Skip to main content

Data as the enabler of predictive maintenance

(Image credit: Future)

In the age of industry 4.0, companies are increasingly using predictive maintenance to prevent failures and optimise their equipment. This can cut maintenance costs by up to 20 per cent, but predictive maintenance strategies are only ever as good as the intelligence, or data, that is gathered about the state of the equipment that needs to be maintained. Data is indeed the most precious asset in predictive maintenance and its quality can mean the difference between a successful or catastrophic deployment.

Adopting a ‘data oriented’ approach

With the development of the Internet of Things (IoT), companies in the industrial sector can now collect increasingly large volumes of data from a variety of systems. This can be used as part of a continuous improvement process that assesses performance and makes better decisions. Data collection and analysis now make it possible to anticipate problems, optimise the lifespan of company equipment and ultimately improve the customer experience.

By deploying predictive maintenance, companies can use as much data as possible from the equipment in order to optimise it, make it more reliable and provide continuity along the entire product journey, ending in customer delivery. The data collected makes it possible to re-apply the analyses performed at each stage of the product's life cycle to improve and optimise its use, making data the ultimate key to success in predictive maintenance.

Capturing the critical data for future insights

A constant feed of intelligent data is vital for making a continuous stream of assessments of events that could happen in the near future.

It is impossible to conduct predictive maintenance without a reliable feed of high quality, historical data. By aggregating and cross referencing the data that has been gathered from an array of sensors on multiple components, organisations can build up a picture of the state of the entire estate of machinery. By correlating this with historical information, previous parameters and measured behaviours, it is possible for the management system to learn about the system being monitored. This ‘machine learning’ allows the system to understand what’s normal and what isn’t, plus spot potential problems in development.

By predicting events – such as a component wearing out, or knowing when a gasket is about to be blown – the management system can limit damage and ensure continuity. This is only possible if the company can continuously extract and store data. It is essential to know the full history of a machine’s previous behaviours before you can properly make comparisons with today’s events. The capture of data is a crucial process.

One significant challenge is that that not all industrial equipment provides direct or even partial access to the data. Indeed, some of the older equipment might have no sensors at all. Companies must therefore ensure that equipment is properly connected; the installation of appropriate sensors is critical. Older equipment especially should be linked to sensors that monitor their temperature, vibration and movement. Understandably, the older the equipment, the more likely it is to break down, but the less likely it is to have been built with sensors included.

The quality of connectivity is also critical in the capture of data throughout the production process. This is an area ripe for development, with numerous start-ups developing boxes that can be easily added to equipment in order to provide monitoring and connectivity.

The next challenge is to make sense of all this information. Since it will mean different things to different people, everyone must look at it in a different way. All data strategies must be differentiated on the basis of each company’s machines, processes, industry and a number of other variables.

Data processing across the sectors

Every vertical sector has its own unique set of data issues. A weapons manufacturer does not face the same challenges as a car maker. The nuances lie both in the sensitivity of the data and in its relevance to the context.

In addition, the level of maturity of companies in data collection can vary wildly. In the manufacturing sector for example, equipment is at a relatively early stage of digital transition, with more and more sensors being installed thanks to the rise of IoT. In comparison, the aeronautics sector already relies heavily on sensors. Indeed, some airliners are equipped with thousands of sensors, producing large volumes of data that need to be managed, analysed and acted upon.

Finally, data access is a decisive factor in the success of predictive maintenance. This necessity has encouraged industrial companies to group all their data together, with the aim of maintaining reliability and consistency while respecting safety criteria.

The construction of a relevant and effective predictive maintenance model will inevitably depend on how successfully all parties can cooperate. But it is in everyone’s interests to do so, especially at a time when organisations must take every opportunity to optimise the use of machines and equipment to increase productivity and reduce their costs.

Maurizio Canton, VP Customer Success, TIBCO (opens in new tab)

Maurizio Canton is VP Customer Success at TIBCO. He has more than 25 years experience in IT, working for several software vendors, such as TIBCO Software, IBM, Siebel, SOA Software and Red Hat.