In several manufacturing fields, there is the need for the classification of raw material surface in order to reach the expected quality level of the end product. This task depends more on subjective factors linked to the experience and skills of the worker than on well defined and measurable parameters. The increase of automation in manufacturing processes raises the demand for quick answers in the evaluation of quality, to avoid process slow-downs. As a consequence, human operators are asked to work in very stressing conditions causing a high rate of mistakes.
The goal of this project is the development of an intelligent system performing the following functions :
- on and off-line surface inspection
- on and off-line quality classification
- quality assessment of produced materials following classification rates.
The main theme is represented by the application of neural network and fuzzy logic techniques to evaluate data provided by conventional inspection devicec such as a vision system. The research will concentrate on the development of a common classification methodology independent from inspected materials.
The system will be demonstrated in three different fields :
- in a plywood factory, where it will be applied to the process of grading veneers. The automatic and intelligent classification will enable productivity to increase by 10% and allow material treatments to be optimized according to the stated quality.
- in a textile factory, to state the quality of surface finish in velvet production. The system will inspect a portion of the fabric strip during the fibre raising phase, in order to classify parameters that are suitable in further developments to steer the global finishing process.
- in a steel industry, in three different ways, following the main steps in the use of steel sheet, from production to application in the forming process. Reducing waste of faulty cold mill rolls from 25% down to 10% per production unit should be achievable by proper analysis and classification of surface roughness, avoiding faulty steel sheet production of hundreds of tons.
The intended research will be performed through the following steps :
- analysis of measuring conditions and selection of inspection devices;
- definition of parametric models and classification methodology;
- development of the logical interface;
- definition of the system architecture;
- software and hardware implementation;
- modules integration and prototype installation.
Funding SchemeCSC - Cost-sharing contracts
48100 Mungia (Bizkaia)
CF2 1XH Cardiff