Simply put, machine vision is the ability of a computer to “see” and process information about what it “sees;” a camera system is used to acquire an image, which is then processed and used to complete a task. For example, a seam between two parts can be located and laser welded together. The exact position of the seam may vary due to part stack up tolerances; if so, the vision system is used to adjust the welding tool path automatically to match the actual location of the parts.
Using one or more video cameras, analog-to-digital conversion (ADC), and digital signal processing (DSP), a machine vision system produces data that gets sent to a computer or robot controller. The data sent is based on the set-up process, whereby one “teaches” the machine-vision system, defining what is good or bad, or outside of limits. The camera/computer visually looks at a part and is programmed to indicate whether the process should proceed, stop, or adjust. If it is outside of boundaries, the program is triggered to throw up a flag.
Laser marking/welding, resistance welding, and cutting systems are starting to incorporate machine-vision as an additional quality check. It is often used for inspection, reading barcodes, or to ensure optical character or part alignment.
The complexity and cost of the system is dependent on the specific vision application, not just the cost of the vision hardware. It is worth noting that considerable experimentation with lighting, vision parameters, and testing with a quantity of sample parts is a key part of developing a successful vision process. This is particularly true for vision-based inspection. Lighting plays a significant role in the effectiveness of the vision system. Illumination of the target feature must be carefully controlled by the system builder so that the vision system can consistently and correctly interpret what it is seeing. In particular, the direction, color, and intensity of the lighting source must optimize the visibility of the target features, and stray or ambient light from the production environment must not be allowed to interfere with this process. The vision application test lab at AMADA MIYACHI is used to test and specify appropriate lighting and camera settings for a successful vision system implementation.
Another important but often overlooked aspect for effective vision system operation is the consistency of the part-to-part appearance to the camera. Variations in surface condition (shiny vs matte finish, for example) can cause variations in the vision signal which may cause errors in interpretation. It is important to fine-tune the lighting and vision system parameters using a large quantity of the target parts. This up-front work pays dividends in the success of the vision system when finally deployed in production.
What is machine vision used for?
There are many uses for machine vision. The key applications typically found in laser welding, laser marking, and laser micromachining include:
- Part alignment/orientation
- Mark verification
Part alignment and orientation is required for many applications. Examples include detecting the weld seam and aligning a weld path to the parts; determining the position and rotation of a part to enable a mark to be placed exactly at the correct location and orientation; and using part or tool fiducials to either align parts for micromachining or position feature locations.
Mark verification provides a good / no good assessment of the mark, which may be text or barcodes or both. In addition to a basic read function, the barcode can be read and verified to match with quality aspects of MIL and IEC standards.
Inspection is becoming a more important aspect of vision. Parts can be inspected while still on the machine or records of the part weld, mark of feature can be captured and saved to a data base.
What are the benefits of machine vision?
Increased throughput – In many cases a vision system can provide corrective commands in about one second-much faster than even a trained operator could assess and make the necessary manual corrections to the system.
Decreased scrap – Machine vision can correct for part manufacturing tolerances. Regardless of how parts are made, there are tolerances on all the dimensions. While machine vision does not give manufacturers a blank slate, it can help fine tune and improve results. Especially for high-priced components, where failure can mean thousands of dollars, it is critical to have the checks and balances offered by a vision system.
Automation to eliminate human error – In some cases, automation by machine vision can eliminate human error. This is particularly the case when the part can be identified – either by shape, barcode, or otherwise. Here, a vision system can either select the correct program to run or issue a flag that an unexpected event has occurred. It is also worth noting that the vision system also removes operator to operator variance and removes operator fatigue from the equation.
Examples of machine vision in AMADA MIYACHI systems
AMADA MIYACHI integrates machine vision into many systems, including laser welding, laser marking, laser micromachining and resistance welding.
Machine vision systems are very common in laser welding systems, where they are frequently used to identify reference points or seam tracking, to determine if tolerances are off, or to ensure that the right parts have been loaded.
In resistance welding, machine vision is often integrated to inspect electrodes and identify when they need cleaning or replacement. This ensures quality, repeatable welds over the lifetime of the system. Combined with AMADA MIYACHI’s weld checkers, this provides a high level of certainty that the weld was successfully applied.
Setting up a machine vision system for laser or resistance welding, laser marking or micromachining
Three main components are required to set up a machine vision system: camera, lighting, and software.
When we set up a machine vision system, we begin with a camera that can acquire images of sufficient resolution. Special lighting must often be used to acquire an acceptable image. These cameras are typically mated with a programming suite that an engineer will use to develop the vision program. The program must acquire and process the image, identify/communicate what is good/bad, and give instructions on how to proceed, depending on the image collected.
Due to the amount of effort needed to develop these programs, machine vision systems are used most often in high value parts in medical and automotive applications. They are most often used to ensure the right part was loaded and as part of a series of quality/safety checks before and after the process.
One recent example was a 3 x 4 array (3 columns by 4 rows) of small oval-shaped parts to be used for a medical ablation application. In the middle of the oval, 2 pins stuck up straight – but not dead center of the oval. The machine vision system we developed checked that all parts were filled into the array and that the 2 pins were placed at zero degrees and 180 degrees so they would be properly offset when turned.
Keep an eye out for our next post on machine vision, which will focus on potential issues in your machine vision system, and how to resolve them.