The manufacturing industry has been going through a transformation in recent years. What we see now is an enhanced network of technology, systems, and machinery to increase productivity and im-prove quality and efficiency. This has resulted in massive demand for big data, smart objects, and AI (artificial intelligence).
In a production environment that is becoming increasingly competitive, quality has become a vital dif-ferentiating factor. Lack of quality can significantly reduce a factory’s competitiveness, which has very real financial consequences. In a recent poll, however, nearly 20 percent of manufacturers admitted they don’t have the capacity to properly measure non-quality. This is where modern AI systems can play a crucial role.
Even companies that can measure non-quality might be capturing only a percentage of possible de-fects. Four criteria are typically used to measure non-quality:
It is easy to see why reducing non-quality has become so important for companies that want to re-duce costs and at the same time increase performance.
In a 2017 study, Infosys found that AI and machine learning is regarded by 75 percent of manufac-turers as a core factor in transformation. Around 57 percent of respondents regarded the cognitive tasks controlled by AI as a key factor.
Artificial intelligence reinforces a company’s current human resources and infrastructure and gives it the ability to pro-actively discover faults and mistakes that not only reduce the quality of the products being manufactured but also weaken the whole production chain.
The deep learning and machine learning algorithms used by modern AI systems contribute to the in-creasing automation of quality control. This in turn helps to significantly reduce the risk of faulty parts.
The increasing level of automation in modern manufacturing plants represents a significant challenge to traditional quality control systems that are still based on human beings’ limited ability to visually de-tect quality deviations. AI, on the other hand, is by its very nature not restricted by occupational health, physiological, or variability constraints - provided that it is trained on high-quality data.
More reliable and more systematic than the human eye, AI systems can pick up inferior quality right from the selection of raw material to the final stages of production. The Center for Industrial Research in Quebec has, for example, developed a system that uses mathematical models, sensors, and digi-tal vision technology to manage quality control of wood shavings used in the production of paper. The system measures not only dimensions but also freshness and the presence of contaminants and other faults.
The same revolution is currently occurring on assembly lines. In the motoring industry, for example, most faults typically take place in the bodywork painting department. With modern AI-driven optical inspection solutions, however, faults can be automatically (and rapidly) detected regardless of the type or color of paint being used.
AI-powered quality control can also be implemented right at the end of a production line to check whether products comply with a certain set of standards. A well-known pizza chain has, for example, installed an AI-powered video control system that in a split second detects whether a pizza looks the way it is supposed to. The system checks factors such as shape, size, and the distribution of ingredi-ents. If the yellow, green, and red peppers are, for example, unevenly distributed across the pizza, it will be rejected.
Using a combination of machine learning algorithms and computer vision, modern AI systems are now able to not only pick up faulty products but also to anticipate defects. For these systems to work properly, however, it is important to train them on data sets that are both relevant and precise. The old adage is still true: garbage in, garbage out.
Drop us a line and we will get back to you