Prediction of gas molecular weight an‎d density using temperature, pressure an‎d accousting velocity; Analysis with COMSOL an‎d Artificial Intelligence

11/21/2025 12:00:00 AM

Prediction an‎d determination of the thermodynamic properties of gases present in flare systems are of great importance in the oil, gas, an‎d petrochemical industries. Accurate identification of these properties enables a reliable an‎d well-documented eva‎luation of the flowing gas type an‎d a precise analysis of its physical an‎d thermodynamic behavior. Beyond enhancing efficiency an‎d process optimization, this knowledge plays a crucial role in improving safety, reducing potential hazards, an‎d ensuring the stable performance of flare systems. Among various gas properties, molecular weight an‎d density are fundamental parameters whose precise determination is essential for thermodynamic analysis, system design, an‎d process control. In this study, based on the available data for temperature, pressure, an‎d sound velocity, an accurate prediction of these properties in the flare system has been performed. The sound velocity, as a sensitive parameter reflecting physical behavior, was obtained through detailed acoustic simulations in COMSOL software under both flowing an‎d non-flowing conditions. Subsequently, the AGA‑10 equation, recognized as a reliable model in gas thermodynamics, was employed to calculate the molecular weight an‎d density with high accuracy. This equation can predict gas properties under various operational conditions based on well‑established physical an‎d thermodynamic relationships. Finally, an intelligent approach based on machine learning an‎d a multilayer perceptron (MLP) neural network was developed to identify the gas type in the flare system. By analyzing the input data—including sound velocity, molecular weight, an‎d density—this method accurately classifies the gas type, marking a significant step toward the intelligent monitoring an‎d control of flare systems. The results showed that the calculation error for sound velocity was less than 4% for light gases an‎d less than 7% for heavy gases. Moreover, the predicted molecular weight an‎d density values for industrial flare gases exhibited a relative deviation of less than 8% compared with experimental data, confirming the high performance an‎d reliability of the proposed model in predicting gas properties across a wide range of molecular compositions.

وزن مولكولي گازها , دانستيه گازها , سرعت صوت در گازها , شبيه سازي مبدل فراصوت , روابط AGA-10 , شبكه عصبي MLP

Gas molecular weight , gas density , speed of sound in gases , ultrasonic transducer simulation , AGA-10 correlations , MLP neural networks

Soheil Mofavaz

Dr. Seyed Hasan Hashemabadi