神经式网络

Neural network technology is used to acquire the hidden knowledge from the results to realize the effective reasoning and application of knowledge. And rough set theory is applied to reduce and unitize the experiential knowledge to find out the important attributes affecting the decision-making and extract the expert design knowledge from the simulation and experimental results.

运用神经网络技术从试验数据中获取隐式知识，实现对知识的联想、推理等高效运用，同时运用粗集理论对经验性知识进行约简，从而找出对决策信息具有重要影响的属性，以及隐含在试验结果中的专家级设计知识。

In addition to,the implicit camera system model is made with BP neural network simulating the relationship between the 3D objects and its images in the stereo system,which avoiding the system errors caused by unperfected mathematic relation.

又运用BP神经网络来模拟立体视觉系统三维空间与二维图像平面之间的物、像对应关系，建立了双目立体视觉系统的摄像机隐式标定模型，避免了因数学模型的不完善而带来的系统误差。

Database of physical and chemical properties for solid wastes were constructed through investigation for single typical component and their mixture of solid wastes. The particulate trajectory model and heat transfer model were developed, based on the studies of Characteristics of solid wastes movement and heat mass transfer in rotary kiln. Pyrolysis experiments of solid wastes in lab and pilot scale rotary kiln pyrolyzers were performed. Mechanism of pyrolysis for typical solid wastes was analyzed. The characteristics of pyrolytic products, such as physical and chemical properties, composition and combustibility, etc, were investigated. A neural network model for the prediction of yields and properties of pyrolysis products was developed. Then, the potential applications of pyrolytic products and the substitution of pyrolytic fuels for corresponding commercial furls were investigated. The optimization of energy recovery and utilization for different wastes through pyrolysis was analyzed.

通过对典型固体废物组分及其混合物特性的分析，建立了固体废物的化学特性分析数据库，并据此进行了物理分类；对固体废物在回转窑内的运动和传热特性进行了研究，建立了固体物料的随机颗粒滚动理论模型和传热模型；在小型和中试回转式热解炉上进行了实验；对各种典型固体废物在回转窑内的热解机理和主要热解产物的性质进行了研究；利用神经网络等方法建立了各典型固体废物的热解产物的产率及特性的数学预测模型；对固体废物热解产物的物性、成分、燃烧特性等开展了研究，分析了其与现有商业燃料匹配的可能性；对热解产物的多种应用性能进行了研究，对不同废物热解中的能源回收和利用进行了优化分析。

To avoid the collision accidents of ships at sea, with the universalizing and application of computer, and the development and maturity of the technology of Expert System and Artificial Intelligent, this paper developed an intelligent collision avoidance expert system for navigation. The main research work centers on several aspects, which will be represented as below:By the navigation rules" understanding and analyzing, and by the navigation experience and navigation samples" collecting and trimming, we put forth and build a multi-unit and layering KB systematic structure, and implement the KBM. According to features of different knowledge, we adopt multifarious KR, such as: frame KR, production rule KR, procedure KR. We also build a multi-inference system, which based on analog inference, forward illation inference, conversion inference and meta-rule inference. At the same time, we develop each reasoning algorithm. For some problems in collision avoidance region during the building of the expert system, we put forth and build a set of models to solve them using neural network technology.

为了避免船舶间碰撞事故的发生，结合计算机技术在各类船舶中的普及与应用、专家系统技术及人工智能技术的发展、成熟，本文研制、开发了一种具有一定智能的航海避碰专家系统，主要内容包括以下几个方面：通过对航海规则的理解与分析，对航海经验、航海实例的搜集、整理，提出并建立起了航海避碰专家系统的多元分层知识库体系结构，并实现了知识库的管理；根据不同知识的特点，分别采用了框架、产生式规则、过程等多种知识表示方法；提出了一种基于类比推理、正向演绎推理、换位推理及元级推理等的多种推理机制，并建立起了相应的推理算法；引入神经网络技术，针对在建立专家系统过程中所遇到的有关避碰领域内的一些难题，提出并建立了相应解决问题的模型，其中包括：船舶类型的识别，会遇态势的分类及避碰危险的评估。

When based on dual wavelength chromatographic data analysis, a new method, namely dual wavelength characteristic information analysis, used for the base line correction, determination of number of components and region of pure components signal.

但色谱峰的重叠给多组分的同时分析带来了困难，因此针对色谱重叠峰的解析提出了不少方法，如曲线拟合、神经网络、小波变换、窗口因子分析、直观推导式演进特征投影分析等 [1～ 7] 。

In our work here,a model of data mining was developed,which got its foundation from Artificial Neural Networks.hi fact,this kind of model might be called an infant protocol for DSS,which accepts final users raw data, integrates various sources of data,cleanses them from garbages,then normalizes them into an intermediatary data file.

我们在这里设计并实现了一个基于具有模糊式输出的人工神经网络的数据挖掘模型，这个模型实际上是一个具有智能的决策支持系统的雏形，能够对用户所给的原始数据进行处理，利用已经获得的知识对用户数据进行判断，并以适当的方式向用户解释系统决策的结果。

With this measure, a new method for calculating the minimum embedding segment dimension of a linear segment sequence is proposed. With this method, the MESD of the linear segment representation of a real financial time series is calculated.

该方法实现了对时间序列的基于基本变化形态的合理分割，保证转换后的每一个符号都能代表一个基本的、相对独立的变化模式，为最终挖掘结果的有效性和可视化提供保证；同时，针对数据挖掘数据量大，对算法的计算效率和在线性能要求高的特点，提出了一个实现时间序列分段线性化的增量式更新算法；并利用神经网络模糊聚类算法实现了完全数据驱动的在线聚类分析。

In the feedback BP neural network design process,using the dynamic response analysis method,can effective analysis of system stability.

在反馈式BP神经网络的设计过程中，采用动态响应分析的方法，可有效地分析系统的稳定性。

Experimental study had been carried out by using Mandarin emotional speech database recoded and emotional speech database respectively.

首先以隐马尔可夫模型和人工神经网络为基础，设计了三种分类器；然后用改进的排序式选举算法，实现对三种分类器的融合。

The towed buoy method acquires relative position information by analyzing the towed cables' 3-D spatial shapes,the free buoy method sets up and trains neural network models,based on the output information obtained by inputting the collected data into above models,extends the position of buoy toward.

拖曳式浮标法通过分析拖曳线缆在水中的三维空间形态获得相对位置信息；自由式浮标法为建立与训练神经网络模型，通过将采集数据输入此模型获得的输出信息，将浮标的位置向水下延伸。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。