Adaptive control of a CSTR with a neural network model

An adaptive control algorithm with a neural network model, previously propo sed in the literature for the control of mechanical manipulators, is applie d to a CSTR (Continuous Stirred Tank Reactor). The neural network model use s either radial Gaussian or "Mexican hat" wavelets as basis functions. This work shows that the addition of linear functions to the networks significa ntly improves the error convergence when the CSTR is operated for long peri ods of time in a neighborhood of one operating point, a common scenario in chemical process control. Then, a quantitative comparative study based on o utput errors and control efforts is conducted where adaptive controllers us ing wavelets or Gaussian basis functions and PID controllers (IMC tuning wi th fixed parameters and self tuning PID) are compared. From this comparativ e study, the practicality and advantages of the adaptive controllers over f ixed or adaptive PID control is assessed. (C) 2000 Elsevier Science Ltd. Al l rights reserved.