Process monitoring, which is also called state estimation, is very important for implementation of on-line control strategies . Dissolved oxygen (DO) and pH are the most commonly measured parameters using electrochemical sensors . However, some key state variables, such as biomass concentration, may not be measured directly due to the lack of suitable sensors or high costs. In recent years, lots of efforts have been involved in on-line software sensor (softsensor) development. The key concept of softsensor techniques is to estimate unmeasured states from measured states. Unmeasured states are normally inaccessible or difficult to measure by means of biosensors or hardware sensors, while measured states are relatively easy to monitor on-line using reliable well-established instruments. Based on this philosophy, several softsensor techniques have been proposed in the literature , namely:
• estimation using elemental balances ;
• filtering techniques (Kalman filter, extended Kalman filter) .
The first two methods suffer from the inaccuracies of available instruments and models. The third method requires much design work and prior estimates of measurement noise and model uncertainty characteristics. It also suffers from some numerical problems and convergence difficulties due to the approximation associated with model linearization.
Was this article helpful?