When it comes to the chemical products that Kaneka produces, improper or incomplete drying of resins will cause quality issues, which is a problem as the resin needs to be dry before it’s sold to manufacturers. Within drying installations, the operator adjusts the operating conditions under several constraints (i.e., temperature, exhaust humidity, feed rate, etc.).
Kaneka faced challenges due to the manual nature of the resin drying process, resulting in waste. Kaneka’s goal was to standardize operations across their production facilities and optimize resources to avoid that waste.
How Dataiku Helped
The team’s objective is to predict the future state of the resin drying process with gradient boosting according to the temperature setpoint. Using Dataiku, real-time information from the equipment is transmitted to the database. The new workflow in Dataiku uses the temperature setting in various parts of the dryer as model input. The ML model then determines an intermediate feed rate and the dryer exhaust humidity.
The following goals are defined to operate in optimum conditions:
- Increase intermediate supply.
- Ensure exhaust humidity meets requirements.
- Minimize heater steam usage.
Through inverse analysis, the algorithm computes the optimum parameters in line with the targets.
With AI prediction and the calculation of optimum conditions, feedback is provided to the equipment, which will update the temperature settings to improve production and steam usage without any manual intervention.
A guidance screen further enables employees in the production department to monitor the current and past conditions to check for inconsistencies.
The Dataiku Difference
By establishing workflows and automation in Dataiku, the Kaneka team has been enabled to automate temperature adjustments in-process and has eliminated the need to assign people to tedious monitoring tasks. Through a Dataiku dashboard, important information can be displayed and monitored. Current temperature control can be seen at a glance, allowing team members to focus on other tasks.
Using Dataiku to improve drying operations has also benefited Kaneka by:
- Improved steam consumption (cost reduction).
- Increased production capacity through constant operation under optimum conditions.
- Standardization and automation have reduced the need for manual operations.
Lasting Benefits
The data analysis project and resulting ML model implementation occurred much faster than previous “from scratch” attempts. Kaneka estimates it took half the labor of a custom in-house solution. Further, graphical representation of the models facilitated communication with field engineers for validation.
Beyond the models utilized for automation, Dataiku also empowered new teams to analyze additional metrics of production. These new insights accelerated momentum across the organization to look for more operational efficiency improvements. Dataiku has a proven and continued positive ROI for Kaneka. As an end-to-end platform, Dataiku is critical to Kaneka’s daily operations and viewed as essential for further efficiency improvements.
