Deep learning as a quality enabler in the semiconductor industry
Chips are becoming increasingly complex because of more features, while their sizes keep shrinking. At the same time, customers are constantly pushing for faster processes. These are very specific and contradictory challenges that quality control in the semiconductor industry is facing.
Deep learning (DL) has proven to be a perfect tool for image processing and pattern recognition, as automatic feature-learning abilities of deep methods have made them superior. Robovision offers a best-in-class AI platform that can help corporations to adopt and integrate it in a complex runtime production environment.
Quality inspection for deep sub-micron
When in semicon we talk about small, we are talking nanometer-small—especially for wafer production. The most advanced inspection tools on the market today use E-beam technology that identifies defects in-line in nodes at sub-10 nanometers (nm). These defects are the root cause of die failure and need to be detected as early as possible.
As Robovision is able to interface with DUV, E-beam and optical solutions, it works seamlessly with the devices that indentifies defects down to that scale. We enable manufacturers to timely defect and classify flaws in the production process. Thus, the number of Defective Parts Per Million (DPPM) can be significantly reduced and they can increase production yield as a result.
Robovision AI: 95%+ accuracy in ADC
Robovision AI tackles the challenge of traditional rule-based systems. It uses deep learning to classify the defects. By adding more data over time, such as examples of specific defects, the accuracy continues to improve.
Rather than a rule-based system, the AI technology classifies the image to not only spot the presence of a defect, but also to classify these defects automatically. It does so at high speed with a high accuracy score. Robovision AI has proven to be able to achieve 100% accuracy in ADC on training data. And we can achieve 95%+ on production data at 40,000 multi-view defects per hour.
Why the Semiconductor Industry Chooses Robovision
Advanced integrated circuit semiconductors consist of hundreds of layers. Each of these layers of a wafer requires a new model based on a unique dataset. Our specialised algorithm enables automated wafer-map pattern classification. Thanks to our platform’s role management, operators can train the algorithm themselves for each layer. They don’t need hard-to-learn coding skills or the intervention of a data scientist.
Not only does this enable a faster time to market, this approach also enables the fast switching of production lines.
Robovision AI offers a full menu or you can pick and choose à la carte. We can integrate existing algorithm libraries into our platform via an SDK or we can use our own library. Operator and supervisor user roles can be easily customised to accommodate the fab workflow. The API allows for integration of our platform’s capabilities in any existing environment. Alternatively, you can select for an end-to-end solution by Robovision: the choice is yours and anything goes.
Visual inspection ensures quality by detecting defects early in the front-end (wafer fabrication) and back-end (assembly and test) production process. This is frequently conducted during production—for example, using cameras, microscopes, or scanning-electron microscopes. Those images are still commonly evaluated manually by operators for potential defects, however, leaving them subject to error and backlogs and driving up costs.
Robovision helps companies develop modern deep-learning systems for wafer inspection that can be trained to detect and classify defects on wafers automatically. With an accuracy on par with, or better than, human inspectors, companies can obtain early insights on potential process or tool deviations, allowing them to detect problems earlier and improve yields, all the while reducing costs.
Perform every step of the AI life cycle, without ever encountering a line of code. Domain experts and business users can contribute their knowledge and help develop better AI, without having to understand the ins and outs of it.
The Graphical User Interface of our platform actually won the prestigious Henry Van De Velde award in the category business innovation because it manages to unleash the power of AI with just a few easy clicks.
Training an AI model is just the beginning. Maintaining them and monitoring their performance is actually the hard part. Our Robovision AI platform’s streamlined end-to-end AI life cycle also solves the issue of AI scalability due to maintenance challenges. Our AI platform makes every step as smooth and intuitive as possible, from data management to model deployment.
An integrated solution for quality inspection in semicon
Robovision selects the best-in-class to partner for specific needs.
NVIDIA‘s invention of the GPU in 1999 sparked the growth of the PC gaming market, redefined modern computer graphics, and revolutionised parallel computing. More recently, GPU deep learning ignited modern AI—the next era of computing—with the GPU acting as the brain of computers, robots, and self-driving cars that can perceive and understand the world.
Netherlands-based Prodrive Technologies designs and manufactures electronics, software, and mechatronic solutions. In close cooperation with customers, Prodrive Technologies develops and delivers a wide range of high-tech products, systems and solutions, doing everything in-house to cater to many different markets. Prodrive Technologies owns flexible and fully automated manufacturing plants or factories in the Netherlands, U.S., and China where it effectively optimises product designs for each customer.