DEEP NEURAL NETWORKS COMBINING MULTI-TASK LEARNING FOR SOLVING DELAY INTEGRO-DIFFERENTIAL EQUATIONS

Deep neural networks(DNNs)are effective in solving both forward and inverse problems for nonlinear partial differential equations(PDEs).However,conventional DNNs are not effective in handling problems such as delay differential equations(DDEs)and delay integrodifferential equations(DIDEs)with constant delays,primarily due to their low regularity at delayinduced breaking points.In this paper,a DNN method that combines multi-task learning(MTL)which is proposed to solve both the forward and inverse problems of DIDEs.The core idea of this approach is to divide the original equation into multiple tasks based on the delay,using auxiliary outputs to represent the integral terms,followed by the use of MTL to seamlessly incorporate the properties at the breaking points into the loss function.Furthermore,given the increased training dificulty associated with multiple tasks and outputs,we employ a sequential training scheme to reduce training complexity and provide reference solutions for subsequent tasks.This approach significantly enhances the approximation accuracy of solving DIDEs with DNNs,as demonstrated by comparisons with traditional DNN methods.We validate the effectiveness of this method through several numerical experiments,test various parameter sharing structures in MTL and compare the testing results of these structures.Finally,this method is implemented to solve the inverse problem of nonlinear DIDE and the results show that the unknown parameters of DIDE can be discovered with sparse or noisy data.

Application of neural networks for permanent magnet synchronous motor direct torque control

Neural networks require a lot of training to understand the model of a plant or a process. Issues such as learning speed, stability, and weight convergence remain as areas of research and comparison of many training algorithms. The application of neural networks to control interior permanent magnet synchronous motor using direct torque control （DTC） is discussed. A neural network is used to emulate the state selector of the DTC. The neural networks used are the back-propagation and radial basis function. To reduce the training patterns and increase the execution speed of the training process, the inputs of switching table are converted to digital signals, i.e., one bit represent the flux error, one bit the torque error, and three bits the region of stator flux. Computer simulations of the motor and neural-network system using the two approaches are presented and compared. Discussions about the back-propagation and radial basis function as the most promising training techniques are presented, giving its advantages and disadvantages. The system using back-propagation and radial basis function networks controller has quick parallel speed and high torque response.

Test and Experiment Study on Synchronous Cooperative Design System Over Networked Manufacturing Platform

Based on the analysis on the previous research in virtual manufacturing and virtual enterprises,this paper pro- vides a novel architecture of networked manufacturing system around the cooperative design.The key technologies for synchronous cooperative design in networked manufacturing platform,such as the cooperative mechanism,cooperative rules,control authority conveyed,cooperative efficiency,are detailed,with which a synchronous cooperative design system is developed.Due to the cooper- ative efficiency is the major bottleneck of the synchronous cooperative design over Internet,this research details the test and experi- ment to demonstrate the practicality of the system.Finally the advantages of the system are illustrated compared with current soft- ware tools.

Novel Sequential Neural Network Learning Algorithm for Function Approximation

A novel sequential neural network learning algorithm for function approximation is presented. The multi-step-ahead output predictor of the stochastic time series is introduced to the growing and pruning network for constructing network structure. And the network parameters are adjusted by the proportional differential filter （PDF） rather than EKF when the network growing criteria are not met. Experimental results show that the proposed algorithm can obtain a more compact network along with a smaller error in mean square sense than other typical sequential learning algorithms.

Noise-induced synchronous stochastic oscillations in small scale cultured heart-cell networks

This paper reports that the synchronous integer multiple oscillations of heart-cell networks or clusters are observed in the biology experiment. The behaviour of the integer multiple rhythm is a transition between super- and sub- threshold oscillations, the stochastic mechanism of the transition is identified. The similar synchronized oscillations are theoretically reproduced in the stochastic network composed of heterogeneous cells whose behaviours are chosen as excitable or oscillatory states near a Hopf bifurcation point. The parameter regions of coupling strength and noise density that the complex oscillatory rhythms can be simulated are identified. The results show that the rhythm results from a simple stochastic alternating process between super- and sub-threshold oscillations. Studies on single heart cells forming these clusters reveal excitable or oscillatory state nearby a Hopf bifurcation point underpinning the stochastic alternation. In discussion, the results are related to some abnormal heartbeat rhythms such as the sinus arrest.

QoS Adaptive Topology Configuration in Synchronous Wireless Sensor Networks

By using hyper-graph theory,this paper proposes a QoS adaptive topology configuration(QATC) algorithm to effectively control large-scale topology and achieve robust data transmitting in synchronous wireless sensor networks.Firstly,a concise hyper-graph model is abstracted to analyze the large-scale and high-connectivity network.Secondly,based on the control theory of biologic 'Cell Mergence',a novel self-adaptive topology configuration algorithm is used to build homologous perceptive data logic sub-network ...

Synchronous hybrid transport network design

Delay,as an inevitable real-world phenomenon,is usually ignored in transport network design.A model of urban hybrid transport system with stochastic delay was created on the basis of the idealized public transport system design.After formulating the total trip time cost composed of accessing time in the sub-region of the city,waiting time at the public transport station,and in-vehicle time in the public transit network,the analytical properties of the total trip time cost function were investigated.The results show that in the urban hybrid transport network design,the total trip time cost reaches its approximate minimum in a δ-neighbourhood of buffer time of 1.5 min,and that through modelling optimal delay in hybrid transport system,the maximal synchronization can be achieved and operational efficiency and passenger satisfaction can be improved.The proposed modelling and analytical investigations are attempts to contribute to more realistic modelling of future idealized public transport system that involves more practical constraints.

SEQUENTIAL DIAGNOSIS FOR A CENTRIFUGAL PUMP BASED ON FUZZY NEURAL NETWORK

A sequential diagnosis method is proposed based on a fuzzy neural network realized by ＂the partially-linearized neural network （PNN）＂, by which the fault types of rotating machinery can be precisely and effectively distinguished at an early stage on the basis of the possibilities of symptom parameters. The non-dimensional symptom parameters in time domain are defined for reflecting the features of time signals measured for the fault diagnosis of rotating machinery. The synthetic detection index is also proposed to evaluate the sensitivity of non-dimensional symptom parameters for detecting faults. The practical example of condition diagnosis for detecting and distinguishing fault states of a centrifugal pump system, such as cavitation, impeller eccentricity which often occur in a centrifugal pump system, are shown to verify the efficiency of the method proposed in this paper.

Damping properties of silicone rubber/polyacrylate sequential interpenetrating networks

Silicone rubber/polyacrylate sequential interpenetrating polymer networks(IPNs) were prepared by silicone rubber sheet dipped into the solution composed of different acrylate monomers and benzoyl peroxides(BPOs) for different time at room temperature and then acrylate polymerized at 80℃for 2 h. The molecular structure and damping properties of sequential IPNs were studied by means of FT-IR and dynamic mechanical analysis(DMA), respectively. The FT-IR spectrum shows that polyacrylate distributes unevenly along the thickness direction of IPNs, i.e. the concentration of polyacrylate decreases from the midst to the surface of the IPNs. The DMA shows that cold crystallization of silicone in the temperature range from -47℃to -30℃is reduced and loss factor of IPNs is improved after interpenetrating with polyacrylate. This suggestes that IPNs can be used as damping materials.

Analysis of Permanent Magnet-assisted Synchronous Reluctance Motor Based on Equivalent Reluctance Network Model

In this paper,the equivalent reluctance network model(ERNM)is used to calculate the magnetic circuit of a permanent magnet-assisted synchronous reluctance motor(PMASynRM)and calculate no-load air-gap magnetic field and electromagnetic torque.Iteration method is used to solve the relative permeability of iron core.A novel reluctance network model based on actual distribution of the magnetic flux inside the motor is established.The magnetomotive force(MMF)generated by armature winding affects the relative permeability of iron core,which is considered in the calculation of ERNM to improve the accuracy when the motor is under load.ERNM can be used to measure air-gap flux density,no-load back electromotive force(EMF),the average value of motor torque,the armature winding voltage under load,and power factor.The method of calculating the motor performance is proposed.The results of calculation are consistent with finite element method(FEM)and the computational complexity is much less than that of the FEM.The results of ERNM has been verified,which will provide a simple method for motor design and analysis.

A New Sequential Detection Based on Hopfield Neural Network in Frequency Selective Fading Channels

ＡＮｅｗＳｅｑｕｅｎｔｉａｌＤｅｔｅｃｔｉｏｎＢａｓｅｄｏｎＨｏｐｆｉｅｌｄＮｅｕｒａｌＮｅｔｗｏｒｋｉｎＦｒｅｑｕｅｎｃｙＳｅｌｅｃｔｉｖｅＦａｄｉｎｇＣｈａｎｎｅｌｓＷｅｎｇＪｉａｎｆｅｎｇ；ＢｉＧｕａｎｇｇｕｏ（ＳｏｕｔｈｅａｓｔＵｎｉｖｅｒｓ...

Sequential Detection Based on Hopfieid Neural Network

A sequence detector based on Hopfield Neural network(HNN) is presented, which is used to estimate the transmitted sequences from the received signals in mobile communications. In order to avoid the convergence of HNN in local minima, a decreasing step algorithm (DSA) is proposed to search the optimum sequence quickly on the basis of the traditional simulated annealing (SA) algorithm. Computer simulation results show that the new HNN detector provides almost the same performance as that of the Viterbi detector while needs less computations and memory capacity, thus it is more feasible in hardware implementation and long constraint convolutional decoding.

Synchronization investigation of the network group constituted by the nearest neighbor networks under inner and outer synchronous couplings

A new synchronization technique of inner and outer couplings is proposed in this work to investigate the synchro- nization of network group. Some Haken-Lorenz lasers with chaos behaviors are taken as the nodes to construct a few nearest neighbor complex networks and those sub-networks are also connected to form a network group. The effective node controllers are designed based on Lyapunov function and the complete synchronization among the sub-networks is realized perfectly under inner and outer couplings. The work is of potential applications in the cooperation output of lasers and the communication network.

Power-law statistics of synchronous transition in inhibitory neuronal networks

We investigate the relationship between the synchronous transition and the power law behavior in spiking networks which are composed of inhibitory neurons and balanced by dc current. In the region of the synchronous transition, the avalanche size and duration distribution obey a power law distribution. We demonstrate the robustness of the power law for event sizes at different parameters and multiple time scales. Importantly, the exponent of the event size and duration distribution can satisfy the critical scaling relation. By changing the network structure parameters in the parameter region of transition, quasicriticality is observed, that is, critical exponents depart away from the criticality while still hold approximately to a dynamical scaling relation. The results suggest that power law statistics can emerge in networks composed of inhibitory neurons when the networks are balanced by external driving signal.

On Point countable Sequential Neighborhood Networks

In this paper we prove that a space X with point countable sequential neighborhood network if and only if it is α 4 space with point countable cs network.

Artificial Monitoring of Eccentric Synchronous Reluctance Motors Using Neural Networks

This paper proposes an artificial neural network for monitoring and detecting the eccentric error of synchronous reluctance motors.Firstly,a 15 kWsynchronous reluctance motor is introduced and took as a case study to investigate the effects of eccentric rotor.Then,the equivalent magnetic circuits of the studied motor are analyzed and developed,in cases of dynamic eccentric rotor and static eccentric rotor condition,respectively.After that,the analytical equations of the studied motor are derived,in terms of its air-gap flux density,electromagnetic torque,and electromagnetic force,followed by the electromagnetic finite element analyses.Then,the modal analyses of the stator and the whole motor are performed,respectively,to explore the natural frequency and the modal shape of the motor,by which the further vibrational analysis is possible to be conducted.The vibration level of the housing is furtherly studied to investigate its relationship with the rotor eccentricity,which is validated by the prototype test.Furthermore,an artificial neural network,which has 3 layers,is proposed.By taking the air-gap flux density,the electromagnetic force,and the vibrational level as inputs,and taking the eccentric distance as output,the proposed neural network is trained till the error smaller than 5%.Therefore,this neural network is obtaining the input parameters of the tested motor,based on which it is automatically monitoring and reporting the eccentric error to the upper-level control center.

Synchronous Control of Complex Networks with Fuzzy Connections

This article is based on the T-S fuzzy control theory and investigates the synchronization control problem of complex networks with fuzzy connections. Firstly, the main stability equation of a complex network system is obtained, which can determine the stability of the synchronous manifold. Secondly, the main stable system is fuzzified, and based on fuzzy control theory, the control design of the fuzzified main stable system is carried out to obtain a coupling matrix that enables the complex network to achieve complete synchronization. The numerical analysis results indicate that the control method proposed in this paper can effectively achieve synchronization control of complex networks, while also controlling the transition time for the network to achieve synchronization.

Dynamical behaviors in discrete memristor-coupled small-world neuronal networks

The brain is a complex network system in which a large number of neurons are widely connected to each other and transmit signals to each other.The memory characteristic of memristors makes them suitable for simulating neuronal synapses with plasticity.In this paper,a memristor is used to simulate a synapse,a discrete small-world neuronal network is constructed based on Rulkov neurons and its dynamical behavior is explored.We explore the influence of system parameters on the dynamical behaviors of the discrete small-world network,and the system shows a variety of firing patterns such as spiking firing and triangular burst firing when the neuronal parameterαis changed.The results of a numerical simulation based on Matlab show that the network topology can affect the synchronous firing behavior of the neuronal network,and the higher the reconnection probability and number of the nearest neurons,the more significant the synchronization state of the neurons.In addition,by increasing the coupling strength of memristor synapses,synchronization performance is promoted.The results of this paper can boost research into complex neuronal networks coupled with memristor synapses and further promote the development of neuroscience.