Real-time neural network earthquake profile predictor
AssigneeThe Regents of the University of California
InventorsRichard R. Leach, Farid U. Dowla
Life of this patent
Estimating the term…
Abstract
A neural network has been developed that uses first-arrival energy to predict the characteristics of impending earthquake seismograph signals. The propagation of ground motion energy through the earth is a highly nonlinear function. This is due to different forms of ground motion as well as to changes in the elastic properties of the media throughout the propagation path. The neural network is trained using seismogram data from earthquakes. Presented with a previously unseen earthquake, the neural network produces a profile of the complete earthquake signal using data from the first seconds of the signal. This offers a significant advance in the real-time monitoring, warning, and subsequent hazard minimization of catastrophic ground motion.
Claims27
A method for predicting an earthquake profile, comprising:
- training an artificial neural network (ANN), comprising:
- inputting at least one seismogram into the ANN to produce an output;
- calculating an error between the ANN output and the seismogram;
- adjusting the internal weights of the ANN to reduce the error in the calculating step;
- repeating the preceding steps until the error is reduced to a level where the ANN can generalize sufficiently to produce an accurate output for a previously unseen input;
- detecting an earthquake with three mutually orthogonal ground motion detectors which will produce real-time earthquake data;
- transmitting into the ANN the real-time earthquake data from the three mutually orthogonal ground motion detectors;
- and
- feedforwarding the real-time earthquake data in the ANN to produce a predicted earthquake profile.
An apparatus for predicting an earthquake profile, comprising:
- at least one ground motion detector, comprising three mutually orthogonal components, wherein each component is identical and is selected from a group consisting of seismometers, accelerometers and geophones, said detector being battery powered and producing an output signal;
- an amplifier to receive said output signal and produce a three channel amplified analog signal;
- an analog-to-digital (A-D) converter board comprising at least three channels electrically configured to receive said three channel amplified analog signal;
- a computer to house said A-D converter board, said computer comprising an event detection algorithm, wherein an incoming seismic signal is preprocessed to produce a seismogram having time series and frequency information;
- and
- an artificial neural network which feeds forward said seismogram to produce a complete earthquake profile.
Dependencies are inferred from the claim text.