Papers by Keyword: Time-Series

Abstract: Time series AR models (AutoRegressive models) are established with the use of the colored noise filtered by vibration signals,so as to study vibration characteristics of the magnetorheological damper under different working currents and frequencies .Because vibration signals are non-stationary and nonlinear , signal processing tool of trispectrum is introduced.The trispectrum are utilized for analyzing dynamic frequency characteristics of the magnetorheological damper. The result shows that the trispectrum slices can quantificationally describe nonlinear coupling, restrain gaussian noise and reserve phase components. Characteristics of the trispectrum is closely related to the current and frequency .The trispectrum slices can be used to describe laws of vibration system effectively.

Abstract: The paper used Matlab to program for single-layer BP neural network, selected the annual energy consumption data as the training sample and inspection sample to train BP neural network, then, predict the future China's energy consumption quantity.

Abstract: Dynamic time warping(DTW) distance is the most effective similarity measurement methods in time-series data mining area. Because of the high computational complexity of DTW, it is not suitable for huge amounts of data. Thus, some DTW distance lower bound methods are proposed, which can improve the efficiency of DTW distance calculation. In this paper, we summarized the existing lower bound of DTW method and then proposed a closely related field with the application and exacted Lower Bounding distance measure with Segmentation(eLBS). Experiment results show that this method meet the need of non-omission.

Abstract: This paper introduces a method to apply time series analysis in dam deformation monitoring and prediction. We provide a simplified AR prediction model, which is relatively optimized in fitting constructive dynamic deformation features, analyzing deformation data and predicting deformation trend. We use this AR model in a certain dam’s deformation data processing, and prove it is an effective dynamic deformation prediction model.

Abstract: The paper has made a maximum Lyapunov exponent and Lyapunov exponent spectrum entropy analysis of ASCE Benchmark using non-linear theory and chaos time sequence. The maximum Lyapunov exponents in the two kinds of structural monitored data are both over zero, indicating that in the structural system chaos phenomenon has appeared. And, experiments have shown that the maximum Lyapunov exponent is sensitive of the amount of samples and the time delay. So, to compute the chaos index, the amount of samples and the time duration are of importance. Meanwhile, the Lyapunov exponent spectrum entropy is effective to measure the chaotic characteristic of the system, but ,the entropy is less sensitive to state recognition more than the max Lyapunov exponent.

Abstract: Analysis of rock acoustic emission characteristics in complete stress-strain process under uniaxial compression shows that the evolution process of rock deformation and damage can be divided into five stages—compaction, linear elastic deformation, accelerate inelastic deformation, damage and development, plastic flow. Acoustic emission characteristics of rock can well reflect its compressive deformation and damage evolution process. Acoustic emission time series of typical plastic coal is comparatively dispersed and presents group shock type. The acoustic emission frequency events and their energy of coal are bigger around the damage point, but the maximum amplitude and the secondary maximum amplitude have not much difference. The acoustic emission events and their energy of very uniform and dense fragile limestone are mainly concentrated in a very short period of time before its damage and the acoustic emission time series manifests as isolated shock type. For sandstone whose homogeneous degree lies between typical plastic coal and very uniform and dense fragile limestone, its acoustic emission time series manifest as main shock type.

Abstract: Owing to the distinctiveness of business operation mode and profit acquisition style between commercial website and entity enterprise, traditional customer equity measurement model is not applicable to that of commercial website.Therefore,this paper combines the basic theory of customer equity with the features of commercial website and puts forward customer equity measurement of commercial website on the basis of ARIMA time series model,then further choose Sohu company as a case and calculate its customer equity value. Study results show that this model is not only accurate but highly feasible.

Abstract: This paper presents an analysis of the nonlinear characteristics of a bridge structure and the chaos of BHM(bridge health monitoring) information. Chongqing Masangxi bridge’s BHM information is analyzed by using the max Lyapunov index with Wolf and correlation dimension with G_P algorithm. The results show:1) all of the max Lyapunov index is nonnegative ;2)the correlation dimension is non-integeral and greater than 2.These proves that the bridge structure is in the chaos. Meanwhile, with the evolution of time, the index of chaos is sensitive with status of structure system and varies in different key sections of bridge structure. These findings lay a solid foundation for the further development of bridge safety assessment and prediction when non-linear chaotic theory is utilized to analyze the bridge health monitoring information.

Abstract: Support vector machine (SVM) is gaining popularity on time series analysis due to its advanced theory foundation. The introduction of the hidden information on the basis of SVM is called support vector machine plus (SVM+). However, the hidden information which provides something closely associated with the time series increases the difficulty of training SVM model. In this paper, a new time series regression method GA-RSVM+ is put forward, in which Genetic Algorithm (GA) is used to search the optimal combination of free parameters. The experimental result shows that GA-RSVM+ can accurately determine the parameters on its own and achieve best regression precision. This method has a clear advantage in the regression analysis of time series.

Abstract: To improve the accuracy of load forecasting is the focus of the load forecasting. As the daily load by various environmental factors and periodical, this makes the load time series of changes occurring during non-stationary random process. The key of improving the accuracy of artificial neural network training is to select effective training sample. This paper based on the time series forecasting techniques’ random time series autocorrelation function to select the neural network training samples. The method of modeling is more objective. By example, the comparison with autoregressive (AR) Model predictions and BP Artificial Neural Network (ANN) predicted results through error analysis and confirmed the proposed scheme good performance.