Modelling 3-D cutting dynamics with neural networks

Identification of 3-D cutting dynamics requires an expensive experimental set-up and complicated analysis. Recently, time series methods were used to model cutting dynamics. This approach allows a simpler experimental set-uup and estimates the discrete transfer functions used for simulation and/or calculation of frequency domain characteristics of the system. In this paper, the use of neural networks is proposed to model the 3-D cutting dynamics. Neural networks can be trained using the same experimental set-up used for the time series methods. However, several time series models (for different cutting speeds) can be represented with a single neural network, and cutting forces can be studied for varying cutting speed conditions. Also, four neural networks were used to store the frequency domain characteristics of the thrust direction cutting force. In this study, the estimation errors for the neural networks were less than 7% of the defined range (the difference between the maximum and minimum of the data).     

Modelling process mean and variation with MLP neural networks

Most industrial processes are intrinsically noisy and non-deterministic. To date, most multilayer perceptron (MLP)-based process models were established for process mean only. This paper proposes an approach to modelling the mean and variation of a non-deterministic process simultaneously using a MLP network. The input neurons consist of process variables and one additional neuron for the Z value. The corresponding output responses are calculated based on , s_p/√k.

The process variance s_p² is determined by pooling the individual sample variances for k experimental conditions. Each sample variance is calculated from the replicated data. The effects of a number of hidden neurons and learning algorithms are studied. Two learning algorithms are applied. They are the back-propagation with momentum (BPM) and Fletcher-Reeves (FR) algorithms. The effectiveness of the proposed approach is tested with a fictitious process and an actual manufacturing process. The test results are provided and discussed.     

Revisiting Hume-Rothery’s Rules with artificial neural networks

Hume-Rothery’s breadth of knowledge combined with a quest for generality gave him insights into the reasons for solubility in metallic systems that have become known as Hume-Rothery’s Rules. Presented with solubility details from similar sets of constitutional diagrams, can one expect artificial neural networks (ANN), which are blind to the underlying metals physics, to reveal similar or better correlations? The aim is to test whether it is feasible to predict solid solubility limits using ANN with the parameters that Hume-Rothery identified. The results indicate that the correlations expected by Hume-Rothery’s Rules work best for a certain range of copper or silver alloy systems. The ANN can predict a value for solubility, which is a refinement on the original qualitative duties of Hume-Rothery’s Rules. The best combination of input parameters can also be evaluated by ANN.     

On the chatter frequencies of milling processes with runout

The detection of undesirable vibrations in milling operations is an important task for the manufacturing engineer. While monitoring the frequency spectra is usually an efficient approach for chatter detection, since these spectra typically have a clear and systematic structure, we show in this paper that the stability of the cutting process cannot always be determined from solely viewing the frequency spectra. Specifically, the disturbing effect of the tool runout can sometimes prevent the proper determination of stability. In this paper, we show these cases can be classified by alternative analysis of the vibration signal and the corresponding Poincaré section. Floquet theory for periodic systems is used to explore the influence of runout on the structure of milling chatter frequencies. Finally, the results from theoretical analysis are confirmed by a series of experimental cutting tests.     

Process modeling with neural networks for pulsed GMAW braze welds

In this work, an improved weld-bead geometry and reduced sheet metal distortion were desired in a weld-brazing process used to finish exterior sheet metal prior to painting. Numerical modeling techniques are widely used to model welding processes, but accurate methods for modeling pulsed-welding processes are lacking. Instead, a continuous empirical model of the welding process was developed using a statistically designed experiment and a neural network for data analysis. Graphical methods were used to simulate the process, and robust parameter design techniques were used to analyze response surfaces produced by the model. Using the response surfaces, possible welding procedures were generated and tested. A new process that resulted in improved joint quality was implemented in production.     

Monitoring drill conditions with wavelet based encoding and neural networks

Encoding of thrust force signals of microdrilling operations with wavelet transformations and classification of estimated coefficients with adaptive resonance theory (ART2)-type neural networks are proposed for detection of severe tool damage just before complete tip breakage occurs. The coefficients of the wavelets were classified both directly and after a secondary encoding to reduce the humber of inputs. Direct classification of the wavelets was found to be more reliable in the sixty-one cases studied. The proposed approach was also tested with two sampling intervals. Large sampling intervals were used to inspect complete drilling cycles. Smaller sampling intervals were used to focus on thrust force variations during the motion of the machine tool table when it is driven by a stepping motor. It was found that the data collected at smaller sampling intervals were easier to classify to detect severe damage to the tool.     

Method for chatter detection with standard PLC systems

The aim of growing productivity together with increasing quality demands in machining leads to milling processes that are near their stability limits. The remaining stability reserves become smaller and the risk of unstable processing conditions like chatter increases. Unstable processes cause unwanted vibrations with high amplitudes in bearing loads. As a result, the lifetime of the bearings of the main spindle is reduced. Besides this, the surfaces of unstable processed work-pieces have unwanted chatter marks and do not fulfill the quality demands. To basically avoid unstable processing,a continuous monitoring of the process state is necessary. In this paper, a method for monitoring cutting processes using a standard programmable logic controller which is integrated in the drive controller of the machine tool spindle unit is presented. The method is real time-capable and based on the hypothesis that unstable process conditions result in a modulation of the amplitudes of the cutting forces. To detect this, an order tracking method is implemented, which uses a recursive parameter estimation algorithm together with inherently given signals of the drive controller. It is shown that the characteristic property of the used estimation algorithm and allowed aliasing lead to a reliable chatter detection even at sub-sampling. Finally some results of the first experimental investigation of the method are given.     

Prediction of flow stresses at high temperatures with artificial neural networks

1　ＩＮＴＲＯＤＵＣＴＩＯＮＧｅｎｅｒａｌｌｙａｌｌｏｙｍａｔｅｒｉａｌｓｗｉｌｌｐｒｅｓｅｎｔｓｔａｂｌｅｆｌｏｗｆｅａｔｕｒｅａｔｈｉｇｈｔｅｍｐｅｒａｔｕｒｅ ｐｌａｓｔｉｃｄｅｆｏｒｍａｔｉｏｎ ,ｎａｍｅｌｙ,ｕｎｄｅｒｃｅｒｔａｉｎｔｅｍｐｅｒａｔｕｒｅｓａｎｄｓｔｒａｉｎｒａｔｅｓｔｒｕｅｓｔｒｅｓｓｅｓ (σ)ｗｉｌｌｎｏｔａｐｐａｒｅｎｔｌｙｃｈａｎｇｅｗｉｔｈｔｈｅｃｏｎｔｉｎｕｏｕｓｉｎｃｒｅａｓｉｎｇｏｆｓｔｒａｉｎｓ (ε)ａｆｔｅｒｔｒｕｅｓｔｒａｉｎｓａｒｅｂｅｙｏｎｄｓｏｍ…     

Drill wear monitoring using neural networks

The primary objective of this research is to monitor drill wear on-line. In this paper, drill wear monitoring is carried out by measuring the thrust force and torque signals. In order to identify the tool wear conditions based on the signal measured, a neural network, using a cumulative back-propagation algorithm, is adopted. This paper also describes the experimental procedure used and presents the results obtained for establishing the neural network. The inputs to the neural network are the mean values of thrust force and torque, spindle rotational speed, feedrate and drill diameter. The neural network is trained to estimate the average drill wear. It is confirmed experimentally that the tool wear can be accurately estimated by the trained neural network. The accuracy of tool wear estimation using the neural network is superior to that using other regression models.     

Evolving unipolar memristor spiking neural networks

Neuromorphic computing – brain-like computing in hardware – typically requires myriad complimentary metal oxide semiconductor spiking neurons interconnected by a dense mesh of nanoscale plastic synapses. Memristors are frequently cited as strong synapse candidates due to their statefulness and potential for low-power implementations. To date, plentiful research has focused on the bipolar memristor synapse, which is capable of incremental weight alterations and can provide adaptive self-organisation under a Hebbian learning scheme. In this paper, we consider the unipolar memristor synapse – a device capable of non-Hebbian switching between only two states (conductive and resistive) through application of a suitable input voltage – and discuss its suitability for neuromorphic systems. A self-adaptive evolutionary process is used to autonomously find highly fit network configurations. Experimentation on two robotics tasks shows that unipolar memristor networks evolve task-solving controllers faster than both bipolar memristor networks and networks containing constant non-plastic connections whilst performing at least comparably.     

Task-dependent evolution of modularity in neural networks

There exist many ideas and assumptions about the development and meaning of modularity in biological and technical neural systems. We empirically study the evolution of connectionist models in the context of modular problems. For this purpose, we define quantitative measures for the degree of modularity and monitor them during evolutionary processes under different constraints. It turns out that the modularity of the problem is reflected by the architecture of adapted systems, although learning can counterbalance some imperfection of the architecture. The demand for fast learning systems increases the selective pressure towards modularity.     

On-line drilling chatter recognition and avoidance using an ART2—A neural network

In the paper, an adaptive resonance theory (ART2-A) neural network is applied to on-line recognition and avoidance of drilling chatter. It is shown that the ART2-A neural network can adaptively learn the features of the thrust force spectrum in a drilling process. As a result, drilling chatter can be automatically detected when a chatter feature starts to appear in the thrust force spectrum. Once chatter is detected, a spindle speed regulation method to suppress chatter is used. Experiments show that this new developed system can monitor and suppress drilling chatter efficiently even under varying cutting conditions.     

基于分子动力学结合神经网络的Au表面能计算方法

利用嵌入原子模型,采用分子动力学方法计算了贵金属Au低指数晶面及部分简单高指数晶面的表面能.同时,采用Levenberg-Marquardt算法,建立了Au表面能的BP神经网络模型;结合分子动力学模型的计算数据,通过大量数据的自学习训练,完成神经网络模型对Au高指数晶面表面能的预测.计算结果表明:该方法具有较高的预测精度,能正确预言低指数晶面表面能的排序,Au各晶面的表面能随其晶面与(111)密排面夹角的增大呈现先增大后减小的特点.     

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.     

Reuse of information in multi-layer sigma-pi neural networks

This new work is an extension of existing research into artificial neural networks (Neville and Stonham, Connection Sci.: J. Neural Comput. Artif. Intell. Cognitive Res., 7, pp. 29–60, 1995; Neville, Neural Net., 45, pp. 375–393, 2002b). These previous studies of the reuse of information (Neville, IEEE World Congress on Computational Intelligence, 1998b, pp. 1377–1382; Neville and Eldridge, Neural Net., pp. 375–393, 2002; Neville, IEEE World Congress on Computational Intelligence, 1998c, pp. 1095–1100; Neville, IEEE 2003 International Joint Conference on Neural Networks, 2003; Neville, IEEE IJCNN'04, 2004 International Joint Conference on Neural Networks, 2004) are associated with a methodology that prescribes the weights, as opposed to training them. In addition, they work with smaller networks. Here, this work is extended to include larger nets. This methodology is considered in the context of artificial neural networks: geometric reuse of information is described mathematically and then validated experimentally. The theory shows that the trained weights of a neural network can be used to prescribe the weights of other nets of the same architecture. Hence, the other nets have prescribed weights that enable them to map related geometric functions. This means the nets are a method of ‘reuse of information’. This work is significant in that it validates the statement that, ‘knowledge encapsulated in a trained multi-layer sigma-pi neural network (MLSNN) can be reused to prescribe the weights of other MLSNNs which perform similar tasks or functions’. The important point to note here is that the other MLSNNs weights are prescribed in order to represent related functions. This implies that the knowledge encapsulated in the initially trained MLSNN is of more use than may initially appear.     

Automatic chatter detection in grinding

Two methods for automatic chatter detection in outer diameter plunge feed grinding are proposed. The methods employ entropy and coarse-grained information rate (CIR) as indicators of chatter. Entropy is calculated from a power spectrum, while CIR is calculated directly from fluctuations of a recorded signal. The methods are verified using signals of the normal grinding force and RMS acoustic emission. The results show that entropy and CIR perform equally well as chatter indicators. Based on the normal grinding force, they detect chatter in its early stage, while only cases of strong chatter are detected based on RMS acoustic emission.     

Context-free and context-sensitive dynamics in recurrent neural networks

Continuous-valued recurrent neural networks can learn mechanisms for processing context-free languages. The dynamics of such networks is usually based on damped oscillation around fixed points in state space and requires that the dynamical components are arranged in certain ways. It is shown that qualitatively similar dynamics with similar constraints hold for aⁿbⁿcⁿ , a context-sensitive language. The additional difficulty with aⁿbⁿcⁿ , compared with the context-free language aⁿbⁿ , consists of 'counting up' and 'counting down' letters simultaneously. The network solution is to oscillate in two principal dimensions, one for counting up and one for counting down. This study focuses on the dynamics employed by the sequential cascaded network, in contrast to the simple recurrent network, and the use of backpropagation through time. Found solutions generalize well beyond training data, however, learning is not reliable. The contribution of this study lies in demonstrating how the dynamics in recurrent neural networks that process context-free languages can also be employed in processing some context-sensitive languages (traditionally thought of as requiring additional computation resources). This continuity of mechanism between language classes contributes to our understanding of neural networks in modelling language learning and processing.     

Artificial neural networks in steel-mushy aluminum pressing bonding

1　ＩＮＴＲＯＤＵＣＴＩＯＮＦｏｒｓｔｅｅｌ ａｌｕｍｉｎｕｍｂｏｎｄｉｎｇ ,ｉｆａｌｕｍｉｎｕｍｓｏｌｉｄｆｒａｃｔｉｏｎｉｓ 10 0 % ,ｉｔｉｓｓｔｅｅｌ ａｌｕｍｉｎｕｍｓｏｌｉｄｔｏｓｏｌｉｄｂｏｎｄｉｎｇ ;ｉｆａｌｕｍｉｎｕｍｓｏｌｉｄｆｒａｃｔｉｏｎｉｓ 0 ,ｉｔｉｓｓｔｅｅｌ ａ ｌｕｍｉｎｕｍｓｏｌｉｄｔｏｌｉｑｕｉｄｂｏｎｄｉｎｇ ;ｉｆａｌｕｍｉｎｕｍｓｏｌｉｄｆｒａｃｔｉｏｎｉｓｗｉｔｈｉｎ 0～ 10 0 % ,ｔｈｅｂｏｎｄｉｎｇｉｓｓｔｅｅｌ ｍｕｓｈｙａｌｕｍｉｎｕｍｂｏｎｄｉｎｇ .Ｆｏｒ…     

Fuzzy neural networks for arc welding quality control

0　ＩｎｔｒｏｄｕｃｔｉｏｎＦｕｚｚｙｌｏｇｉｃｃｏｎｔｒｏｌ(ＦＬＣ)ｉｓａｋｎｏｗｌｅｄｇｅｂａｓｅｄｃｏｎｔｒｏｌｓｔｒａｔｅｇｙｔｈａｔｈａｓｓｈｏｗｎｉｔｓｐｒｏｍｉｓｉｎｇａｐｐｌｉｃａｔｉｏｎｉｎｉｎｄｕｓｔｒｉａｌｃｏｎｔｒｏｌｅｎｇｉｎｅｅｒｉｎｇｉｎｒｅｃｅｎｔｙｅａｒｓ.Ｉｔｃａｎｂｅｕｓｅｄｗｈｅｎａｓｕｆｆｉｃｉｅｎｔｌｙａｃｃｕｒａｔｅｍｏｄｅｌｏｆｔｈｅｐｈｙｓｉｃａｌｓｙｓｔｅｍｔｏｂｅｃｏｎｔｒｏｌｌｅｄｉｓｕｎａｖａｉｌａｂｌｅｏｒｗｈｅｎａｐｒｅｃｉｓｅｍｅａｓ…     

Emergent latent symbol systems in recurrent neural networks

Fodor and Pylyshyn [(1988). Connectionism and cognitive architecture: A critical analysis. Cognition, 28(1–2), 3–71] famously argued that neural networks cannot behave systematically short of implementing a combinatorial symbol system. A recent response from Frank et al. [(2009). Connectionist semantic systematicity. Cognition, 110(3), 358–379] claimed to have trained a neural network to behave systematically without implementing a symbol system and without any in-built predisposition towards combinatorial representations. We believe systems like theirs may in fact implement a symbol system on a deeper and more interesting level: one where the symbols are latent – not visible at the level of network structure. In order to illustrate this possibility, we demonstrate our own recurrent neural network that learns to understand sentence-level language in terms of a scene. We demonstrate our model's learned understanding by testing it on novel sentences and scenes. By paring down our model into an architecturally minimal version, we demonstrate how it supports combinatorial computation over distributed representations by using the associative memory operations of Vector Symbolic Architectures. Knowledge of the model's memory scheme gives us tools to explain its errors and construct superior future models. We show how the model designs and manipulates a latent symbol system in which the combinatorial symbols are patterns of activation distributed across the layers of a neural network, instantiating a hybrid of classical symbolic and connectionist representations that combines advantages of both.