Home Solutions Industries Platform About Us Our Team Insights Innovation Lab Contact

Manufacturing Optimization using Data Analytics

In brief: How a premium instant coffee manufacturer used data analytics and predictive models across 32 process parameters to control moisture and quality.

Business Challenge

A premium instant coffee manufacturer faced quality control issues with moisture content. International standards allow maximum 5% moisture, but customer specifications ranged between 2-4%. The intricate manufacturing process required maintaining 32 process parameters in a highly controlled environment, with temperature being a critical control variable.

Current Problem

Temperature settings were adjusted based on operator experience, leading to significant inconsistencies. Requirements underwent a reduction through the supply chain — the "Bull-Whip Effect" — causing excessive product drying and reduced yield.

Analytical Approach

The team examined existing IIOT sensor data and identified that operators were setting targets substantially below requirements. For instance, targeting 1.8% moisture when customers needed 2.0%.

Solution Methodology

  • Data Collection: Captured 32 parameters including ten temperature readings and vacuum pressure from the SCADA system per second.
  • Data Preparation: Developed specialized mapping software to track batch data throughout production.
  • Exploratory Data Analysis: Conducted pattern analysis identifying critical controllable variables affecting moisture.
  • Model Development: Applied Decision Tree algorithms and Multivariate Polynomial Regression, ultimately reducing requirements from 17 to 3 variables.

Recommendations

  • Only two temperatures require adjustment for any target moisture level
  • Provided lookup tables for temperature settings across moisture ranges (1.5%-5.5%)
  • Documented impact rates for every 0.5-degree temperature changes

Results

Implementation streamlined operations and reduced output variation. The solution delivered a benefit of up to USD 800,000 per annum per production line and was extended across multiple facilities.

Related solution: Manufacturing — predictive operations and quality intelligence →

← Previous Article
Optimizing Community Account Activities with Appian and ChatGPT

Ready to let data drive
your manufacturing decisions? ↗

Ready to ignite your progress?

Stand out with Codenatives