According to a report by maintenance specialist Senseye, companies in the manufacturing sector experience an average of 27 hours of downtime per month due to system failures, which results in revenue losses in the multi-million euro range every year.

When it comes to maintenance strategies that allow manufacturers to act before equipment components fail, the terms preventive and predictive maintenance are often used interchangeably. Both methods are superior to reactive maintenance, which keeps equipment running until emergency repairs are required, often ultimately costing companies four to five times as much as proactive maintenance options.

However, preventive and predictive maintenance are not the same thing. Preventive maintenance involves checks at regular intervals that take place regardless of the condition of the equipment. This maintenance strategy relies on best practice guidelines and historical data to give plant managers the best chance of keeping their machines in good working order, but it also requires planned downtime at regular intervals. According to the Operations and Maintenance Practices Guide, predictive maintenance is estimated to save companies around 12 to 18% in costs.

In contrast, predictive maintenance, which relies on real-time data from IIoT-connected devices to identify potential threats before problems occur, only takes action when needed. In this way, repairs are focused on actual problems and are more targeted, meaning that downtime, when necessary, is reduced by 25-30% compared to other maintenance methods.

Storage of large amounts of data

In order to function efficiently, predictive maintenance requires data from sensors that provide information about the condition of the equipment. However, IBM estimates that around 90% of all data collected by sensors remains unused. This means that manufacturers are missing out on opportunities to make well-informed decisions about their equipment while paying expensive money to collect and store data. Data that is collected but not processed or used is known as dark data and is a major challenge for the industry. While sensors to collect data are relatively inexpensive and easy to set up, the real challenge is to process the data to draw conclusions about the condition of the machines.

Data processing can be difficult for many reasons, from understanding the data to transferring it to the relevant department. Data silos occur, for example, when data is processed and relevant patterns are discovered but not shared between the different departments of an organization. This may be because companies do not have the technology required for data transparency. For example, some companies may not have a single integrated data management (IDM) tool or computerized maintenance management system (CMMS), and different teams may be working with different platforms.

Sensors to collect equipment information are the first step in integrating predictive techniques into maintenance strategy. Newer machines typically offer various options for real-time data collection, but legacy equipment can also be retrofitted with low-cost sensors. In fact, predictive maintenance can be an important resource in dealing with aging equipment that requires careful planning to procure obsolete spare parts. However, the depressing amount of dark data in the industry, combined with the problem of data silos, shows that collecting data alone is not enough.

To effectively predict the failure of equipment components, manufacturers should implement technologies that facilitate real-time data processing and give the responsible employees access to the resulting findings.

To eliminate problems with data exchange between different departments, companies should put the merging of information technology (IT) and operational technology (OT) at the top of their list of priorities. It was once logical to run and manage these two areas separately due to their different capabilities and competencies, but due to the increasing digitalization of manufacturing processes, including maintenance, it is now necessary to bring them together.

The OT area collects raw data from PLCs, sensors and other important devices, and the IT area makes sense of this data by recognizing relevant patterns. However, communication and collaboration must be right, both in terms of equipment and within and between the teams.

One of the problems with data from industrial plants is that it becomes less relevant and less accurate as it ages. Edge computing minimizes delays and supports real-time decision making by analyzing data as close as possible to where it is collected.

Data processing at the edge of the network

This data processing "at the edge of the network" also offers support in linking even more equipment to the IIoT - and therefore benefits in terms of cyber security. This is because when data is transferred to and from the cloud, it increases the risk of it being compromised. However, this does not mean that the processing or storage of data in the cloud should be avoided at all costs, but simply that both options should be used in a well-coordinated manner to optimize the results.

Predictive and preventive maintenance will always prove to be more effective than reactive maintenance, despite the existing problems with dark data and the need to change the corporate culture in manufacturing plants. And to be successful in an increasingly digital world, manufacturers need to reduce downtime, cut costs and increase efficiency.

Neil Ballinger, Head of the EMEA region, EU Automation