Sensor Fault Detection in Coupled Liquid Tank System F. Kousar, M. Abid, A. Q. Khan Department of Electrical Engineering, Pakistan Institute of Engineering and Applied Science, P. O. Nilore 45650, Islamabad, Pakistan engr.farzanakousar@gmail.com , mabid@pieas.edu.pk , aqkhan@pieas.edu.pk Abstract — As the field of automation is progressing, safer and reliable systems are highly desirable. Any malfunction in the plant result in reduced efficiency of the plant, reduced quality of the product and sometimes may result into fatalities. Therefore fault detection and process monitoring is becoming an integral part of modern control systems. The coupled liquid tank system is an experimental setup with nonlinear dynamics. The objective of this paper is to develop and implement fault detection techniques for coupled liquid tank system. The proposed scheme makes use of observer based residual generation and norm based residual evaluation. First system was linearized by Jacobian linearization and fault diagnosis system has been designed for the linearized system. Then this algorithm has been implemented on real plant and satisfactory results have been obtained. Keywords—Sensor fault detection, Observer based fault detection, Model based, Couple liquid tank syste. I. INTRODUCTION The widely accepted concept is that a fault is an unexpected change of system function although it may not represent physical failure or breakdown [1]. This paper discusses sensor faults. This is shown in figure below and can be described mathematically as described in [1].        Figure 1: Sensors, outputs and measured outputs [1] In other words, sensor faults disturb output measurements. These faults can create problems if we use measured output to further control some device. There are many different approaches which are used for the purpose of fault diagnosis. There are many survey papers that describe the details of these techniques [2-8]. In [9], a structured residual generation approach has been used for fault detection as well as fault isolation purpose for a coupled tank system. Among all these approaches, model based fault diagnosis techniques are most commonly used. Model based fault diagnosis can be defined as the detection, isolation and characterization of faults in components of a system from the comparison of the system's available measurements, with a priori information represented by the system's mathematical model. Faults are detected by setting a (fixed or variable) threshold on a residual quantity generated from the difference between real measurements and estimates of these measurements using the mathematical model [1]. This paper is organized as follows: Section 2 contains description about the plant, In Section 3; Mathematical model of the system has been derived, in section 4, a brief overview of fault diagnosis has been given. Section 5 deals with residual generation step and section 6 deals with residual evaluation stage. In section 7 contains results after implementing the algorithm on real plant. Section 8 concludes the work. II. DESCRIPTION OF PLANT Coupled Liquid Tank System (CLTS) is an experimental setup with highly nonlinear dynamics and is quite useful to test nonlinear control and fault detection algorithm. It consists of the cylindrical tanks with equal cross sectional area as shown in figure 2. Detailed description of CLTS can be found in [10]. Each tank is equipped with a sensor to measure water level. The plant can be used to simulate several kinds of faults; these include • Sensor faults • Component faults: can be simulated by opening or closing of valves • Actuator faults There is seepage in tanks which can be taken as disturbance. In this paper, we shall discuss the detection of sensor faults. Figure 2: Coupled liquid tank system III. MATHEMATICAL MODELING Cross sectional areas tank 1 and 3 do not vary with water level. While cross sectional area of tank 2 varies with the water level. Each tank basically acts as an integrator. Free flow out of each tank is a non linear function of the level in the tank and orifice discharge coefficient   . Different notations which are used throughout the paper are described in Table 1. 2012 10th International Conference on Frontiers of Information Technology 978-0-7695-4927-9/12 $26.00 © 2012 IEEE DOI 10.1109/FIT.2012.64 318