The three use cases presented in this paper demonstrate how the industry’s most advanced machine vision software and its image preprocessing techniques can lead to exponential improvements in the performance of 2D and 3D vision inspections.
3D imaging technology and smart cameras have enhanced the flexibility and performance of machine vision inspection, which, when combined with advanced image preprocessing tools and intelligent software, can further solve automation that was difficult or nearly impossible in the past Check the application.
Advances in these technologies can lead to higher quality products while reducing waste generation and greatly reducing costs. Defect detection of individual parts early in the manufacturing process allows parts to be replaced or repaired before final assembly of the product, reducing production time and costs.
Through the three practical application cases mentioned below, it is not difficult to find that 2D and 3D imaging in a common software platform, combined with advanced and flexible 2D and 3D image processing algorithms, brings significant advantages to visual system inspection. advantages, and improve the efficiency of software operation. The first application case is the automatic inspection and manufacturing of toilets – given the materials and shapes used in the toilet, the inspection process is undoubtedly complicated; the second application is the robot grasping and inspection of automobile brake discs, which was finally realized after several attempts The automation of this production link and the exploration of multiple solutions; the third inspection application is the lithium battery pole piece used in the manufacture of automobile batteries, with faster inspection speed and higher accuracy.
Quick check
You may not imagine that the manufacturing process of toilets can be so complicated, and it is also difficult to automate production. The ceramics and materials used in toilets have their own unique reflectivity, and quality control of their curved surfaces can be challenging. Even the most basic accessories, such as tanks, urinals, inlets and outlets, and sewage pipes, require careful inspection during the manufacturing process. In addition, designers are trying to enrich toilet features, such as the addition of heated toilet seats, integrated bidet and dual-flush systems in well-known toilet brands, which makes toilet manufacturing more complex.
3D imaging enables real-time height measurement of toilet seats to check for flush fit of toilet lids (image courtesy of Teledyne DALSA).
Since quality checks are essential at several manufacturing nodes, manual assembly and inspection of an ordinary toilet can take as many as 20 people, and the entire manual assembly and inspection process can take 5 to 6 minutes.
If defects are not discovered until the final visual inspection link, the production cycle may be extended. For example, the left side of the toilet is slightly higher than the right side, and a spacer needs to be added to the right side to make the sides level, and this small correction will add an additional 5 to 6 minutes to the labor time.
In a typical toilet production line, we use 20 cameras to complete automated inspection tasks and implement machine operation functions. The inspection cameras used in this process are mainly smart cameras, which can be used for quality inspection, part positioning, label position confirmation or to complete machine pick and place operations.
This type of robotic system relies on machine vision to accurately identify the position of the hole and guide the end-of-arm tool to put the screw into the hole. After the screw is installed, the camera again confirms that the screw is installed in the correct position.
With the help of vision software, a laser triangulation 3D camera performs the final inspection. The software is capable of applying 2D and 3D algorithms to image data, converting the camera’s 3D point cloud data into 2D image data to support conventional image preprocessing and facilitate pattern matching or edge detection. The 2D image data also helps guide the robot to the correct location, and the camera, with its 3D imaging capabilities, will then ensure that all parts are properly placed and aligned when taking pictures from the top and sides of the toilet. The camera also confirms that each screw and tag is in the correct position, and can detect gaps between the toilet lid and the bowl, and between the bowl and the floor.
The combination of traditional machine vision and 3D machine vision helps to automate this production line, and the significant advantages of production automation in shortening production time, reducing labor costs and improving product quality are further highlighted.
Through automation, the time to assemble a toilet on the production line is now reduced from 5-6 minutes to 65 seconds, and the labor required has also been reduced from 20 to just 3. In addition, automation is fundamentally more precise and repeatable, resulting in higher quality products.
Reduce brake disc flyout
In the second application case, 2D and 3D cameras work together to help automate the production of automotive brake discs, thereby increasing production efficiency. The brake discs required for this production process are often coated with anti-corrosion materials, such as zinc or polymers, that improve performance and prolong the life of the brake disc.
High precision is required when gripping a car brake disc (image courtesy of Teledyne DALSA).
In the production process of automobile brake discs, before being introduced into automated equipment, the brake discs are placed in a 1-square-meter silo, and then each disc is manually transferred from the silo by workers and placed on the conveyor belt. Another worker then removes the brake discs one by one from the conveyor belt and takes them to the spray booth to balance the brake discs on the coating rod, which will rotate to ensure that the paint can be distributed evenly. Most coating rooms have glass windows or glass doors for easy observation by workers.
After the coating process is complete, workers remove each brake disc from the coating rod and place it on another conveyor belt to continue production of the brake system. If the working hours are 8 hours a day, workers can coat 200 brake discs with anti-corrosion materials.
The average weight of the brake disc prescribed for a typical car is usually 9.5 kg (about 21 lbs), which can lead to worker fatigue, personal injury from falling brake discs, or back strain from improper operation.
Since this production step is highly repetitive, the second automation use case provides an ideal solution, but it is also difficult to operate – the biggest challenge is accurate positioning and balancing on the coated rod The position of the balance disc, otherwise, the brake disc may fly off the rod and cause damage to the glass doors, windows, etc. of the coating room. If such an accident occurs, the assembly line has to be shut down for maintenance, greatly increasing the cost.
It took seven attempts by different suppliers to figure out how to automate the precise positioning of the brake disc on the coated rod to avoid damaging the workshop.
The successful solution of this challenge relies on the use of 2D and 3D cameras at two production process nodes. Among them, the 3D camera is mainly used to mill out the outline of the groove on the pallet where the brake disc is stacked to help guide the robot arm to transfer the brake disc to the moving conveyor belt. A 3D camera was specifically installed about 2.5 meters away from the silo to capture images of the area and its surroundings.
The information initially detected by the 2D camera can then be used to guide a second robot to pick up the brake disc and place it on a rod in the coating chamber. The 2D camera is installed outside the coating and the tilt angle between the camera and the indoor coating rod is calibrated with the help of software. When the image was generated, the software corrected the orientation of the brake disc to ensure it was perfectly perpendicular to the vertical coating rod before placement. Otherwise, the robot may mistakenly place the brake disc halfway, resulting in a production accident.
Smart cameras power robotic pick-and-place applications for smart manufacturing (image courtesy of Teledyne DALSA).
After the spraying is completed, under the operation of the robot arm, the disc is separated from the coating rod, and then the next operation process is continued.
Ultimately, the use of automation has led to an increase in the number of coated brake discs from 200 to 400 per day, with greater precision and reliability, and a further reduction in the risk of worker safety incidents.
Minimize defects
The third use case relies on preprocessing and cutting-edge algorithms widely used in machine vision software to enable accurate inspection of lithium battery components used in modern electric vehicles.
When inspecting lithium battery pole pieces, we can apply line scan cameras and highly intelligent software to check for tiny imperfections such as dirt, scratches or foreign objects that could lead to defects in the finished battery (Image courtesy of Teledyne DALSA ).
Lithium batteries are layered structures consisting of cells, including electrode sheets, also known as lithium paper. Lithium electrode sheets are made up of several layers of materials, including separators and collector foils, such as aluminum foil on the cathode side and copper foil on the anode side. When producing lithium paper, the separator and collector foils need to be stacked in the following order: separator/anode/separator/cathode/separator. The width of the finished lithium pole pieces is between 100 and 400 mm.
Any defects such as dirt, scratches, foreign objects, bright or dark spots, incomplete cathode/anode sheets can cause the lithium battery to explode, creating a safety hazard.
Quality control of lithium paper is critical in this process, and coil inspection of continuous material is often required. Therefore, the use case requires a line scan camera and highly intelligent software combined with image preprocessing techniques to remove background noise and color.
The cameras available are 4K or 8K line scan cameras, depending on the speed of the production line and the level of inspection accuracy required by the manufacturer. For example, a 4K camera with a 26-KHz line speed can inspect 400mm wide Li-ion battery paper at speeds in excess of 150m/min, and can accurately spot anode and cathode pole pieces for defects on the order of 0.01mm. The 8K line scan camera can double the accuracy of the 4K line scan camera, but its operation speed will be relatively slowed by half.
In practice, automated inspections are not always simply calculated based on hardware specifications. Advanced calibration tools assist with linear, nonlinear and perspective image distortion correction, while cutting-edge software algorithms remove or minimize unwanted noise in images to highlight important patterns, features or edges.
Workers can inspect 15 meters of lithium paper every minute, but cannot always identify all potential defects. If 4K and 8K line scan cameras are used, 60-150 meters of lithium paper can be automatically detected per minute, the productivity is greatly increased from 400% to 800%, and the application cost of new environmental protection technology is relatively reduced.
future software development
The above application cases show how we combine various imaging, preprocessing and software technologies to achieve 1+1+1>3 effects and improve product performance. They always guarantee higher product quality, lower manufacturing costs and faster production rates.
Integrating 2D and 3D data into a single software platform facilitates the automation of the production process and is one of the fundamental capabilities of future machine vision software.
Advances in technology bring more automation tools, and software enables these tools to work together more efficiently.