KI und ai-predict (2)
The service offered by ai-predict consists of solving highly complex problems using AI approaches. This involves focusing on a small number of specialist topics in order to identify and implement suitable AI methods using existing specialist domain knowledge. Customers receive individualized, ready-to-use AI software (ML pipeline).
Our specialist domains are “formulations”, “energy efficiency” and “failure probabilities”
For recipes (ai.recipe) is the approach of using AI to calculate the recipe of a product. The more complex the product, the more extensive the interactions between the ingredients, the more extensive the necessary series of tests in the laboratory. A formulation with 15 ingredients, all of which may also behave non-linearly to one another, results in several billion possible combinations. The AI approach is able to efficiently determine solutions in this complexity. Previous projects have shown that development times can be reduced by more than 80% while maintaining the same quality. The business case was very quickly positive for the customer. When it comes to energy efficiency(ai.energy), the solution approach is to use AI to optimize the energy efficiency of component motion sequences. A component has programmed motion sequences that are often not designed to be energy efficient. By using real data (e.g. mass, time, energy consumption), energy-optimized movement profiles are determined. Previous experience has shown that energy savings in the 2-digit percentage range can be achieved. This means that the business case for the customer is also quickly positive. In the “ai.failsafe” area, failure probabilities were developed in the production areas of large OEMs. The basic idea behind this is that patterns can be identified in the complex processes of production planning and control that indicate a failure. The occurrence of these patterns is gradually recognized during operation, enabling proactive action to be taken and production processes to be prevented in advance. There are active inquiries in this area, which are being pursued but will not be a focus. ai-predict has practical experience in all 3 business areas. Preliminary studies and projects have been successfully implemented. Thanks to our domain knowledge in these areas, we quickly understand the content issues and can thus effectively implement ML pipelines.