Probabilistic characterisation of model error using Gaussian mixture model—With application to Charpy impact energy prediction for alloy steel

A novel approach to characterise the model prediction errors using a Gaussian mixture model is proposed. The motivation for this work lies behind many data models that are developed through prediction error minimisation with the assumption of a normal noise distribution. When the noise is non-normal, which may often be the case in complicated data modelling scenarios, the model prediction errors may contain rich information, which can be further exploited for model refinement and improvement. The key contents presented in this paper include: choosing the relevant variables to form the error data, optimising the number of Gaussian components required for the error data modelling, and fitting the Gaussian mixture parameters using an expectation-maximisation algorithm. Application of the proposed method for further model improvement, within the framework of hybrid deterministic/stochastic modelling, is also discussed. Preliminary results on the real industrial Charpy impact energy data for heat-treated steels show its effectiveness for model error characterisation, and the potential for model performance improvement in terms of prediction accuracy as well as providing accurate prediction confidence intervals.     

通用网在模拟无序信息传输中的应用

传输信息常涉及到不同的难题。当保护信息免受非控制访问时，一个新的似然性可能得以应用，即无序。由于在安全通信中实施无序的兴趣日见增加，已经提出许多设计方案。本文提出了建立非线性动态无序系统模型的方法，并进行了通用网的设计与访真。     

Combined Petri net modelling and AI-based heuristic hybrid search for flexible manufacturing systems—part II. Heuristic hybrid search

This two-part paper presents modelling and scheduling approaches of flexible manufacturing systems using Petri nets (PNs) and artificial intelligence (AI)-based heuristic search methods. In Part I, PN-based modelling approaches and basic AI-based heuristic search algorithms were presented. In Part II, a new heuristic function that exploits PN information is proposed. Heuristic information obtained from the PN model is used to dramatically reduce the search space. This heuristic is derived from a new concept, the resource cost reachability matrix, which builds on the properties of B-nets proposed in Part I. Two hybrid search algorithms, (1) an approach to model dispatching rules using analysis information provided by the PN simulation and (2) an approach of the modified stage-search algorithm, are proposed to reduce the complexity of large systems. A random problem generator is developed to test the proposed methods. The experimental results show promising results.     

压力容器壳体曲面造型与展开误差分析

采用基于特征的多面体造型方法，解决了压力容器壳体曲面造型的问题，同时，提出了一种筒体展开的算法，并进行了误差分析与修正。     

基于边界表示的曲面实体造型系统的数据结构

实体造型是ＣＡＤ系统的核心模块，而二次曲面在工程中常用来描述轴，孔及圆角等形状特征。讨论了在微机上实现的一个基于边界表示的实体模型的数据结构，并针对二次曲面的特点，实现了对自封闭曲面的无二义性拓扩定义，为顺利进行曲面的拼合运算打下了基础。     

Performance comparison of 100Base-T fast ethernet and 100VG-AnyLAN

There are two competing technologies for 100 Mbps Ethernet, namely IEEE 802.3u 100Base-T Fast Ethernet and IEEE 802.12 100VG-AnyLAN. The former technology is based on the same medium access control protocol as the popular IEEE 802.3 CSMA/CD network, while the later is based on a new demand priority protocol. Contrary to common belief, a test conducted recently by Newman and Levy using commercially available adaptor cards and hubs showed that 100Base-T networks outperformed 100VG-AnyLAN networks with similar configurations by a significant margin. From the analytical models developed, we found that the anomaly is due to the larger-than-expected overheads incurred by the 100VG-AnyLAN hub in polling every station in the network in a round robin manner.     

Simulation based on learning methods

The use of simulation technology as a tool for planning and control is of increasing significance in most fields of production. The main part of the expenditure concerning simulation analyses is the modelling of the considered production. Despite the use of modern building-block-oriented modelling technology, this modelling can often not be done by the user, but only by external experts. Against this backdrop, an adaptive simulation system is being developed by the Institute for Industrial Manufacturing and Management (IFF) at the University of Stuttgart. It independently adapts to real production processes, i.e. it learns about the interdependencies of production processes, and, in this way, supports the user in constructing and maintaining the model. In terms of information technology, the research in the field of artificial intelligence, especially in the subdomain of machine learning, is the basis for the realization of such adaptive systems.     

Modelling and FDI of Dynamic Discrete Time Systems Using a MLP with a New Sigmoidal Activation Function

In this paper we investigate the use of the multi-layer perceptron (MLP) for system modelling. A new sigmoidal activation function is introduced and the study is focused at the utilization of this function on a MLP that performs modelling of dynamic, discrete time systems. The role of the activation function in the training process is investigated analytically, and it is proven that the shape of the activation function and it's derivative can affect the training outcome. The method is simulated at a well known benchmark, namely the three tank system, and is incorporated in a Fault Detection and Identification (FDI) method, also applied and simulated at the three tank system. Finally, a comparison is made with an approach that utilizes local model neural networks for system modeling.     

Sensitivity of spectral reflectance to variation in live fuel moisture content at leaf and canopy level

Wildland fires burn large areas of the earth's land surface annually, causing significant environmental damage and danger to human health. In order to mitigate the effects, and to better manage the incidence of such fires, fire behaviour models are used to predict, among other things, the likelihood of ignition, the rate of spread, and the intensity and duration of burning. A key input parameter to these models is the amount of water in the vegetation, described as the fuel moisture content (FMC). A number of studies have shown that vegetation indices (VI) calculated from red and NIR reflectances may be used to map spatial and temporal variation in FMC in a range of fire-prone environments, with varying degrees of success. Strong empirical relationships may be established between VI and FMC over grasslands, yet over shrublands and forests, the relationships are weaker. If FMC is to be estimated with greater accuracy and consistency than is currently achieved, it is necessary to fully understand the relative contribution that spatial and temporal variation in the various biophysical and geometrical variables make to reflectance variability at the leaf and canopy level.This paper uses a modelling approach to investigate the sensitivity of reflectance data at leaf and canopy level to variation in the biophysical variables that are used to compute FMC. At the leaf level, the results show that the sensitivity of reflectance to variation in leaf water and dry matter content, used to compute FMC, is greatest in the SWIR and NIR, respectively. Variation in FMC has no effect in the visible wavelengths. At the canopy level, the results show that the sensitivity of reflectance to variation in leaf water and dry matter content is heavily dependent upon the type of model used and the range of variation over which the variables are tested. In the longer wavelengths of the SWIR, the competing influence of variable leaf area index, fractional vegetation cover, and solar zenith angle is shown to be greater than that at the shorter wavelengths of the SWIR and NIR. Empirical relationships between the normalised difference water index (NDWI) and FMC are shown to be weaker than that with canopy water content. However, when the range of the variables under study is more limited, useful empirical relationships between FMC and remotely sensed VI may be established.     

Characterization of indoor wireless channel at 5 GHz U-NII bands

The paper reports on the experiments undertaken at the University of Wollongong to characterise fading profiles and delay parameters of an indoor wireless channels at 5 GHz U-NII bands. The measurements were undertaken at different locations around the campus with results recorded for a post-processing to calculate the Rician K-factor, the level crossing rate and the average fade duration as well as mean excess delay, rms delay spread, and the coherence bandwidth of the channel. The presented measurement results can be useful in developing a Markov chain based model of the transport channel for IEEE802.11a or HYPRLAN-2 networks. The results also indicate scenarios where the coherence bandwidth of the channel is smaller than the width of the sub-carrier OFDM channels in either of the mentioned systems.