An ECG electrode that has been gelled and run through the curing oven can't be moved. A film sealed inside a flow sensor can't be moved either. With flat, often transparent parts in medical device assembly, you find the position once and then lock it in with adhesive, heat, a cut, or a cover.
At the handoff to the carrier or gripper, a second question comes up. Does the position simply release the station, or does it become a correction value for an axis? This article works through both questions using example application scenarios based on real inquiries.
Key Takeaways
- Flat, often transparent medical parts are positioned once and then fixed by adhesive, heat, cutting, or a cover. The station only needs to know where the part is, so pattern matching is usually enough. You don't need a tolerance-based measurement or an inspection report.
- When you reuse a proven station on a new machine, such as one that finds the position and orientation of a transparent film with pattern matching, carry over the lighting and camera distance too.
- For ECG electrodes, a setup can check position before and after the oven and also find clear adhesive drops. The VS-G Series controller supports up to eight cameras, but multi-spectrum imaging depends on the camera type. Choose the camera for the drops before you design the station.
- The next step decides whether position is a release or a correction value. A cuvette in a carrier needs a pass/fail. A robot picking covers from a hopper needs position and angle.
- Not every position question needs an image. Distance often tells you whether a part is seated or standing up. The LR-Z Series CMOS laser sensor detects up to about 20 inches, depending on the model. Angle, contour, or multiple positions call for a camera.
Position Once, Fix Permanently: Using Position as a Release Before Bonding, Curing, and Cutting
Before a step that locks a part in place, the station doesn't ask how big the part is. It asks where the part is. Pattern matching is usually enough for that. The camera learns the part contour once, finds it again in the image, and outputs position and angle. A tolerance-based measurement and an inspection report aren't required, and that keeps the setup simpler. The field of view can be larger, because it has to include the room the part has when it is placed. The resolution depends on the release limit, not on a measurement uncertainty.
The material is the hard part. Transparent films and clear adhesive drops give little contrast under top lighting. A backlight turns the film outline into a sharp edge. For drops that barely stand out from the background, imaging at several wavelengths helps.
ECG Electrodes Before and After the Oven: Two Positions and a Clear Adhesive Drop
In one example scenario, ECG electrodes are pre-positioned during gel placement, and three checks are planned. The first is the position before the oven. The second is the position after the oven. The third is detection of clear adhesive drops. The plan uses two cameras with ring lights for the two positions and a third camera with multi-spectrum imaging for the drops.
There is a reason to check two positions on the same part. Before the oven, the position decides whether the part can go in. After the oven, it shows whether the part shifted during curing. For the drop, the question here isn't drop quality. It is whether the drop sits where the electrode will be fixed.
A VS-G Series controller fits a setup with three cameras on one controller, because it supports up to eight cameras with compatible area cameras. Multi-spectrum imaging depends on the camera type, and not every camera supports it. Choose the camera for the drops before you design the station.
Transparent Film in a Flow Sensor: Pattern Matching Instead of Measurement
In another example, a new assembly machine for medical device flow sensors repeats a station that was solved years earlier. The station finds the position and orientation of a transparent film about 1 inch long. The field of view is about 2 inches in one direction, and there is only one product type. The position comes from pattern matching, not from a measurement.
Reusing a solved station is often faster than designing a new one. For a new station like this, the VS Series all-in-one camera with built-in AI is a strong fit. With a backlight and pattern search, it finds the film outline and sends position and angle to the PLC. If you repeat a solved station, carry over the lighting and camera distance as well. That way, the learned pattern works without a new setup.
Cut and Place at ±0.008 Inch: Field of View, Web Speed, and Cycle Time
Take an example pharmaceutical application with a cut-and-place machine. A blank is cut from a web and placed. The assumed values are about ±0.008 inch, a field of view of up to about 10 inches, a web moving at about 6.6 feet per minute, and a few dozen cycles per minute. These values are illustrative.
You can check the numbers against each other up front. A 5-megapixel camera with 2,432 pixels across covers a 10-inch field at roughly 0.004 inch per pixel. For ±0.008 inch, that is a thin margin. If it gets too tight, shrink the field to the area where the blank can actually sit. The web speed isn't a problem. With an assumed exposure of 1 millisecond, the web moves the part about one third of a pixel.
Cuvette and Hopper: Position in the Carrier as a Release or a Correction Value
In an example diagnostics assembly machine, cuvettes are checked for presence and position. The position accuracy is still to be defined, at roughly 0.012 inch. Here, position is a release. Either the cuvette sits in the carrier so the next station can work, or the station waits. The IV4 Series vision sensor with built-in AI fits this question. Its position adjustment lets the inspection windows follow the position of the carrier.
Now take a different example, a planned assembly machine for a medical device. Red plastic covers come out of a hopper, and their position must be found before pick and place. The rate is a few hundred parts per minute, the field of view is about 8 x 16 inches, and the tolerance is about ±0.08 inch. Here, position becomes a correction value that the robot needs to grip the part. The VS Series camera finds position and angle and passes them on with its 2D vision-guided robotics function, which connects to many major robot manufacturers. The red color helps, because the part stands out clearly from the hopper surface.
Ten Parts in a Gripper: Measure the Spacing Instead of Trusting the Mechanics
One idea from small-part medical assembly starts earlier in the process. A gripper picks up ten parts at once, with a specific spacing between them. A cylinder sets that spacing. Its speed depends on friction, so the spacing varies. Instead of trusting the mechanics, a distance sensor would measure the actual spacing and adjust the robot position over a travel of up to about 20 inches, to within 0.004 inch. The wish list includes a compact body and IO-Link or Ethernet.
The idea is sound, but the sensor class is still open. High-resolution distance sensors usually have short measuring ranges. A travel of 20 inches at 0.004 inch needs either a sensor whose range covers the whole travel, or a measurement at a fixed reference point that the gripper passes. Check which class meets range and resolution together before you pick any product.
Where Position Hands Off to the Robot and Where the Cell Begins
The cuvette, hopper, and gripper examples show the same boundary. Position as a release or as a correction value is a positioning task. The sensor finds where the part is and passes on a pass/fail or a number. Picking, the cell with feeding and handoff, and the robot cycle time are a separate automation task. The position question ends where the gripper closes.
In many projects, it is still open where the result goes. In the ECG electrode concept, one option was for the controller to send its signals straight to the servo motors, so a separate controller might not be needed. That is a planning idea, not a working solution. The usual route runs through the PLC and the fieldbus, because that is where releases, interlocks, and station records come together. For a medical device line, it is also the route that equipment qualification can trace most easily.
Simple Position Checks With a Laser Sensor: When a Camera Is Too Much
Not every position question needs an image. Whether a part sits in the carrier or stands up, or whether a cover is seated, can often be answered by distance alone. The LR-Z Series CMOS laser sensor detects distances up to about 20 inches, depending on the model, and switches when the part isn't at the expected height. It can register the distance to the background as a reference. Some versions can store a matte, firmly fixed surface behind the part as that reference. A surface that moves isn't compensated. Transparent objects can be detected when there is a background, so guides and fixtures help.
If the question goes beyond height, to angle, contour, or several positions, a camera is the right tool.
Frequently Asked Questions
If the question is where the film is, yes. A backlight turns its outline into an edge, and pattern search outputs position and angle. A dimensional check with a tolerance value is a different task.
With the VS-G Series, it depends on the camera type. For clear adhesive drops, choose the camera type before you design the station, and confirm compatibility with KEYENCE.
It is a release when the station only needs to know whether the part sits correctly in the carrier, like the cuvette. It is a correction value when an axis or a robot needs the position to grip or to adjust, like the covers from the hopper. The cell that does the gripping is a separate task.
One ECG electrode concept considered it, but it was not put into practice there. The usual route runs through the PLC and the fieldbus, such as PROFINET, EtherNet/IP, or EtherCAT. That is where releases and interlocks for the station come together.
When the position question can be answered by a distance, such as whether a part is seated or standing up. The LR-Z Series covers up to about 20 inches, depending on the model. For angle, contour, or multiple positions in the image, a camera is the right class.
- AI-Powered High-Performance Vision System VS-G series
AI-Powered High-Performance Vision System
VS-G series
- Vision System with Built-in AI VS series
Vision System with Built-in AI
VS series
- Vision Sensor with Built-in AI IV4 series
Vision Sensor with Built-in AI
IV4 series
