Intelligent Manufacturing: Keeping a Sharper Eye on Industrial Quality Inspection with Smart Technologies
Produkty, Rozwiązania i usługi dla przedsiębiorstw
How high is your production yield?
This has easily got to be one of the most frequently raised questions in the manufacturing sector, as yield, the percentage of non-defective items from total production, is the sector's life line. A manufacturer with a low yield for example could easily lose orders, whereas an original equipment manufacturer (OEM) whose yield is unable to meet certain expectations could not obtain long-term contracts from R&D and design enterprises.
Therefore, manufacturers are committed to increasing their yield. To achieve this, they need to optimize production technologies, streamline processes, purchase more advanced equipment, and enhance operators' capabilities and expertise. However, no matter how much is optimized, quality defects may still exist.
This is what quality inspection is for, the last line of defense in quality control. Traditionally, quality inspectors have to detect defects by naked eye, which is slow and inefficient. If they work for long periods of time, they may develop eye strain. This can lead to increased mistakes and defective materials escaping from the production line.
To bring more certainties into industrial quality inspection, an indefatigable "inspector" that can identify defects more accurately is required.
A growing number of manufacturers are turning to machines to take over the jobs of humans by equipping machines with vision to identify, inspect, and align the position of parts.
It is worth noting that we are not simply just talking about machine vision in the traditional sense, but something on an entirely new level. Conventional machine-vision-based inspection is oriented to specific scenarios where target features such as edges or corners in an image are manually defined. The process involves finding target features, and completing logical judgment tasks based on the numerical value of whether these features exist and the distance between target features.
Although this method can replace manual work, it has obvious bottlenecks. To begin with, it has limited anti-interference. For tasks involving highly random and complex features that cannot be expressed simply using "edges" and "corners" (for example, dense dot-shaped roughness), the defect detection rate can be low. In addition, it does not support iterative learning in the case of production line changes and process upgrades. Redesign is required every time there is a new type of defect or feature. Algorithm development and debugging are inefficient and the cycle is long.
Faced with complex industrial mechanisms and scenarios, manufacturers now incorporate deep learning into machine vision. This technology automatically extracts and compares complex features through sample images, removing the limitations of manually designed feature rules and enabling AI-powered auto learning. Random defects can be identified. This AI-enhanced machine vision can be applied in a wider range of scenarios.
AI-powered quality inspection brings clear advantages. First, it is fast, accurate, and has strong anti-interference capability. Second, it has high precision and good adaptability, suitable for complex defect scenarios. And finally, it supports iterative learning. The larger the data volume, the higher the precision, and the smarter the model.
Indeed, the use of AI can dramatically enhance the level of automation and intelligence in industrial quality inspection. It is an effective way for manufacturers to improve yield and embrace intelligent manufacturing.
IDC believes that the overall market for industrial AI-powered quality inspection is expected to reach US$958 million by 2025, with a compound annual growth rate (CAGR) of 28.5% from 2021 to 2025. Having been piloted over the past several years, industrial AI-powered quality inspection is now being deployed on a large scale.
Huawei, a typical modern manufacturing enterprise, is an active practitioner of AI-powered quality inspection. It has developed a state-of-the-art Industrial AI-Powered Quality Inspection Solution that capitalizes on the company's leading ICTs, including AI, cloud computing, and big data. The solution draws on the wealth of knowledge that Huawei has gained from its quality inspection practices on over 200 production lines, and offers more than 800 industrial-grade image processing tools. So far, this solution has been applied in industries including automotive, electronics, and more, and has delivered unique value to customers.
Automobile manufacturing consists of complex processes, including stamping, welding, painting, and final assembly. During production, automakers are faced with similar quality inspection challenges, such as too many types of parts, models, and defects. Huawei's Industrial AI-Powered Quality Inspection Solution proves to be helpful in a wide range of use cases.
One example is the automotive gap and flush measurement scenario. Traditional manual measurement uses gap and flush gauges, and is highly inefficient. Huawei's solution makes the process much faster. It automatically scans the area to be tested and extracts gap points before finding a straight line fitting the points and removing outliers. The optimal direction vector will be identified, and the gap direction determined, which will be compared to obtain the gap value. With this solution in place, gap and flush measurement per vehicle takes as short as 53 seconds, with an accuracy of ±0.1 mm. The solution realizes high-precision measurement and accurate defect detection.
Another example is the engine assembly compliance detection scenario, a process to prevent quality risks caused by missing assembly actions or non-standard operation sequences. Huawei's Industrial AI-Powered Quality Inspection Solution combs through all assembly actions based on process requirements and identifies actions that affect product quality. The target identification algorithm identifies key actions and performs logical judgment on whether the sequence is correct. Over 99% of non-standard operations can be detected, which helps ensure standard assembly behaviors. This solution can replace traditional manual self-checks and peer checks during engine assembly, thereby increasing production efficiency.
According to the State Administration for Market Regulation of China, in 2021 alone, 8.73 million vehicles had been recalled nationwide, 15% of which were due to manufacturing defects. It is unequivocal that quality improvement is pivotal to enhancing the capabilities of automotive manufacturing. AI-powered quality inspection plays a unique role in a series of automobile manufacturing scenarios, including surface defect detection, operation compliance detection, wrong/missing/reverse component detection, and production safety monitoring.
In a real-industry scenario, an automaker deployed Huawei's Industrial AI-Powered Quality Inspection Solution to upgrade its production facility. The solution helped the manufacturer analyze in real time whether the workers' assembly behaviors are normal, and also detect defects in produced parts and products. Moreover, the automaker worked with Huawei to fully digitalize its production processes. At its factories, defects per unit have been reduced by 80%, the time to produce a vehicle is six minutes shorter, and order delivery is 20% faster. All of the above enhancements improve the company's market competitiveness.
The automotive industry is just one field among many where AI-powered quality inspection can help manufacturers overcome challenges. IDC pointed out that four major industries where AI-powered quality inspection is applied at present are communications and electronics manufacturing, automobiles and parts, consumer goods, and raw materials, and new application scenarios are emerging.
Huawei's Industrial AI-Powered Quality Inspection Solution addresses challenges in scenarios such as operation compliance detection, defect detection, position alignment, and measurement, and has been proven by in a variety of industrial manufacturing use cases.
Foxconn partnered with Huawei to build a showcase production line for Ascend AI-powered quality inspection of smart PV controllers. AI computing and algorithms are used to check whether silicone grease applied to smart PV controllers has the correct color, whether the amount of silicone is insufficient or missing, and whether nameplates are missing, upside down, or wrongly attached. More than 6000 devices can be inspected every month, with an overall accuracy of above 99%. The shift from automated to intelligent quality inspection significantly improves production efficiency and quality.
Powerleader introduced the Ascend Smart Manufacturing Solution to incorporate AI abilities into processes such as incoming quality control, manufacturing process quality check, and packaging inspection. Since implementation, the solution has delivered a detection accuracy exceeding 99%. It not only helps improve product quality, but also cuts down on production and labor costs.
The Ascend Smart Manufacturing Solution was also adopted by Midea Group in its refrigerator factories for adjustable leg inspection, ecolabel detection, brand logo detection, and condenser application inspection. The solution increases the defect detection accuracy by 10% and greatly improves efficiency.
Huawei's Industrial AI-Powered Quality Inspection Solution, which runs on the Ascend AI software and hardware platform, draws on Huawei's own experience in AI-powered quality inspection on over 200 production lines, and delivers a platform for accurate, automated, and intelligent production quality control in multiple application scenarios. Huawei also provides a low-code development platform to encapsulate operators for typical industrial use cases. Visualized orchestration for different service scenarios enables the efficient development of AI-powered quality inspection applications. On top of that, Huawei's device-edge-cloud synergy solution for industrial manufacturers makes AI model iteration much more efficient by enabling online scenario data acquisition and debugging. Models can be quickly iterated, optimized, and delivered to terminal devices in real time.
In summary, Huawei's Industrial AI-Powered Quality Inspection Solution makes it easier to deploy AI to each production line, helping manufacturers overcome obstacles such as low detection accuracy, complex algorithm development, and difficult O&M. The use of AI will deliver practical value for manufacturers by enabling them to keep a sharper eye on quality inspection, even for detecting the tiniest defects in their products. Enterprises are now embracing a brand-new opportunity to stride faster towards intelligent manufacturing.
