Common problems with predictive maintenance

man in factory checking machine data

Identifying the four most common problems in delivering an effective predictive maintenance strategy, and a number of solutions to overcome these for optimum performance.

There’s no question that predictive maintenance is shaking up the asset management landscape. A far cry from the traditional maintenance strategies we have become accustomed to in recent decades, predictive maintenance leverages artificial intelligence (AI), the cloud, and powerful analytics, to drive real-time insights into asset performance.

Whereas early Computerised Maintenance Management Systems (CMMS) channelled a reactive approach to maintenance, addressing issues once an error or breakdown had already occurred, later preventative models brought more structure through adopting routine approaches to checking equipment. However, neither delivered the level of reliability and confidence necessary to optimise production and performance to a standard commensurate with the demands of modern markets.

According to McKinsey & Company, AI-based predictive maintenance can boost asset availability by up to 20% while reducing inspection costs by 25% and annual maintenance fees by up to 10%. Testament to its potential is a predicted market value of $31,965.49 million by 2027, up from $4,331.56 million in 2019.

A silver bullet?

Despite its potential, predictive maintenance is by no means a panacea. As with any new model, implementing it effectively involves overcoming a number of barriers, the most common of which we will explore below:


1. Predictive Maintenance Data Challenges

For the majority of organisations, there are two main data challenges when considering a predictive maintenance platform. The first is identification of the key indicators for each asset, and the second is collecting data in a structured, consistent way, in order that AI systems can use it effectively.

Selection of data collection sources very much depends on the nature of the asset in question. While sensors are hugely beneficial, they should be considered within a framework of other means, as deploying sensors on every asset can be costly and impractical. A conveyor system for example, might require thermal imaging to assess motor temperature, while a compressor might rely on acoustic sensors to identify potential air leaks. Given the scale and diverse nature of assets in a typical production plant, it’s incredibly important to strike the right balance.  

Once established, it’s crucial to collect the data in the right way. The best PdM platforms leverage AI to transform raw data into actionable insights, which in turn, lead to informed decisions. Through interpreting data using algorithms which are continuously refined, meaningful, real-time insights can be delivered in a click.


2. Technology Skill Gaps

New technology typically requires new skills, and when it comes to those which rely on data science and AI, skills are neither cheap nor easily available. Some of the leading PdM providers in the space are addressing this issue by building data science and AI into cloud-based platforms in order that they can be deployed at scale. Not only does such an approach eliminate the need for costly skills, it reduces the risk of employees leaving the company and taking valuable knowledge with them.

A further benefit is that the embedded AI can identify the specific sensors on a machine that are experiencing issues so that remedial action can be extended straight away. Often, the associated cost impact will be estimated at the same time. Because predictive maintenance replicates the experience and intuition of a team of engineers, manual checks are minimised, so health and safety risk is reduced.

3. Cost of Predictive Maintenance

One of the biggest initial obstacles to adopting predictive maintenance is cost. Early iterations of the technology demanded large levels of up-front investment, and in parallel, a need for expensive skills to operate.

The most important consideration when addressing this “problem” is value. Businesses need to look beyond and toward the substantial savings that PdM can bring, as return on investment (ROI) for predictive maintenance is incredibly high. According to a U.S. Department of Energy study, a well-implemented PdM program can reduce maintenance costs by as much as 30%. Also, thanks to improved asset utilization, it is estimated to have a return on investment (ROI) of 10 times or more. 

Cloud maturity has also helped to mitigate up-front costs, as more and more cloud-based predictive maintenance platforms such as Senseye PdM, have become available. This in turn has extended ROI, and in fact, Senseye’s ROI Lock calculator typically finds that Senseye PdM reduces unplanned downtime by up to 50%, with productivity boosts of 55% and an increase in maintenance accuracy of 85%.


4. PdM Security

Security is perhaps the most prevalent concern for organisations when implementing predictive maintenance as part of an Industrial Internet of Things (IIOT) strategy. Recent ransomware attacks for example have targeted manufacturing facilities, which is unusual as this wasn’t really heard of prior to the prevalence of IIoT. This is largely because associated systems open up processes which would previously have been confined to the knowledge base of an engineering team, and therefore hidden from the outside world.

Thankfully, the new breed of cloud-based PdM platforms are built on the highest security standards and typically represent a lower risk option than developing systems in-house where resources are potentially spread more thinly across disciplines. At Senseye for example, data is encrypted at rest and in transport in line with industry-recommended standards. It is regularly tested and audited by independent security bodies, while internal policies mandate strict governance around data handling, use of equipment and software development to ensure confidential data is handled appropriately.

High reward

Given the vast payback potential of predictive maintenance, it’s perhaps not surprising that deployments involve a number of hurdles. Challenges such as data management, skills, investment and security are of course common to many digitalisation projects. The key to overcoming these is undoubtedly to ensure C-level sponsorship, and embrace projects with a realistic, informed mindset, while working with a best-in-class provider who can help deliver the capabilities to achieve your maintenance goals.

For those who get the strategy right, the rewards, in the form of greater reliability, production and confidence, are there for the taking.

Find more detail on where PdM came from and why vendors - and some buyers - aren’t getting it right in our white paper “Senseye in Depth: Why is Predictive Maintenance so hard?