank welding using ART2 neural network

Laser blank welding is becoming more and more important in the automotive industry and the quality of the weld is critical for a successful application.A fully automated solution is required to inspect the quality of the blanks.This paper presents a vision inspection system with a CMOS camera which uses ART2 network to inspect the defects on-line to obtain the geometry and the quality of the weld seam.The neural network ART2 has the capability of self-learning from the environment. It can distinguish the defects that have been learned before and give new outputs for new defects.So ART2 network is suitable for weld quality inspection in laser blank welding.Additionally,a CO_2 laser is used for the blank butt-welding.     

Friction stir welding of aluminum to copper—An overview

Components made by joining different materials are required in various engineering applications. Fabrication of such components is a challenging task due to the vast difference in mechanical, thermal and electrical properties of the materials being used. Friction stir welding (FSW) is capable of joining dissimilar materials such as aluminum (Al) and copper (Cu) and therefore researchers have used this novel process for dissimilar joining. Consequently, several works pertaining to dissimilar joining, specifically Al-Cu, are available in the literature but they are scattered in different sources, which makes the task of gathering information about dissimilar FSW of Al-Cu cumbersome. This work has been written with an aim to provide all pertinent information related to dissimilar FSW of Al-Cu at one place to ease the problems of researchers. It comprehensively covers and summarizes the topics such as the effect of tool design and geometry, FSW process parameters, FSW strategies on mechanical properties, microstructure and formation of defects during dissimilar FSW of Al-Cu. In addition, it also presents and discusses several variants of dissimilar FSW of Al-Cu. Finally, this work not only puts forth major findings of the previous researchers but also suggests future recommendations for dissimilar FSW of Al-Cu.     

Artificial neural network approach for prediction of stress–strain curve of near β titanium alloy

In the present study, artificial neural network（ANN） approach was used to predict the stress-strain curve of near beta titanium alloy as a function of volume fractions of a and b. This approach is to develop the best possible combination or neural network（NN） to predict the stress-strain curve. In order to achieve this, three different NN architectures（feed-forward back-propagation network,cascade-forward back-propagation network, and layer recurrent network）, three different transfer functions（purelin, Log-Sigmoid, and Tan-Sigmoid）, number of hidden layers（1 and 2）, number of neurons in the hidden layer（s）,and different training algorithms were employed. ANN training modules, the load in terms of strain, and volume fraction of a are the inputs and the stress as an output.ANN system was trained using the prepared training set（a,16 % a, 40 % a, and b stress-strain curves）. After training process, test data were used to check system accuracy. It is observed that feed-forward back-propagation network is the fastest, and Log-Sigmoid transfer function is giving the best results. Finally, layer recurrent NN with a single hidden layer consists of 11 neurons, and Log-Sigmoid transfer function using trainlm as training algorithm is giving good result, and average relative error is1.27 ± 1.45 %. In two hidden layers, layer recurrent NN consists of 7 neurons in each hidden layer with trainrp as the training algorithm having the transfer function of LogSigmoid which gives better results. As a result, the NN is founded successful for the prediction of stress-strain curve of near b titanium alloy.     

Artificial neural network approach for prediction of stress–strain curve of near b titanium alloy

In the present study, artificial neural network(ANN) approach was used to predict the stress–strain curve of near beta titanium alloy as a function of volume fractions of a and b. This approach is to develop the best possible combination or neural network(NN) to predict the stress–strain curve. In order to achieve this, three different NN architectures(feed-forward back-propagation network,cascade-forward back-propagation network, and layer recurrent network), three different transfer functions(purelin, Log-Sigmoid, and Tan-Sigmoid), number of hidden layers(1 and 2), number of neurons in the hidden layer(s),and different training algorithms were employed. ANN training modules, the load in terms of strain, and volume fraction of a are the inputs and the stress as an output.ANN system was trained using the prepared training set(a,16 % a, 40 % a, and b stress–strain curves). After training process, test data were used to check system accuracy. It is observed that feed-forward back-propagation network is the fastest, and Log-Sigmoid transfer function is giving the best results. Finally, layer recurrent NN with a single hidden layer consists of 11 neurons, and Log-Sigmoid transfer function using trainlm as training algorithm is giving good result, and average relative error is1.27 ± 1.45 %. In two hidden layers, layer recurrent NN consists of 7 neurons in each hidden layer with trainrp as the training algorithm having the transfer function of LogSigmoid which gives better results. As a result, the NN is founded successful for the prediction of stress–strain curve of near b titanium alloy.     

An artificial neural network for detecting weld position in arc welding process

A kind of self-organizing artificial neural net work used for weld dete ction is presented in this paper, and its concepts and issues are discussed. The network can transform the weld visual information into typical patterns and match with the weld data collected on-line, and so realize the accurate detection of the weld position in arc welding process.     

Predicting effects of diffusion welding parameters on welded joint properties by artificial neural network

1　ＩＮＴＲＯＤＵＣＴＩＯＮＭｅｔａｌｍａｔｒｉｘｃｏｍｐｏｓｉｔｅｓ (ＭＭＣｓ) ,ｄｕｅｔｏｔｈｅｉｒｏｕｔｓｔａｎｄｉｎｇｐｒｏｐｅｒｔｉｅｓｓｕｃｈａｓｈｉｇｈｓｐｅｃｉｆｉｃｓｔｒｅｎｇｔｈ ,ｓｐｅｃｉｆｉｃｍｏｄｕｌｕｓ ,ｓｉｚｅｓｔａｂｉｌｉｔｙ ,ｈｉｇｈ ｔｅｍｐｅｒａｔｕｒｅ ｒｅｓｉｓｔａｎｔａｎｄｃｏｓｍｉｃ ｒａｙｒｅｓｉｓｔａｎｔ ,ａｒｅｗｉｄｅｌｙｕｓｅｄｉｎａｅｒｏｓｐａｃｅ ,ａｖｉａｔｉｏｎａｎｄｅｌｅｃｔｒｏｎｆｉｅｌｄｓ ,ａｎｄｂｅｃｏｍｅｔｈｅｍａｊｏｒｄｅｖｅｌ…     

紧密对接激光焊焊缝位置神经网络预测

激光焊接过程中,控制激光束准确对中焊缝是获得良好焊件的关键。以低碳钢板紧密对接激光焊(焊缝间隙不大于0.1 mm)作为研究对象,利用磁光传感法摄取焊接过程中焊缝区域磁光图像。分析焊缝区域图像特征,定义并提取紧密对接焊缝位置坐标,以前时刻的焊缝位置及其变化值作为输入量,当前时刻焊缝位置坐标作为输出量,应用神经网络建立焊缝位置的预测模型。试验结果表明,建立的前馈型神经网络能够较好地预测焊缝位置坐标,为激光焊缝及时纠偏和自动跟踪奠定基础。     

Detecting and tracking of welding line in visual plasma robotic welding

The formation of stable back beads in joining of the thick materials is important in order to achieve high-quality welded metal joints. Plasma welding uses the high welding current density, which is suitable for thick materials. The keyhole in the plasma welding depends on the pilot gas and the welding current. The voltage behaviour depends on the keyhole situation. If the torch is moved away from the welding line in conventional GMA welding, the welding voltage and the welding current are changed due to variations of the arc length. But the welding voltage does not change with the arc length in plasma welding, because the welding voltage depends on the situation of the keyhole. The authors tried to observe the weld pool on the top side by using a CCD camera. The timing of the shutter in the CCD camera is investigated to take clear images of the weld pool. As a result, the clear images of the weld pool were taken when the welding current was reduced to 30 A and an interference filter of 950 nm was attached to the CCD camera. The weld pool shape was changed with the torch position in the groove. The image-processing method was developed to detect the top of the weld pool. The torch position was estimated by processing the weld pool images. The digital control was designed to trace the welding line. The performance of the controller was verified by carrying out tracking experiments.     

基于人工神经网络的点焊熔核直径预测

从电极位移曲线上提取出的两个特征值及焊接能量值作为输入值,熔核直径为输出值,建立了基于BP算法的铝合金直流点焊熔核直径预测模型.该模型为隐层结点数为5的三层结构,隐层转移函数为Sigmoid函数,输出层的转移函数为线性函数.对实测结果与仿真结果进行了对比分析,结果表明,45.2%的预测值与实测值相差不超过0.5 mm,77.4%的预测误差不超过1 mm,94.3%的预测误差没有超过2 mm.回归分析结果为A=0.878T 0.982.     

基于PID神经网络的弧焊电源控制器的研究

为了实现弧焊过程弧焊电源输出特性到给定量的快速响应及其智能控制,提出了一种基于PID神经网络的控制器.将PID控制与BP神经网络相结合应用于弧焊过程的智能控制,既有PID控制规律的优点,又具有神经网络很强的信息综合处理能力特点.实验表明,该控制器用于弧焊电源输出特性的智能控制,具有控制响应速度快、超调量小、收敛速度快、稳态精度高的特点,满足弧焊过程控制的要求.     

基于人工神经网络技术的板料拉深成形工艺优化的数值建模

文章利用人工神经网络技术,建立了板料拉深成形工艺参数和性能评价指标之间的映射关系,为各参数的研究和优化提供了重要参考;详细分析了其中的几项关键技术:拉延筋的描述、性能评价指标的建立和正交实验法确定训练样本。最后,对某翼子板进行拉深成形建模,通过对训练后的网络进行测试,网络误差小于5%,验证了所建网络模型的正确性和实用性,为工艺优化奠定了基础。     

人工神经网络在过程工业中的应用

当前,集过程实时监测、故障诊断、模拟、优化、控制以及调度等各层次功能于一体的过程工业生产过程综合自动化成了过程工业界和学术界共同关注的热点之一.与离散产品的制造业相比,由于流程型工业过程具有强非线性的特点,给实现流程工业综合自动化造成很大的困难,因此必须引入新的思路,开发新的方法.人工神经网络是一种模拟人类思维活动的并行分布式的信息处理系统,可用于映射任何连续函数及进行模式识别,同时还具有自学习功能,实现知识的自动获取,自20世纪90年代以来已在过程系统工程领域内受到广泛的瞩目.重点讨论了人工神经网络在过程系统建模、故障诊断以及在线优化等方面的应用,以展示这种方法在流程工业综合自动化中的良好应用前景.     

Neural network modeling for weld shape process of P-GMAW

Weld shape control is a fundamental issue in automatic welding. In this paper, a double side visual system is established for pulsed gas metal arc welding （ P-GMAW） , and both topside and backside weld pool images can be captured and stored continuously in real time. By analyzing the weld shape regulation with the molten metal volume, some topside weld pool characterized parameters （WPCPs） are proposed for determining penetration in butt welding of thin mild steel. Moreover, some BP network models are established to predict backside Weld pool width with welding parameters and WPCPs as inputs.