The performance of three centrifugal pumps designed to operate at a rotational speed of 151.84 rad/s and flow rates of 1, 25, and 45 kg/s is being investigated for both water and non-Newtonian fluids at various rotational speeds and flow rates. The analyses are being conducted experimentally and numerically within the flow rate range of 0.25–55 kg/s and rotational speed values between 52.36 and 151.84 rad/s. Additionally, artificial neural networks (ANN) trained using experimental pump performance data are being tested with experimental and numerical values obtained at a new rotational speed of 130.9 rad/s. The non-Newtonian fluids being tested include CMC 0.2% and CMC 0.4%, comprising carboxy methyl cellulose (CMC) solution and water. The results indicate that the pump's performance when handling non-Newtonian fluids is significantly influenced by the pump's geometry, rotational speed, and flow rate. In design parameters, the head obtained with 0.2% CMC for pump 1 is 3.3% greater than that in water. For pump 2, the highest head is in water according to design parameters. Pump 3 exhibits the highest head at a CMC of 0.4 in design parameters, and this value is 0.81% higher than the value with water. Experimental and numerical results demonstrate good agreement, especially in design parameters. The head obtained from numerical analyses with the RNG k–ε turbulence model for pumps 1, 2, and 3 at design parameters is 3, 10, and 9.83 m, respectively. The corresponding experimental heads are 3, 10, and 9.84 m, respectively. However, discrepancies between these results increase with higher flow rates and the use of non-Newtonian fluids. The compatibility of ANN results with experimental results is better than with numerical results, particularly at higher flow rates than the design condition. Pump performance values estimated by ANNs are 2% lower than the experimental results. This study provides comprehensive experimental data on the use of non-Newtonian fluids in different centrifugal pumps, and it also offers important guidance for future research by comparing ANN and computational fluid dynamics.
Read full abstract