Maintenance of enterprise assets is one of the most important activities for every industrial organization. It safeguards the good health of the various assets as a means of ensuring continuity and efficiency in production operations, while at the same time avoiding costly downtimes. It is also an activity that affects the enterprise budget, as Maintenance and Repair Operations (MRO) represent a significant line of the balance sheet. Therefore, enterprises are constantly in quest for innovative solutions that optimize the cost and the effectiveness of their maintenance operations.
Predictive Maintenance: A Killer Application for Industry 4.0
In principle, assets are serviced and repaired according to one of the following maintenance modalities:
- Reactive maintenance: In this maintenance modality repairs take place when the asset (e.g., the equipment) has already broken down. Following the repair operation, the equipment is restored to its normal condition and can support the production processes. This reactive approach is sometimes called “breakdown maintenance” as it is associated with equipment breakdowns.
- Preventive maintenance: It refers to maintenance that is performed at regular intervals as a means of preventing equipment failures and unexpected downtimes. Preventive maintenance is driven by insights about the nominal expected lifetime of the asset, as a means of servicing it before it breaks down.
- Predictive Maintenance: This is based on the monitoring of the performance and the condition of assets, towards identifying the best time to maintain it before it breaks down. It therefore leverages information about the actual state of the asset, rather than some theoretical life expectancy value provided by the OEM (Original Equipment Manufacturing). Predictive maintenance (PdM) is sometimes called Condition-Based Maintenance (CBM) as it relies on the actual condition of the asset in order to reduce the likelihood of failures.
As evident from the above definitions, preventive maintenance offers significant benefits over reactive maintenance, as it provides a good basis for avoiding costly production stops and unscheduled MRO operations. However, preventive maintenance leads still to a sub-optimal utilization of assets, as it occurs well in advance before the asset’s breakdown. Predictive maintenance can alleviate this limitation of preventive maintenance and provides a basis for optimal asset utilization and that increased OEE (Overall Equipment Efficiency), when compared to other maintenance models.
Despite these known benefits of predictive maintenance, most organizations dispose with preventive maintenance for the majority of their assets. This is largely because there are still no easy ways for monitoring the condition of the assets in a credible way, but also for predicting their future state as reflected in parameters like the Remaining Useful Life (RUL) and the End of Life (EoL) of the asset.
During the last couple of years, the accelerated adoption of digital technologies in the shopfloor and the advent of the fourth industrial revolution (Industry 4.0) are disrupting the enterprise maintenance landscape through enabling more organizations to develop, deploy and operate credible predictive maintenance solutions. This is largely due to the availability or more digital data about the assets, but also due to the ability of enterprises to analyse them in faster and more effective ways. In particular, thanks to the deployment of cyber physical systems and internet connected devices in the shopfloor, enterprises can now collect a wealth of digital data about the condition of their assets, such as data from vibration, acoustic, ultrasonic, temperature, energy consumption, oil analysis and thermal imaging sensors. This data can be combined with other maintenance related datasets such as product quality information and equipment master data that are available in business information systems like ERP (Enterprise Resource Planning) and Asset Management systems. The analysis of these datasets can produce predictive insights for the condition of the assets, including credible information about their RUL and EoL parameters. This analysis can nowadays be performed faster than ever before, due to the abundance of low-cost computing resources. The latter enable the timely execution of predictive analytics and artificial intelligence algorithms, which provide insights on the status of the asset, along with recommendations for optimizing maintenance scheduling and maximizing OEE.
Predictive maintenance can be seen as the “killer” application of Industry 4.0 for two main reasons. First, it can be applied to the majority of enterprise assets, for which digital data can be collected and analysed. Second, it’s an application of interest to all industrial sectors, including for example manufacturing, energy, oil & gas, mining and construction, as well as for public infrastructure projects. Hence, it is up to date one of the most popular Industry 4.0 applications for enterprise organizations.
Anatomy of Digital Predictive Maintenance Infrastructures and Applications
The typical elements of a predictive maintenance application in the Industry 4.0 era are:
- A maintenance data collection infrastructure: This infrastructure is in charge of collecting data about the various assets, including for example data from sensors and cyber physical systems in the shopfloor. Data can be collected through either wired or wireless sensors, depending also on the available industrial networking infrastructure. A data collection infrastructure for predictive maintenance should be able to collect and bring together different types of data, such as sensor data streams with high ingestion rates, static data residing in business information systems and databases, as well as data from the equipment itself.
- A data storage and management infrastructure: Predictive maintenance solutions rely on the analysis of very high volumes of maintenance related data, from a variety of different sources and sometimes of high velocity (e.g., streaming data) i.e. what we typically call Big Data. It’s therefore important to ensure the proper storage and management of such Big Data in data storage infrastructures such as historian databases, data warehouses and data lakes.
- Predictive analytics algorithms: A PdM system entails the processing of maintenance datasets towards deriving predictive insights about the status of the assets. Such insights require the deployment of predictive analytics algorithms, which fall in the realm of machine learning (ML) systems. In many cases Deep Learning (DL) and Artificial Intelligence techniques are also employed, given the need to identify complex degradation or defect patterns for the assets.
- Visualization solutions: The predictive insights about the condition of the assets need to be properly visualized and presented to maintenance professionals and service technicians. To this end, PdM deployments comprise advanced visualization solution that presents information about the health of the assets in ergonomic and user-friendly ways. In their simplest form, visualization solutions are comprised by one or more dashboards. Nevertheless, more advanced solutions like Augmented Reality visualizations are also possible.
Based on these elements it is possible to develop and deploy solutions that present maintenance engineers, service technicians and other end users with timely and detailed information about the current and the future health condition of industrial assets. This information can be used to support maintenance related processes and workflows such as Failure Mode, Effects & Criticality Analysis (FMECA) and Reliability Centered maintenance (RCM). For example, PdM enables the extraction of quantitative information about the failure probability of the assets, which is a cornerstone for FMECA analysis. Based on quantitative probabilities, maintenance professionals can present credible estimations regarding potential failures, as well as their impact on industrial operations. Likewise, PdM can be used to extract hidden patterns of degradation of the assets in a way that unveils the root causes of the failures, which is another important element of FMECA processes. Furthermore, it’s also possible to use predictive insights in order to optimize maintenance programs as part of an RCM discipline. The latter optimization can be based on the identification of the best point in time for maintaining or repairing an asset.
Enabling Advanced Maintenance Functionalities
Predictive maintenance can be combined with other Industry 4.0 applications in order to provide greater efficiencies and OEE optimizations. Here are some characteristic examples:
- Automatically Scheduling Maintenance Tasks: Predictive insights about the assets (e.g., RUL and EoL) can be combined with information for production orders and schedules in order to decide the optimal dates for performing the maintenance. In essence, this optimization takes into account constraints stemming from the practical use of the assets in order to achieve the best possible OEE.
- Integration with Digital Automation: In addition to presenting insights to maintenance engineers and service technicians, predictive maintenance applications can be integrated with automation functionalities in order to instantly influence production in ways that optimize the condition of the asset. For example, in cases where early signs of a machine’s degradation are identified, the rate of its operation could be decreased as a means of prolonging its EoL.
- Executing What-if Simulations about the Condition of the Assets: Maintenance professionals can greatly benefit from digital simulations about the future behaviour of the assets. In particular, it is possible to develop digital twins of the assets, as a means of simulating what-if scenarios about the use and the maintenance of equipment.
- Leveraging Augmented Reality for Remote Maintenance: Predictive maintenance can be also combined with Augmented Reality visualizations that indicate to maintenance professionals which maintenance and service actions to perform on the asset. In this way, it is possible for OEMs to support their customers from remote i.e. through providing remote on-line instructions to the workers, which saves precious time and travel resources.
These functionalities are generally advanced and currently in their infancy. However, they will be increasingly deployed, as industrial organization advance their digital transformation and level of Industry 4.0 adoption.
The Main Challenges of PdM Deployments
Despite the emergence and rise of Industry 4.0, developing, integrating and deploying an effective PdM solution remains challenging and requires a well-balanced team of solution vendors, system integration and experienced consultants. This is largely due to the fact that there is a need for confronting the following challenges:
- Fragmented Datasets: PdM is mainly about collecting, consolidating and analysing large amounts of maintenance related datasets. Nevertheless, such datasets are in more cases fragmented in different systems within an enterprise (e.g., ERP, MRP, Quality Management systems) and/or stem from a wide variety of sensors and devices. Therefore, the task of bringing together data from so many different sources, platforms and formats, remains extremely challenging. To alleviate the challenges there is a need for constructing a proper data management and data integration plan, while at the same time developing a scalable Big Data infrastructure that addresses the need of the organization.
- Talent Gap in Big Data & AI: Nowadays, there is a globally proclaimed gap in Big Data analytics and AI, which makes it very difficult to create a truly competent data science team to deal with predictive maintenance projects. Thus, it is still challenging to analyse data towards deriving predictive insights for the condition of assets.
- AI Bias and Explainability: Data driven predictive maintenance can in principle provide insights about the root causes of assets’ failures, which can help enterprises improve their knowledge about the efficiency and the shortcomings of their maintenance practices. This however requires transparent AI algorithms that operate in a way understandable to humans rather than in a “black-box” function. Devising and validating explainable and transparent AI algorithms is one of the main challenges for the future of predictive maintenance. Likewise, there is a need for devising properly functioning algorithms that do not suffer from bias. The latter is a matter of having appropriate datasets, as well as of experts with domain knowledge in the data science team. This is also challenging as such domain experts are hard to get.
- Limited Integration with Automation and Lack of Closed Loop Processes: Early predictive maintenance systems emphasize on presenting predictive insights to maintenance professionals, as a means of driving their maintenance actions and decisions. In the future, predictive insights will be used to drive the operation of automation systems with much less human intervention. In this way, they will enable closed-looped processes that interact with the field. Currently automation and maintenance operate in isolation and their integration remain challenging as a result of the lack of interoperability between predictive maintenance systems and digital automation platforms.
Existing technologies provide the means for confronting these challenges. Nevertheless, exploiting available technologies requires a competent consulting team with strong knowledge and skills in industrial maintenance and in the design and implementation of Industry 4.0 solutions.
SAP’s Solution for Predictive Maintenance
As part of its portfolio of Intelligent Asset Management solutions, SAP provides a collection of powerful tools for building and delivering Predictive Maintenance solutions. In particular, SAP’s Predictive Maintenance and Service solutions provide the means for integrating operations and analytics on a single platform. The solutions are empowered by the unique in-memory computing technology of the SAP HANA platform, which provides flexibility in integrating and using business and operational data, from structured, unstructured and semi-structured data sources. The solutions are also empowered by an integrated service centre, which centralizes data management and enables the creation of innovative and holistic maintenance strategies.
SAP’s Predictive Maintenance and Service solutions deliver benefits to various stakeholders, including:
- Owners and operators of equipment, who can analyse the operational trends of their assets in order to minimize unplanned downtimes and gain knowledge that helps them streamline maintenance operations with business goals.
- Service providers that can improve the speed and quality of the services that they provide to asset owners, while at the same time exploiting opportunities for value added services like Remote Maintenance based on Augmented Reality.
- Asset manufacturers (including OEMs) that can benefit from data-driven insights that can help them improve the design of their assets, lower warranty costs and create innovative business models such as Maintenance as a Service (MaaS).
SAP’s Predictive maintenance solutions are provided in two pricing models, including:
- Cloud-based pricing: This is based on a utility-based model that is charged based on the volume of devices and data storage that comprises your maintenance solutions.
- On-premise pricing: This model is based on a perpetual license model in-line with the client’s specific requirements for data storage.
Conclusion & Future Outlook
At the dawn of Industry 4.0 enterprises are provided with opportunities for transforming their maintenance operations from reactive and preventive maintenance to predictive maintenance. This transformation is set to exploit the proclaimed benefits of predictive maintenance towards maximizing equipment uptime, improving service performance levels and ultimately optimizing OEE. It is also set to enable enterprises to anticipate equipment malfunctions and service degradation, in order to apply proactive remedies that alleviate or even prevent the effects of such malfunctions.
SAP’s solutions offer unique functionalities in terms of an enterprise’s ability to monitor assets in real-time and running predictive analytics about the future condition of assets and their impact on business operations. Moreover, SAP provides flexibility in the packaging, delivering and pricing of these solutions based on both licensed and pay-as-you-go models. Nevertheless, a competent and experienced consulting team is a key prerequisite for getting the most out of your SAP Predictive Maintenance deployment. Gentackle facilitates enterprises to develop, deploy and fully leverage predictive maintenance solutions in ways that maximizing business benefits and the returns on your PdM investments.