Artificial intelligence expands the coverage of a user’s assets.
by Simon Kampa
September 3, 2019

The world of industry is undergoing a period of profound change. Technology has changed the sector almost beyond recognition over the last 20 years, and the introduction of smart systems that work together to improve efficiency and productivity has led to conversations about the advent of a fourth industrial revolution. Predictive maintenance is the latest example of what is possible in this new era.

Predictive maintenance is by no means a new concept. Its origins lie in the defense and aerospace sectors more than 30 years ago, when new safety rules and regulations made constant assessment of mission critical components an obligatory, although expensive, task. It was a laborious, time-consuming process, where machines and individual components had to be checked thoroughly by expert data analysts in order to spot potential problems before they affected performances.

Advances in artificial intelligence (AI) and big data, plus the growing ubiquity of the Industry 4.0, are advancing this area of endeavor, making an automated form of predictive maintenance available to a far wider range of factory sizes and sectors.

Image 1. Smart factory environmentImage 1. Smart factory environment (Images courtesy of Senseye)

Collecting Data

In the past, gathering data from machines such as pumps, industrial motors and heating, ventilation and air conditioning (HVAC) systems took a substantial amount of time. Data scientists had to track, monitor and analyze data gathered from key components and machine parts manually in order to make meaningful predictions.

The introduction of the Industry 4.0, where connected machines and sensors can gather data and communicate it elsewhere for analysis, was a major step forward in this regard. Instead of relying on people to manually take
readings and collect information from machines, this would happen automatically.

The scalability this provides expands the coverage of an organization’s assets, and instead of looking only at the most critical assets, predictive maintenance analysis could be done organization-wide for the first time. Typically, the number of assets one person can monitor would vary from between 50 to 100 a day. With automated data gathering and analysis, the number of assets monitored can grow into the thousands.

Smart Analytics Predict the Future

As it is now easier to gather large amounts of data at relatively low cost, there has been a rush of organizations tracking vital machine statistics across a range of industries and use cases. But, while around two-thirds of industrial organizations are collecting data, relatively few are using it productively.

The next big step in the data journey, for many, is to find the right applications and software to sort through the data and put it to use.

Predictive maintenance algorithms, for instance, are being used increasingly in industrial settings to compare readings on things such as machine vibration, pressure, temperature, torque and electrical current against known faults to spot emerging problems and predict if and when a machine will fail, months in advance.

AI plays a key role here. Rather than depending on data scientists to complete the laborious process of producing bespoke algorithms for each monitored device, organizations have turned to applying relatively generic condition monitoring algorithms that then adapt themselves to the devices they are tracking. These algorithms then learn the precise characteristics of the machines they are monitoring and fine-tune the performance to become increasingly predictive over time as they are fed more data.

Benefits for All Sectors

This approach allows industrial organizations to start quickly making sense of their data and to provide actionable insights such as an indication of the remaining useful life for each machine to improve how they plan their maintenance activities. The impact on unplanned downtime has proved transformative.

Downtime is one of the biggest expenses for any plant. Manufacturers lose thousands of dollars every minute factories are offline. Knowing when a machine is going to fail, ahead of time, means problems can be addressed before they affect production schedules.

image 2 predictive maintenanceImage 2. Predictive maintenance algorithms are being used in industrial settings.

Rather than bombarding engineers with large, complex data sets, alerts can be set to provide information about matching failure models and the remaining useful life of each piece of machinery. Engineers can look at a simple dashboard each morning to see where efforts would be best applied and when.

Industry 4.0 is a key part of this development. The ability to gather data automatically and use self-learning algorithms has made predictive maintenance much more accessible, freeing up time and resources to be used elsewhere.

It is now possible to automate 90 percent of the analytic tasks performed by diagnostic system engineers, allowing them to introduce large scale predictive maintenance programs to monitor all assets without an increase in labor costs.

Advantages to Predictive Maintenance

The benefits of Industry 4.0 and the kind of predictive maintenance approaches it provides access to are being embraced by leading manufacturers and other industrial organizations around the world. In many sectors, organizations have cut their levels of unplanned downtime in half and reduced maintenance costs by as much as 40 percent. The potential for the future will no doubt be transformative.