Model validation and higher order sliding mode controller design for a research reactor S.H. Qaiser a , A.I. Bhatti b , Masood Iqbal c, * , R. Samar b , J. Qadir c a Center for Advanced Studies in Engineering, 19 Ataturk Avenue, Islamabad, Pakistan b Mohammad Ali Jinnah University, Islamabad, Pakistan c Nuclear Engineering Division, Pakistan Institute of Nuclear Science and Technology, P.O. Nilore, Islamabad, Pakistan article info Article history: Received 17 June 2008 Received in revised form 24 October 2008 Accepted 29 October 2008 Available online 10 December 2008 abstract The paper pertains to model validation and novel higher order sliding mode controller design for a nuclear research reactor. Sliding mode controllers for nuclear reactors were reported before but higher order sliding mode controllers have added advantage of reduced chattering. As a first step of model development a simulation model of control rod drive mechanism (CRDM) has been developed using SIMULINK Ò . This model has been validated with a lab based CRDM model, which is similar to Pakistan Research Reactor-1 (PARR-1) CRDM system. The nonlinear model of PARR-1 has been tuned and validated with experimental data. This model has been subsequently used for higher order sliding mode controller design and performance evaluation. Certain parameter values have also been recalculated to ensure model accuracy. Based on the validated model a robust nonlinear controller for controlling output power by manipulating control rod position has been developed and simulated. The new controller showed improved performance as compared to the classical PID controller. Ó 2008 Elsevier Ltd. All rights reserved. 1. Introduction Accurate mathematical model of a system is essential for the correct development of control system. If a plant is not modeled correctly then consequently designed controller will exhibit poor performance. Nuclear reactors are highly complex, nonlinear, time-varying, and constrained systems, so accurate model devel- opment and validation is a very difficult task. Also plant parame- ters vary with time, for example, the nuclear plant characteristics vary with operating power levels, ageing effects and changes in nu- clear core reactivity with fuel burnup. These parameter variations make accurate model development a difficult task. In this paper nonlinear and linear models found in literature (Edwards et al., 1990) are modified and correspondingly tuned for a local 10 MW swimming pool type research reactor PARR-1. PARR-1 was first made critical on 21st December 1965. The instrumentation and control system was upgraded in 1986 and the reactor power from 5 MW to 10 MW was upgraded in 1992 and also the high enriched uranium (HEU) fuel was replaced with light en- riched uranium (LEU) fuel. There are five control rods. The reactor becomes critical when these rods are withdrawn about 56% from their fully inserted position. After reactor becomes critical, a single rod is used to regulate the reactor power (Pakistan Research Reac- tor-1, Final Safety Analysis Report, 1991). Being a research reactor a huge body of research literature exists but so far no formal attempt made for analytic control system design is known to the authors. For any such exercise a dynamic model is essential. This paper is aimed at validating a novel model for PARR-1 and the design and simulation of a novel higher order sliding mode controller for PARR-1. The reactor modeling consisted of various steps. Various models for reactors can be found in literature e.g. (Huang and Edwards, 2002; Glasstone and Sesonske, 1986; Shtessel, 1998). These models vary in size and complexity. The model best suited for the reactor under consideration was chosen. Once the model structure is se- lected, the next step would be to populate the chosen model struc- ture with PARR-1 parameters. These parameters were acquired through different resources and procedures (Pakistan Research Reactor-1, Final Safety Analysis Report, 1991; Iqbal et al., 1997; Ansari et al., 1994). Once the populated model is obtained then this model is appended with CRDM. This model was simulated using MATLAB Ò . Reactor trials were designed for model validation. From reactor trials test vectors (experimental data) were generated to tune and validate the constructed model. CRDM part was properly tuned to conform the model response to the acquired test vectors. Various reactor parameters were evaluated for PARR-1. To verify model fidelity the SIMULINK Ò model for CRDM was used with experimental data. This model validation is fundamental in ana- lytic controller design as no such attempt is known to be previ- ously reported. After model validation a second order sliding mode controller is designed and simulated for regulating output power. The nonlinear model of a system covers a wide operating range and helps in better control of the system. The linear model behaves well at the operating point, where the model is linearized, 0306-4549/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.anucene.2008.10.005 * Corresponding author. Tel.: +92 51 2207264; fax: +92 51 9290275. E-mail address: masiqbal@hotmail.com (M. Iqbal). Annals of Nuclear Energy 36 (2009) 37–45 Contents lists available at ScienceDirect Annals of Nuclear Energy journal homepage: www.elsevier.com/locate/anucene