In-process angle measurement and control for flexible sheet metal manufacture

A new method of controlling springback in small-radius pressbrake bending operations has been developed. This method provides a more accurate bending process, necessary for the further development of precision, small-lot sheet metal assembly manufacture. The pursuit of this research has led to the development of an inexpensive, high-resolution, on-line angle sensor. In addition, a simplified analytic model of the bending process was developed to predict springback in terms of material and tooling geometry variables. Finally, a springback control system has been developed with demonstrated accuracy of one-third of a degree in right-angle bends for cold-rolled steel samples covering a range of material properties.     

Design of a sensor for on-line measurement of loaded bend angle for pressbrake control

Controlling springback in small radius pressbrake bending operations is motivated by the need to produce small lot parts of high quality. A new technique for springback control has been developed based on a simplified analytic model of material and tooling geometry variables. This technique requires the on-line measurement of loaded angle with a robust, high resolution optical sensor which is insensitive to material surface finish. The design of the sensor minimizes systematic error due to placement on the press bed. Loaded angle measurement accuracy of less than one arc minute is achieved. In combination with a press ram position control scheme, this sensor provides a more accurate bending process necessary for the further development of precision, small-lot sheet metal assembly manufacture.     

Deep Learning in Sheet Metal Bending With a Novel Theory-Guided Deep Neural Network

Sheet metal forming technologies have been intensively studied for decades to meet the increasing demand for lightweight metal components. To surmount the springback occurring in sheet metal forming processes, numerous studies have been performed to develop compensation methods. However, for most existing methods, the development cycle is still considerably time-consumptive and demands high computational or capital cost. In this paper, a novel theory-guided regularization method for training of deep neural networks (DNNs), implanted in a learning system, is introduced to learn the intrinsic relationship between the workpiece shape after springback and the required process parameter, e.g., loading stroke, in sheet metal bending processes. By directly bridging the workpiece shape to the process parameter, issues concerning springback in the process design would be circumvented. The novel regularization method utilizes the well-recognized theories in material mechanics, Swift’s law, by penalizing divergence from this law throughout the network training process. The regularization is implemented by a multi-task learning network architecture, with the learning of extra tasks regularized during training. The stress-strain curve describing the material properties and the prior knowledge used to guide learning are stored in the database and the knowledge base, respectively. One can obtain the predicted loading stroke for a new workpiece shape by importing the target geometry through the user interface. In this research, the neural models were found to outperform a traditional machine learning model, support vector regression model, in experiments with different amount of training data. Through a series of studies with varying conditions of training data structure and amount, workpiece material and applied bending processes, the theory-guided DNN has been shown to achieve superior generalization and learning consistency than the data-driven DNNs, especially when only scarce and scattered experiment data are available for training which is often the case in practice. The theory-guided DNN could also be applicable to other sheet metal forming processes. It provides an alternative method for compensating springback with significantly shorter development cycle and less capital cost and computational requirement than traditional compensation methods in sheet metal forming industry.      

口腔正畸弓丝矫治器自动制备成形中的 回弹补偿方法

在传统的口腔正畸临床治疗中,弓丝矫治器主要依赖于正畸医师手工弯制来完成,这不仅要求医师有长时间的训练过程,同时难以达到理想的个性化及精度要求,也增加了患者复诊次数及椅旁等待时间。据此,我们研发机器人系统来实现自动化的正畸弓丝矫治器制备。其中的一个难点问题是,弓丝材料的高弹性引起的回弹现象严重影响了弓丝矫治器的成形精度。因此,提出一种基于过弯预补偿模型和回弹力在线检测的过弯补偿方法,来消除弯制过程中的弓丝回弹现象。该方法首先通过一个自行设计的弯制回弹装置测量回弹前后的角度差,并将回弹后的角度作为目标成形角,以此建立成形角过弯量预补偿模型;然后通过力传感器进行弓丝回弹力在线检测,建立基于零力状态判断的自动弯制控制方法。上述方法在所研发的弓丝弯制机器人系统中进行弯制实验测试。结果表明,它能够最大限度地减小回弹效应,实现满足临床精度要求的正畸弓丝自动弯制成形。     

Response surface methodology for design of sheet forming parameters to control springback effects

This paper deals with the optimization of tools geometry in sheet metal forming in order to reduce the springback effects after forming. A response surface method (RSM) based on diffuse approximation is used; this technique has been proved more efficient than classical gradient based methods since it requires fewer iterations and convergence is guaranteed especially for nonlinear problems. A new improved Inverse Approach for the stamping simulation based on DKTRF shell element is presented. In the new version, the strains and stresses due to bending and unbending effects are calculated analytically from the final workpiece, especially on the die entrance radii for curvature changes. The bending/unbending moments and the final shape are used to calculate springback using a second incremental Approach based on the Updated Lagrangian Formulation. The benchmark on the “U” bending problem of NUMISHEET’93 has been used to validate the method, good results on the elimination of springback have been obtained. The final results are validated using STAMPACK^® and ABAQUS^® commercial codes.     

Integration of feedforward neural network and finite element in the draw-bend springback prediction

To achieve accurate results, current nonlinear elastic recovery applications of finite element (FE) analysis have become more complicated for sheet metal springback prediction. In this paper, an alternative modelling method able to facilitate nonlinear recovery was developed for springback prediction. The nonlinear elastic recovery was processed using back-propagation networks in an artificial neural network (ANN). This approach is able to perform pattern recognition and create direct mapping of the elastically-driven change after plastic deformation. The FE program for the sheet metal springback experiment was carried out with the integration of ANN. The results obtained at the end of the FE analyses were found to have improved in comparison to the measured data.     

Springback equation of small curvature plane bending

A new analytical method for springback of small curvature plane bending is addressed with unloading rule of classical elastic-plastic theory and principle of strain superposition.We start from strain analysis of plane bending which has initial curvature,and the theoretic derivation is on the widely applicable basic hypotheses.The results are unified to geometry constraint equations and springback equation of plane bending,which can be evolved to straight beam plane bending and pure bending.The expanding and...     

自适应模糊神经网络在板料弯曲回弹预测中的应用

回弹是板料冲压成形中影响工件质量的重要因素，因为它是一个多变量相互作用的高度非线性问题，至今在解析和数值方法中未能找到一个很有效的解决途径。该文提出利用自适应模糊神经网络(ANFIS)对非线性问题的良好逼近能力，采用基于有限元方法获得训练样本，经训练后得到具有回弹预测能力的ANFIS模型。实验验证了该方法的有效性。     

Multi-objective optimization of stainless steel 304 tube laser forming process using GA

Laser forming is one of the most recent forming processes developed which uses a laser beam to induce a deliberate thermal stress on a workpiece to form a sheet metal. Accordingly, bending tubes using laser beam have attracted the attention of many engineers. In this paper, we studied the effects of various laser beam parameters on the tube bending process. To investigate the effects of all the parameters, we performed a large number of analyses and generated applicable tube laser bending data. We utilized Taguchi design of experiment method to manage the finite element simulation of the laser forming process. Subsequently, to have an easier, but more flexible and more complete laser forming data bank, we employed artificial neural networks to predict the required tube bending parameters for the proposed forming criteria. Finally, we used genetic algorithm programming to solve the multi-objective optimization with respect to the laser forming parameters. The objectives include maximum bending angle, minimum ovality, minimum thickening, and minimum forming energy consumption. The results from this study indicate that we can use applied data tables to find the optimum tube laser forming parameters. The outcome of the numerical experiments is consistent with the existing literature on the laser forming process.

     

Optimization of variable blank holder force trajectory for springback reduction via sequential approximate optimization with radial basis function network

Springback is one of the major defects in sheet metal forming. Variable blank holder force (VBHF) approach is one of the effective ways for the springback reduction. In this paper, the VBHF trajectory is optimized to reduce the springback by a sequential approximate optimization (SAO) with radial basis function (RBF) network. The U-shaped forming in NUMISHEET’93 is employed to determine an optimum VBHF trajectory, for example. In this paper, the bending moment is taken as the objective function. The tearing of sheet during the forming is considered as the design constraint, and the forming limit diagram (FLD) is employed to evaluate the design constraint quantitatively. It has been found from numerical results that the optimal VBHF trajectory can drastically reduce the springback in comparison with various VBHF trajectories. Through the theoretical examination and numerical simulation, the springback reduction of metal forming by the VBHF trajectory is discussed.     

Optimal process design of sheet metal forming for minimum springback via an integrated neural network evolutionary algorithm

The process of sheet metal forming is characterized by various process parameters. Accurate prediction of springback is essential for the design of tools used in sheet metal forming operations. In this paper, an evolutionary algorithm is presented that is capable of handling single/multiobjective, unconstrained and constrained formulations of optimal process design problems. To illustrate the use of the algorithm, a relatively simple springback minimization problem (hemispherical cup-drawing) is solved in this paper, and complete formulations of the algorithm are provided to deal with the constraints and multiple objectives. The algorithm is capable of generating multiple optimal solutions in a single run. The evolutionary algorithm is combined with the finite element method for springback computation, in order to arrive at the set of optimal process parameters. To reduce the computational time required by the evolutionary algorithm due to actual springback computations via the finite element method, a neural network model is developed and integrated within the evolutionary algorithm as an approximator. The results clearly show the viability of the use of the evolutionary algorithm and the use of approximators to derive optimal process parameters for metal forming operations.     

基于误差分治的神经网络验证

随着神经网络技术的快速发展, 其在自动驾驶、智能制造、医疗诊断等安全攸关领域得到了广泛应用, 神经网络的可信保障变得至关重要. 然而, 由于神经网络具有脆弱性, 轻微的扰动经常会导致错误的结果, 因此采用形式化验证的手段来保障神经网络安全可信是非常重要的. 目前神经网络的验证方法主要关注分析的精度, 而易忽略运行效率. 在验证一些复杂网络的安全性质时, 较大规模的状态空间可能会导致验证方法不可行或者无法求解等问题. 为了减少神经网络的状态空间, 提高验证效率, 提出一种基于过近似误差分治的神经网络形式化验证方法. 该方法利用可达性分析技术计算非线性节点的上下界, 并采用一种改进的符号线性松弛方法减少了非线性节点边界计算过程中的过近似误差. 通过计算节点过近似误差的直接和间接影响, 将节点的约束进行细化, 从而将原始验证问题划分为一组子问题, 其混合整数规划(MILP)公式具有较少的约束数量. 所提方法已实现为工具NNVerifier, 并通过实验在经典的3个数据集上训练的4个基于ReLU的全连接基准网络进行性质验证和评估. 实验结果表明, NNVerifier的验证效率比现有的完备验证技术提高了37.18%.     

Research and prediction of coordination reactions between CPA-mA and some metal ions using artificial neural networks

The complex relationship between maximum absorption wavelength (lambda(max)), molar absorptivity (epsilon) of the coordination compounds formed from m-acetyl-chlorophosphonazo (CPA-mA) and the metal ions, the acidity of coordination reaction, some properties of metal ions and the properties of more than 20 coordination compounds were studied using artificial neural networks with extended delta-bar-delta EDBD back learning algorithms in this paper. Six parameters: the pH of coordination reactions, metal ion radius (R), relative atomic weight (Wt), ionic electronic energy (E), metal ion standard Gibbs' free energy (deltaG0) and hard-soft acid-base dual scale (f) were used as input parameters, to predict the lambda(max) and epsilon of the coordination compounds. The structures of networks and the learning times were optimized. The best networks structure is 6-7-2. The optimum number of learning times is about 160,196. It is shown that the maximum relative error is no more than 6% in the testing set. The trained networks are used to simulate the complicated relations between the metal ion properties, coordination reaction conditions and the properties of coordination compounds. This optimized networks have been used for the prediction of the lembda(max) and epsilon of coordination compounds formed from Tb3+, Ho3+ with CPA-mA separately and with satisfactory results.     

Comparison of regression and artificial neural network model for the prediction of springback during air bending process of interstitial free steel sheet

This paper compares the regression and neural network modeling for predicting springback of interstial free steel sheet during air bending process. In this investigation, punch travel, strain hardening exponent, punch radius, punch velocity and width of the sheet were considered as input variables and springback as response variable. It has been observed that the ANN modeling process has been able to predict the springback with higher accuracy when compared with regression model.     

An intelligent approach to monitor and control the blanking process

Sensory fusion can be used to infer and analyse complex phenomena and detect changes in a process from a set of measurements. This paper proposes to use Neural Network modelling to monitor product quality of blanking by assessing changes in tool geometry, material quality and tool configuration. The approach may also be used to detect internal and external fractures in the products of blanking. The neural network model is fed with representative parameters, extracted from acoustic signals emitted during fracture, the corresponding waveform and force/displacement data. The neural network model is used, after training to correlate the extracted features of various signals to changes in blanking process parameters such as tool geometry, material hardness and tools clearance. A computerised data-acquisition system, using specialised software and hardware is used to draw from several transducers to record data of the experiments. A total of 192 experiments were performed, using different blanking configurations and materials. This data is used to train the neural network model, which may be integrated with other elements of the control system, to provide a fully automated, real-time supervisory system for blanking. The system could be used to sort components, shutdown the press or alert operators in the event of manufacturing-related defects in the operating cycle.     

Statistical,Evolutionary, and Neurocomputing Clustering Techniques: Cluster-Based vs Object-Based Approaches

Modern day computers cannot provide optimal solution to the clustering problem. There are many clustering algorithms that attempt to provide an approximation of the optimal solution. These clustering techniques can be broadly classified into two categories. The techniques from first category directly assign objects to clusters and then analyze the resulting clusters. The methods from second category adjust representations of clusters and then determine the object assignments. In terms of disciplines, these techniques can be classified as statistical, genetic algorithms based, and neural network based. This paper reports the results of experiments comparing five different approaches: hierarchical grouping, object-based genetic algorithms, cluster-based genetic algorithms, Kohonen neural networks, and K-means method. The comparisons consist of the time requirements and within-group errors. The theoretical analyses were tested for clustering of highway sections and supermarket customers. All the techniques were applied to clustering of highway sections. The hierarchical grouping and genetic algorithms approaches were computationally infeasible for clustering a larger set of supermarket customers. Hence only Kohonen neural networks and K-means techniques were applied to the second set to confirm some of the results from previous experiments.     

Modeling of a joint-type flexible endoscope based on elastic deformation and internal friction

Flexible endoscopes are widely used in minimally invasive surgical robot systems. Various kinematic models have been developed for describing the deformation of such endoscopes. For joint-type flexible endoscopes, most existing models neglect the effect of internal friction and cannot precisely show the shape.

In this paper, we propose a new nonlinear bending model. The rubber tube and metal net at each joint are approximated as a tube under elastic deformation and are assigned an equivalent bending stiffness. The internal friction force is also taken into account to build the moment balance equation at each joint. Groups of experiments were performed to validate the nonlinear model. The results closely confirm the model’s predictions. The model’s tip position error during the bending and unbending phases are 1.48?±?0.99?mm and 1.68?±?0.91?mm respectively; the bending angle errors are ?5.50?±?2.54° and 1.68?±?3.66°, respectively The model can also take account of the hysteresis effect of the bending, which is quite common for cable-driven flexible robots. Moreover, the model has good computational efficiency, making it suitable for real-time control.     

Constructive approximation to real function by wavelet neural networks

We present a type of single-hidden layer feed-forward wavelet neural networks. First, we give a new and quantitative proof of the fact that a single-hidden layer wavelet neural network with n + 1 hidden neurons can interpolate n + 1 distinct samples with zero error. Then, without training, we constructed a wavelet neural network X _a (x, A), which can approximately interpolate, with arbitrary precision, any set of distinct data in one or several dimensions. The given wavelet neural network can uniformly approximate any continuous function of one variable.     

Molecular distance geometry methods: from continuous to discrete

Distance geometry problems (DGP) arise from the need to position entities in the Euclidean K‐space given some of their respective distances. Entities may be atoms (molecular distance geometry), wireless sensors (sensor network localization), or abstract vertices of a graph (graph drawing). In the context of molecular distance geometry, the distances are usually known because of chemical properties and nuclear magnetic resonance experiments; sensor networks can estimate their relative distance by recording the power loss during a two‐way exchange; finally, when drawing graphs in two or three dimensions, the graph to be drawn is given, and therefore distances between vertices can be computed. DGPs involve a search in a continuous Euclidean space, but sometimes the problem structure helps reduce the search to a discrete set of points. In this paper we survey some continuous and discrete methods for solving some problems of molecular distance geometry.     

Next generation stamping dies — controllability and flexibility

In sheet metal forming, external energy is transferred to sheet metal through a set of tooling to plastically deform a workpiece. The design of the tooling and its associated forming process parameters play important roles in this manufacturing process since they directly affect the quality and cost of the final product. With increasing demands from customers, government regulations, and global competition, the controllability and flexibility of stamping dies have been challenged. In this paper, we will summarize the research activities conducted at the Advanced Materials Processing Laboratory at Northwestern University in the area of sheet metal forming. An overview of our approach towards the system will be given followed by a summary of individual projects in the areas of failure prediction, design and control of a variable binder force, and the segmented die design with local adaptive controllers.