Milacron leverages Artificial Intelligence (AI) for Predictive Machine Operations
Shares Real Results Achieved
Milacron is a global leader in the manufacturing, distribution, and service of sophisticated plastic processing equipment. Last year, Milacron rolled out “M-Powered”, a suite of IoT applications that allows Milacron users to access and monitor key performance indicators and provide remote diagnostics and maintenance services to their users. At the same time, Milacron launched a predictive initiative that uses advanced analytics and Artificial Intelligence to capture value from machine data to improve their overall asset utilization.
“It is important that Milacron customers make money with our machines. So, our mission is to help customers get the maximum utility from the capital that they spend,” says Giovanni Spitale, President, Customer Service & Support at Milacron.
The project, led by Industrial IoT leader ei3 with support from Milacron engineering, targeted areas of injection machines where there are opportunities to predict failures, and where predictions can be used to alert service teams to turn unplanned downtime into planned downtime, thus improving machine availability and asset utilization.
The project followed ei3‘s proven roadmap for putting data science to work on industrial machines to predict outcomes and improve performance. First, the data was reviewed to discover what parts of the process showed a correlation with downtime events. This was done using machines that have been connected to ei3 for years. The large amount of data from this installed base of connected equipment provided the starting point to discovery, by finding points that have “interesting” correlations with equipment downtimes.
- Once these interesting correlations are found, the team created detailed algorithms by exploring the areas of interest using advanced toolchains that apply machine learning and AI methods to identify emerging trends and changes.
- After the algorithms are created, they were then run against production machine data using the ei3 secure private cloud to perform a proof-of-concept. This test deployment demonstrated the real-world applicability of using these algorithms to predict equipment failures. The trial also evaluated the effectiveness of different ways of presenting the prediction information to machine operators and maintenance personnel.
Using this prediction discovery roadmap, the ei3 data science team created prediction algorithms for three common plastics injection machine failures – Heater Bands, Hydraulic Pumps, and Feed Screws. The trial demonstrated that by applying a combination of Industrial IoT and Artificial Intelligence to create heath-scores for each component, it is possible to deliver a robust prediction service that shows the estimated useful lifetime of each part, and more importantly, the presence of an impending part failure.
With this project completed, Milacron is now able to successfully detect when a monitored part shows signs of impending failure. These predictions will lead to new service models where Milacron can take steps to pre-emptively order and deliver the spare part to the customer to avoid the loss of productivity and improve process stability.
Giovanni adds, “some of our customers have delivery penalties where a day of downtime can cost them a million dollars. So, if we can reduce the risk of a pump failure being the root cause for that million-dollar penalty, it creates a perfect win-win situation. It allows Milacron to better position their spare parts inventory and helps the customer operate efficiently.”
This case shows the benefits that come from bringing the latest Artificial Intelligence technologies together with best practices in machine design and engineering. Milacron sees these benefits and is leading the plastics industry’s digital transformation through its collaboration with Industrial IoT leader ei3. The examples described are just the beginning and the opportunities that will be created by Milacron M-Powered will continue to grow.
Access the full presentation: