N-policy for redundant machining system with double retrial orbits using soft computing techniques

The present study is concerned with the performance prediction of a double retrial orbit redundant repairable machining system. Both primary and secondary orbits are available as waiting/buffer space for the failed units. In these orbits, the failed units can reside and make re-attempts for the repair. As per N-policy, if there are no units in the orbits for the repairing job, the repairman goes on vacation and further starts the repair job when N-failed units are accumulated. The objective of this investigation is to evaluate the transient and steady-state distributions of the queue length of failed units under N-policy. The matrix analytic and matrix recursive methods are utilized for solution purpose while an adaptive neuro-fuzzy inference system (ANFIS) is employed for validating the feasibility of designing the AI-based controller. The harmonic search (HS) and particle swarm optimization (PSO) methods have been implemented for the cost optimization purpose so as to evaluate the optimal design parameters. The outputs of study provides critical insights into optimal system performance and improving the repair policy. Furthermore, a practical application of this investigation is demonstrated in a telecommunications network traffic system, where the proposed methods can be utilized to manage the maintenance issues of routers in the network traffic.