基于Sugeno模糊积分神经网络分类器融合方法在手写数字识别中的应用

神经网络是模式识别中一种常见的分类器。针对同一个分类问题,构建多个分类器并把多个分类器进行融合可以提高分类系统的分类正确率、改善系统的稳健性。首先介绍了Sugeno模糊积分及Sugeno模糊积分神经网络分类器融合方法的一般原理,而后将其应用于手写数字识别,通过实际的案例验证了该融合方法的有效性和可行性。

A Hybrid Compensation Scheme for the Input Rate-Dependent Hysteresis of the Piezoelectric Ceramic Actuators

A hybrid compensation scheme for piezoelectric ceramic actuators(PEAs)is proposed.In the hybrid compensation scheme,the input rate-dependent hysteresis characteristics of the PEAs are compensated.The feedforward controller is a novel input rate-dependent neural network hysteresis inverse model,while the feedback controller is a proportion integration differentiation(PID)controller.In the proposed inverse model,an input ratedependent auxiliary inverse operator(RAIO)and output of the hysteresis construct the expanded input space(EIS)of the inverse model which transforms the hysteresis inverse with multi-valued mapping into single-valued mapping,and the wiping-out,rate-dependent and continuous properties of the RAIO are analyzed in theories.Based on the EIS method,a hysteresis neural network inverse model,namely the dynamic back propagation neural network(DBPNN)model,is established.Moreover,a hybrid compensation scheme for the PEAs is designed to compensate for the hysteresis.Finally,the proposed method,the conventional PID controller and the hybrid controller with the modified input rate-dependent Prandtl-Ishlinskii(MRPI)model are all applied in the experimental platform.Experimental results show that the proposed method has obvious superiorities in the performance of the system.

Rolling bearing fault diagnosis based on data-level and feature-level information fusion

To address the limitation of single acceleration sensor signals in effectively reflecting the health status of rolling bearings,a rolling bearing fault diagnosis method based on the fusion of data-level and feature-level information was proposed.First,according to the impact characteristics of rolling bearing faults,correlation kurtosis rules were designed to guide the weight distribution of multi-sensor signals.These rules were then combined with a weighted fusion method to obtain high-quality data-level fusion signals.Subsequently,a feature-fusion convolutional neural network(FFCNN)that merges the one-dimensional(1D)features extracted from the fused signal with the two-dimensional(2D)features extracted from the wavelet time-frequency spectrum was designed to obtain a comprehensive representation of the health status of rolling bearings.Finally,the fused features were fed into a Softmax classifier to complete the fault diagnosis.The results show that the proposed method exhibits an average test accuracy of over 99.00%on the two rolling bearing fault datasets,outperforming other comparison methods.Thus,the method can be effectively utilized for diagnosing rolling bearing faults.

Intelligent method to develop constitutive relationship of Ti-6Al-2Zr-1Mo-1V alloy

The isothermal compression tests were carried out in the Thermecmastor-Z thermo-simulator at temperatures of 800, 850, 900, 950, 1000 and 1050 ℃ and the strain rates of 0.01, 0.1, 1 and 10 s-1. The influence of deformation temperature and strain rate on the flow stress of Ti-6Al-2Zr-IMo-IV alloy was studied. Based on the experimental data sets, the high temperature deformation behavior of Ti-6A1-2Zr-IMo-IV alloy was presented using the intelligent method of artificial neural network （ANN）. The results indicate that the predicted flow stress values by ANN model is quite consistent with the experimental results, which implies that the artificial neural network is an effective tool for studying the hot deformation behavior of the present alloy. In addition, the development of graphical user interface is implemented using Visual Basic programming language.

Decentralized Control Based on FNNSMC for Interconnected Uncertain Nonlinear Systems

A new type controller, fuzzy neural networks sliding mode controller (FNNSMC), is developed for a class of large scale systems with unknown bounds of high order interconnections and disturbances. Although sliding mode control is simple and insensitive to uncertainties and disturbances, there are two main problems in the sliding mode controller (SMC): control input chattering and the assumption of known bounds of uncertainties and disturbances. The FNNSMC, which incorporates the fuzzy neural networks (FNN) and the SMC, can eliminate the chattering by using the continuous output of the FNN to replace the discontinuous sign term in the SMC. The bounds of uncertainties and disturbances are also not required in the FNNSMC design. The simulation results show that the FNNSMC has more robustness than the SMC.

Function chain neural network prediction on heat transfer performance of oscillating heat pipe based on grey relational analysis

As for the factors affecting the heat transfer performance of complex and nonlinear oscillating heat pipe (OHP),grey relational analysis (GRA) was used to deal with the relationship between heat transfer rate of a looped copper-water OHP and charging ratio,inner diameter,inclination angel,heat input,number of turns,and the main influencing factors were defined.Then,forecasting model was obtained by using main influencing factors (such as charging ratio,interior diameter,and inclination angel) as the inputs of function chain neural network.The results show that the relative average error between the predicted and actual value is 4%,which illustrates that the function chain neural network can be applied to predict the performance of OHP accurately.

Correlation of Vapour Liquid Equilibria of Binary Mixtures Using Artificial Neural Networks

In this paper, a back propagation artificial neural network （BP-ANN） model is presented for the simultaneous estimation of vapour liquid equilibria （VLE） of four binary systems viz chlorodifluoromethan-carbondioxide, trifluoromethan-carbondioxide, carbondisulfied-trifluoromethan and carbondisulfied-chlorodifluoromethan. VLE data of the systems were taken from the literature for wide ranges of temperature （222.04-343.23K） and pressure （0.105 to 7.46MPa）. BP-ANN trained by the Levenberg-Marquardt algorithm in the MATLAB neural network toolbox was used for building and optimizing the model. It is shown that the established model could estimate the VLE with satisfactory precision and accuracy for the four systems with the root mean square error in the range of 0.054-0.119. Predictions using BP-ANN were compared with the conventional Redlich-Kwang-Soave （RKS） equation of state, suggesting that BP-ANN has better ability in estimation as compared with the RKS equation （the root mean square error in the range of 0.115-0.1546）.

Research on the Correlation Between Physical Examination Indexes and TCM Constitutions Using the RBF Neural Network

Objective To establish correlation models between various physical examination indexes and traditional Chinese medicine(TCM)constitutions,and explore their relationships based on the radial basis function(RBF)neural network.Methods The raw data of physical examination indexes and TMC constitutions of 650 subjects who underwent a physical examination were cleaned,classified and sorted,on the basis of which valid data were retrieved and categorized into a training dataset and a test dataset.Subsequently,the RBF neural network was applied to the valid samples in the training set to establish correlation models between various physical examination indexes and TCM constitutions.The accuracy and the error margin of the correlation model were then verified using the valid samples in the test set.Results Of all selected samples,the highest accuracy rates were 80% for the blood lipid index-TCM constitution model;100% for the renal function index-TCM constitution model;100% for the blood routine(male)index-TCM constitution model;88.8% for the blood routine(female)index-TCM constitution model;84.1%for the urine routine index-TCM constitution model;and 100% for the blood transfusion index-TCM constitution model.Conclusions The samples selected in this study suggested that there is a strong correlation between physical examination indexes and TCM constitutions,making it feasible to apply the established correlation models to TCM constitution identification.

Generalized Nonlinear Irreducible Auto-Correlation and Its Applications in Nonlinear Prediction Models Identification

There is still an obstacle to prevent neural network from wider and more effective applications, i.e., the lack of effective theories of models identification. Based on information theory and its generalization, this paper introduces a universal method to achieve nonlinear models identification. Two key quantities, which are called nonlinear irreducible auto-correlation (NIAC) and generalized nonlinear irreducible auto-correlation (GNIAC), are defined and discussed. NIAC and GNIAC correspond with intrinstic irreducible auto-(dependency) (IAD) and generalized irreducible auto-(dependency) (GIAD) of time series respectively. By investigating the evolving trend of NIAC and GNIAC, the optimal auto-regressive order of nonlinear auto-regressive models could be determined naturally. Subsequently, an efficient algorithm computing NIAC and GNIAC is discussed. Experiments on simulating data sets and typical nonlinear prediction models indicate remarkable correlation between optimal auto-regressive order and the highest order that NIAC-GNIAC have a remarkable non-zero value, therefore demonstrate the validity of the proposal in this paper.

Neural Network Model for Boiling Heat Transfer of R22 and Its Alternative Refrigerants inside Horizontal Smooth Tubes

Accurate prediction of refrigerant boiling heat transfer coefficients is important for the design of evaporators. The generalized correlations have different forms, and could not provide satisfactory results for R22 and its alternative refrigerants R134a, R407C and R410A. This study proposes to use artificial neural network (ANNs) as a generalized correlation model, selects the input parameters of ANNs on the basis of the dimensionless parameter groups of existing correlations, and correlates the in-tube boiling heat transfer coefficients of the above four refrigerants. The results show that the ANNs model with the input and output based on the Liu-Winterton correlation has the best result. The root-mean-square deviations in training and test are 15.5% and 20.2% respectively, and approximately 85% of the deviations are within ±20%, which is much better than that of the existing generalized correlations.

Expansion of Edlen Equation based on cascade-correlation learning architecture

Due to the limitation of Edlen Equation to compensate for air refractivity in ordinary air pressure, an experiment to study the relationship between air refractivity and temperature, along with its pressure, is designed and carried out from ordinary pressure to low pressure. The expansion of Edlen Equation is achieved by using the cascade-Correlation learning method, and a neural network architecture model. The applied accuracy of neural network is the same as that of Edlen Equation in an ordinary pressure zone.

Constitution Identification System for Traditional Chinese Medicine (TCM) Based on Correlation Between TCM Constitution and Physical Examination Indices

Objective Identification of one’s constitution based on a combination of features and back propagation neural network theory is needed in modern medicine and traditional Chinese medicine(TCM).We describe a method to identify one’s constitution based on TCM constitution classification and a physical index model.Methods We created a constitution identification system based on neural network using Visio Studio development tool.We report the initial implementation of the system,the accuracy of which was verified using actual data.Results We found a relatively strong correlation between TCM constitution and physical indicators.Conclusion Finally,our report describes a possible application of the proposed system.

An Incremental Time-delay Neural Network for Dynamical Recurrent Associative Memory

An incremental time-delay neural network based on synapse growth, which is suitable for dynamic control and learning of autonomous robots, is proposed to improve the learning and retrieving performance of dynamical recurrent associative memory architecture. The model allows steady and continuous establishment of associative memory for spatio-temporal regularities and time series in discrete sequence of inputs. The inserted hidden units can be taken as the long-term memories that expand the capacity of network and sometimes may fade away under certain condition. Preliminary experiment has shown that this incremental network may be a promising approach to endow autonomous robots with the ability of adapting to new data without destroying the learned patterns. The system also benefits from its potential chaos character for emergence.

Research on red tide occurrence forecast method based on deep learning

As a marine disaster,red tides have a serious impact on marine fisheries,ecology,economy,human production and life.Red tides have been widely concerned by researchers for a long time.However,due to its complex formation mechanism,red tide forecasting is extremely challenging.Aiming at addressing problem of red tide forecasting,this paper collects the marine monitoring data before and after the occurrence of red tide in Xiamen sea area,and analyzes the correlation between multiple environmental factors and the red tide occurrence by combining the methods of Pearson correlation coefficient,Scatter matrix,and multiple correlation coefficient.The fusion method of LSTM and CNN based on deep learning are applied to mine the temporal dependence of environmental factors and find the local features of sequence data,then predict the occurrence of red tides.In the Xiamen No.1 and Xiamen No.2 datasets,the RMSE and MAE errors of this method are reaching 0.5218 and 0.5043,respectively.The forecast probability of red tide occurrence was further determined through the collaborative comparison model.The final forecast accuracy of the two datasets is 67.58%and 63.49%,respectively.This study provides exploratory experience for the analysis and forecasting of red tides,which proves the feasibility of applying deep learning methods to red tide forecasting.

A forward-inverse dynamics modeling framework for human musculoskeletal multibody system

Multibody musculoskeletal modeling of human gait has been proved helpful in investigating the pathology of musculoskeletal disorders.However,conventional inverse dynamics methods rely on external force sensors and cannot capture the nonlinear muscle behaviors.Meanwhile,the forward dynamics approach is computationally demanding and only suited for relatively simple tasks.This study proposed an integrated simulation methodology to fulfill the requirements of estimating foot-ground reaction force,tendon elasticity,and muscle recruitment optimization.A hybrid motion capture system,which combines the marker-based infrared device and markerless tracking through deep convolutional neural networks,was developed to track lower limb movements.The foot-ground reaction forces were determined by a contact model for soft materials,and its parameters were estimated using a two-step optimization method.The muscle recruitment problem was first resolved via a static optimization algorithm,and the obtained muscle activations were used as initial values for further simulation.A torque tracking procedure was then performed by minimizing the errors of joint torques calculated by musculotendon equilibrium equations and inverse dynamics.The proposed approach was validated against the electromyography measurements of a healthy subject during gait.The simulation framework provides a robust way of predicting joint torques,musculotendon forces,and muscle activations,which can be beneficial for understanding the biomechanics of normal and pathological gait.

An experimental investigation into electromyography, constitutive relationship and morphology of crucian carp for biomechanical “digital fish”

Currently, the integrated biomechanical studies on fish locomotion come into focus, so it is urgent to provide reliable and sys- tematic experimental results, and to establish a biomechanical ＂digital fish＂ database for some typical fish species. Accord- ingly, based on the control framework of ＂Neural Control - Active Contraction of Muscle - Passive Deformation＂, the elec- tromyography （EMG） signals, the mechanical properties and the constitutive relationship of skin, muscle, and body trunk, as well as morphological parameters of crucian carp, are investigated with experiments, from which a simplified database of bio- mechanical ＂digital fish＂ is established. First, the EMG signals from three lateral superficial red muscles of crucian carp, which was evolving in the C-start movement, were acquired with a self-designing amplifier. The modes of muscle activity were also investigated. Secondly, the Young＇s modulus and the reduced relaxation function of crucian carp＇s skin and muscle were de- termined by failure tests and relaxation tests in uniaxial tensile ways, respectively. Viscoelastic models were adopted to deduce the constitutive relationship. The mechanical properties and the angular stiffness of different sites on the crucian carp＇s body trunk were obtained with dynamic bending experiments, where a self-designing dynamic bending test machine was employed. The conclusion was drawn regarding the body trunk of crucian carp under dynamic bending deformation as an approximate elastomer. According to the above experimental results, a possible benefit of body effective stiffness increasing with a little energy dissipation was discussed. Thirdly, the distribution of geometric parameters and weight parameters for a single experi- mental individual and multiple individuals of crucian carp was studied with experiments. Finally, considering all the above re- suits, generic experimental data were obtained by normalization, and a preliminary biomechanical ＂digital fish＂ database for crucian carp was established.

Stability switches in a Cohen-Grossberg neural network with multi-delays

In this paper, we consider a Cohen-Grossberg neural network with three delays. Regard- ing time delays as a parameter, we investigate the effect of time delays on its dynamics. We show that there exist stability switches for time delays under certain conditions and the conditions for the existence of periodic oscillations are given by discussing the associated characteristic equation. Numerical simulations are given to illustrate the obtained results and interesting network behaviors are observed, such as multiple stability switches of the network equilibrium and synchronous （asynchronous） oscillations.