In-situ neural network process controller for copper chemical mechanical polishing

Process control is one of the key methods to improve manufacturing quality. This research proposes a neural network based run-to-run process control scheme that is adaptive to the time-varying environment. Two multilayer feedforward neural networks are implemented to conduct the process control and system identification duties. The controller neural network equips the control system with more capability in handling complicated nonlinear processes. With the system information provided by this neural network, batch polishing time (T) an additional control variable, can be implemented along with the commonly used down force (p) and relative speed between the plashing pad and the plashed wafer (v). Computer simulations and experiments on copper chemical mechanical polishing processes illustrate that in drafting suppression and environmental changing adaptation that the proposed neural network based run-to-run controller (NNRTRC) performs better than the double exponentially weighted moving average (d-EWMA) approach. It is also suggested that the proposed approach can be further implemented as both an end-point detector and a pad-conditioning sensor.     

Fault detection and classification in chemical processes based on neural networks with feature extraction

Feed forward neural networks are investigated here for fault diagnosis in chemical processes, especially batch processes. The use of the neural model prediction error as the residual for fault diagnosis of sensor and component is analyzed. To reduce the training time required for the neural process model, an input feature extraction process for the neural model is implemented. An additional radial basis function neural classifier is developed to isolate faults from the residual generated, and results are presented to demonstrate the satisfactory detection and isolation of faults using this approach.     

基于建模误差PDF形状的间歇过程数据驱动模型

间歇过程的优化控制依赖于过程精确的数学模型,数据驱动的建模方法是目前间歇过程模型研究中的热点问题.突破传统数据驱动建模方法中采用均方差(mean squared error,MSE)作为准则函数的思想,提出一种新颖的间歇过程数据驱动建模方法,引入了概率密度函数(probability density function,PDF)控制的概念,构造间歇过程模型误差控制系统,将模型的可调参数作为控制系统的输入,模型误差PDF的形状作为控制系统的输出,从而把开环模型参数辨识问题转化为模型误差PDF形状的闭环控制问题.通过可调参数控制模型误差PDF的空间分布状态,不仅能够保障模型精度,还可控制模型误差的空间分布状态,从而消除模型中的有色噪声.仿真实验表明,基于模型误差PDF形状的间歇过程数据驱动模型具有较好的建模精度、鲁棒性和泛化能力,为间歇过程的数据驱动建模提供了一条新途径.     

Neural networks approach to the determination of the machining parameters

A neural networks based approach to determine the appropriate machining parameters such as speed, depth of cut and feed is proposed in this study. In this approach neural networks were used for building automatic process planning systems. Training of neural networks was performed with back propagation method by using data sets sampled in a standard handbook. These networks consist of simple processing, elements or nodes capable of processing information in response to external inputs. This approach saves computing time and storage space. In addition, it provides easy extendability as new data become available. Currently, the system provides three neural networks: for turning, for milling and for drilling operations. The performance of the trained neural network for drilling is evaluated to examine how well it predicts the machining parameters. Test results show that the neural network for the turning operation is able to predict the machining parameter values within an acceptable error rate.     

<Emphasis Type="Bold">Application of Neural Networks in Injection Moulding Process Control</Emphasis>

In order to produce precise injection moulding products, a closed-loop controller is employed instead of the open-loop control of a traditional injection moulding machine for monitoring the filling and post-filling phases of the injection processes. Since the injection moulding system has complicated and variable dynamics, the classical control theory is difficult to implement for the precise injection moulding processes. Here, two intelligent neural network control strategies are employed to adjust the injection speed of the filling phase and control the nozzle pressure of the post-filling phase. Since the neural controller has learning ability to track the variation of the injection processes, this control strategy has the advantages of adaptivity and robustness for general purpose application to an injection moulding machine. The experimental results show that this controller has good performance in the actual injection moulding processes.     

Data-driven model reference control of MIMO vertical tank systems with model-free VRFT and Q-Learning

This paper proposes a combined Virtual Reference Feedback Tuning–Q-learning model-free control approach, which tunes nonlinear static state feedback controllers to achieve output model reference tracking in an optimal control framework. The novel iterative Batch Fitted Q-learning strategy uses two neural networks to represent the value function (critic) and the controller (actor), and it is referred to as a mixed Virtual Reference Feedback Tuning–Batch Fitted Q-learning approach. Learning convergence of the Q-learning schemes generally depends, among other settings, on the efficient exploration of the state-action space. Handcrafting test signals for efficient exploration is difficult even for input-output stable unknown processes. Virtual Reference Feedback Tuning can ensure an initial stabilizing controller to be learned from few input-output data and it can be next used to collect substantially more input-state data in a controlled mode, in a constrained environment, by compensating the process dynamics. This data is used to learn significantly superior nonlinear state feedback neural networks controllers for model reference tracking, using the proposed Batch Fitted Q-learning iterative tuning strategy, motivating the original combination of the two techniques. The mixed Virtual Reference Feedback Tuning–Batch Fitted Q-learning approach is experimentally validated for water level control of a multi input-multi output nonlinear constrained coupled two-tank system. Discussions on the observed control behavior are offered.     

Practical applications of artificial neural networks in chemical process development

Artificial neural networks (ANN) have the ability to map non-linear relationships without a-priori information about process or system models. This significant feature allows the network to “learn” the behavior of a system by example when it may be difficult or impractical to complete a rigorous mathematical solution. Recently ANN technology has been leaving the academic arena and placed in user-friendly software packages. This paper will offer an introduction to artificial neural networks and present a case history of two problems in chemical process development that were approached with ANN. Both optimal PID control tuning parameters and product particle size predictions were constructed from process information using neural networks. The ANN provides a rapid solution to many applications with little physical insight into the underlying system function. The amount of data preparation and performance limitations using a neural network will be discussed. However, the properly applied ANN will generally provide insight to which variables are most influential to the model and evolve dynamically to the minimum performance surface squared error. Neural networks have been used successfully with non-linear dynamic systems and can be applied to chemical process development for system identification and multivariate optimization problems.     

Uneven batch data alignment with application to the control of batch end-product quality

Batch processes are commonly characterized by uneven trajectories due to the existence of batch-to-batch variations. The batch end-product quality is usually measured at the end of these uneven trajectories. It is necessary to align the time differences for both the measured trajectories and the batch end-product quality in order to implement statistical process monitoring and control schemes. Apart from synchronizing trajectories with variable lengths using an indicator variable or dynamic time warping, this paper proposes a novel approach to align uneven batch data by identifying short-window PCA&PLS models at first and then applying these identified models to extend shorter trajectories and predict future batch end-product quality. Furthermore, uneven batch data can also be aligned to be a specified batch length using moving window estimation. The proposed approach and its application to the control of batch end-product quality are demonstrated with a simulated example of fed-batch fermentation for penicillin production.     

A novel approach based on artificial neural network for calibration of multi-hole pressure probes

Imperfections in the manufacturing process of flow measuring probes affect their measuring behavior. Nevertheless, in order to provide the highest possible accuracy, each individual multi-hole pressure probe has to be calibrated before using them in turbomachinery. This paper presents a novel method based on artificial neural networks (ANN) to predict the flow parameters of multi-hole pressure probes. A two-stage ANN approach using multilayer perceptron (MLP) is proposed in this study. The two-stage prediction approach involves two MLP networks, which represent the calibration data and the prediction error. For a given set of inputs, outputs from both networks are combined to estimate the measured value. The calibration data of a 5-hole probe at RWTH Aachen was used to develop and validate the proposed ANN models and two-stage prediction approach. The results showed that the ANN can predict the flow parameters with high accuracy. Using the two-stage approach, the prediction accuracy was further improved compared to polynomial functions, i.e. a commonly used method in probe calibration. Furthermore, the proposed approach offers high interpolation capabilities while preventing overfitting (i.e. failure to fit new data). Unlike polynomials, it is shown that the ANN based method can provide accurate predictions at intermediate points without large oscillations.     

基于专家知识融入的模糊神经元网络结构及在镗削颤振判别中的应用

在制造系统状态监控中采用神经元网络作为模式识别器,已被证明是一种行之有效的方法。但是,由于制造系统中的加工情况复杂、包含的信息量大,所以神经元网络的学习需要大量样本才能保证它的准确性。本文利用模糊集理论,将专家知识转化为神经元网络可直接处理的模糊ｉｆ－ｔｈｅｎ规则,利用专家知识作为典型样本对模糊神经元网络进行训练,这样节省了大量获取样本的时间,同时又不降低网络的准确性。将之应用于镗削加工中颤振的判别,取得了良好的效果。     

Neural networks implementations to control real-time manufacturing systems

The main objective of advanced manufacturing control techniques is to provide efficient and accurate tools in order to control machines and manufacturing systems in real-time operations. Recent developments and implementations of expert systems and neural networks support this objective. This research explores the use of neural networks to control several manufacturing systems in real-time operations: robot manipulators, tool changes, conveyor systems and machine faults diagnosis. The main barrier to wide implementation of neural networks is the huge computation resources (times and capacities) required to train a network. This research represents the use of a multi-layer architecture of networks (input layer, several hidden layers and an output layer) to define single-valued inter-relationships between system participants and to avoid the need for long training processes. The use of neural networks to control the above-mentioned systems was evaluated from the following parameters: the architectures, network training methods, efficiencies and accuracies of networks to perform the task of control. Several conclusions related to neural network implementations to manufacturing systems were produced: (1) the multi-layer architecture fits the complexity of manufacturing systems; (2) neural networks are efficient to control real-time operations of machines; (3) machines which were controlled by neural networks performed accurate results; and (4) the use of several hidden layers can replace the need for long training processes and saves on computation resources.     

一类间歇式反应釜温度控制方法

针对反应釜内温度在反应釜反应过程控制中的重要性,运用神经网络、专家控制系统、模糊控制系统进行了系统温度控制,与反应物料无关,实现了对反应釜温度的准确控制。经过对两例反应物对象的模拟仿真,运用主体神经网络的运算,同时利用专家系统经初始状况对系统初始化,并利用模糊控制系统对系统优化给予加速。系统仿真结果表明,该系统控制范围广、控制稳定性好,是一种提高反应釜控制效果的有效方法。     

A neural network model based predictive control approach: application to a semi-batch reactor

Neural networks can be considered to be new modelling tools in process control and especially in non-linear dynamical systems cases. Their ability to approximate non-linear functions has been very often demonstrated and tested by simulation and experimental studies. In this paper, a predictive control strategy of a semi-batch reactor based on neural network models is proposed. Results of a non-linear control of the reactant temperature of a semi-batch reactor are presented. The process identification is composed of an off-line phase that consists in training the network, and of an on-line phase that corresponds to the neural model adaptation so that it fits any modification of the process dynamics. Experimental results when using this method to control a semi-batch reactor are reported and show the great potential of this strategy in controlling non-linear processes.     

Fuzzy neural network-based in-process mixed material-caused flash prediction (FNN-IPMFP) in injection molding operations

This paper describes the development of a fuzzy neural network-based in-process mixed material-caused flash prediction (FNN-IPMFP) system for injection molding processes. The goal is to employ a fuzzy neural network to predict flash in injection molding operations when using recycled mixed plastics. Major processing parameters, such as injection speed, melt temperature, and holding pressure, are varied within a small range. The vibration signal data during the mold closing and injection filling stages was collected in real-time using an accelerometer sensor. The data was analyzed with neural networks and fuzzy reasoning algorithms, in conjunction with a multiple-regression model, to obtain flash prediction threshold values under different parameter settings. The FNN-IPMFP system was shown to predict flash with 96.1% accuracy during the injection molding process.     

Artificial neural network based on genetic algorithm for emissions prediction of a SI gasoline engine

This paper proposes a hybrid learning of artificial neural network (ANN) with the nondominated sorting genetic algorithm-II (NSGAII) to improve accuracy in order to predict the exhaust emissions of a four stroke spark ignition (SI) engine. In the proposed approach, the genetic algorithm (GA) determines initial weights of local linear model tree (LOLIMOT) neural networks. A multi-objective optimization problem is determined. A sensitivity analysis is performed on NSGA-II parameters in order to provide better solutions along the optimal Pareto front. Then, a fuzzy decision maker and the technique for order preference by similarity to ideal solution (TOPSIS) are employed to select compromised solutions among the obtained Pareto solutions. The LOLIMOT-GA responses are compared with the provided by radial basis function (RBF) and multilayer perceptron (MLP) neural networks in terms of correlation coefficient R ².     

Cascade control of superheated steam temperature with neuro-PID controller

In this paper, an improved cascade control methodology for superheated processes is developed, in which the primary PID controller is implemented by neural networks trained by minimizing error entropy criterion. The entropy of the tracking error can be estimated recursively by utilizing receding horizon window technique. The measurable disturbances in superheated processes are input to the neuro-PID controller besides the sequences of tracking error in outer loop control system, hence, feedback control is combined with feedforward control in the proposed neuro-PID controller. The convergent condition of the neural networks is analyzed. The implementation procedures of the proposed cascade control approach are summarized. Compared with the neuro-PID controller using minimizing squared error criterion, the proposed neuro-PID controller using minimizing error entropy criterion may decrease fluctuations of the superheated steam temperature. A simulation example shows the advantages of the proposed method.     

Modelling Deformation of Hydroformed Circular Plates Using Neural Networks

The process of applying fluid pressure to form metal sheets into desired shapes is widely used in the industry and is known as hydroforming. Similar to most other metal forming processes, hydroforming leads to non-homogeneous plastic deformation of the workpiece. Predicting the amount of deformation caused by any sheet metal forming process leads to better products. In this paper, a model is developed to predict the amount of deformation caused by hydroforming using an artificial intelligence technique known as neural networks. The data used to design the neural network model is collected from an apparatus that was designed and built in our laboratory. The neural network model has a feedforward architecture and uses Powell’s optimisation techniques in the training process. Single- and two-hidden-layer feedforward neural network models are used to capture the nonlinear correlations between the input and output data. The neural network model was able to predict the centre deflection, the thickness variation, and the deformed shape of circular plate specimens with good accuracy. ID="A1"Correspondance and offprint requests to: Dr M. Karkoub, Mechanical and Industrial Engineering Department, College of Engineering and Petroleum, Kuwait University, PO Box 5969, Safat 13060, Kuwait     

FDM成形件精度预测模型的建立

熔融沉积成形(FDM)是快速成型(RP)最有发展前途的工艺之一,掌握提高成形件精度的控制方法是推广其应用的重要途径。在分析FDM成形件精度影响因素的基础上,提出应用误差反向传播(BP)神经网络建立预测精度模型的方法。将主要影响因素作为BP神经网络模型的输入参数,并根据最小预测误差选择输入层和中间层的维数,确定了BP模型结构。利用多组实验数据进行模型训练,建立了BP神经网络模型。模型预测与实验测量的对比结果表明,模型的预测误差在6%以内,具有很高的预测精度,可以指导实际应用。     

A nonlinear regression model-based predictive control algorithm

This paper presents a unique approach for designing a nonlinear regression model-based predictive controller (NRPC) for single-input-single-output (SISO) and multi-input-multi-output (MIMO) processes that are common in industrial applications. The innovation of this strategy is that the controller structure allows nonlinear open-loop modeling to be conducted while closed-loop control is executed every sampling instant. Consequently, the system matrix is regenerated every sampling instant using a continuous function providing a more accurate prediction of the plant. Computer simulations are carried out on nonlinear plants, demonstrating that the new approach is easily implemented and provides tight control. Also, the proposed algorithm is implemented on two real time SISO applications; a DC motor, a plastic injection molding machine and a nonlinear MIMO thermal system comprising three temperature zones to be controlled with interacting effects. The experimental closed-loop responses of the proposed algorithm were compared to a multi-model dynamic matrix controller (MPC) with improved results for various set point trajectories. Good disturbance rejection was attained, resulting in improved tracking of multi-set point profiles in comparison to multi-model MPC.