عدي البطاط

In an oil refinery, everything happens in large proportions: fluids flow in tons per hour, temperatures are measured in hundreds to thousands of degrees, and electricity is generated in megawatts. A small glitch or mistake in the system can seriously damage the entire plant and workers or cause a loss of revenue. Due to mission critical processes, oil and gas businesses have already implanted thousands of sensors in and around their physical systems. Raw sensor data is continuously collected and needed to be processed in the real time to predict error occurrence. In this thesis I have presented innovative algorithm based on neural networks to detect and classify the errors. For eva‎luation I use real data from oil refinery and extend it by LSTM network. The result of simulations shows that my proposed algorithm can detect error occurrence with the accuracy of 99.7% and classify detected error with the accuracy of 100%.

پردازش داده‌هاي پالايشگاه نفت , تشخيص خطا

Oil refinery data processing , error detection

Oday AL-Batat

Dr. Mohammad Riyahi