确定学习与基于数据的建模及控制

确定学习运用自适应控制和动力学系统的概念与方法,研究未知动态环境下的知识获取、表达、存储和利用等问题.针对产生周期或回归轨迹的连续非线性动态系统,确定学习可以对其未知系统动态进行局部准确建模,其基本要素包括:1)使用径向基函数(Radial basis function,RBF)神经网络;2)对于周期(或回归)状态轨迹满足部分持续激励条件;3)在周期(或回归)轨迹的邻域内实现对非线性系统动态的局部准确神经网络逼近(局部准确建模);4)所学的知识以时不变且空间分布的方式表达、以常值神经网络权值的方式存储,并可在动态环境下用于动态模式的快速识别或者闭环神经网络控制.本文针对离散动态系统,扩展了确定学习理论,提出一个根据时态数据序列对离散动态系统进行建模与控制的框架.首先,运用确定学习原理和离散系统的自适应辨识方法,实现对产生时态数据的离散非线性系统的未知动态进行局部准确的神经网络建模,并利用此建模结果对时态数据序列进行时不变表达.其次,提出时态数据序列的基于动力学的相似性定义,以及对离散动态系统产生的时态数据序列(亦可称为动态模式)进行快速识别方法.最后,针对离散非线性控制系统,实现了基于时态数据序列对控制系统动态的闭环辨识(局部准确建模).所学关于闭环动态的知识可用于基于模式的智能控制.本文表明确定学习可以为时态数据挖掘的研究提供新的途径,并为基于数据的建模与控制等问题提供新的研究思路.

Multi-dimension and multi-modal rolling mill vibration prediction model based on multi-level network fusion

Mill vibration is a common problem in rolling production,which directly affects the thickness accuracy of the strip and may even lead to strip fracture accidents in serious cases.The existing vibration prediction models do not consider the features contained in the data,resulting in limited improvement of model accuracy.To address these challenges,this paper proposes a multi-dimensional multi-modal cold rolling vibration time series prediction model(MDMMVPM)based on the deep fusion of multi-level networks.In the model,the long-term and short-term modal features of multi-dimensional data are considered,and the appropriate prediction algorithms are selected for different data features.Based on the established prediction model,the effects of tension and rolling force on mill vibration are analyzed.Taking the 5th stand of a cold mill in a steel mill as the research object,the innovative model is applied to predict the mill vibration for the first time.The experimental results show that the correlation coefficient(R^(2))of the model proposed in this paper is 92.5%,and the root-mean-square error(RMSE)is 0.0011,which significantly improves the modeling accuracy compared with the existing models.The proposed model is also suitable for the hot rolling process,which provides a new method for the prediction of strip rolling vibration.

Real-time road traffic states estimation based on kernel-KNN matching of road traffic spatial characteristics

The accurate estimation of road traffic states can provide decision making for travelers and traffic managers. In this work,an algorithm based on kernel-k nearest neighbor(KNN) matching of road traffic spatial characteristics is presented to estimate road traffic states. Firstly, the representative road traffic state data were extracted to establish the reference sequences of road traffic running characteristics(RSRTRC). Secondly, the spatial road traffic state data sequence was selected and the kernel function was constructed, with which the spatial road traffic data sequence could be mapped into a high dimensional feature space. Thirdly, the referenced and current spatial road traffic data sequences were extracted and the Euclidean distances in the feature space between them were obtained. Finally, the road traffic states were estimated from weighted averages of the selected k road traffic states, which corresponded to the nearest Euclidean distances. Several typical links in Beijing were adopted for case studies. The final results of the experiments show that the accuracy of this algorithm for estimating speed and volume is 95.27% and 91.32% respectively, which prove that this road traffic states estimation approach based on kernel-KNN matching of road traffic spatial characteristics is feasible and can achieve a high accuracy.

Preliminary analysis of spatiotemporal pattern of global land surface water

Land surface water(LSW) is one of the most important resources for human survival and development, and it is also a main component of global water recycling. A full understanding of the spatial distribution of land surface water and a continuous measuring of its dynamics can support to diagnose the global ecosystem and environment. Based on the Global Land 30-water 2000 and Global Land 30-water 2010 products, this research analyzed the spatial distribution pattern and temporal fluctuation of land surface water under scale-levels of global, latitude and longitude, continents, and climate zones. The Global Land 30-water products were corrected the temporal inconsistency of original remotely sensed data using MODIS time-series data, and then calculated the indices such as water area, water ration and coefficient of spatial variation for further analysis. Results show that total water area of land surface is about 3.68 million km2(2010), and occupies 2.73% of land area. The spatial distribution of land surface water is extremely uneven and is gathered mainly in mid- to high-latitude area of the Northern Hemisphere and tropic area. The comparison of water ratio between 2000 and 2010 indicates the overall fluctuation is small but spatially differentiated. The Global Land 30-water products and the statistics provided the fundamental information for analyzing the spatial distribution pattern and temporal fluctuation of land surface water and diagnosing the global ecosystem and environment.