Farzan Rashidi
Center of Excellence for Control and Intelligent Processing
Department of Electrical and Computer Engineering
Tehran University
emails: farzan_rashidi@yahoo.com, f.rashidi@ece.ut.ac.ir
Abstract:
Sensor failures are a major cause of concern in many industrial systems such as engine-performance monitoring . In this and similar applications the quantity which must be measured may be physically difficult to access, e.g. temperature. Moreover, the sensors utilized in these situations are susceptible to changes in physical parameters. For instance, the time constant of the sensors may change over time. This fact, may affect the performance of the closed-loop control system whose signals are based on the feedback signals measured from the sensors. Multiple measurements of the same parameter with more than one sensor can be helpful in decreasing the impact of the unfavorable effects, while the correlation between the different sensors can result in a less vulnerable signal. In other words, through a combination of distributed sensing and measurement, deficiencies in any of the parameters of the sensing system can be addressed. Industrial applications of sensory signal fusion algorithms makes the subject important from the standpoint of safety and robust performance. Thus, enhancing the ability of the system to compensate for time delays in the feedback loop is highly motivated. To ameliorate the effects of faults of the sensor fusers, in non-deterministic and novel situation, some stochastic and probabilistic methods have been proposed, while some traditional filtering methods and soft computing intelligent algorithms have also been put forward. In this paper, a biologically motivated algorithm originating in emotional processes in the limbic system of the mammalian brain -termed BEL- is used to stabilize a control system which suffers from time delays in sensory feedback signals. The practical problems with sensor physical parameters' variations are usually inevitable. One of the most common effects produced due to these variations is deviation of nominal time constant of the sensor, which depends on the physical conditions under which the sensors function. The changes in the time constants of the sensors reflect unwanted -and usually unknown- delays which may deteriorate the performance of the closed loop control system that relies on the sensory feedbacks of the affected sensors. In this paper, we show that using a module to intelligently fuse the feedback signals could compensate for such effects and improve the performance of the closed-loop system and even stabilize it in the cases where the system goes unstable. The signal fuser module used here, BEL, works based on the values of the different sensory signals and a feedback of the fused one to generate the emotional cue, and then tries to improve the fused signal to ameliorate the effects of delayed ones.