At SKF AI, we want to make AI-Driven Industrial Analytics and its benefits to process accessible to all manufacturers, no matter how many data scientists they have on staff. To that end, our SKF Enlight AI maintenance 4.0 solution is built on Automated Machine Learning (AutoML). Freddy Hernandez, Director Connected Technologies at SKF explains.
Unscheduled downtime impacts manufacturers of all sizes across every industry. Though there are different estimations of how unscheduled downtime affects production on a yearly basis, the International Society of Automation estimates that factories lose between 5-20% of overall annual production time due to these unforeseen setbacks.
The actual cost of unscheduled downtime is much greater than the minutes or hours a machine is unable to produce its allotted quota. Among the invisible expenses that manufacturers absorb are labor and overtime for plant workers and technicians, costs associated with the ordering of new machine parts and the transportation of this equipment, and damaged production output from machines that manufacture lower quality materials before breaking down.
As factory owners increasingly look to Big Data, Artificial Intelligence and Machine Learning to revolutionize processes and efficiency on the plant floor, one Industrial IoT (IIoT) technology that has been heralded as a pioneering maintenance solution is AI-Driven Industrial Analytics.
Many factory assets are already fitted with sensors that capture Big Data, such as vibrations and temperature, to monitor their wellbeing in real-time. AI-Driven Industrial Analytics harness this data to build complex Machine Learning models that can anticipate machine downtime. The advance warning empowers technicians to schedule maintenance, order parts, and minimize the labor and output losses that unscheduled downtime necessarily entails.
The promise of AI-Driven Industrial Analytics – increased uptime, reduced O&M costs – has the potential to be a manufacturing gamechanger. Nevertheless, according to a 2018 McKinsey report, only 30% of all IIoT pilots are successfully scaled across the entire organization. In the case of AI-Driven Industrial Analytics, two formidable obstacles standing in the way of adoption are the global shortage in data scientists and a difficulty in scaling any solution across not one but dozens or even hundreds of factories.
Machine Learning is a difficult science that requires a high level of discipline expertise, and many AI-Driven Industrial Analytics solutions require a team of data scientists to keep the system running smoothly. In each step of the process, data scientists are responsible for configuring the Machine Learning’s models. There is no preprogrammed guidebook for these labor-intensive decisions regarding aspects such as model selection and configuration or hyperparameter optimization.
At SKF AI, we want to make AI-Driven Industrial Analytics and its benefits to process accessible to all manufacturers, no matter how many data scientists they have on staff. To that end, our Enlight AI maintenance 4.0 solution is built on Automated Machine Learning (AutoML).
As its name suggests, AutoML automates aspects of Machine Learning by having algorithms, not data scientists, make model selection, validation and other configurations which were previously performed by human data scientists.
This approach enables us to reduce many of the repetitive tasks performed by data scientists, improve the performance and accuracy of the Machine Learning models, and empower non-data scientists to make use of the insights generated. Using Correlative Pattern Recognition, algorithms analyze all the anomalous data to find broader hidden patterns. Once a deviation is detected, the system sends out an alert to plant technicians, who are then able to schedule a planned downtime, order parts and book repairs – all in advance.
Technicians using our solution don’t need to become “citizen data scientists” to reap of the benefits of Machine Learning insights. Instead, an intuitive user interface provides information on the impending machine breakdown and empowers technicians to respond in real time using their existing skillsets.
Ever smarter AI is empowering us to offer a digital transformation product that combines advanced Machine Learning analysis with expert knowledge.
One benefit of this hybrid relationship between experts and AI-Driven Industrial Analytics is its impact on wrench time. Wrench time is the time that maintenance workers are actively engaged on the job with “tool-in-hand.” A substantial amount of wrench time is made up of waiting: waiting to arrive at the inoperative machine, waiting for parts, waiting for additional backup. Enlight AI significantly reduces the wait time by providing in-depth insights into two core maintenance issues, Root Cause Analysis and Time-to-Failure.
Though maintenance and reliability personnel may be skilled at repairing machines, they aren’t always capable of diagnosing why a machine has faltered. In automated AI-Driven Industrial Analytics solutions, the algorithm discerns deviant sensor behaviors or correlations of these behaviors. Identifying the abnormal sensor activity provides insight into the Root Cause Analysis.
Tracing the breakdown to the behavior of a specific sensor in a piece of equipment helps identify the precise repair activity that is required for remediation. From the technician’s perspective, less time is wasted on guesswork and testing, the machine can be repaired more quickly, and the technician can return to his or her scheduled tasks faster.
Using Big Data for AI-Driven Industrial Analytics also provides earlier indications of deterioration or impending asset failure. By identifying Time-to-Failure in advance, non-wrench time activities can be reduced or eliminated because of upfront planning. Rather than waiting for parts or extra support to arrive, repair workers’ time can be allocated to other routine maintenance tasks. When technicians can follow a set schedule, instead of being randomly paged to malfunctioning machines, the need for overtime can be significantly reduced.
As manufacturers ponder the concept of the Smart Factory and what that might mean for them, SKF is pioneering a tangible solution for the here and now. Maintenance continues to dominate Industry 4.0 conversations, and our expansion into the maintenance services sector aims to provide new value to our customers.
In a time in which hacking, and data breaches have dominated the news, we’re using secure and encoded channels to stream manufacturers’ data to our cloud. While it may have been a non-issue a decade ago, data ownership has become a pressing issue in its own right.
Maintenance 4.0 is a critical component of a manufacturer’s digitization journey, and the introduction of AI-Driven Industrial Analytics within the factory has the potential to significantly improve the employee experience, machine longevity, production quality and output, revenue and savings. Enlight AI provides a customizable maintenance solution for increased uptime and reduced operational costs. We look forward to helping manufacturers implement this predictive companion in and across factories.