Related Vendors

Datron AG

Verein Deutscher Werkzeug- maschinenfabriken e. V. (VDW)

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

AI reduces waste in injection molding, for example

At the very beginning, according to the advice of IWU researcher Klimant, one should ask how significant the efficiency gains through AI in their own production could actually be. This question cannot be answered universally without further consideration. The potential depends heavily on the specific process and the associated optimization possibilities. Klimant adds, "In the field of plastics processing, for instance in the widely used injection molding, reject rates of 20 to 30 percent can occasionally occur!" At the same time, this presents a great lever for efficiency gains through AI. But even in processes that are already running stably, AI can also be beneficial. Klimant mentions, for example, predictive maintenance to benefit from longer tool lifespans. According to Klimant, artificial intelligence can also play an important role in addressing the shortage of skilled workers: "We implicitly store knowledge in the AI. This knowledge can be used to train new personnel—particularly when employees retire." This AI knowledge storage also offers new opportunities for automation—particularly for automated quality controls.

Artificial Intelligence is a Kind of Black Box

The researcher also defines artificial intelligence as follows: "When we talk about AI, we usually mean machine learning as a subset of AI. It is capable of learning independently from training data." This involves an empirical method that learns relationships without knowing the analytical connections. Simply put: humans learn from experience! But through AI, process parameters in production are optimized and fed back into process control via an automated control system. "Artificial intelligence can be considered like a black box, where input values flow in and predictions come out," Klimant explains. As an example, a forming process can be cited, where an acoustic signal is measured, and the AI determines whether the process was successful or not. In principle, at the end stands a digital system that can be connected to controls via existing interfaces. This allows AI to influence control algorithms at various points.

Everything Stands and Falls with Computing Power for AI

For AI to be successfully used in production, hardware with very high computing power is sometimes required. "First, it's necessary to distinguish between the training phase and the usage phase (inference). The training phase is always more computationally intensive but is carried out offline. In the usage phase, classical methods like the Support Vector Machine often suffice with edge devices," Klimant notes. It is different, however, in the case of image processing. These AI models require more computing power in both the training and usage phases. "The application cycle also plays a crucial role here," the researcher explains. For example, if a result is needed every five seconds, significantly more computing power will be required compared to a cycle time of 30 seconds. Excluded from this are the evaluations of language models, as they require powerful hardware for computational processes—ranging from well-performing consumer graphics cards to specialized AI cards.