Camera system for detecting cracks in corrugated roof tiles on 2 production lines, for a construction materials factory

How the brief came about

The task in this case was to inspect double-corrugated roof tiles for cracks using cameras while the tiles travel on the conveyor rollers of 2 production lines placed side by side. The customer is a construction materials factory producing tiles in several colours, from grey, red and green through to pearlescent.

The tile colours make this task harder than inspecting a single-colour workpiece. Lighting that shows cracks clearly on one tile colour may not suit another, so the system has to adjust the lighting to match the product being made.

Requirements and constraints

The system has to inspect tiles as they travel on the conveyor rollers. The main requirements and site conditions to handle are as follows.

• Inspect each of the 2 production lines placed side by side, with an inspection machine frame installed straddling the conveyor rollers.
• Support 2 sizes of double-corrugated tile: 50x120 cm and 50x150 cm.
• Support 12 tile colours, each of which needs its own sensor tuning and light level setting.
• The project requirements set a target of detecting cracks of approximately 0.5 cm.
• Light from outside the cabinet must not interfere with the tile detection point or the imaging point, because external light strikes the tiles at some points on site.

How does this system inspect tiles of several colours for cracks?

This system inspects tiles of several colours for cracks by using LED lighting with brightness levels set separately for each colour, then having the camera and image processing unit at each point look for cracks in the images. Each line has cameras with image processing units, 4 sets per line for 8 sets in total, and every set connects to the central machine (Center) over the LAN network.

We chose to adjust the lighting rather than use a single light value for every colour, because surfaces of different colours reflect light differently. PWM-controlled LED lighting can be set in fine steps and the same values can be recalled for reuse, so the system was designed to store the light level for each colour for reuse when the production colour changes.

1. 1

   Detects tiles passing through

   A light sensor (opto sensor) detects that a tile is passing through the inspection point.

2. 2

   Light according to tile colour

   The LED lighting uses the brightness level set for the colour in production. The central machine sends light level adjustment commands to the image processing units, with a test record on line B.

3. 3

   Capture images and look for cracks

   The image sensor camera captures images of the tile surface, then the image processing unit checks each image using the crack detection program.

4. 4

   Display results and store reports

   The program screen lets users select the product Model Name, enter the Lot Number and open the Report menu to view inspection results.

Deliverables

We delivered crack inspection sets for 2 production lines, from the machine frame and circuits through to the program and documents. The main items are as follows.

• Inspection machine frame straddling the conveyor rollers of the 2 lines, designed by us along with its circuits and wiring.
• 8 sets of image sensor cameras and image processing units, 4 per line.
• Light sensors (opto sensors) for detecting tiles passing through.
• PWM-controlled LED lighting with settings for 12 tile colours.
• A central machine (Center) connected to every set over the LAN network.
• The crack detection program, plus a screen for selecting the product model, entering the lot number and opening inspection reports.
• Handover documents with manuals, and a process capability (Cpk) report form in the project document set.

Results confirmed by on-site testing

The confirmed results are that we installed the crack inspection sets on each of the 2 production lines and tuned the sensors to 12 tile colours, and that the system passed the equipment check, the test run and testing against the program manual on site in 2017. The work was delivered with documents and manuals, and the project has been closed.

This case has no business performance figures, so we report only results backed by documents. Light adjustment commands from the central machine have a test record on line B, while for the target of detecting cracks of approximately 0.5 cm there is not yet any document showing test results directly against this target. The system in this case detects cracks and displays inspection reports. Removing tiles from the line is not among the confirmed results of this work.

What we learned

This work offers the following principles for camera inspection of workpiece surfaces on lines producing several colours.

• Light settings belong to each product, not to the machine. When one line produces several colours, light levels must be stored separately by colour, and colours added later must be tuned and tested again before use.
• The workpiece detection point needs the same controlled lighting conditions as the imaging point. In this work, light from outside the cabinet striking the tiles prevented the light sensor from detecting tiles at some points.
• An inspection target must come with a test method that measures the same value as the target, and the test must be repeated when the scope of work changes. In this work, the report form in the document set did not directly show results against the crack size target.

How this approach can apply to your factory

If your factory needs to inspect for cracks or defects on the surface of sheet materials travelling on conveyor rollers, you can use the same approach as this case: detect the workpiece, light it according to the product colour, capture images and look for defects with the image processing unit at each point. What has to be adjusted is the camera distance, the light settings and the inspection method when the workpiece or defect differs from this case.

• Compile the full list of product models and colours before starting, using the same names the factory actually uses.
• Define from the outset the defect size to be detected and an acceptance test method that measures the same value.
• Keep external light from reaching both the workpiece detection point and the imaging point.
• Plan for site upkeep, including heat inside the cabinet, dust that needs cleaning and the stability of each camera set's network cable.
• Set user training dates in the delivery plan from the outset.