Measurements of dynamic behavior of a multistage flash water desalination system

This study focuses on measuring the process dynamics for the multistage flash desalination process (MSF) in an industrial unit with a capacity of 4546 m3/d. This is a novel addition to the literature because previous studies are limited to theoretical analysis of process dynamics or controller tuning, as well as conceptual design of conventional or advanced control systems. The measurements evaluate the performance of seven control loops, which include the pressure, temperature, and flow rate of the heating steam; the pressure of the vacuum ejector; and flow rates of the brine recycle, make-up seawater, and cooling seawater. All measurements start from steady-state conditions. The system is then set on manual where all control units are disengaged. Subsequently, only one control valve is adjusted by ±15% of its steady- state setting. A total of 14 experiments were performed involving simultaneous measurements of the system variables. Measurements showed non-linear behavior where increasing or decreasing the valve settings did not provide similar trends. Analysis of results shows that one of the most sensitive variables is the distillate level in the last stage: the distillate trays either were flooded or became dry. The brine level in the last flashing stage was also found to be sensitive to valve settings where level increase resulted in higher product salinity. The results and analysis presented provide a better understanding in system fault analysis which could be caused by improper operating conditions. These data are essential to propose, design, and evaluate advanced/comprehensive control systems for the MSF process.

Experimental measurements and analysis are presented for the dynamic response of a MSF desalination system with 19 flashing stages and a production capacity of 4546 m3/d. The experiments were performed upon variations in the flow rates of the brine recycle, feed seawater, cooling seawater, and heating steam, temperature and pressure of heating steam, and pressure of the vacuum ejector. Analysis of the results shows the following: 1. In all experiments system operation was terminated because of the increase or decrease of the brine or distillate levels. In all cases system performance deteriorated rapidly with a reduction in the distillate production rate and an increase in its salinity or reduction in the brine level with vapor flow-through across the stages. 2. The brine recycle flow rate had a very strong effect on the system performance where its increase resulted in a reduction in the brine level in the last stage and a reduction of the system temperature. 3. Increase of the brine level in the last stage was associated with an increase in the feed seawater flow rate or decrease in the brine recycle flow rate. In both cases, an increase in the brine level in the last stage increased the product salinity beyond the design value. 4. Decrease in the brine level in the last stage improved the flashing efficiency and resulted in an increase in the level of the distillate product to high values that result in tray flooding and product loss. Further decrease in the brine level resulted in vapor flow-through across the stages and rapid deterioration in operating conditions where the brine circulation stream flow inside the condenser tubes did not reach the design temperature before entering the brine heater. 5. Increase in the heating steam temperature or pressure resulted in increasing the system temperature, which increased the flashing rate or the net product flow rate. Also, this resulted in a decrease of the brine level that might result in vapor leaking across the stages. 6. Increase in the ejector vacuum pressure resulted in an increase in the entrainment rate of the heating steam as well as the vapor formed in the flashing stages. This reduced the system temperature and, as a result, it reduced the product flow rate and resulted in increasing the brine level. 7. The transfer functions and time domain for data fitting included ramp (first order), stable second order, unstable third order, and a combined function of the ramp and the stable second order. Fitting of data gave R2 values higher than 90%.