神经网络ART模型在故障诊断中的应用

讨论了目前最成功的一种无导师神经网络模型──自适应谐振理论ＡＲＴ。分析了ＡＲＴ的工作原理，给出了ＡＲＴ的具体算法（已在ＰＣ－４８６上用Ｃ语言实现）；指出了ＡＲＴ的实质，并以“有轨自动物料搬运小车系统”为例详述了ＡＲＴ在故障诊断中的工作过程，获得了很好的结果。     

一种面向模式分类的修正的ART1神经网络

本文提出了一种关于ｎ维布尔模式分类的，基于欧氏距离的判定规则，在ＡＲＴ１神经网络的基础上，设计了一种实现这一判定规则的神经网络（ＭＡＲＴ１）及其学习规则，并给出了计算机仿真结果。     

一种主存数据库的存取方法

本文给一个主存数据库管理系统ＡＲＴｓ－ＭＭＤＢＳ，着重讨论了它的数据库管理机制与存取方法并具体阐述了它的一种新的索引结构－－ＳＢ树。     

神经计算在故障诊断中的应用

故障诊断的基本问题是故障的分类问题。神经网络有很强的分类能力，并且有着特有的联想记忆、并行推理和抗干扰的能力，这些使得它成为研究故障诊断的一个新的课题。本文介绍了神经网络几种典型的模型：ＢＰ网、Ｈｏｐｆｉｅｄ、网、ＢＡＭ网和ＡＲＴ网在故障诊断中的应用。     

SCCM并行计算模型及其上的优化算法设计

本文提出了基于同步无缓冲通信的多计算机系统的一种新的并行计算模型－－ＳＣＭＭ模型，给出了其上的一些算法的优化设计例子，并在Ｔｒａｎｓｐｕｔｅｒ多机系统上设计和实现了图像重建里的ＡＲＴ算法。     

PARTER：一种适用于进行纹理识别的并行系统

９４０３４ＰＡＲＴＥＲ：一种适用于进行纹理识别的并行系统／／ＭｉｃｒｏｐｒｏｃｅｓｓｏｒｓａｎｄＭｉｃｒｏｓｙｓｔｅｍｓ－ｌ９９３，１７（２）．－９３～９９ＰＡＲＴＥＲ是一种使用并行处理技术，完成数字图象上静态纹理识别的系统。该系统装配有训练程序，能...     

一种改进的ART1算法及其在人像识别中的应用

本文通过对自适应共振理论（ＡＲＴ）ＡＲＴ１的研究,提出了一种改进的ＡＲＴ１算法,这种算法不仅具有ＡＲＴ１算法的所有优点,而且降低了ＡＲＴ１算法的识别识误差。该算法在人像识别中取得了令人满意的效果。     

对FTART算法的研究及改进

ＦＴＡＲＴ（ｆｉｅｌｄｔｈｅｏｒｙ－ｂａｓｅｄＡＲＴ）算法结合了ＡＲＴ（ａｄａｐｔｉｖｅｒｅｓｏｎａｎｃｅｔｈｅｏｒｙ）算法、ＡＲＴＭＡＰ算法、域理论的思想，以样本在实例空间中出现的概率为启发信息修改学习中生成的分类，采用了不同于其它算法的解决样本间的冲突和动态扩大分类区域的方法．本文在对ＦＴＡＲＴ算法的研究的基础上进行了改进，使算法在学习连续函数的映射时更加有效．同时给出了算法的测试结果和对测试结果的分析，测试表明，ＦＴＡＲＴ算法在模式识别和连续函数映射的学习方面具有比较好的性能.     

HART通信多站监控仪HM301及其应用

ＨＡＲＴ协议是智能仪表数字通信事实上的工业标准。本文介绍了ＨＡＲＴ通信多站监控仪ＨＭ３０１的硬件和软件，功能特点及其应用。文中介绍了ＨＡＲＴ协议物理层、数据链层和应用层的实现方法。实践应用表明：ＨＭ３０１是一种先进，方便而又经济的智能仪表ＨＡＲＴ通信应用方式     

一种单片机管理的外总线CRT适配器

本文研究了一种新型ＣＲＴ适配器，依靠自带的８０３１单片机完成ＣＲＴ显示器的控制与管理，适配器中固化了国家标准（ＧＢ２３１２－８０）中１～８７区的所有汉字和符号，它采用Ｃｅｎｔｒｏｎｉｓ３６芯标准接口，命令和参数全部用ＡＳＣＩＩ码传递，任何主机只需象连接打印机或绘图仪那样向适配器发送命令，即可在ＣＲＴ上显示字符、汉字以及各种图形，因而特别适合在微机化仪器仪表与工业控制系统中使用。     

融合DSTM和USTM方法的主题模型

当前监督或半监督隐藏狄利克雷分配（latent Dirichlet allocation,LDA）模型多数采用DSTM（down-stream supervised topic model）或USTM（upstream supervised topic model）方式加入额外信息,使得模型具有较高的主题提取和数据降维能力,然而无法处理包含多种额外信息的学术文档数据。通过对LDA及其扩展模型的研究,提出了一种将DSTM和USTM结合的概率主题模型ART（author ＆amp; reference topic）。ART模型分别以USTM和DSTM方式构建了文档作者和引用文献的生成过程,因此可以对既包含作者信息又包含引用文献信息的文档进行有效的分析处理。在实验过程中采用Stochastic EM Sampling 方法对模型参数进行了学习,并将实验结果与Labeled LDA和DMR模型进行了对比。实验结果表明,ART模型不仅拥有高效的文档主题提取和聚类能力,同时还拥有优良的文档作者判别和引用文献排序能力。     

Observations and problems applying ART2 for dynamic sensor patterninterpretation

This paper discusses characteristics of the ART2 (adaptive resonance theory) information processing model which emerge when applied to the problem of interpreting dynamic sensor data. Fast learn ART2 is employed in a supervised learning framework to classify process “fingerprints” generated from multi-sensor trend patterns. Interest in ART2 was motivated by the ability to provide closed classification regions, uniform hyperspherical clusters, feature extraction, and on-line adaption. Sensor data interpretation is briefly discussed with an emphasis on the unique attributes of the problem and the interaction with ART2 information processing principles. Pattern representations, e.g., time domain, which encode information in both magnitude and direction of the input vector are shown to be fundamentally incompatible with ART2. Complement coding is shown to solve this problem when the feature extraction capability of the ART2 network is disabled. Complement coding is also shown to preserve the clustering characteristics of the process “fingerprints” which are otherwise lost using the ART2 directional similarity measure. These issues are illustrated using an ART2-based monitoring system for a dynamically simulated chemical process     

Syndrome-decoding algorithms for static-diagnosis models

The problem of finding fault patterns consistent with a given syndrome is discussed for graph-theoretical diagnosis models such as the fault-diagnosis and self-diagnosis models. The fault-diagnosis model consists of two types of vertices, fault units and measurements, and is expressed by a bipartite graph. Faulty states of a fault unit always imply abnormal states of all the measurements which are adjacent to the unit, otherwise a measurement remains normal. A self-diagnosis model consists of one type of unit which has the capability of testing other units and being tested itself. The testing relation is represented by a directed arc; this produces test outcomes which are invalid if the testing unit is faulty. The inverse system which yields a fault pattern from a corresponding syndrome for fault-diagnosis models is studied and a syndrome-decoding algorithm is proposed which works for some class of diagnosis models with observation noise. The algorithm uses a similar measure to the syndrome-decoding algorithm of error-correcting codes which use the Hamming distance. Another measure is presented for the self-diagnosis model expressed by a directed graph and this measure is characterized by a ranking method.     

Generation of a diagnosis model for hybrid-electric vehicles using machine learning

Online fault-diagnosis on system level for complex mechatronic systems takes multiple sensor measurements of the various components into account and contributes to a significantly increased system reliability by tracking down faults in the system at run time, enabling fault-specific recovery actions, such as reconfigurations. Ongoing efforts in the technological development of automobiles, especially in the field of driver assistance systems, yield more and more safety-critical systems, e.g., breaking control systems, and thus generate a high demand for reliable online diagnosis systems. In order to perform fault-diagnosis on system level, the interrelations between all measurements must be determined, which is a challenging and often demanding task done by human system experts. In this paper we present a systematic approach based on machine learning to establish online diagnosis for a hybrid-electric vehicle model in the context of the DAKODIS research project. With this paper we publish the Matlab/Simulink HEV research platform including a fault injection framework and data processing algorithms for active fault-diagnosis and recovery evaluations.     

Adaptive random testing based on distribution metrics

Random testing (RT) is a fundamental software testing technique. Adaptive random testing (ART), an enhancement of RT, generally uses fewer test cases than RT to detect the first failure. ART generates test cases in a random manner, together with additional test case selection criteria to enforce that the executed test cases are evenly spread over the input domain. Some studies have been conducted to measure how evenly an ART algorithm can spread its test cases with respect to some distribution metrics. These studies observed that there exists a correlation between the failure detection capability and the evenness of test case distribution. Inspired by this observation, we aim to study whether failure detection capability of ART can be enhanced by using distribution metrics as criteria for the test case selection process. Our simulations and empirical results show that the newly proposed algorithms not only improve the evenness of test case distribution, but also enhance the failure detection capability of ART.     

Mirror adaptive random testing

Recently, adaptive random testing (ART) has been introduced to improve the fault-detection effectiveness of random testing for non-point types of failure patterns. However, ART requires additional computations to ensure an even spread of test cases, which may render ART less cost-effective than random testing. This paper presents a new technique, namely mirror ART, to reduce these computations. It is an integration of the technique of mirroring and ART. Our simulation results clearly show that mirror ART does improve the cost-effectiveness of ART.     

An ART1 microchip and its use in multi-ART1 systems

Recently, a real-time clustering microchip neural engine based on the ART1 architecture has been reported. However, that chip rendered an extremely high silicon area consumption of 1 cm(2), and consequently an extremely low yield of 6%. Redundant circuit techniques can be introduced to improve yield performance at the cost of further increasing chip size. In this paper we present an improved ART1 chip prototype based on a different approach to implement the most area consuming circuit elements of the first prototype: an array of several thousand current sources which have to match within a precision of around 1%. Such achievement was possible after a careful transistor mismatch characterization of the fabrication process (ES2-1.0 mum CMOS). A new prototype chip has been fabricated which can cluster 50-b input patterns into up to ten categories. The chip has 15 times less area, shows a yield performance of 98%, and presents the same precision and speed than the previous prototype. Due to its higher robustness multichip systems are easily assembled. As a demonstration we show results of a two-chip ART1 system, and of an ARTMAP system made of two ART1 chips and an extra interfacing chip.     

An extension of a fuzzy reputation agent trust model (AFRAS) in the ART testbed

With the introduction of web services, users require an automated way of determining their reliability and even their matching to personal and subjective preferences. Therefore, trust modelling of web services, managed in an autonomous way by intelligent agents, is a challenging and relevant issue. Due to the dynamic and distributed nature of web services, recommendations of web services from third parties may also play an important role to build and update automated trust models. In this context, the agent reputation and trust (ART) testbed has been used to compare trust models in three international competitions. The testbed runs locally and defines an ART appraisal domain with a simulation engine, although the trust models may be applied to any kind of automated and remote services, such as web services. Our previous works proposed an already-published trust model called AFRAS that used fuzzy sets to represent reputation of service providers and of recommenders of such services. In this paper we describe the extension required in the trust model to participate in these competitions. The extension consists of a trust strategy that applies the AFRAS trust model to the ART testbed concepts and protocols. An implementation of this extension of AFRAS trust model has participated in the (Spanish and International) 2006 ART competitions. Using this ART platform and some of the agents who participated, we executed a set of ART games to evaluate the relevance of trust strategy over trust model, and the advantage of using fuzzy representation of trust and reputation.     

用于模式识别的ART-2神经网络算法的改进

针对模式识别中模式有序输出的要求,对ＡＲＴ－２神经网络的算法进行了改进和调整,提出了ＡＲＴ－２神经网络的改进算法,通过对改进算法与原算法的识别试验结果进行比较,表明该改进算法对模式的有序输出是可行的和有效的。