Simulation model of a heated two-tank system with fault emulation

Abstract

Modelling and simulation of benchmark systems play a key role in advancing research control strategies and fault detection and isolation (FDI) techniques. Benchmark systems, especially with fault emulation capabilities are essential for developing FDI algorithms. An in-house developed heated two-tank system (HTTS) is capable of fault emulation, and a basic simulation model of the system also exists. The simulation model is however not representative of the physical plant. This study therefore improves the simulation model by redefining the underlying modelling approach and addresses the hardware limitations by specifying several hardware improvements. The improved model uses first-principle modelling approaches and first-order-plus-deadtime transfer functions. Parameter estimation was implemented using direct analysis or genetic algorithms under normal and faulty conditions. The tank and storage dynamics were modelled using first-principle modelling, whilst the control valves, heat pump, and gas heater were modelled by means of parametrised transfer functions. Model validation across a variety of operating conditions and scenarios, using both visual and statistical metrics showed significant improvements over the previous model. The simulation model had an improvement of tank level normalised root mean squared error (NRMSE) of 26.7% and an improvement on tank temperature of 2541%. Finally, based on the analysis of the hardware and the improved model, the hardware upgrade specifications were suggested. These hardware improvements will enhance the HTTS but will of course influence the system dynamics and will require a second iteration of simulation model validation. Overall, this study delivers an improved simulation model capable of accurately reproducing the HTTS dynamics and suggests possible upgrades to the physical system. These physical upgrades and improved model form a solid foundation for future FDI research.