Alberto García-Villoria

The preventive maintenance scheduling problem involves sequencing the inspection of M machines within a cyclic policy of length T periods, where each machine must be inspected at least once. In the case addressed here, and during each period t, the maximum capacity for maintenance tasks is limited to one. The machines to be maintained can generate two types of costs: the maintenance cost, incurred in the periods when the machine is inspected, and the operational cost, which arises in the periods when a machine operates without being inspected. Both costs are directly proportional to the number of periods elapsed since the last maintenance of the machine. For this type of problem, it is common to use cycle lengths predetermined by external factors (such as days in a month or weeks in a year). However, this article introduces a mathematical program that calculates the ideal cycle length T and the optimal maintenance frequencies. The results show that in 69,44% of the instances, an ideal T different from predetermined T value is found. The average execution time is 106,96 seconds. Once the ideal cycle length T is obtained, a matheuristic algorithm is then used to sequence the maintenance tasks for each machine. It is observed that in 65,6% of the instances, the average cost of the maintenance policy is lower than when using predetermined T values, achieving an average cost reduction of 2,41%. Key Words: Mathematical programming, preventive maintenance, matheuristic algorithms, activity sequencing.

Multi-operator assembly lines are used in manufacturing industries that process a high volume of large parts, such as the automotive industry. An assembly line consists of a sequence of stations through which parts flow.

Multi-manned assembly lines are widely used in manufacturing industries that process a high-volume of large-sized workpieces. At each workstation there are multiple workers simultaneously performing different tasks with the possibility of interfering with each other, leading to an increase in the processing task times. This paper studies this type of problem: the multi-manned assembly line balancing problem with dependent task times (MALBP-DTT). As discussed in recent literature, the HEUR_PART procedure presents the best behaviour for solving MALBP-DTT. In this paper, HEUR_PART is improved by using the Empirically Adjusted Greedy Heuristics (EAGH) procedure along with a new procedure (named “EAGH-CKTL”) that is presented in this paper and is based on using EAGH combined with the cocktail of heuristics concept. EAGH and EAGH-CKTL are used to design new priority rules for solving MALBP-DTT through the HEUR_PART steps. In particular, EAGH-CKTL is applied for building new priority rules that have good performance as part of a cocktail of heuristics. The computational experiments show the efficiency of using both EAGH and EAGH-CKTL in the process of designing efficient priority rules: one of the priority rules designed with EAGH presents better performance than any other rule proposed in the literature for HEUR_PART, while another rule designed with EAGH-CKTL evidences a remarkable improvement in the HEUR_PART results when added to its original cocktail of heuristics. Keywords: combinatorial optimisation; assembly line balancing; multi-manned workstations; dependent task times; EAGH; cocktail of heuristics.