As a global player in tire reinforcement, Kordsa reinforces one out of every three automobile tires and two out of every three aircraft tires worldwide. In such manufacturing operations, even small inefficiencies add up to large costs. One of the most significant such costs in this industry is that which results from unplanned stops associated with electronic failures. Such failures can lead up to 20% loss in up-time of equipment lines as well as substantial raw material and quality losses which happen before and during these stops.

In principle, if such failures could be predicted, they could be prevented. However, two main challenges need to be considered: 1. predictions need to be made several hours in advance of the failure event, to give crews enough time to perform preventative actions 2. the root cause of the future failure also needs to be identified, together with the prediction. These challenges, as well as the complex, non-linear relationships between sensor data and observed outcomes put the goal out of reach of existing machine learning solutions. This is where Kaizen's proven deep learning-based solution promised to be a perfect fit.

We integrated our solution on top of the existing sensor data infrastructure, which records a continuous stream of temperature, pressure, and electric current values from all equipment. After our robust data cleaning process removed sensor malfunctions and anomalies, our model was able to predict more than 70% of faults with an average of 6-8 hours advance notice, which gives the company ample time to plan its maintenance operation to prevent these stops. Furthermore, we produce diagnostic insights, which helps ground crews pinpoint the root cause of the issue and take on-point precautions to prevent occurrence of failures.