Fault correction of algorithm implementation for intelligentized robotic multipass welding process based on finite state machines

Abstract

• Modeling the typical welding process via finite state machines for process monitoring. • Rules are presented to judge whether the identified feature points are effective fault detection and diagnosis. • Fault detection and diagnosis for the method of laser stripe extraction to obtain higher ratios. • Implementation of fault detection and diagnosis based on finite state machines for more stable running of the peripheral software system. The intelligentized robotic multipass welding process (IRMWP) involves adjustments of welding parameters, posture adjustments of the welding torch, real-time decision making of the tracking point, etc. It constructs a typical mixed logical dynamical (MLD) process. In order to smooth the welding process with thick steel plates, the typical process of implementing two continuous welds with T-joints is first modeled based on finite state machines (FSMs) at software and hardware levels. This process consists of four stages: preparation, determination and adjustment, tracking, and return stage. In these stages, some algorithms that are integrated into the peripheral software system (PSS) terminate the welding process when empirical parameters set in them are inappropriate. Since the weld profile extraction algorithm is the prerequisite for subsequent operations, this paper then presents a strategy to adaptively alter the empirical parameters arranged in this algorithm. The strategy implements fault detection and diagnosis (FDD) for the extraction process. Welding experiments are conducted under the framework of the proposed model, and results show that the proposed method leads to better stability of the PSS and higher ratios of successful weld profile extraction, over 95%. This research is of practical significance for strengthening the stability of the IRMWP with thick steel plates and improving welding quality.