AI- Autonomous production

Production technology is increasingly characterised by the use of computer-supported knowledge management systems. A distinction is made between expert systems, which pursue a deductive learning approach based on empirical knowledge, and inductive optimization methods, which use mathematical models (algorithms) and initially do not require expert knowledge. The use of inductive learning methods (deep learning, machine learning) has the potential to represent the relevant technology in the future within the framework of autonomous production. Autonomy in this context means that the future production unit (machine network including automation technology along the entire production chain) will be controlled and regulated by a computer in a fully automatic and highly efficient and effective way, fully automatically and without human intervention, with regard to productivity, quality and profitability.

Self-learning electronic data processing systems master the increasing complexity in the environment of dynamically acting manufacturing systems much better than individual people or even whole groups of people are able to do, due to the currently widespread available computing power of the associated hardware. In contrast to the individual human being, they work continuously, increasingly error-free, reliably and achieve maximum manufacturing robustness with the help of predictive, i.e. anticipatory, assistance systems while maximizing productivity. This does not necessarily mean a reduction in personnel deployment in production, but rather that people will be able to support this system to the greatest extent possible through its many possibilities. In this type of system, humans will be able to act either as workers without authority to issue instructions or, as proven experts, as plausibility checkers of the numerically made decisions.  

Computer-controlled production requires concrete optimisation goals, otherwise the classical learning methods of artificial intelligence or machine learning cannot be applied. Such goals can be:

1. Maximizing the number of good parts per hour = max {production output}
2. Maximisation of profitability (€) = max {earnings per hour}

Priorities must be assigned. The top priority may be, for example, to reduce unit costs. Without the simultaneous setpoint value specification of the number of pieces to be produced per time unit, the system may not work properly.