155 Fuzzy Logic Control

RO desalination is a nonlinear process, which has to operate under specific conditions that are of utmost importance for plant operation optimization. The nonlinearity relates to changing process characteristics such as feed total dissolved solids and pH, which in turn cause changes to product quality and quantity parameters such as salt rejection and recovery. An RO desalination system is usually designed based on a defined set of data analysis such as flow, temperature, and feed water composition. However, in reality, plant operation has to be flexible in order to respond to changing variables. Integration of key process information into the control strategy's decisionmaking and prediction can yield an increase in the lifetime of the membranes, availability, and efficiency and optimize plant performance.

The nature of the membrane separation process and the characteristics of the membrane system impose a number of constraints on the system [19], [20]. These constraints require continuous monitoring and control if the system is to perform economically over a long period. The main operational constraints for RO desalination are as follow:

1. Pretreatment control for suspended solids to obtain biological and chemical stability;

2. Operation between a minimum temperature to provide required flux and a maximum temperature allowed by the membrane specifications;

3. Operation between minimum brine flow to avoid concentration polarization and maximum flow with respect to desired recovery;

4. Operation at a pressure to obtain desired mass transfer and equalization of pressure drop; and

5. Chemical characteristics of feed water and dynamics of mass transfer.

Loss in salt rejection and loss of permeate flow are the main problems encountered in RO plant operation. It is of utmost importance that corrective measures are taken as early as possible [21]. Some of the parameters, such as temperature and feed water salinity change naturally. Other parameters may change as a consequence of other changes that are present in the system. Once the problem has been identified, causes must be identified and corrective measures must be taken by the system. The RBFN predictions of recovery and salt rejection described in previous sections were used as part of the input to the FIS. Measured values from sensor information make up the remaining input values to the FIS. These inputs include the following:

1. Temperature

2. Feed TDS

3. Feed pH

4. Feed flow

5. Feed pressure

6. Brine pressure

7. Permeate conductivity

8. Salt rejection

9. Recovery 10. Predicted Scale Index

The fuzzy system was implemented using Mamdani architecture [22] with the following FIS properties:

And method: min

Or method: max

Implication: min

Aggregation: max

Defuzzification: centroid

The hardware system configuration is shown in Figure 15.15 and consisted of a fuzzy controller and a Siemens PLC interface to the controlled plant. A programmable logic controller (PLC) controlled input-output interface signals, and provided for data transfer to the fuzzy controller. The PLC included two 16-channel analog input and output cards and eight 32-channel digital input and output cards. The CPU used 32-bit architecture for all arithmetic and comparison operations and an expanded register set. Figure 15.16 shows the schematic diagram of the developed fuzzy controller. Figure 15.17 shows the process layout of the prototype RO plant built at FAU research laboratories with feed intake from the Atlantic Ocean.

Fuzzy PLC Controlled

Controller Plant

Sample/hold

0 0

Post a comment