Periodic Reporting for period 2 - IoTwins (Distributed Digital Twins for industrial SMEs: a big-data platform)
Berichtszeitraum: 2021-03-01 bis 2022-08-31
The main objective of IoTwins is to effectively reduce the barriers to the adoption of distributed and hybrid digital twins for SMEs (and not only) operating in the industrial domains of advanced manufacturing and facility management. IoTwins twins are distributed because they can run at both cloud/HPC side and edge nodes close to datasources, so to enable latency/traffic reduction and increased reliability when needed; they are hybrid because they are able to be fed by both offline/online data from monitored physical systems and synthetic data generated by simulation tools running at HPC/cloud resource-rich deployment environments.
To this purpose, IoTwins has already concentrated (and will continue to do that in the second half of the project) its development and innovation efforts along two primary directions: i) the IoTwins framework and ii) the IoTwins testbeds. On the one hand, IoTwins has defined a common reference architecture for the runtime support of distributed hybrid digital twins and developed highly flexible, modular, and industry-oriented components (at both the platform and the AI service layers) to implement several incarnations of this reference architecture, in practical deployment scenarios where there is the need to integrate with legacy solutions and already deployed equipment and components. On the other hand, as planned, IoTwins has almost completed the first prototyping of high-TRL 7 testbeds (4 in the Industry4.0 application domain and 3 in the facility management domain) based on the IoTwins reference architecture and components, while 2 first testbeds about replicability/scalability (to demonstrate the capability of the IoTwins approach to be replicated over larger/smaller scale deployment scenarios easily) are already under first development.
To this purpose, IoTwins has already concentrated (and will continue to do that in the second half of the project) its development and innovation efforts along two primary directions: i) the IoTwins framework and ii) the IoTwins testbeds. On the one hand, IoTwins has defined a common reference architecture for the runtime support of distributed hybrid digital twins and developed highly flexible, modular, and industry-oriented components (at both the platform and the AI service layers) to implement several incarnations of this reference architecture, in practical deployment scenarios where there is the need to integrate with legacy solutions and already deployed equipment and components. On the other hand, as planned, IoTwins has almost completed the first prototyping of high-TRL 7 testbeds (4 in the Industry4.0 application domain and 3 in the facility management domain) based on the IoTwins reference architecture and components, while 2 first testbeds about replicability/scalability (to demonstrate the capability of the IoTwins approach to be replicated over larger/smaller scale deployment scenarios easily) are already under first development.
The technical work performed within the first 18 months of the project, even if some limited deviations have been necessary mainly due to the pandemic situation, has achieved the primary planned objectives.
About the technical work packages two primary categories of innovation activities have been performed, namely the work about the IoTwins framework and the work about the realization of the different testbeds.
In the first category, WP2 and WP3 have intensely worked on the common understanding, definition, design, implementation, and first validation of the IoTwins framework, which has already started to be used by the IoTwins testbeds, as detailed below. In particular and in short, WP2 has worked on identifying the requirements, selecting the technologies, and developing the mechanisms needed in the IoTwins framework to distributedly manage digital representations of physical assets in both the manufacturing and facility management domains. The primary results already obtained by WP2 are: i) identification of requirements and growth of common understanding about distributed and hybrid digital twins among the partners; ii) definition of the reference architecture for the IoTwins framework, capable of fulfilling the need for extreme flexibility, adaptivity, and extensibility to embrace the different requirements and characteristics of the multiple IoTwins testbeds; iii) preparation of instantiations of the IoTwins platform for different use cases, with differentiated solutions to be integrated at edge nodes and at the backbone infrastructure (cloud and HPC resources). This has also required the development and extension of orchestration and security solutions, as in the Proof of Concept demonstrator of the IoTwins reference architecture, which for example uses the Indigo PaaS Orchestrator, IAM for authentication, and Mesos clusters at the edge.
By passing to WP3, as planned, this work package has worked on the definition, design, and implementation of the AI service part of the IoTwins framework, with the primary goal of developing AI services that can be re-used and can benefit multiple IoTwins use-cases (and others in the future). The IoTwins AI services have already started to be tailored to the specific needs of some testbeds, which are adopting and integrating them in their workflows; they are already available in a common repository, structured into classes for easier usability (8 general services, 4 anomaly detection ones, 8 time-series specific one, 1 about model introspection, and 1 optimization service, developed so far). This is an initial working version of the IoTwins AI services; they will be modified and extended by following the evaluation and feedback stage of the following months. Let us recall the crucial point that the provided services are especially devised for the specific needs of industrial partners and of SMEs, thus making them quite different from the general-purpose services that can be offered by “traditional” and commercial cloud service providers. IoTwins AI services were developed as a collection of self-contained Docker containers; their code is available at the following Git repository: https://gitlab.hpc.cineca.it/iotwins/ai-services/.
About all the testbed activities (WP4, WP5, and WP6), very careful attention was posed to correctly communicating the methodology and the significant advantages of adopting the common IoTwins framework, also by trying to win the natural resistance to innovation associated with the fact that several industrial testbeds were not developed from scratch (similarly to what happens almost always in the industrial domain and to what will be regular for SMEs adopting IoTwins solutions in the future). The details about the activity progress for each single testbed (4 in the advanced manufacturing domain, 3 in the facility management domain, and two for replicability/scalability currently in their early stage of development) are reported in the WP-level technical reports.
About the technical work packages two primary categories of innovation activities have been performed, namely the work about the IoTwins framework and the work about the realization of the different testbeds.
In the first category, WP2 and WP3 have intensely worked on the common understanding, definition, design, implementation, and first validation of the IoTwins framework, which has already started to be used by the IoTwins testbeds, as detailed below. In particular and in short, WP2 has worked on identifying the requirements, selecting the technologies, and developing the mechanisms needed in the IoTwins framework to distributedly manage digital representations of physical assets in both the manufacturing and facility management domains. The primary results already obtained by WP2 are: i) identification of requirements and growth of common understanding about distributed and hybrid digital twins among the partners; ii) definition of the reference architecture for the IoTwins framework, capable of fulfilling the need for extreme flexibility, adaptivity, and extensibility to embrace the different requirements and characteristics of the multiple IoTwins testbeds; iii) preparation of instantiations of the IoTwins platform for different use cases, with differentiated solutions to be integrated at edge nodes and at the backbone infrastructure (cloud and HPC resources). This has also required the development and extension of orchestration and security solutions, as in the Proof of Concept demonstrator of the IoTwins reference architecture, which for example uses the Indigo PaaS Orchestrator, IAM for authentication, and Mesos clusters at the edge.
By passing to WP3, as planned, this work package has worked on the definition, design, and implementation of the AI service part of the IoTwins framework, with the primary goal of developing AI services that can be re-used and can benefit multiple IoTwins use-cases (and others in the future). The IoTwins AI services have already started to be tailored to the specific needs of some testbeds, which are adopting and integrating them in their workflows; they are already available in a common repository, structured into classes for easier usability (8 general services, 4 anomaly detection ones, 8 time-series specific one, 1 about model introspection, and 1 optimization service, developed so far). This is an initial working version of the IoTwins AI services; they will be modified and extended by following the evaluation and feedback stage of the following months. Let us recall the crucial point that the provided services are especially devised for the specific needs of industrial partners and of SMEs, thus making them quite different from the general-purpose services that can be offered by “traditional” and commercial cloud service providers. IoTwins AI services were developed as a collection of self-contained Docker containers; their code is available at the following Git repository: https://gitlab.hpc.cineca.it/iotwins/ai-services/.
About all the testbed activities (WP4, WP5, and WP6), very careful attention was posed to correctly communicating the methodology and the significant advantages of adopting the common IoTwins framework, also by trying to win the natural resistance to innovation associated with the fact that several industrial testbeds were not developed from scratch (similarly to what happens almost always in the industrial domain and to what will be regular for SMEs adopting IoTwins solutions in the future). The details about the activity progress for each single testbed (4 in the advanced manufacturing domain, 3 in the facility management domain, and two for replicability/scalability currently in their early stage of development) are reported in the WP-level technical reports.
Already in its first 18 months, IoTwins has contributed to the progress beyond the state-of-the-art along different perspectives: a novel and highly flexible reference architecture for distributed hybrid digital twins at high TRL has been originally defined; a repository of originally developed AI & simulation services, ready to be integrated in IoTwins platform incarnations, has been put in place; 4 innovative testbeds for the manufacturing domain and 3 for the facility management domain are under finalization, all exploiting the original concept of distributed hybrid digital twins and the high-usability integration with HPC/cloud resources for resource-demanding tasks.
The expected results until the end of the project include the refinement of the IoTwins framework based on the testbed evaluation and feedback, as well as the completion and thorough validation of all the 12 testbeds, including the replicability/scalability ones. Special attention will be dedicated to amplifying at maximum the industrial and societal impact of the project. Indeed, IoTwins aims at enabling SMEs in the manufacturing and facility management/service sectors to access edge/cloud-enabled big data analytics and AI services to create hybrid digital companions to improve their production process and optimize the management of their facilities.
The expected results until the end of the project include the refinement of the IoTwins framework based on the testbed evaluation and feedback, as well as the completion and thorough validation of all the 12 testbeds, including the replicability/scalability ones. Special attention will be dedicated to amplifying at maximum the industrial and societal impact of the project. Indeed, IoTwins aims at enabling SMEs in the manufacturing and facility management/service sectors to access edge/cloud-enabled big data analytics and AI services to create hybrid digital companions to improve their production process and optimize the management of their facilities.