www.ptreview.co.uk
03
'25
Written on Modified on
Quality Control is the top AI driver in manufacturing for a good reason
Interview with Daniel Sperlich, Strategic Product Manager for Controllers in the EMEA region.
emea.mitsubishielectric.com
[Source: Mitsubishi Electric Europe]
As Gartner forecasts, as many as half of the world's manufacturing companies will already be supporting themselves with AI tools for quality control of their products or components by the end of this year.
From a European perspective, quality control remains a top priority in AI-driven automation. According to “The 2024 AI in European Manufacturing Report by Makerverse”, 79% of European manufacturers believe AI will significantly enhance production efficiency over the next five years. A closer look at the data reveals that quality control and inspection lead the way in automation priorities, with 59% of industrial companies focusing on AI to improve this area—by far the most significant investment in AI-driven manufacturing.
Undoubtedly, the emphasis on implementing AI solutions in product verification and quality control has specific reasons. We ask Daniel Sperlich, Strategic Product Manager for Controllers in the EMEA region at Mitsubishi Electric, about these, wanting to find out what prospects and opportunities the move to next-generation quality control opens up for manufacturing plants.
Journalist: Automated AI quality control is one of the fastest-growing areas of production automation. Why are European companies so eager to invest in these types of solutions?
Daniel Sperlich: Indeed, we are observing significant interest in AI-supported automated visual inspection systems. I can say that quality control is the top AI driver in manufacturing for a good reason. Firstly, quality control is a process that directly impacts customer satisfaction and brand reputation. Every defective product that reaches the customer generates complaint costs and can damage the company's image. Rebuilding it is often very difficult and requires a series of corrective actions.
Secondly, previously employees were manual inspectors who had to check every single product. With AI visual inspection, it largely frees up these employees to focus on other activities, while guaranteeing operation with ultimate precision 24 hours a day. We know that, as humans, we have certain limitations regarding our ability to maintain concentration for extended periods, and that's why we're looking with interest at tools that can effectively support us in this regard. So I'm totally not talking about replacing human quality control, but shifting the task from 100% inspection including good products to only inspecting defective products, resulting in fewer products to be examined over a 1-day worktime. This minimizes the risk of missing defects to almost zero.
I think it's also worth mentioning the labor shortage problem constantly plaguing the industry. I recently came across a Eurobarometer survey showing that twenty-two percent of employers in European heavy industry say they can't find the workers they need. This usually means very limited quality control capabilities and even the need to slow down the production process.
Improved quality control, therefore, also means maintaining higher plant productivity and overall production cost optimization. Products with defects are immediately reported and taken off the line, reducing the risk of producing entire batches of defective products. This has a huge impact on the OEE indicator.
Journalist: What specific production problems can be solved through automated visual inspection?
Daniel Sperlich: The list is quite extensive. Take our VIXIO system, for example. It can detect all sorts of defects - scratches, cracks, discoloration, incorrect textures. It checks assembly completeness - whether all components are in place.. Moreover, thanks to deep learning, the system learns to recognize increasingly new types of defects and is prepared for varying production conditions.
MELSOFT MaiLab analysis solution provides continuous monitoring of all welding points. By analyzing all welding parameters in real-time - at a rate of 1,000 welding parameters per second - the tool can detect even the smallest changes in patterns and deviations that indicate potential quality problems before they cause defects or flaws. The system achieves 97% accuracy in defect detection and 100% coverage of welding points, which can result in a 65% reduction in manual inspection time and a 40% decrease in welding defects.
Journalist: You mentioned deep learning. What makes AI-based systems superior to any other vision systems?
Daniel Sperlich: Traditional vision systems operate based on rigidly defined rules and parameters. They work well for simple, repetitive tasks. But when there's greater product variability, different types of defects, varying lighting conditions they start to become very complex and require a lot of time for thorough consideration and testing-. Deep learning systems like VIXIO can learn and you can easily adopt new product variants or improve the AI Accuracy. They just need to be shown enough examples of good and defective products, and they'll "learn" to recognize defects, even previously undefined ones. Moreover, they become increasingly effective over time with more good product pictures and knowledge gained in practice.
Journalist: What does the implementation process for such a system look like? Does it require significant changes to the existing production line?
Daniel Sperlich: We strive to implement in a totally non-invasive model. What does this mean? The VIXIO can be integrated with virtually any production line. The system consists of an industrial camera, lighting, IPC computing unit, and software and can be installed beside the already running production. Then we conduct the system training process, using samples of good and defective products. In many cases, our customers already have sample pictures that we can use even prior to the installation to prove the effectiveness of the Inspection software. The AI Creation phase normally takes only a few minutes, until it can be deployed and first tests can be started. Of course, we also provide comprehensive technical support and operator training.
Journalist: Which industries most frequently turn to automated visual inspection solutions?
Daniel Sperlich: Initially, the automotive and electronics industries dominated, where quality requirements were particularly high. But currently, we're seeing great interest from practically all industrial sectors. VIXIO is used in the production of packaging, metal products, plastics, building materials, and food products. It's valuable anywhere where product quality is crucial and can be assessed based on visual characteristics.
I think we can highlight here the example of VIXIO potential in practice. The case is about the body panel inspection in automotive. After the press of the chassis parts, the whole part is then inspected on both sides. If anything happened during the process, may it be a slight displacement of the metal sheet, material defects etc. had to be inspected by the worker at the final station. Together with robots, that task is now possible to automate with AI visual inspection.
Journalist: How do you see the future of quality control in manufacturing?
Daniel Sperlich: The trend is clearly moving towards full automation and AI integration. We're seeing increasing interest in predictive quality control, where systems not only detect defects but also predict potential quality issues before they occur. The integration of visual inspection with other Industry 4.0 technologies - for example data analytics platforms like MELSOFT MAILAB. Since not all defects can be visually detected, process parameters also play an important role in predicting quality losses that, if combined, provide a powerful, allrounder inspection platform that can adapt and counteract, before serial defects can form. This creates comprehensive Quality Management Systems (QMS) that can significantly improve production efficiency and product quality.
Journalist: What advice would you give to companies considering investing in automated visual inspection?
Daniel Sperlich: First, clearly define your quality control needs and challenges. Start with areas where automation of inspection processes has the biggest impact, for example in manual inspection stations. Consider not just the immediate cost savings but also long-term benefits like improved product quality, reduced waste, and enhanced customer satisfaction. It's also important to choose a solution that can grow with your business and can simply adapt to the production line and changing needs. Finally, ensure you have good support from your technology provider - proper implementation and training are crucial for success and allow rapid adoption of new products or AI improvement.
Sources:
The 2024 AI in European Manufacturing Report by Makerverse: 2024 AI in European Manufacturing Report - MakerVerse
Eurobarometer survey: European Year of Skills: Survey highlights skills shortages in small and medium-sized enterprises (SMEs) - European Commission
From a European perspective, quality control remains a top priority in AI-driven automation. According to “The 2024 AI in European Manufacturing Report by Makerverse”, 79% of European manufacturers believe AI will significantly enhance production efficiency over the next five years. A closer look at the data reveals that quality control and inspection lead the way in automation priorities, with 59% of industrial companies focusing on AI to improve this area—by far the most significant investment in AI-driven manufacturing.
Undoubtedly, the emphasis on implementing AI solutions in product verification and quality control has specific reasons. We ask Daniel Sperlich, Strategic Product Manager for Controllers in the EMEA region at Mitsubishi Electric, about these, wanting to find out what prospects and opportunities the move to next-generation quality control opens up for manufacturing plants.
Journalist: Automated AI quality control is one of the fastest-growing areas of production automation. Why are European companies so eager to invest in these types of solutions?
Daniel Sperlich: Indeed, we are observing significant interest in AI-supported automated visual inspection systems. I can say that quality control is the top AI driver in manufacturing for a good reason. Firstly, quality control is a process that directly impacts customer satisfaction and brand reputation. Every defective product that reaches the customer generates complaint costs and can damage the company's image. Rebuilding it is often very difficult and requires a series of corrective actions.
Secondly, previously employees were manual inspectors who had to check every single product. With AI visual inspection, it largely frees up these employees to focus on other activities, while guaranteeing operation with ultimate precision 24 hours a day. We know that, as humans, we have certain limitations regarding our ability to maintain concentration for extended periods, and that's why we're looking with interest at tools that can effectively support us in this regard. So I'm totally not talking about replacing human quality control, but shifting the task from 100% inspection including good products to only inspecting defective products, resulting in fewer products to be examined over a 1-day worktime. This minimizes the risk of missing defects to almost zero.
I think it's also worth mentioning the labor shortage problem constantly plaguing the industry. I recently came across a Eurobarometer survey showing that twenty-two percent of employers in European heavy industry say they can't find the workers they need. This usually means very limited quality control capabilities and even the need to slow down the production process.
Improved quality control, therefore, also means maintaining higher plant productivity and overall production cost optimization. Products with defects are immediately reported and taken off the line, reducing the risk of producing entire batches of defective products. This has a huge impact on the OEE indicator.
Journalist: What specific production problems can be solved through automated visual inspection?
Daniel Sperlich: The list is quite extensive. Take our VIXIO system, for example. It can detect all sorts of defects - scratches, cracks, discoloration, incorrect textures. It checks assembly completeness - whether all components are in place.. Moreover, thanks to deep learning, the system learns to recognize increasingly new types of defects and is prepared for varying production conditions.
MELSOFT MaiLab analysis solution provides continuous monitoring of all welding points. By analyzing all welding parameters in real-time - at a rate of 1,000 welding parameters per second - the tool can detect even the smallest changes in patterns and deviations that indicate potential quality problems before they cause defects or flaws. The system achieves 97% accuracy in defect detection and 100% coverage of welding points, which can result in a 65% reduction in manual inspection time and a 40% decrease in welding defects.
Journalist: You mentioned deep learning. What makes AI-based systems superior to any other vision systems?
Daniel Sperlich: Traditional vision systems operate based on rigidly defined rules and parameters. They work well for simple, repetitive tasks. But when there's greater product variability, different types of defects, varying lighting conditions they start to become very complex and require a lot of time for thorough consideration and testing-. Deep learning systems like VIXIO can learn and you can easily adopt new product variants or improve the AI Accuracy. They just need to be shown enough examples of good and defective products, and they'll "learn" to recognize defects, even previously undefined ones. Moreover, they become increasingly effective over time with more good product pictures and knowledge gained in practice.
Journalist: What does the implementation process for such a system look like? Does it require significant changes to the existing production line?
Daniel Sperlich: We strive to implement in a totally non-invasive model. What does this mean? The VIXIO can be integrated with virtually any production line. The system consists of an industrial camera, lighting, IPC computing unit, and software and can be installed beside the already running production. Then we conduct the system training process, using samples of good and defective products. In many cases, our customers already have sample pictures that we can use even prior to the installation to prove the effectiveness of the Inspection software. The AI Creation phase normally takes only a few minutes, until it can be deployed and first tests can be started. Of course, we also provide comprehensive technical support and operator training.
Journalist: Which industries most frequently turn to automated visual inspection solutions?
Daniel Sperlich: Initially, the automotive and electronics industries dominated, where quality requirements were particularly high. But currently, we're seeing great interest from practically all industrial sectors. VIXIO is used in the production of packaging, metal products, plastics, building materials, and food products. It's valuable anywhere where product quality is crucial and can be assessed based on visual characteristics.
I think we can highlight here the example of VIXIO potential in practice. The case is about the body panel inspection in automotive. After the press of the chassis parts, the whole part is then inspected on both sides. If anything happened during the process, may it be a slight displacement of the metal sheet, material defects etc. had to be inspected by the worker at the final station. Together with robots, that task is now possible to automate with AI visual inspection.
Journalist: How do you see the future of quality control in manufacturing?
Daniel Sperlich: The trend is clearly moving towards full automation and AI integration. We're seeing increasing interest in predictive quality control, where systems not only detect defects but also predict potential quality issues before they occur. The integration of visual inspection with other Industry 4.0 technologies - for example data analytics platforms like MELSOFT MAILAB. Since not all defects can be visually detected, process parameters also play an important role in predicting quality losses that, if combined, provide a powerful, allrounder inspection platform that can adapt and counteract, before serial defects can form. This creates comprehensive Quality Management Systems (QMS) that can significantly improve production efficiency and product quality.
Journalist: What advice would you give to companies considering investing in automated visual inspection?
Daniel Sperlich: First, clearly define your quality control needs and challenges. Start with areas where automation of inspection processes has the biggest impact, for example in manual inspection stations. Consider not just the immediate cost savings but also long-term benefits like improved product quality, reduced waste, and enhanced customer satisfaction. It's also important to choose a solution that can grow with your business and can simply adapt to the production line and changing needs. Finally, ensure you have good support from your technology provider - proper implementation and training are crucial for success and allow rapid adoption of new products or AI improvement.
Sources:
The 2024 AI in European Manufacturing Report by Makerverse: 2024 AI in European Manufacturing Report - MakerVerse
Eurobarometer survey: European Year of Skills: Survey highlights skills shortages in small and medium-sized enterprises (SMEs) - European Commission
