What is Smart Data?

Smart-Data is an innovative approach to industrial data issues, in order to guarantee upstream the production of relevant data and downstream an exploitation adapted to the targeted objectives.

Objectives

Smart-Data aims to make the production and use of data related to the technical operations of material transformation more reliable. Indeed, field observations show that a considerable amount of technical data produced is unusable, because it is not representative.

The objectives of the Smart Data approach for processing industries are various:

• determine measurable parameters of interest
• develop and adapt tools and methods to specific industrial challenges
• feed a technical data strategy in “full awareness” of the issues and risks of measurement bias
• make technical and managerial teams aware of these issues

Principle of use

Smart-Data consists first of all of looking at the data produced with a critical eye. It is a question of questioning the relevance of the data produced in relation to the properties, behaviors and industrial phenomena of material transformation.

The means used to produce these data also need to be questioned, not only as is usual from the angle of their precision and/or repeatability, but above all from the angle of their representativeness in relation to the problem(s) faced by the data are intended to constitute rationalization values.

The relevance of upstream data

Measurement standards from quality control and other common laboratory measurement techniques sometimes generate data that are not representative of the actual behavior of the material.

The determination of the measurable parameters of interest by considerations of industrial science phenomena operating in the processes or applications is therefore a key issue in the technical data strategy.

Downstream data mining

The exploitation of representative data makes it possible to envisage effective correlations and to develop predictivity.

However, it often remains necessary to carry out mathematical processing of the data produced to extract the key parameters of the phenomena of interest. At this stage too, phenomenology and the analysis of existing models provide key elements.

Valuable assets

Mastering the representativeness of measured data opens the way to many improvements at different levels of the organization:

• Adapt internal standards
• Develop predictive approaches and thus make the critical stages of industrialization or implementation more reliable
• Establish specifications for raw materials
• Open paths of innovation