It is part of the Technica Group Steel 9 : Factory-wide control, social and environmental issues.
The goal of our project is to develop new methodologies and tools to help plants to improve the quality of their products and reduce their manufacturing costs by focusing primarily on three quality criteria: appearance surface, internal health and mechanical properties.
These tools should allow:
- Optimize the manufacturing process by identifying the main causes of bad quality
- Predict the quality of the product as soon as possible to better characterize it and reduce the cost
To achieve this goal and to make a major breakthrough in the application of data mining approach in the steel industry, we propose to contribute to new research areas recently developed in the field of data mining.
These new approaches are designed to extract knowledge from huge amount of complex data: sensorial time series of very large number of parameters (several hundred) registered for a substantial period of time (2-3 years) and a high frequency (1-10Hz). Indeed, only summary information (e.g.; casting speed average) was used for statistical analysis. To analyse automatically and massively these sensorial time series data, we propose a comprehensive solution built around five main axes
Develop an analytical server to speed up models construction, optimize their managements and improve exchanges between process experts and data mining experts
Descriptive and predictive analysis for the identification of the causes of non-quality and a better prediction
Develop algorithms for constructing advanced indicators to better represent process phenomena that may affect the quality