D. Fefer

Time series prediction with neural networks: a case study of two examples

A new approach to optimal presentation of a real world problem to a neural network that predicts chaotic time series is presented in the following paper. The proposed approach takes advantage of two measures of deterministic chaos in time series-the correlation exponent and the mutual information function. Its successfulness is most promising and is demonstrated on two time series.<<ETX>>

A learning algorithm for self-calibration of a voltage calibrator

An algorithm either to extend the calibration period or to reduce the measurement uncertainty of a DC voltage reference module is presented. This module is used either as a transfer, independent, or working standard, or as a reference module incorporated into a larger measuring system. The basic idea is that the deviation history of measured voltage differences of reference elements of a group reference module during the calibration period can be used as a learning period for a neural network. This neural network, when created, can numerically correct particular reference elements later in the working period. Results were obtained by simulation and evaluated on the basis of empirical data and simulated input functions. Hardware solutions to model this algorithm are discussed. >

An adaptive internal calibration of a voltage calibrator

An algorithm for adaptive internal calibration of a precision voltage calibrator is discussed. Due to the built-in metrology organization and real-time data processing of measurement results, improved traceability is obtained. An evaluation of internal calibration results indicates that a learning control algorithm is feasible, thus extending the normal calibration recall cycle. >

Case study of the predictive models used for improvement of the stability of the DC voltage reference source

The aim of this paper is to present a non-typical application of predictive models for voltage correction in a high precision solid-state DC voltage reference source (DCVRS). Several types of neural networks are trained until the correlation dimension and the leading Lyapunov exponent of the predicted signals reach the values of the same invariant measures of the original signals. The predictive models are used as a segment in the software controlled VRE. A control loop is implemented to reduce the sensitivity of the reference source which contributes to enhancement of the robustness of the system and thereby the stability of the reference voltage.

Geomagnetic and strong static magnetic field effects on growth and chlorophyll a fluorescence in Lemna minor.

The geomagnetic field (GMF) varies over Earths surface and changes over time, but it is generally not considered as a factor that could influence plant growth. The effects of reduced and enhanced GMFs and a strong static magnetic field on growth and chlorophyll a (Chl a) fluorescence of Lemna minor plants were investigated under controlled conditions. A standard 7 day test was conducted in extreme geomagnetic environments of 4 µT and 100 µT as well as in a strong static magnetic field environment of 150 mT. Specific growth rates as well as slow and fast Chl a fluorescence kinetics were measured after 7 days incubation. The results, compared to those of controls, showed that the reduced GMF significantly stimulated growth rate of the total frond area in the magnetically treated plants. However, the enhanced GMF pointed towards inhibition of growth rate in exposed plants in comparison to control, but the difference was not statistically significant. This trend was not observed in the case of treatments with strong static magnetic fields. Our measurements suggest that the efficiency of photosystem II is not affected by variations in GMF. In contrast, the strong static magnetic field seems to have the potential to increase initial Chl a fluorescence and energy dissipation in Lemna minor plants.

Case study of the predictive models used for stability improvement of the DC voltage reference source

The aim of this paper is to present a a typical application of predictive models for voltage correction in a high-precision solid-state DC voltage reference source (DCVRS). Several types of neural networks are trained until the invariant measures of dynamics, such as correlation dimension and leading Lyapunov exponent of the predicted signals, reach the values of the same invariant measures of the original signals. The predictive models are used as a segment in the software-controlled voltage reference element (VRE). A control loop is implemented to reduce the interference sensitivity of the reference source which contributes to enhancement of the robustness of the system and thereby the stability of the reference voltage.

Industrial experience with an intelligent AC/DC standard

An intelligent AC/DC voltage standard has been constructed to meet stringent demands for calibration in laboratories as well as in demanding industrial environments. The use of a microcomputer MC enables programmable output parameters, storage of calibration constants, and integration into a complex measuring and testing system. Stability of AC and DC parameters, reliability of operation, and reliability of metrological parameters were achieved by real-time data processing. The prototype exhibited the following characteristics: an output frequency range from 30 to 400 Hz with a resolution of 1 Hz, four voltage ranges from 0.1 to 1000 V with 12-bit resolution, and output power of 50 V.A with limitations of output voltage U/sub max/=1000 V/sub RMS/ and output current I/sub max/=5 A/sub RMS/. >

Evaluation of an alternatively designed digital phase angle standard

A phase reference source, that uses a direct digital frequency synthesizer as a digital sine sample generator was designed and a prototype was built. The paper explains how the experimental verification of the findings that were previously obtained via theoretical analysis and computer simulations was performed on the prototype. The output phase angle deviations are under investigation, especially the circumstances at which they occur. Some preliminary measurement results are presented that confirm the previously gained knowledge regarding the dependencies of the phase modulation intensity.

Deterministic predictive models for DC voltage reference source control

In this paper, we use predictive models for voltage correction in a high-precision DC voltage reference source (DCVRS) based on Zener diodes. Voltage reference elements, which compose the DCVRS, are improved by implementing a control loop with built-in predictors. Thus, the sensitivity of the system is reduced and, thereby, the stability of the DCVRS is improved. The predictive abilities of two different paradigms, neural-network-based predictors and difference equation predictors obtained by equation discovery system LAGRAMGE, are compared.

Comparison of neural networks to statistical techniques for prediction of time series generated by nonlinear dynamic systems

The following paper is focused on comparison of neural networks to statistical techniques for time series prediction. Four statistical models, the ARIMA, the exponential smoothing, the exponential growth and the bilinear model are compared to two neural network architectures, the hierarchical multilayer perceptron and the ontogenic cascade correlation network. The intercomparison was done on two examples, a generic and a real-world one. The results of analyses were most promising from the neural networks point of view