Client Story
Machine Learning assisted paint recipe correction system to achieve right-first-time color shades
Share
Business Problem
DataTheta is working with a leading solvent and water-based paint manufacturer. The quantity of tinter or a coloring agent dictates the color intensity of paints. A marginal excess (or less) of tinting elements will spoil the batch or commands an expensive rework. The optimal addition of a tinter is considered an art, accomplished only by the trained operators. Achieving the paint shades the right first time (without any rework) with assistive technology is the only foolproof solution that brings economic value to the table.
Data Solution
The historical tinter addition data from the various mixing units are collected and stored. Data validity checking procedures were applied on 30 input variables such as tinter quantity, mixer speed, etc. The data discovery procedures enabled the identification of the pertinent variables that have a “cause-and-effect” relationship with the final shade of paint. A mathematical model was built using an indigenous algorithm that assists the operator to recommend the precise quantity of various tinters to be added to achieve the preferred paint shade.
Implementation
The data from various sources (from ERP and machines) are orchestrated to the data warehouse using data pipelines. Strict data validity measures are implemented to ensure the cleanliness and integrity of the data. The pre-trained models are deployed in the SAP Data Intelligence cloud and the model versions are updated and stored automatically. The models are field tested, and the coherence of the models is evaluated before deployment. Production grade models and their persistence and consistency for various boundary conditions and process limits are monitored and compensated continuously.
KPI
Recipe Optimization
Duration
14 Weeks
End user
Plant Operator
Business Group
Manufacturing Operations
Technology Stack
Value creation
This Machine Learning based assistive technology enabled the business to remove the trial and error approach in the recipe standardization process to maintain the shade of the paints. In this business setting, the extra or rework means more resource consumption that challenges sustainability and economic focus. This system has removed the human guesswork (or empirical tuning) of the process for the recipe adjustment to software-based standardization.
Ready to get started?
From global engineering and IT departments to solo data analysts, DataTheta has solutions for every team.