Robustness analysis of radial base function and multi-layered feed-forward neural network models

In this paper, two popular types of neural network models (radial base function (RBF) and multi-layered feed-forward (MLF) networks) trained by the generalized delta rule, are tested on their robustness to random errors in input space. A method is proposed to estimate the sensitivity of network outputs to the amplitude of random errors in the input space, sampled from known normal distributions. An additional parameter can be extracted to give a general indication about the bias on the network predictions. The modelling performances of MLF and RBF neural networks have been tested on a variety of simulated function approximation problems. Since the results of the proposed validation method strongly depend on the configuration of the networks and the data used, little can be said about robustness as an intrinsic quality of the neural network model. However, given a data set where ‘pure’ errors from input and output space are specified, the method can be applied to select a neural network model which optimally approximates the nonlinear relations between objects in input and output space. The proposed method has been applied to a nonlinear modelling problem from industrial chemical practice. Since MLF and RBF networks are based on different concepts from biological neural processes, a brief theoretical introduction is given.     

Short-term hydrothermal optimisation with congestion and quality of service constraints

In short-term hydrothermal coordination (STHC), the transmission network is typically modelled through a DC power flow. However, this modelling can lead to inoperable solutions when verifying with AC power flow. A methodology that includes an AC power flow model to overcome the problem applied to STHC is presented. The approach takes into account issues such as congestion management and control of quality of service, which are often present in large and weakly meshed networks - the typical pattern of Latin American electrical power systems. Generalised Benders- decomposition together with more traditional and well-known optimisation techniques, is used for this problem. The master problem stage defines the generation levels by considering the inter-temporal constraints, whereas the sub-problem stage determines both the active and reactive economical dispatches for each step of the load curve. It meets the electrical constraints (nodal balance, transmission limits and voltage levels) through a modified AC optimal power flow. The methodology was proved over a nine-busbar hydrothermal system and the solution found was validated with a quasi-exhaustive enumeration procedure to prove the optimality of the solution. Also proved over large system was the feasibility to realistic systems.     

基于递归神经网络的传感器非线性动态建模

根据动态校准实验结果建立传感器的动态数学模型,以研究传感器的动态性能,是动态测试的一个重要内容。讨论了递归神经网络模型在传感器动态建模中的应用,给出了递归神经网络模型的结构及相应的训练算法。由于其反馈特征,使得递归神经网络模型能获取系统的动态响应。该方法特别适用于传感器非线性动态建模,而且避免了传感器模型阶次的选择的困难。试验结果表明,应用递归神经网络对传感器进行动态建模是一种行之有效的方法。     

Multi‐bit Boolean model for chemotactic drift of Escherichia coli

Dynamic biological systems can be modelled to an equivalent modular structure using Boolean networks (BNs) due to their simple construction and relative ease of integration. The chemotaxis network of the bacterium Escherichia coli (E. coli ) is one of the most investigated biological systems. In this study, the authors developed a multi‐bit Boolean approach to model the drifting behaviour of the E. coli chemotaxis system. Their approach, which is slightly different than the conventional BNs, is designed to provide finer resolution to mimic high‐level functional behaviour. Using this approach, they simulated the transient and steady‐state responses of the chemoreceptor sensory module. Furthermore, they estimated the drift velocity under conditions of the exponential nutrient gradient. Their predictions on chemotactic drifting are in good agreement with the experimental measurements under similar input conditions. Taken together, by simulating chemotactic drifting, they propose that multi‐bit Boolean methodology can be used for modelling complex biological networks. Application of the method towards designing bio‐inspired systems such as nano‐bots is discussed.Inspec keywords: cell motility, microorganisms, Boolean functionsOther keywords: multibit Boolean approach, conventional BNs, high‐level functional behaviour, steady‐state responses, chemoreceptor sensory module, drift velocity, chemotactic drifting, multibit Boolean methodology, complex biological networks, bio‐inspired systems, multibit Boolean model, chemotactic drift, dynamic biological systems, equivalent modular structure, Boolean networks, simple construction, chemotaxis network, bacterium Escherichia coli, biological systems     

Implicit constitutive modelling for viscoplasticity using neural networks

Up to now, a number of models have been proposed and discussed to describe a wide range of inelastic behaviours of materials. The fatal problem of using such models is however the existence of model errors, and the problem remains inevitably as far as a material model is written explicitly. In this paper, the authors define the implicit constitutive model and propose an implicit viscoplastic constitutive model using neural networks. In their modelling, inelastic material behaviours are generalized in a state-space representation and the state-space form is constructed by a neural network using input–output data sets. A technique to extract the input–output data from experimental data is also described. The proposed model was first generated from pseudo-experimental data created by one of the widely used constitutive models and was found to replace the model well. Then, having been tested with the actual experimental data, the proposed model resulted in a negligible amount of model errors indicating its superiority to all the existing explicit models in accuracy. © 1998 John Wiley & Sons, Ltd.     

Effective sampling trajectory optimisation for sensitivity analysis of biological systems

Sensitivity analysis has been widely applied to study the biological systems, including metabolic networks, signalling pathways, and genetic circuits. The Morris method is a kind of screening sensitivity analysis approach, which can fast identify a few key factors from numerous biological parameters and inputs. The parameter or input space is randomly sampled to produce a very limited number of trajectories for the calculation of elementary effects. It is clear that the sampled trajectories are not enough to cover the whole uncertain space, which eventually causes unstable sensitivity measures. This paper presents a novel trajectory optimisation algorithm for the Morris‐based sensitivity calculation to ensure a good scan throughout the whole uncertain space. The paper demonstrates that this presented method gets more consistent sensitivity results through a benchmark example. The application to a previously published ordinary differential equation model of a cellular signalling network is presented. In detail, the parameter sensitivity analysis verifies the good agreement with data of the literatures.Inspec keywords: genetics, differential equations, sensitivity analysis, biology, sampling methods, optimisationOther keywords: biological systems, metabolic networks, genetic circuits, Morris‐based sensitivity calculation, ordinary differential equation, sampling trajectory optimisation, sensitivity analysis, parameter sensitivity analysis, cellular signalling network     

On-line identification of control chart patterns using self-organizing approaches

Effective identification of unnatural control chart patterns (CCPs) is an important issue in statistical process control, as unnatural CCPs can be associated with specific assignable causes adversely affecting the process. The intention of this paper is to develop an automatic CCP identification system using self-organizing approaches—neural network and decision tree (DT) learning. Recently, back-propagation networks (BPNs) have been widely used in the research field of CCP identification. However, one of the major limitations of conventional BPN is in dealing with dynamic patterns that vary over time, such as CCPs. This limitation is one of the major reasons for the false classification problem commonly encountered in the BPN-based CCP identification schemes in the literature. A time-lagging input algorithm is proposed in this research to enhance the performances of the BPN-based CCP identifiers. Additionally, DT learning is employed as a novel approach to the CCP identification problem. The simulation experiments demonstrate that both the BPN-based system with time-lagging input and the DT-based system perform better than the conventional BPN-based system in terms of identification accuracy and speed. The proposed time-lagging input algorithm can greatly improve the identification speed and stability of the BPN-based CCP identifier. Besides, the empirical comparison indicates that the DT-based system outperforms the BPN-based system with respect to classification capability in an on-line CCP identification scheme. Moreover, the learning time of the DT-based system is much shorter than that of the BPN-based system.     

Structural and practical identifiability analysis of S‐system

In the field of systems biology, biological reaction networks are usually modelled by ordinary differential equations. A sub‐class, the S‐systems representation, is a widely used form of modelling. Existing S‐systems identification techniques assume that the system itself is always structurally identifiable. However, due to practical limitations, biological reaction networks are often only partially measured. In addition, the captured data only covers a limited trajectory, therefore data can only be considered as a local snapshot of the system responses with respect to the complete set of state trajectories over the entire state space. Hence the estimated model can only reflect partial system dynamics and may not be unique. To improve the identification quality, the structural and practical identifiablility of S‐system are studied. The S‐system is shown to be identifiable under a set of assumptions. Then, an application on yeast fermentation pathway was conducted. Two case studies were chosen; where the first case is based on a larger state trajectories and the second case is based on a smaller one. By expanding the dataset which span a relatively larger state space, the uncertainty of the estimated system can be reduced. The results indicated that initial concentration is related to the practical identifiablity.Inspec keywords: biochemistry, differential equations, microorganisms, cellular biophysics, fermentationOther keywords: structural identifiability analysis, practical identifiability analysis, S‐system, system biology, biological reaction networks, ordinary differential equations, local snapshot, state trajectories, estimated model, partial system dynamics, identification quality, yeast fermentation pathway, relatively larger state space     

Quantized hopfield networks for reliability optimization

The use of neural networks in the reliability optimization field is rare. This paper presents an application of a recent kind of neural networks in a reliability optimization problem for a series system with multiple-choice constraints incorporated at each subsystem, to maximize the system reliability subject to the system budget. The problem is formulated as a nonlinear binary integer programming problem and characterized as an NP-hard problem. Our design of neural network to solve efficiently this problem is based on a quantized Hopfield network. This network allows us to obtain optimal design solutions very frequently and much more quickly than others Hopfield networks.     

Optimal control policy for probabilistic Boolean networks with hard constraints

It is well known that the control/intervention of some genes in a genetic regulatory network is useful for avoiding undesirable states associated with some diseases like cancer. For this purpose, both optimal finitehorizon control and infinite-horizon control policies have been proposed. Boolean networks (BNs) and its extension probabilistic Boolean networks (PBNs) as useful and effective tools for modelling gene regulatory systems have received much attention in the biophysics community. The control problem for these models has been studied widely. The optimal control problem in a PBN can be formulated as a probabilistic dynamic programming problem. In the previous studies, the optimal control problems did not take into account the hard constraints, i.e. to include an upper bound for the number of controls that can be applied to the captured PBN. This is important as more treatments may bring more side effects and the patients may not bear too many treatments. A formulation for the optimal finite-horizon control problem with hard constraints introduced by the authors. This model is state independent and the objective function is only dependent on the distance between the desirable states and the terminal states. An approximation method is also given to reduce the computational cost in solving the problem. Experimental results are given to demonstrate the efficiency of our proposed formulations and methods.     

Neural networks and modelling in vacuum science

The paper is an attempt to describe how neural networks may be used as an approximation-modelling tool. A brief survey of the evolution of the approximation theory and neural networks is presented. Practical applications are based on modelling of vacuum science problems, especially the modelling of a cold cathode pressure gauge. The problem of approximation of wide range functions, that are one of the characteristics of vacuum science problems, is introduced. Parameters such as pressure or cathode current span over several decades and neural networks are not suitable for any approximation of such functions; therefore, two strategies need to be introduced, and these are described. The approximation made by the neural network is obtained by the training process. The models obtained by several independent repetitions of training processes performed on the same training set lead to slightly different results. Therefore the definition of training stability is introduced and described. Finally, some practical hints regarding the neural network synthesis (design) are given.     

Distributed supply chain network modelling and simulation: integration of agent-based distributed simulation and improved SCOR model

Agent-based distributed simulation is an efficient methodology for modelling and analysing such complex adaptive systems as dynamic supply chain networks. However, it lacks an acceptable generic standard. Supply chain operations reference (SCOR) model is a cross-functional framework widely accepted as an industry standard. It provides the standard processes, performance metrics, best practices and associated software functionalities for modelling, evaluating and improving supply chain networks. However, it is a static tool. Integration of agent-based distributed simulation and SCOR model can exploit their advantages to form a generic methodology for modelling and simulation of a wide range of supply chain networks. Therefore, this paper proposes a methodology for distributed supply chain network modelling and simulation by means of integration of agent-based distributed simulation and an improved SCOR model. The methodology contains two components: a hierarchical framework for modelling supply chain network based on the improved SCOR model and agent building blocks integrating the standard processes from the SCOR model. The hierarchical framework provides an approach for structure modelling in any level with different granularities based on the improved SCOR model, and allows rapidly mapping a supply chain network into the structure model of a multi-agent system; while agent building blocks are quite useful and convenient to fill the structure model to fulfil its function modelling. With the approach of structure modelling and function filling, not only can the process of agent-based supply chain network modelling be accelerated, but also the built models can be reused and expanded. Because the hierarchical framework is based on the conceptual framework of SCOR model and agent building blocks integrate the standard processes from SCOR model, the proposed methodology is more generic. In addition, the issues of sub-model synchronisation and data distribution management in the agent-based distributed simulation implementation are taken into consideration and the corresponding solutions for these issues are proposed. Finally, an example of a supply chain network is modelled and implemented to illustrate the proposed methodology and related solutions.     

Modeling the thermal decomposition of friction composite systems based on yarn reinforced polymer matrices using artificial neural networks

The presented work deals with the application of artificial neural networks in the modelling of the thermal decomposition process of friction composite systems based on polymer matrices reinforced by yarns. The thermal decomposition of the automotive clutch friction composite system consisting of a polymer blend reinforced by yarns from organic, inorganic and metallic fibres impregnated with resin, as well as its individual components, was monitored by a method of non‐isothermal thermogravimetry over a wide temperature range. A supervised feed‐forward back‐propagation multi‐layer artificial neural network model, with temperature as the only input parameter, has been developed to predict the thermogravimetric curves of weight loss and time derivative of weight loss of studied friction composite system and its individual components acquired at a fixed constant heating rate under a pure dry nitrogen atmosphere at a constant flow rate. It has been proven that an optimized model with a 1‐25‐6 architecture of an artificial neural network trained by a Levenberg‐Marquardt algorithm is able to predict simultaneously all the analyzed experimental thermogravimetric curves with a high level of reliability and that it thus represents the highly effective artificial intelligence tool for the modelling of thermal stability also of relatively complicated friction composite systems.     

Trajectory control in the world coordinate system by an adaptive forecasting algorithm

A new approach to trajectory control of an industrial robot is proposed which aims to reduce the error in the trajectory, under the assumption that the position of the hand is known or can be measured in Cartesian space. An on-line system identification technique, referred to as the adaptive linear modelling technique, and a forecasting algorithm to generate corrective commands were used. A laser measurement system with a laser source and a photodetector was developed to detect the error between the actual and desired trajectories in Cartesian space. By implementing the proposed methods a 70% improvement in controlling the trajectory error was achieved.     

航天器太阳帆板伸展过程最优控制的粒子群优化算法

讨论航天器太阳帆板伸展过程中航天器姿态运动的最优控制问题。利用多体动力学方法导出带太阳帆板航天器姿态运动方程。在系统角动量为零的情况下,带太阳帆板航天器系统的姿态运动控制问题可转化为无漂移系统的非完整运动规划问题。在非完整运动规划中引入粒子群优化算法,通过控制太阳帆板伸展运动可同时获得航天器姿态的期望位形。数值仿真表明,该方法对太阳帆板伸展过程中航天器主体姿态控制是有效的。     

Short messages

This paper has three purposes. The first one is to explain to a general audience what is involved in retrieving a web page or performing some other complex network transaction, and what can make it slow, and why the problem of slowness is likely to get worse as networked applications become more complex. The second is to describe, to those who program networked applications, certain facts that we have learnt from modelling communication networks, notably the fact of heavy-tailed distributions in traffic, which may allow more efficient applications to be written. The third is to describe to network modellers an interesting class of problems relating to algorithm design for communication networks.     

Crosstalk between pathways enhances the controllability of signalling networks

The control of complex networks is one of the most challenging problems in the fields of biology and engineering. In this study, the authors explored the controllability and control energy of several signalling networks, which consisted of many interconnected pathways, including networks with a bow‐tie architecture. On the basis of the theory of structure controllability, they revealed that biological mechanisms, such as cross‐pathway interactions, compartmentalisation and so on make the networks easier to fully control. Furthermore, using numerical simulations for two realistic examples, they demonstrated that the control energy of normal networks with crosstalk is lower than in networks without crosstalk. These results indicate that the biological networks are optimally designed to achieve their normal functions from the viewpoint of the control theory. The authors’ work provides a comprehensive understanding of the impact of network structures and properties on controllability.Inspec keywords: genetics, numerical analysis, control theoryOther keywords: signalling network controllability, interconnected pathways, bow‐tie architecture, structure controllability, biological mechanisms, cross‐pathway interactions, numerical simulations, biological networks, control theory, gene regulatory network     

基于卷积神经网络的生物式水质监测方法

生物式水质监测通常是先通过提取水生物在不同环境下的应激反应特征,再进行特征分类,从而识别水质。针对水质监测问题,提出一种使用卷积神经网络(CNN)的方法。鱼类运动轨迹是当前所有文献使用的多种水质分类特征的综合性表现,是生物式水质分类的重要依据。使用Mask-RCNN的图像分割方法,求取鱼体的质心坐标,并绘制出一定时间段内鱼体的运动轨迹图像,制作正常与异常水质下两种轨迹图像数据集。融合Inception-v3网络作为数据集的特征预处理部分,重新建立卷积神经网络对Inception-v3网络提取的特征进行分类。通过设置多组平行实验,在不同的水质环境中对正常水质与异常水质进行分类。结果表明,卷积神经网络模型的水质识别率为99.38%,完全达到水质识别的要求。     

Hybrid Micropower Source for Wireless Sensor Network

Wireless sensor networks have become a very significant enabling technology in many applications and the use of environmental energy is a feasible source for low-power wireless sensor networks. The challenges of developing a power supply including generation or conversion, storage, and power management are manifold to extend the lifetime of a wireless sensor network. The objective of this research is to develop an intelligent hybrid power system to realize a self-sustaining wireless sensor node. The photovoltaic and thermoelectric generators are adopted as energy converters. The lithium ion battery and ultracapacitor are used as reservoirs. An intelligent power management system has been developed to control the power distribution. The design data and experimental results show that the hybrid micropower source can extend the lifetime of a sensor network.     

Neural network modelling of cold-cathode gauge parameters

This article describes modelling of the operating characteristics of a cold-cathode ionisation gauge (CCG). The gauge characteristics were measured on a gauge comparison UHV calibration system with a test chamber, an extractor gauge, a spinning rotor gauge, and a gas manifold with a precise leak valve. Discharge intensity was measured vs. anode voltage at different pressures selected in the range from 1×10⁻⁹ to 1×10⁻⁵ mbar, and vs. pressure at different operating voltages ranging from 1.2 to 9 kV. In all cases the magnetic flux density was the same and amounted to about 0.13 T. The CCG exhibits an extremely low thermal outgassing rate and a low measurement limit. Therefore, it is suitable for pressure measurements in the ultrahigh vacuum range; however, it has a significant disadvantage. The discharge current vs. the pressure characteristic is non-linear and, in some cases, even discontinuous.The measured CCG characteristics were used as an input for the artificial neural network, which was used to generate a non-linear CCG input-output function used for linearisation purposes. It is generally known and strictly proven that neural networks are capable of learning and building any kind of real and non-polynomial input-output function. Furthermore, it was also mathematically proven that the single hidden neural layer system can learn any function. Other authors have reported that the learned function characteristics are not always continuous.In our experimental work, no mapping discontinuities in the formed model were detected. Despite the fact that learning of the input-output characteristics can be obtained by the neural networks with only one hidden layer, we have used the multilayer neural networks that exhibit a faster convergent and smoother learning process. The neural networks were trained to perform the transfer function between the input gauge parameters and the pressure. The neural networks are a suitable solution for CCG characteristics modelling and thus offer the possibility to overcome the disadvantages of the CCG.