Prediction Of Compressive Strength Of Different Concrete Mixes Subjected To Sulfate Attack Using Artificial Neural Networks / by Aya Mahmoud Anwar Taha ; Under the Supervision of Prof. Dr. Osama Abdalgafour HodHod

Thesis (M.Sc.)-Cairo University, 2023.

Bibliography: pages 120-126.

Use of the Artificial Neural Network to estimate compressive strength loss as a
result of sulfate attack, neural network model demonstrated great validity. ANNs have
the capability to quickly understand the input-output relationships for any complex
problem. Design charts were created which make it easy to predict how much
compressive strength would decrease after any certain time and under different sulfate
concentration for various concrete mixtures. The compressive strength of several types
of concrete containing partial cement replacement were predicted using artificial neural
network (ANN) models. A large amount of data with a variety of ranges are required to
build an ANN model so that it may be tested and trained. The model's input data, which
were gathered from earlier studies, examined the impact of sulfate attack on concrete.
There is no doubt that sulfates have a dangerous effect on compressive strength of
concrete structures and its damage takes a long time to show up. Experiments and tests
consume a lot of labor ,waste material and cost. However, by the help of these previous
experiments the data were used to predict the compressive strength of different concrete
mixtures including cement content, water aggregates, coal gangue powder waste glass
powder , silica fume , blast furnace slag and other additives; subjected to sulfate attack
by a mean of artificial intelligence. Artificial Neural Network were used to simulate
the previous experiments and understand the underlying pattern to predict the
compressive strength of concrete at different ages. Six models were trained with
various mixtures of concrete exposed a range of 0 to 20 percent concentration of
sodium and magnesium sulfate exposure to forecast new outputs of compressive
strength which were then compared to the real values of compressive strength to get a
correlation factor more than 90 percent which shows a good accuracy of these models.

تم بناء نماذج الشبكة العصبية الاصطناعية للتنبؤ بقوة الضغط لأنواع مختلفة من الخرسانة المعرضة لنسبة مئوية من كبريتات الصوديوم والمغنيسيوم. البيانات المستخدمة في النموذج والتي تم جمعها من باحثين سابقين ، درست تأثير الكبريتات على الخرسانة. تم تحديد مقاومة الانضغاط لمدة 7 إلى 730 يومًا بشكل تجريبي للعينات التي تحتوي على ما يصل إلى 20 ٪ من كبريتات الصوديوم وما يصل إلى 20 ٪ من التعرض لكبريتات المغنيسيوم. تم استخدام ما مجموعه 353 عينة كمجموعات بيانات لتدريب واختبار الشبكات. تم تنويع معاملات الإدخال ، مثل محتوى الأسمنت ، والركام ، ونسبة الماء إلى الأسمنت ، وخبث الفرن ، ونوع الركام والنسبة ، لكل تجربة. تم نمذجة ست مجموعات من البيانات باستخدام الشبكات العصبية الاصطناعية بخلطات مختلفة للخرسانة للتنبؤ بالمقاومة , كان معامل التقارب لمقاومة الضغط المتوقعة هو واحد تقريبًا لجميع النماذج الستة عند مقارنتها بالقيم الحقيقية وهذه تعتبر دقة جيدة لهذه النماذج

Issued also as CD

Text in English and abstract in Arabic & English.