Model-free control system

It is apparent that a predictive model is the hearth a control algorithm. Unless the relationship between process inputs and process performance is known, deviations can be detected, but effective corrective action cannot be taken. However, fast, error-free monitoring is also essential unless inputs and state variables can be quickly and accurately quantified, the control system is blind and devoid of value. 

Early approaches to fault diagnosis were often based on the so-called physical redundancy [11], i.e., the duplication of sensors, actuators, computers, and softwares to measure and/or control a variable. Typically, a voting scheme is applied to the redundant system to detect and isolate a fault. The physical redundant methods are very reliable, but they need extra equipment and extra maintenance costs. Thus, in the last years, researchers focused their attention on techniques not requiring extra equipment. These techniques can be classified into two general categories, model-free data-driven approaches and model-based approaches. 

Various performance indices have been suggested [54, 53, 149, 20, 148] and several approaches have been proposed for estimating the performance index for SISO systems, including the normalized performance index approach [53], the three estimator approach [175[, and the filtering and correlation analysis (FCOR) approach [115[. A model free approach for linear quadratic CPM from closed-loop experiments that uses spectrum analysis of the input and output data has been suggested [136]. Implementation of SISO loop based CPM tools for refinery-wide control loop performance assessment has been reported [294]. 

Process models are also important components of reactor control schemes. Kiparissides et al. [17] and Penlidis et al. [16] have used reactor models for control simulation studies. Particle number and size characteristics are the most difficult latex properties to control. Particle nucleation can be very rapid and a strong function of the concentration of free emulsifier, electrolytes and various possible reagent impurities. Hence the control of particle number and the related particle surface areas can be a difficult problem. Even with on-line light scattering, chromatographic [18], surface tension and/or conversion measurements [19], control of nucleation in a CSTR system can be difficult. The use of a pre-made seed or an upstream tubular reactor can be utilized to avoid nucleation in the CSTR and thereby imjHOve particle number control as well as increase the number of particles formed [20-22]. Figures 8.6 and 8.7 illustrate open-loop CTSR systems for the emulsion polymerization of methyl methacrylate with and... 

Ramakrishnan and Zukoski (2000) extended the work of Rosenbaum et al. and tested the ability of different pair potentials to characterize the interactions and phase behavior of STA. The strength of interaction was controlled by dispersing STA in different salt concentrations. The experimental variables used in characterizing the interactions were the osmotic compressibility (dP/dp), the second virial coefficient(.82), relative solution viscosity and the solubility. Various techniques were then developed to extract the parameters ofthe square well, the adhesive hard sphere and the Yukawa pair potentials that best describe the experimental data. As mentioned before, the adhesive hard sphere potential describes the solution thermodynamics only where the system is weakly attractive but as would be expected fails when long range repulsions come into play at low salt concentrations. Model free representations were then presented which offer the opportunity to extract pair potential parameters (F/g. 19-8). 

The computing system must be fed input data if it is to generate any output. Three items of information are available from each loop set point, measurement, and controller output. (The error signal cannot be used in the steady-state model because it has no steady-state value.) Of this information, the set points are most useful because they represent the exact demands on the control system and are free from feedback transients, lying outside the loops. 

---