An application of fuzzy inference for studying the dependency of roll force and roll torque on process variables in cold flat rolling

In this work, the roll force and roll torque in a cold flat rolling process are modelled using first order Takagi–Sugeno fuzzy models. The fuzzy models predict the most likely lower and upper estimates of the roll force and roll torque. Although the fuzzy models can be based on the experimental data, in the present work, the required data is generated by radial basis function neural networks. The neural networks, in turn, are trained by a finite element method-based code. It is demonstrated that the coefficients of the linear crisp function used to represent the output variables in the fuzzy inference system can be used for assessing the sensitivity of these variables with respect to the process variables. An algorithm to detect and suppress the outliers in the data is proposed. The effectiveness of the proposed algorithm is demonstrated through an example.     

Simplified theories of flat rolling-I. The calculation of roll pressure, roll force and roll torque

The classical theories of rolling continue to be widely used in practice although finite element methods can provide a more detailed analysis of the deformation during rolling. Both a homogeneous deformation solution and an inhomogeneous method for calculating the roll pressure are reformulated as a development of the classical theories of flat rolling. The new solutions do not require the gradient of the stress-strain curve d(2k)/dφ. This simplifies the computations, leads to consistent estimates of the main rolling parameters including the calculation of the roll torque and allows more general yield stress functions to be readily included in solutions for hot or cold rolling. Data for the rolling of annealed copper is used to establish the consistency and accuracy of the new solutions.     

Analyses for roll force and torque in ring rolling, with some supporting experiments

Analyses are developed for predicting roll force and torque during ring rolling between plain cylindrical rolls. A nomograph is given to allow easy application of the theory. Conventional analyses for flat rolling are not generally appropriate to ring rolling, whereas the present theory considers the particular roll geometry and rolling conditions encountered in the process.Measurements of roll force and torque, taken during ring rolling experiments, are shown to be in good agreement with the theoretical predictions. The experiments are carried out on a new experimental mill having a mandrel which retracts axially for loading and unloading the ring, and is symmetrically supported between roller bearings within a four-column frame during rolling. This arrangement is believed to offer certain advantages over most commercial machines.     

Improved method for grinding force prediction based on neural network

The work presented in this paper is an investigation of the prediction of amplitudes of the specific grinding force components. An improved method for artificial neural networks (ANNs) establishment is proposed here allowing accurate prediction of specific normal and tangential grinding forces. This method can determine the optimal set of inputs to be used for these ANN. This set of inputs is composed of significant factors and interactions among factors that could possibility offer the best learning and generalization of ANNs simultaneously. A 48-run experimental design (MED) is used in this research to train the ANNs and a total of 81 experiments are conducted to test the generalization performances of ANNs. Results have indicated that the developed ANNs show low deviations from the training data, and acceptable deviations from the testing data. In addition, the accuracies of these ANNs are found to be significantly better than those of other approaches used for modelling of the specific grinding force components. These approaches use regression models, power models, genetic algorithms or the common ANNs for which only factors of the MED are usually used in the input layer.     

Reduction of the roll force during lubricated cold rolling of aluminium strips

Lubricants, especially blended for cold rolling of aluminium, have been studied in terms of their effects on the roll-separating forces during the process. Synthetic and mineral oils, with and without additives, were investigated. Two general observations are made. The first concerns the roll force in the pass, averaged over the length of the workpiece. The second deals with the variation of the force as the process continues. In both cases, the roll-separating forces change in noticeable ways, depending on the lubricant. Tests, designed to identify some of the causes of these changes, were conducted. It was found that the effect of the lubricants on the roll-separating forces can be divided into three groups. The forces measured while using the synthetic ester based oil, with or without additives, decreased as the rolling process continued. The forces with the other base oils without additives increased slightly or remained essentially unchanged with additives. The changing lubrication conditions in the roll gap are expected to cause the variations of the roll forces.     

基于信息优化的加工过程神经网络控制

为了提高加工过程生产率和保证加工精度,以加工过程的恒切削力控制作为研究对象,将信息论原理和神经网络智能控制理论应用于加工过程控制,以信息熵作为加工过程智能控制系统的性能测度能统一各级性能指标,将神经网络作为加工过程控制输入和系统输出的信息传输通道,确定了神经网络基于信息优化的目标函数,推导出了信息优化的三层BP神经网络学习算法,提出了恒力切削过程中基于信息优化的神经网络控制系统框架.通过加工过程的仿真实例证明,与传统自适应神经网络控制方法相比,基于信息优化的神经网络控制方法收敛精确,速度快,振荡小,系统超调量小,具有较好的综合性能.研究结果为信息理论应用于加工过程控制提供了有效途径.     

A numerical method of obtaining the deformed shape of the roll in the cold rolling process

A numerical method which determines the shape of an elastic roll when the pressure distribution is known, is described. The method uses the principle of superposition, by replacing the pressure profile by equibased overlapping isosceles triangles whose individual deflection profile are known. The results show that if narrow-gauge steel strip is rolled, considerable elastic distortion of the roll results, it being quite different from the normally assumed circular arc.     

Fuzzy algorithm for calculating roll speed variation based on roll separating force in hot rolling

Thickness control of hot-rolled strips has become an important issue in recent years because of the need for improving the quality of the hot-rolled strip. For this purpose, various thickness control systems such as finishing mill set-up (FSU), automatic gauge control (AGC), and looper control system, have been developed at steel works. Although these systems have greatly improved the quality of the strip thickness, there still exists a small amount of thickness deviation. It is difficult to adequately control by applying conventional thickness control techniques since hot rolling process is a highly nonlinear system in which many process parameters are coupled. In this study, a fuzzy algorithm to calculate the roll speed variations was developed in order to improve the thickness uniformity of hot-rolled strips. Since the strip thickness is mostly affected by the magnitude of roll separating force depending on the roll speed, the strip thickness deviation between the desired and actual thicknesses can be reduced by controlling roll speed. In order to carry out this investigation, slab analysis was carried out to determine the relation between roll separating force and roll speed for various process parameters such as roll speed, reduction ratio, strip entry thickness, and front and back tensions. From the production data, the effective stress-strain rate relations of the materials used in slab analyses were acquired. Based on the analytical results, the relation between roll separating force and roll speed was approximated by a log function. A fuzzy algorithm was developed to determine variations in roll speed according to variations of roll separating force, depending on various ranges of rolling temperature, reduction ratio, front and back tensions, and strip thickness. In addition, simulations to predict roll speed variations for a small amount of thickness deviation were carried out at continuous finishing mills consisting of seven stands and the calculated roll speed variations were found to be reasonable. Thus, the developed fuzzy algorithm might be useful in reducing the thickness deviation in the actual hot rolling mills.     

A comparison of methods to estimate the roll torque in thin strip rolling

An energy balance method is used to calculate the roll torque associated with non-circular thin strip rolling and the results for the torque are compared with the moments of the forces exerted on the roll by the strip. An example is given in which the results of the two methods are quite different. This probably indicates that the axis of the roll is not at the point about which the moment is taken. The energy balance method is also used to estimate the torque associated with circular arc rolling. The results are compared with those calculated with Hill's formula in one example and, in another example, the comparison is made between the results from the energy method formulated in this paper and a commercial package which also uses both an energy method and Hill's formula.     

Analysis of the roll hunting force due to hardness in a hot rolling process

A hot rolling operation is performed to alter the thickness of a metal by passing the material through a pair of rollers, forming a gap that is somewhat narrower than the thickness of the material. Therefore, the quality of the product is a function of the pressure applied by the rollers. However, in this process, a roll hunting force occurs in which the rolling force is irregularly changed during the rotation of the rollers due to various complex mechanisms, which include roll surface hardness, difference in rotational speed between rolls, heat treatment conditions, and roll wear. In this study, roll wear tests were conducted to analyze the roll hunting force caused by variation in the hardness of the work roll. The friction coefficient of the work roll was then examined based on hardness. Then, a two-dimensional finite element model was constructed to investigate the roll hunting force as a function of the change in friction coefficient of the work roll. This finite element model was verified in relation to the theoretical rolling expression. Finite element model analysis was performed for three friction coefficients, and the effect of the roll hunting force was determined based on the reduction ratio and temperature. In addition, the wear depth of the work roll by the hardness was predicted. The influence of the abrasion of the work roll on the hunting force was analyzed.

     

基于数据场聚类的模糊神经网络在发酵过程中的应用

针对传统模糊神经网络方法中模糊规则难以提取、网络结构优化时间过长以及易于早熟的问题,提出了一种基于数据场聚类的免疫模糊神经网络方法.该方法将物理学中场的理念引入到抽象的数域空间,通过模拟对象在虚拟数据场中的相互作用实现数据对象的自组织聚类,提取模糊规则建立初始模糊神经网络模型,并运用免疫遗传算法优化构成隶属函数的网络结构.以实验室赖氨酸发酵过程关键生物量参数软测量为例,进行了仿真验证.结果表明,与常规方法相比,该方法具有较好的建模精度和实用性.     

赖氨酸发酵过程关键参数的模糊神经网络逆软测量研究

针对微生物发酵关键生物量参数(基质浓度、菌丝浓度和产物浓度)难以直接测量的问题,以赖氨酸发酵过程为研究对象,采用基于"虚拟子系统"的模糊神经网络逆系统软测量方法对关键生物量参数进行在线估计.假定在发酵过程内部存在一个以不可直接测量参数为输入,直接可测参数为输出的"虚拟子系统",并建立 "虚拟子系统" 的数学模型.再构造"虚拟子系统"的模糊神经网络逆系统,将逆系统串接在"虚拟子系统"后构成复合伪线性系统,得到动态软测量模型,实现不可直接测量参数的在线估计.实验结果表明:该方法能很好地实时估算赖氨酸发酵过程关键参数,为进行赖氨酸发酵过程补料优化控制打下良好的基础.     

An experimental determination of the roll pressure distributions in cold rolling

The paper describes the design of transducers and experiments to measure the pressure distribution in the roll gap during cold metal rolling. Results are presented for aluminium, copper and mild steel. The pressure distributions obtained experimentally conform to cold rolling theory in certain circumstances, but there are notable exceptions. The experiments indicate that there is no sticking between the specimen and rolls.     

A study on the fault diagnosis of the 3-d roll shape in cold rolling

The metal processing system usually consists of various components such as motors, work rolls, backup rolls, idle rolls, sensors, etc. Even a simple fault in a single component in the system may cause a serious damage on the final product. It is, therefore, necessary to diagnose the faults of the components to detect and prevent a system failure. Especially, the defects in a work roll are critical to the quality of strip. In this study, a new 3-D diagnosis method was developed for roll shape defects in rolling processes. The new method was induced from analyzing the rolling mechanism by using a rolling force model, a tension model, the Hitchcock’ s equation, and measurement of the strip thickness, etc. Computer simulation shows that the proposed method is very useful in the diagnosis of the 3-D roll shape.     

基于神经网络-模糊集的多元过程质量异常检测及分类

传统的多元控制图用于多元质量过程监控难以给出异常相关的进一步信息.而对于与过程异常关联的质量特性或其组合的检测及其偏移量的测定或定性,对于异常原因的快速诊断、纠正措施的及时制定意义重大,从而减少过程异常导致的不合格产品量.本文提出一种利用神经网络和模糊集技术对多元过程质量异常进行检测及分类的方法,神经网络模块利用其模式识别功能对过程偏移信号作出解释,确定引发异常的质量特性或其组合;模糊分类模块利用其模糊聚类功能对神经网络的输出信号加以分类,确定异常质量特性或其组合的偏移程度.并以ARL为评价指标,与多元T2控制图做了比较.     

基于人工神经网络预测磨合磨损的最佳分形维数

基于BP人工神经网络的L-M算法,建立了磨合磨损的分形参数预测模型。将该模型用于销一盘磨合磨损试验。对最佳分形维数进行了准确预测。该模型收敛速度快、误差小,输出结果与实验结果有极好的吻合性。