Seven frequently asked questions on artificial intelligence in quality assurance
In many areas “artificial intelligence” (AI for short) and machine learning is now used. Especially in manufacturing, automated processes are replacing manual visual inspections. The introduction of Vision AI (artificial intelligence) technologies is being driven by manufacturers looking for ways to improve productivity and throughput while reducing product defects and production costs. Although AI-based technologies are not a panacea here, they are a clear advantage in terms of efficiency and accuracy in image analysis.
At DENKweit we work with manufacturers to help them achieve their production metrics. Our technology and its access is specialised for use in mass production.
In the following you will find questions that are frequently asked by manufacturers:
What are the capabilities of Vision AI using machine learning processes compared to traditional image analysis?
Computer or image processing systems are best suited when simple logic-based rules need to be evaluated, such as orientation, presence of parts, measurement, etc. However, when it comes to the variability and complexity of product parts, sometimes under widely varying conditions, traditional image processing often fails. In addition, the effort to adjust or optimise such systems is high. AI-based systems can handle much more at the same time. Anomalies such as defects on surface level, deformations, scratches, dents etc. can be detected automatically and faster. Learning, if needed, is easy, because you only have to tell the AI what you want it to detect. The AI then learns its way independently. In addition, AI-based systems can analyse large amounts of data in a few seconds. Translated with www.DeepL.com/Translator (free version)
How many data/how many images do I need to get started with Vision AI?
Our technology requires only a small amount of data/images to be trained to provide highly accurate detection. A clear advantage of our method. For some applications, 15 sample images are sufficient to generate a high level of recognition within production.
Usually our solutions need less than 100 sample images. In order to achieve this high level of efficiency, we decided to use a cloud interface. Our technologies in the background guarantee a high quality of your very individual AI. Using high computing power, it is no problem for DENKweit to seamlessly integrate the latest technologies.
Does a permanent connection to the Internet have to be available?
A short answer: No. Only the image evaluations are created or trained on our servers. Further training or changes can be made here at any time. Only a small number of sample images need to be uploaded. The trained AI is integrated into the production and runs there completely offline. A further trained or improved AI can then be realised by simply exchanging a single file.
Can Vision AI be added to an existing quality inspection setup? What if no cameras are installed in production?
You may have a quality inspection facility with hardware and cameras that you know and trust. Our software can be integrated into your existing system, allowing you to expand your inspection system without the expensive and time-consuming process of installing a new vision system with machine learning.
If you do not have visual inspection systems, we can fall back on experienced partners. Some of our partners have decades of experience with automated inspection systems. We are also happy to work with a partner of your choice.
Does an AI expert in machine learning on staff have to be?
No. Our solution is simple and intuitive. This enables existing personnel to create Vision AI applications for quality assurance in a very short time and with few data. During development and further development, we focus on ensuring that our systems can be operated by everyone.
What are the best applications for detecting faults in production?
There are many applications for detecting anomalies in the quality inspection process. In electrical manufacturing, detection can identify any deviation from the “acceptable” part, and users do not need to predict every variation of a defect; contract packagers can seamlessly change their lines to accommodate a new product or package products in a different way; in bottling plants, surface defects such as dented or misaligned bottle caps can be easily identified; and in metal manufacturing, surface scratches or the quality of a weld seam can be assessed. Translated with www.DeepL.com/Translator (free version) There are no limits to possible applications.
Are there free demo tests or can a proof of concept be made?
Of course. You can always make a free demo with us. We will be happy to advise you and talk about your required solution. Simply contact us.
Would you like to benefit from the advantages of machine learning? Just contact us!
Our services:
Integration and training of AI-based image evaluation, integration of the required camera technologies with our partners, replacement of manual visual inspection
