Using AI-based quality inspection for ceramic tiles to catch tricky visual defects
Modern tile production lines run at 200+ pieces per minute. At that speed, manual inspection becomes a bottleneck—and defects slip through. For tile manufacturers, every defective tile that ships is a customer complaint waiting to happen. At Enao Vision, our AI-based quality inspection has scanned over 70 million items in just over half a year for our customers, including tile manufacturers. Here’s how AI-based machine vision systems can be integrated into the production line to improve processes and reduce waste.
Why 100% inspection on the line matters
An ongoing problem with manual inspection for ceramic tiles is that defects can happen to part of the run. Kiln temperature fluctuation over a 10-minute window can cause defects before stabilizing. A glaze applicator clog or a contamination event can affect a few tiles before clearing up.
It’s too expensive and inefficient to have 100% manual checks. This is where automated quality inspections come in. They can consistently perform past 8 hour shifts and flag defects for human operators to focus on. With AI-based solutions today, machine inspections have become more affordable and easy to integrate into the production line than ever. It’s not an either-or between automated and manual inspection. We believe the two should work together.
Catching surface defects on ceramic tiles
Surface defects are the most obvious quality failures to catch before shipping: pinholes, blistering, and crawling, to name a few. The challenge with manual or spot inspection is that by the time they’re caught, the run might be done.
With automated quality inspection integrated into production lines, the defect patterns will show up almost immediately. Moreover, it supports human operators to check root causes: issues with the firing process or body composition, trapped gas during molding, contamination or glaze adhesion problems. Catching them is only half the job—classifying them correctly enables process correction. This is now easy for human operators to do by adding labels and descriptions, as well as drawing boxes, on interfaces that are as familiar as the daily apps people use.
Before even selecting an automated quality control solution, installing the right lighting that can improve defect detection. Lighting that is brighter, with enough contrast, but from angles that don’t produce glare will help both human operators and machine vision systems to accurately scan for surface defects. A standardized inspection lighting at 300 lux per ISO 10545-2, with tiles viewed perpendicular from at least 1 meter distance is optimal, and training operators on the specific defect types also increases consistency. Investing in an AI-based quality control can now bring the added advantage of training a model once and adding to its accuracy over time with more data, instead of retraining every new employee.
Catching crazing on tiles with AI-based quality control
Crazing—fine spider-web cracks in the glaze—is particularly problematic because it can appear after firing, during storage, after installation, or even months later. Using the latest technologies to catch these micro-level details that slips through manual inspection will dramatically improve operational costs.
ISO 10545-11 specifically addresses crazing resistance testing, but standard testing catches only the most severe mismatches. Moreover, cracks that can trap dirt and bacteria make tiles unsuitable for hygienic applications like bathrooms or commercial kitchens. Keeping the highest quality standards keeps customers loyal and opens up new markets in a competitive landscape.
Where AI-based machine vision systems can help is detecting fine surface cracking. Specialised cameras for quality control that have lenses to catch defects beyond the human eye, as well as correctly set up lighting can improve defect detection rates.
When speaking to an AI quality control provider, we suggest you ask them about lighting recommendations and setup support because it can significantly impact the quality of the images. In addition, whether your provider has the industry’s highest resolution cameras may not be necessary: test solutions on the line to understand how their hardware interacts with the software provided. For example, Enao Vision sends a starter kit with a 5G hotspot iPhone as your camera and a mount, and can provide customized lighting recommendations based on your setup to ensure you’re able to test within hours.
Detecting colour variation and performing shade matching on the production line
While defects create hazards, cosmetic variations can also have a big impact. Automated scanners that match colour gradients can ensure that tiles from the same production run ultimately match on a customer’s walls or floor.
Shade variation is one of the most common customer complaints—and one of the hardest to catch with human inspection. Our eyes adapt to gradual changes, so operators may not notice drift that's obvious when tiles from the start and end of a run are placed side by side. Most manufacturers grade tiles by shade (A, B, C) and batch them accordingly, but manual shade grading is subjective, and inconsistent grading leads to mixed batches.
AI-based quality control solutions are trained and continuously improved through a library of defects. This means that solutions can be trained consistently with physical reference standards and compare tiles directly from the data library, rather than relying on individual operator memory. Objective shade assessment across every tile will help manual operators detect drift before it exceeds tolerance and enabling tighter shade batching.
Detecting dimensional defects with machine vision systems
Dimensional defects such as warping, edge curvature, and size variation (caliber) affect installation, but can be visually detected by AI solutions.
With the latest technologies, industrial monitoring solutions can be installed at various stages to support quality control. For example, temperature monitors can be installed for heating and cooling to reduce the chances of warping and defects can be caused by automated dimensional measurement systems. AI-powered quality control is both a monitoring solution and a last-check before products get shipped.
Where visual inspection with machine vision systems can be particularly useful is catching edge chips, corner damage, and edge straightness issues that dimensional systems miss. These systems can be trained to track diagonal and edge warpage to ensure flatness. For example, Enao Vision is installed on our customers’ production line to complement dimensional measurement by catching visual edge defects—chips, nipped corners, and surface irregularities near edges—that automated calipers don't detect.
Getting Started
Manual inspection on production lines is an unnecessary compromise in the latest generation of automated quality control. Manufacturers today can choose from a variety of solutions providers, from full-packages that include proprietary cameras and special training to customizable solutions that can be tested on production lines within hours. New technologies and pricing models also give manufacturers more flexibility than ever to make the best use of budgets to test providers before committing for years. To get started, you can check out our list of leading machine vision system providers.
Enao Vision’s iPhone-based solution can be tested for the first month free of charge and can come with a complete starter kit. Try Enao Vision for free on one of your ceramic tile production lines to see how it works.