IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. XX, NO. XX, NOV. 2019 1 A Survey of Predictive Maintenance: Systems, Purposes and Approaches Yongyi Ran∗, Xin Zhou∗, Pengfeng Lin†, Yonggang Wen∗, Ruilong Deng ‡ Nanyang Technological University ∗†, Zhejiang University † ∗{yyran, zhouxin, ygwen}@ntu.edu.sg, †{linp0010}@e.ntu.edu.sg, ‡{dengruilong}@zju.edu.cn Abstract—This paper provides a comprehensive literature re- to prevent unexpected outages, improve overall reliability and view on Predictive Maintenance (PdM) with emphasis on system reduce operating costs. architectures, purposes and approaches. In industry, any outages The evolution of modern techniques (e.g., Internet of things, and unplanned downtime of machines or systems would degrade or interrupt a company’s core business, potentially resulting in sensing technology, artificial intelligence, etc.) reflects a tran- significant penalties and unmeasurable reputation loss. Existing sition of maintenance strategies from Reactive Maintenance traditional maintenance approaches suffer from some assump- (RM) to Preventive Maintenance (PM) to Predictive Mainte- tions and limits, such as high prevent/repair costs, inadequate nance (PdM). RM is only executed to restore the operating or inaccurate mathematical degradation processes and manual state of the equipment after failure occurs, and thus tends to feature extraction. With the trend of smart manufacturing and development of Internet of Things (IoT), data mining and cause serious lag and results in high reactive repair costs. PM Artificial Intelligence (AI), etc., PdM is proposed as a novel type is carried out according to a planned schedule based on time or of maintenance paradigm to perform maintenances only after process iterations to prevent breakdown, and thus may perform the analytical models predict certain failures or degradations. unnecessary maintenance and result in high prevention costs. In this survey, we first provide a high-level view of the PdM In order to achieve the best trade-off between the two, PdM system architectures including the Open System Architecture for Condition Based Monitoring (OSA-CBM), cloud-enhanced is performed based on an online estimate of the “health” and PdM system and PdM 4.0, etc. Then, we make clear the can achieve timely pre-failure interventions. PdM allows the specific maintenance purposes/objectives, which mainly comprise maintenance frequency to be as low as possible to prevent cost minimization, availability/reliability maximization and multi- unplanned RM, without incurring costs associated with doing objective optimization. Furthermore, we provide a review of too much PM. the existing approaches for fault diagnosis and prognosis in PdM systems that include three major subcategories: knowledge The concept of PdM has existed for many years, but based, traditional Machine Learning (ML) based and DL based only recently emerging technologies become both seemingly approaches. We make a brief review on the knowledge based and capable and inexpensive enough to make PdM widely ac- traditional ML based approaches applied in diverse industrial cessible [4]. PdM typically involves condition monitoring, systems or components with a complete list of references, while fault diagnosis, fault prognosis, and maintenance plans [5]. providing a comprehensive review of DL based approaches. Finally, important future research directions are introduced. The enabling technologies have the enhanced potential to detect, isolate, and identify the precursor and incipient faults Index Terms —Predictive maintenance, fault diagnosis, fault of machinery equipment and components, monitor and predict prognosis, machine learning, deep learning the progression of faults, and provide decision-support or automation to develop maintenance schedules. Specifically, the I. INTRODUCTION emerging technologies enhance PdM in the following aspects: arXiv:1912.07383v1 [eess.SP] 12 Dec 2019 Maintenance as a crucial activity in industry, with its signifi- 1) IoT for data acquisition: IoT enables gathering a huge and cant impact on costs and reliability, is immensely influential to increasing amount of data from multiple sensors installed a company’s ability to be competitive in low price, high quality on machines or components [6]. and performance. Any unplanned downtime of machinery 2) Big data techniques for data (pre-)processing: Big data equipment or devices would degrade or interrupt a company’s techniques have been revolutionizing intelligent mainte- core business, potentially resulting in significant penalties and nance by turning the big machinery data into actionable unmeasurable reputation loss. For instance, Amazon experi- information, e.g., data cleaning and transforming, feature enced just 49 minutes of downtime, which cost the company extraction and fusion, etc. $4 million in lost sales in 2013. On average, organizations lose 3) Advanced Deep Learning (DL) methods for fault diag- $138,000 per hour due to data centre downtime according to a nosis and prognosis: In recent years, more and more DL market study by the Ponemon Institute [1]. It is also reported approaches are invented and getting matured in terms that the Operation and Maintenance (O&M) costs for offshore of classification and regression. The larger number of wind turbines account for 20% to 35% of the total revenues of layers and neurons in an DL network allow the abstrac- the generated electricity [2] and maintenance expenditure in tion of complex problems and enable more accuracy of oil and gas industry costs ranging from 15% to 70% of total fault diagnosis and prognosis (e.g., remaining useful life production cost [3]. Therefore, it is critical for companies to prediction). At the same time, the huge amount of data develop a well-implemented and efficient maintenance strategy is able to offset the complexity increase behind DL and IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. XX, NO. XX, NOV. 2019 2 improve its generalization capability. systems, including four categories of DL architecture: auto- 4) Deep Reinforcement Learning (DRL) for decision mak- encoder, Deep Belief Network (DBN), CNN and Recurrent ing: The breakthrough of DRL and its variants provide a Neural Network (RNN). Most efforts of this survey are promising technique for effective control in complicated aimed at fault identification and classification other than fault systems. DRL is able to deal with highly dynamic time- prognostics. Khan et al. [11] present a simple architecture variant environments with a sophisticated state space of system health management and review the applications of (such as AlphaGo [7]), which can be leveraged to provide auto-encoder, CNN and RNN in system health management. In decision support for a PdM system. addition, a series of survey papers focus on the fault diagnosis 5) Powerful hardwares for complex computing: With the for a specific type of components or equipment, e.g., bearing rapid development of semiconductor technology, the pow- [12, 13], rotating machinery [14], building systems [15], wind erful hardwares, such as graphics processing unit (GPU) turbines [16]. In [12], Zhang et al. systematically summarize and tensor processing units (TPU), can significantly ex- the existing literature employing machine learning (ML) and pedite the evolution process and reduce the required time data mining techniques for bearing fault diagnosis. Liu et al. of DL algorithms. For example, Sun et al. [8] achieve a [14] provide a comprehensive review of AI algorithms in rotat- 95-epoch training of ImageNet/AlexNet on 512 GPUs in ing machinery fault diagnosis from the perspectives of theories 1.5 minutes. and industrial applications. There also exist several survey Although PdM becomes a promising approach to decrease papers that focus on fault prognosis. Remadna et al. [17] the downtime of machines, improve overall reliability of sys- present a generic Prognostic and Health Management (PHM) tems, and reduce operating costs, the high complexity, automa- architecture and the applocations of DL in fault prognostics. tion and flexibility of modern industrial systems bring new The involved DL approaches in this survey only comprise challenges. Specifically, three fundamental problems should CNN, DBN and auto-encoder. Lei et al. [18] deliver a review be well considered in the context of PdM: of machinery prognostics following four processes of the prognostic program, namely data acquisition, HI construction, 1) PdM system architectures: With the advent of Industry HS division and Remaining Useful Life (RUL) prediction. The 4.0, a variety of techniques have been involved in indus- model-based and data-driven approaches for RUL prediction trial systems, e.g., advanced sensing techniques, cloud are summarized in this survey paper. computing, etc. In order to design efficient, accurate The aforementioned survey papers have given interesting and universal maintenance systems by embracing these reviews related to the filed of fault diagnosis and prognosis, emerging techniques, PdM systems should: a) be com- but they have the following limitations: 1) Most of the existing patible with various industrial standards, b) be easy to survey papers [10–18] only focus on reviewing the existing integrate with the emerging or future techniques, and fault diagnosis and/or prognosis approaches, most of which are c) satisfy the basic requirements of PdM, e.g., data equipment specific. There is no clear way provided to select, collecting, fault diagnosis and prognosis, etc. design or implement a holistic