AX AI Process Advance Framework

Once extracted thousands of lines of existing data and mapped them to process flows, we created a cyber-physical model, able to predict and optimize the production process and calculate correct set points for operations in real-time.

YOUR CHALLENGE

  • You are operating a complex production process
  • The output is dependent of multiple, complex correlated inputs and boundaries
  • Limitations in output due to (semi) manual adjustment procedures

OUR SOLUTION

  • Investigate and define metering concept for critical input parameters
  • Train dedicated AI models that correlate and predict input to output KPIs
  • (Semi) Automate process adjustment procedures based on AI output

Emission Cost Cutting

CEMENT

Once extracted thousands of lines of existing data and mapped them to process flows, we created a cyber-physical model, able to predict and optimize the cement production process and calculate correct set points for operations in real-time.

Automatic Steel Strip Positioning Error Correction

STEEL

Installation of high-speed, high-resolution cameras to record steel strips during rolling in mill to provide input for subsequent AI based image analysis, classification of their position and dynamic derivation of the positioning error.

Management of Billet Temperature

STEEL

Installation of quality inspection system to collect quality information and relevant boundary conditions as input for AI. Training of AI model to correlate and predict output quality and derive recommendations and set points for operators in real time.

Break Sensitivity Indicator

PUL&PAPER

Development of a web break sensitivity indicator using basic principles of case-based reasoning with a linguistic equations approach and basic fuzzy logic able to combine on-line measurement data with expert knowledge.

Fuzzy Logic Control of Thermo-Mechanical Pulping

PULP&PAPER

Dedicated sensors and controllers observe the fiber length and smoothly begin to adjust the freeness set point if the fiber length develops outside specified boundary parameters. A second controller keeps freeness at a desired level by changing the set point of the total specified energy.

High Quality Output Based on Automated Inspection

STEEL / PULP & PAPER

Implementation of AI based Convolutional Neural Network model, for supervised identification and classification of defects in steel strips. It correlates defect classes with production parameters and boundary conditions for automated recommendation derivation.