Verification of value-adding results through applied Machine Learning for industrial processes

Purpose and goal

The project aimed to create a knowledge-giving example for Swedish industry regarding applied Machine Learning (ML) in an industrial application. The project has developed, commissioned and evaluated a test application where Machine Learning is applied to Erasteel AB in Långshyttan to detect deviations in the production process of steel strip and wire according to project plan. Each processed steel bar that passes through the plant is automatically analyzed by the project´s application and deviations are reported.

Expected results and effects

The overall project goals and effects have been achieved by: A ML-application has been developed and implemented in a industrial process for deviation detection. New knowledge and understanding has been detected by project partners about the strengths and weaknesses of the technology and has given rise to new ideas and future applications. The result has been continuously communicated via Automation Region, PiiA and above all through outreach activities to companies in Swedish industry.

Planned approach and implementation

The project has been carried out in five partially parallel phases: Information, Development / adaptation of ML method to fit a industrial process, Implementation of method at Erasteel, Analysis of results, Communication and dissemination av results. All phases have been completed according to plan.