Today's SME lot-size-one manufacturers are struggling with production efficiency compared to large manufacturers, due to the fact that they do not have enough resources to use the most modern tools for managing production processes, such as MOM or MES systems, not to mention AI solutions. The entry barrier for them is, on the one hand, the cost of such solutions and, on the other, the lack of human resources to exploit the full potential of these tools. Among all MOM deployment preparatory steps, production technologies data input is the most significant barrier to entry, as it is highly labour-intensive and at the same time is critical for the successful deployment of the system. This very often discourages SMEs from investing in solutions that automate and optimize production forecasting and planning.
In many industries, use of CAD tools at the beginning of the production process is a standard. Based on such design, the entire production technology is planned, i.e. the resources needed (workers, machines, robots, materials, semi-products), the individual tasks and production steps and their execution time. Entering data from the CAD projects into the software supporting manufacturing operations management is highly time-consuming, as is further manual planning of required resources and time needed. Such an approach also results in a high degree of inaccuracy in the abovementioned predictions and is fraught with additional time and workload should any changes or deviations from the original assumptions occur. The more variable is the production, the more effort needs to be put into the technology preparation process and its later adoption to changing circumstances.

In AIM2ASSIST project, by applying a deep neural prediction model built on the dedicated neural architecture involving the convolutional neural units (CNNs), we are developing an AI-based component that will forecast necessary resources and/or tasks for a workpiece production and eventually expected duration of the production process. This will be supplemented by two other components developed by us within the project, that will function for dataset generation for the AI-based component. The three components will be integrated with the MASTA's manufacturing operations management system to facilitate access to the solution for manufacturers.

The project has a strongly experimental character. It will allow MASTA to develop, deploy and validate a Proof-of-Concept of the solution in a real operating environment of metal assembling products.