Contacto
Vision News > The Machine vision blog > Artificial Vision System Maintenance: What to Check
19/11/2026

Artificial Vision System Maintenance: What to Check

An artificial vision system rarely fails all at once. More often, it gradually loses accuracy. Dust builds up on the lens, lighting intensity drops, a mounting bracket loosens due to vibration, and suddenly the production line starts rejecting good products or letting defective ones through.

That is why artificial vision system maintenance works best as a short, regular routine rather than an intervention after something goes wrong. These checks help you detect performance drift before it leads to downtime or a customer complaint.

Why a Vision System Becomes Less Reliable Over Time

The production environment changes, even when the system itself does not. Temperatures fluctuate, dirt accumulates, packaging batches arrive with a different sheen, and machinery wears down. Every small change shifts the conditions under which the inspection system was originally configured.

The tricky part is that the system keeps running. No alarm goes off, but the safety margin becomes narrower. Regular inspections are precisely what allow you to measure that margin.

What to Check When Maintaining an Artificial Vision System

Optics and Protective Windows 

Clean the lens and protective glass using the method recommended by the manufacturer. A film of dust, oil or condensation reduces contrast and causes false rejects that can look like a software problem.

Lighting

Check that the light intensity matches the reference level. LEDs age, and a gradual loss of light changes what the camera sees without anyone noticing. If the system uses lamps or diffusers, check their condition and cleanliness too.

Mountings and Mechanical Positioning

Vibration can loosen mounting brackets, and a shift of just a few millimetres can alter the field of view. Check that the camera, lighting and reflectors are still in their calibrated positions. Marking the correct position with a physical reference makes this much easier to verify.

Wiring, Connectors and Environment

Inspect connectors, moving cables and seals. Also check the temperature inside the control cabinet and its ventilation, as heat and dust shorten the service life of electronic components.

Calibration and Repeatability

Run a reference part through the system—ideally both a known good part and a known defective one—and check that the results match those recorded during commissioning. If the system performs dimensional measurements, verify calibration using a reference standard.

What to Check in the Software and Data

Back up the inspection program and trained models before making any changes. If the platform allows it, as is the case with Cortex, Bcnvision Group’s artificial vision software, keep a version history so you can roll back if necessary.

Then review the inspection statistics from recent months. A sustained rise in rejects, or a sudden drop, needs an explanation. If you use deep learning, consider retraining the model with images of new products or batches. A model trained on packaging that is no longer used will lose accuracy.

If your problem is the rejection of good products, it is worth reading about how to address false rejects in pharmaceutical serialisation, as many of the same causes also occur in other industries.

Rejectors and Rejection Mechanisms

Accurate detection is of little use if the rejection mechanism fails. Check that the rejector operates at the right time and with the right force, and that the confirmation sensors respond correctly. A rejection that is not carried out becomes a defect that reaches the customer.

When a Maintenance Contract Makes Sense

You can carry out these checks in-house if you have trained staff and enough time. When the line is critical, or the system includes 3D, deep learning or multiple inspection points, support from the manufacturer is often worthwhile.

Bcnvision Group offers maintenance contracts and technical support and resources to ensure that inspections, backups and adjustments are documented. If you would rather train your own team, the company also offers in-person and online training courses.

A Routine That Protects Inspection Performance

Artificial vision system maintenance does not require lengthy shutdowns. It requires consistency, a record of what is measured, and the habit of always comparing results with the initial baseline. If you would like us to review how your installation is performing, you can contact the technical team.

You might also be interested in

Bottle case-packing inspection at the end of the line

A case of bottles leaves the production line, is sealed, palletised and...

Drying trays and flexible packaging in the food industry

The water that remains on the surface of a tray or flexible...

Industrial Rejectors: How to Choose the Right System

Detecting a defective product is only part of automated quality control. To...

The Role of Deep Learning in Battery Manufacturing for Electric Vehicles

Battery manufacturing for electric vehicles is at a decisive turning point. Improving...

CheckCan, AI-powered can inspection for quality control, format changes and bottom closure inspection

On can production lines, speed leaves no room for error. Every container...

Most common problems in container capping and how to detect them in time

In a production line, container capping may seem like a simple stage:...

Maximising efficiency and savings through computer vision in industrial filling control

Efficiency within a bottling plant is not merely measured by line speed...

3D Vision and AI for Tuna Cutting with Automated Portioning

The food industry is undergoing a technological revolution that enhances precision, efficiency,...

Unsupervised Neural Networks: How They Work and Their Application in Intelligent Label Inspection

Industrial automation has evolved from simple programmed systems to technologies capable of...

Can Drying and Rust Prevention: A Key Challenge in the Food Packaging Industry

In the beverage and food industry, production line efficiency does not depend...

Machine vision solutions for the automotive industry

The automotive industry is facing increasingly complex challenges, such as the need...

Figure 1: Principle of TOF measurements. Phase shift measurement. Image extracted from https://ebuah.uah.es/dspace/bitstream/handle/10017/21313/TFM-Rufo-Merino-2013.pdf?sequence=5&isAllowed=y

3D TOF Cameras

What is a TOF camera and how does it work? Applications in...