Time Window Technique Research Articles

Intelligent detection of underground openings and surrounding disturbed zones

This study aims to develop an intelligent method to detect subsurface anomalies for prediction and mitigation of geologic risks. We trained a convolutional neural network (CNN) model using the seismic data from 6 field cases of regular pipes and irregular cavities and a sliding time window technique to augment the datasets, and tested the CNN model using another 2 field cases. We derived probability distribution as an indicator of anomaly existence and validated high-value and low-value probability distributions corresponding to underground openings and surrounding disturbed zones, respectively. We found that the characteristics of subsurface anomalies determines the effective seismic features and influences the CNN model performance. Finally, we applied the CNN model to investigate a circular-bored tunnel and surrounding disturbed zones. Our results demonstrate that the CNN model is fast and accurate to detect the horizontal locations of subsurface anomalies, while the accuracy of vertical locations depends on the estimation of P-wave velocity. The intelligent method has the potential to identify hidden risks at early stages and to mitigate subsequent geologic hazards.

A combined model based on POA-VMD secondary decomposition and LSTM for ultra-short-term wind power forecasting

As the application of wind power expands, precise prediction of wind energy becomes essential for the effective plan and reliable functioning in the realm of the power system. Aiming to enhance wind power utilization efficiency and minimize error relating to ultra-short-term wind power forecasting, a novel model grounded in sliding time window, Pelican optimization algorithm-variational mode decomposition (POA-VMD) secondary decomposition, sample entropy calculation, sequence reconstruction, and long short-term memory (LSTM) prediction is introduced in this paper. First, in the training set, the sliding time window technique is employed to identify the optimal parameters for the forecasting algorithm, aiming to closely replicate the actual forecasting performance. Subsequently, the VMD algorithm is enhanced through optimization with the POA. This involves utilizing POA to dynamically ascertain the optimal parameters [k, α] for VMD, allowing for an adaptive decomposition of the raw wind power data sequence and effectively diminishing data noise. After calculating each modal's sample entropy, the modal with the highest sample entropy is further decomposed using POA-VMD. The decomposed sequence is predicted using LSTM to get the final prediction. The experiment ultimately demonstrated that the introduced model markedly improves the accuracy of forecasting. By adding POA-VMD secondary decomposition residuals, the prediction errors, as measured by mean absolute error, root mean square error, and mean absolute percentage error, are decreased by 52.03%, 30.34%, and 39.87%, respectively, and coefficient of determination (R2) is increased by 7.75%.

An integrated model for crude oil forecasting: Causality assessment and technical efficiency

In light of the central role of crude oil in the economy and the complex mechanisms involved in forecasting crude oil prices, this study proposes a two-stage model that optimally selects driving predictors for crude oil price forecasting by integrating Granger causality test (GCT) and stochastic frontier analysis (SFA). In the first stage, GCT is used to perform causality assessments for 92 predictors across eight categories of factors (demand, supply, inventory, financial market, macroeconomy, economic policy uncertainty, geopolitical risk, and technical indicator). In the second stage, SFA is employed to assess the forecasting power of the preliminarily selected predictors in terms of technical efficiency by using multiple evaluation measures. By collecting a data sample which spans a 21-year period from January 1, 2000 to December 31, 2020, we conduct a comprehensive empirical study by employing rolling time window technique. The empirical results demonstrate that the two-stage model significantly outperforms eight competing models in terms of four forecasting techniques (linear regression, artificial neural network, support vector regression, and random forest). The proposed model's outperformance is robust to different time windows, different forecast horizons, alternative proxies of crude oil prices, and different business conditions. We also explore the time-varying characteristics of predictors for crude oil price forecasting and confirm that financial factors remain vital determinants affecting oil prices.

New Optimized Time Window Technique Based MFCC Feature Extraction for Text-Independent Speaker Verification

New Optimized Time Window Technique Based MFCC Feature Extraction for Text-Independent Speaker Verification

Fast wireless sensor for anomaly detection based on data stream in an edge-computing-enabled smart greenhouse

Edge-computing-enabled smart greenhouses are a representative application of the Internet of Things (IoT) technology, which can monitor the environmental information in real-time and employ the information to contribute to intelligent decision-making. In the process, anomaly detection for wireless sensor data plays an important role. However, the traditional anomaly detection algorithms originally designed for anomaly detection in static data do not properly consider the inherent characteristics of the data stream produced by wireless sensors such as infiniteness, correlations, and concept drift, which may pose a considerable challenge to anomaly detection based on data stream and lead to low detection accuracy and efficiency. First, the data stream is usually generated quickly, which means that the data stream is infinite and enormous. Hence, any traditional off-line anomaly detection algorithm that attempts to store the whole dataset or to scan the dataset multiple times for anomaly detection will run out of memory space. Second, there exist correlations among different data streams, and traditional algorithms hardly consider these correlations. Third, the underlying data generation process or distribution may change over time. Thus, traditional anomaly detection algorithms with no model update will lose their effects. Considering these issues, a novel method (called DLSHiForest) based on Locality-Sensitive Hashing and the time window technique is proposed to solve these problems while achieving accurate and efficient detection. Comprehensive experiments are executed using a real-world agricultural greenhouse dataset to demonstrate the feasibility of our approach. Experimental results show that our proposal is practical for addressing the challenges of traditional anomaly detection while ensuring accuracy and efficiency.

Data-Driven Deep Learning-Based Attention Mechanism for Remaining Useful Life Prediction: Case Study Application to Turbofan Engine Analysis

Accurately predicting the remaining useful life (RUL) of the turbofan engine is of great significance for improving the reliability and safety of the engine system. Due to the high dimension and complex features of sensor data in RUL prediction, this paper proposes four data-driven prognostic models based on deep neural networks (DNNs) with an attention mechanism. To improve DNN feature extraction, data are prepared using a sliding time window technique. The raw data collected after normalizing is simply fed into the suggested network, requiring no prior knowledge of prognostics or signal processing and simplifying the proposed method’s applicability. In order to verify the RUL prediction ability of the proposed DNN techniques, the C-MAPSS benchmark dataset of the turbofan engine system is validated. The experimental results showed that the developed long short-term memory (LSTM) model with attention mechanism achieved accurate RUL prediction in both scenarios with a high degree of robustness and generalization ability. Furthermore, the proposed model performance outperforms several state-of-the-art prognosis methods, where the LSTM-based model with attention mechanism achieved an RMSE of 12.87 and 11.23 for FD002 and FD003 subset of data, respectively.

Dynamic financial distress prediction based on class-imbalanced data batches

This study proposes two approaches for dynamic financial distress prediction (FDP) based on class-imbalanced data batches by considering both concept drift and class imbalance. One is based on sliding time window and synthetic minority over-sampling technique (SMOTE) and the other is based on sliding time window and majority class partition. Support vector machine, multiple discriminant analysis (MDA) and logistic regression are used as base classifiers in the experiments on a real-world dataset. The results indicate that the two approaches perform better than the pure dynamic FDP (DFDP) models without class imbalance processing and the static FDP models either with or without class imbalance processing.

Neurothrombectomy for Acute Ischemic Stroke Across Clinical Trial Design and Technique: A Single Center Pooled Analysis.

Introduction: The practice of endovascular therapy has evolved dramatically over the last 10 years with randomized clinical trials investigating the benefit of thrombectomy in select patient populations based on time of presentation, imaging criteria, and procedural technique. We sought to understand the benefit of thrombectomy in patients treated within the context of a clinical trial at a single academic center.Methods: Patient-level data recorded in case forms and core-lab adjudicated data were analyzed from patients enrolled in RCTs investigating the benefit of endovascular thrombectomy over medical management (IMSIII, MR RESCUE, ESCAPE, SWIFT PRIME, and DAWN) between 2007 and 2017 at a single academic referral center.Results: A total of 134 patients (intervention group, n = 81; medical group, n = 53) were identified across five clinical trials (IMSIII, n = 46; MR RESCUE, n = 4; ESCAPE, n = 24; SWIFT PRIME, n = 14; DAWN, n = 46). There were no significant differences between the treatment arm and control arm in terms of age, gender, baseline NIHSS, ASPECTS, and site of occlusion. Rates of good outcome were superior in the intervention group with early neurological recovery (NIHSS of 0–1 or increase NIHSS of 8 points at 24 h) at a higher rate of 49% vs. 17% (p = <0.001) and higher rates of functional independence (90 day mRS 0–2 of 53% vs. 26%, p = 0.002). In multivariate logistic regression analysis, lower NIHSS and younger age were predictors of good outcome. There were comparable rates of good outcome irrespective of clinical trial, imaging selection criteria (CTP vs. MRI), early vs. late time window (0–6 h vs. 6–24 h) and procedural technique (Merci vs. Solitaire/Trevo device). There were no differences in rates of sICH, PH-2 or mortality in the intervention group vs. medical group.Conclusions: At a large academic center, the benefit of endovascular therapy over medical therapy is observed irrespective of clinical trial design, patient selection or procedural technique.

3-D joint geodetic and strong-motion finite fault inversion of the 2008 May 12, Wenchuan, China Earthquake

SUMMARYWe present a source inversion of the 2008 Wenchuan, China earthquake, using strong-motion waveforms and geodetic offsets together with 3-D synthetic ground motions. We applied the linear multiple time window technique considering geodetic and dynamic Green's functions computed with the finite-element method and the reciprocity and Strain Green's Tensor formalism. All ground motion estimates, valid up to 1 Hz, accounted for 3-D effects, including the topography and the geometry of the Beichuan and Pengguan faults. Our joint inversion has a higher moment (M0) than a purely geodetic inversion and the slip distribution presents differences when compared to 1-D model source inversions. The moment is estimated to be M0 = 1.2 × 1021 N·m, slightly larger than other works. Our results show that considering a complex 3-D structure reduces the size of large areas of 10 m slip or greater by distributing it in wider zones, with reduced slips, in the central portion of the Beichuan and the Pengguan faults. Finally, we compare our source with a relocated aftershock catalogue and conclude that the 4–5 m slip contours approximately bound the absence or presence of aftershocks.

Design of de-noising FEM-FIR filters for the evaluation of temporal and spatial derivatives of measured displacement in elastic solids

Abstract This paper presents an FEM-FIR filter de-noising measured displacement to estimate temporal and spatial derivatives of displacement in an elastic solid. The filter is formulated by two independent inverse problems, which are defined as the minimization of the difference between measured and de-noised displacement in the temporal and spatial domains. The regularization functions are introduced in the minimization problems to eliminate noise in measured displacement. The L2-norm of the acceleration and the strain energy of an elastic solid are utilized as the regularization functions in the temporal and spatial domains, respectively, to impose the integrability of the temporal and spatial derivatives. The moving time-window technique is adopted in the temporal filter, and a usual finite element procedure is applied to discretize the minimization problems. The temporal and spatial filters are combined together to form a single, unified filter, which not only de-noises measured displacement in both domains but also reconstructs the velocity field. The acceleration field is obtained by the 1st-order central finite difference of the reconstructed velocity field. The strain field is evaluated at the Gauss points of each finite element using the de-noised nodal displacement. The regularization factors, which control the degree of de-noising strength, are determined by the desired accuracy of the filters at the target frequency. It is shown through a numerical simulation study that the proposed filters are capable of de-noising measured displacement effectively and yielding accurate temporal and spatial derivatives.

CC_TRS: Continuous Clustering of Trajectory Stream Data Based on Micro Cluster Life

The rapid spreading of positioning devices leads to the generation of massive spatiotemporal trajectories data. In some scenarios, spatiotemporal data are received in stream manner. Clustering of stream data is beneficial for different applications such as traffic management and weather forecasting. In this article, an algorithm for Continuous Clustering of Trajectory Stream Data Based on Micro Cluster Life is proposed. The algorithm consists of two phases. There is the online phase where temporal micro clusters are used to store summarized spatiotemporal information for each group of similar segments. The clustering task in online phase is based on temporal micro cluster lifetime instead of time window technique which divides stream data into time bins and clusters each bin separately. For offline phase, a density based clustering approach is used to generate macro clusters depending on temporal micro clusters. The evaluation of the proposed algorithm on real data sets shows the efficiency and the effectiveness of the proposed algorithm and proved it is efficient alternative to time window technique.

Empirical modeless decomposition and modal identification method for nonstationary vibration processes

An empirical data decomposition method is developed in this study to accurately identify modeless component and intrinsic vibration modal terms separately from vibration processes, which are characterized by nonlinearity and nonstationarity. Two key innovations related to this method are proposed. One of them is associated with the definition of root mean square instantaneous amplitude, which is adopted as a meaningful supplement for vibration mode identification. It also acts as a critical criterion to visualize the nonstationarity of some vibration processes. The other innovation is the use of a predetermined maximum intrinsic time window and cubic spline interpolation technique for the empirical data decomposition method with the aim to decompose a modeless component from nonstationary processes. Unlike in other decomposition methods, a modeless component should be decomposed prior to modal analysis. The requirement of using spurious harmonics to represent nonlinear and nonstationary signals is fully eliminated via these two key innovations, thereby increasing the accuracy of the subsequent modal identification. Given the well-designed decomposition method, energy aliasing in data processing can also be effectively avoided. Significant conclusions about the complicated vibration processes composed of nonstationary and intrinsic modal terms are drawn by using the proposed empirical modeless decomposition and modal identification method.

Detection of occupancy profile based on carbon dioxide concentration pattern matching

Abstract Occupancy is one of major factors influencing indoor microclimate. The aim of this work was to determine the influence of this factor on indoor air quality (IAQ) on the basis of CO 2 concentration measurements and statistical analysis. We wanted to identify periods of time when IAQ was strongly affected by the occupancy described by the given profile. The proposed approach consisted of several stages. The CO 2 concentration was measured and recorded in the form of univariate time series. Then, the relationship between occupancy and internal structure of the time series was disclosed. There were distinguished segments based on sample periodogram. Each segment was associated with a particular occupancy profile. In order to detect how human factor represented by a given occupancy profile influences IAQ we proposed to use pattern matching. In this approach there was examined the similarity between segments of the time series and the pattern of CO 2 variability, which represented a selected occupancy profile. The analysis was performed in time domain using moving time window technique. The similarity was judged based on two types of indexes, namely correlation coefficients and distance measures. It was shown that our approach may be applied to successfully detect a particular occupancy profile. The best performance was achieved when using angular distance as the similarity index. In this case we reached 82% true positive and 22% false positive detections. The proposed method may be applied in diagnostics problems to reveal sources of indoor air quality problems.

Extreme learning machine for time sequence classification

In this paper, a new framework to effectively classify the time sequence is developed. The whole time sequence is divided into several smaller sub-sequence by means of the sliding time window technique. The sub-sequence is modeled as a linear dynamic model by appropriate dimension reduction and the whole time sequence is represented as a bag-of-systems model. Such a model is very flexible to describe time sequence originated from different sensor source. To construct the bag-of-systems model, we design the codebook by using the K-medoids clustering algorithm and Martin distance between linear dynamic systems. Such a technology avoids the problem that linear dynamic systems lie in non-Euclidean manifold. After obtaining the represented of time sequence, an extreme learning machine is utilized for classification. Finally, the proposed method is verified on some benchmark and shows that it obtains promising results.

A Fuzzy Sequential Pattern Mining Algorithm Based on Independent Pruning Strategy for Parameters Optimization of Ball Mill Pulverizing System

This paper presents a fuzzy sequential pattern mining algorithm based on independent pruning strategy for parameters optimization of ball mill pulverizing system. Based on the Apriori-alike process, the proposed algorithm uses the independent pruning strategy to mine the fuzzy sequential patterns, which could enhance the efficiency of the algorithm. Then, the optimal values of the process variables are determined by a searching method with the mined sequential patterns. The improved fuzzy sequential pattern support and the fuzzy sequential pattern confidence are adopted to ensure the accuracy of the mined sequential patterns. Moreover, the sliding time window technique is used to ensure the completeness of mining results. The experimental results for parameters optimization of ball mill pulverizing system also verify that the proposed algorithm could determine the optimal values correctly and the running time is not long. In addition, the proposed algorithm has been put into practice successfully and the statistic data show that the pulverizing capability of ball mill pulverizing system is increased and the energy consumption would be reduced. DOI: http://dx.doi.org/10.5755/j01.itc.43.3.5180

Analyzing the time variance of orchestral instrument directivities

Thirteen-channel anechoic recordings of 20-s musical excerpts for violin, flute, and bass trombone were analyzed to determine how significantly an instrument’s directivity changes as it is played. A clearer understanding of the time variance of instrument directivities may help in the advancement of room acoustical computer modeling, as the source properties are crucial to the simulation’s accuracy. Previous research has accurately documented the static spatial characteristics of instrument directivities. A multi-channel auralization technique incorporating time-varying source behavior has been developed, but the time variance of the directivities has yet to be explored. In this project, a time-windowing technique was utilized to calculate the directivity index throughout each channel recording for each instrument. The results were analyzed in terms of maximum directivity index and a sampling of complete directivity patterns, and finally used to explore quantification methods. The time variations displayed by each instrument’s directivity were unique in terms of magnitude, direction, and frequency. Calculating the average change in directivity index for each channel at each frequency band was found to be a suitable method for summarizing the results. [Work supported by a UNL Undergraduate Creative Activities and Research Experience Grant.]

Continuous-cyclic variations in the b-value of the earthquake frequency-magnitude distribution

Seismicity of the Earth (M ≥ 4.5) was compiled from NEIC, IRIS and ISC catalogues and used to compute b-value based on various time windows. It is found that continuous cyclic b-variations occur on both long and short time scales, the latter being of much higher value and sometimes in excess of 0.7 of the absolute b-value. These variations occur not only yearly or monthly, but also daily. Before the occurrence of large earthquakes, b-values start increasing with variable gradients that are affected by foreshocks. In some cases, the gradient is reduced to zero or to a negative value a few days before the earthquake occurrence. In general, calculated b-values attain maxima 1 day before large earthquakes and minima soon after their occurrence. Both linear regression and maximum likelihood methods give correlatable, but variable results. It is found that an expanding time window technique from a fixed starting point is more effective in the study of b-variations. The calculated b-variations for the whole Earth, its hemispheres, quadrants and the epicentral regions of some large earthquakes are of both local and regional character, which may indicate that in such cases, the geodynamic processes acting within a certain region have a much regional effect within the Earth. The b-variations have long been known to vary with a number of local and regional factors including tectonic stresses. The results reported here indicate that geotectonic stress remains the most significant factor that controls b-variations. It is found that for earthquakes with Mw ≥ 7, an increase of about 0.20 in the b-value implies a stress increase that will result in an earthquake with a magnitude one unit higher.

Performance reliability estimation method based on adaptive failure threshold

In the process of performance reliability estimation for individual equipments, due to the lack of a large amount of empirical information and experimental data, failure thresholds and degradation models are difficult to be determined directly. To solve this problem, a novel performance degradation reliability based on an adaptive failure threshold is proposed. First, a pattern discrimination model of degradation failure, which is combined with the one-class SVM path solution algorithm, is developed to obtain nonlinear failure threshold at any time. Second, we adopt the sliding time-window technique to extract statistical samples from degradation data series respectively and establish a dynamic kernel estimation model to continuously estimate the conditional probability density function of these samples. The probability distribution which exceeds the adaptive failure threshold is regarded as a reliability indicator. We successfully apply our method to evaluate the performance reliabilities of the bearing and the high-pressure descaling pump. Results show that the method is available to adaptively yield the failure threshold and estimate reliability for individual equipments without empirical information. Moreover, the method can dynamically adjust probability density function to meet the different statistical sample to overcome the limitation of parameter distribution model.

Analysis and Simulation of Strong Earthquake Ground Motions Using ARMA Models

The acceleration record of an earthquake ground motion is a nonstationary process with both amplitude and frequency content varying in time. The paper presents a general procedure for the analysis and simulation of strong earthquake ground motions based on parametric ARMA models. Structural design spectra are based on smoothed linear response spectra obtained from different events scaled by their peak values. Such an approach does not incorporate other characteristics of the excitation represented by measured data. This study investigate the use of non-stationary models which can be considered characteristic and representative of specific historical earthquakes. An earthquake record is regarded as a sample realization from a population of such samples, which could have been generated by the stochastic process characterized by an Autoregressive Moving Average (ARMA) model. This model is capable of reproducing the nonstationary amplitude as well as the frequency content of the earthquake ground accelerations. The moving time-window technique is applied to synthesize the near field earthquakes, Boumerdes-1, Boumerdes -2, and Boumerdes -3 2003 recorded on dense soils in Algeria. This model, is based on a low-order, time-invariant ARMA process excited by Gaussian white noise and amplitude modulated using a simple envelope function to account for the non-stationary characteristics. This simple model gives a reasonable fit to the observed ground motion. It is shown that the selected ARMA (2,1) model and the algorithm used for generating the accelerograms are able to preserve the features of the real earthquake records with different frequency content. In this evaluation, the linear and non linear responses of a given soil layer have been adopted. This study suggests the ability to characterize the earthquake by a minimum number of parameters.

BUILDING USER TASK SPACE BY MINING TEMPORALLY PROXIMATE DESKTOP ACTIONS

Awareness of the user's current context, that is, the task the user resides in, can allow for various ways of supporting their work. In this article we present a novel approach for mining user's regular tasks on the basis of temporal proximity of users’ desktop actions without user intervention in a completely automatic, unobtrusive, and unsupervised manner. The proposed method is based on time-based clustering of a user's desktop actions and, in contrast to previous approaches, it does not enforce fixed time constraints on the mined actions, and typical transition times, when they exist, emerge from the logged activity data for that specific user. The performance of our technique was evaluated on a large data set of 724 days of desktop work by five knowledge workers. The results showed that our approach performed very well and was able to cluster those temporally proximate user actions that commonly used fix time window techniques fail to deal with. Notably, the main improvement is in the recall of the tasks, where our approach recalled 10.18% more actions than the predefined time window technique.