ZDMP aims at providing an extendable platform for supporting factories with a high interoperability level to cope with the concept of connected factories to reach the zero defects goal. In this context, ZDMP will allow end-users to connect their systems (i.e. shopfloor and ERP Systems) to benefit from the features of the platform. These benefits include products and production quality assurance among others. SARKKIS in collaboration with INESC will develop a sub-project entitled AI for Robotic Welding Parametrization and Inspection - AI4R.WELD.
Welding is one of the traditional applications of industrial robots. However, its application to flexible automated productions is still limited. The main reasons are the limitations in robot programming and parameterization. The AI4R.WELD pilot combines automatic robot collision-free program generation with advanced sensing and machine learning for welding parametrization. The approach is human-centric and innovative promoting its usage and standardization for welding applications in a wide range of application areas. Fast and intuitive welding parametrization is the missing link for truly effective robotized welding.
The AI4R.WELD project aims to develop a flexible and user-centric solution for smart manufacturing to expedite the integration of robotic systems into digital manufacturing for welding processes. This results in three main objectives:
According to the ISO 9000, welding is defined as a “special process” which means that its quality cannot be readily verified, and its successful application requires specialist management, personnel and procedures above and beyond those that are considered for general quality systems. These lead to the development of specific procedures, notably the publication of ISO 3834; but checking the quality of welding is not easy. The evidence for this is the various stringent qualifications and conditions specified for welding operators and instruments stringent quality management, induced by the numerous serious accidents in ships, bridges and infrastructure caused by welding problems.
The proposal scope is to provide smart tools, integrated in the ZDMP ecosystem to assist across the setup and production stage of welding operations, ensuring sustainable and efficient productions, applying these to robotic systems, considering human perspective and expertise allied to advance sensing and machine learning.
AI4R.WELD aims to optimise the reduction of raw material usage while also minimising the energy consumption of welded steel fabrication. Robotized Welding is one of the most demanding technological processes to tune due to the numerous variables that are in play, such as the materials to be welded, the wire, gases, external conditions, welding technologies, torch paths/techniques among many others. Manual welding operations are enabled by the best sensor/learning combo available (the human’s senses and cognitive capabilities). Notwithstanding, current robotic welding parametrisation is achieved through educated guess decision making based on an operator expertise, which results waste during trials and due to production quality problems. The automatic parametrisation and quality fault detection for these applications is still neglected due to data analysis limitations. Hence, we believe that machine learning (ML) allied to artificial intelligence (AI) techniques for welding parametrisation and quality control is a business opportunity that deserves to be exploited.
From here, considering mass customisation fabrication, flexibility is an important factor to consider. The ability to reconfigure work cells for new parts and fixtures and parts requires dynamic digitalisation framework able to cope with environment modifications. This is both a symptom and a benefit of the awareness for cyber-physical systems, for “Digital Twins”. Integrated with this paradigm is the robotic integration (cyber-physical agents able to increase productivity). Currently the adoption of robotic welding systems by mass customisation fabrication is still very low due to, among others, the lack of intuitive human machine interfaces that integrate easy programming and quality control. The consortium believes that the key to streamline robotic welding relies on the usage of smart sensing, capable of collecting extracting information, e.g., temperature, humidity, 3D shapes, position and orientation combined with machine learning to automatically adapt robots behaviour and specific process parametrisation according to the environment.
Considering ZDMP, AI4R.WELD contributes with a new and easy-to-use solution for zero-defect smart manufacturing processes that include welding. This solution can be applied to a wide range of applications related to construction applications. The outputs of AI4R.WELD are two-folded: (1) a service providing the correct parametrisation for welding tasks and (2) a product that promotes the usage of smart sensing for quality control for efficient robotic welding.
The members of this consortium believe that the relation with ZDMP is mutual beneficial. While providing these innovative solutions, the consortium will integrate several ZDMP modules such as the Digital Twin and the Non-Destructive Inspection. Then, to further developed the current state of the developed software, AI4R.WELD will integrate the message broker provided by ZDMP to standardize the information exchange between all modules and enabling the service deployment for current and future ZDMP endeavors.