Periodic Reporting for period 1 - OPTIMAI (Optimizing Manufacturing Processes through Artificial Intelligence and Virtualization)
Periodo di rendicontazione: 2021-01-01 al 2022-06-30
The advent of Industry 4.0 has certainly increased productivity and efficiency in manufacturing. Nonetheless, global competition as well as environmental concerns, call for major improvements in European factories that on one hand need to increase their productivity and on the other minimize their ecological footprint. In this respect, OPTIMAI is developing an industrial ecosystem based on Smart Instrumentation, Artificial Intelligence, Augmented Reality and Digital Twins that vertically optimizes multiple stages of production including planning, setup and manufacturing.
Through its developments OPTIMAI aspires to increase the competitiveness of European industries allowing factories to produce more with less, that is increase their production yield and throughput while reducing resources’ utilization. To accomplish this goal, there are complementary research directions that OPTIMAI pursues. One of these is closed-loop AI control and actuation solutions, where AI methods are applied on shop floor sensorial data to timely notify workers on emerging defects or even automatically reconfigure production parameters to avoid or contain defects.
In addition, OPTIMAI develops human-centric technologies for the shopfloor, enhancing workers abilities through Augmented Reality (AR) wearables and Human-Machine Interaction (HMI). Using AR glasses workers are able to see configuration suggestions and notifications, overlaid on their field of view whereas computer vision methods provide contextual analysis and gesture recognition functionalities for easier (re)configuration of machines. Regarding production planning, OPTIMAI develops an environment for the creation of Digital Twins for production lines that is enriched with virtual sensors models and AI-based simulation methods. In this framework engineers can virtually test various production configurations without having to waste production resources in time-consuming pre-production runs. In addition to technological advances, OPTIMAI aims to establish a legal and ethical framework that will on one hand promote workers rights in the OPTIMAI ecosystem and on the other allow industry to harness the potential of the developed groundbreaking technologies.
Through its developments OPTIMAI aspires to increase the competitiveness of European industries allowing factories to produce more with less, that is increase their production yield and throughput while reducing resources’ utilization. To accomplish this goal, there are complementary research directions that OPTIMAI pursues. One of these is closed-loop AI control and actuation solutions, where AI methods are applied on shop floor sensorial data to timely notify workers on emerging defects or even automatically reconfigure production parameters to avoid or contain defects.
In addition, OPTIMAI develops human-centric technologies for the shopfloor, enhancing workers abilities through Augmented Reality (AR) wearables and Human-Machine Interaction (HMI). Using AR glasses workers are able to see configuration suggestions and notifications, overlaid on their field of view whereas computer vision methods provide contextual analysis and gesture recognition functionalities for easier (re)configuration of machines. Regarding production planning, OPTIMAI develops an environment for the creation of Digital Twins for production lines that is enriched with virtual sensors models and AI-based simulation methods. In this framework engineers can virtually test various production configurations without having to waste production resources in time-consuming pre-production runs. In addition to technological advances, OPTIMAI aims to establish a legal and ethical framework that will on one hand promote workers rights in the OPTIMAI ecosystem and on the other allow industry to harness the potential of the developed groundbreaking technologies.
One of the first activities for OPTIMAI was the elicitation of user requirements following a co-creation process that involved industrial end-users, technology providers as well as legal experts in the consortium. Based on these requirements and a thorough analysis of state of the art technologies, user arcs were created that led to the elaboration of the foreseen use cases for each pilot site. The next step was the definition of the OPTIMAI Architecture based on the RAMI 4.0 model which is the basis for the ongoing development of technological components.
In terms of software components, the OPTIMAI middleware has been developed and deployed, providing an API that allows the integration of multiple sensors and modules. Integration of sensors is a continuous process where several sensors are used for testing and development in the laboratory before being installed in the studied production lines. Indicatively machine vision cameras and 3D lasers have been installed in the end-users shop floors while novel soft sensing methodologies have been proposed that produce accurate prediction with low instrumentation cost. Since AI is a focal point for OPTIMAI several AI pipelines have been developed for Zero Defect Manufacturing and are being progressively deployed to support the project’s use cases.
As human-centered production is another goal of OPTIMAI, a prototype for Augmented Reality glasses has been developed, enhanced with Computer Vision functionalities that can analyze the workers field of view and provide relevant notifications coming from the middleware. For example, information regarding a particular defect is overlaid on a manufactured part, whereas workers’ gestures can be recognized facilitating her interaction with the OPTIMAI ecosystem. In a similar line of research, Digital Twins for the pilot sites are developed, with the digitization of the TVES shop floor having been completed whereas certain manufacturing steps for MTCL and KLEE have been virtualized. Another important result has been the Decision Support System (DSS) that has been designed, with its first version being deployed for the early notification on identified defects and the provision of recommendations about production setup. Finally a blockchain layer has been added for recording and tracing production failures as well as important configuration actions.
In terms of software components, the OPTIMAI middleware has been developed and deployed, providing an API that allows the integration of multiple sensors and modules. Integration of sensors is a continuous process where several sensors are used for testing and development in the laboratory before being installed in the studied production lines. Indicatively machine vision cameras and 3D lasers have been installed in the end-users shop floors while novel soft sensing methodologies have been proposed that produce accurate prediction with low instrumentation cost. Since AI is a focal point for OPTIMAI several AI pipelines have been developed for Zero Defect Manufacturing and are being progressively deployed to support the project’s use cases.
As human-centered production is another goal of OPTIMAI, a prototype for Augmented Reality glasses has been developed, enhanced with Computer Vision functionalities that can analyze the workers field of view and provide relevant notifications coming from the middleware. For example, information regarding a particular defect is overlaid on a manufactured part, whereas workers’ gestures can be recognized facilitating her interaction with the OPTIMAI ecosystem. In a similar line of research, Digital Twins for the pilot sites are developed, with the digitization of the TVES shop floor having been completed whereas certain manufacturing steps for MTCL and KLEE have been virtualized. Another important result has been the Decision Support System (DSS) that has been designed, with its first version being deployed for the early notification on identified defects and the provision of recommendations about production setup. Finally a blockchain layer has been added for recording and tracing production failures as well as important configuration actions.
There have been significant scientific achievements for OPTIMAI that reach beyond state of the art. One of them is the development of a soft sensing methodology that uses AI to replicate the accuracy of a high end sensor from a low end one thus reducing the instrumentation costs by an order of magnitude. Essentially, AI is used to capture the correlation between the measurements of these sensors and infer accurate estimations. Another significant advance has been the use of Reinforcement Learning (RL) as an automated approach for equipment calibration where RL monitors the industrial process at hand and decides on the optimal calibration actions. There has been other important advances in the domains of Digital Twins and simulation as well as in Augmented Reality that are both used to improve human performance allowing workers and engineers to carry out their tasks more effectively and with less effort.
Integrating this advances into a common industrial ecosystem, OPTIMAI aspires to increase overall productivity with the timely detection, prediction and containment of defects while at the same time improving the quality of produced goods. Additionally OPTIMAI aims to optimize the use of production resources by minimizing generated waste particularly during pre-production runs and are typically error-prone. To achieve this, Digital Twins are used as a production planning framework. Further focusing in pre-production, OPTIMAI creates methodologies for the (semi)-automated configuration of production equipment that will enable factories to ramp-up their production faster and consequently reduce time to market.
It terms of socio-economic impact, OPTIMAI will reduce the environmental footprint of the studied use cases by minimizing the generated waste and containing defects before they propagate downstream in subsequent production stages. It will also increase the competitiveness of factories both by improving production quality and also by allowing faster setup and calibration for new products. Last but not least, OPTIMAI develops human-centric technologies based on Augmented Reality that enhance human capabilities and overall performance via decision support, advanced human-machine interfaces and smart, human-triggered actuation. Using Augmented Reality wearables, OPTIMAI will visualize on top of real objects of interest, relevant information and notifications while minimizing the cognitive load for workers. Then using smart actuation and gesture recognition software workers are able to change and optimize machine parameters with minimal effort.
Integrating this advances into a common industrial ecosystem, OPTIMAI aspires to increase overall productivity with the timely detection, prediction and containment of defects while at the same time improving the quality of produced goods. Additionally OPTIMAI aims to optimize the use of production resources by minimizing generated waste particularly during pre-production runs and are typically error-prone. To achieve this, Digital Twins are used as a production planning framework. Further focusing in pre-production, OPTIMAI creates methodologies for the (semi)-automated configuration of production equipment that will enable factories to ramp-up their production faster and consequently reduce time to market.
It terms of socio-economic impact, OPTIMAI will reduce the environmental footprint of the studied use cases by minimizing the generated waste and containing defects before they propagate downstream in subsequent production stages. It will also increase the competitiveness of factories both by improving production quality and also by allowing faster setup and calibration for new products. Last but not least, OPTIMAI develops human-centric technologies based on Augmented Reality that enhance human capabilities and overall performance via decision support, advanced human-machine interfaces and smart, human-triggered actuation. Using Augmented Reality wearables, OPTIMAI will visualize on top of real objects of interest, relevant information and notifications while minimizing the cognitive load for workers. Then using smart actuation and gesture recognition software workers are able to change and optimize machine parameters with minimal effort.