Neuro compensated sliding mode control with nonlinear surfaces for pipe crack sealing manipulator

Abstract

This work focuses on ensuring control over the pipe crack sealing manipulator (PCSM) to seal the cracks within concrete pipes. PCSM is a tree-shaped robot featuring five specialised branches, suggesting a multifunctional design for crack-sealing operation within the pipe. PCSM demonstrates versatility in navigating vertical and horizontal pipes. A CAD model featuring the pipe, target crack, and PCSM has been modelled in SolidWorks, Utilising crack data from SolidWorks, the inverse kinematics and dynamics of the model are being simulated in Simulink for precise control. Observation and simulations study in SoildWorks revealed that only the fifth branch of PCSM successfully executed crack repairs over a substantial length. The tracing of the actual crack trajectory for effective sealing within the concave pipe becomes more challenging in the presence of disturbances and uncertainty in the system. To address this issue, sliding mode control (SMC) is employed with a nonlinear surface (SMCNS), which proves effective in handling external disturbances and uncertainty. To enhance control performance, a neural network (NN) compensator is combined with SMCNS and the proposed controller is called neuro-compensated sliding mode control with the nonlinear surface (NCSMCNS). The incorporation of a neural network with a nonlinear surface leads to the convergence of tracking error to zero, a conclusion validated through the use of Lyapunov theory. Furthermore, the performance of the proposed controller has been compared with SMC (linear and non-linear surface) and neuro-compensated SMC (linear surface).