Improved identification and autotuning of PI controllers for MIMO processes by relay techniques

A technique for on-line identification and tuning is proposed to be used in the framework of a MIMO autotuning procedure. The proposed technique does not suffer from the risks of instability and the lack in performance of common tuning techniques in MIMO autotuning. Identification is accomplished through an extension of the well known ATV autotune identification method and requires only few additional tests in order to obtain some more knowledge about the process. The resulting model, which describes with good precision the process in a region of frequencies around the critical point, is then used for tuning: the integral time is found as a function of the model time constants and delay, while the gain is computed in order to give a desired value of the closed-loop resonance peak. Examples of application show that advantages over other proposed techniques can be retained for processes having different dynamic characteristics.     

PID自适应控制

本文提出将ＰＩＤ继电自整定与神经网络相结合，共同完成ＰＩＤ自适应控制。以一个两层线性网络构造ＰＩＤ控制器，将由ＰＩＤ继电自整定法获取的ＰＩＤ参数值做适当的修正后作为网络权的初值，实现对系统的在线控制。     

An expert autotuner for multiloop SISO controllers

In this paper, an on-line expert autotuner for a class of 2-input-2-output multivariable process control applications is proposed. The autotuning controller, which uses a pattern-recognition technique is designed with a view to its practical implementation in multivariable processes. The main idea of the autotuning methodology is to use the observed multiloop responses with reference to the single-loop responses such that proper detuning of the SISO controllers is achieved. Customized identification techniques in SISO and MIMO environments based on closed-loop responses are developed for this application. Simulation results for a range of 2-input-2-output multivariable processes characterized by the Relative Gain (RG) and the relative dynamics are used to evaluate the performance of the autotuning controller under different conditions. The time response of the autotuning controller is compared to that of Biggest Log Modulus Tuning (BLT) method with a few distillation column models proposed in the literature.     

A comparative study of recent relay autotuning methods for multivariable systems

A typical procedure for designing multivariable controllers is the following: build a model for the multivariable process, choose the control structure, calculate the control parameters, test the controller (possibly with simulation) and then retune controller parameters as necessary. This procedure is complex and time consuming even for scalar control loops. For multivariable controllers, the procedure is even more daunting. Automation of the design method is and has been a concern of many researchers. There has been a large number of papers on relay autotuning of control systems. The choice of relay feedback to solve the design problem is justified by the possible integration of system identification and control into the same design strategy, giving birth to relay autotuning. In this paper, nine different relay autotuning methods for multivariable systems are compared. Most of these methods have common basics but they may differ in the tuning procedure, convergence, identification method, control structure and performance achievement. The paper summarizes these methods and investigates the advantages and drawback of each algorithm.     

Shape factor of relay response curves and its use in autotuning

Identification of process parameters using single relay feedback test is mostly used in practice. Limit cycle data along with shape factor of the response curves are important to identify the correct model structure and corresponding model parameters. Time domain analytical expressions are helpful in deriving conditions to estimate process parameters accurately. Second order plus dead time (SOPDT) processes are utilized to represent all different shapes with minimal number of model parameters. Provided with analytical expressions for relay feedback responses, identification algorithms are formulated and three different model structures are categorized. The autotuning procedure consists of the following steps. First, a relay feedback test is conducted and relay response is recorded for analysis. If the response is not symmetric we do a biased relay test to restore the symmetry, otherwise, we proceed for system identification. After finding out model structures and parameters suitable tuning rules are suggested for different ranges of dead time to time constant ratio (D/τ) and damping coefficient (ξ) values. Closed-loop performances of the identified systems with and without measurement noise are found to be satisfactory.     

Autotuning process controller with improved load disturbance rejection

This paper presents an autotuning process controller aimed at providing efficient rejection of load disturbances in a class of situations that are quite typical in process control, and not easy to treat with most standard autotuning controllers, especially when not only the duration of the load disturbance response, but also the peak deviation of the controlled variable is an issue. The regulator structure is not fixed a priori; this is a peculiarity with respect to the main research stream on autotuning regulators, that refers essentially to fixed-structure (PID) regulators. Both simulation and laboratory examples are reported, to show the advantages of the proposed autotuning controller.     

基于改进全卷积神经网络的医疗数据表面重建算法

为了实现对医疗数据的快速检测和分类识别,需要对医疗数据进行表面重建设计,首先,提出一种基于改进全卷积神经网络的医疗数据表面重建算法.采用无线射频识别技术进行医疗数据的大数据采样,对RFID采集的医疗数据进行信息融合处理,采用多元回归分析方法提取医疗数据的相关性统计特征量,然后,针对医疗数据中的冗余特征采用匹配滤波检测器进行冗余滤波处理,对提纯后的医疗数据采用相空间重构技术实现医疗数据重构,最后,对重构数据采用改进全卷积神经网络分类器进行分类识别,实现医疗数据的表面重建和自动识别.仿真结果表明,所提方法的医疗数据冗余特征处理效果较好,数据分类精度可高达90%以上,且医疗数据重建误差小,耗时少.     

A neural network PI controller tuner

The development of a neural network system for tuning proportional and integral (PI) feedback controllers is presented. The tuning process includes an initial gain setting procedure and a fine tuning procedure. The initial gain settings are obtained by using the standard Ziegler-Nichols tuning rules based on the openloop step response of the process. The fine tuning procedure is performed iteratively by using a neural network. The neural network suggests adjustments to the proportional gain and integrator time based on the closed-loop controlled system response. Four parameters are defined to describe the response characteristics. They are the normalised peak rise time, normalised overshoot, normalised peak to peak height, and normalised final error. These four parameters are used as inputs to the neural network. The tuning knowledge of the neural network is extracted from the tuning of a representative process. Finally, examples covering a wide range of process dynamics are tested to demonstrate the excellent performance of the tuner.     

基于特征融合的恶意加密流量识别

随着加密技术的全面应用, 越来越多的恶意软件同样采用加密的方式隐藏自身的网络活动, 导致基于规则和特征的传统方法无法满足准确性和普适性的要求. 针对上述问题, 提出一种层次特征融合和注意力的恶意加密流量识别方法. 算法具备层次结构, 依次提取数据包的特征和会话流的特征, 前一阶段设计全局混合池化方法进行特征融合; 后一阶段使用注意力机制提高BiLSTM网络分析序列关系的能力. 最终, 实验采用CIC-AndMal 2017数据集进行验证, 结果表明: 模型设计合理, 相比TextCNN模型和HST-MHSA模型, 漏报率分别降低5.8%和2.6%, 加权F1值分别提高4.7%和3.5%, 在恶意加密流量识别和分类方面体现良好的优化效果.     

Neural network controller using autotuning method for nonlinearfunctions

An autotuning method for the optimum sigmoid function of neural networks is proposed. It is based on the steepest descent method. Simulated results using a learning-type direct controller confirm both the practicality and the characteristics of the autotuning method.     

A linear model based on Kalman filter for improving neural network classification performance

Neural network has been applied in several classification problems such as in medical diagnosis, handwriting recognition, and product inspection, with a good classification performance. The performance of a neural network is characterized by the neural network's structure, transfer function, and learning algorithm. However, a neural network classifier tends to be weak if it uses an inappropriate structure. The neural network's structure depends on the complexity of the relationship between the input and the output. There are no exact rules that can be used to determine the neural network's structure. Therefore, studies in improving neural network classification performance without changing the neural network's structure is a challenging issue. This paper proposes a method to improve neural network classification performance by constructing a linear model based on the Kalman filter as a post processing. The linear model transforms the predicted output of the neural network to a value close to the desired output by using the linear combination of the object features and the predicted output. This simple transformation will reduce the error of neural network and improve classification performance. The Kalman filter iteration is used to estimate the parameters of the linear model. Five datasets from various domains with various characteristics, such as attribute types, the number of attributes, the number of samples, and the number of classes, were used for empirical validation. The validation results show that the linear model based on the Kalman filter can improve the performance of the original neural network.     

Identification Recurrent Type 2 Fuzzy Wavelet Neural Network and L2‐Gain Adaptive Variable Sliding Mode Robust Control of Electro‐Hydraulic Servo System (EHSS)

An electro‐hydraulic servo system (EHSS) is a kind of system with the characteristics of time‐variant, serious nonlinearity, parameter and structural uncertainty, and uncertain load disturbance in most cases. These characteristics make it very difficult to realize highly accurate control by conventional methods. In order to solve the above problems, this paper introduces a recurrent type 2 fuzzy wavelet neural network to approximate the unknown nonlinear functions of the dynamic systems through tuning by the desired adaptive law. Based on the identification by recurrent type 2 fuzzy wavelet neural network, a L2 gain design method, combining gain adaptive variable sliding mode control with H infinity control, is proposed for load disturbance, thereby accommodating uncertainties that are the main factors affecting system stability and accuracy in EHSS. In this algorithm, a recurrent type 2 fuzzy wavelet neural network is employed to evaluate the unknown dynamic characteristics of the system and gain adaptive variable sliding mode control to compensate for evaluating errors, and H infinity control to suppress the effect on system by load disturbance. The experiment results show that the proposed system L2 gain design method can make the system exhibit strong robustness to parameter variation and load disturbance.     

Genetic Tuning of PID Controllers Using a Neural Network Model: A Seesaw Example

When genetic algorithms (GAs) are applied for PID parameter tuning, since the PID parameters are adjusted almost randomly, it is possible that the plant will be damaged due to abrupt changes in PID parameters. To solve this problem, a neural network will be used to model the plant and the genetic tuning procedure will be performed on the neural network instead of the plant. After determining the PID parameters in this off-line manner, these gains are then applied to the plant for on-line control. Moreover, considering that the neural network model may not be accurate enough, a method is also proposed for on-line fine-tuning of PID parameters. To show the validity of the proposed method, a seesaw system that has one input and two outputs will be used for experimental evaluation     

Model predictive control using an extended ARMarkov model

The original ARMarkov identification method explicitly determines the first μ Markov parameters from plant input–output data and approximates the slower dynamics of the process by an ARX model structure. In this paper, the method is extended to include a disturbance model and an ARIMAX structure is used to approximate the slower dynamics. This extended ARMarkov model is then used to formulate a predictive controller. As the number of Markov parameters in the model varies from one to P (prediction horizon)+1, the controller changes from generalized predictive control (GPC) to dynamic matrix control (DMC). The advantages of the proposed ARM-MPC are the consistency of the Markov parameters estimated by the ARMarkov method, independent tuning of the controller for servo and regulatory responses and the ability to combine the characteristics of GPC and DMC. The theoretical results are illustrated through simulation examples.     

基于多级小波神经网络的毒品爆炸物能量 色散识别①

利用小波神经网络自适应学习分类的优点,提出将多个小波神经网络并联使用,改进小波网络结构,在每个小波特征空间中确定小波神经元个数和初始化合适的小波基,用多级小波神经网络对毒品爆炸物的X光能量色谱的进行了识别分类。实验表明,用多级小波神经网络可以实现对不同种类毒品爆炸物的识别和鉴定,为X光能量色散技术用于毒品爆炸物的检测和识别提供了一种有效的方法。     

基于降噪自编码神经网络的事件相关电位脑电信号分析方法

提出一种基于降噪自编码神经网络事件相关电位分析方法,首先建立3层神经网络结构,利用降噪自编码对神经网络进行初始化,实现了降噪自编码深度学习模型的无监督学习.从无标签数据中自动学习数据特征,通过优化模型训练得到的权值作为神经网络初始化参数.其次,经过有标签的样本进行网络参数的微调即可完成对神经网络的训练,该方法有效解决了神经网络训练中因随机选择初始化参数,而导致网络易陷入局部极小的缺陷.最后,利用上述神经网络对第3届脑机接口竞赛数据集Data set Ⅱ(事件相关电位脑电信号)进行分类分析.实验结果表明:利用降噪自编码迭代2500次训练神经网络模型,在受试者A和受试者B样本数据叠加5次、10次、15次3种情况下获得的分类准确率分别为73.4%, 87.4%和97.2%.该最高准确率优于其他分类方法,比竞赛第1名联合支持向量机(SVM)分类器(ESVM)提高了0.7%,为事件相关电位脑电信号提供了一种深度学习分析方法.     

基于Matlab的自组织神经网络在油气层识别中的应用研究

介绍了自组织竞争网络和自组织影射网络的原理,对自组织竞争网络和自组织影射网络的优缺点进行了比较。采用大庆的油气层数据建立网络模型,对网络结构的参数进行了优化并对输入样本进行了聚类分析。数据分析表明自组织竞争网络和自组织影射网络都有较好的聚类结果,自组织竞争网络较自组织影射网络方法识别出的结果更客观可靠,是油气层识别的一种有效方法。     

基于双池化特征加权结构CNN的图像分类

传统的池化方式会造成特征信息丢失,导致卷积神经网络中提取的特征信息不足。为了提高卷积神经网络在图像分类过程中的准确率,优化其学习性能,本文在传统池化方式的基础上提出一种双池化特征加权结构的池化算法,利用最大池化和平均池化2种方式保留更多的有价值的特征信息,并通过遗传算法对模型进行优化。通过训练不同池化方式的卷积神经网络,研究卷积神经网络在不同数据集上的分类准确率和收敛速度。实验在遥感图像数据集NWPU-RESISC45和彩色图像数据集Cifar-10上对采用几种池化方式的卷积神经网络分类结果进行对比验证,结果分析表明：双池化特征加权结构使得卷积神经网络的分类准确率有很大程度的提高,同时模型的收敛速度得到进一步提高。     

Adaptive Neuro-fuzzy Control System by RBF and GRNN Neural Networks

Recently, adaptive control systems utilizing artificial intelligent techniques are being actively investigated in many applications. Neural networks with their powerful learning capability are being sought as the basis for many adaptive control systems where on-line adaptation can be implemented. Fuzzy logic, on the other hand, has proved to be rather popular in many control system applications due to providing a rule-base like structure. In this paper, an adaptive neuro-fuzzy control system is proposed in which the Radial Basis Function neural network (RBF) is implemented as a neuro-fuzzy controller (NFC) and the General Regression neural network (GRNN) as a predictor. The adaptation of the system involves the following three procedures: (1) tuning of the control actions or rules, (2) trimming of the control actions, and (3) adjustment of the controller output gain. The tuning method is a non-gradient descent method based on the predicted system response which is able to self-organize the control actions from the initial stage. The trimming scheme can help to reduce the aggressiveness of the particular control rules such that the response is stabilized to the set-points more effectively, while the controller gain adjustment scheme can be applied in the cases where the appropriate controller output gain is difficult to determine heuristically. To show the effectiveness of this methodology, its performance is compared with the well known Generalized Predictive Control (GPC) technique which is a combination of both adaptive and predictive control schemes. Comparisons are made with respect to the transient response, disturbance rejection and changes in plant dynamics. The proposed control system is also applied in controlling a single link manipulator. The results show that it exhibits robustness and good adaptation capability which can be practically implemented.     

Automated nonlinear system modelling with multiple neural networks

This article discusses the identification of nonlinear dynamic systems using multi-layer perceptrons (MLPs). It focuses on both structure uncertainty and parameter uncertainty, which have been widely explored in the literature of nonlinear system identification. The main contribution is that an integrated analytic framework is proposed for automated neural network structure selection, parameter identification and hysteresis network switching with guaranteed neural identification performance. First, an automated network structure selection procedure is proposed within a fixed time interval for a given network construction criterion. Then, the network parameter updating algorithm is proposed with guaranteed bounded identification error. To cope with structure uncertainty, a hysteresis strategy is proposed to enable neural identifier switching with guaranteed network performance along the switching process. Both theoretic analysis and a simulation example show the efficacy of the proposed method.