Periodic Reporting for period 2 - MONSOON (MOdel based coNtrol framework for Site-wide OptmizatiON of data-intensive processes)
Période du rapport: 2018-04-01 au 2019-09-30
Process industries are characterized by intense use of raw resources and energy, thus providing a context where even small optimizations can lead to savings both in terms of economic and environmental costs. This is especially true for specific industrial processes such as aluminium smelting or injection moulding, characterized by production in high volumes divided among distributed production units, across several lines, plants or even sites.
Predictive modelling techniques can be especially effective in optimizing processes in such context, but their application is not straightforward for several reasons including e.g. the high cost of integrating new sensors or actuators into legacy production, difficulties in monitoring physical parameters in harsh conditions, interoperability issues, difficulties in application fusing and correlating information collected at different SCADA levels, challenges in defining KPIs, etc. As a consequence, the deployment of model-based predictive functions in such production environment at a sustainable cost or with sufficient reliability is not always feasible, resulting in optimization potentials remaining untapped.
In past markets characterized by lower international competition, stable demand, relatively low labour cost and high abundance of raw materials, industry was able to remain viable just through progressive improvements in production technology, organization and logistics. The change in global competition and resources availability calls instead for a drastic re-invention and re-design of production processes and sites. Enabling benefits by integrating innovations in the installed process base is a fundamental step to help process industries transitioning from the current model oriented to the production of goods by consuming resources, to newer “circular” models. In this perspective, resource, cost and environmental sustainability is considered, monitored and optimized at all times, resulting in benefits for industries and society as a whole.
MONSOON project aims at establishing data-driven methodology and tools to support identification and exploitation of optimization potentials through model based predictive controls. The data lab enables multidisciplinary teams to jointly model, develop, simulate, verify, deploy and evaluate distributed predictions and controls. This will help plants in meeting their optimization.
MONSOON achieved to develop a flexible and scalable infrastructure able to support a model-based development environment dedicated to easier the study, the creation and the test of function dedicated to improve and optimize processes. MONSOON solution has been tested by data scientists and end users in order to better align the finalization of the requirements. In fact, data scientists created new or improved existing predictive functions, regularly cooperating with the domain experts. The predictive functions for the different scenarios have been created and tested by using MONSOON platform. Nevertheless, the consortium faced and overcame several technical and not technical challenges related to the production and plant needs and schedule. All the challenges gave to the consortium many important lessons learned.
Predictive modelling techniques can be especially effective in optimizing processes in such context, but their application is not straightforward for several reasons including e.g. the high cost of integrating new sensors or actuators into legacy production, difficulties in monitoring physical parameters in harsh conditions, interoperability issues, difficulties in application fusing and correlating information collected at different SCADA levels, challenges in defining KPIs, etc. As a consequence, the deployment of model-based predictive functions in such production environment at a sustainable cost or with sufficient reliability is not always feasible, resulting in optimization potentials remaining untapped.
In past markets characterized by lower international competition, stable demand, relatively low labour cost and high abundance of raw materials, industry was able to remain viable just through progressive improvements in production technology, organization and logistics. The change in global competition and resources availability calls instead for a drastic re-invention and re-design of production processes and sites. Enabling benefits by integrating innovations in the installed process base is a fundamental step to help process industries transitioning from the current model oriented to the production of goods by consuming resources, to newer “circular” models. In this perspective, resource, cost and environmental sustainability is considered, monitored and optimized at all times, resulting in benefits for industries and society as a whole.
MONSOON project aims at establishing data-driven methodology and tools to support identification and exploitation of optimization potentials through model based predictive controls. The data lab enables multidisciplinary teams to jointly model, develop, simulate, verify, deploy and evaluate distributed predictions and controls. This will help plants in meeting their optimization.
MONSOON achieved to develop a flexible and scalable infrastructure able to support a model-based development environment dedicated to easier the study, the creation and the test of function dedicated to improve and optimize processes. MONSOON solution has been tested by data scientists and end users in order to better align the finalization of the requirements. In fact, data scientists created new or improved existing predictive functions, regularly cooperating with the domain experts. The predictive functions for the different scenarios have been created and tested by using MONSOON platform. Nevertheless, the consortium faced and overcame several technical and not technical challenges related to the production and plant needs and schedule. All the challenges gave to the consortium many important lessons learned.
The first period in the project was shaped by the work on project initiation, requirements engineering and specification, prototypes development and pilot definition.
Specific effort was devoted to the definition of the main vision and related context scenarios as well as an initial set of requirements, for both domains. A work with an Extended Stakeholder Group (ESG) - experts with strong skills in one or more technology fields related to the MONSOON platform - was initiated. In the first months of the project (ramp up phase), a minimal ICT/IoT infrastructure to enable continuous data collection from the production floor in both domains and storage into a scalable, cloud-oriented platform has been developed and deployed. In parallel, reference architecture was designed, based on scenarios and user interviews, and built upon the concept of cross-sectorial data lab.
During the second period, MONSOON components were developed in all their parts, finalized, tested and deployed in the real plants. The data scientists created new or improved existing predictive functions (PFs), using the MONSOON solution. PFs for the different scenarios have been created, tested and deployed in the pilot sites, along with visual analytics tools. Workshops and online meetings have been organized with the end users to give more technical explanation and training about the tools. Dissemination activities focused on different events and B2B connections. The consortium organized two major events – a joint with the other SPIRE-02-2016 projects and a final Workshop (involving the same projects), to present MONSOON. ESG interaction was further bound by few online meetings and one physical meeting. The consortium developed a MONSOON Final Exploitation plan agreed among the partners.Finally from the standardization point of view, a CWA document has been created and published.
Specific effort was devoted to the definition of the main vision and related context scenarios as well as an initial set of requirements, for both domains. A work with an Extended Stakeholder Group (ESG) - experts with strong skills in one or more technology fields related to the MONSOON platform - was initiated. In the first months of the project (ramp up phase), a minimal ICT/IoT infrastructure to enable continuous data collection from the production floor in both domains and storage into a scalable, cloud-oriented platform has been developed and deployed. In parallel, reference architecture was designed, based on scenarios and user interviews, and built upon the concept of cross-sectorial data lab.
During the second period, MONSOON components were developed in all their parts, finalized, tested and deployed in the real plants. The data scientists created new or improved existing predictive functions (PFs), using the MONSOON solution. PFs for the different scenarios have been created, tested and deployed in the pilot sites, along with visual analytics tools. Workshops and online meetings have been organized with the end users to give more technical explanation and training about the tools. Dissemination activities focused on different events and B2B connections. The consortium organized two major events – a joint with the other SPIRE-02-2016 projects and a final Workshop (involving the same projects), to present MONSOON. ESG interaction was further bound by few online meetings and one physical meeting. The consortium developed a MONSOON Final Exploitation plan agreed among the partners.Finally from the standardization point of view, a CWA document has been created and published.
MONSOON has combined known best practices methodologies for model based site-wide control and integrate them within the collaborative Data Lab methodology. Advances lie in the use of the data-driven methodology, exploiting since the beginning of the project real data from field installation.
MONSOON addressed progress beyond state of the art in the following topics: methodologies for the multi scale modelling , techniques for early malfunctioning detection, integration of heterogeneous systems for model based plant-site monitoring and optimization, data analysis techniques, multi-Scale Deep Learning, methodologies for Life Cycle Management, multi-level Analytics and Multi-Modal Visualization.
MONSOON has provided the following major outcomes.
** Multi-scale control methodology - Data-driven methodology (based on machine and deep learning algorithms) to perform large-scale data analysis for predictive control.
** Real-time Plant Operations Platform - SW framework allowing to collect data and interact with process industry systems, to implement predictive control and life cycle management.
** Cross-sectorial Data Lab - Distributed big data storage and data analytics platform.
** Semantic framework - Proposal for the standard to formalize data analysis process for predictive control and maintenance and simplify communication between experts and data scientists.
** Analytics and visualization tools - Fuse data coming from disjoint plant levels to detect complex patterns of manufacturing processes and provide useful information.
** Integrated LC Management Tools - Integration of LC management tools to access data from ERP and MES system and to feed LC targets and elaborated metrics back into the control infrastructure.
Based on the technical results, MONSOON expected impact directly relates to the wide / comprehensive / sustainable growth of the European productivity ecosystem. The change in global competition is actually driving a drastic re-design of production processes and sites. In this scenario, having the possibility to reach the same results with “few” ICT-driven investments represents an enormous potential, enabling immediate benefits for the company involved, as well as indirect positive impacts on the European production tissue and on the wider society.
MONSOON addressed progress beyond state of the art in the following topics: methodologies for the multi scale modelling , techniques for early malfunctioning detection, integration of heterogeneous systems for model based plant-site monitoring and optimization, data analysis techniques, multi-Scale Deep Learning, methodologies for Life Cycle Management, multi-level Analytics and Multi-Modal Visualization.
MONSOON has provided the following major outcomes.
** Multi-scale control methodology - Data-driven methodology (based on machine and deep learning algorithms) to perform large-scale data analysis for predictive control.
** Real-time Plant Operations Platform - SW framework allowing to collect data and interact with process industry systems, to implement predictive control and life cycle management.
** Cross-sectorial Data Lab - Distributed big data storage and data analytics platform.
** Semantic framework - Proposal for the standard to formalize data analysis process for predictive control and maintenance and simplify communication between experts and data scientists.
** Analytics and visualization tools - Fuse data coming from disjoint plant levels to detect complex patterns of manufacturing processes and provide useful information.
** Integrated LC Management Tools - Integration of LC management tools to access data from ERP and MES system and to feed LC targets and elaborated metrics back into the control infrastructure.
Based on the technical results, MONSOON expected impact directly relates to the wide / comprehensive / sustainable growth of the European productivity ecosystem. The change in global competition is actually driving a drastic re-design of production processes and sites. In this scenario, having the possibility to reach the same results with “few” ICT-driven investments represents an enormous potential, enabling immediate benefits for the company involved, as well as indirect positive impacts on the European production tissue and on the wider society.