Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Use of artificial neural networks to determine the groundwater levels in the Danube catchment area

Galanis Dimitrios

Full record

---

Summary

The monitoring of groundwater, in terms of their quantity and quality, is mandatory based on Greek and European legislation. However, their accurate monitoring of their level as well as their quality characteristics, is not possible, in every point of the respective study area. Instead, suitable computer models are designed that can give an accurate estimate of both quantity and quality.In recent years, instead of using traditional arithmetic models, based on the approximate solution of the underground flow equation, the use of artificial intelligence and in particular artificial neural networks (A.N.N.), for the estimation of both quantitative and qualitative characteristics of groundwater is gaining advantage.In the present work, a brief record of the most important innovations in the field of A.N.N., as well as all the developments that have established them as the main tool for the use of artificial intelligence in various fields of science, is listed. Afterwards, an analysis of the training methods used in machine learning is been made, followed up with the analysis of the basic equations used by the respective training algorithms.In the next part of the study, a brief report of the Danube river basin (which was used as study area) is been made. While the method used to organize the data in order to be able to be used in machine learning, is been described in detail. Followed up with a description of the MATLAB A.N.N. tool and the selection of the appropriate algorithm as well as the proper architecture of the A.N.N.A number of scenarios are been examined, in order to achieve a correct simulation of the groundwater levels in the whole study area. Followed up, by a validation part, in which a comparison of the predicted data with unused real data is been made.In conclusion, there is a review of this study. Including observations from the experimental results, as well as suggestions for changes and avoidance points for future studies in order to significantly improve the assessment of A.N.N.

Available Files

Services