Skip to main content

Energy-aware Factory Analytics for Process Industries

Periodic Reporting for period 1 - FACTLOG (Energy-aware Factory Analytics for Process Industries)

Reporting period: 2019-11-01 to 2020-10-31

Digital transformation of the process industry is very challenging. The penetration and wide adoption of Artificial Intelligence (AI), Big Data and Internet of Things (IoT) create some areas of automation and improved decision making. Aspects like Energy consumption, process performance can be easily monitored and optimized with the use of Industry 4.0 concepts. For instance, industry sector accounted for 25.35% of the EU-28 total final energy consumption in 2015 being the second most energy consuming sector behind transport. Energy consumption is a critical cost factor in the process industry and can be reduced in different ways: a) by transforming energy sources to “greener” and less fuel/fossil dependant, and b) to optimize operations in a way that they can use less energy.

In the digital world, different tools can be used for optimized operations. The usage of Digital Twins is considered one of the most promising trends where we can model process industry assets in the digital world and be able to better monitor the performance and assess impact and test optimization scenarios.

Complementary to the Digital Twins concept, the use of AI can enhance the ability to understand behavioural trends, predict issues and possible anomalies. Predictive analytics offer a wide range of functionalities such as predictive maintenance, energy optimization and others which result in significant cost reductions including energy efficiency.

FACTLOG aims to address the above challenges in the process industry by developing a technology pipeline based on the concept of Enhanced Digital Twins (ECT): Digital Twins with cognition capabilities, able to understand and alter their behaviour and enhanced with optimization capabilities. The project will deliver this pipeline with configuration and integration services and will ensure adoption in different process contexts. FACTLOG will test the offerings into five different process industry environments.
The main work done in the first year of the project was the definition of the business cases/application scenarios of FACTLOG in the process industry. We did a thorough analysis of the five pilot cases and we concluded with different problems that can be solved with the use of ECT: from energy monitoring, to predictive maintenance, process scheduling and self-configurable process units.

In addition, the project defined a cognitive framework based on the concept of Digital Twins, where the capabilities of ECT are defined and with detailed description of how this framework supports the realization of the business scenarios.

Further to this, the project defined an overall technical architecture with the necessary components and interfaces among them. We started modelling the analytics services which support the cognition process in each of the five pilots.

Last we started discussions with external stakeholders to expand our focus beyond the project pilots and see the market potentials of FACTLOG solution. We got initial interest and some basic challenges, and we expect to continue the discussion to maximise the project impact.
Whereas in traditional large enterprises digital twins are mostly used as “digital replicas” of physical assets (machines) that support better design of products (through simulations), data-driven manufacturing requires more “cognitive augmentation” of assets for enabling data-driven continuous process improvement. FACTLOG’s Enhanced Cognitive Twins have a (cognition-like) ability for understanding/resolving the “unknown unknowns” (i.e. as discussed above, the situations which can neither be modelled by design, e.g. encapsulated in numerical models, nor experienced in reality, i.e. behaviour in past data). FACTLOG incorporates a solution that will have the capability to “reason” about the problem (e.g. level of the tool degradation for each machine and the type of product/material) based on real-time data from the current process. This cognition capability is supported by the so-called “just-in-time simulation” of the process status in some suspicious/unusual situations in order to calculate the confidence that an unusual situation leads to an anomaly that should be resolved. It then incorporates optimisation methods enabling the digital twin to recover the actual system’s performance and resource efficiency.

Short-term decision support level implies real-time (and/or reactive) optimization, as the performance of a manufacturing process or control system depends mostly on its ability to rapidly adapt schedules to current circumstances. These dynamic (re-)optimization problems, where the optimization process is performed gradually in specific time intervals and evolves dynamically in an effort to incorporate the occurrence of new information or the update of old ones, can be approached by reactive and proactive frameworks. Such approaches have already been designed and are to be deployed in at least four of FACTLOG’s industrial sites, facing a diversity of operational environment, from on-spec timely recovery of Liquefied Petroleum Gas (LPG) refinement to effective scheduling of textile weaving and to maintenance driven manufacturing.

FACTLOG’s progress beyond existing knowledge relies not only on optimisation and analytics as individual components but on their tight integration that relies heavily on process modelling and knowledge graphs to ensure replicability and transferability to manufacturing environments analogous or closely related to the ones examined within the FACTLOG’s pilots. This happens through a message bus and integrating all information systems including the EDTs.
In socio-economic terms, FACTLOG introduces enhanced automation that will support process industries and manufacturing in terms of cost-reduction and resource efficiency, thus bringing tangible benefits to companies associated with FACTLOG through its industrial clusters and associations. Details are already accessible in deliverables D1.3 and D2.1.