New modified smith predictor designs

Abstract

By combining several existing modified Smith predictor structures, presented in a literature review, a generalised form for the Smith predictor is proposed. From this structure, excellent servo and regulator responses can be obtained by specifying three controller dynamic elements. Two new modified Smith predictor structures are presented with their associated tuning rules. The results of the simulations indicate that the two new designs give better overall responses than the original Smith predictor.

Implementation of new modified Smith predictor designs

“…This structure is fully explained by Sourdille and O'Dwyer [2]. From this article, it is found that the controllers are given by equations (1), (2) and (3) and that the associated tuning rules, depending on the index…”

“…This modified Smith predictor is presented in detail by Sourdille and O'Dwyer [3]. This article explains the step by step procedure used to obtain the controller transfer functions given by equations (4), (5) and (6); the associated tuning rules depending on the index…”

“…This comparison is presented in detail by Sourdille and O'Dwyer [4]. Table 9 shows the number of simulations in which improvement in response was detected, when the first modified Smith predictor was used instead of the second modified Smith predictor, with the responses evaluated using the relevant indices.…”

“…Implementation is an important issue for dead time compensators as it validates the design. The paper reviews the two modified Smith predictor structures developed previously by the authors [2][3][4], presents the simulation results and finally compares the different implementation results for the Smith predictor and the two modified Smith predictors, for the control of the PT326 process trainer. The original contributions compared to [2][3][4] are the validation of the tuning rules developed for the modified Smith predictor structures and the proof that the modified Smith predictor structures achieve better servo and regulator responses than the Smith predictor on a real process.…”

Abstract only

“…A potential drawback is the necessity of a distributed feedback channel, which, however, could be approximated through similar infrastructures. As future work we identify the use of a Smith predictor [22,30] to compensate for dead time due to delays and thereby stabilize the controller as well as an evaluation of fuzzy controllers [4] to verify if they are able to yield acceptable results in our system.…”