With the development of performance-based earthquake engineering, the risk-informed assessment framework has received broad recognition over the world, of which the probability seismic fragility ...analysis is an important step. The classic seismic fragility adopts the lognormal assumption and forms a parametric derivation. With the development of fragility theory, researchers are hoping to seek out non-parametric approaches to express the intrinsic fragility in a pure analytical form without any distribution assumptions. Besides, how to keep the calculation efficiency (e.g., combining with cloud approach) and how to consider the non-stationary stochastic responses (e.g., combining with non-stationary stochastic excitation model) are critical aspects in fragility that deserve further attention of researchers. In this paper, a kernel density estimation (KDE) based non-parametric cloud approach is proposed for efficient seismic fragility estimation of structures under non-stationary excitation. First, the methodology framework of the efficient approach is illustrated. Then, the procedures of non-stationary stochastic seismic response of structures and KDE-based non-parametric cloud approach for efficient seismic fragility are demonstrated. After that, an application example via a three-span-six-story reinforced concrete frame is given for implementation, followed with a parametric analysis of critical factors. During the process, the classic parametric linear-regression based cloud approach (cloud-LR) and benchmark Monte-Carlo-simulation based cloud approach (cloud-MCS) are also incorporated for validation. In general, the analysis verifies the effectiveness of the non-parametric cloud-KDE approach without requiring more computation work (i.e., same as the parametric cloud-LR approach and much less than the benchmark cloud-MCS approach). Meanwhile, the non-parametric cloud-KDE approach indicates a comparable accuracy with the classic fragility approaches (i.e., less deviation than the parametric cloud-LR approach and much closer to the benchmark cloud-MCS approach), and with the increase of stochastic cloud-point number, the corresponding fitting degree of cloud-KDE approach is growing better. The research provides a new sight for the development of non-parametric seismic fragility approach, and the corresponding findings can be further combined with the probabilistic hazard and risk analysis for a non-parametric assessment procedure in performance-based earthquake engineering.
•Develop a kernel-density-based non-parametric methodology for seismic fragility estimation.•Combine the cloud approach and the non-stationary excitation for both efficiency and accuracy.•Incorporate the parametric cloud-LR and the benchmark cloud-MCS approaches for validation.•Provide a new sight for the development of non-parametric seismic fragility estimation.
Wildlife-vehicle collisions (WVCs) pose a serious global issue. Factors influencing the occurrence of WVC along roads can be divided in general into two groups: spatially random and non-random. The ...latter group consists of local factors which act at specific places, whereas the former group consists of globally acting factors. We analyzed 27,142 WVC records (roe deer and wild boar), which took place between 2012 and 2016 on Czech roads. Statistically significant clusters of WVCs occurrence were identified using the clustering (KDE+) approach. Local factors were consequently measured for the 75 most important clusters as cases and the same number of single WVCs outside clusters as controls, and identified by the use of odds ratio, Bayesian inference and logistic regression. Subsequently, a simulation study randomly distributing WVC in clusters into case and control groups was performed to highlight the importance of the clustering approach. All statistically significant clusters with roe deer (wild boar) contained 34% (27%) of all records related to this species. The overall length of the respective clusters covered 0.982% (0.177%) of the analyzed road network. The results suggest that the most pronounced signal identifying the statistically significant local factors is achieved when WVCs were divided according to their occurrence in clusters and outside clusters. We conclude that application of a clustering approach should precede regression modeling in order to reliably identify the local factors influencing spatially non-random occurrence of WVCs along the transportation infrastructure.
•Factors influencing WVCs can be divided in into two groups: spatially random and non-random.•Local (spatially non-random) factors were measured for the 75 clusters and 75 controls.•WVCs divided according to cluster occurrence returned the best signal for local factors.•Coldspots should not be used as controls and single WVCs as cases.•Application of the KDE+ approach to WVC data should precede regression modeling.
In archaeological mobility studies, non-local humans and animals can be identified by means of stable strontium isotope analysis. However, defining the range of local 87Sr/86Sr ratios is ...prerequisite. To achieve this goal, proxy-based mixing models have recently been proposed using 87Sr/86Sr ratios measured in modern local vegetation, water and soil samples. Our study complements earlier efforts by introducing archaeological animal bones as an additional proxy. We then evaluate the different modelling approaches by contrasting proxy-results generated for the county of Erding (Upper Bavaria, Germany) with a comprehensive set of strontium measurements obtained from tooth enamel of late antique and early medieval human individuals (n = 49) from the same micro-region. We conclude that current mixing models based on environmental proxies clearly underestimate the locally bioavailable 87Sr/86Sr ratios due to the limited sample size of modern environmental specimens and a suit of imponderables inherent to efforts modelling complex geobiological processes. In sum, currently available mixing models are deemed inadequate and can therefore not be recommended.
Display omitted
•An improved 87Sr/86Sr mixing model for humans has been developed.•Mixing models calculated with limited sets of proxies are usually unreliable.•Mixing model analyses depend strongly on the component sampling strategy.•Caution is required when using modern samples for past mobility studies.•Kernel density analyses are helpful for determining local 87Sr/86Sr ratio ranges.
•Daily rainfall and evaporation is downscaled using MMM-KDE and k-nearest neighbor approach.•Climate change impact on hydrology is studied using ArcSWAT and an ensemble of GCMs.•Future projections at ...a catchment level shows an increase in the water stress.•Changes in the average flows are not significant, but low flows are of concern.•Increase in irrigation demand and reduction in groundwater recharge are projected.
This work evaluates the impact of climate change on the water balance of a catchment in India. Rainfall and hydro-meteorological variables for current (20C3M scenario, 1981–2000) and two future time periods: mid of the 21st century (2046–2065) and end of the century (2081–2100) are simulated using Modified Markov Model-Kernel Density Estimation (MMM-KDE) and k-nearest neighbor downscaling models. Climate projections from an ensemble of 5 GCMs (MPI-ECHAM5, BCCR-BCM2.0, CSIRO-mk3.5, IPSL-CM4, and MRI-CGCM2) are used in this study. Hydrologic simulations for the current as well as future climate scenarios are carried out using Soil and Water Assessment Tool (SWAT) integrated with ArcGIS (ArcSWAT v.2009). The results show marginal reduction in runoff ratio, annual streamflow and groundwater recharge towards the end of the century. Increased temperature and evapotranspiration project an increase in the irrigation demand towards the end of the century. Rainfall projections for the future shows marginal increase in the annual average rainfall. Short and moderate wet spells are projected to decrease, whereas short and moderate dry spells are projected to increase in the future. Projected reduction in streamflow and groundwater recharge along with the increase in irrigation demand is likely to aggravate the water stress in the region under the future scenario.
Highly Improving the Accuracy of Clustering (HIAC) algorithm is designed to enhance clustering accuracy by introducing a gravitational force between data objects, drawing them closer together, and ...employing a decision graph to establish a weight threshold for differentiating neighbor classes and outliers. Despite its strengths, HIAC faces two shortcomings: (1) its inability to generate effective decision graphs for small-scale datasets and (2) the non-smooth probability curve within the decision graph, making threshold determination by visual inspection both difficult and imprecise. This study presents an improved adaptive algorithm based on Kernel Density Estimation (KDE-AHIAC). This approach automatically selects the bandwidth based on the density and distribution of the data, utilizing the kernel density function to create a decision graph that applies to any dataset. For threshold selection, we introduce an adaptive calculation method that leverages the smoothness and continuity of the kernel density curve, replacing the observational approach. Additionally, we incorporate an outlier test model using Analysis of Similarity (ANOSIM) to avert misclassification of valid samples as outliers. Through comprehensive experimentation, we tested KDE-AHIAC and found that it offers notable improvements over HIAC. KDE-AHIAC enhances the clustering accuracy of the dataset by 66.05% compared to the original data and by 6.22% over HIAC.
•The real-time fracture process of fissured red sandstone specimen is monitored.•Two indices of RA and AF are performed to classify the different fracture modes.•The adequacy of the fracture mode ...classification is evaluated by Kernel Density Estimation (KDE).•KDE can identify and visualise the high concentration regions of RA and AF values.•RA values can serve as an early warning for ultimate failure of rocks.
To study the classification of fracture modes of rocks during the cracking process, uniaxial compression tests were conducted on intact and flawed red sandstone specimens. Meanwhile, both acoustic emission (AE) and digital image correlation (DIC) technologies were adopted to monitor and record the real-time cracking process of the specimens tested. In this study, the interevent time function F(τ) (AE events rate) was utilized to distinguish the transition from microcracking to macrocracking for the tested specimens. An AE parameter analysis method based on two indices of RA (rise time/amplitude) and AF (AE counts/duration) values was performed to classify the different cracking modes during the loading process. The classification results of the cracking modes coincide with the cracking type of macrocracks captured by the photographic system. Moreover, the adequacy of the crack classification was also evaluated by the kernel density estimation (KDE) function, a nonparametric density estimation method. KDE is used as a parametric model to overcome the randomness found in the data set generated by AE testing and can well identify and visualize the high concentration regions of RA and AF values. A continuous increase of RA values (more than 400 ms/v) can serve as an early warning for the ultimate failure of red sandstone. The results of this present investigation can be applied in the health monitoring of rock engineering.
Display omitted
•A method is developed for day-ahead 96-step probabilistic wind speed forecasts.•This work analyzes error characteristic of Numerical Weather Prediction results.•Error correction ...models are established based on structured neural networks.•Mixture kernel density estimation is used for joint probabilistic information.•Case studies of wind farm in China certify the effectiveness of proposed method.
At present, wind forecast based on Numerical Weather Prediction is widely recognized and applied for a safer and more sufficient usage of wind sources. However, because of the unescapable inherent errors of numerical techniques, there are many negative cases of forecasts. Thus, aiming to quantize and evaluate the inherent errors of physical outcomes, this paper analyzes the characteristic of residuals between numerical results and actual measured data in statistical way, designs combined non-linear and non-parameter algorithms to correct original prediction values, and achieves probabilistic one-day-ahead 96-step wind speed forecasts. The concise process of the method can be described as followings. Firstly, this work utilizes autocorrelation analysis to verify the non-noise attribute of error sequences. Based on the characteristic, adaptive and structured error correction models of nonlinear autoregressive with exogenous inputs network are established to acquire deterministic optimized outcomes. Then, aiming to calculate conditional error boundaries of different confidence levels, mixture kernel density estimation is adopted step by step to estimate joint probability density of corrected values and revised errors. The results on test set show the correction considering inherent errors of numerical techniques can integrate the physical with statistical information effectively and enhance the forecast accuracy indeed.
Fault diagnosis for rolling elements in rotating machinery persistently receives high research interest due to the said machinery's prevalence in a broad range of applications. State-of-the-art ...methods in such setups focus on effective identification of faults that usually involve a single component while rejecting noise from limited sources. This article studies the data-based diagnosis of mixed faults coming from multiple components with an emphasis on model robustness against a wide spectrum of external perturbation. A dataset is collected on a rotor and bearing system by varying the levels and types of faults in both the rotor and bearing, which results in 48 machine health conditions. A duplet classifier is developed by combining two 1-D convolutional neural networks (CNNs) that are responsible for the diagnosis of the rotor and bearing faults, respectively. Experimental results show that the proposed classifier can reliably identify the onset and nature of mixed faults. In addition, one-vs-all classifiers are built using the features generated by the developed 1-D CNNs as predictors to recognize previously unlearned fault types. The effectiveness of such classifiers is demonstrated using data collected from four new fault types. Finally, the robustness and ability to reject external perturbation of the duplet classification model are analyzed using kernel density estimation. The code for the proposed classifiers is available at https://github.com/siyuanc2/machine-fault-diag .
Numerous facets of scientific research implicitly or explicitly call for the estimation of probability densities. Histograms and kernel density estimates (KDEs) are two commonly used techniques for ...estimating such information, with the KDE generally providing a higher fidelity representation of the probability density function (PDF). Both methods require specification of either a bin width or a kernel bandwidth. While techniques exist for choosing the kernel bandwidth optimally and objectively, they are computationally intensive, since they require repeated calculation of the KDE. A solution for objectively and optimally choosing both the kernel shape and width has recently been developed by Bernacchia and Pigolotti (2011). While this solution theoretically applies to multidimensional KDEs, it has not been clear how to practically do so.
A method for practically extending the Bernacchia–Pigolotti KDE to multidimensions is introduced. This multidimensional extension is combined with a recently-developed computational improvement to their method that makes it computationally efficient: a 2D KDE on 105 samples only takes 1 s on a modern workstation. This fast and objective KDE method, called the fastKDE method, retains the excellent statistical convergence properties that have been demonstrated for univariate samples. The fastKDE method exhibits statistical accuracy that is comparable to state-of-the-science KDE methods publicly available in R, and it produces kernel density estimates several orders of magnitude faster. The fastKDE method does an excellent job of encoding covariance information for bivariate samples. This property allows for direct calculation of conditional PDFs with fastKDE. It is demonstrated how this capability might be leveraged for detecting non-trivial relationships between quantities in physical systems, such as transitional behavior.
•A multidimensional, fast, and robust kernel density estimation is proposed: fastKDE.•fastKDE has statistical performance comparable to state-of-the-science kernel density estimate packages in R.•fastKDE is demonstrably orders of magnitude faster than comparable, state-of-the-science density estimate packages in R.•A Python-based implementation of fastKDE is available at https://bitbucket.org/lbl-cascade/fastkde.