In this paper, an incremental learning model called Resource Allocating Network with Long-Term Memory (RAN-LTM) is extended such that the learning is conducted with some autonomy for the following functions: 1) data c...In this paper, an incremental learning model called Resource Allocating Network with Long-Term Memory (RAN-LTM) is extended such that the learning is conducted with some autonomy for the following functions: 1) data collection for initial learning, 2) data normalization, 3) addition of radial basis functions (RBFs), and 4) determination of RBF cen-ters and widths. The proposed learning algorithm called Autonomous Learning algorithm for Resource Allocating Network (AL-RAN) is divided into the two learning phases: initial learning phase and incremental learning phase. And the former is further divided into the autonomous data collection and the initial network learning. In the initial learning phase, training data are first collected until the class separability is converged or has a significant dif-ference between normalized and unnormalized data. Then, an initial structure of AL-RAN is autonomously determined by selecting a moderate number of RBF centers from the collected data and by defining as large RBF widths as possible within a proper range. After the initial learning, the incremental learning of AL-RAN is conducted in a sequential way whenever a new training data is given. In the experiments, we evaluate AL-RAN using five benchmark data sets. From the experimental results, we confirm that the above autonomous functions work well and the efficiency in terms of network structure and learning time is improved without sacrificing the recognition accuracy as compared with the previous version of AL-RAN.展开更多
To improve the nonlinear approximating ability of cerebellar model articulation controller(CMAC), by introducing the Gauss basis functions and the similarity measure based addressing scheme, a new kind of fuzzy CMAC...To improve the nonlinear approximating ability of cerebellar model articulation controller(CMAC), by introducing the Gauss basis functions and the similarity measure based addressing scheme, a new kind of fuzzy CMAC with Gauss basis functions(GFCMAC) was presented. Moreover, based upon the improvement of the self organizing feature map algorithm of Kohonen, the structural self organizing algorithm for GFCMAC(SOGFCMAC) was proposed. Simulation results show that adopting the Gauss basis functions and fuzzy techniques can remarkably improve the nonlinear approximating capacity of CMAC. Compared with the traditional CMAC,CMAC with general basis functions and fuzzy CMAC(FCMAC), SOGFCMAC has the obvious advantages in the aspects of the convergent speed, approximating accuracy and structural self organizing.展开更多
针对一类不确定系统的跟踪控制,设计了一种将GBF-CMAC(cerebellar model articulation controller with Gauss basis function)与滑模控制相结合的控制系统。利用符号距离和分层结构减少了神经网络所需存储器的数量,并提出了一种神经网...针对一类不确定系统的跟踪控制,设计了一种将GBF-CMAC(cerebellar model articulation controller with Gauss basis function)与滑模控制相结合的控制系统。利用符号距离和分层结构减少了神经网络所需存储器的数量,并提出了一种神经网络参数的自适应学习律。将设计的控制器用于含有不确定性和欠驱动结构的高阶柔性直线结构系统的跟踪控制,并与一般滑模控制和积分滑模控制进行了比较。实验结果表明,所设计的控制器不仅具有较好的鲁棒性,而且改善了滑模控制存在的抖振问题。同时通过调整神经网络的参数对抖振进行控制,实现了抖振和跟踪性能之间的最优选择。展开更多
针对高阶柔性直线系统的跟踪控制问题,设计了一种将改进的小脑模型关节控制器(Cerebellar Model Articulation Controller,CMAC)神经网络与传统控制器并联的控制方案。基于参数自适应算法和梯度下降法提出了一种高斯基函数CMAC神经网络...针对高阶柔性直线系统的跟踪控制问题,设计了一种将改进的小脑模型关节控制器(Cerebellar Model Articulation Controller,CMAC)神经网络与传统控制器并联的控制方案。基于参数自适应算法和梯度下降法提出了一种高斯基函数CMAC神经网络的参数自适应更新律;利用符号距离的概念减少了神经网络的输入维数;并利用自适应遗传算法对神经网络的学习率进行了优化。仿真实验表明改进的高斯基函数CMAC具有比传统CMAC更好的学习能力,控制方案实现了高阶柔性直线系统的无差跟踪控制。展开更多
In this paper, we propose a simple learning algorithm for non\|linear function approximation and system modeling using minimal radial basis function neural network with high generalization performance. A hybrid algori...In this paper, we propose a simple learning algorithm for non\|linear function approximation and system modeling using minimal radial basis function neural network with high generalization performance. A hybrid algorithm is constructed, which combines recursive n \|means clustering algorithm with a simple recursive regularized least squares algorithm (SRRLS). The n \|means clustering algorithm adjusts the centers of the network, while the SRRLS constructs a parsimonious network which makes the generalization performance of the network well. The SRRLS algorithm needs no matrix computing, so it has a lower computational cost and no ill\|conditional problem. Because of the recursive manner, this algorithm is suitable for on\|line applications. The effectiveness of this algorithm is demonstrated by two benchmark examples.展开更多
The method of determining the structures and parameters of radial basis function neural networks(RBFNNs) using improved genetic algorithms is proposed. Akaike′s information criterion (AIC) with generalization error t...The method of determining the structures and parameters of radial basis function neural networks(RBFNNs) using improved genetic algorithms is proposed. Akaike′s information criterion (AIC) with generalization error term is used as the best criterion of optimizing the structures and parameters of networks. It is shown from the simulation results that the method not only improves the approximation and generalization capability of RBFNNs ,but also obtain the optimal or suboptimal structures of networks.展开更多
Due to their inherent imperfections, it is hard to use the static neural networks for nonlinear time-varying process modeling and prediction, and the minimal resource allocation network (MRAN) is difficult to be reali...Due to their inherent imperfections, it is hard to use the static neural networks for nonlinear time-varying process modeling and prediction, and the minimal resource allocation network (MRAN) is difficult to be realized for its too many regulation parameters. A new sequential learning algorithm for radial basis function (RBF) neural networks based on local projection named Local Projection Network (LPN) is proposed in this paper. The results of validation for several benchmark problems with the new algorithm show that the presented LPN not only has the same level as M-RAN in network size and precision of the outputs, but also has fewer regulation parameters and is more predictable.展开更多
Radial Basis Function Neural Network(RBFNN)ensembles have long suffered from non-efficient training,where incorrect parameter settings can be computationally disastrous.This paper examines different evolutionary algor...Radial Basis Function Neural Network(RBFNN)ensembles have long suffered from non-efficient training,where incorrect parameter settings can be computationally disastrous.This paper examines different evolutionary algorithms for training the Symbolic Radial Basis Function Neural Network(SRBFNN)through the behavior’s integration of satisfiability programming.Inspired by evolutionary algorithms,which can iteratively find the nearoptimal solution,different Evolutionary Algorithms(EAs)were designed to optimize the producer output weight of the SRBFNN that corresponds to the embedded logic programming 2Satisfiability representation(SRBFNN-2SAT).The SRBFNN’s objective function that corresponds to Satisfiability logic programming can be minimized by different algorithms,including Genetic Algorithm(GA),Evolution Strategy Algorithm(ES),Differential Evolution Algorithm(DE),and Evolutionary Programming Algorithm(EP).Each of these methods is presented in the steps in the flowchart form which can be used for its straightforward implementation in any programming language.With the use of SRBFNN-2SAT,a training method based on these algorithms has been presented,then training has been compared among algorithms,which were applied in Microsoft Visual C++software using multiple metrics of performance,including Mean Absolute Relative Error(MARE),Root Mean Square Error(RMSE),Mean Absolute Percentage Error(MAPE),Mean Bias Error(MBE),Systematic Error(SD),Schwarz Bayesian Criterion(SBC),and Central Process Unit time(CPU time).Based on the results,the EP algorithm achieved a higher training rate and simple structure compared with the rest of the algorithms.It has been confirmed that the EP algorithm is quite effective in training and obtaining the best output weight,accompanied by the slightest iteration error,which minimizes the objective function of SRBFNN-2SAT.展开更多
Aim To detect sensor failure in control system using a single sensor signal. Methods A neural predictor was designed based on a radial basis function network(RBFN), and the neural predictor learned the sensor sig...Aim To detect sensor failure in control system using a single sensor signal. Methods A neural predictor was designed based on a radial basis function network(RBFN), and the neural predictor learned the sensor signal on line with a hybrid algorithm composed of n means clustering and Kalman filter and then gave the estimation of the sensor signal at the next step. If the difference between the estimation and the actural values of the sensor signal exceeded a threshold, the sensor could be declared to have a failure. The choice of the failure detection threshold depends on the noise variance and the possible prediction error of neural predictor. Results and Conclusion\ The computer simulation results show the proposed method can detect sensor failure correctly for a gyro in an automotive engine.展开更多
The crack fault is one of the most common faults in the rotor system,and researchers have paid close attention to its fault diagnosis.However,most studies focus on discussing the dynamic response characteristics cause...The crack fault is one of the most common faults in the rotor system,and researchers have paid close attention to its fault diagnosis.However,most studies focus on discussing the dynamic response characteristics caused by the crack rather than estimating the crack depth and position based on the obtained vibration signals.In this paper,a novel crack fault diagnosis and location method for a dual-disk hollow shaft rotor system based on the Radial basis function(RBF)network and Pattern recognition neural network(PRNN)is presented.Firstly,a rotor system model with a breathing crack suitable for a short-thick hollow shaft rotor is established based on the finite element method,where the crack's periodic opening and closing pattern and different degrees of crack depth are considered.Then,the dynamic response is obtained by the harmonic balance method.By adjusting the crack parameters,the dynamic characteristics related to the crack depth and position are analyzed through the amplitude-frequency responses and waterfall plots.The analysis results show that the first critical speed,first subcritical speed,first critical speed amplitude,and super-harmonic resonance peak at the first subcritical speed can be utilized for the crack fault diagnosis.Based on this,the RBF network and PRNN are adopted to determine the depth and approximate location of the crack respectively by taking the above dynamic characteristics as input.Test results show that the proposed method has high fault diagnosis accuracy.This research proposes a crack detection method adequate for the hollow shaft rotor system,where the crack depth and position are both unknown.展开更多
文摘In this paper, an incremental learning model called Resource Allocating Network with Long-Term Memory (RAN-LTM) is extended such that the learning is conducted with some autonomy for the following functions: 1) data collection for initial learning, 2) data normalization, 3) addition of radial basis functions (RBFs), and 4) determination of RBF cen-ters and widths. The proposed learning algorithm called Autonomous Learning algorithm for Resource Allocating Network (AL-RAN) is divided into the two learning phases: initial learning phase and incremental learning phase. And the former is further divided into the autonomous data collection and the initial network learning. In the initial learning phase, training data are first collected until the class separability is converged or has a significant dif-ference between normalized and unnormalized data. Then, an initial structure of AL-RAN is autonomously determined by selecting a moderate number of RBF centers from the collected data and by defining as large RBF widths as possible within a proper range. After the initial learning, the incremental learning of AL-RAN is conducted in a sequential way whenever a new training data is given. In the experiments, we evaluate AL-RAN using five benchmark data sets. From the experimental results, we confirm that the above autonomous functions work well and the efficiency in terms of network structure and learning time is improved without sacrificing the recognition accuracy as compared with the previous version of AL-RAN.
文摘To improve the nonlinear approximating ability of cerebellar model articulation controller(CMAC), by introducing the Gauss basis functions and the similarity measure based addressing scheme, a new kind of fuzzy CMAC with Gauss basis functions(GFCMAC) was presented. Moreover, based upon the improvement of the self organizing feature map algorithm of Kohonen, the structural self organizing algorithm for GFCMAC(SOGFCMAC) was proposed. Simulation results show that adopting the Gauss basis functions and fuzzy techniques can remarkably improve the nonlinear approximating capacity of CMAC. Compared with the traditional CMAC,CMAC with general basis functions and fuzzy CMAC(FCMAC), SOGFCMAC has the obvious advantages in the aspects of the convergent speed, approximating accuracy and structural self organizing.
文摘针对一类不确定系统的跟踪控制,设计了一种将GBF-CMAC(cerebellar model articulation controller with Gauss basis function)与滑模控制相结合的控制系统。利用符号距离和分层结构减少了神经网络所需存储器的数量,并提出了一种神经网络参数的自适应学习律。将设计的控制器用于含有不确定性和欠驱动结构的高阶柔性直线结构系统的跟踪控制,并与一般滑模控制和积分滑模控制进行了比较。实验结果表明,所设计的控制器不仅具有较好的鲁棒性,而且改善了滑模控制存在的抖振问题。同时通过调整神经网络的参数对抖振进行控制,实现了抖振和跟踪性能之间的最优选择。
文摘针对高阶柔性直线系统的跟踪控制问题,设计了一种将改进的小脑模型关节控制器(Cerebellar Model Articulation Controller,CMAC)神经网络与传统控制器并联的控制方案。基于参数自适应算法和梯度下降法提出了一种高斯基函数CMAC神经网络的参数自适应更新律;利用符号距离的概念减少了神经网络的输入维数;并利用自适应遗传算法对神经网络的学习率进行了优化。仿真实验表明改进的高斯基函数CMAC具有比传统CMAC更好的学习能力,控制方案实现了高阶柔性直线系统的无差跟踪控制。
基金This work is supported by the International Pioneering Center of TEDATianjinP.R China
文摘In this paper, we propose a simple learning algorithm for non\|linear function approximation and system modeling using minimal radial basis function neural network with high generalization performance. A hybrid algorithm is constructed, which combines recursive n \|means clustering algorithm with a simple recursive regularized least squares algorithm (SRRLS). The n \|means clustering algorithm adjusts the centers of the network, while the SRRLS constructs a parsimonious network which makes the generalization performance of the network well. The SRRLS algorithm needs no matrix computing, so it has a lower computational cost and no ill\|conditional problem. Because of the recursive manner, this algorithm is suitable for on\|line applications. The effectiveness of this algorithm is demonstrated by two benchmark examples.
文摘The method of determining the structures and parameters of radial basis function neural networks(RBFNNs) using improved genetic algorithms is proposed. Akaike′s information criterion (AIC) with generalization error term is used as the best criterion of optimizing the structures and parameters of networks. It is shown from the simulation results that the method not only improves the approximation and generalization capability of RBFNNs ,but also obtain the optimal or suboptimal structures of networks.
基金the National Natural Science Foundation of China (Grant No.50076008).
文摘Due to their inherent imperfections, it is hard to use the static neural networks for nonlinear time-varying process modeling and prediction, and the minimal resource allocation network (MRAN) is difficult to be realized for its too many regulation parameters. A new sequential learning algorithm for radial basis function (RBF) neural networks based on local projection named Local Projection Network (LPN) is proposed in this paper. The results of validation for several benchmark problems with the new algorithm show that the presented LPN not only has the same level as M-RAN in network size and precision of the outputs, but also has fewer regulation parameters and is more predictable.
基金This work is supported by Ministry of Higher Education(MOHE)through Fundamental Research Grant Scheme(FRGS)(FRGS/1/2020/STG06/UTHM/03/7).
文摘Radial Basis Function Neural Network(RBFNN)ensembles have long suffered from non-efficient training,where incorrect parameter settings can be computationally disastrous.This paper examines different evolutionary algorithms for training the Symbolic Radial Basis Function Neural Network(SRBFNN)through the behavior’s integration of satisfiability programming.Inspired by evolutionary algorithms,which can iteratively find the nearoptimal solution,different Evolutionary Algorithms(EAs)were designed to optimize the producer output weight of the SRBFNN that corresponds to the embedded logic programming 2Satisfiability representation(SRBFNN-2SAT).The SRBFNN’s objective function that corresponds to Satisfiability logic programming can be minimized by different algorithms,including Genetic Algorithm(GA),Evolution Strategy Algorithm(ES),Differential Evolution Algorithm(DE),and Evolutionary Programming Algorithm(EP).Each of these methods is presented in the steps in the flowchart form which can be used for its straightforward implementation in any programming language.With the use of SRBFNN-2SAT,a training method based on these algorithms has been presented,then training has been compared among algorithms,which were applied in Microsoft Visual C++software using multiple metrics of performance,including Mean Absolute Relative Error(MARE),Root Mean Square Error(RMSE),Mean Absolute Percentage Error(MAPE),Mean Bias Error(MBE),Systematic Error(SD),Schwarz Bayesian Criterion(SBC),and Central Process Unit time(CPU time).Based on the results,the EP algorithm achieved a higher training rate and simple structure compared with the rest of the algorithms.It has been confirmed that the EP algorithm is quite effective in training and obtaining the best output weight,accompanied by the slightest iteration error,which minimizes the objective function of SRBFNN-2SAT.
文摘Aim To detect sensor failure in control system using a single sensor signal. Methods A neural predictor was designed based on a radial basis function network(RBFN), and the neural predictor learned the sensor signal on line with a hybrid algorithm composed of n means clustering and Kalman filter and then gave the estimation of the sensor signal at the next step. If the difference between the estimation and the actural values of the sensor signal exceeded a threshold, the sensor could be declared to have a failure. The choice of the failure detection threshold depends on the noise variance and the possible prediction error of neural predictor. Results and Conclusion\ The computer simulation results show the proposed method can detect sensor failure correctly for a gyro in an automotive engine.
基金Supported by National Natural Science Foundation of China (Grant No.11972129)National Science and Technology Major Project of China (Grant No.2017-IV-0008-0045)+1 种基金Heilongjiang Provincial Natural Science Foundation (Grant No.YQ2022A008)the Fundamental Research Funds for the Central Universities。
文摘The crack fault is one of the most common faults in the rotor system,and researchers have paid close attention to its fault diagnosis.However,most studies focus on discussing the dynamic response characteristics caused by the crack rather than estimating the crack depth and position based on the obtained vibration signals.In this paper,a novel crack fault diagnosis and location method for a dual-disk hollow shaft rotor system based on the Radial basis function(RBF)network and Pattern recognition neural network(PRNN)is presented.Firstly,a rotor system model with a breathing crack suitable for a short-thick hollow shaft rotor is established based on the finite element method,where the crack's periodic opening and closing pattern and different degrees of crack depth are considered.Then,the dynamic response is obtained by the harmonic balance method.By adjusting the crack parameters,the dynamic characteristics related to the crack depth and position are analyzed through the amplitude-frequency responses and waterfall plots.The analysis results show that the first critical speed,first subcritical speed,first critical speed amplitude,and super-harmonic resonance peak at the first subcritical speed can be utilized for the crack fault diagnosis.Based on this,the RBF network and PRNN are adopted to determine the depth and approximate location of the crack respectively by taking the above dynamic characteristics as input.Test results show that the proposed method has high fault diagnosis accuracy.This research proposes a crack detection method adequate for the hollow shaft rotor system,where the crack depth and position are both unknown.