Application and Results

This section describes two separate case studies. The first case study is a four-tank experimental apparatus. This system has two measurements and two manipulated variables with up to 4 external unmeasured disturbances. The second case study is a CO₂ absorber simulation. This system has ten measurements, one manipulated input, and up to sixteen unmeasured disturbances. Control and estimation results from both case studies will be presented.

Figure 2 shows a general schematic for configuring the system. Each box in the diagram represents a separate physical computer or apparatus. The four-tank process is controlled by a Bailey Process Station apparatus, which can then be controlled by a PC. In Figure 2, each line is a flow of information and each block of text represents a computational process. Information travels between physical entities using TCP/IP and between processes on a PC using a Dynamic Data Exchange (DDE) protocol. For the computational processes, the indented text represents a subroutine of the process. MATLAB code has been developed to allow direct use of CPLEX and LOQO applications for solving LP, QP, and mixed integer problems. Independent MATLAB processes can pass and receive messages from a Java based application, allowing for non-blocking (continuous) operation of both processes. The Parallel Matlab Interface (PMI) toolbox also allows for communication between MATLAB processes, but receiving messages becomes a blocking task and the structure is parent-child instead of peer-to-peer.

  

Figure 2: Schematic of the control system