Artificial neural network for predicting the performance of reverse osmosis desalination plants

العناوين الأخرى

الشبكات العصبية الاصطناعية للتنبؤ بأداء محطات التحلية بالتناضح العكسي

مقدم أطروحة جامعية

Jasim, Diya Jumah

مشرف أطروحة جامعية

Jasim, Ala Abd al-Razzaq

أعضاء اللجنة

Muhammad, Ahmad Abd
Sulaymunn, Abass H.
al-Fayiz, Mustafa M.

الجامعة

جامعة البصرة

الكلية

كلية الهندسة

القسم الأكاديمي

قسم الهندسة الكيميائية

دولة الجامعة

العراق

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2012

الملخص الإنجليزي

-Modeling and simulation of reverse osmosis system is considered as the main step for improving the quantity and quality of water production.

The conventional methods have depended on simulating the theoretical results relative to experimental data then degree of convergence is specified.

These methods are depended on arranging the mass balance equations and the equations of RO technology.

The profession method for testing the performance of RO system are depended on Artificial Neural Network.

This method can be applied to specified the best model which can be used to simulate the RO system under various operating conditions.

In this work, two cases have tested and investigated.

The first case is applied to construct ANN model for domestic RO system, when the experimental result is used as a training data in ANN model.

The second case is used to predict the weight percent of ANN model on operating parameters such as, feed pressure, feed temperature and feed concentration.

The weight of each parameters have specified and discussed it's effect on the performance of RO-membranes.

This case is simulated depending on the results of (ROSA72) software for FILMTEC membrane type (BW30-4040), DOW company.

For specifying the optimum ANN models, feed-forward back propagation has been used with one hidden layer, and employed Levenberg-Marquardt back propagation.

Different numbers and types of transfer functions for neurons in hidden layers have been explored in for selecting the best network that can be used to describe the experiment and software results.

The effects of the transfer function and neural network architectures on the ANN performance, as reflected by the (MSE), (R) and Epoch are evaluated and discussed.

ABSTRACT VII For, both cases, a comparison study between the experimental results (Target) and simulation results (Output) has carried out for finding the best ANN model.

The weight percent of ANN model for each parameters have evaluated for showing the effect of these parameter on the performance of RO system.

For case one, the weight percent of operating pressure, feed temperature and fouling (operating time) have varied between (25.84 to 89.3%), (5 to 1.5%) and (69.13 to 9.1%) respectively.

While for case two the weight percent of operating pressure, feed temperature and feed concentration have 64.3%, 6.5% and 29.2% respectively.

Neural network provides a possible method for handling any complex problem by providing adequate training data and sufficient number of nodes to represent the internal features and relationships that connect input and output parameters.

The overall results by using ANN models for both cases approved the ability of ANN model for describing and predicting the performance Of RO system under any operating conditions.

التخصصات الرئيسية

العلوم الهندسية والتكنولوجية (متداخلة التخصصات)

الموضوعات

• تحلية المياه
• المياه المالحة
• محطات معالجة المياه
• الشبكات العصبية(الحاسبات الإلكترونية)

عدد الصفحات

117

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-317077