Opportunistic maintenance policies for multi-machine production systems with quality and availability improvement

Highlights

• A joint model for a series multi-machine production system is developed.

• A quantitative index is introduced to characterize the condition of the machines.

• A hybrid performance factor is developed to evaluate the joint effect of maintenance.

• Economic dependence and availability dependence are considered in the PM schedule.

Abstract

Effective maintenance policies can improve the production system performance by alleviating product quality degradation and increasing system reliability and availability. However, reliability, product quality degradation, and availability are conflicting optimization objectives and are rarely considered together in maintenance policy optimization. Aiming at optimizing these three objectives simultaneously, this paper proposes a joint model for opportunistic maintenance (OM), and product quality and availability improvement of a series multi-machine production system (MMPS). First, a condition index (CI) is developed to characterize the condition of individual machines based on monitoring data, and the corresponding threshold functions are developed to realize maintenance decision-making for individual machines. Then, considering maintenance dependence in the MMPS, based on the CI, the dual dynamic threshold functions are introduced: the preventive maintenance (PM) and the OM threshold functions, and a novel dual-dynamic-thresholds-based OM policy is developed. Whenever the CI of a machine reaches its PM threshold, all the other machines whose CI reaches the OM zone will be maintained together with this machine. Finally, a case study and comparisons are performed to show that the proposed model can simultaneously provide good production economic performance, high quality, and availability.

Introduction

Globalization and fierce market competition make manufacturing enterprises face great challenges in maintaining profits and competitive predominance. The machines in a production system undergo wear and degradation with age and usage. The degradation and wear can increase the risk of machine malfunction and deteriorate the product quality of the production system. Maintenance is used to mitigate the production system degradation and keep high product quality and production availability [1]. However, non-optimal maintenance activities can bring large maintenance costs and/or lead to long system downtime, resulting in low availability and large production losses. The maintenance problem is challenging as production system reliability and availability, and product quality are inherently conflicting objectives [2], which calls for a trade off by an effective maintenance policy with also lowest possible costs.

In the past few decades, the integration of two by two of the three key production system performances: reliability, product quality, and availability, has been extensively studied [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18]. Through the combination of statistical process control (SPC) and preventive maintenance (PM), some scholars have researched the combination of maintenance optimization and quality improvement from different aspects, such as control charts and acceptance sampling [9], [10], [11]. Yet, these studies do not address underlying mechanism of the impact of machine or system degradation on quality deviations. Jin and Shi [12] proposed the state-space model to establish the formation mechanism of product feature deviation in the production process. On this basis, Jin and Chen [13] proposed the QR-Co-Effect method to analyze the relationship between component reliability and product quality deviation. Then, based on the response surface model [19] and Taguchi’s loss function [20], Chen and Jin [14] further studied product quality degradation for the production system. However, these studies mainly focus on the relationship between product quality degradation and single machine system. The impact of the time required for maintenance actions that could exacerbate system unavailability and lead to further loss of profit, is often ignored.

As one of the key system performances, availability has been extensively studied in the reliability field [2], [15], [16], [17], [18], [21], [22], [23]. However, there is limited research on single-machine PM decision-making that integrates reliability, product quality degradation, and availability, simultaneously, let alone the multi-unit production system. It is clear that reliability and product quality, and availability are conflicting objectives: frequent maintenance activities can keep high reliability and product quality but would bring large maintenance costs and lead to frequent system downtime which results in low availability and thus more production losses; compromising reliability for cost efficiency would bring worse product quality. Besides, maintenance optimization integrating reliability, product quality degradation, and availability is a multi-objective optimization, whose methods of solution are mainly classified in two categories: traditional optimization methods [15], [17] and intelligent algorithms [2], [16], [21]. However, in practice for some traditional methods such as the weighted method, the determination of weights cannot be quantified, and for intelligent algorithms, it takes too much time to solve the target, which is not conducive to the practical application of maintenance policies. Therefore, a hybrid performance factor that combines traditional methods and intelligent algorithms is developed in this paper to achieve the multi-objective optimization.

Note that the studies mentioned above mainly involve single machine systems. A multi-machine production system (MMPS) is a system containing multiple machines working together to yield a final product. Such systems are quite common in the production industry, such as continuous flow manufacturing systems, ferrite phase shifting unit manufacturing systems and automatic stamping production lines [24], [25], [26]. For MMPS, an important issue that needs to be addressed is the maintenance dependence among units [27], [28]. To exploit the dependence, when maintenance activity on one machine in the MMPS leads to system downtime or partial downtime, this downtime opportunity can be used to schedule other machines for simultaneous maintenance. This strategy is called opportunistic maintenance (OM) strategy. Over the past decades, various OM strategies have been proposed for a wide variety of multi-component/unit systems with different maintenance dependence [29], [30], [31], [32], [33]. However, in these OM studies, cost rate is the only optimization objective. The OM policies simultaneously optimized by reliability, product quality degradation, and availability are rarely developed, which will lead to worse product quality and system availability decrease and, thus, considerable production losses.

Meanwhile, these OM policies can be generally divided into two categories according to the OM criterion. One category is short-term cost-saving grouping opportunity maintenance, that is, grouping all units in the system with all possible combinations and determining the optimal combinations by maximizing the cost-saving [30], [32]. The other category leads to implementing OM by defining an OM zone, which refers to the status threshold (i.e. hazard rate, age, etc.) or maintenance time window [29], [31]. In this category, when a unit in the multi-unit system is stopped for PM, other units that fall into the OM zone are maintained together. However, for the first category, when the number of units contained in the system increases, the computational complexity of the short-term cost-savings method experiences exponential growth, since the OM alternatives need to be enumerated. For the second category, there are usually only one or two common constant OM zones for all units in the system. In this way, some units may become a bottleneck to reduce system maintenance costs. Besides, although cost optimization at the system level may be achieved, at the unit level it may cause ‘over maintenance’ or ‘under maintenance’.

In this paper, a novel opportunistic PM policy with condition index (CI) and dynamic threshold functions (curves) is proposed for a serial MMPS, where PM is performed to improve the reliability, product quality, and availability of the individual machine and the MMPS, simultaneously. First, the CI and the PM threshold function are developed to evaluate the condition of machines in the MMPS and to implement PM decision-making of the individual machine. A hybrid performance factor is proposed to achieve the simultaneous optimization of the three objectives: reliability, product quality, and availability. The most effective maintenance schedule for the each individual machine is obtained by optimizing the corresponding hybrid performance factor. Then, the OM threshold function is defined to develop the OM zone (the OM criterion). The OM zone of each machine is variable and separately determined with the hybrid performance factor. Finally, based on the OM zone, the tentative PM schedule of all machines are further adjusted to obtain the optimal PM schedule of the whole MMPS.

The remainder of this paper is organized as follows. Section 2 gives the problem description and some assumptions. In Section 3, the PM model for each machine is formulated. Section 4 develops the OM zone and describes the opportunistic PM scheduling methodology. A case study and a comparison analysis are conducted in Section 5 to demonstrate the validity of the proposed model. Section 6 presents some discussions and conclusions.