Checking while making
Quality problems are cheaper to deal with the earlier they are found. A part with a wrong dimension, a missing component or a damaged surface costs more the further it travels down the line. Industrial sensors let a plant check products as they are made, often on every unit, rather than relying only on samples measured later in a lab.
The idea is the same as in process control. A sensor takes a measurement, a controller compares it with an acceptable range, and something happens as a result: a part is accepted, rejected, flagged or sent for rework. The broader practice of manufacturing quality control also includes procedures, sampling plans and people, and sensors are one tool within it.
What gets checked
Inspection sensors typically answer a few basic questions.
- Is it there? Presence sensors confirm that a part, label, cap or component exists in the right place.
- Is it the right size? Dimension checks verify length, width, height, diameter or thickness against tolerances.
- Is it positioned and oriented correctly? Many automated steps depend on a part sitting where the machine expects it.
- Does it look right? Surface checks look for scratches, dents, stains, color differences or print errors.
- Is it complete and filled correctly? Level, weight and count checks cover packaging and assembly.
Simple sensors that do a lot
Many effective checks use plain devices. Photoelectric sensors detect whether a light beam is blocked or reflected, which is enough to confirm presence or count items on a conveyor. Inductive sensors detect metal parts, and limit switches and proximity devices confirm that something has reached a position. Weighing systems and load cells check that a pack contains roughly the right amount of product. These are inexpensive, fast and easy to maintain, which is why they remain very common.
Measuring dimensions
For dimensions, plants use several approaches, chosen by the precision needed and the part’s shape.
Contact probes touch the part and report its position, which suits parts that must be measured precisely and can tolerate contact. Laser displacement sensors project light onto a surface and calculate distance, allowing thickness, height or flatness to be measured without touching the part. Laser line scanners and similar devices capture a profile of a part as it moves past. Position encoders on moving axes can also report how far a part or tool has traveled, supporting measurements built from motion. For more complex shapes, 3D scanners capture a full surface and compare it to the design model, although these are more often used for sampling than for checking every unit at line speed.
Vision systems
A machine vision system pairs a camera, lighting and software. The camera captures an image of each part, and the software extracts measurements or looks for features and flaws. Vision systems can read printed codes, check labels, verify assembly, measure edges and detect surface defects that simple sensors cannot.
Results depend heavily on setup. Consistent lighting is usually the most important factor, since shadows and reflections can fool software. The camera must be mounted rigidly, because vibration blurs images and shifts measurements. Parts must be presented in a repeatable way, and the optics must suit the size of the feature being checked.
Newer systems may use trained software models to judge appearance. How well they work typically depends on the quality and variety of the examples they were trained on, and they still need monitoring over time. Wider approaches that use AI quality inspection are discussed in separate articles.
From detection to action
A measurement alone does not improve quality. The system needs a defined response:
- Reject. A diverter, air jet or pusher removes the bad part from the line.
- Flag or hold. The part is marked and set aside for human review.
- Alert. Repeated failures trigger an alarm so operators can investigate.
- Feed back. Trends in measurements can adjust the process automatically or prompt a person to do so. For instance, a slow drift in a dimension might point to tool wear or a changing temperature.
That last step is where inspection starts to improve the process itself instead of just sorting its output. Consistent results from the line, sometimes described as product consistency, generally come from controlling the process, with inspection confirming that control is working.
Keeping measurements trustworthy
An inspection sensor that is wrong is worse than none, because it creates false confidence or rejects good product.
- Calibration. Gauges and sensors need regular checks against known references such as certified master parts. The interval depends on the device, the tolerance and the manufacturer’s guidance.
- Cleanliness. Dust, coolant, adhesive and condensation on lenses and windows are common causes of false readings.
- Validation. Plants often test a system with known good and known bad parts to confirm it catches what it should.
- Review of rejects. Looking at what the system rejects shows whether it is too strict, too lenient or drifting.
Limits to keep in mind
No inspection method finds everything. Some defects are internal and invisible to a camera or laser. Setup can be time-consuming when products change frequently. False rejects waste good material, and false accepts let defects through. Inspection also adds equipment and upkeep cost, so plants usually weigh it against how costly or risky a defect would be, particularly in sectors with strict requirements.
Install and service these devices according to the manufacturer’s manuals, with trained personnel, and follow lockout and tagout style procedures and local regulations, since inspection stations often sit beside moving machinery. Laser-based devices in particular carry their own safety requirements. This article is general information, not engineering advice.