The role of data-driven predictive maintenance in industry 4.0

5 mins read

By Pablo Rios, Business Development Manager (Energy, Utilities and Manufacturing), Keepler Data Tech

Manufacturing, as with almost every industrial sector in the economy, is up against an extremely challenging economic backdrop. 

From the pandemic fallout and its huge impact on the functioning of supply chains to intense geopolitical tensions which are exacerbating issues and driving up the cost of powering manufacturers' day to day operations, there doesn’t appear to be much respite on the horizon. 

Firms are having to make difficult decisions. Faced with little choice but to optimise performance and reduce cost bases wherever possible, important modernisation agendas, including sustainability and net zero strategies, risk being side-lined as businesses remain in survival mode. 

Added to this, the UK Purchasing Managers’ Index fell to a 29-month low of 46.2 in October 2022, with new orders falling at the fastest rate since May 2020 when the economy was in almost total lockdown. As a result, manufacturing firms are reporting job losses for the first time since December 2020 as they bid to get a grip of their finances. 

One of the most important cost bases that manufacturers are looking to make efficiencies in is maintenance. 

Downtime is becoming a bigger problem – according to Senseye’s True Cost Of Downtime 2022 Report, large manufacturing and industrial firms lost £1.3 trillion from unplanned downtime in 2020, some 50% more than it did in the 2019-20 reporting period. An hour of downtime in the automotive sector, for instance, now sets companies back an eye-watering £1.7 million. 

Many manufacturers have turned to preventative maintenance as a means of stemming the tide, yet recent years have uncovered a series of limitations which is holding firms back. 

The first is a lack of knowledge of assets. When carrying out maintenance on a regular basis for the sake of it, the opportunity to perform more acute, condition-based monitoring is lost. 

Although these maintenance schedules are planned, emergency remedial work is often required due to a relative lack of awareness on the status of the equipment days or weeks in advance. 

These limitations have a severe impact on productivity, both in terms of quantity due to interrupted production caused by machine downtime, and quality due to one or more machines not functioning properly. Manufacturers also face increased operational costs due to not performing maintenance in a timely manner based on the analysis of condition data. 


Enter data-driven predictive maintenance  

Fortunately, there is a better way of reducing the stifling costs of downtime in the form of predictive maintenance. 

Put simply, predictive maintenance is a proactive technological method that tracks equipment performance in real time or near real-time and predicts machine failure – this means manufacturers can fix equipment before it causes any downtime. 

Powered by IoT sensors and advanced analytics, predictive maintenance constantly analyses the conditions of equipment during normal operation, gathering and processing enormous amounts of data to present a constant picture of health of the machine in question. When irregularities are detected, owners of the machinery are immediately given the real-time information they need to take remedial action. 

Indeed, this form of dynamic monitoring and the detection of potential faults before they happen is widely regarded as an essential component of the industry 4.0 revolution that continues to transform the sector. According to research from McKinsey, the industry could benefit from savings of up to $630 billion per year in 2025 if IoT-based predictive maintenance was deployed more effectively. 

Manufacturers are also benefiting from improved yields, reduced repair costs and extended asset lifespans - McKinsey also stating that capital investment on equipment could be reduced by as much as 3-5% due to more effective maintenance regimes. 

It is therefore no surprise that the market for predictive maintenance is set to soar through the course of this decade. Valued at around $6.4 billion in 2021, by 2030 it is forecast to expand by more than 10 times to $67.2 billion, growing at a CAGR of almost 30%.  


Challenges on the road to implementation

Although the business case for implementing predictive maintenance practices is becoming increasingly difficult to ignore, the process of doing so is not straightforward. 

Based on our experience in supporting customers with these projects, manufacturing firms need to consider a range of factors before and during implementation, including some of the more common issues that they may be faced with. 

Let’s start at the beginning of an implementation. Firstly, data about issues may not exist because they have not been reported, or the data history is not substantial enough to have a descriptive sample of issues. Meanwhile, assets which are slowly degraded already could make it more challenging to identify anomalies over time, and external conditions can also affect correct identification. 

Firms also need to consider how data is going to be communicated between their assets and receiving systems (such as public cloud) which can throw up issues that require some kind of offline data storage mechanism, management of reconnections and data forwarding. All of this, of course, also needs to be carried out in line with regulatory frameworks that cover the security and confidentiality of the data established by the industry.

Arguably the most significant factor to consider, however, is the fact that predictive models should not serve as a replacement for human beings. Human expertise is essential to getting the most out of predictive maintenance strategies – they are key to understanding the process, designing and modelling the algorithms that enable predictions to be made, and offering more information and insight to maintenance teams to underpin their decision making. 


Three top tips to emerge from the storm

None of these challenges are insurmountable, and those which approach the implementation of predictive maintenance in the right way will avoid many if not all of them. 

We’ve helped a huge range of manufacturing firms along their journeys, and while each case is unique, there are a few tips we can share that are widely applicable. 

First up, think big but start small. Depict a business case for a pilot or proof-of-concept with few assets (maybe even one). This will allow you to analyse the quality of data sensors to make sure you can run a bigger project later. In doing so, manufacturers will be in a win-win position – the pilot will either demonstrate value or, if it is not successful, firms will fail fast and thus reduce waste in the investment.

We also recommend making use of cloud computing services to speed up the project. The public cloud will provide you with agility through high-level managed services (PaaS), and the cost of the investment will be shorter than running on-premises. 

Finally, consider leveraging the support of an external partner. A good partner with the appropriate technical capabilities will be able to demonstrate a track record of real experience in productive environments, as well as a high degree of specialisation in the use and optimisation of cloud services to work with data securely. In addition, the right third-party support will also enable manufacturers to overcome any of the aforementioned sticking points along the journey, should they arise. 

With further economic uncertainty ahead of us in 2023, the manufacturing sector needs to get creative to make the efficiencies required to emerge from the storm intact. 

Operational efficiency and reducing the damage caused by downtime will be essential to this end. With predictive maintenance, firms stand every chance of making the sort of gains they need to both navigate the tough period ahead and future proof their businesses for the long term.