Ersin Tulunay

Forecasting of ionospheric critical frequency using neural networks

Multilayer perception type neural networks (NN) are employed for forecasting ionospheric critical frequency (foF2) one hour in advance. The nonlinear black-box modeling approach in system identification is used. The main contributions: 1. A flexible and easily accessible training database capable of handling extensive physical data is prepared, 2. Novel NN design and experimentation software is developed, 3. A training strategy is adopted in order to significantly enhance the generalization or extrapolation ability of NNs, 4. A method is developed for determining the relative significances (RS) of NN inputs in terms of mapping capability.

Temporal and spatial forecasting of ionospheric critical frequency using neural networks

The ionospheric critical frequency, ƒoF2, is forecast 1 hour in advance by using artificial neural networks. The value ƒoF2 at the time instant k of the day is designated by ƒ(k). The inputs used for the neural network are the time of day; the day of year; season information; past observations ƒoF2, the first difference Δ1 (k) = ƒ(k) = ƒ(k − 1); the second difference Δ2 (k) = Δ1 (k) − Δ1 (k − 1); the relative difference RΔ(k) = Δ1(k)/ƒ(k); geomagnetic indices Kp, ap, Dst, sunspot number, and solar 10.7-cm radio flux; and the solar wind magnetic field components By and Bz. This paper gives a new method, and it is the first application of neural networks for modeling both temporal and spatial dependencies. In order to understand the physical characteristics of the process and determine how important a particular input is, a test which shows the relative significance of inputs to the neural networks is performed at the output. The performance of a neural network is measured by considering errors. For the errors to be more meaningful, training and test times and times for comparison with other results are selected from the same solar activity period. Among the various neural network structures, the best configuration is found to be the one with one hidden layer with five hidden neurons, giving an absolute overall error of 5.88%, or 0.432 MHz.

Temporal and spatial forecasting of the foF2 values up to twenty four hours in advance

Abstract Radio waves of a wide range of frequencies from very low frequency (VLF) to high frequency (HF), (broadly 3 to 30 MHz) can be propagated to great distances via the ionosphere. Since the largest variability occurs in the F-region the objective of this paper is to demonstrate a neural network model with the backpropagation algorithm which is designed to forecast the foF2 values of the highly nonlinear ionosphere up to 24 hours in advance. In other words, the model forecasts all values from 1 to 24 hours ahead. By using foF2 data for three European Ionospheric stations this neural network based model can forecast foF2 values both in time and in space for those three stations. The model seems promising for practical work since the root mean square errors involved are within reasonable limits.

An attempt to model the influence of the trough on HF communication by using neural networks

Trough is an interesting phenomenon in characterizing the behavior of the ionosphere, especially during disturbed conditions. The subject, which was introduced around the 1970s, is still attracting attention, especially during recent years. In HF communication, in particular, over the midlatitude ionospheric regions the electron density trough exhibits a phenomenon of abrupt gradients of electron densities in space and time which are directly reflected to foF2. Thus the performances of HF communications are directly affected. In this work an attempt has been made for the modeling to quantify the influence of the ionospheric midlatitude electron density trough on the ionospheric critical frequency foF2 by using neural networks. Data sets are used from the ground stations that include observations in the trough region. It has been demonstrated that the neural-net based approaches are promising in modeling of the ionospheric processes. Data generated by using statistical relationships are used to train the neural network. Then the trained neural network is used to forecast the ionospheric critical frequency, foF2, values 1 hour in advance for the cases when the probability of influence of the trough is high. Preliminary results will be presented to discuss the suitability of the neural-network-based approach in the modeling of complex processes such as the influence of the trough on foF2. The basic contributions of this work are 1) generation and organization of significant data for teaching complex processes, 2) neural-network-based modeling of a highly complex nonlinear process such as the influence of the trough on foF2 forecasting, and 3) general demonstration of learning capability by calculating cross correlations and general demonstration of reaching a proper operating point by calculating errors (that is, during the optimization process the neural network reaches the global minimum by using the gradient descent method).

Neural networks and cascade modeling technique in system identification

The use of the Middle East Technical University Neural Network and Cascade Modeling (METU-NN-C) technique in system identification to forecast complex nonlinear processes has been examined. Special cascade models based on Hammerstein system modeling have been developed. The total electron content (TEC) data evaluated from GPS measurements are vital in telecommunications and satellite navigation systems. Using the model, forecast of the TEC data in 10 minute intervals 1 hour ahead, during disturbed conditions have been made. In performance analysis an operation has been performed on a new validation data set by producing the forecast values. Forecast of GPS-TEC values have been achieved with high sensitivity and accuracy before, during and after the disturbed conditions. The performance results of the cascade modeling of the near Earth space process have been discussed in terms of system identification.

Teaching and Research in Industrial Process Control and Instrumentation

Abstract Some of the important aspects of teaching and research in industrial process control and instrumentation are considered based on the vast experience and in the light of some new developments in teaching.

The Use of Look Ahead Factoring in Neurocontrol Using Forward Plant Dynamics

Abstract Almost all of the neurocontrollers reported in the literature use inverse plant dynamics in the generation of control actions. In this work forward plant dynamics are used and a novel method based on look head factoring for the control of dynamic plants with imperfect neurocontrollers is presented

Two possible approaches for ionospheric forecasting to be employed along with the IRI model

Ionospheric forecasting is a popular research area required by telecommunication and navigation system planners and operators. The problem is challenging because ionospheric processes are nonlinear. Data-driven techniques are of particular interest since they overcome most of these difficulties. In this work, two possible ionospheric forecasting approaches have been considered to be employed along with the IRI model. The authors reported these approaches previously. Ionospheric critical frequency values have been forecast using Fuzzy inference and Neural Networks considering the two possible approaches, METU-FNN and METU-NN. In parallel, the foF2 values have been calculated based on the IRI model.

Some Basic Points in Control Education For Industry

Abstract Control is a subject that heavily relies on mathematics and physics. High level theoretical treatments are more easily understood and digested when accompanied by concrete examples. In this study, two important phenomena, namely linearization and integration are chosen as two cases that are important in industrial processes. Examples of teaching approaches are demonstrated around these two cases. Some simple examples are chosen, for easy demonstration, without the loss of generality. The approaches are believed to be effective in teaching engineering students and especially engineers working in industry.

A Twenty Two Year Experience in Teaching Process Control and Instrumentaion

Abstract In the Electrical and Electronic Engineering Department of the Middle East Technical University, Process Control and Instrumentation courses have been taught since 1974. The courses and the laboratory experiments have been basically designed by the author. In this paper this twenty-two year of experience between 1974-1996 is summarized.