基于CFNN的电火花加工工艺效果预测模型

针时常规神经网络和模糊神经网络的不足,介绍了一种具有快速算法的补偿模糊神经网络,并根据电火花加工的工艺特点及其复杂性,建立了基于补偿模糊神经网络的电火花加工工艺效果预测模型,可实现指定加工条件下的工艺效果预测.仿真结果显示了其良好的预测精度,其性能优于常规模糊神经网络.     

自适应蚁群差分进化算法

针对传统差分进化算法在求解高维复杂问题时存在通用性差、鲁棒性低、收敛速度慢和求解精度低等问题,提出一种基于蚁群算法的自适应多模式差分变异策略.算法在每代进化中,个体根据各变异进化模式上的信息素大小,采用轮盘赌选择策略选择变异算子,并根据各变异算子对优化所做贡献的大小对信息素进行动态更新,贡献大的变异算子可以获得更多被选择的机会,使得各变异算子发挥其最大性能,从而提高算法的收敛速度和通用性.对5个高维的benchmark函数进行算法验证,实验结果表明,该算法很好的提高了差分进化算法的通用性和鲁棒性,有效地克服了收敛速度慢和早熟等问题.     

一类自适应混沌文化算法

混沌变异进化算法忽略了混沌规律性,未充分利用知识来提高算法的局部收敛能力.为此,借鉴文化算法的双层进化结构,在文化算法的进化引导函数中引入自适应混沌变异策略,提出一种自适应混沌文化算法.利用进化过程隐含知识控制变异尺度,使知识引导个体能跳出局部较优解,在保证种群多样性的同时,实现进化后期的精细搜索.仿真结果表明,该算法可以有效提高进化收敛速度,具有较好的计算稳定性.     

进化神经网络中的变异算子研究

针对进化神经网络中遗传算法收敛速度慢和容易早熟这两个难题,提出了一个启发性的变异算子.该算子采用了自适应的变异率和启发式的变异位的选择策略.在多代无进化时,通过提高变异率扩大搜索范围,同时减小变异量进行更细致的搜索.求解XOR问题的实验表明,该算法既具有很快的收敛速度又能自动维持群体的多样性.     

变异自适应混沌粒子群算法求解线性超定方程

针对粒子群算法容易陷入局部最优值和全局局部搜索平衡能力较差的问题,提出来变异自适应混沌粒子群算法.根据进化代数调节惯性权重和加速因子,新算法较好平衡了粒子群算法中的全局和局部搜索模型,利用变异因子可以使得粒子跳出局部最优值,保证种群后期仍然具有进化潜力.新算法在5个测试函数上和标准粒子群算法,自适应粒子群算法,混沌粒子群算法进行了比较,通过三种比较标准,结果说明了新算法具有较快的收敛速度,和较强的进化潜力.通过对线性超定方程组的求解,说明了新算法在数学方面具有较高的使用价值.     

基于混沌序列的自适应进化规划算法

针对进化规划在进化搜索过程中存在较多非法个体这一问题,提出一种新的自适应进化规划算法.该算法根据混沌序列的特性,将混沌序列引入到进化规划的变异算子中.理论分析和实验表明,与基于Guassian变异的进化规划相比,该算法不仅具有良好的快速收敛性和鲁棒性,而且能够有效地消除非法解,计算简单、容易实现.     

改进进化策略及其在神经网络训练中的应用

进化策略是一类策略参数自适应进化算法。文章提出了一种改进进化策略(MES),MES采用基于个体排序的随机自适应Gaussian-Cauchy混合变异策略,将Gaussian和Cauchy变异算子结合起来以达到全局探索和局部搜索之间的动态平衡。此外,MES还使用重组算子以进一步提高算法的性能。将该算法用于多层前向神经网络训练,数值仿真结果显示了该算法的有效性。     

混沌变异进化规划及在数字水印中的应用

针对现有的数字水印算法都是根据人类视觉模型理论来寻找嵌入位置的问题,本文提出一种新的变尺度进化规划算法.该算法根据混沌序列的特性,将混沌引入到进化规划的变异算子中,并利用提出的进化规划算法采寻找数字图像中水印的嵌入部位.理论分析和实验结果表明,该算法具有良好的快速收敛性和鲁棒性,并可以自适应寻找水印的最优嵌入部位,以最大限度地减少因水印的嵌入而对原始图像造成的破坏.     

基于模糊推理的进化规划用于神经网络设计

将模糊技术和进化计算技术融合,提出一种新的进化规划算法.该方法不仅能够加快普通进化规划算法的收敛性,而且能够有效地保证种群的多样性.将该方法用于神经网络的优化设计,计算机实验结果表明该方法性能稳定并减少了学习时间,较前人的研究成果更为有效.     

一类自适应混沌文化算法

混沌变异进化算法忽略了混沌规律性,未充分利用知识来提高算法的局部收敛能力.为此,借鉴文化算法的双层进化结构,在文化算法的进化引导函数中引入自适应混沌变异策略,提出一种自适应混沌文化算法.利用进化过程隐含知识控制变异尺度,使知识引导个体能跳出局部较优解,在保证种群多样性的同时,实现进化后期的精细搜索.仿真结果表明,该算法可以有效提高进化收敛速度,具有较好的计算稳定性.

     

Neural-Network-Based Numerical Control for Milling Machine

We describe a device which uses a neural network to generate part-programs for milling, drilling and similar operations on machining centres, on the basis of 2D, 2.5D or 3D geometric models of prismatic parts, without operator intervention. The neural network consists of networks for prediction of milling strategy, for prediction of surface quality and for the optimisation of technological parameters in milling. We introduce the surface complexity index (SCI) for identifying surfaces which are very difficult to machine. The SCI takes the surface roughness and machining strategy into account. Teaching and testing of the NN is described. The device, which can be retrofitted to a CNC controller, can be trained from a set of typical parts and will then generate new NC part-programs. A case study of a tool used in the automotive supplier industry shows how a milling strategy is proposed, according to set constraints.     

A Bayesian network model for surface roughness prediction in the machining process

The literature reports many scientific works on the use of artificial intelligence techniques such as neural networks or fuzzy logic to predict surface roughness. This article aims at introducing Bayesian network-based classifiers to predict surface roughness (Ra) in high-speed machining. These models are appropriate as prediction techniques because the non-linearity of the machining process demands robust and reliable algorithms to deal with all the invisible trends present when a work piece is machining. The experimental test obtained from a high-speed milling contouring process analysed the indicator of goodness using the Naïve Bayes and the Tree-Augmented Network algorithms. Up to 81.2% accuracy was achieved in the Ra classification results. Therefore, we envisage that Bayesian network-based classifiers may become a powerful and flexible tool in high-speed machining.     

Modeling and adaptive force control of milling by using artificial techniques

The contribution discusses the use of combining the methods of neural networks, fuzzy logic and PSO evolutionary strategy in modeling and adaptively controlling the process of ball-end milling. On the basis of the hybrid process modeling, off-line optimization and feed-forward neural control scheme (UNKS) the combined system for off-line optimization and adaptive adjustment of cutting parameters is built. This is an adaptive control system controlling the cutting force and maintaining constant roughness of the surface being milled by digital adaptation of cutting parameters. In this way it compensates all disturbances during the cutting process: tool wear, non-homogeneity of the workpiece material, vibrations, chatter, etc. The basic control principle is based on the control scheme (UNKS) consisting of two neural identifiers of the process dynamics and primary regulator. An overall procedure of hybrid modeling of cutting process used for creating the CNC milling simulator has been prepared. The experimental results show that not only does the milling system with the design controller have high robustness, and global stability, but also the machining efficiency of the milling system with the adaptive controller is 27% higher than for traditional CNC milling system.     

多传感器数据融合技术在刀具状态监测中的应用

提出了一种基于混合智能融合技术进行铣刀磨损量监测和预测方法。利用多传感器对切削力和振动信号进行监测,通过频率变换提取切削力特征量,采用小波包分解技术提取振动信号特征量。通过信号特征值的组合,分别探讨了几种混合智能数据融合技术——小波神经网络、遗传神经网络、遗传小波神经网络对刀具磨损量的预测效果。试验分析表明:提出的几种基于多传感器的混合智能数据融合技术均能够有效地完成刀具磨损量监测和预测,同时,对这几种数据融合技术各自的特点进行了比较分析。     

基于深度卷积神经网络的刀具寿命动态预测研究

刀具寿命预测对提高工件加工精度和生产加工效率具有重要意义.同工况下同型号刀具监测信号数据分布不一致,导致历史寿命预测模型对刀具寿命预测效果有限.鉴于此,提出一种基于深度卷积神经网络(DCNN)的刀具寿命动态预测方法.首先,利用DCNN挖掘历史刀具监测信号的退化趋势特征,构建刀具寿命预测模型,并加入注意力机制对DCNN输出进行加权,加强对刀具寿命特征的学习,提高寿命预测准确度;然后,通过基于KL散度对刀具监测信号数据分布不一致进行检测,从而在已有刀具寿命预测模型的基础上进行更新迭代;最后,利用迭代后的模型再次进行刀具寿命预测.所提出方法很好地体现了刀具实际加工过程对刀具寿命的影响,以铣削数据集为例验证了所提出方法的有效性.     

Microstructuring of non-conductive silicon carbide by electrical discharge machining

The electrical discharge machining process is an established process for machining materials regardless of their mechanical properties. Thus this process is especially attractive for materials which are hard to machine with conventional machining methods. The only requirement a material has to fulfil is having a certain electrical conductivity. Ceramic materials, (e.g. zirconia, silicon nitride or silicon carbide) exhibit excellent mechanical properties but are mostly electrically non-conductive. This can be compensated by an applied, electrically conductive assisting electrode. With this modification, the electrical discharge machining of non-conductive ceramic material is enabled. In this study the micro electrical discharge machining of non-conductive sintered silicon carbide is investigated. The drilling process shows instabilities due to the excessive generation of carbon products. A stabilisation of the process up to the maximum depth of 420 μm is realized by two approaches: adapting process parameters and adapting the tool electrode geometry. An analysis of the amount of infeed used in a milling process shows that an infeed of 15 μm has the best material removal rate to tool wear rate ratio. A maximum material removal rate of 3.58 × 10^?3 mm³/min is achieved. Detached microstructures with an aspect ratio of 30 are machined. A conducted surface analysis indicates that the present removal mechanism is thermally induced spalling. Furthermore no heat affected zone is present in the machined near-surface area.     

基于混合智能计算的铣刀状态监测

提出了一种基于混合智能融合技术进行铣刀磨损量监测和预测方法。利用多传感器对切削力和振动信号进行监测,通过频率变换提取切削力特征量,采用小波包分解技术提取振动信号特征量。通过信号特征值的组合,分别探讨了几种混合智能数据融合技术-小波神经网络,遗传神经网络,遗传小波神经网络对刀具磨损量的预测效果。实验分析表明,提出的几种基于多传感器的混合智能数据融合技术均能够有效地完成刀具磨损量监测和预测,同时对它们各自的特点进行了比较分析。     

Recurrent ANN for monitoring degraded behaviours in a range of workpiece thicknesses

This paper presents the use of artificial neural networks (ANN) to diagnose degraded behaviours in wire electrical discharge machining (WEDM). The detection in advance of the degradation of the cutting process is crucial since this can lead to the breakage of the cutting tool (the wire), reducing the process productivity and the required accuracy. Concerning this, previous investigations have identified different types of degraded behaviours in two commonly used workpiece thicknesses (50 and 100 mm). This goal was achieved by monitoring different functions of characteristic discharge variables. However, the thresholds achieved by these functions depended on the thickness of the workpiece. Consequently, the main objective of this work is to detect the degradation of the process when machining workpiece of different thicknesses using one unique empirical model. Since artificial neural network techniques are appropriate for stochastic and non-linear nature processes, its use is investigated here to cope with workpieces of different thicknesses. The results of this work show a satisfactory performance of the presented approach. The satisfactory performance is shown by two ratios: the validation ratio, which ranges between 85% and 100%, and the test ratio, which results between 75% and 100%.     

The evolutionary development of roughness prediction models

The vigorous expansion of wind energy power generation over the last decade has also entailed innovative improvements to surface roughness prediction models applied to high-torque milling operations. Artificial neural networks are the most widely used soft computing technique for the development of these prediction models. In this paper, we concentrate on the initial data transformation and its effect on the prediction of surface roughness in high-torque face milling operations. An extensive data set is generated from experiments performed under industrial conditions. The data set includes a very broad set of different parameters that influence surface roughness: cutting tool properties, machining parameters and cutting phenomena. Some of these parameters may potentially be related to the others or may only have a minor influence on the prediction model. Moreover, depending on the number of available records, the machine learning models may or may not be capable of modelling some of the underlying dependencies. Hence, the need to select an appropriate number of input signals and their matching prediction model configuration.A hybrid algorithm that combines a genetic algorithm with neural networks is proposed in this paper, in order to address the selection of relevant parameters and their appropriate transformation. The algorithm has been tested in a number of experiments performed under workshop conditions with data sets of different sizes to investigate the impact of available data on the selection of corresponding data transformation. Data set size has a direct influence on the accuracy of the prediction models for roughness modelling, but also on the use of individual parameters and transformed features. The results of the tests show significant improvements in the quality of prediction models constructed in this way. These improvements are evident when these models are compared with standard multilayer perceptrons trained with all the parameters and with data reduced through standard Principal Component Analysis practice.     

Neural networks for estimating the tool path length in concurrent engineering applications

This paper deals with the development of a neural computing system that can predict the cutting tool path length for milling an arbitrary pocket defined within the domain of a product design, in a computer numerically controlled (CNC) setting. Existing computer aided design and manufacturing systems (CAD/CAM) consume significant amounts of time in terms of data entry pertaining to the geometries and subsequent modifications to them. In the concurrent engineering environment, where even the designer needs information from the CAD/CAM systems, such time-consuming processes can be expensive. To alleviate this problem, a neural network system can be used to estimate machining time by predicting cost-dependent variables such as tool path length for the pocket milling operation. Pockets are characterized and classified into various groups. A randomized design is described so that the training samples that have been chosen represent the domain evenly. An appropriate network was built and trained with the sample pocket geometries. The analysis of the performance of the system in terms of tool path length prediction for new pocket geometries is presented.