Recurrent Neural Network for Approximate Earthquake Time and Location Prediction Using Multiple Seismicity Indicators

Ashif Panakkat, Hojjat Adeli: Computer Aided Civil and Infrastructure Engineering 24(4): 280-292 (2009)

Abstract:A computational approach is presented for predicting the location and time of occurrence of future moderate-to-large earthquakes in an approximate sense based on neural network modeling and using a vector of eight seismicity indicators as input. Two different methods are explored. In the first method, a large seismic region is subdivided into several small subregions and the temporal historical earthquake record is divided into a number of small equal time periods. Seismicity indicators are computed for each subregion for each time period and their relationship to the magnitude of the largest earthquake occurring in that subregion during the following time-period is studied using a recurrent neural network. In the second more direct approach, the temporal historical earthquake record is divided into a number of unequal time periods where each period is defined as the time between large earthquakes. Seismicity indicators are computed for each time-period and their relationship to the latitude and longitude of the epicentral location, and time of occurrence of the following major earthquake is studied using a recurrent neural network.

Get Full Article

A probabilistic neural network for earthquake magnitude prediction

Hojjat Adeli, Ashif Panakkat: Neural Networks 22(7): 1018-1024 (2009)

Abstract: A probabilistic neural network (PNN) is presented for predicting the magnitude of the largest earthquake in a pre-defined future time period in a seismic region using eight mathematically computed parameters known as seismicity indicators. The indicators considered are the time elapsed during a particular number (n) of significant seismic events before the month in question, the slope of the Gutenberg–Richter inverse power law curve for the n events, the mean square deviation about the regression line based on the Gutenberg–Richter inverse power law for the n events, the average magnitude of the last n events, the difference between the observed maximum magnitude among the last n events and that expected through the Gutenberg–Richter relationship known as the magnitude deficit, the rate of square root of seismic energy released during the n events, the mean time or period between characteristic events, and the coefficient of variation of the mean time. Prediction accuracies of the model are evaluated using three different statistical measures: the probability of detection, the false alarm ratio, and the true skill score or R score. The PNN model is trained and tested using data for the Southern California region. The model yields good prediction accuracies for earthquakes of magnitude between 4.5 and 6.0. The PNN model presented in this paper complements the recurrent neural network model developed by the authors previously, where good results were reported for predicting earthquakes with magnitude greater than 6.0.

Get Full Article

Recent Efforts in Earthquake Prediction (1990–2007)

Ashif Panakkat, Hojjat Adeli: ASCE Natural Hazards Review 9(2): 70-81 (2008)

Abstract: From an emergency-management and hazard-preparedness perspective, long-term seismic risk analysis of a region is of utmost importance. There have been a good number of efforts during the past 15 years or so in earthquake prediction, and this paper is an attempt to present a state-of-the-art review of the subject. The most significant recent efforts in predicting the three earthquake parameters, namely, the time of occurrence, epicentral location, and the magnitude of future earthquakes are reviewed. Prediction studies can be broadly grouped based on the basic approach, which vary from purely theoretical geophysics, to genetic mutations and biology, to statistical, mathematical, and computational modeling of earthquake parameter data recorded in historical catalogs of seismic regions. The papers reviewed in this article are classified into two groups: (1) studies based on recording and analyzing earthquake precursors (seismic monitoring); and (2) studies based on historic earthquake data analysis. The complexity of the earthquake prediction problem notwithstanding, the authors believe that the scientific community should pursue research on the subject vigorously. It is hoped that this article fuels additional interest in the subject.

Get Full Article

Neural Network Models for Earthquake Magnitude Prediction Using Multiple seismicity Indicators.

Ashif Panakkat, Hojjat Adeli: International Journal of Neural Systems 17(1): 13-33 (2007)

Abstract: Neural networks are investigated for predicting the magnitude of the largest seismic event in the following month based on the analysis of eight mathematically computed parameters known as seismicity indicators. The indicators are selected based on the Gutenberg-Richter and characteristic earthquake magnitude distribution and also on the conclusions drawn by recent earthquake prediction studies. Since there is no known established mathematical or even empirical relationship between these indicators and the location and magnitude of a succeeding earthquake in a particular time window, the problem is modeled using three different neural networks: a feed-forward Levenberg-Marquardt backpropagation (LMBP) neural network, a recurrent neural network, and a radial basis function (RBF) neural network. Prediction accuracies of the models are evaluated using four different statistical measures: the probability of detection, the false alarm ratio, the frequency bias, and the true skill score or R score. The models are trained and tested using data for two seismically different regions: Southern California and the San Francisco bay region. Overall the recurrent neural network model yields the best prediction accuracies compared with LMBP and RBF networks. While at the present earthquake prediction cannot be made with a high degree of certainty this research provides a scientific approach for evaluating the short-term seismic hazard potential of a region.

Get Full Article

Modeling Transition to Turbulence in Eccentric Stenotic Flows

Sonu S. Varghese, Steven H. Frankel, Paul F. Fischer: Journal of Biomechanical Engineering 130: 014503-1 (2008)

Abstract: Mean flow predictions obtained from a host of turbulence models were found to be in poor agreement with recent direct numerical simulation results for turbulent flow distal to an idealized eccentric stenosis. Many of the widely used turbulence models, including a large eddy simulation model, were unable to accurately capture the poststenotic transition to turbulence. The results suggest that efforts toward developing more accurate turbulence models for low-Reynolds number, separated transitional flows are necessary before such models can be used confidently under hemodynamic conditions where turbulence may develop.

Get Full Article

Direct Numerical Simulation of Stenotic Flows, Part 1: Steady Flow

Sonu S. Varghese, Steven H. Frankel, Paul F. Fischer: Journal of Fluid Mechanics 582: 253-280 (2007)

Abstract: Direct numerical simulations (DNS) of steady and pulsatile flow through 75% (by area reduction) stenosed tubes have been performed, with the motivation of understanding the biofluid dynamics of actual stenosed arteries. The spectral-element method, providing geometric flexibility and high-order spectral accuracy, was employed for the simulations. The steady flow results are examined here while the pulsatile flow analysis is dealt with in Part 2 of this study. At inlet Reynolds numbers of 500 and 1000, DNS predict a laminar flow field downstream of an axisymmetric stenosis and comparison to previous experiments show good agreement in the immediate post-stenotic region. The introduction of a geometric perturbation within the current model, in the form of a stenosis eccentricity that was 5% of the main vessel diameter at the throat, resulted in breaking of the symmetry of the post-stenotic flow field by causing the jet to deflect towards the side of the eccentricity and, at a high enough Reynolds number of 1000, jet breakdown occurred in the downstream region. The flow transitioned to turbulence about five diameters away from the stenosis, with velocity spectra taking on a broadband nature, acquiring a -5/3 slope that is typical of turbulent flows. Transition was accomplished by the breaking up of streamwise, hairpin vortices into a localized turbulent spot, reminiscent of the turbulent puff observed in pipe flow transition, within which r.m.s. velocity and turbulent energy levels were highest. Turbulent fluctuations and energy levels rapidly decayed beyond this region and flow relaminarized. The acceleration of the fluid through the stenosis resulted in wall shear stress (WSS) magnitudes that exceeded upstream levels by more than a factor of 30 but low WSS levels accompanied the flow separation zones that formed immediately downstream of the stenosis. Transition to turbulence in the case of the eccentric stenosis was found to be manifested as large temporal and spatial gradients of shear stress, with significant axial and circumferential variations in instantaneous WSS.

Get Full Article

Direct Numerical Simulation of Stenotic Flows, Part 2: Pulsatile Flow

Sonu S. Varghese, Steven H. Frankel, Paul F. Fischer: Journal of Fluid Mechanics, 582: 281-318 (2007)

Abstract: Direct numerical simulations (DNS) of stenotic flows under conditions of steady inlet flow were discussed in Part 1 of this study. DNS of pulsatile flow through the 75% stenosed tube (by area) employed for the computations in Part 1 is examined here. Analogous to the steady flow results, DNS predicts a laminar post-stenotic flow field in the case of pulsatile flow through the axisymmetric stenosis model, in contrast to previous experiments, in which intermittent disturbed flow regions and turbulent breakdown were observed in the downstream region. The introduction of a stenosis eccentricity, that was 5% of the main vessel diameter at the throat, resulted in periodic, localized transition to turbulence. Analysis in this study indicates that the early and mid-acceleration phases of the time period cycle were relatively stable, with no turbulent activity in the post-stenotic region. However, towards the end of acceleration, the starting vortex, formed earlier as the fluid accelerated through the stenosis at the beginning of acceleration, started to break up into elongated streamwise structures. These streamwise vortices broke down at peak flow, forming a turbulent spot in the post-stenotic region. In the early part of deceleration there was intense turbulent activity within this spot. Past the mid-deceleration phase, through to minimum flow, the inlet flow lost its momentum and the flow field began to relaminarize. The start of acceleration in the following cycle saw a recurrence of the entire process of a starting structure undergoing turbulent breakdown and subsequent relaminarization of the post-stenotic flow field. Peak wall shear stress (WSS) levels occurred at the stenosis throat, with the rest of the vessel experiencing much lower levels. Turbulent breakdown at peak flow resulted in a sharp amplification of instantaneous WSS magnitudes across the region corresponding to the turbulent spot, accompanied by large axial and circumferential fluctuations, even while ensemble-averaged axial shear stresses remained mostly low and negative. WSS levels dropped rapidly after the mid-deceleration phase, when the relaminarization process took over, and were almost identical to laminar, axisymmetric shear levels through most of the acceleration phase.

Get Full Article

Numerical Modeling of Pulsatile Turbulent Flow in Stenotic Vessels

Sonu S. Varghese, Steven H. Frankel, Paul F. Fischer: Journal of Biomechanical Engineering 125: 445-460 (2003)

Abstract: Pulsatile turbulent flow in stenotic vessels has been numerically modeled using the Reynolds-averaged Navier-Stokes equation approach. The commercially available computational fluid dynamics code (CFD), FLUENT, has been used for these studies. Two different experiments were modeled involving pulsatile flow through axisymmetric stenoses. Four different turbulence models were employed to study their influence on the results. It was found that the low Reynolds number k–omega turbulence model was in much better agreement with previous experimental measurements than both the low and high Reynolds number versions of the RNG (renormalization-group theory) k-epsilon turbulence model and the standard k-epsilon model, with regard to predicting the mean flow distal to the stenosis including aspects of the vortex shedding process and the turbulent flow field. All models predicted a wall shear stress peak at the throat of the stenosis with minimum values observed distal to the stenosis where flow separation occurred.

Get Full Article

Relative Visibility of Internally and Externally illuminated on-premise signs

Philip M. Garvey, Chandrashekar Ramaswamy, Ramy Ghebrial, Miguel De la Riva, and Martin T. Pietrucha

Abstract: The purpose of this study was to evaluate the relative performance of internally and externally illuminated on-premises signs. To do this, the performance of six signs that differed in mode of illumination, text and background colors, and contrast orientation (i.e.,light letters on a darker background and dark letters on a lighter background) was evaluated. These signs were field tested with older and younger motorists in both daytime and night conditions. The two measures of effectiveness were sign recognition distance and legibility distance. Based on the results, the distance at which drivers can begin to read a sign’s message as a function of the type of illumination was calculated. These distances were then converted to time at various approach speeds to determine the amount of time that motorists will have to read the sign content. Results showed that the internally illuminated signs provided significantly longer visibility distances and longer available reading times than externally illuminated signs.

Get Full Article

Emergency Road Flare Effectiveness in enhancing the “Safety Zone

M. de la Riva, P. M. Garvey, M. T. Pietrucha, R. Ghebrial and C. Ramaswamy

Abstract: Motorist behavior (passing speed, lateral separation from a disabled vehicle, and lane distribution) was evaluated under several roadway scenarios using various safety flare configurations. The treatments were installed on a flat and straight segment of a four-lane limited-access divided highway (I-99) with a posted speed limit of 65 mph. Roadway sensors were used to unobtrusively observe passing vehicle behavior. Data were collected for a period of two weeks, one week of baseline data and one week of test data. A total of over 7,000 vehicles passed through the test location during the evaluation. A significant number of vehicles passing the flare configurations moved from the right to the left lane (in the baseline condition 86.2 percent of vehicles were in the right lane whereas in the test conditions this was reduced to only 8.5 percent); vehicles that remained in the right lane moved on average 27 inches further away from the shoulder compared to baseline, and operating speeds were reduced by an average of 9.0 mph (a 15 percent reduction from baseline). In traffic scenarios with police presence and light bar activated, the addition of roadway safety flares produced an extra traffic safety benefit. In traffic scenarios without police presence, the use of roadway flares provided a safety benefit equivalent to having a patrol car present with light bar activated.

Get Full Article