By Dr. Simon Kampa, CEO and founder of Senseye
Condition monitoring is by no means a new concept for manufacturers. However, the practice that first began in the aerospace and defense sectors 30 years ago is barely recognizable today. Once, it entailed huge armies of data scientists manually gathering vast quantities of data; experts in analysis would then painstakingly review in order to spot any signs of faults. This labor-intensive process made it an expensive undertaking and inaccessible for most manufacturing organizations.
The rapid development of Industry 4.0—including cloud and machine-learning technologies—has changed this. Smart sensors and machines that can read their own vital statistics have automated the collection of data, while advanced software is able to identify patterns and anomalies in machine health.
By reading machine-data outputs for vibration, temperature and the amount of electrical current drawn, it is now possible to meticulously assess the health of machines and spot emerging problems up to six months before they might affect production. This allows the organization to put in place more precise, more appropriate maintenance schedules.
Rather than create bespoke algorithms for each machine type, factories can start with a series of generic algorithms that can be fine-tuned by machine learning. The algorithms analyze data from machine sensors and cloud platforms to diagnose problems, predict failures and assess the remaining useful life of machinery.
The introduction of automatic trend-recognition algorithms is another example of progress in this space. These algorithms monitor gradual changes in the condition of industrial machinery over long periods of time, while spotting small (but significant) variations in vibration, pressure, temperature, torque, electrical current and other sources that indicate deterioration in machine health.
While traditional approaches to condition monitoring required organizations to recruit data scientists, automation enables manufacturers to achieve the benefits of condition monitoring without having to find these increasingly scarce and expensive professionals.
Organizations that already employ data scientists can now leave mundane day-to-day monitoring to the computer, allowing the human experts to focus their attention on the most complex issues that require more creativity and lateral thinking.
We’re in the early days of machine-learning algorithms. We’re still uncovering all of their capabilities. This much is clear—automating the gathering of data has made condition monitoring much more accessible, economical and available to a wide range of sectors.