Predictive maintenance in industry: using IIoT and ML to prevent equipment failure


Sooner or later, all machines break down, but with a wide range of consequences. A sudden failure of the coffee maker can ruin your mood and your morning. Consequently, an unexpected malfunction in a power plant has the potential to leave thousands of people in total darkness for hours, and cause a loss of several million Euros.

For example, the average cost of unplanned downtime in the energy, manufacturing, transport and other sectors is €250,000 per hour, or €2 million per working day. 

Indeed, to avoid costly breakdowns and mitigate the damage caused by breakdowns, companies need an effective maintenance policy. This article presents the strategies available, the advantages of the most advanced predictive approach and the resources required to implement it.

Maintenance strategy: corrective vs. preventive vs. predictive

First of all, there are three main types of maintenance strategy to which a company can adhere: corrective, preventive and predictive. Each option has its advantages and disadvantages, so let's take a closer look.

maintenance strategy

Reactive maintenance: solve the problem when it happens

Major benefits: low maintenance costs, reduced permanent staff, minimal planning required

Main disadvantages: high repair costs, safety risks, potentially greater damage to machines

Thus, reactive maintenance, also known as corrective maintenance, from operation to failure, means that actions are taken when the equipment is already down. This approach saves time and money on planning and support services. It can be applied to redundant, easy-to-repair, non-critical assets. Let's say light bulbs are replaced only after they've burned out.

While corrective maintenance requires no up-front costs, it proves very costly in the long term, taking into account overtime, reduced asset life, reputational damage and safety risks. According to Marshall Institute estimates, the reactive approach costs companies up to 5 times more than proactive types of maintenance.

Preventive maintenance: repair everything on schedule

Major benefits: increased efficiency and equipment lifecycle, reduced likelihood of breakdowns, cost savings

Major drawbacks: no way to rule out catastrophic failures, increased work intensity and planned downtime, extra time on planning

Thus, preventive maintenance triggers regular inspections of equipment to mitigate deterioration and reduce the likelihood of breakdowns. Planned activities such as lubrication or filter changes extend the useful life of assets and increase their efficiency. All this translates into savings. Studies show that average savings from planned maintenance amount to 12 to 18% compared with reactive maintenance.

However, preventive measures cannot entirely exclude the possibility of catastrophic breakdowns. What's more, this practice requires additional planning and human resources. Often, the frequency of checks is higher or lower than necessary to guarantee reliability.

Predictive maintenance (PdM): don't fix what isn't about to break

Major benefits: reduced maintenance time and costs, longer asset life, reduced safety, environmental and quality risks

Main disadvantages: the need for organizational changes, major investments in hardware, software, expertise and staff training


Indeed, predictive maintenance has become possible with the advent of Industry 4.0, the fourth industrial revolution driven by automation, machine learning, real-time data and interconnectivity. More or less similar to preventive maintenance, PdM is a proactive approach to machine maintenance. The difference is that a company plans activities based on constant condition monitoring. Once unhealthy trends have been identified, damaged parts are repaired or replaced to avoid more costly breakdowns.

The benefits of PdM for companies include :

-Reduced maintenance costs,

-Extended equipment life,

-Reduced downtime,

-Increased production capacity and enhanced security. According to a Deloitte Insights report, "PdM promises".

-Reduction in maintenance planning time from 20 to 50%,

-Increase in equipment uptime and availability from 10 to 20%, and

5-10% reduction in overall maintenance costs.


But these improvements require significant investment in IT infrastructure and expertise - namely, in industrial IoT (IIOT) sensors, analytics software with machine-learning capabilities; the services of data scientists and IT specialists, and staff training. A company needs to build an entire ecosystem to support prevention activities. 

In the following sections, we'll look at when these efforts make sense, and what exactly it takes to implement LOP.

Predictive maintenance use cases in industry and success factors

In the first place, this cost- and technology-intensive strategy is justified by high-value, mission-critical equipment that must always be operational. Obviously, PdM is too expensive and inefficient for components that can be down for hours or even days without affecting the production cycle. Anything in between requires further deliberation to make the right choice.

Currently, the most efficient PdM is used in the following industrial sectors:

-manufacturing plants,

power plants,



-oil and gas industry, and

-logistics and transport.

Tasks you can solve with predictive maintenance

Whatever your industry, when you decide to implement it, you need to clearly understand that PdM only applies to tasks of a predictable nature. The PdM strategy can answer five main questions:


  • -Probability of a failure within a given period of time.
  • -Remaining useful life of the asset. In other words, how long will the machine run before it breaks down?
  • -Probable cause of a given problem.
  • -The asset with the greatest risk of breakdown.
  • -Which activities will solve the problem most effectively?

Key factors for PdM

The PdM strategy rests on several major pillars, and will simply not work if they are absent or insufficient.

-Dons. The more data you have, the more informed decisions your maintenance staff can make. To generate accurate forecasts, you need to collect and process real-time data from sensors; historical maintenance and breakdown records, equipment metadata and even external information, such as weather conditions.

Competence. In addition to training staff to understand PdM processes and work with new equipment, you need additional technical expertise. The effectiveness of predictive maintenance depends largely on the following specialists:

          ++software and cloud engineers to integrate all the parts of a predictive maintenance IT puzzle into an end-to-end solution and orchestrate their work;

          ++data experts to prepare data, select, tune and train machine learning models, and interpret results; and

          ++reliability engineers to use the results provided by data scientists to improve equipment efficiency and safety.

So, given the complexity of the IT infrastructure required to run PoM activities, you may also need to involve an enterprise architect. This expert will assess your current systems; consult you on available technologies; and help you address key software and integration challenges.

-IT infrastructure. PdM uses several hardware and software modules, as well as cloud technologies. All the components of the IT system are essential for predictive maintenance activities, so we'll look at them in more detail below.

Ecosystem or computerized predictive maintenance technologies that drive PdM

Firstly, predictive maintenance involves a constant flow of data from physical assets; analyzing real-time information against historical records, predicting outcomes and; mitigating or preventing potential failures and downtime. thus, the smooth workflow is made possible by the orchestrated working of multiple systems and software solutions.

IIoT system

IIoT devices or smart sensors that monitor the condition of equipment are at the heart of predictive maintenance. These integrated or external hardware elements capture physical parameters and translate them into digital signals.

Parameters used in predictive maintenance include, but are not limited to:

  • -vibration,
  • -current and voltage sensors,
  • -lubricant quality,
  • -liquid levels,
  • -Temperature,
  • -pressure,
  • -sound levels and frequency, and
  • -chemical content.


In effect, sensors transmit signals to data storage via an IoT gateway; a physical device or software program acting as a bridge between hardware and cloud installations. It pre-processes and filters IIoT data, reducing its quantity before sending it to the data center. In addition, the gateway ensures connectivity; enhances security; and enables translation between different messaging protocols.

Computer-aided maintenance management (CMMS)

A CMMS system is another important piece of software behind PdM. It enables control and analysis of all maintenance-related information, such as repair schedules; breakdown history, spare parts usage and maintenance activities; as well as equipment specifications and technical requirements. In this way, historical data accumulated over the years creates a solid basis for accurate forecasts.

Central data storage

Firstly, you need large, scalable storage to aggregate both real-time sensor data and historical CMMS data.

Indeed, it should be noted that very few companies have sufficient resources to store sensor information in on-site data centers. So cloud solutions, namely IoT and IIoT middleware platforms, are a better choice for collecting and storing large amounts of data. You can easily increase and reduce their capacity according to the volume of data transferred and the number of sensors connected.

Analytical solution with machine learning capabilities

It's not enough to collect massive data sets from different sources. The predictive maintenance strategy requires powerful analytical tools based on machine learning algorithms. Generating predictions with ML involves several phases.

  • -An analysis engine determines the equipment's normal state of health on the basis of historical data (learning phase).
  • Data-driven models are applied to continuously monitor the equipment's health index.
  • -Once the system identifies signs of wear, it warns of problems so that technicians can take action before a breakdown occurs.

Sometimes, a predictive maintenance module is an integral part of the CMMS, but this is not always the case. In addition, this module can be difficult to customize and lack machine learning capabilities, resulting in less accurate prognosis.

Ecosystem or computerized predictive maintenance technologies that drive PdM

As analysis technologies become more powerful and sensors cheaper, predictive maintenance is gaining in popularity and use. To understand the value that PdM can bring to your business without painful financial cost; So, you can start with a "proof of concept", applying the strategy to a machine or product line. The process will involve several steps.

  • -Identify a critical piece of equipment.
  • -Define the parameters to be monitored and install the appropriate sensors.
  • -Engage an external team of data science experts to collect data and create machine learning models capable of extracting meaningful information.
  • -Run the proof of concept for a few months to evaluate predictions against existing maintenance processes.
  • -Estimate annual savings.

Indeed, if implementing PdM proves to reduce costs and improve efficiency, roll it out progressively across the entire enterprise, calling on SYRAM's experts to choose the right solutions, create customized software components and help with integrations.

Author: NAJI Faouzi

Follow our thoughts on social networks

We frequently publish on social networks (LinkedinTwitter and Medium) our innovations and the new functionalities of our industrial management solutions.

We would be happy to share with you the latest trends in industrial management 4.0 through high quality content that you can share.