Slate Inspection

Key details

  • Product quality and cost efficiency are of pivotal importance

  • Camera and laser automation of a labour-intensive task (real-time quality control)

  • Seamlessly integrated into the existing production process

  • Sorting of in/out-of-specification slate

  • OMS assisted in the process of efficiently manufacturing slate in the UK

  • Allowing our customer to remain competitive in a global market which is good for the UK economy and better from an ecological standpoint.

Why?

More than 2 million houses in the UK have a slate roof. A large number of these buildings were roofed before 1919 and need replacing every 30 years. Slate is a multi-million business and product quality is key alongside efficiency. Any company producing items every few seconds which have to be in specification is likely to need robotics. A robot will do what it is told such as remove defective items. As a result most automated systems will require a degree of intelligence based upon sensors to approve or reject products. In this case the item to be inspected was Slate used in roofing. The slate, being a natural material, can be of differing thickness, width, length. It may also be chipped and have missing edges or other damages. Any human can easily identify such a defect but cannot realistically be expected to do this every second and interface to a robot intent on getting the products out of the factory door.

Slate production.jpg

Solution

This is where a complex machine vision and non-contact sensing system comes in. Shadow imagery and laser distance measurement provide the sensing necessary to be able to do real-time quality control.

How?

Lasers are part of the solution. A typical short range laser distance measuring sensor can measure a distance very accurately several thousand times per second. Cameras are similarly capable of obtaining a snapshot of an object far faster than a human can blink. The combination of these two sensors with some appropriate lighting form the basis for a system that can check a slate in less than a second and communicate a reject to the robotic handling system.

For this application the thickness of the slate had to be measured in several places. This was achieved by temporarily resting the slate on three supports and mounting the lasers above the support. Three supports will always ensure that the slate was held by each point so with a calibration the lasers were able to determine whether the slate was within acceptable thickness bounds and the average thickness. As a natural product that is cleaved apart quite a bit of variation is possible. Too thick and the slate will not easily mate with another slate leaving gaps between on a roof. Too thin and the slate may fail early or could also not marry up as well as required.

The camera used was capable of checking several aspects of the shape of the slate. The width of the slate and its length had to be within certain boundaries and two types of slate had to be sorted on the fly depending on the length. In addition any missing areas of slate in any location had to be identified, measured and rejected if out-of-specification. The following images illustrate both situations.

Out of specification

The system had to be linked to the production process (see below) such that in the fraction of time that the slate was in position an image was taken of the slate silhouetted against the bright background the image could be collected. The image processing then had to quickly remove any extraneous and unimportant features and compute the dimensions of the slate and detect any areas where the slate was damaged. Pass/fail information had to then be linked to the other dimensional measurement data in order to enable the control of the sorting process.

Production setting

The outcome of the project was a system that assisted in the process of efficiently manufacturing slate in the UK to enable a locally produced natural product to remain competitive in a global market which is good for the UK economy and better from an ecological standpoint.