Reduction of shape defectsAchievement of skill transferYield improvementProductivity improvement

Data Science

DL-ASC: Deep Learning for Automatic Shape Control System

Departures from dependence on the individual abilities of rolling operators in shape control operation

Solution Point

Automatically selects and learns only operation that conforms to the shape and quality of the target product from operation by operators.
As a result, automatically selects excellent operation, and utilizes AI to reproduce the operation of excellent operators.
Can be added to the existing control system by retrofitting.

Features 01

Expected Problems

・Is shape control unstable in multi-high single-stand mills?

・Is yield poor because it's impossible to improve shape defects?

・Are there variations in shape control accuracy due to individual differences in the skills of operators?

Features 02

Technology highlights

Point01　Learns only excellent operations from past data.

Point02　As a result, the operations of excellent operators are learned automatically.

Point03　Total system renovation is not necessary because functions can be retrofitted on existing systems.

Features 03

Proposed Solutions (Examples)

・Addition of functions to existing multi-high single stand mills.

・Improvement of accuracy by application of a learning model that incorporates the know-how of JFE Steel operators.

Links 01