Discussion on Some Problems of ExpertSystem for Rock Mass Classification

On the basis of discussing present situation for classification of rock masses and necessity of building expert system of rock masses.The intelligent design software for classification of rock masses CRES has been completed. In this paper the CRES is expounded systematically in the following,such as knowledge_base,data_base,reasoning machine,learning machine,and so on.Finally,the CRES is used for verifying 30 engineering examples and is used to classify rock masses of two roadways successfully .The CRES has been connected to SBSDES.     

Risk of erosion in peat soils – an investigation using Bayesian belief networks

Peat is an important carbon sink in the context of climate change. However, well‐documented examples suggest that risk of peat erosion is widespread and significant. Our understanding of peat vulnerability to erosion is commonly constrained by the complexity of drivers, and their interactions, in this process. However, the key constraints are: limited, consistent and comprehensive quantitative data relating to this process and, more significantly, the explicit relationships between the occurrence of peat erosion and its causes and drivers. Bayesian belief networks (BBNs) provide a methodology for integrating qualitative and quantitative knowledge. BBNs can capture and structure available knowledge and rationalize complex interactions, where empirical data are limited or poorly compatible and processes are complex or uncertain. In this article we explore the BBN potential to advance our understanding and to identify gaps in current knowledge. BBN has been demonstrated to be a useful tool in structuring and utilizing currently available knowledge, often with limited evidence, of peat’s actual exposure to erosive forces. Despite considerable research into peat erosion processes and understanding the inherent vulnerability of peat, results presented indicate clear gaps in knowledge regarding the role of land management, spatially explicit data related to land management as well as limited evidence of the relevant relationships between many of the variables. The attention of further research will focus on these gaps. The BBN approach provides a framework in which different scenarios of biophysical, climatic and land management (social and economic) conditions can examine and assess the probability of erosion.     

证据推理遥感分类背景下遥感影像证据支持度生成方法的对比

选择黑龙江省洪河自然保护区的ETM+遥感影像,对比分析了基于最小距离法(MD)和频率分布法(FD)两种不同证据支持度生成方法的证据推理遥感分类精度.结果表明,在使用前3个或更少的波段情况下,MD方法证据支持度质量比MD方法好,而在前4个或更多的波段情况下,MD方法生成的证据支持质量更高.证据支持度的合理与否直接影响证据...     

基于EMD和IDAR的灌区年降水量预测研究

针对降水量这样一种非线性、非平稳序列,研究经验模态分解方法(EMD)和信息扩散近似推理方法（IDAR）在年降水量预测中的组合应用,解决资料序列不充分情况下的区域降水量预测问题。首先,通过EMD方法对具有典型非线性与非平稳特征的年降水量时间序列进行处理,分解出包含原信号不同特征尺度的分量数据系列;然后应用信息扩散近似推理技术对各降水量分量间的复杂非线性关系进行描述,建立当前趋势以及相邻年份之间的预测规则,并进行预测。以文献案例灌区长系列降水资料为样本进行实例计算,并与其它预测方法进行了对比。结果表明：基于EMD和信息扩散近似推理的预测方法效果较好,误差绝对值和为1.30,优于人工神经网络、线性自回归方法以及单纯信息扩散近似推理的统计结果。同时,为了验证该方法的适用性,将该方法应用于文峪河灌区的降雨量预测,取得了满意的效果。研究中发现：信息扩散近似推理可将样本点转换成模糊集,部分弥补了由于数据的不完备性所造成的信息空白,并可将矛盾模式转换成兼容模式。而EMD方法可有效分解具有非线性、非平稳特征的降水序列,保留其原序列在空间（或时间）各种尺度上的分布规律。两者结合对解决样本不完备的非平稳序列的预测问题是非常有价值的。通过与其它预测方法比较,发现该模型能够很好地光滑样本数据以及能够较好地发掘知识,有较高的预测精度和推广应用价值。     

Product ontology screening and its application in variant design

A product ontology screening method which can be used in variant design is proposed. With the support of complete ontology view of product resources, the product functional requirement and structural characteristics are converted to demand constraints, which are used to screen product ontology. For different screening hierarchies and objects, different screening rules such as pattern matching, case matching combined with manual decision and auto matching on the basis of case-based reasoning are adopted. The semi-automatic screening method with human-computer cooperation has impressive effectiveness on the improvements of the accuracy of design schemes and the flexibility and efficiency of design process. Finally, a variant design supporting system for building door and window is developed, which proves the feasibility and effectiveness of the method.     

云制造环境下知识云能力量化评估方法研究

云制造是一种智能化制造新模式,知识是支撑云制造服务系统运行的核心。但对知识云及其服务能力、影响因素和评估方法的研究尚不够深入。在建立知识云能力整体特征评估体系基础上,提出了一种基于云推理的量化评估方法。通过对知识能力评估指标的云化处理,采用知识能力云的期望和超熵对服务效能满意度进行定量评价,并构建了知识能力服务效能评估云推理发生器,实现通过知识能力变化云描述服务满意度的变化情况,为最大限度保障云制造系统知识资源与知识能力共享和按需使用提供智能化支持。实验表明,该方法较好地实现了知识云能力的最大效能服务过程,对云制造环境下知识云能力服务效能综合评估进行了有益的探索。     

船舶碰撞危险综合评价系统的比较研究

应用模糊推理系统、神经网络系统以及基于自适应神经网络的模糊推理系统对船舶碰撞危险评价进行了研究,并对这三种推理机制进行了比较。结果表明,基于自适应神经网络的模糊推理系统既能表达模糊语言、利用已有的知识,又具有前向网络的学习能力,是一种非常有前途的评价方法。     

基于神经网络的模糊逻辑推理系统模型

采用多层前向神经网络模型及BP学习算法, 建立了基于神经网络的模糊逻辑推理方法和基于神经网络的模糊系统模型, 从而将模糊逻辑和神经网络集成融合为一个有效的功能性系统。     

AN INEXACT REASONING INFORMATION MANAGEMENT SYSTEM

The author introduces some basic ideas and methods of artificial intelligence into an information management system,and therefore proposes a design for the information management system which has the ability of inexact reasoning based on the certainty knowledge. According to the design, the system can also be made with the flexibility of change and expansion, and good adaptability for application. This paper describles the design and its programing skill in detail.     

基于范例推理的人机界面设计系统

针对人机界面设计与评价在一定程度上依赖于专家知识和以往经验的特点,将基于范例推理(CBR)引入到人机界面设计中,提出了基于范例推理的人机界面设计与评价系统架构,并分析了架构的组成。研究了实现系统所采用的范例知识表示、相似性度量算法等关键技术。在拖拉机人机界面设计中的应用结果表明,依据提出的架构及其关键技术,易于实现人机界面设计知识的重用和创新,提高设计工效。