Artificial intelligence: a modern approach
Artificial intelligence: a modern approach
Robust model-based fault diagnosis for dynamic systems
Robust model-based fault diagnosis for dynamic systems
Computational Intelligence in Control Engineering
Computational Intelligence in Control Engineering
Fault Diagnosis: Models, Artificial Intelligence, Applications
Fault Diagnosis: Models, Artificial Intelligence, Applications
Discredibility detection of a sensor in a control loop via computational intelligence
AIC'05 Proceedings of the 5th WSEAS International Conference on Applied Informatics and Communications
Hi-index | 0.00 |
The gradual loss of the credibility of sensors measuring control variables in control loops (so-called control variable sensor discredibility) may cause serious problems in control. As it will be shown, if such a sensor produces biased data, which are not correct but not totally wrong, then there is a danger that the controlled variable exceeds limits of a tolerance range without displaying any difference from the set point. However, this is not the only negative consequence, sometimes, undesirable side effects may occur and they remain unrecognized. The problem of these indirect unrecognized impacts becomes important especially in combustion processes. In case of a pilot stoke-fired boiler, which has been used for experiments in the CTU labs, the sensor discredibility problem is related with function of the oxygen probe used to accurate control of the burning air. Discredibility of this sensor will affect hidden increase of penalized gaseous emissions, above all CO and NOx. The general aim of our research is to enhance the function of the controller so that it will be able to warn the operator about changes in the properties of the control variable sensor. This paper introduces model-based control variable sensor discredibility detection method using evolutionary algorithms and documents the results obtained from testing the methods in simulated examples. A newly enlarged method also includes the ability to approximate the time until the control variable sensor discredibility.