A light on the tiles

A novel automated system based on machine vision, which can be effectively used in tile sorting to increase the quality and the value of production, is changing the tile making process in Europe.

Quality control in ceramic tile manufacturing is hard, labour intensive work and is performed in a harsh industrial environment with noise, extreme temperature and humidity. This control can be divided into colour analysis, surface defect detection and verification of dimensions. Outline is a product that addresses defect detection and verification of dimensions.

The reduction of space needed to accumulate tiles that are manufactured 24 hours per day but are sorted only during two shifts is one benefit of the system. Because the system works all day, buffer stocks are not required and fixed costs for storing and moving tiles are reduced.

Another benefit of the system is the repeatability of results. This is because the automated system always uses the same criteria, does not get tired and it is not influenced by environmental conditions. This stability allows a higher value for batches and increases customer satisfaction.

The continuous measurement of surface and dimension defects gives information on the manufacturing process condition and helps kiln operators to optimise temperature profile, speed and other operating parameters.

The preliminary evaluation of benefits shows that Outline may pay for itself in less than one year.

The quality and value of a batch of tiles are affected by the presence of surface defects (scrapes, broken corners, homogeneity, etc) and differences in dimension, shape and flatness. Problems with dimension include differences between width and length and defects in shape which are manifested as 'cushion' or 'trapezium' effects.

International bodies and tile manufacturers have defined standards for defects and for each tile format. They have also specified procedures to be used to verify compliance with these standards during production processes. The main problem is the detection of surface defects and the measurement of dimensions in an effective and repeatable way on the production line.

According to the manufacturer, Outline demonstrates that an automated system based on vision can be effectively used in tile sorting to increase the quality and the value of production. The system aims to replace the judgement of human operators in defect detection by an automated system.

Checking surface defects online is a difficult task because of environment and process characteristics. Tiles are moved from the oven to sorting locations, at speeds ranging from 40 to 140 pieces per minute depending on size, and operators have a very limited time to make decisions.

Humans can only work effectively on this job for short periods and many different operators are needed to check the same batch of tiles. Continuity over time is not guaranteed and may result in overall poor quality, which may cause customers to complain or even to reject the batch.

Dimensions and flatness are normally checked online with automated systems but the accuracy of systems in current widespread use is not sufficient to allow secure certification of the tiles.

This is due to the structure of these devices based on an array of cameras to verify width and length, which are adapted manually to the size of the tile being manufactured. Tiles are then sampled from the line and then examined in the laboratories with certified instruments that guarantee the required precision.

The system uses linear and matrix cameras, together with dedicated lighting solutions to enhance imaging and facilitate defect detection. The equipment can be installed on existing manufacturing lines without the need for modifications and can manage tiles with different dimensions (eg, 10 x 10 , 30 x 30, 40 x 40 cm, etc.). Measurement precision, sorting speed and accuracy are adequate for the most demanding standards and the characteristics of production lines.

The software is portable and can run on standard PC or Pentium-based multiprocessor systems or event scaleable, 16 to 512 elements SIMD (single instruction multiple data) parallel processing subsystems.

According to the software developers, the eventual goal of Outline is a correct classification that is repetitive and with the same level of accuracy over the time which can guarantee better production yield and avoid most acceptance problems with customers.