A leading pulp and fibre manufacturing company was facing issues with maintaining consistent cloth quality due to variations in oil presence in cotton fibre. The company’s current process involved conducting chemical tests in a lab to determine the oil presence unit (OPU) in the cotton fibre. To improve the process and reduce dependency on manual testing, the company decided to implement an AI/ML solution that would use IoT sensors to predict OPU.



The company implemented an IoT solution that used sensors to measure production rate, viscous flow, machine speed, machine dryer speed, multiple dryers’ temperature, soft finish dozing flows, and soft finish circulation. The input to the process was raw material that created synthetic cotton fibre, and the system used IOT data to predict OPU. The AI/ML solution was developed in collaboration with client and the solution was deployed on an EC2 instance. The end-users could access the solution through a Grafana dashboard that displayed OPU predictions at 1-hour intervals, to avoid fluctuations.


IOT data was fetched using API’s from aspen servers in plant every 10 mins and fed into Influx DB (timeseries database) maintained on EC2 Instance. The predictions were generated by the AI algorithm using this sensor data and stored back into Influx DB. The trained models were pushed to a git repository and refreshed in the Docker container for ML in production, instead of creating a fresh docker image every new release.



The virtual sensing project with more than 90% accuracy has provided a reliable and efficient way to predict OPU in cotton fibre, which has reduced the dependency on manual testing and improved the overall cloth quality. The solution has been deployed in plants in two plants in India with plans to expand to international plants soon. The end-users are satisfied with the performance of the solution, and the company plans to adopt AI to measure and improve other quality metrics.