Throughout the history of technological innovation, factories have always been drivers and early adopters of new technologies. From the early Industrial Revolution to today, the factory has seen significant change and advancement, ultimately striving for higher productivity, greater efficiency, and increased safety.

Today, the factory is once again undergoing a technological revolution thanks to modern advancements in robotics and AI. While the impact of robotics in the factory can be felt in almost all facets of the production process, some of the most significant benefits are occurring in quality control.

In this article, I’ll be discussing traditional challenges in quality control and how modern factories are solving these challenges with robotics.

Historically, a difficult challenge in the production line is how to detect and correct issues with manufactured goods to ensure the highest quality of product possible — a process known as quality control.

Quality control has traditionally been an extremely manual and tedious process. The conventional quality–control process consists of human operators who stand in front of production lines and manually observe each produced good, checking for anomalies or defects. If an anomaly occurs, it is the operator’s job to immediately stop the machine, remove the damaged goods, and then begin a root–cause analysis of the error.

While this approach has been met with varying degrees of success, using human operators ultimately limits the efficiency of the production line, as the speed of production is bottlenecked by the speed of the manual operators. Naturally, a human quality–control operator is prone to error, and the rate of errors only increases as the production line speeds up. In both cases, a human operator is a limiting factor to the line’s performance.

To further illustrate the shortcomings of traditional quality control, consider the example of a beverage bottling plant. Here, an operator checks each produced bottle as it comes down the production line, looking for errors such as physical damage or underfilled bottles. Now imagine an anomaly is found, and a large number of bottles are suddenly coming out with defects. This is a likely scenario because human operators often cannot detect anomalies until it is too late, and the error has already cascaded throughout the system.

In this situation, the operator must stop the production line, search for the root cause of the error, and attempt to resolve the manufacturing defects. Not only is this detrimental because of the number of products that cannot be sold because of defects, it forces the line to be down while the issue is being resolved. This results in lost time, production output, and, ultimately, money.

Luckily, many existing challenges in quality control are today being solved by coupling robotics with data analytics.

Specifically, production lines today are benefiting from the marriage of robotics and computer vision to help achieve more effective quality control. In place of human operators, these production lines leverage advanced systems that consist of robotic arms, such as pick–and–place machines, coupled with high–resolution cameras. These cameras work together to capture a complete image of each product on the line that can then be used for visual analysis and quality control.

Instead of sending these pictures to a human operator to observe, modern quality–control systems leverage computer–vision techniques to detect anomalies and defects in the images of each product. Feeding each image into a pre–trained machine–learning model, the quality–control system can then automatically detect the presence of anomalies in production without the need for any human intervention.

There are many benefits here, key among them an increase in the speed and accuracy of quality control. Compared with human operators, robotic and computer–vision–based systems can more quickly identify anomalies and defects, allowing for the production line to operate at a faster pace. As a corollary to this, robotic–based systems can help identify anomalies earlier than human operators, which helps prevent a cascade of errors throughout the system. When combined with advanced analytics from system sensors, these systems can not only detect issues but also use system analytics to find the root cause.

Consider our example of the bottling plant. Instead of using human operators, a bottling plant equipped with robotic systems leveraging computer vision will be able to perform fast localization of a detected anomaly before its effects can compound and become too detrimental to production. Further, the robotic system can expedite the associated repair, leading to less overall downtime for the production line, less wear on machines, and more profits for the company.

The potential of robotics in quality control is undoubtedly significant, but there is still a number of technical challenges to address.

One obvious challenge is how to handle the massive amounts of data produced by the vision systems on the production line, as well as the computational power required to run the associated computer–vision algorithms. Further, achieving low–latency, real–time inference and responses from our quality–control systems requires high-performance computing on the edge.

Beyond the hardware challenges, there are also system–level challenges involved in appropriately handling anomalies and other deviations from the norm. Engineers will need to design their systems in such a way as to maintain the proper flow of the production line and optimize efficiency while minimizing downtime. This can be a challenge, as failures and anomalies are not easy to predict, making it difficult to prepare appropriate actions and responses in advance.

As modern factories are quickly being revolutionized by budding technologies such as robotic automation, no application is benefiting more than quality control.

Leveraging the marriage of robotics, computer vision, and advanced analytics, modern quality–control systems can lead to higher degrees of automation and ultimately more efficiency for the production line.

Krishnan Ramanathan is an optical engineer at Omron Industrial Automation in Europe. His role centers around R&D and field application developments, where he collaborates directly with customers to understand their application needs and provide suitable safety solutions using input-logic-output alignment combined with other technologies. He also works on integrating cutting-edge technologies like AI, RTLS, and augmented reality within industrial safety sensors.

You must Register or Login to post a comment.

This site uses Akismet to reduce spam. Learn how your comment data is processed.