Tire Technology International - March 2024 Issue
Machine vision
Trinity Francis 2024-04-06 10:10:39
LINE OF SIGHT
Discovering the latest uses of automated inspection systems – and how their role in tire production continues to advance
Machine vision technologies continue to evolve and enhance the tire production process. Be it catching defects early, seeing where humans can’t, or monitoring tire building machines themselves, the application of machine vision has grown throughout the tire factory. Whether machine vision will ever be good enough to replicate the highly complex abilities of trained human senses is widely debated, but machine vision developers are striving to bring their technology to a point where the vast majority of human inspection becomes redundant.
“With high-quality image data, machines can detect non-conformities more precisely than humans,” explains Jin Uk Shin, manager of the digital intelligence team at Hankook. “Vision, sensing and inspection technologies collectively play crucial roles in identifying cosmetic defects or non-conformities.”
But to get to a stage where machine vision starts to be on par with – or better than – human inspection, sensor manufacturers must rely on emerging technologies to improve their systems.
As Marián Šrámek, managing director at Micro-Epsilon, explains, the development of machine vision has always faced the same challenges: “It is a battle of the required parameters that have not changed over the years: the resolution of measurement versus material coverage, against the line speeds – and at the same time, the long-term stability of the sensor.”
Šrámek argues that new computer chips will facilitate improved speeds and resolution, but sensor stability is down to the robustness of the hardware – which must be able to withstand significant temperature changes and 24/7 operation.
"Even the definitions of product defects on the side of customers are very vague” Marián Šrámek, managing director, Micro-Epsilon
“Continental needs machine vision hardware that is robust enough to run in a tire production environment,” agrees Steve Howat, Continental’s general manager of technical services for the UK and Ireland.
“Moreover,” he continues, “the imaging hardware must be especially adapted to black surfaces as this is crucial for the quality control of black tires.” Black surfaces make optical sensing particularly difficult if the area is not sufficiently illuminated.
“We’re always looking for lenses with better optical performance,” says Sophia Nilsson, strategic product manager for 3D Vision at Sick. “Our cameras run at very high speeds. With the cameras we have today we can run at 4-6kHz; we’re planning to double this speed quite soon.”
These extremely short exposure times mean the cameras need a lot of light to accurately detect defects against a black surface. To remedy this, Sick is introducing powerful lasers to ensure adequate lighting is provided. However, Nilsson notes, “It’s hard to get a strong enough laser without having a very high laser class, and that presents safety issues.”
In the future, Alain Klein, global account manager at Sick, hopes to eradicate the need for lasers in this process. Instead, LED or other types of lighting will be used to illuminate the area cameras need to evaluate. In this sense, Klein and Nilsson agree that machine vision manufacturers are pushing the limits of existing technologies.
Machine learning
Interpreting vision data is an area where technology is just starting to be explored and optimized for tire manufacturing. As Nilsson explains, over the past 10 years more 3D vision solutions have been introduced, and the past five years have seen a boost in the use of artificial intelligence and deep learning.
“We’re seeing that machine vision is getting easier,” she says. “No one wants a complicated solution that’s time-consuming to set up, so deep learning is very fruitful.”
Tire makers are optimistic about these developments. “In the past, we identified defects through manual inspection and ran the risk of human error and inconsistencies in judgment,” Shin says. “However, the emergence of vision and AI technology has facilitated faster and more accurate detection, showing great promise for future applications.”
But while AI opens new possibilities, implementing these programs can be challenging. “Even the definitions of product defects on the side of customers are very vague and, in principle, vary greatly depending on the tire manufacturers,” says Šrámek. He warns that visual inspection providers must be prepared to significantly alter AI algorithms to tailor this software to each tire maker.
Hankook is looking for its suppliers to deliver three technologies in this area. First, the AI must be able to learn visual defects and detect these accurately.
“Such non-conformities are rare in actual practice and thus require an augmentation technology that can resemble reality using small amounts of data for proper machine learning,” Shin explains.
Second, detected defects must be automatically assigned a label. This helps the tire maker track common defects and optimize production processes to minimize these.
“No one wants a complicated solution that’s time-consuming to set up, so deep learning is very fruitful” Sophia Nilsson, strategic product manager, 3D Vision, Sick
As Alexander Bissle, product manager for tire and rubber at Erhardt+Leimer, explains, this is a growing trend in the use of AI: “There is no opportunity to make error classification without the use of AI.”
Lastly, Hankook is looking for suppliers to demonstrate a choice of, and difference in, the performance of various AI algorithm models.
“It would also be instrumental to be able to choose functions such as classification, object detection and segmentation depending on the required level of object detection,” Shin explains.
Personnel touch
Although a large part of the tire production process already features some level of automation, AI and machine learning are likely to facilitate higher levels of human-free operations over the coming years.
“There is no doubt that, in 10 years’ time there will be fully automated visual inspection of tires as a standard part of the production process,” Šrámek says. This will coincide with fewer humans actively involved in production. However, there will still be a requirement for staff to support production in technician and maintenance roles, for which existing staff can be retrained.
“The challenge is to find the right balance between 100% manual and 100% automatic inspection” Alain Klein, global account manager, Sick
Continental is also planning to increase automation in its tire manufacturing facilities. “Mandatory inspections include checking of tire uniformity and geometry with the help of parameters such as radial and lateral force variation, radial and lateral runout, conicity, ply steer and sidewall bulge as well as indentations,” says Howat.
“Continental is running several development activities to continuously automate all these inspection processes.”
The limitation in this approach is the financial viability of introducing more sophisticated sensing systems and software to interpret this data in the pursuit of automation.
As Klein argues, “Everything is possible; it’s a matter of cost. The challenge is for the tire industry to find the right balance between 100% manual inspection and 100% automatic inspection.”
In reality, Klein sees the implementation of complete automation as too expensive. This means there’s likely to be a middling point where tire makers are willing to invest in additional machine vision and sensors to enable automation without affecting the cost of the product.
As hardware and software continue to improve, machine vision becomes even more key to enable tire makers to optimize the production process. The industry is trending toward greater use of machine vision and better data interpretation to realize the potential of this highly detailed information. Ensuring the security of this data is paramount and something that tire makers are concerned about.
“They are reluctant to bring image data to the cloud but maybe they will have to due to the amount of data they need to analyze,” Klein says.
Regardless of the way it’s implemented, machine vision will certainly take on an increasingly important role in tire manufacturing, especially in conjunction with emerging technologies such as AI and machine learning.
BIG JOBS, BIG DATA
How do sensing requirements change outside of PCR applications?
For larger tires, machine vision and inspection requirements differ slightly from those for car tires, to ensure the same levels of accurate inspection.
“For car tires we use different systems, such as fixed-point laser sensors and sheet-of-light laser systems to check the tire uniformity and geometry characteristics,” explains Steve Howat at Continental. “Truck tire inspections require additional x-ray technology to check the several steel breaker and reinforcement layers.”
As well as checking the structural integrity of a tire’s internal layers, sensing equipment must be able to assess the external surface of a tire.
“The main differences are only in the measurement ranges of the inspection systems,” says Marián Šrámek at Micro Epsilon.
“In the case of more complex machines, the mechanics can fundamentally change as well, which must be able to work with other required load capacities and loads,” he continues.
Again, the requirement for efficient inspection of truck tires needs both hardware and software adaptations for the best results.
“When dealing with objects of varying sizes that deviate from the image range, the auto-focus lens zooms in and out so the camera lens can adjust the image frame accordingly,” says Hankook’s Jin Uk Shin. This enables tire manufacturers to switch between different tire sizes without having to manually adjust vision and sensing aids.
To improve overall visibility, “We can calibrate the vision camera to rotate, rather than stay in a fixed position, based on the objective to accommodate various sizes,” adds Shin. “These steps are incorporated during the planning and design phase to tailor the system to fit the purpose.”
Cameras can be moved or additional optical sensors installed to accurately cover a larger tire. However, as Sick’s Alain Klein notes, “Using a robotic arm, we can change camera position, but this takes more time to complete the inspection as well as more calculation power from the computer side.”
This leaves the tire maker to decide whether more sensors or moving sensors are a better solution to assess truck tires.
SECOND JOB
Using machine vision to increase the longevity of tires – inside the factory and on the road
Machine vision is an essential part of the retreading process, to ensure a tire does not have any major defects that would make it ineligible for retreading. This is not a new application for advanced vision systems, but Sick’s Alain Klein proposes that tire manufacturers could use the technology outside of the retreading factory.
“Manufacturers are trying to differentiate themselves from their competitors, not only with the tire itself but also the services they are able to provide their customers,” he explains.
This is especially true of fleet customers. Machine vision can be employed in the use phase of a tire to monitor wear and ensure tires are sent for retreading at the right time.
Although the service would be used to promote a manufacturer’s retreading services, machine vision in the use phase can improve tire life. If a tire is identified to be wearing unevenly it can, for example, be changed to a different position on the vehicle or have its inflation adjusted. This can result in cost saving for operators by making their vehicles as fuel efficient as possible. It’s also an important safety check to ensure no tire is close to experiencing a major failure.
“This is a new possibility where cameras could be used to generate revenue for the tire manufacturers – not with tires, but with information,” says Klein. It creates an opportunity for tire manufacturers to add value for their customers and form a new business model.
Data collection during tire life will also support the tire maker in identifying common faults or issues that can be prevented with adequate maintenance or by redesigning an aspect of the tire.
Greater visibility in this stage will mean more tires can benefit from a second or even third life. This has a direct impact on tire makers’ Scope 3 emissions, reducing the environmental impact of their tires during use.
©UKi Media & Events. View All Articles.