2024: Validation of a prototype of a software robot for automatic modelling, planning and optimiztion of production processes based on artificial intelligence in Aktaprint OÜ
PROJECT NAME: Adjustment and validation of a prototype of an AI-based software robot for automatic modelling, planning and optimization of production processes based on artificial intelligence in Aktaprint OÜ.
CHALLENGE: Adjusting and validating a prototype of an automatic modelling and optimization software robot based on Artificial Intelligence, which takes into account the complexity of production planning in this field, streamlines production processes, improves efficiency and thereby increases the company’s competitiveness.
PURPOSE: To adjust and validate a prototype of a software robot, designed for the most effective modelling and optimization of the company’s various products and production processes.
AI TECHNOLOGY TO BE VALIDATED: a wide range of machine learning models were tried and compared: decision tree, random forest, logistic/linear/polynomial regression, etc. Monte Carlo Tree Search and Simulated Annealing proved to be the most suitable.
PROJECT ACTIVITIES: In the implementation of the demo project, three stages with corresponding activities can be distinguished:
1) mapping the requirements of the product needed by the company; 2) adjusting the architecture and design of the software prototype and validating the prototype; 3) testing and validating the software prototype.
EXPECTED RESULTS: The selected AI solution found the most suitable variant for large-scale orders (the most frequently occurring size class in practice) in hundreds of times less time than the created algorithm (Brute Force, which is guaranteed to find the optimal (cheapest) bow assembly solution). In the test, the estimates of 96% of the different orders were within 5% compared to the Brute Force solution.