Remaining useful life prediction and predictive maintenance strategies for multi-state manufacturing systems considering functional dependence

https://doi.org/10.1016/j.ress.2021.107560Get rights and content

Highlights

  • Connotation of functional dependence of manufacturing components is defined.

  • A novel mission reliability oriented RUL prediction method for manufacturing systems is developed.

  • A RUL driven predictive maintenance approach for manufacturing systems is presented.

  • An application example of a serial manufacturing system is implemented.

ABSTRACT

The performance states of the manufacturing equipment and the quality states of the manufactured products are important indicators for the operational state evaluation and maintenance decision of the multi-state system. Further, the performance degradation of manufacturing components shows some dependence on the decline in product quality. However, the traditional remaining useful life (RUL) prediction and maintenance strategy of manufacturing system are limited to the dependence of the manufacturing components performance degradation. Based on the RUL prediction model that considers the components dependence for product quality requirements, a system predictive maintenance method based on the component functional importance is proposed. First, the connotation of degradation mechanism, functional dependence and RUL for manufacturing system is expounded. Second, a mission reliability oriented RUL prediction method for manufacturing systems is developed based on the functional dependence of components. Third, an approach for average maintenance cost calculation is proposed based on dynamic RUL prediction after each maintenance action, and the functional importance is applied to prioritize the predictive maintenance component-sets. Finally, the case results show that the proposed approach can ensure the ability of manufacturing systems to complete production tasks with high quality product, and reduce the maintenance cost in the production cycle simultaneously.

Introduction

Manufacturing systems are the physical carriers in realizing the production, and the dynamic monitoring and RUL prediction of its health states are crucial to manufacturer profit [1]. In intelligent manufacturing, massive on-line monitoring data is applied to perform the system reliability analysis through system degradation modeling, and then the system RUL prediction and maintenance can be effectively optimized [2]. Lei et al. [3] summarized the current researches of mechanical system health prediction systematically, including data acquisition, health index construction, health stage division, and RUL prediction. The current RUL prediction methods can be divided into two types: model-driven methods [4] and data-driven ones [5,6]. Also, degradation-model based RUL prediction of electromechanical products is commonly used in manufacturing engineering [7], among which Markov models [8, 9] are widely used in components degradation state monitoring and RUL prediction. The model of data-driven RUL prediction method is established by using features extracted from monitoring conditions, then it is trained through a large amount of monitoring data to achieve RUL prediction [10]. Si et al. [11] summarized statistical data-driven RUL prediction methods, which are divided into the processes based on directly observable states and the other processes based on states without directly observation.

When developing the degradation/reliability modeling and making maintenance decisions, component dependence has been the research focus for a complex multi-state system composed of multiple components. In previous research, component dependence was usually classified into three ways: structural [12], stochastic [13], and economic dependence [14]. Liu et al. [15] proposed reliability modeling and preventive maintenance of load sharing systems with degrading components. Zhao [16] established reliability modeling of load-sharing systems with continuously degrading components. Considering the maintenance resources shared by multiple components in the system, some scholars have also proposed the concept of resource dependence [17]. However, the functional characteristics of the manufacturing system determines that the influence of components do not only exist in themselves but also in their function output, namely, work-in-process (WIP). Quality-reliability chain (QR-chain) model was established to describe the interaction between the manufacturing components and WIP, the process quality decreased as components degrade and the WIP with dimensional deviation exceeding the limit accelerated the degradation of downstream manufacturing components [18]. The description of the degradation process of WIP in the production process has been extensively studied, among which the study of the stream of variation (SOV) [19] is most in-depth. The SOV model describes the transfer and accumulation of dimensional deviations during manufacturing, possibly effectively reflecting the influence of manufacturing components on the state degradation of WIP [20]. Abellan and Jose [21] applied the SOV model to the manufacture of ceramic floor tiles. He et al. [22] modeled the transmission and accumulation of the assembly variations in the key quality characteristics (KQCs) for multistage assembly processes.

Meanwhile, the performance degradation of manufacturing machine can be represented by the transmission and accumulation of WIP manufacturing deviations during production process from another perspective. On the basis of QR chain, He [19, [23], [24], [25]] has established the product reliability-oriented reliability-quality-reliability chain (RQR-chain) that extended the QR chain, to characterize the transfer relationship of manufacturing system reliability(R), manufacturing process quality(Q) and product reliability(R). The risk formation mechanism of product quality accidents was elaborated based on RQR chain [23], where the key quality negative events (KQNEs) affecting KQCs directly and the non-key quality negative events (NKQNEs) affecting KQCs indirectly under the collective effect of CCF (the common cause failure) & CF (the cascading failure) from the manufacturing process, jointly leaded to accelerated degradation of product KQCs during the usage phase. Further, a reliability loss-oriented product assembly quality risk model is proposed based on the RQR chain [24], which extended the quantification of RPN value by quantifying the undetectable rate of the assembling system, the occurrence possibility of process variations, and the product reliability loss in order to characterize the assembly quality risk. Zhao et al. [25] defined the probability of machine performance degradation to failure as the basic risk, and defined the probability that a manufacturing system cannot complete the number and quality requirements of production tasks in operation as the explicit and implicit risks, respectively. Then, D-S evidence theory was used to integrate the above three fuzzy heterogeneous risk indicators. In a word, these studies provide a solid foundation to expound the functional dependence of multi-state manufacturing systems.

With the development of dynamic RUL prediction technology for manufacturing equipment, research on condition-based maintenance [26] (CBM) and predictive maintenance [27] is also rapidly developing. Chang and Lee [28] derived the wear life distribution with the Markov chain Monte Carlo method in the manufacturing processes of semiconductor wafers and developed an early stage data-based maintenance strategy. Furthermore, an RUL prediction method considering the influence of imperfect maintenance activities on the degradation level and the degradation rate is proposed [29]. In addition, integrated predictive maintenance strategy became a focus that considering the quality of the produced products affected by the machine deterioration [30,31]. Fakher et al. [32] considered the failure type that the system produces a fraction of nonconforming items and analyzed the trade-offs between maintenance, quality and production. Wang et al. [33] developed an integrated model of production, quality, and maintenance to avoid the non-conforming items produced by machines that suffered quality failures and minimized the total cost.

Although previous literature shows that RUL prediction and health management are significant for complex manufacturing systems, few studies have established RUL prediction models and predictive maintenance strategies that consider the functional output of manufacturing systems. To address this problem, it is a prerequisite to establish a dependence model among the manufacturing system, multi-components, and WIP, which is called functional dependence in this paper. Therefore, this paper develops a task-oriented analysis framework for manufacturing system reliability to proactively achieve system-level RUL prediction and predictive maintenance strategy optimization. The main contributions are as follows.

  • (1)

    The definition of hard failure and soft failure of the manufacturing system are put forward, which is the foundation for the establishment of product quality oriented functional dependence among the manufacturing components.

  • (2)

    A novel mission reliability oriented RUL prediction method for manufacturing systems is developed that considers the functional dependence of the manufacturing components and soft failures.

  • (3)

    Aiming at improving the mission reliability of the manufacturing system, the functional importance of manufacturing components is proposed. Based on the dynamic RUL prediction and function importance ranking of the manufacturing components, the multi-component system predictive maintenance strategy is optimized considering the average maintenance cost.

The remainder of the paper is organized as follows: Section 2 introduces the basics of RUL prediction considering functional dependence for a manufacturing system. Section 3 presents an RUL prediction approach with the aid of the mission reliability model for manufacturing systems. Section 4 presents an optimization strategy of predictive maintenance considering RUL for manufacturing systems. Section 5 provides a case study to validate the effectiveness of the presented approach. Section 6 concludes this paper and presents future research directions.

Section snippets

Basics of RUL modeling considering functional dependence for manufacturing system

The premise of RUL prediction is the definition and identification of failure modes. Existing RUL prediction methods focus on the failure mode recognition and performance degradation prediction at the level of manufacturing equipment, which lacks the consideration of the WIP quality decline caused by equipment degradation at the manufacturing system level. Therefore, from the perspective of hard failures at the equipment function level and soft failures at the system function level, failure

RUL prediction approach based on mission reliability of manufacturing system

The mission profile of manufacturing system refers to the completion of the production tasks with specified quantity and quality requirements under the established production environment. Therefore, the mission reliability of manufacturing system can be applied to measure the probability without the hard and soft failures, which is better than the basic reliability that only focuses on the component hard failure. Based on the redefined hard failure and soft failure, the mission reliability

Optimization strategy of predictive maintenance considering RUL for manufacturing system

The maintenance of manufacturing components aims to ensure that the manufacturing system can continuously and steadily complete the required functions. Under the advanced manufacturing mode, dynamic evaluation of components performance states and system RUL is essential to accurate predictive maintenance. From the perspective of maintenance cost, the length of time that the manufacturing system can effectively maintain the functional output is an important evaluation index for maintenance

Background

Many enterprises at present have adopted advanced fault diagnosis technology to monitor the physical signals of manufacturing components, such as machine tool bearings, and predict the RUL of the machines to reduce unexpected downtime. However, for high-precision products that require high quality in the manufacturing process, such as aerospace components and engine components, enterprises need to bear the losses caused by unqualified products. Practice has proved that independent condition

Conclusion

The proposed method provided a novel idea for predictive health management of complex multi-state manufacturing systems. First, the hard failure and the soft failure of manufacturing system were redefined from the perspective that the product quality decline were considered as the functional failure of the manufacturing system, and the connotation of the functional dependence of manufacturing components was put forward. Second, based on the redefined failure mode of the manufacturing system,

Author statement

Xiao Han: Data curation, Writing- Original draft preparation. Zili Wang: Methodology, Supervision. Min Xie: Reviewing and Editing. Yihai He: Conceptualization, Methodology, Supervision. Yao Li: Visualization, Investigation. Wenzhuo Wang: Validation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was supported by a grant from Research Grant Council of Hong Kong under a theme-based project grant (T32-101/15-R) and a GRF (CityU 11203519), the grant of National Natural Science Foundation of China (No. 71971181 & No. 72071007).

References (37)

Cited by (101)

View all citing articles on Scopus
View full text